Molecular Mechanisms Underlying Neurotransmitter Release

Reconstitution of SNARE-dependent membrane fusion

The realization that the membrane proximal SNARE motifs of synaptobrevin and syntaxin-1 bind in a parallel fashion led to the proposal that formation of the SNARE complex brings the synaptic vesicle and plasma membranes together and induces membrane fusion (58) (Figure 2c). In this model, the energy of formation of this highly stable complex provides the energy to drive membrane fusion. The model was supported by pioneering reconstitution assays showing lipid mixing between synaptobrevin-containing liposomes and syntaxin-1-SNAP-25-containing liposomes (165) and subsequent research has provided ample evidence that the neuronal SNAREs alone can cause membrane fusion [reviewed (18, 67, 148, 189)]. This research has included the development of diverse additional methods that monitor fusion between small unilamellar vesicles (SUVs) and giant unilamellar vesicles (GUVs) (93), between single pairs of vesicles (78, 187), between vesicles and supported or suspended lipid bilayers (42, 60, 76), and between nanodiscs and vesicles (10, 136), planar bilayers (9) or cells (174).

Each one of these methods has its own advantages but it is also important to consider their limitations when interpreting the results. Bulk assays with liposomes that use fluorescence resonance energy transfer (FRET) are relatively easy to perform and have provided critical insights into the mechanisms underlying neurotransmitter release (88, 155, 165) and other types of membrane traffic, particularly yeast vacuolar fusion (101, 143). However, these assays suffer from a slow time resolution, which hinders dissection of steps such as docking and fusion, or detection of intermediates, and do not mimic the flat nature of the plasma membrane. Single vesicle-single vesicle fusion assays provide a faster time resolution and have also yielded important insights (39, 79, 81, 82). Some data have suggested that fusion of single vesicles with flat bilayers occurs faster than between two vesicles under comparable conditions (74), perhaps because the former mimic the geometry of synaptic vesicle fusion better or because of unknown technical reasons. It is also important to note that, in both bulk and single vesicle fusion assays monitored by fluorescence spectroscopy, it is critical to monitor not only lipid mixing but also content mixing to ensure that actual fusion without leakiness takes place, as large amounts of lipid mixing can occur with very little content mixing (25, 78, 196).

Assays that monitor fusion by electrophysiological methods (9, 60, 174) offer even faster time resolution [in the 10 μs time scale] than fluorescence-based methods and hence are most promising as optimization of the reconstitution systems recapitulate synaptic vesicle fusion more accurately. These methods have allowed the observation of fusion pores and indicate that one SNARE complex is sufficient to form a fusion pore, but the size of the fusion pore increases with the number of SNARE complexes (9, 60). These observations agree with the notion that the number of assembled SNARE complexes dictates release probability. Moreover, comparable data were obtained in nanodisc-liposome, nanodisc-flat bilayer and liposome-flat bilayer assays (9, 60, 136), and there is evidence that a minimum of 2–3 SNARE complexes is normally required for neurotransmitter release but even a single SNARE complex can support release (103, 139). This correlation and the consistency of the data are important to support the biological relevance of the results, as the mere observation of membrane fusion in vitro does not mean that the underlying mechanism resembles that occurring in vivo.

A different picture emerged from cryo-electron microscopy (cryo-EM) images of bulk SNARE-dependent fusion reactions between liposomes or between liposomes and GUVs showing the formation of large, extended interfaces (13, 39, 61) (illustrated by the diagram of Figure 2d) that constitute intermediates in the fusion pathway (171). Because fusion and the disappearance of the extended interfaces in these experiments are much slower than that of neurotransmitter release, it is unlikely that these intermediates occur in vivo. Indeed, it appears that other components of the release machinery may limit the formation of these extended interfaces (39, 53). It is unclear whether the extended interfaces can lead to the fusion pores observed in liposome-flat bilayer fusion assays (60), but it is possible that fusion might occur by more than one mechanism. Thus, the observed fusion pores were estimated to involve a few SNARE complexes, which may lead initially to the formation of small, point-of-contact interfaces, whereas formation of a large number of SNARE complexes might lead to the extended interfaces. Hence, formation of the extended interfaces observed by cryo-EM might arise because the large amount of SNAREs present in the liposomes. Note that, although the SNARE-to-lipid ratios used in these experiments (13, 39, 61) were comparable to physiological levels, other key components of the release machinery such as Munc18–1 and Munc13–1 may limit the number of SNARE complexes that are formed in a primed vesicle [(177), see below], preventing the formation of extended interfaces.

