Turbocharging synaptic transmission: 12 SNAREpins are required for rapid release of reconstituted synaptic vesicles

Abstract

Synaptic transmission occurs synchronously with real-world events, far faster than vesicle fusion for hormone release or membrane biogenesis, all mediated by soluble N-ethylmaleimide–sensitive factor attachment protein receptor (SNARE) complexes. How SNAREs cooperate to achieve synchronous neurotransmitter release is a long-standing mystery. Rapid release (<7 milliseconds) has been reconstituted from purified synaptic SNAREs, SNARE-assembling chaperones, and calcium ion sensors in a fully-defined, genetically validated system that enables single-molecule counting in docked vesicles before release. SNAREpin complexes (12 ± 0.3) are found in each such ready-release vesicle, suggesting a regular structure. Several genetic conditions (including point mutation of the synaptic vesicle protein Synaptophysin from a Synaptopathy patient and human and mouse disease mutations of the synaptic vesicle protein vesicle-associated membrane protein-2 (VAMP2) reduce the number of SNAREpins to 6 ± 0.3 and result in profoundly delayed release over 0.1 to 1 seconds. Omitting Synaptophysin, whose hexamers preassemble 12 copies of VAMP2, also yields ~6 SNAREpins and delays release.

Synaptophysin-VAMP2 interfaces control SNAREpin number, regulating synchronous complete release versus fusion pore flickering.

INTRODUCTION

Synaptic transmission must operate on millisecond timescales to match real-world demands, requiring the release of neurotransmitters at even faster speeds. Synaptic vesicles (SVs) containing the neurotransmitters undergo an elaborate “priming” process to prepare them for the necessary ultrafast (<1 ms) release, which occurs following exposure to Ca²⁺ ions. While the proteins needed for the priming process are well known, how they organize and how they achieve ultrafast (“turbo-charged”) release in structural and biophysical terms is still largely unknown.

Some of the priming proteins are dedicated chaperones [Munc (Mammalian uncoordinated)-13, Munc18, and Synaptophysin (Syp)] that initiate the assembly of synaptic soluble N-ethylmaleimide–sensitive factor attachment protein (SNAP) receptor (SNARE) complexes (composed of VAMP, Syntaxin, and SNAP25) into four-helical bundles termed SNAREpins, which then further self-assemble (“zipper”) to ultimately fuse the vesicle to the plasma membrane (PM). Other priming proteins are specialized clamps and sensors (Complexin and Synaptotagmin), which capture and stabilize the highly energized near-release (“primed”) state (1). Clamps thus prevent release before Ca²⁺ ions arrive but also trigger release when they do (2, 3).

The simple fact that neurotransmitter release occurs at least 10² to 10³ times faster than other forms of membrane fusion suggests that an unusually large amount of potential energy derived from SNARE assembly is stored in the confined interface of the SV with the PM, thereby “turbocharging” release relative to ordinary fusion (1). The structure of the protein assembly at this interface is therefore of great interest.

Cryo–electron microscopy (cryo-EM) tomography, averaging hundreds of SV–membrane interfaces, revealed a regular sixfold symmetric ringlike organization of protein density (4, 5). In particular, the sixfold rotational symmetry suggested that there is a fixed stoichiometry among the known constituents of the interface (SNAREpins, chaperones, and clamps) in which each is present in a multiple of six copies. This symmetry, along with the structures of separately characterized component complexes—Synaptotagmin-SNAREpin (6), Complexin-bridged paired SNAREpins (7), and Ca²⁺-sensitive Synaptotagmin rings (8, 9)—has enabled a highly constrained molecular model of the interface to be constructed (1, 10). This (“buttressed rings”) model (Fig. 1A) specifically predicts that the interface contains a single hexamer of Syp (anchored in the vesicle) at its center, from which emanate 12 SNAREpins organized as six pairs, each pair being placed symmetrically upon an inner ring of 18 copies of Synaptotagmin, which is, in turn, surrounded by an outer ring of six copies of Munc13 (Fig. 1, B and C).

Fig. 1. A buttressed ring model.

(A) Two concentric rings of SNARE chaperons, Munc13 and Synaptotagmin-1 (Syt1) C2Bs (top view). Six MUN domains (gray) are hypothetically arranged on the PM end-to-end manner using its intrinsic 60° bend. The arc length of this Munc13 covers ~21 nm (solid black arc), which can cover only three Syt1 C2B molecules (black dashed arc). Therefore, a ring of six Munc13 molecules encompasses a Syt1 ring of 18 C2B domains (pink and light purple), where every third Syt1 C2B (light purple) would make the primary binding interface. (B) From the SV side, Syp hexamer (olive green), consisting of 12 VAMP2 (blue) molecules (yet to form SNAREpin), is arranged on the top of the Munc13-Syt1 ring (PM side), where each vertex of Syp contains a VAMP2 dimer (top view). These VAMP2 molecules will transition to C upon SNAREpin formation. (C) We suggest that these 12 VAMP2 molecules are available to form 12 SNAREpins (a four-helical bundle consisting of one Syntaxin1A, SNAP25, and a VAMP2). Six SNAREpins (dark green) are directly bound to Syt1 C2B, which we refer to as central SNAREpins, whereas the rest six SNAREpins (light green) are directly bridged to central SNAREpin via complexin (cyan) and are referred to as peripheral SNAREpins. This molecular model was generated using the crystal structures, 5W5C (79) and 3RL0 (7). We removed the transmembrane domains (TMDs) of SNAREs and Syt1 for clarity.

In addition to predicting precise geometry and stoichiometry, the buttressed ring model suggests that the SNAREpins in each pair are asymmetrically positioned, a “central” SNAREpin closer to the incipient fusion pore and its conjugate “peripheral” counterpart further out radially. Each central SNAREpin is directly bound to every fourth Synaptotagmin in the ring, and their Syntaxin transmembrane domains (TMDs) are confined within the ring. The model predicts that the TMDs of the paired peripheral SNAREpins are located outside the ring (1).

The central SNAREpins are predicted to be directly clamped against further zippering by attachment to the Synaptotagmin ring; they are directly gated by Ca²⁺, which releases them when it dissociates the ring (10). The peripheral SNAREpins are indirectly clamped by the bridging Complexin molecule, which emanates from the surface of the central SNAREpin and inserts its accessory helix into the zippering end of the peripheral SNAREpin (7). When Ca²⁺ releases the central SNAREpin, the mechanically attached accessory helix will be removed from the peripheral SNAREpin, unleashing its zippering force. We speculated that the predicted six peripheral SNAREpins are directly responsible for the turbocharge (1). The central SNAREpins, simply by virtue of their inner positioning, will complete zippering first and thereby open the fusion pore. The peripheral SNAREpins will not open the pore but can exert forces to accelerate its opening and expansion (1).

Our goal in this work is to test these predictions and, by implication, the closely related buttressed ring hypothesis. To accomplish this, it was necessary to simultaneously measure SNAREpin stoichiometry in docked vesicles, to control and vary the resulting number of these SNAREpins, and to measure the effects on the kinetics of release. This kind of precision is beyond the reach of current technologies available for physiological systems and can only be achieved in the context of fully controllable, comprehensive cell-free functional reconstitution of vesicle priming and ultrafast release.

We recently reported the development and thorough genetic validation of such a fully defined cell-free reconstitution that contains only purified SNAREs, chaperones, clamps, sensors, and lipids (11–16). This system is assembled on a chip containing hundreds of microfabricated holes, including a giant unilamellar vesicle (GUV)–derived bilayer covering the holes to mimic the presynaptic PM, containing Syntaxin-Munc18, palmitoylated SNAP25, and phosphatidylinositol 4,5-bisphosphate (PIP₂). The chaperone Munc13 and Complexin are added in the buffer above the chip, along with small unilamellar vesicles (SUVs) containing physiomimetic copy numbers of VAMP2, Synaptotagmin, and Syp. They are fluorescently labeled in their bilayers with an Atto-phospholipid dye, in their internal volume with Sulforhodamine B (SRB), or both. This allows us to monitor vesicle docking, clamping, and release in real time (with a time resolution of 7 ms) and in a massively parallel fashion.

The aqueous volume below the suspended bilayer is refractive index matched to the chip to enable evanescent wave–based total internal reflectance fluorescence (TIRF) imaging. This enables the number of copies of a targeted fluorescently tagged protein bound to each docked vesicle to be simultaneously measured using single-molecule methods combined with TIRF (17, 18). Together, these measurements allow us to determine the precise protein content at the interface of primed vesicles and to relate the observed stoichiometry to the resulting kinetics of release on a single-vesicle basis. Using a free, “suspended” bilayer is an essential feature to enable ultrafast fusion.

Here, we compare the kinetics of Ca²⁺-triggered release of content from cell-free primed vesicles in the complete chaperone system containing either 12 (+Syp) or 6 (-Syp) SNAREpins. We also engineered mutations in VAMP2 and Syp that reduce the number of SNAREpins to ~6 by interfering with dimerization of VAMP via its TMD, and we identified naturally occurring disease-associated mutations in these proteins that have the same effect. Reducing the SNAREpin number from 12 to 6 systematically suppresses the turbocharge fast kinetics, which causes initial fusion pore opening to be delayed >100-fold, increasing up to ~1 s in some cases. Fusion pore expansion is similarly delayed. Unexpectedly, the fusion pores also appear to repeatedly flicker open and closed, reminiscent of some aspects of “kiss and run” behavior reported in certain physiologic experiments (19, 20), suggesting that fusion pore expansion is hindered.

