Molecular mechanisms of fast neurotransmitter release

1. Introduction

During synaptic transmission, Ca²⁺ influx into the presynaptic terminal triggers neurotransmitter release. This process involves sensing Ca²⁺, and subsequently fusing neurotransmitter-filled synaptic vesicles with the presynaptic membrane in less than a millisecond (122, 139). In the neuron, neurotransmitters are released “synchronously” upon action potential arrival in the synaptic terminal, “asynchronously” with a certain delay after an action potential, and “spontaneously” without an action potential; each of these processes likely involves specific sets of synaptic proteins (69). Many, if not most, of the key factors of the core synaptic fusion machinery have been identified, including SNAREs (Soluble N-ethylmaleimide sensitive factor Attachment protein REceptor), NSF (N-ethylmaleimide-sensitive factor), SNAP (soluble NSF adaptor protein), synaptotagmin (abbreviated as Syt), complexin (abbreviated as Cpx), Munc18 (mammalian uncoordinated-18) and Munc13 (mammalian uncoordinated-13). Yet, important questions remain: how is the process tightly regulated by Ca²⁺, and how is it triggered within less than a millisecond, much faster than any other biological membrane fusion process? We will first summarize the current knowledge about each individual factor in an encyclopedic fashion (Sections 1–7), and then discuss recent insights into the cooperation and interplay between these factors for neurotransmitter release (Sections 8–10). The reader may wish to skip to these sections first and consult the sections on the individual factors when more information is needed.

2. SNAREs

In this review, we focus on the neuronal SNAREs and their most extensively studied isoforms. Prior to membrane fusion, synaptobrevin-2 (also called VAMP2 – Vesicle Associated Membrane Protein 2) on the synaptic vesicle, and syntaxin-1A and SNAP-25A on the plasma membrane initially form a trans SNARE complex with the transmembrane domains of synaptobrevin-2 and syntaxin-1A in their respective membranes (Fig. 1). Moreover, individual syntaxin-1A molecules form clusters in the plasma membrane that are segregated from phosphatidylinositol-4,5-bisphosphate (PIP2) domains (149). During fusion, the SNARE complex completely zippers into the fully assembled cis SNARE complex where the soluble core consists of a parallel four α-helix bundle (144). The synaptobrevin-2 and syntaxin-1A helices of the cis SNARE complex extend into the merged membrane (136). The topology of the core of the SNARE complex is a left-handed four-helix coiled coil where the core consists of layers of primarily hydrophobic residues, except for an ionic layer at the center of the complex consisting of three glutamine amino acid residue and one arginine residue (referred to as ionic or zero layer) (39, 144). In addition to this parallel four α-helix bundle configuration, SNAREs can form a variety of non-canonical stoichiometries and configurations in vitro (reviewed in ref. (17)), and we will discuss potential physiological roles of these non-canonical assemblies in Section 10.

The assembly of the SNARE complex is thought to provide the energy necessary for membrane fusion (144, 158). Single-molecule optical and magnetic trap pulling experiments suggest that the free energy that is released by the zippering of one SNARE complex is approximately 36 k_BT (42, 103), and that a half-zippered intermediate exists under “loaded” conditions (i.e., when the juxtamembrane regions of the SNARE complex are steadily pulled by a force apparatus), suggesting that the trans SNARE complex may consist of a partially zippered complex. However, we note that direct imaging of the trans SNARE complex has not yet been achieved. Nevertheless, the estimated free energy that is released by formation of one SNARE complex is somewhat less than the energy that is required to overcome the hydration-force barrier for the formation of a lipid stalk (i.e., the dehydration energy of ~40–90 k_BT that is required to induce stalk formation) (2). Additional energy may be required for pore formation and expansion (28). Consistent with this notion, SNAREs mediate membrane exchange when reconstituted into proteoliposomes (158). As little as one SNARE complex (148) is sufficient for ensemble lipid mixing of proteoliposomes, but at least two synaptobrevin-2 molecules, and presumably two SNARE complexes, are required for fast Ca²⁺-triggered exocytosis (131). Although these lipid mixing experiments demonstrate that SNAREs mediate membrane exchange, content mixing is the proper correlation for neurotransmitter release since lipid mixing can occur in the absence of full fusion due to metastable hemifusion diaphragms (35, 54, 78). Moreover, dequenching of the fluorescent dyes used in lipid mixing assays may occur due to other phenomena (186). The first genuine content mixing experiment for precisely timed Ca²⁺-triggered fusion with synaptic proteins was reported by (78, 79) (Fig. 2) (for a review of other assays and more details see (18)). Over the years, this single-vesicle fusion assay has progressed to include more molecular components with the goal of a more complete physiological reconstitution (35, 80–82) (Section 10).