In this context, the use of nanodiscs as one of the fusion partners is attractive because it allows good control of the number of SNARE complexes that can mediate fusion. However, synaptobrevin was normally incorporated into the nanodiscs, which does not mimic the geometry of synaptic vesicle fusion. Moreover, data obtained with small nanodiscs need to be interpreted with caution given the limited number of lipid molecules that can be incorporated in each nanodisc. For instance, experiments with 6 nm nanodiscs led to the conclusion that exocytotic fusion pores are composed of both lipids and proteins (10), but such nanodiscs are expected to contain only 30–40 lipids in each leaflet and it is therefore highly unclear whether the results can be extrapolated to physiological fusion pores. Geometrical considerations become even more important as other components of the release machinery are incorporated in reconstitutions assays because the actions of large proteins such as Munc13–1 (see below) may be impossible to recapitulate with nanodiscs even if they have diameters of 30 nm.

How do the SNAREs induce membrane fusion?

Although it is now well established that SNARE complexes can induce membrane fusion, the underlying mechanism remains highly enigmatic. Original models envisaged long, continuous helices formed by the SNARE motif, the short juxtamembrane region and the TM region of both synaptobrevin and syntaxin-1 coming progressively together and forcing membrane fusion as the SNARE complex ‘zippers’ from the N- to the C-terminus (58, 150, 165). Although such cartoons are widespread in the literature, the strong bend of the helices in the juxtamembrane sequences depicted in these models is highly unlikely based on conformational grounds. These sequences are expected to remain largely unstructured when the four helix bundle zippers fully between two membranes (as depicted in the diagram of Figure 2c, left). Note also that the free energy of SNARE complex assembly is estimated to be between 35 and 68 k_BT (51, 83) and, therefore, formation of a few SNARE complexes or perhaps even just one can in principle overcome the energy required for membrane fusion, but it is unclear how this energy is applied to the membranes. Most of the free energy is spent when the four-helix bundle is fully zippered, which should bring the two membranes within a distance of few nanometers. However, bringing two membranes into contact does not necessarily lead to fusion. It thus seems likely that the SNAREs do more than bringing the membranes together to cause fusion, and the likely culprits are the juxtamembrane regions. These sequences contain abundant basic residues that are expected to interact with the negatively charged head groups and thus help bring the membranes even closer. Moreover, the synaptobrevin juxtamembrane region contains tandem tryptophan residues that can insert into lipid bilayers and perturb their packing. Indeed, liposome fusion in reconstitution assays is strongly affected by the position of these tryptophan residues with respect to the TM region (63), and Ca²⁺-evoked neurotransmitter release is considerably impaired by mutation of the two tryptophans to alanine (96). As discussed below, other proteins (Sec17 and synaptotagmin-1) likely accelerate membrane fusion also by perturbing lipid bilayers, suggesting that proteins must perform two key actions to fuse membranes: bring them into close proximity and perturb bilayer packing to facilitate the formation of non-bilayer intermediates that lead to fusion (169).

NSF and SNAPs

N-ethylmaleimide sensitive factor (NSF) is a soluble protein that was shown in 1988 to be required for transport of cargo vesicles to Golgi cisternae and was proposed to assemble into a multisubunit ‘fusion machine’ (91). Soluble NSF attachment proteins (SNAPs) were later shown to mediate binding of NSF to membrane sites (167), and to form a complex with synaptobrevin, syntaxin-1 and SNAP-25, leading to their designation as SNAREs [for SNAP receptors, (142)]. It was later shown that the SNAREs form a complex by themselves (the SNARE complex) and that NSF disassembles the SNARE complex through its ATPase activity, with the help of SNAPs (141). These and other results (8, 97) led to the current view that the main function of NSF/SNAPs is to disassemble the cis-SNARE complexes that result after membrane fusion to recycle the SNAREs. NSF/SNAPs also disassemble 2:1 syntaxin-1/SNAP-25 complexes and improperly assembled SNARE complexes with antiparallel helices (32, 88). Moreover, NSF/SNAPs were recently shown to disassemble trans-SNARE complexes in vitro (115, 185), which correlates with the finding that primed synaptic vesicles can be de-primed and N-ethylmaleimide impairs such de-priming (59). Not surprisingly, NSF and αSNAP completely inhibit SNARE-mediated liposome fusion in reconstitution assays (85, 88). In fact, αSNAP potently inhibits fusion by itself, in part because it binds to syntaxin-1-SNAP-25 heterodimers, preventing binding to synaptobrevin (Fig. 1a), and in part because it blocks fusion by pre-formed trans-SNARE complexes (111, 144) (Fig. 1b).