RESULTS

Background

In the “complete” version of the reconstitution (Fig. 2A), which we use throughout this paper, the three synaptic SNAREs [VAMP2, Syntaxin1A (Stx1A), and SNAP25] are added separately from each other along with the chaperones (Munc13, Munc18, and Syp) and the fusion clamps (Synaptotagmin and Complexin). Our previous work (18) used a simplified system in which the chaperones Munc13 and Munc18 were omitted, and the requirements for them bypassed using preassembled Stx1A-SNAP25 (t-SNAREs). In this simplified system, there were also 12 ± 1 SNAREpins per vesicle when Syp was included in the vesicles. Without Syp, the number of SNAREpins varied widely, ranging from 5 to 35 per vesicle. This result was readily explainable because, mechanistically, Syp can cooperatively chaperone 12 SNAREpins as a group because it self-assembles into hexamers, in which each monomer binds two copies of the VAMP (18, 21–24).

Fig. 2. Syp templates 12 SNAREpins synergistically with Munc13.

(A) A schematic representation of the SNAREpin counting method using TIRF microscopy. SUVs labeled with 2% Atto 465 DOPE, containing Syt1 (not shown) and VAMP2 (blue) ± Syp (yellow four-helical) proteins, are docked in the presence of either dark (black) or Alexa Fluor 647–labeled Complexin-1 (Cpx1; red). The bilayer, labeled with 0.1% Atto 465 DOPE, includes Munc18 (blue sphere)–Syntaxin1A (red), palmitoylated SNAP25 (green), 1,2-dihexanoyl-sn-glycerol (DHG) (yellow), and Munc13L (blue). (B) Diagram illustrating the photobleaching of the vesicle and its associated Cpx1 molecules. When the labeling efficiency was below 25%, the photobleaching curve for Cpx1 revealed discernible steps. However, when labeling efficiency exceeded 25%, the photobleaching curve was better fit by an exponential decay model, as individual bleaching steps became indistinguishable. This allowed for the estimation of the number of Cpx1 molecules under the vesicle. (C) Histograms showing the number of Cpx1⁶⁴⁷ molecules underneath a vesicle at different labeling efficiencies (25, 50, and 85%). Mean ± SD of means are reported for each labeling efficiency. Red dashed lines denote the predicted binomial distribution (for N = 12) for the corresponding labeling percentage. Number of vesicles > 300, and number of independent experiments = 4 for each labeling percentage. (D) Plot showing the mean ± SD of the means for each labeling efficiency, along with previously published values (without Syp; 25, 50, and 75% labeling data were plotted; red) for comparison. All data were fitted with a linear model fixing the intercept at 0, and the final SNAREpin counts are shown with mean ± error of the fit.

We also reported that when Syp is omitted from otherwise complete Munc13/Munc18-containing reconstitutions (25), only 6 ± 1 SNAREpins were assembled under each stably docked vesicle. This result suggested that another chaperone (besides Syp) can independently control the number of SNAREpins to 6. That chaperone was suspected to be Munc13 because Munc13 is known to assemble into hexamers on lipid bilayers (26, 27). This was confirmed using a Munc13 mutant that blocks hexamer assembly, which prevented the six SNAREpins from assembling under the vesicles, although the mutation spares intrinsic chaperone activity (assembly of SNARE complexes in solution) (25). The simplest interpretation is that Munc13 alone assembles only the six central SNAREpins and that Syp further enables Munc13 to assemble an additional six peripheral SNAREpins. Our model (13) also predicts that vesicles primed without peripheral SNAREpins should lack turbocharge, resulting in a delayed fusion pore opening and expansion and therefore slower content release.

Stably docked vesicles assembling in complete reconstitutions contain 12 SNAREpins

Complexin [Complexin-1 (Cpx1)] binds half-zippered VAMP2 and Syntaxin1A in a 1:1 molar ratio, providing a proxy for SNAREpin formation (28–32). We labeled Cpx1 with Alexa Fluor 647 at its sole naturally occurring cysteine-105 using a maleimide-dye derivative. This enabled us to simultaneously visualize the docking of vesicles (labeled with 2 mol % Atto 465–DOPE; see Materials and Methods) and the binding of Cpx1 in complete reconstitutions (Fig. 2A). After 5 to 10 min at 27°C to enable vesicles to dock on suspended bilayers (see Materials and Methods), the unbound Cpx1 was washed out, and the number of Cpx1 molecules underneath each vesicle was quantified (Fig. 2B). Cpx1 was intentionally labeled with different molar efficiencies (25, 50, and 85%) for these experiments to facilitate subsequent analysis (Fig. 2C). With 25% labeling, individual molecules could be counted by step-photobleaching. For higher labeling efficiencies, photobleaching traces were fitted with exponential decay curves to calculate the starting total numbers [detailed in Bera et al. (18, 25)]. The mean ± SD for each distribution is plotted with its corresponding labeling percentage and linearly fitted, keeping the intercept at 0. The final SNAREpin per vesicle was obtained from the slope of the linear fit. In the presence of Syp (Syp), this value was calculated to be 12 ± 0.3 (mean ± error of the fitting), yielding the final SNAREpins per vesicle in this experiment (Fig. 2D, squares), consistent with the anticipated value of 12. As expected, when Syp was omitted from the reconstitution, 6 ± 0.3 SNAREpins were found per docked vesicle (Fig. 2D, circles).

From the perspective of the buttressed ring model (Fig. 1A), these represent the six Synaptotagmin ring–gated central SNAREpins because (as will be seen) release is still Ca²⁺ dependent. The assembly of these six SNAREpins was previously shown to depend on the ability of Munc13 to self-assemble into lateral hexamers as distinct from the intrinsic chaperone activity of Munc13 monomers (25). These vesicles containing only six SNAREpins were less stably docked than their 12 SNAREpin counterparts (fig. S1A), suggesting that the peripheral SNAREpins (present only when Syp was added) also stabilize the central SNAREpins, perhaps due to the bridging Complexin. DHG (1,2-dihexanoyl-sn-glycerol) enhanced the stability of vesicle clamping (fig. S1B). Upon addition of Ca²⁺, these clamps were released, resulting in fusion as monitored by lipid mixing. Fusion probability increased linearly with Ca²⁺ concentration, reaching ~90% at 100 μM Ca²⁺, and was independent of Syp (fig. S1C).

We observed that complexin is recruited within ~100 ms of vesicle docking, with comparable binding kinetics across vesicles despite variability in docking times, indicating that complexin binding is not rate limiting (fig. S2). Moreover, complexin recruitment is largely independent of vesicle size, particularly in the presence of Syp (fig. S3).

In addition to quantifying Cpx1, we introduced soluble SNAP25 counting as an independent proxy for estimating SNAREpin numbers. SNAP25 is an integral component of the SNARE complex and associates with each SNAREpin at a 1:1 molar ratio (25). Using a similar labeling and analysis strategy (see Materials and Methods), we measured an average of 6.2 ± 0.7 SNAP25 molecules per vesicle in the absence of Syp and 10.4 ± 1.6 SNAP25 molecules per vesicle in the presence of Syp (maleimide Alexa Fluor 647 labeling efficiency of 85%; mean ± SEM; fig. S4). The close agreement between these two independent counting approaches supports a similar number of SNAREpins per vesicle and is consistent with the existence of an organized SNAREpin architecture, potentially arising from the chaperoning activities of Syp, Synaptotagmin-1 (Syt1), and Munc13.

Synaptophysin and 12 SNAREpins enable turbocharged release

We thus identified two distinct conditions where SNAREpin numbers are tightly regulated, one producing 12 and the other 6 SNAREpins per vesicle. We next sought to test the hypothesis that SNAREpin numbers influence the kinetics of release. For this purpose, we used the dye SRB as a content marker. Vesicles loaded with SRB, prepared with or without Syp, were allowed to dock to suspended bilayers in otherwise complete reconstitutions (movie S1). Then, fusion was triggered by adding 100 μM (final) Ca²⁺. SRB was intentionally trapped in the vesicles at high concentrations (typically ~7 mM, corresponding to ~18,000 SRB molecules inside an average vesicle of ~100-nm radius) at which concentration the dye’s fluorescence is largely but not completely self-quenched (fig. S5). As a result, as the dye exits the fusion pore, the fluorescence of the pixel containing the vesicle will increase at first due to dequenching as the remaining dye concentration is reduced and then eventually decrease to below the starting level as SRB continues to exit after self-quenching no longer occurs. We measure the total time to complete the dequenching experimentally for each released vesicle, (∆t). The efficiency of trapping of SRB varies considerably among vesicles (even the same size vesicles). As a result, the dequenching profiles (change in fluorescence intensity) vary among individual release events. Therefore, the total fusion time (∆t) is the most reliable quantitative measure of release profiles.

Dequenching events were monitored by TIRF after triggering fusion with Ca²⁺ (Fig. 3, A and B). The high signal-to-noise ratio enabled us to observe content release kinetics at ~7-ms (frame rate of 148 Hz) time resolution. Vesicles containing Syp exhibited a sharp, single-frame fluorescent intensity increase, followed by a rapid decrease, indicative of complete fusion at or faster than the frame rate (Fig. 3B, top panel, and movie S2). In contrast, vesicles without Syp demonstrated a much slower, gradual, and occasionally stepwise increase and decrease in fluorescence intensity (Fig. 3, B and C, and movie S3). These kinetics suggest very narrow pore openings with slower expansion, continuous high-frequency pore flickering, or both. Approximately 85% of Syp-containing vesicles completed content release within a single frame (~7 ms), whereas vesicles lacking Syp required 35 ± 13 ms (∆t) for completion. In both cases, the probability of vesicle release (the fraction of vesicles responding to Ca²⁺) was very high, 0.88 ± 0.06 and 0.9 ± 0.05 in the presence or the absence of Syp, respectively.

Fig. 3. 12 SNAREpins turbocharge the release.