3. Synaptotagmin

In addition to SNAREs, synaptotagmins (Syts) constitute an evolutionary conserved family of proteins (105) that are composed of an N-terminal single transmembrane-spanning domain, a variable juxtamembrane linker and two C-terminal cytoplasmic C2 domains, termed C2A and C2B, respectively (112). In this review we focus on synaptotagmin-1 (Syt1), which resides on the synaptic vesicle and is vital for synchronous Ca²⁺-triggered synaptic vesicle fusion (41, 43).

Syt2 and Syt9 are also involved in Ca²⁺-triggered synchronous neurotransmitter release for different subsets of neurons (167), while Syt7 mediates asynchronous release (5, 160). In addition, Syt1 and Syt7 together act as redundant Ca²⁺ sensors for neuroendocrine exocytosis (125, 141), and they also participate in postsynaptic AMPA receptor exocytosis during long term potentiation (LTP) (164).

Syt1 interacts with anionic membranes and SNARE complexes in both Ca²⁺-dependent and Ca²⁺-independent manners (7, 15, 16, 22, 26, 31, 41, 73, 77, 111, 152, 156, 184). Syt1 function and membrane binding is specific to Ca²⁺, i.e., other divalent cations, such as Sr²⁺, only weakly trigger synaptic vesicle fusion (38). Under certain experimental in vitro conditions, (i.e., in the presence of Ca²⁺), the Syt1 C2B domain or the C2A-C2B fragment binds to curved membranes or favors deformation of the membrane (60, 93), along with membrane clustering (3). Computer simulations support the notion that both C2 domains cooperate to induce membrane bending (165). Moreover, the linker between the C2 domains is important for Syt1 function in neurons (6, 87), and both C2 domains can insert into the same membrane at the same time upon Ca²⁺-binding in vitro (59). Additionally, the juxtamembrane linker domain—the domain between the transmembrane domain and the C2 domains—is important for Syt1 function in fusion assays (83) and in neurons (85). Thus the relative locations of the C2 domains are critical for proper function, but the precise location of Syt C2 domains before and after Ca²⁺-triggering remains to be directly visualized.

While mutations in the Ca²⁺-binding sites of the Syt1 C2A domain can partially rescue Syt1-deficient mammalian neurons, mutations in the Ca²⁺-binding sites of the Syt1 C2B domain cannot (107, 130, 168). The mutations in the Syt1 C2B domain not only block Ca²⁺-evoked synchronous release in an intact synapse, but also dominantly inhibit the ability of endogenous wild-type Syt1 to trigger fusion, consequently increasing spontaneous miniature release when co-expressed (84, 164, 185).

SNAREs and Syt1 alone are sufficient to promote Ca²⁺-triggered proteoliposome lipid mixing (147). However, as mentioned above, lipid mixing does not necessarily correlate with neurotransmitter release since lipid mixing can occur without content mixing. While SNAREs and Syt1 can indeed promote full fusion (i.e., content mixing) upon Ca²⁺-triggering (35, 78) with a ratio of Ca²⁺-dependent fusion to Ca²⁺-independent fusion of ~ 10 (82), Ca²⁺-triggered fusion with this minimal system is relatively inefficient. As discussed in Section 10, a more complete reconstitution greatly increases the efficiency and synchrony of Ca²⁺-triggered fusion.

The first crystal structure of a Syt C2 domain (143) revealed that there are no large conformational changes upon Ca²⁺ binding. Rather, the C2 domain acts as an electrostatic switch where Ca²⁺-binding neutralizes the negative charges in that region as suggested by NMR experiments (127). Structures of complexes between Syts and the SNARE complex have been extremely difficult to obtain due to the multiple binding modes of the components of the complex (15, 26), and it had been questioned if SNARE-Syt interactions are functionally important (110). It is only recently that crystal structures of SNARE/Syt1 and SNARE/Cpx1/Syt1 complexes have been obtained and functionally validated (184, 185) (Section 9).