Cryo-EM structures of the 20S complex formed by NSF, αSNAP and the neuronal SNAREs provided seminal insights into the SNARE complex disassembly mechanism (26, 65, 166, 190) (Figure 3a). NSF contains an N-terminal domain that binds to αSNAP and two nucleotide binding domains of the AAA family, which are called D1 and D2 and mediate the formation of hexameric rings by NSF. NSF interacts directly with the N-terminus of SNAP25 through a pore formed by the D1 ring (166). Two to four αSNAP molecules bound to the SNAREs were observed in these structures, although the higher stoichiometry may be favored on membranes because of co-localization. When four αSNAP molecules are bound, they cover almost the entire surface of the SNARE four-helix bundle, interacting with the SNAREs primarily through electrostatic interactions and suboptimal packing, which explains the lack of specificity of the NSF/SNAP machinery. Suboptimal packing arises in part because the elongated αSNAP structure has a right-handed twist whereas the SNARE four-helix bundle has a left-handed twist, suggesting that αSNAP binding might already destabilize the interactions between the SNAREs. SNARE complex disassembly is believed to involve a power stroke when ATP hydrolysis by the D1 domain causes large conformational changes that propagate to the NSF N-terminal domain and αSNAP, ultimately tearing the SNARE four-helix bundle apart (65, 190). A hydrophobic N-terminal loop of αSNAP that inserts into membranes and accelerates disassembly of membrane-anchored cis-SNARE complexes (170) is located next to the C-terminus of the SNARE four-helix bundle in the structure of the 20S complex (Figure 3a), indicating that the acceleration arises because simultaneous interactions of four αSNAP molecules with the membrane and the SNAREs can cooperate to form the 20S complex.

Open in a separate window

Figure 3.

The 20S complex. (a) Cryo-EM structure of the 20S complex formed by NSF, αSNAP, synaptobrevin (red), syntaxin-1 (yellow) and SNAP-25 (green) (190) (PDB ID 3J96). The four molecules of αSNAP that bind around the SNARE four-helix bundle are colored in orange and brown in an alternative fashion. Similarly, the six molecules of NSF are colored in blue and cyan in an alternative fashion. The positions of the D1, D2 and N-domains of NSF, as well as the N-terminal hydrophobic loop of αSNAP, are indicated. (b) Model proposing that the core machinery that mediates yeast vacuolar fusion is a 20S complex formed by Sec18 (blue), Sec17 (orange) and the four vacuolar SNAREs (yellow, green and red) (143). Note that the small ellipse at the membrane-proximal end of each αSNAP molecule represents its N-terminal hydrophobic loop and insertion of the loop into the bilayers is postulated to make a key contribution to membrane fusion.

Synaptotagmin-1 and Ca²⁺/phospholipid binding

Synaptotagmins are a family of proteins that contain tandem C₂ domains and includes Ca²⁺ sensors involved in diverse forms of regulated secretion (172). Here I focus on the synaptic vesicle protein synaptotagmin-1, the Ca²⁺ sensor that triggers synchronous neurotransmitter release (48), but it is important to keep in mind its interplay with other Ca²⁺ sensors that may compete for binding to common targets. Thus, synchronous release is abrogated in synaptotagmin-1 KO mice but asynchronous release is enhanced, and this asynchronous component is almost abolished by knockdown of synaptotagmin-7 (5), showing that synaptotagmin-7 acts as a Ca²⁺ sensor in asynchronous release and illustrating a functional competition between synaptotagmins-1 and −7. Conversely, Synaptotagmin-1 KO mice exhibit an enhancement in spontaneous release that is not altered by synaptotagmin-7 knockdown (5). This increase likely arises because of an inhibitory activity of Ca²⁺-free synaptotagmin-1 (see below) and/or because of an interplay with other Ca²⁺ sensors such as Doc2b that remains controversial (35, 40, 107). Synaptic vesicle priming is impaired in synaptotagmin-1 KO mice (27) but is not abolished as in Munc18–1 KO or Munc13–1/2 DKO mice (158, 159). These results likely arise because Munc13 and Munc18–1 are essential for assembly of trans-SNARE complexes in the presence of NSF and αSNAP, whereas synaptotagmin-1 is not essential but markedly stimulates assembly (115).