(A) A cartoon representation of a content release (SRB) experiment is shown. SUVs [containing VAMP2 (blue) and Syt1 (orange) ± Syp] loaded with SRB are docked on the bilayer in the presence of Cpx1 (black), Munc18 (blue sphere)–Stx1A (red), Munc13L (blue), and DHG (yellow). Fusion was triggered with the addition of 100 μM Ca²⁺ (final), while SRB dequenching was monitored with a 532-nm laser at 7-ms time resolution under TIRF. Total fusion time was calculated from the data depicted in the right-side panel. (B) Dequenching curves of the content release are shown. Two examples of vesicle fusion in each case (±Syp) are plotted together, depicting synchronous fusion. In the presence of Syp, SRB was dequenched completely within 7 to 14 ms (top panel), while in the absence of Syp, the dequenching event takes place over ~70 ms. The black dotted line indicates the arrival of Ca²⁺. a.u., arbitrary units. (C) Total fusion time was calculated from the dequenching starting to completion and plotted as histograms with 7-ms binning. Number of vesicles for each case (n) >150 and number of independent experiments (N) >4. The mean ± SDs of each distribution are reported.

To measure the precise time interval between the arrival of Ca²⁺ and the commencement of release, we included a lipid-linked Ca²⁺ sensor, Calcium Green C24, on the suspended bilayer as reported previously (fig. S6A) (11, 13). Simultaneous imaging of the calcium indicator and SRB release shows that the onset of SRB dequenching occurs precisely at the arrival of Ca²⁺ (fig. S6, B and C, and movie S4).

To further rule out the possibility that differences in release kinetics arise from variations in Ca²⁺ exposure, we used the fast and highly specific Ca²⁺ chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′ tetraacetic acid (BAPTA) (33, 34) to generate a brief pulse of free Ca²⁺ (Fig. 4A). Although vesicles at different spatial locations initially experience distinct Ca²⁺ concentrations, Ca²⁺ rapidly equilibrates within ~10 ms (Fig. 4B, inset). From this equilibrated state, buffering reduces free Ca²⁺ to ~1 μM in less than 1 ms (Fig. 4B; see Materials and Methods). Consequently, the overall timescale of Ca²⁺ exposure closely approximates that of a physiological Ca²⁺ transient (10 to 20 ms). Under these conditions of reduced total Ca²⁺ exposure, the overall probability of release was markedly lower (~20%) and independent of Syp (Fig. 4C, inset). The kinetic characteristics of the release events that did occur were unchanged (Fig. 4C); rapid/complete with Syp (12 SNAREpins) and slow/gradual without Syp (6 SNAREpins). We note that pulsatile exposure to Ca²⁺ can be thought of as mimicking the physiological situation of a single action potential, whereas continuous exposure (Fig. 3) more closely mimics the result of a train of action potentials. In both cases, release is profoundly delayed when Syp is absent and only six SNAREpins are present.

Fig. 4. Content release kinetics in the presence of fast Ca²⁺ chelator, BAPTA.

(A) Schematics of the content release experiment in the presence of Ca²⁺ chelator. SRB-loaded SUVs (containing VAMP2 and Syt1 ± Syp) were docked first in the presence of Cpx1, Munc13L, and DHG when the suspended bilayer also contained Stx1A-Munc18 and palmitoylated SNAP25. After stably docking these vesicles, 100 μM BAPTA (final concentration) was added. The fusion of docked SUVs was triggered with 100 μM Ca²⁺ while monitoring dequenching events under TIRF. (B) Numerical calculation (from equilibrated state) predicts that BAPTA can chelate out Ca²⁺ very rapidly, bringing the concentration to lower than 1 μM within 1 ms of exposure, resembling a single action potential pulse in the active zone. The inset figure depicts the local concentration of free Ca²⁺ as a function of time. Diffusion of Ca²⁺ was considered as the cylindrical model with various radii (0.6, 1.2, 1.8, 2.4, and 3.0 mm), and 100 μM Ca²⁺ was calculated to be readily quenched by BAPTA. See Materials and Methods for the detailed calculation. (C) Representative dequenching curves in the presence of BAPTA for +Syp (yellow) and –Syp (blue) vesicles. The green dotted line denotes the arrival of Ca²⁺. Time resolution: 7 ms. The inset shows the fusion probability of docked vesicles upon addition of 100 μM Ca²⁺ in the presence of an equimolar concentration of BAPTA. Vesicle numbers were 61 (+Syp) and 44 (−Syp), pooled from three independent experiments. Horizontal lines indicate the mean ± SD; n.s., not significant.

The nature of the alteration to the fusion pore in the absence of Syp will require further investigation with much better time resolution than the current optical monitoring permits. However, initial insights can be gleaned by comparing the kinetics of phospholipid and soluble content release from the same vesicles in the presence or absence of Syp (fig. S7A). In Syp-containing vesicles (fig. S7B and movie S5), lipid mixing and SRB dequenching were completed simultaneously within the first frame (~7 ms). In the absence of Syp (fig. S7C and movie S6), SRB dequenching and vesicle phospholipid (Atto 647 PE) efflux occurred gradually but with two distinct behaviors. For ~65% of the vesicles (example 1), much but not all of the lipid escapes immediately with little release of soluble content, and then the rest escapes gradually together with most of the SRB content. For ~35% of the vesicles (example 2), lipid and content are gradually released together from the outset. Example 1 is most simply interpreted as a form of transient hemifusion, followed by repeated rapid fusion pore opening and closing (flickering) with a generally higher frequency than the (147 Hz) frame rate while the vesicle remains in place. Example 2 is most simply interpreted as flickering without hemifusion. The characteristic times required for 99% content release (t₉₉) were calculated from the SRB dequenching curves in solution (fig. S8; see Materials and Methods) and were determined to be 8 ms in the presence of Syp and 48 ms in its absence.

We previously reported that Syp assembled 12 SNAREpins using a simplified preassembled t-SNAREs (18). We observed that 12 SNAREpins under this condition also produced similar turbocharged release (currently defined as within a single 7-ms time frame) (fig. S9).

12 SNAREpins are needed for the turbocharge

The loss of the turbocharge when Syp is removed is readily explained by the reduction in force when the number of SNAREpins is reduced from 12 to 6, but it could also be that Syp itself, acting directly as a protein separate from the number of SNAREpins, somehow powers the turbocharged kinetics. To make this distinction forcefully would require a mutated version of Syp that assembles only 6 SNAREpins rather than 12, so that Syp could be physically present even as the number of SNAREpins is reduced to 6.

With this in mind, we found a previously unknown homozygous missense variation (SYP:c.95 T > C, p.Leu32Pro) in an evolutionarily conserved leucine residue in a TMD of Syp from a female patient diagnosed with ataxia (fig. S10A; see Materials and Methods). AlphaFold Multimer (33) suggested that Leu³² interacts with the VAMP2 TMD (fig. S10B), and this encouraged us to examine whether the function of Syp^L32P might be compromised in our functional biochemical assays despite the unavailability of sequence information from the parents and siblings.

We first tested whether Syp^L32P was affected in its ability to form hexamers or complexes with VAMP2 in reconstituted lipid bilayers (Fig. 5A). Wild-type (WT) Syp efficiently assembles into hexamers, each of which binds 11 ± 3 copies of VAMP2. While Syp^L32P still assembles into hexamers (Fig. 5B, top panel), these now only bind 5 ± 3 copies of VAMP2 (Fig. 5B, bottom panel). This reduction in VAMP2 binding is also evident in Western blot analyses of FLAG bead–immunoprecipitated complexes despite comparable levels of protein expression (fig. S10C).

Fig. 5. Syp^L32P produces 6 SNAREpins.

(A) Schematic representation of the cocluster analyses of the Syp^L32P-VAMP2 complex on supported bilayer. Syp^L32P–GFP (green fluorescent protein; purified from Expi293) and VAMP2 (purified from bacteria)–Alexa Fluor 647–conjugated proteins were reconstituted into the same SUVs and burst onto the glass surface to make a supported bilayer. (B) Histogram representation of the coclusters. The number of molecules in each cluster was determined upon photobleaching and fluorescence intensity. Syp^L32P (top panel; green) produced 5 ± 1 (mean ± SD), while corresponding VAMP2 (bottom panel; red) produced 5 ± 3 molecules per cluster. N > 3 and n > 150 for each class. (C) SNAREpin counting with Syp^L32P-VAMP2^wt-Syt1 vesicles. Vesicles were docked in the presence of 85% Alexa Fluor 647–labeled Cpx1, Munc13L, and DHG on the suspended bilayer containing Munc18-Stx1A and palmitoylated SNAP25. Upon docking, excess Cpx1 was removed by washing, and the number of Cpx1 molecules was counted upon photobleaching. N > 3 and n > 120. (D) Content release kinetics of docked Syp^L32P containing vesicle with ~7-ms time resolution. Two representative examples of the dequenching curves are shown; one population produced a very rapid response with slower release kinetics, while other populations showed a slow response to Ca2+ arrival yet with faster release kinetics. The black dotted line marks the arrival of Ca²⁺. (E) A histogram population of the kinetics of the release. Bin size: 7 ms. N > 3 and n > 120.