4. Complexin

The cytoplasmic protein complexin (Cpx) also plays critical roles in neurotransmitter release. Specifically, synchronous evoked neurotransmitter release depends on Cpx (99), and this activating role of Cpx is conserved across all species and different types of Ca²⁺-induced exocytosis studied to date (21, 55, 61, 67, 94, 97, 119, 173, 175, 177). Cpx also regulates spontaneous release, although this effect is less conserved among species and experimental conditions: for example, in Drosophila, spontaneous release increases with knockout of Cpx compared to wildtype neurons (65, 171). Likewise, knockdown in cultured cortical mouse neurons increases spontaneous release compared to wild-type neurons, although knockout of Cpx only affects spontaneous release depending on the particular neuronal cell type (67, 97, 146, 175).

We focus here on the homolog complexin-1 (Cpx1) of the mammalian Cpx family whose primary sequence is highly conserved (96%) in mouse, rat, and human. Cpx1 consists of four domains: the N-terminal domain is important for activation of synchronous Ca²⁺-triggered release in murine neurons (97, 170, 172) and in isolated chromaffin cells (33); the accessory domain regulates spontaneous release (97); the central domain is required for all functions of Cpx1 and it binds with high affinity (~10 nM) to the neuronal ternary SNARE complex (24, 99, 109); and the C-terminal domain is involved in vesicle priming and binds to anionic membranes in a curvature sensitive fashion (24, 47, 67, 97, 133).

Structurally, in isolation, both the N- and C-terminal domains of Cpx1 are largely unstructured, while the accessory and central domains have α-helical propensity (108). The central domain of Cpx1 binds to the groove formed by the synaptobrevin-2 and syntaxin-1A α-helices in the SNARE complex (14, 24). This interaction is anti-parallel, i.e., the direction of the α-helix of the accessory domain of Cpx1 is anti-parallel to the direction of the α-helices in the ternary SNARE complex.

The accessory domain of Cpx1 is not required for activation of Ca²⁺-triggered single-vesicle fusion with reconstituted neuronal SNAREs and Syt1 (81), and mutations of this domain have no effect on evoked release in neurons (176). However, mutations of the Cpx1 accessory domain affect spontaneous release (25, 176), and elimination of the accessory domain increases Ca²⁺-independent single-vesicle fusion (content mixing) (81) compared to wild-type control. In another study, introduction of an α-helix breaking mutation into the accessory domain of Cpx1 increased spontaneous release in C. elegans (117).

Functional studies, along with molecular modeling suggested that Cpx1 may inhibit full ternary SNARE complex formation by preventing the C-terminal part of synaptobrevin-2 from binding to the syntaxin-1A/SNAP-25A subcomplex of the ternary SNARE complex (45, 46). However, a subsequent crystal structure of a mutant of Cpx1 (D27L, E34F, R37A) in complex with a partially truncated SNARE complex (containing a truncated synaptobrevin-2 fragment), along with isothermal titration calorimetry (ITC) and light scattering experiments instead found that Cpx1 bridges two partially truncated SNARE complexes: one partial SNARE complex binds to the central domain of Cpx1 while another partial SNARE complex binds weakly to the accessory domain of the same Cpx1 molecule (PDB IDs 3RK3 and 3RL0) (75, 76). Although this weak interaction of the accessory domain is observable by ITC (75, 116), it was not observed by NMR (146), perhaps due to differences in constructs (75, 116). Finally, single-molecule experiments demonstrated that the accessory domain also weakly interacts with the syntaxin-1A/SNAP-25A (binary) complex (27). Moreover, these single-molecule studies showed that both binding modes of the Cpx1 accessory helix are possible, reconciling the previous results. However, the biological context of these weak interactions involving the accessory domain remains to be elucidated.