The C₂A and C₂B domains of synaptotagmin-1 form most of its cytoplasmic region and bind three and two Ca²⁺-ions, respectively, through loops at the top of β-sandwich structures (49, 149, 156) (Figure 7b). These loops also mediate Ca²⁺-dependent phospholipid binding, which involves a Ca²⁺-induced switch in the electrostatic potential (135), coordination of Ca²⁺ by the phospholipid head groups and a combination of ionic and hydrophobic interactions between the loops and the lipids (28, 188). Mutations that increased or decreased the apparent Ca²⁺ affinity of synaptotagmin-1 binding to membranes led to parallel changes in the Ca²⁺ sensitivity of neurotransmitter release (48, 123), which demonstrated that synaptotagmin-1 is the major Ca²⁺ sensor that triggers release and that Ca²⁺/phospholipid binding is key for this function. Mutations in residues of the C₂B domain that bind Ca²⁺ or phospholipids, including hydrophobic residues that insert into the lipid bilayer, cause stronger phenotypes than analogous mutations in the C₂A domain because they cause dominant negative effects, but mutations in the C₂A domain can also cause severe disruption of release (14, 89, 106, 146), showing that cooperation between the two C₂ domains is crucial for synaptotagmin-1 function.

The synaptotagmin-1 tandem C₂ domains can bind not only to one membrane but to two membranes simultaneously in the presence of Ca²⁺ due to the presence of multiple basic sequences on their surface, particularly that of the C₂B domain (3) (Figure 7b). This observation led to the proposal that upon Ca²⁺ binding synaptotagmin-1 can cooperate with the SNAREs in bringing the membranes together, which was supported by diverse evidence (157, 181). Among the basic sequences, particularly important for synaptotagmin-1 function are R398,R399 at the bottom of the C₂B domain, which were implicated in membrane bridging (3, 181), and a polybasic sequence on the side of the C₂B domain β-sandwich, which binds to PIP₂-containing membranes in the absence of Ca²⁺ (6). Both of these sequences also mediate binding to the SNARE complex in vitro (see below). Interestingly, the drastically different effects of distinct mutations in the polybasic region on neurotransmitter release correlate well with their distinct effects on Ca²⁺-dependent binding to PIP₂-containing membranes (see below) but not with their indiscriminate effects on Ca²⁺-independent PIP₂ binding (16, 160). Hence, although binding of synaptotagmin-1 to PIP₂ likely stabilizes the primed states of synaptic vesicles (Figure 1e,​,f),f), what appears to be particularly critical for release is the Ca²⁺-dependent interaction with PIP₂.

Synaptotagmin-1-SNARE-complexin coupling in neurotransmitter release

Unraveling how the actions of synaptotagmin-1, complexins and the SNAREs are coordinated to trigger fast neurotransmitter release upon Ca²⁺ influx has been hindered by the difficulty of distinguishing which among the many interactions that have been described between synaptotagmin-1 and the SNAREs are biologically relevant and which arise merely from the promiscuity of these proteins (126). This uncertainty also arose from the observation that a complexin-1 fragment spanning the accessory and central helices and a synaptotagmin-1 fragment spanning its two C₂ domains (C₂AB) bind simultaneously to the SNARE complex in solution, but compete for binding to membrane-anchored SNARE complex (152, 176). Elucidation of three synaptotagmin-1-SNARE complex structures (16, 193, 194) first added to the confusion because they revealed three different binding modes, but ultimately proved fundamental to understand the behavior of these proteins in vitro and to bring a clearer picture of how their functions are coupled in vivo.

A structure determined by NMR spectroscopy revealed a dynamic binding mode involving ionic interactions between the polybasic region of the synaptotagmin-1 C₂B domain and a polyacidic area of the SNARE complex (16) (Figure 7c). In contrast, a crystal structure revealed binding of the C₂B domain to the SNARE complex through a so-called primary interface involving two regions of the C₂B domain, one formed largely by E295 and Y338, and another including R398,R399 (193) (Figure 7d). This interface was also observed in a subsequent crystal structure that included a complexin-1 fragment, but the crystals contained an additional interface involving binding of an α-helix at the bottom of the C₂B domain to a groove of the SNARE complex, continuing the complexin-1 helix (referred to as tripartite interface) (194) (Figure 7e). A recent detailed analysis of syntaptotagmin-1-SNARE complex interactions in solution and on nanodiscs, together with previously available data, suggested that the key physiological interaction between synaptotagmin-1 and the SNARE complex is mediated by the primary interface (160). This study led to a model whereby this interaction occurs in the primed states of synaptic vesicles (Figures 1e,​,f)f) and is dissociated upon Ca²⁺ influx to initiate synaptic vesicle fusion (Figures 1f,​,gg,​,hh and ​and7f7f,​,g).g). This model challenges the long-held view that Ca²⁺ stimulates synaptotagmin-1-SNARE complex interactions and is supported by the following arguments.