Stably docked vesicles produced in reconstitutions of vesicle priming containing the Syp^L32P had a reduced number of SNAREpins (6 ± 1), indistinguishable from 6 (Fig. 5C). As was the case when only six SNAREpins were present as a result of omitting Syp altogether (Figs. 3 and 4), SRB content release with Syp^L32P was gradual (Fig. 5D). Although broadly similar to release in the absence of Syp, there was more variability in the release kinetics among individual Syp^L32P-containing vesicles, many of which fell into two distinct classes having ∆t values of ~35 and ~100 ms, respectively. Other examples of variable behavior include rapid onset, followed by gradual release (Fig. 5E, vesicle 1) and greatly delayed initial fusion pore opening (Fig. 5E, vesicle 2) in some cases even up to ~1 s. A population analysis of lipid mixing was profoundly delayed (fig. S11) into the ~1-s range (t_1/2 ~ 0.9 s; time resolution: 0.15 s), likely reflecting an envelope of variably delayed fusion events with Syp^L32P. Further detailed understanding of these fusion pores (and analogous pores in the next section on VAMP2 TMD mutations) is beyond the scope of this report and will require improved time resolution.

Engineered and disease-causing dimerization mutations in the VAMP2 TMD eliminate the turbocharge and reduce the number of SNAREpins

To take a complementary approach to test the hypothesis that the number of SNAREpins is the primary determinant of ultrafast release, we sought to reduce the number of SNAREpins independently of any alterations in Syp. For this purpose, we mutated the TMD domain of VAMP2, which is known to dimerize in detergent solutions to engineer a nondimerizing version (34, 35). The simplest structural explanation of the 2:1 VAMP2:Syp binding stoichiometry is that each copy of Syp binds one such VAMP2 dimer, perhaps one copy of VAMP2 directly and the other indirectly. If so, a mutation in the VAMP2 TMD that prevents dimerization might retain direct binding while eliminating indirect binding, resulting in a 1:1 stoichiometry with six copies of VAMP2 (mutant) now bound to each Syp hexamer.

As a starting point, we used AlphaFold Multimer to generate a model of the Syp₁-VAMP2₂ complex and noted three residues in the VAMP2 TMD (I97, L107, and L111) forming extensive contacts in the predicted VAMP dimer (fig. S12). Earlier in vitro and simulation studies had revealed two of these residues (L107 and L111) as important for dimerization (34–37). Therefore, we expressed and purified VAMP2^3X, carrying mutations in all three TMD positions (I97A, L107A, and L111A). Comparative native mass spectrometric (nMS) analysis of the WT and mutant indeed established that dimerization in detergent was completely prevented upon triple mutation (Fig. 6, A and B).

Fig. 6. nMS of VAMP2 mutants in detergent solution.

nMS analyses of the bacterially purified recombinant VAMP2 (WT and mutants) were performed from Triton X-100 detergent supplemented with cholesterol. WT VAMP2 (A) produced a significant dimer population (denoted with two purple spheres), while mutants [3X (B), I102N (C), and A104T (D)] produced only monomeric (marked with one purple sphere) population under the same condition. Corresponding monomeric and dimeric mass are indicated.

To measure the effect of these TMD dimerization mutations on the stoichiometry of VAMP2-Syp complexes in native lipid bilayers (27), we coreconstituted purified Syp–GFP (green fluorescent protein) and either WT-VAMP2 or VAMP2-3X (each labeled with Alexa Fluor 647) into SUVs containing 20 mol % cholesterol to stabilize Syp hexamers (22, 38, 39). The SUVs were then applied to glass slides to form continuous supported bilayers (Fig. 7A). We reconstituted SUVs containing very-low surface concentrations (~10 molecules/μm²) of Syp and VAMP2 to avoid the overlap of emissions from separate but randomly nearby molecules in the same optical pixel. By contrast, the surface densities of VAMP2 and Syp in SV membranes are ~10³ times greater (~9000 and ~4000 molecules/μm², respectively) (40). Thus, any complexes detected in our experiments are likely to exist in native SVs.

Fig. 7. VAMP2 mutants produce ~6 SNAREpins with slower release kinetics.

(A) Representative photobleaching traces of the Syp^wt-GFP (green) and VAMP2^3X (red) coclusters. A black dashed line marks the steps. Syp-GFP was purified from the ExpiHEK-293 cell, and VAMP2 mutants were bacterially purified and labeled with Alexa Fluor 647. Both proteins were coreconstituted into SUVs to form a supported bilayer. (B) Histogram analyses of the coclusters (Syp^wt-GFP: VAMP2^mutants) on supported bilayer. Each cluster was subjected to photobleaching to estimate the total number of molecules in each cluster. Histograms were generated keeping bin size at 1. N ≥ 3 and n > 150 for each condition. Mean ± SEM are reported. (C) SNAREpins under each vesicle with various VAMP2 mutants were counted with Alexa Fluor 647–labeled Cpx1 as described above. N = 3 and n > 120. (D) Content release kinetics (SRB dequenching) of the docked vesicle containing VAMP2 mutants. The vertical dashed black line denotes the arrival of Ca²⁺. Time resolution: ~7 ms. (E) Histogram representation of the total fusion time (∆t) for each VAMP2 mutant. Bin size: 7 ms. N > 3 and n > 70. Mean ± SD are plotted.

We located coclusters (i.e., single pixels containing both GFP and Alexa Fluor 647) and then quantified the number of copies of VAMP2 and Syp present in each using step-wise photobleaching (Fig. 7A). In the control experiment (Fig. 7B), Syp was present mainly as hexamers in complexes with 11 ± 4 copies of WT-VAMP2, indistinguishable from 12 as reported earlier (27). As hoped, VAMP2^3X resulted in only one copy of this dimerization mutant per Syp monomer (6 ± 1 copies of VAMP2^3X per Syp hexamer). This twofold difference in stoichiometry between WT and 3X mutant was invariant over at least a 100-fold range in surface concentrations (fig. S13).

We next confirmed that the reduced stoichiometry is reflected in the number of SNAREpins assembling during reconstituted vesicle priming. v-SUVs were reconstituted containing Syp and either VAMP2 or VAMP2^3X and incubated with vesicles incubated suspended bilayers in the complete chaperone system (Fig. 7C). The VAMP2^3X vesicles assembled 6 ± 2 SNAREpins per vesicle based on counting Cpx (Fig. 7C), fully half of the corresponding result (11 ± 2) for VAMP2^WT (Fig. 2D).

Last, we investigated the kinetics of content release from the two types of vesicles (Fig. 3B). Once again, now with six SNAREpins (resulting from mutations preventing VAMP dimerization and WT-Syp), we observed both delayed and progressive content release with likely fusion pore flickering (Fig. 7D) extending over periods (∆t) averaging 160 ± 30 ms (Fig. 7E).

In a second approach, which would not depend on protein engineering, we sought out naturally occurring disease-causing mutations within the VAMP2 TMD that might be located in the same region (107/11) where two of the three engineered mutations in VAMP2^3X cluster. Many de novo mutations in VAMP2 have been found linked in patients having a broad spectrum of neurological and related developmental disorders (41–43), but few have been found in the TMD. One is A104T (a heterozygous variant in a male patient with neurodevelopmental disorder; see Materials and Methods), and another (in mice) is I102N, producing a “restless” phenotype that results from deficiencies in paired-pulse facilitation (44). Each of these mutations markedly reduces VAMP2 dimerization in detergent solution (Fig. 6, C and D).

Correspondingly (Fig. 7D), Syp hexamers reconstituted in supported bilayers bound only ~6 copies of these nondimerizing naturally occurring mutant VAMP2 proteins (7 ± 2 copies of VAMP2^I102N and 6 ± 2 copies in the case of VAMP2^A104T); in addition, only ~6 SNAREpins (Fig. 7C) were assembled under stably docked, primed Syp-containing vesicles (6 ± 3 copies of VAMP2^I102N and 6 ± 2 copies in the case of VAMP2^A104T). Both disease-causing mutations exhibited progressive content release with apparent fusion pore flickering (Fig. 7, D and E), closely resembling the results obtained with all other conditions where docked vesicles contain only six SNAREpins. The differences observed among the various mutations will require further study and are outside the scope of this report.

In summary, Syp normally templates 12 SNAREpins, enabling the extremely rapid exocytosis of neurotransmitters. This turbocharge is lost by reconstituted synaptic vesicles whenever there are only six SNAREpins, a pathological condition that can be caused naturally by certain disease mutations of VAMP2 or purposefully by engineering mutations in the TMD of this v-SNARE to prevent its dimerization, by a naturally occurring disease-associated variant of Syp, or by simply removing Syp altogether (while retaining Munc13 to template the remaining six SNAREpins).

Increase in spontaneous fusion

The SNAREpin numbers and the altered kinetics of release following exposure to Ca²⁺ that we report for the various VAMP2 mutants (and for Syp^L32P) all concern the population of vesicles that are stably docked in the absence of Ca²⁺ for the experimental period (generally up to 10 min). With WT-VAMP2 and WT-Syp, almost all (~90%) of docked, immobile vesicles are stable during or beyond this period, and only a minority (5 to 10%) fuse spontaneously in the absence of Ca²⁺ (fig. S1). In contrast, whenever there are six SNAREpins, the rate of spontaneous fusion is strongly elevated (fig. S14), and there are now two distinct populations: an unstable population (70 to 80%) that fuses in the absence of Ca²⁺ entirely within ~5 s; and a stable population, the majority of which (>90%) is released by Ca²⁺. We were unable to reliably measure SNAREpin numbers for the unstable population due to their instability. This is the reason why we only report numbers for the stably docked population. This result is readily explained by the buttressed ring model: Whenever the six peripheral SNAREpins can form, their bonding to the central SNAREpins (via both the bridging complexin and now, additionally, dimerization of their VAMP TMDs within the SV bilayer) adds stability to the fusion clamp overall, thereby reducing the rate of spontaneous release by the observed ~10 fold.