The C-terminal domain of Cpx1 binds to phospholipids (67, 126) and it is important for vesicle priming in neurons (33, 34, 67). Moreover, Cpx1 without the C-terminal domain does not reduce spontaneous release, but it still activates Ca²⁺-triggered release in neuronal cultures (67) and in a reconstituted system (82). The C-terminal domain is sensitive to membrane curvature, and it may thus localize Cpx1 to the synaptic membrane (47, 133, 163). Interestingly, the curvature sensing value of the C-terminal domain is comparable to that of α-synuclein (47, 63), i.e., a value that corresponds to the relatively high curvature of synaptic vesicles with a typical diameter of 40 nm. In support of this notion, replacement of the C-terminal domain with plasma membrane-localizing elements, results in increased spontaneous neurotransmission and a prolonged synchronous decay time-constant of NMDA-type glutamate receptor evoked postsynaptic currents in cultured neurons (47).

In conjunction with neuronal SNAREs and Syt1, Cpx1 increases the Ca²⁺-triggered amplitude and synchrony of proteoliposome fusion, and it also suppresses Ca²⁺-independent single vesicle fusion (content mixing) (82), resulting in a 50–100-fold ratio of Ca²⁺-triggered fusion to Ca²⁺-independent fusion. Moreover, the functions of the four domains of Cpx1 have been reproduced in vitro (81, 82). By varying the Cpx1 concentration, these single-vesicle studies suggest that the regulatory effect on Ca²⁺-independent fusion and the facilitating role on Ca²⁺-triggered fusion are governed by distinct molecular mechanisms involving different subsets of the four domains of Cpx1. Similarly, genetic experiments in Drosophila also suggest distinct mechanisms for activation of fast synchronous release and regulation of spontaneous release (25).

5. Munc18

Sec1/Munc18 (SM) proteins are also required components of all membrane trafficking pathways as exemplified by the block of synaptic vesicle fusion upon Munc18–1 knockout in mice (151). Following disassembly of SNARE complexes with NSF and SNAP, Munc18–1 captures syntaxin-1A, locking it into a “closed” conformation and preventing reassembly of the ternary SNARE complex (91). Munc18–1 binds tightly (K_d ~1–4 nM) to syntaxin-1A in which the N-terminal three helix bundle of syntaxin-1A (H_abc domain) forms a cis conformation with the syntaxin-1A SNARE motif (H3 domain) (19, 23, 104), in turn hindering SNARE complex formation (114, 174).

Moreover, Munc18 co-localizes with syntaxin and SNAP-25 in fixed cultured neurons and forms nano-clusters in the plasma membrane (113). Interestingly, in neuroendocrine cells, transfection of syntaxin alone results in its mislocalization, whereas co-transfection of both Munc18 and syntaxin results in proper trafficking to the plasma membrane (100, 123). Thus, one function of Munc18 may be to stabilize or “chaperone” syntaxin, which also agrees with the aforementioned knockout studies of Munc18 in mice that showed a reduced level of syntaxin-1 in the plasma membrane (13). The crystal structure of the Munc18–1/syntaxin-1A complex revealed extensive interactions of syntaxin-1A in the closed conformation with the concave surface formed by domains 1 and 3 of Munc18–1 (104), explaining the inability of the syntaxin-1A H3 domain in this closed conformation to interact with SNAP-25A and synaptobrevin-2.

Binding of Munc18–1 to the closed conformation of syntaxin-1A is not always required, as the syntaxin-1A H_abc domain is not essential for evoked synaptic vesicle fusion in cultured syntaxin-1A-deficient neurons, although it is required for spontaneous synaptic vesicle fusion (183). In marked contrast, the N-terminal residues of syntaxin-1A are required for binding of the ternary SNARE complex to Munc18–1, and may thereby serve as a recruiting factor for Munc18–1 (118). Munc18–1 binds to the assembled ternary SNARE complex via the syntaxin-1A N-terminal residues, with syntaxin-1A in an “open” conformation (128) (32, 58, 169). In this open conformation, the H3 domain of syntaxin-1A interacts with its partner SNARE motifs to form the SNARE complex (144), while the syntaxin-1A H_abc domain is presumably flexibly linked to the four-helix core of the SNARE complex (51, 106).

6. Munc13

Munc13–1 and its homologues are primarily brain-specific, cytoplasmic proteins in the presynaptic terminal that are involved in synaptic vesicle priming and short-term synaptic plasticity (138, 139). Munc13–1 is expressed in all neurons of the rat central nervous system, while Munc13–2 and Munc13–3 are mostly restricted to different brain regions. The ubiquitously expressed splice variant ubMunc13–2 and the Munc13–4 isoform are predominantly expressed in non-neuronal cells (72). Munc13–4 only includes the C2B, MUN, and C2C domains (Domain functions will be detailed below). The UNC-13 and Dunc-13 homologs of Munc13–1 lack the N-terminal C2A domain, as do a brain-specific splice-variant of bMunc13–2 and the Munc13–3 isoform. Here we will focus on the Munc13–1 isoform.