Extensive evidence from initial studies showed that Ca²⁺ stimulates synaptotagmin-1-SNARE interactions in solution, but it is now clear that binding to soluble SNARE complex involves the polybasic and primary interfaces rather than the Ca²⁺ binding loops (160). Thus, such stimulation can be attributed to a Ca²⁺-induced increase in the positive electrostatic potential of the C₂B domain. Ca²⁺ also stimulates binding of synaptotagmin-1 C₂AB to SNARE complex anchored on membranes containing phosphatidyl serine (PS) (36), which also involves the polybasic and primary interfaces (160). This finding explains the competition of Ca²⁺-bound C₂AB with the complexin-1 fragment for binding to membrane-anchored SNARE complex (152), as simultaneous binding would lead to steric clashes of the complexin-1 fragment with the membrane (16). Note however that this competition is most likely irrelevant because C₂AB does not bind to SNARE complex anchored on nanodiscs containing PS and PIP₂ in the presence of Ca²⁺ under physiological conditions including ATP (160), consistent with previous results (110). Moreover, Ca²⁺-dependent binding of C₂AB to PS-PIP₂-containing membranes occurs with very high affinity and is strongly disrupted by a R322E/K325E mutation but not by a K324E/K326E mutation in the polybasic region (160), in correlation with the effects of these mutations on neurotransmitter release (16). These results suggest that the very tight interaction of the C₂B domain with PS-PIP₂-containing membranes induced by Ca²⁺ involves an approximately perpendicular orientation incompatible with SNARE complex binding (Figure 7g). It is worth noting that the effects of the R322E/K325E and K324E/K326E mutations on release also correlated with disruption of SNARE complex binding through the polybasic interface in solution (16). However, it now appears that binding of the C₂B domain polybasic region to the SNARE complex is not physiologically relevant, as it seems much more likely that the disruption of release caused by the R322E/K325E mutation arises because of impairment of the much tighter interaction of the C₂B domain with PS-PIP₂ -containing membranes.

In the absence of Ca²⁺, C₂AB does bind to SNARE complex anchored on PS-PIP₂-containing nanodiscs under physiological conditions including ATP (160). Such binding is dominated by the primary interface, likely because it allows simultaneous binding of the C₂B domain polybasic region to PIP₂ on the membrane (Figure 7f) as suggested by a low-resolution cryo-EM structure of C₂AB bound to SNARE complex on lipid nanotubes (56). This arrangement is compatible with complexin-1 binding to the SNARE complex and is expected to orient the complexin-1 AH toward the vesicle membrane (Figure 7f), which would hinder final zippering of the SNARE complex and synaptic vesicle fusion. Hence, this model suggests a mechanism by which synaptotagmin-1 and complexin-1 promote the formation of a primed state where the SNARE complex is close to fully assembled and ready for release, but fusion is inhibited (Figure 1f) until the arrival of Ca²⁺ relieves the inhibition by inducing dissociation of synaptotagmin-1 from the SNARE complex (Figure 1g, ​,7g).7g). The tight coupling between synaptotagmin-1 and complexin-1 functions predicted by this model is strongly supported by mutagenesis studies in Drosophila (69) [but see (34)]. The notion that Ca²⁺ dissociates synaptotagmin-1 from the SNARE complex was also suggested by the cryo-EM studies on lipid nanotubes mentioned above (56). Note also that the strong impairments of evoked neurotransmitter release caused by E295A/Y338W and R398Q/R399Q mutations designed to disrupt binding through the primary interface supported the physiological importance of this interface (193). However, while the R398Q/R399Q mutation abrogated the inhibition of spontaneous release caused by synaptotagmin-1 and increased asynchronous release, the E295A/Y338W mutant still clamped spontaneous release and did not increase asynchronous release. These findings were explained by the observation that the E295A/Y338W mutation actually enhances binding of synaptotagmin-1 to the SNARE complex, whereas the R398Q/R399Q mutation abrogates this interaction (160). Overall, these results support the notion that synaptotagmin-1 binding to the SNARE complex through the primary interface is critical to generate the primed state that is ready for fast release (Figure 1f), but this interaction needs to be released upon Ca²⁺ influx to allow release (Figure 1g,​,hh).