Quantitative modeling of the role of peripheral SNAREpins in turbocharging synchronous release

To help understand more precisely how the presence of peripheral SNAREpins could affect the kinetics of opening the fusion pore, we revisited a mechanical coupling model of the action of SNAREpins for membrane fusion (45, 46). Each SNAREpin is modeled as a spring that can be in two states, half zippered and fully zippered. These two states were evidenced experimentally by SNAREpin stretching using optical tweezers (47). When a SV is primed, all SNAREpins are placed in the half-zippered state. SNAREpin cooperativity is the result of their indirect connection through the membrane. Upon release of the clamped primed state, i.e., when calcium ions enter the cytosol in the vicinity of the SV, the SNAREpins pull the SV toward the PM and, during the vesicle movement, stochastically switch from one state to the other, the fully zippered state becoming predominant as the vesicle gets closer to the membrane.

The initial opening of the fusion pore occurs when the SNAREpins have brought the vesicle close enough to overcome the short-range hydration and protrusion forces that prevent spontaneous fusion, i.e., ~2 nm (48). In the initial model, a concentric ring of six SNAREpins (Fig. 8A), corresponding to the central SNAREpins, acts alone on the vesicle. The fusion time, defined as the time between clamp release and initial opening of the fusion pore, is displayed as the black line in Fig. 8B. The model predicts that it does not vary with the radius, $r_{\mathrm{c}}$, of the ring and remains close to 500 μs. With 12 central SNAREpins, the fusion time is predicted to increase more than 1000-fold to ~700 ms, again independently of $r_{\mathrm{c}}$ (red line in Fig. 8B). This was expected because the optimum number is 3 to 6 (45).

Fig. 8. A mechanistic model for peripheral SNAREpins to turbocharge the release.

(A) Schematic representation of the various ways of SNAREpin assembly in ringlike manner. Left: Only six SNAREpins (central) are assembled as a ringlike manner with a central ring radius of r_c. In the middle panel, all 12 SNAREpins (6 central + 6 peripheral) are positioned on a single ring with a radius of r_c. In the right panel, six central and six peripheral SNAREpins are positioned as a pair and positioned on a double concentric ring, where central SNAREpins are assembled on a ring with r_c and peripheral SNAREpins are modeled as pair to the central SNAREpin. As peripheral one emanates radially outward, therefore, the difference in the ring radius is considered to be δ_r. (B) A numerical calculation considering SNAREpins are clamped at half-zippered state and fusion pore opening time is calculated with varying δ_r. Notably, for a single ring consisting of either 6 (black) or 12 (red) SNAREpins, fusion time (defined as time required to release the clamp and open the fusion pore) was ~400 μs and 700 ms, respectively. However, when SNAREpins are considered as pair, the fusion time decreased with increasing δ_r. For δ_r > 4 nm (probable distance in physiology is close to 5 nm, one Syt1 C2B diameter), fusion time can reach to 20 μs. (C) A molecular model depicting six SNAREpins arranged on the Syt1 ring. This ring, consisting of 6 central SNAREpins, is hypothesized to create pore flickering and partial release, while 12 SNAREpins, as a pair of 6 central + 6 peripheral (D), are predicted to turbocharge the release. All these molecular models are generated using crystal structures 5W5C (79) and 3RL0 (7). A SV with a 40-nm diameter was shown. TMDs and other auxiliary proteins are removed for clarity.

To incorporate an outer ring of peripheral SNAREpins into a revised version of the mechanical coupling model, we simply added six peripheral SNAREpins in a ring of radius, $r_{\mathrm{c}}+{\mathrm{\delta }}_{\mathrm{r}}$. Because of this larger radius, the peripheral SNAREpins are more stretched than the central SNAREpins. As a result, the energy minima of their half-zippered and fully zippered states are shifted toward shorter PM–SV distances. We computed the fusion time for $r_{\mathrm{c}}$ and ${\mathrm{\delta }}_{\mathrm{r}}$ in the range of 6 to 10 and 2 to 7 nm, respectively. The results are displayed in Fig. 8B. The model predicts that when ${\mathrm{\delta }}_{\mathrm{r}}$ ≥ 4 nm, fusion is accelerated compared to a central SNAREpin ring without peripheral SNAREpins, independent of $r_{\mathrm{c}}$. This identifies the separation ${\mathrm{\delta }}_{\mathrm{r}}$ between the inner and outer SNAREpin rings as a critical variable to determine the extent of turbocharge.

The TMDs of the Syntaxin subunits of the central and peripheral SNAREpins are predicted to be inside and outside of the Synaptotagmin ring, respectively (13). The minimum value of ${\mathrm{\delta }}_{\mathrm{r}}$ is therefore the thickness of a Syt-C₂B domain in the ring (~4 nm) plus a half-diameter for each of the two Syntaxin TMDs (~1 nm), i.e., ${\mathrm{\delta }}_{\mathrm{r}}$ ≥ 5 nm. For an 18-member ring of Syt-C₂B, $r_{\mathrm{c}}$ is 10 nm (8). When $r_{\mathrm{c}}$ = 10 and ${\mathrm{\delta }}_{\mathrm{r}}$ = 5 nm, the fusion time is reduced to ~5 μs, which is ~100-fold faster than without peripheral SNAREpins. The initial opening of a fusion pore is of course only the starting point of the fusion process. Although we did not model it, the positive cooperativity between peripheral and central SNAREpins will likely continue during the expansion of the fusion pore (13).

DISCUSSION

The complete, spontaneous zippering of a single pure SNAREpin between a vesicle and a bilayer provides enough energy to initially open a reversible fusion pore (49, 50), and three such SNAREpins are needed to irreversibly open the pore (51). From an energetic perspective, opening a fusion pore requires overcoming ~30 k_BT of unfavorable activation energy (52). Where, exactly, does this energy come from when a primed SV is released? The maximum potential energy available from a single isolated synaptic SNAREpin corresponds to the energy released when its three constituent SNAREs self-assemble in solution into fully assembled four-helical bundle SNARE complexes, ~100 k_BT (47, 53).

Most of this assembly has already taken place when an SV reaches the ready-release state. Various indirect measurements imply that the four-helical bundle is at least ~70% zippered and possibly as much as ~95% zippered at this stage, so as little as ~5 k_BT per SNAREpin may be incrementally available during release to open the fusion pore initially (47, 54). Where did the balance of the ~100 k_BT per SNAREpin in energy go? We suggest that it is stored as potential energy within the three-dimensional structure assembled at the SV-PM interface, including internal pressure and related physical stresses on the involved lipid bilayers. This could total as much as ~1200 k_BT for 12 SNAREpins. We estimated that these stresses will raise the internal pressure in an SV to approximately 3 atm, enough to initially propel neurotransmitters out of an opening fusion pore in a concentrated jet (1). The exceptionally small size of SV likely contributes to maximizing internal pressure.

We suggest that most of this extraordinary quantity of stored energy (roughly equivalent to ~100 adenosine 5′-triphosphates per SV) is released in what amounts to an explosive process for which Ca²⁺ ions are the trigger. An explosion releases contained energy when it can no longer be contained by an enclosing structure, and the forces released typically propel materials at extreme speeds. It is easy to imagine how two bilayers may be forcibly driven together in this manner, taking advantage of the entirety of the energy once stored in the exploding structure and now released in the explosion.

It is also easy to imagine that, under various artificial or pathological conditions, we identified where there are only six SNAREpins, there is no longer enough stored potential energy to achieve an explosive threshold. Still, even with just six SNAREpins (six being greater than three), a fusion pore will open, although more slowly and perhaps reversibly. This is essentially what we have observed (Figs. 3 to 7).

The buttressed ring structural model (1, 10) of the arrangement of proteins under the SV in the ready-release state offers a more granular explanation of this energetic overview. Six pairs of SNAREpins are symmetrically bound to an inner ring composed of ~18 Synaptotagmin C₂B domains, closely surrounded by an outer ring of six copies of Munc13. One SNAREpin of each pair (the Central SNAREpin) is bound to a Syt-C₂B domain. This association prevents further zippering by preventing the radially inward movement of the SNAREpin toward the SV-PM contact site that is required for continued zippering. Each peripheral SNAREpin is paired to its central SNAREpin by a bridging molecule of Complexin (and now additionally by their dimerized VAMP TMDs) in such a way that the Cpx accessory helix prevents peripheral SNAREpin from further zippering (7); however, peripheral SNAREpins are not themselves bound to Syt-C₂B.

Overall, this geometry creates two concentric and mechanically coupled rings of SNAREpins, an inner ring of six central SNAREpins and an outer ring of six paired peripheral SNAREpins (Fig. 8). Both rings of SNAREpins are expected to be released simultaneously when the integrity of the inner ring of Syt is reduced, as one or more of its C₂B subunits exit the ring upon binding sufficient Ca²⁺. Straightforward physical principles then imply distinct and synergistic roles to each ring that together achieve synchronous neurotransmitter release. Specifically, the inner ring of SNAREpins opens the fusion pore, while the outer ring of SNAREpins greatly accelerates (“turbocharges”) this process and transiently maintains the pressure inside the SV during neurotransmitter release by continuous zippering (1). It is the force exerted by the peripheral SNAREpins of the outer ring that propels the central SNAREpins of the inner ring to open the pore, and critically, this force continues after the pore opens to drive its irreversible expansion. The synergy between the two rings of SNAREpins and the need for their radial separation to achieve a turbocharge are quantitatively captured in the mechanical coupling model (Fig. 8), in which zippering SNAREs are modeled with energetically matched springs (45). When there are only six SNAREpins (and thus only one ring), this effect vanishes, and the pore opens more slowly and reversibly (Fig. 3: 6 SNAREpins) rather than at once and irreversibly (Fig. 3: 12 SNAREs). However, within this model, we cannot distinguish a fully closed fusion pore from a hemifusion state or from a topologically open but extremely narrow pore that is too small to permit dye passage. Resolving these possibilities will require future studies.