Early studies found that neurons in mice lacking Munc13–1 exhibit severely limited activity-dependent as well as hypertonic sucrose-induced neurotransmitter release (4). Moreover, deletion of Munc13–1 results in substantially larger distances between synaptic vesicles and active zones as observed by electron microscopy (50, 62, 159). The priming phenotype of Munc13–1 knockout is much stronger than that of other priming factors, including Cpx1 (175), i.e., the size of the readily releasable pool of synaptic vesicles is much smaller in Munc13–1 knockout mice compared to Cpx1 knockout mice. Moreover, Munc13–1, together with Munc18–1, prevents NSF-dependent de-priming of synaptic vesicles (53). Finally, Munc13–1 and −2 double-knockout mice have essentially no release-competent synaptic vesicles (150) and exhibit a severe docking defect (62).

Below, we briefly summarize structural and functional studies for each domain of Munc13–1. Munc13–1 consists of a C1 domain (C1), a calmodulin-binding domain, two C2 domains (C2A, C2B), and a MUN domain, and another C2 domain (C2C). The crystal structure of the C1C2BMUN fragment of Munc13–1 revealed inter-domain interactions between the C1, C2B, and MUN domains and a folded linker region between the C1 and C2B domains (166). An electron cryo-tomography (cryoET) study of vesicles with bound C1C2BMUN revealed filamentous features that suggest hinge bending of this fragment, with the straight conformation being consistent with the crystal structure (44).

Functionally, the C2A domain of Munc13–1 interacts with the Rab3 effector RIM, forming a tripartite complex (36) which is important for the priming function of Munc13–1 and for the presynaptic localization in mice (68). Likewise, the C2A domain of UNC-13 in C. elegans is important for regulating both evoked and spontaneous release, as well as its precise localization to the active zone (182).

Munc13–1 and the ubiquitously expressed isoform Munc13–2 contain a conserved calmodulin-binding domain. When calmodulin binding is disrupted by mutation, evoked release upon a single action potential is unchanged, but short-term plasticity and Ca²⁺-dependent modulation of the readily-releasable pool is reduced (66).

The diacyglycerol (DAG)-phorbol C1 binding domain regulates Munc13–1 functions in short-term synaptic plasticity and priming of synaptic vesicles, although mutation of the site does not affect isolated evoked neurotransmitter release (120). The C2B domain binds to phospholipids in a Ca²⁺-dependent fashion, with particular specificity for phosphatidylinositol phosphate (PIP) and phosphatidylinositol bisphosphate (PIP2). Additionally, an interaction between the C2B domain and presynaptic voltage-gated calcium channels modulates use-dependence of the channels (20). Interestingly, deletion of the C2B domain enhances Ca²⁺-triggered exocytosis in C. elegans neurons, suggesting an auto-inhibitory function of this domain (102).

The MUN domain of Munc13–1 has two major roles in vesicle fusion: catalyzing the transit of syntaxin-1A from the syntaxin-1A/Munc18–1 complex into the ternary SNARE complex (10, 91, 178) and promoting the proper assembly of the SNARE complex, in particular the parallel configuration of syntaxin-1A and synaptobrevin-2 (80). Supporting this latter role, the so-called LE mutant of syntaxin-1A (syntaxin-1A^LE) (37) bypasses the requirement of Munc13–1 for the transit of syntaxin-1A into the ternary SNARE complex in presence of SNAP-25A and synaptobrevin-2 in vitro (90). It accomplishes this by preferentially adopting an open conformation of syntaxin-1A (155). However, this open syntaxin-1A mutant can only partially restore neurotransmitter release in unc-13 mutants of C. elegans (50), because it does not promote the proper assembly of the SNARE complex, i.e., SNARE complexes with the syntaxin-1A LE mutant are scrambled (80); this observation explains the poor rescue of Munc13 deletion by this mutant. Earlier studies in C. elegans suggest a similar function of full-length unc-13 (121), although the role of the MUN domain has not been directly studied in that system.