The C₂B domain-SNARE complex binding mode involving the tripartite interface (Figure 7e) was interesting because it provided an explanation for the finding that complexins are required for the dominant negative effect caused by mutations that abolish Ca²⁺ binding to the synaptotagmin-1 C₂B domain (194). However, binding through the primary interface with concomitant binding of complexin-1 (Figure 7f) can also explain this observation. Indeed, a screen for mutations that abrogate this dominant negative effect in Drosophila yielded a large number of mutations in the primary interface but none in the tripartite interface (57). Conversely, a quintuple mutation that disrupts the primary interface did not abolish the dominant negative effect in mice (194). The basis for these different results is unclear. It is also worth noting that binding of the C₂B domain to the complexin-1-SNARE complex through the tripartite interface could not be detected so far in solution by NMR spectroscopy at 85 μM concentration even after disrupting binding to the polybasic and primary interfaces (160). Moreover, in the arrangement of Figure 7f, ​,itit is difficult to envision how a second C₂B domain can bind simultaneously to the SNARE complex through the tripartite interface and to PIP₂ on the plasma membrane through the polybasic region. Thus, binding through the tripartite interface would require dissociation of the stronger interaction of the C₂B domain with PIP₂ on the plasma membrane, although it is plausible that binding to the tripartite interface is stabilized by interactions of C₂B with the vesicle membrane (18). Overall, it seems premature to rule out the relevance of the tripartite interface, but also to accept it without further evidence.

An intriguing property of synaptotagmin-1 is the ability to form oligomeric rings of 11–26 units that are disrupted upon Ca²⁺ binding, which suggested a model whereby these rings inhibit release and Ca²⁺ disassembles the rings, activating neurotransmitter release (161). A concern about the significance of this property is that most of the ring images reported so far were obtained by negative stain EM, but such rings were not observed in cryo-electron tomography images of reconstituted liposomes containing synaptotagmin-1 [(53) and our unpublished results]. Nevertheless, the formation of ring-like structures has also been supported by the observation of symmetrical densities between synaptic vesicles and the plasma membrane in tomographic images of presynaptic terminals (117). Moreover, the relevance of synaptotagmin-1 rings is supported by the effects caused by a mutation (F349A) that disrupts ring formation on liposome fusion assays in vitro and on neurotransmitter release in neurons (119, 151). However, F349 is intimately involved in forming the tripartite interface between the synaptotagmin-1 C₂B domain and the SNARE complex (194). Hence, the physiological effects of the F349A mutation might arise because of disruption of the tripartite interface.

Clearly, further research will be required to establish whether the synaptotagmin-1 rings occur in vivo. If they do, ring formation (not shown in Figure 1 for simplicity) can cooperate with the inhibitory activity arising from the synaptotagmin-1-SNARE-complexin-1 assembly to prevent premature release. After Ca²⁺ disrupts this assembly (and the putative oligomeric ring), synaptotagmin-1 may cooperate with the SNAREs in inducing membrane fusion by helping bridge the membranes and/or inducing membrane curvature (3, 94). It is also plausible that synaptotagmin-1 accelerates fusion by perturbing the lipid bilayers upon insertion of its Ca²⁺-binding loops, much as the Sec17 hydrophobic loop likely stimulates vacuolar fusion (169). Reconstitution experiments suggest that, in addition to initiating membrane fusion, synaptotagmin-1 plays a key role in formation and expansion of the fusion pore (37, 81). However, the mechanism underlying the role of synaptotagmin-1 in fusion remains as one of the most crucial enigmas in this field.

I would like to thank all the members of my laboratory, as well as Thomas Sudhof, Christian Rosenmund, Diana Tomchick, Shuzo Sugita, William Wickner, Reinhard Jahn, Axel Brunger and Yeon-Kyun Shin, for many fruitful discussions, and Guillaume David, Klaudia Jaczynska, Axel Brunger and William Wickner for insightful comments on the manuscript. Work on this field in my laboratory is supported by grant I-1304 from the Welch Foundation and NIH Research Project Award R35 NS097333.