The buttressed ring model (1, 10) was formulated by piecing together a combination of structures of individual fusion machinery and their complexes, combined with insights from biochemistry and genetics. As we have seen, the model has proven to have excellent predictive value so far. However, the predicted assemblies of concentric rings of Munc13 and Synaptotagmin at the interface of ready-release vesicles with paired peripheral and central SNAREpins have yet to be directly observed, and testing the model in this manner remains a major goal. We observed that SNAREpins assemble in two successive waves of six each during priming for ready release (18), readily explained by the assembly of central followed by peripheral SNAREpins.

We envisioned that central and peripheral SNAREpins are linked only by the bridging molecule of Complexin, which was originally observed in a crystal structure (7). In particular, the functional role of the accessory helix in stabilizing this trans-SNAREpin complex was first confirmed by solution binding and later by numerous genetic and biochemical studies that amply established its role in minimizing spontaneous release (55–57). The current study has revealed a second mechanism for stabilizing these same SNAREpin pairs: dimerization via the TMDs of their constituent VAMP2 (figs. S11 and S14).

We wish to emphasize the fundamental concept embodied in this pairing of SNAREpins, separate from the other details of the buttressed ring hypothesis. Pairing assures that each SNAREpin that opens a fusion pore will receive a turbocharge from its partner. Pairing is at least bivalent, synergistically stabilized by two cross-links: one by Complexin and the other between the two VAMP2-TMDs in the SV. Disrupting either cross-link (Cpx accessory helix or VAMP2 dimerization mutations) removes stabilizing binding energy from the prefusion/clamped/ready-release state, accelerating spontaneous release and reducing the steady-state size of the ready-release pool of SVs. It is likely that Syp, by binding these VAMP2 dimers (Fig. 6), further stabilizes overall SNAREpin pairing. The fact that the Syp^L32P mutant protein, which binds VAMP2 monomers rather than dimers in bilayers (Fig. 5), also enables spontaneous release (fig. S11) further underscores the basic importance of SNAREpin pairing in controlling the mode and speed of neurotransmitter release.

The observation of apparent flickering (reversible opening and closing) of fusion pores in our study under essentially pathological conditions (i.e., with mutant proteins including disease-causing mutations) raises the possibility of whether such behavior can also be a part of the physiological repertoire. Certainly, this behavior is well established in closely related synaptic systems, such as dense core vesicle (DCV) exocytosis under physiological conditions (58). These involve essentially the same or very similar proteins as SVs but use much larger vesicles. The larger DCVs enable direct electrophysiological measurements of fusion pore dynamics that are not yet possible for SVs in native synapses (59). However, a considerable body of less direct but still powerful evidence from optical approaches suggests that similar flickering can occur in neural networks in culture, where it can be pharmacologically and physiologically regulated (19, 20).

Syp has been largely ignored since the late 1990s, when mouse genetic models failed to show a strong phenotype (60). Further investigation of this negative result involving knocking out multiple Syp-related proteins, as well as Syp itself, has recently revealed strong synaptic phenotypes (61). Biochemical and structural studies are also suggesting specific functional roles. Several cryo-EM structures also recently revealed that Syp binds to the SV Vacuolar-type adenosine triphosphatase complex, suggesting an intriguing relationship with neurotransmitter loading (62, 63). Another recent study revealed that upon neurotransmitter loading, SVs expand their volume. This depends on the presence of Syp, which increases the elasticity of the SV membrane (64). We further hypothesize that Syp may increase membrane tension, resulting in stretching of the membrane at the vesicle-PM interface and potentially accelerating pore opening. Last, Syp templates 12 SNAREpins (27), and, as we show here, this results in turbocharged and complete release.

More speculatively, all of this raises the intriguing possibility that there may be physiological regulation that somehow “toggles” a synapse between alternate modes of fusion: full fusion with a turbocharge involving 12 SNAREpins and explosive neurotransmitter release through an irreversible fusion pore or alternately extended slow quantal release involving only 6 SNAREpins resulting from short-lived opening and closing fusion pores. Syp—which templates 12 SNAREpins during vesicle priming and which, by means of a single point mutation (fig. S15), can switch to templating 6 SNAREpins—would seem to be the natural platform for this and other possible regulation of the fusion pore.

Previous studies indicate that membrane fusion requires multiple SNARE complexes. In PC12 cells, 5–8 Syntaxin TMDs are estimated to align to open the fusion pore, while neurotoxin cleavage experiments suggest 10–15 SNARE complexes participate in SV exocytosis (65, 66). Although 50–70 copies of Syntaxin and SNAP25 are recruited around each granule in PC12 cells, only a small subset is dynamically exchanged during exocytosis (67, 68). Approximately six SNAP25 molecules undergo conformational changes prior to fusion in chromaffin cells, comparable to the ~6 Munc13 molecules docked beneath insulin granules (69, 70). In our in vitro reconstitution system, ~12 copies of Cpx1 and SNAP25 cluster upon vesicle docking. Together, despite differences in systems and methodologies, these findings support the stoichiometric assembly of multiple SNARE complexes, with SNARE organization tracking Munc13 during docking and priming.

Last, most vesicle fusion events in eukaryotic cells occur spontaneously, supporting essential functions such as continuous metabolism, cell growth, and membrane biogenesis. Examples include endoplasmic reticulum (ER)–to–Golgi transport, receptor-mediated endocytosis, autophagy, and exosome production—all of which take place spontaneously on the scale of seconds. We suspect that these processes likely rely on six SNAREpins assembled by analogous hexamers of Munc13 homologs, which are known to chaperone SNARE complex formation (71, 72). These systems are not expected to include peripheral SNAREpins, as Complexin is neither involved in nor expressed in most of these contexts.

MATERIALS AND METHODS

Materials

The following cDNA constructs were used in this study: full-length VAMP2 (mouse His⁶-SUMO-VAMP2, residues 1 to 116), full-length VAMP2^3X (VAMP2-His⁶, residues 1 to 116 with I97A, L107A, and L111A mutations), full-length t-SNARE complex (mouse His⁶-SNAP25B, residues 1 to 206 and rat Syntaxin1A, residues 1 to 288), full-length SNAP25 (mouse His⁶-SNAP25b, residues 1 to 206), Synaptotagmin (rat Synaptotagmin1-His⁵, residues 57 to 421); Munc13 C1-C2B-MUN-C2C domain (referred as Munc13 or Munc13L interchangeably) (rat His¹²-Munc13-1, residues 529 to 1735 with residues 1408 to 1452 replaced by the sequence EF and 1532 to 1550 deleted), Munc18/Syntaxin complex (rat His⁶-SUMO Munc18/Syntaxin1), and Complexin 1 (human His⁶-Complexin 1, residues 1 to 134). All VAMP2 mutations (VAMP2 I102N and A104T) were generated using a site-directed mutagenesis kit (New England Biolabs). We also generated a construct expressing Syp-FLAG and Strep tag-VAMP2 (separated by a T2A sequence) for copurification.

Lipids, 1,2-dioleoyl-sn-glycero-3-(phospho-l-serine) (DOPS), 1,2-dioleoyl-glycerol-3-phosphocholine (DOPC), 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl), PIP₂, and cholesterol were purchased from Avanti Polar Lipids (Alabaster, AL). Atto 647N DOPE and Atto 465 DOPE were purchased from ATTO-TEC GmbH (Siegen, Germany). Alexa Fluor 647 C2 Maleimide was purchased from Thermo Fisher Scientific (Waltham, MA).

Protein expression and purification

VAMP2 (WT and mutants) and preassembled t-SNAREs, Munc18-Stx1A complex, Munc13L, SNAP25, Cpx, Syp, and Syt1 recombinant proteins were expressed and purified as described previously (16, 73, 74). Native Syp-VAMP complex was purified as described before (18, 22). Purified SNAP25 was palmitoylated as described earlier (16). Syp-GFP was cloned into the pEG BacMam vector (Addgene, #160451), and the L32P mutation was generated using site-directed mutagenesis. Both WT and L32P Syp-GFP were expressed and purified from ExpiHEK-293 cells. The plasmid was used to transfect Expi293F cells (Thermo Fisher Scientific, USA) following the manufacturer’s protocol. Briefly, 1 μg of DNA per 1 ml of cell culture was mixed in a 1:3 ratio of DNA: Polyethylenimine transfection reagent for 30 min at room temperature (RT) and then added to approximately 2.5 to 3 × 10⁶ cells with >95% viability. Cells were grown at 37°C under 8% CO₂ with orbital shaking for ~48 hours posttransfection. The cells were centrifuged at 500g for 5 min, washed twice with phosphate-buffered saline (PBS), and frozen. Harvested pellets were stored at −80°C until needed. The cell pellets (from 50 ml of culture) were thawed and resuspended in 8 ml of 1× PBS with protease inhibitors (Roche, Indianapolis, USA). Cells were lysed using a Dounce homogenizer on ice and then centrifuged in a Beckman SW55 rotor at 100,000g for 1 hour. The pellet was then resuspended in 50 mM Hepes (pH 7.4), 175 mM NaCl, 0.2% cholesterol (Avanti, ovine catalog no. 700000P), 10 μl of deoxyribonuclease (10 mg/ml), 1 mM phenylmethylsulfonyl fluoride, 0.5 mM tris(2-carboxyethyl)phosphine (TCEP) buffer, and protease inhibitors. Triton X-100 was added to a final volume of 0.2% (v/v) and incubated overnight on a rotating wheel at 4°C. The lysate was centrifuged at 48,000g for 30 min using a Beckman JA-20 rotor to pellet the insoluble fraction, and the solubilized protein was incubated with 0.5 ml of FLAG resin (GeneScript, FLAG Resin, catalog no. L00432) at 4°C for 4 to 6 hours on a rotating wheel. The resin is then washed multiple times with 50 mM Hepes (pH 7.4), 175 mM NaCl, 0.2% cholesterol, 0.5 mM TCEP buffer, and protease inhibitors. Protein was then eluted using the same buffer with 0.2 mg/ml (final) of 3× FLAG peptide (APExBIO, USA). The purity of the protein was regularly checked with SDS–polyacrylamide gel electrophoresis after Coomassie staining. Western blot of Syp^WT/L32P and VAMP2^WT complexes was performed with equal total protein loading, using anti-Syp (Synaptic Systems, catalog no. 101011) or anti-VAMP2 (CL69.1; Synaptic Systems, catalog no. 104211) antibodies.