Finally, the C2C domain of Munc13–1 also plays a functional role in evoked release in chromaffin granule secretion and neurotransmission in C. elegans (92, 137), and in evoked release in priming in mouse neurons (88). It also enhances ensemble lipid and single-vesicle content mixing experiments (80, 88).

Similar to Munc13, CAPS proteins are members of an extended Munc13/CAPS family of proteins that are also important for priming of synaptic vesicles (64). They contain one Munc13 homology domain (MHD) that comprises part of the MUN domain, and a pleckstrin homology domain that interacts with PIP2. It is likely that the MHD domain in CAPS proteins is partially responsible for their function in synaptic priming (30). Both Munc13s and CAPS proteins are probably jointly important for synaptic vesicle priming. In support of a partially redundant function of both molecules, the C1C2BMUN fragment of Munc13 can substitute for deletion of CAPS in Ca²⁺-triggered dense core vesicle fusion with a planar supported bilayer (74).

7. NSF and SNAPs

Historically, the ATPase NSF was first discovered as a key factor for inter-compartmental transport in eukaryotic cells and, subsequently, SNAREs were found to be substrates for NSF (9, 134). Independently, SNAREs were discovered to be important for neurotransmitter release (11, 12, 140). In concert with the adaptor protein, SNAP, the ATPase NSF disassembles the SNARE complex into individual proteins upon ATP hydrolysis (51, 98, 135). NSF is a member of the AAA+ family consisting of two ATPase rings (known as Type II AAA+), and an N-terminal domain. Most eukaryotic organisms encode only one NSF gene. By contrast, there are three homologues of SNAP, with αSNAP being the most widely studied. Although earlier work suggested that disruption of the SNARE complex by ATP hydrolysis drives fusion (135), subsequent work clarified that it is actually the SNARE complex formation that drives fusion (51, 98). For more structural details of NSF and SNAPs we refer to a recent review (180).

8. Cooperation between Syt1, Cpx1 and SNAREs for evoked release

9. Implications for spontaneous release

Syt1, Cpx1, and SNAREs are also important for spontaneous release (33, 65). More specifically, both the primary SNARE/Syt1 and tripartite SNARE/Cpx1/Syt1 interfaces are required for the suppression of spontaneous release in neurons, i.e., elimination of either interface resulted in an increase in mini frequency (185). Why is there an increase in spontaneous fusion frequency when either interface is disrupted? At a structural level, it could simply mean that membranes can get closer when one or more components of the SNARE/Cpx1/Syt1 complex are missing, consistent with the closer membrane separation distance observed by cryoET of proteoliposomes when Cpx1 is absent (44). However, there is an observation that counters this simple membrane separation distance argument: the probability of Ca²⁺-independent fusion did not increase when Syt1 is absent in single-vesicle fusion experiments with SNAREs and Cpx1 only (82). Despite this, these experiments should be repeated with full reconstitution (80), since in the absence of full reconstitution, improperly formed SNARE complexes may obscure the more subtle effect of suppression of Ca²⁺-independent fusion.

We would like to emphasize that Ca²⁺-independent fusion of single-vesicle experiments cannot strictly be compared with spontaneous release in neurons. First, Ca²⁺-independent fusion probabilities are normalized to the number of associated vesicles. Second, the frequency of spontaneous release measured in electrophysiological experiments depends on both the number of functional synapses as well as the number of synaptic vesicles that are capable of undergoing spontaneous fusion. Third, there is likely another Ca²⁺-sensor that competes with Syt1 in binding to the tripartite interface (142). Elimination of the primary interface or presence of the certain Syt1 C2B mutants would affect the binding equilibrium, possibly explaining the increase of spontaneous release in neurons (185).

10. Cooperation between Munc18 and Munc13 for proper SNARE assembly

11. Perspectives and Outlook

There are other important factors that were not reviewed here. In particular, factors that are involved in tethering, regulation through Rab G-proteins, and voltage gated Ca²⁺ channels (68, 157). Clearly, complexes with these factors also need to be imaged at near atomic resolution, and functionally reconstituted. Moreover, there are many proteins in synaptic vesicles with little known function that may deserve closer examination. Ultimately, studying the effects of disease and aging on the synaptic neurotransmitter release machinery and its regulations may provide clues for entirely new therapeutic developments.