SNAP25 was expressed from a cDNA encoding mouse His₆-SNAP25b (residues 1 to 206) cloned into a pET28b vector containing an N-terminal thrombin cleavage site. Endogenous cysteine residues (Cys⁸⁵, Cys⁸⁸, Cys⁹⁰, and Cys⁹²) were mutated to glycine, and Gln²⁰ was substituted with cysteine to generate a single labeling site for maleimide-based fluorophore conjugation.

Protein labeling

Bacterially expressed and purified complexin (Cpx1^1–134), SNAP25b (Q20C, C85G, C88G, C90G, and C92G), VAMP2 (S28C and C103A), and corresponding mutants (3X, I102N, and A104T) were incubated at 4°C overnight with Alexa Fluor 647 C2 Maleimide dye (Thermo Fisher Scientific, catalog no. A20347) with a 1:5 molar ratio in the presence of 5% dimethyl sulfoxide. Excess dye was removed twice, each time using an equal volume of the dye removal resin (Thermo Fisher Scientific, catalog no. 22858); then, furthermore, protein was loaded into Superdex 75; and peak fractions (monitored both with 280- and 651-nm wavelength) were concentrated using Amicon Ultra-4 (3-kDa cutoff; Merck). Protein concentrations were checked with the Bradford assay, dye concentration was measured from the absorbance (Alexa Fluor 647 at 651 nm), and labeling efficiency was calculated with at least three repeats.

Suspended bilayer and vesicle preparation

Suspended bilayers were prepared as described previously (15, 17), with a composition of 81% DOPC, 15% DOPS, 3% PIP₂, and 1% Atto 465 PE for single-vesicle confocal experiments and 81.9% DOPC, 15% DOPS, 3% PIP₂, and 0.1% Atto 465 PE for single-vesicle TIRF experiments. Briefly, the lipids were dried under N₂ gas, followed by vacuum desiccation for 2 hours. Next, lipids were rehydrated with Munc18/Syntaxin1A and palmitoylated SNAP25 (1:3200 protein:lipid input ratio) in 5× buffer [125 mM Hepes, 600 mM KCl, and 1 mM TCEP (pH 7.4)] supplemented with 2% Triton X-100 for 30 min. Subsequently, the samples were mixed with SM-2 Bio-Beads (Bio-Rad) for 30 min on a rotary mixer and dialyzed overnight in 5× buffer.

Suspended lipid bilayers for single-vesicle confocal experiments were formed by drying and rehydration to form GUVs as described previously (15, 17). Briefly, 4-μl drops of the proteoliposome solution collected after dialysis were dried on Mattek dishes and rehydrated twice. In the second rehydration, the sample was diluted fivefold to 20 μl with distilled water and added to a cleaned silicon chip containing 1× buffer [25 mM Hepes, 120 mM KCl, and 1 mM TCEP (pH 7.4)] supplemented with 5 mM MgCl₂. The bilayer was extensively washed with 1× buffer supplemented with 1 mM MgCl₂.

SNAREpin counting

All the complexin counting experiments underneath docked vesicles were performed with TIRF (Nikon and Abbelight) on a reconstituted planar suspended bilayer on the silicon surface. The design of the silicon chip and the method to create a suspended bilayer compatible with single-molecule imaging were described in detail elsewhere (15, 17). Briefly, Munc18-Stx1A– and palmitoylated SNAP25–containing GUVs (~82% DOPC, 15% DOPS, 3% PIP₂, and 0.1% Atto 465 DOPE) were prepared using the osmotic shock protocol and burst onto freshly plasma-cleaned Si/SiO₂ chips containing 5-μm diameter holes in the presence of Hepes buffer [25 mM Hepes (pH 7.4), 140 mM KCl, and 1 mM TCEP] supplemented with index matching 45% OptiPrep (STEMCELL Technologies) and 5 mM MgCl₂. After waiting 20 min, the suspended bilayer was extensively washed with Hepes buffer containing 1 mM MgCl₂. Before adding any vesicles, the bilayer was completely bleached, and then 0.5 μM total complexin (with various labeling efficiencies), 0.5 μM DHG (short-chain, water-soluble diacylglycerol 1,2-hexanoyl-sn-glycerol), and 0.2 μM Munc13L were added and mixed well. Then 100 nM SUVs (63% DOPC, 15% DOPS, 20% cholesterol, and 2% Atto 465 DOPE) containing different proteins were added and incubated for 5 to 10 min at RT. Lipid-to-protein ratios were kept constant across all SNAREpin counting and content-release kinetics experiments, with Syp^WT/L32P at 1:200, VAMP2^{WT or mutants} at 1:100, and Syt1 at 1:250 (protein:lipid input ratio). When enough vesicles docked (typically within 10 min), to lower the background noise, excess complexin was removed by washing three to four times with Hepes buffer. Both vesicle and complexin were imaged simultaneously with 488- and 633-nm lasers. Complexin signals were subjected to photobleaching using a 633-nm laser with a 100-ms acquisition time. All the data were analyzed with ImageJ. The number of complexin signals underneath a vesicle was calculated from the steps of the photobleaching curves. To determine the total complexin number (especially when labeling percentages were > 25%), photobleaching curves were fitted with an exponential decay curve with Eq. 1

$$I \left(t\right)=I_0{e}^{\frac{-t}{\mathrm{\tau }}}+B$$ (1)

where I(t) denotes the intensity at time t, I₀ is the initial intensity before bleaching, τ is the time constant, and B is the background. Hence, the copy number, N, can be obtained from Eq. 2

$$N=\frac{I_0}{i}$$ (2)

where i is the average unit intensity of a single fluorophore as determined from the small clusters from the same experiment. All the complexin counting data were converted to histograms with a 1 bin size. The histograms were fitted with Poisson’s distribution using OriginPro. Mean ± SEM of the various labeling percentages (25, 50, and 85%) were linearly fitted. The slope of the fit provides the estimates of the total complexin counts under each docked vesicle.

Model for predicting the SNAREpin distribution underneath a vesicle

We predicted that there are exactly $N_{\mathrm{S}}$ SNAREpins that form in the contact between the vesicle and the membrane, and a single Complexin molecule binds to each SNAREpin. Only a fraction $\left(f\right)$ of the Complexin molecules are labeled. Because Complexin randomly binds to SNAREpin, the number of fluorescent dyes, $n,$ follows a binomial distribution

$$p\left(n\right)=\left(\begin{array}{c}{N}_{\mathrm{S}}\\ n\end{array}\right)f^n{(1-f)}^{N_{\mathrm{S}}-n}$$ (3)

Using $N_{\mathrm{S}}=12$ and the three labeling fractions used, approximately 0.25, 0.5, and 0.85, the dye copy number distributions can be directly predicted and compared with the experimental ones. The results are presented in Fig. 2 for $f=0.25,0.5,$ and $0.85$. The predictions are consistent with the observed distributions.

Content release kinetics

SUVs were prepared as described previously (11, 12), with a composition of 48% DOPC, 15% DOPS, 35% cholesterol, and 2% Atto 647N PE. Briefly, the lipids were dried under N2 gas, followed by vacuum desiccation. Next, the bilayer samples were rehydrated with VAMP2^WT/Mutants (1:100 protein:lipid input ratio), Synaptotagmin1 (1:250), and Syp^WT/L32P (1:200) in 1× buffer (supplemented with 40 mM sulforhodamine-B) for 30 min. Subsequently, the samples were diluted to a final volume of 150 μl, mixed with SM-2 Bio-Beads (Bio-Rad) for 30 min on a rotary mixer, and then dialyzed overnight in 1× buffer. The samples were subjected to gel filtration using a CL-4B column to remove the unincorporated dye. Dequenching of the SRB dye inside SUVs was routinely checked by adding 1%n-octyl-β-D-glucoside and recording fluorescence using a plate reader. Typically, 7 to 10 mM SRB was loaded into ~100-nm vesicles (radius) (fig. S5).

SRB-loaded SUVs (with various proteins) were docked on the suspended bilayer containing either Munc18-Stx1A and palmitoylated SNAP25 in the presence of Cpx1, Munc13L, and DHG as described above or preassembled t-SNAREs, as described earlier (13, 16, 18). After 10 min of incubation, excess vesicles were removed with Hepes buffer prewarmed at 37°C. Before adding Ca²⁺, the frame was adjusted to the maximum area compatible with a 147-Hz frame rate (~7-ms time resolution). Fusion was triggered with 100 μM Ca²⁺ (final) while monitoring the vesicle with a 532-nm laser. For monitoring fusion in the presence of BAPTA, 100 μM BAPTA (final) was added and mixed well, and the volume of the reaction mixture was adjusted to a minimum of 150 μl. A total of 100 μM Ca²⁺ was injected with a pipette (10 μl of 1.5 mM stock solution) directly on the laser spot. Numerical calculation for BAPTA chelating out free Ca²⁺ was done using the following Eq. 4

$$C_{\text{Ca}}\left(t\right)=\frac{C_0}{1+C_0{k}_{\text{on}}t}$$ (4)

Free Ca²⁺ concentration, C_Ca, is considered a function of time (t). C₀, initial Ca²⁺ concentration, = 100 μM; and k, on rate of the Ca-BAPTA binding, = 0.550 μm⁻¹ ms⁻¹. We ignored the off rate, as it is very slow and for simplicity.

To calculate the Ca²⁺ concentration (Fig. 4B, inset) due to free diffusion, we considered Ca²⁺ diffusion within a cylindrical area of various radii R and height h. The concentration of free Ca²⁺ at a distant r, from the center of this cylinder and time t after addition, was calculated using the following Eq. 5

$$c_f(r,t)=\frac{c_0}{2\mathit{Dt}}{\int }_0^{R}x{e}^{-\frac{r^2+x^2}{4\mathit{Dt}}}{I}_0\left(\frac{\mathit{rx}}{2\mathit{Dt}}\right)\mathit{dx}-C_0$$ (5)

where $C_0=100 \text{μM}$, which can be immediately quenched by BAPTA, there should be a duration at a distance r, t(r), during which c(r, t) remains above the threshold concentration c_t and I₀ is the modified Bessel function of the first kind.

For monitoring Ca²⁺ arrival, a high-affinity Ca²⁺ sensor, Calcium Green (k_D of ~75 nM) conjugated to a lipophilic 24-carbon alkyl chain (Calcium Green C24), was introduced into the bilayer. Both 532-nm (Ca-Green) and 647-nm (lipid mixing) channels were monitored simultaneously at a 147-Hz frame rate using a dual viewer. Similarly, lipid mixing and SRB dequenching were monitored with 647-nm and 532-nm lasers simultaneously at a 147-Hz frame rate using a dual-viewer. All the movies were analyzed using ImageJ and a time series analyzer plugin. We refrained from background subtraction for SRB dequenching events, as SRB dyes spread out almost instantaneously along the well below increasing the background intensity simultaneously.

Diffusion of fluorescent lipids from the vesicle through the fusion pore

The rate of diffusion of the fluorescent lipids from the vesicle to the pore-spanning membrane can be estimated using Fick’s law. The fluorescent lipid diffusion flux is

$$J_{\text{lip}}=-D_{\text{lip}}\nabla c_{\text{lip}}\left(t\right)=-D_{\text{lip}}\frac{c_{\text{lip}}\left(t\right)}{l_{\mathrm{p}}}$$ (6)

where $c_{\text{lip}}\left(t\right)$, $D_{\text{lip}}$, and $l_{\mathrm{p}}$ are respectively the fluorescent lipid concentration in the vesicle at time $t$, the lipid diffusion coefficient, and the pore length. The number of fluorescent lipids flowing through the pore is

$$\frac{\mathrm{\partial }{n}_{\text{lip}}\left(t\right)}{\mathrm{\partial }t}=2\mathrm{\pi }{r}_{\mathrm{p}}{J}_{\text{lip}}=-\frac{2\mathrm{\pi }{r}_{\mathrm{p}}{D}_{\text{lip}}}{l_{\mathrm{p}}}{c}_{\text{lip}}\left(t\right)$$ (7)

where $r_{\mathrm{p}}$ is the pore radius. Hence, the variation in time of the fluorescent lipid concentration in the vesicle follows

$$\frac{\mathrm{\partial }{c}_{\text{lip}}\left(t\right)}{\mathrm{\partial }t}=\frac{1}{4\mathrm{\pi }{R}^2}\frac{\mathrm{\partial }{n}_{\text{lip}}\left(t\right)}{\mathrm{\partial }t}=-\frac{r_{\mathrm{p}}{D}_{\text{lip}}}{2R^2{l}_{\mathrm{p}}}{c}_{\text{lip}}\left(t\right)$$ (8)

The lipid concentration decreases exponentially with a characteristic time

$${\mathrm{\tau }}_{\text{lip}}=\frac{2R^2{l}_{\mathrm{p}}}{r_{\mathrm{p}}{D}_{\text{lip}}}$$ (9)

Release of SRB from the vesicle

For the rate of release of SRB from the vesicle to the pore-spanning membrane, the same calculation is valid. It can be estimated using Fick’s law. The SRB diffusion flux is

$$J_{\text{SRB}}=-D_{\text{SRB}}\nabla c_{\text{SRB}}\left(t\right)=-D_{\text{SRB}}\frac{c_{\text{SRB}}\left(t\right)}{l_{\mathrm{p}}}$$ (10)

where $c_{\text{SRB}}\left(t\right)$ and $D_{\text{SRB}}$ are respectively the SRB concentration in the vesicle at time $t$ and SRB diffusion coefficient. The number of SRB released through the pore is

$$\frac{\mathrm{\partial }{n}_{\text{SRB}}\left(t\right)}{\mathrm{\partial }t}=\mathrm{\pi }{(r_{\mathrm{p}}-r_{\text{SRB}})}^2{J}_{\text{SRB}}=-\frac{\mathrm{\pi }{(r_{\mathrm{p}}-r_{\text{SRB}})}^2{D}_{\text{SRB}}}{l_{\mathrm{p}}}{c}_{\text{SRB}}\left(t\right)$$ (11)

where $r_{\text{SRB}}$ is the Stokes radius of SRB. Hence, the variation in time of the SRB concentration in the vesicle follows

$$\frac{\mathrm{\partial }{c}_{\text{SRB}}\left(t\right)}{\mathrm{\partial }t}=\frac{3}{4\mathrm{\pi }{R}^3}\frac{\mathrm{\partial }n\left(t\right)}{\mathrm{\partial }t}=-\frac{3{(r_{\mathrm{p}}-r_{\text{SRB}})}^2{D}_{\text{SRB}}}{4R^3{l}_{\mathrm{p}}}{c}_{\text{SRB}}\left(t\right)$$ (12)

In addition, the SRB decreases exponentially with a characteristic time

$${\mathrm{\tau }}_{\text{SRB}}=\frac{4R^3{l}_{\mathrm{p}}}{3{(r_{\mathrm{p}}-r_{\text{SRB}})}^2{D}_{\text{SRB}}}$$ (13)

The ratio between these two characteristic times is

$$\frac{{\mathrm{\tau }}_{\text{lip}}}{{\mathrm{\tau }}_{\text{SRB}}}=\frac{3{(r_{\mathrm{p}}-r_{\text{SRB}})}^2{D}_{\text{SRB}}}{2R{r}_{\mathrm{p}}{D}_{\text{lip}}}$$ (14)

Using typical values ($D_{\text{SRB}}\sim 350 {\text{μm}}^2{\mathrm{s}}^{-1}$ in water, $r_{\text{SRB}}\sim 0.2 \text{nm}$, $l_{\mathrm{p}}\sim 10 \text{nm}$, $R\sim 25 \text{nm}$, and $r_{\mathrm{p}}\sim 2 \text{nm}$), then, ${\mathrm{\tau }}_{\text{SRB}}\sim 0.1 \text{ms}$, considering $D_{\text{lipid}}\sim 5 {\text{μm}}^2{\mathrm{s}}^{-1}$ and ${\mathrm{\tau }}_{\text{lipid}}\sim 1 \text{ms}$. The fusion characteristic time was calculated using the SRB dequenching curve in solution and the raw dequenching data from the experiments.

Cocluster analyses of Synaptophysin and VAMP2 on supported lipid bilayers

The evaluation of cocluster formation of Syp and VAMP2 on a supported lipid bilayer was previously elaborated in Bera et al. (18). Briefly, purified GFP-Syp^WT/L32P and VAMP2^WT/Mutants (Alexa Fluor 647, maleimide labeling; for these experiments, VAMP2^S28C,C103A variant was used as WT to get only one fluorophore per protein) were mixed and incubated overnight at 4°C. SUVs (60 mol % DOPC, 15% DOPS, 5% DOPE–polyethylene glycol, molecular weight 2000, and 20% cholesterol) were initially dried with nitrogen gas for 5 min and further kept under a vacuum for another 2 hours. This dried lipid was dissolved in vesicle buffer [25 mM Hepes (pH 7.4), 125 mM KCl, 0.2 mM TCEP, and 0.2% Triton X-100]. From the overnight incubated mixture of GFP-Syp and VAMP2 (Alexa Fluor 647), the proteins were reconstituted into this lipid mixture with a protein:lipid ratio of 1:50,000, 1:5000, and 1:500 for a final volume of 150 μl. A total of 150-mg prewashed SM-2 Bio-Beads (Bio-Rad, Hercules, CA) were incubated with this mixture for another 30 min at room temperature with gentle shaking. The liposomes were dialyzed overnight with a 6- to 8-kDa cutoff against a detergent-free vesicle buffer. The following day, 150 μl of 80% Nycodenz (in the detergent-free vesicle buffer) was mixed with the sample and then loaded into a 0.8-ml tube (Beckman Coulter, catalog no. 344090), overlaid with 250 μl of 30% Nycodenz, followed by 50 μl of detergent-free vesicle buffer. Tubes were centrifuged in an SW55 rotor at 48,000 rpm for 4 hours in the Beckman ultracentrifuge. Proteoliposomes were collected from the gradient near the buffer/30% Nycodenz interface. The supported bilayer was created by Mg²⁺ (10 mM)–induced bursting of the collected liposomes in ibidi glass-bottom chambers (ibidi GmbH, Germany). The bilayer was extensively washed with the detergent-free vesicle buffer with Mg²⁺ (1 mM). The bilayer was imaged with a TIRF microscope (Nikon and Abbelight) using 488- and 633-nm lasers, and photobleaching of the protein cluster was analyzed with ImageJ. Equations 1 and 2 were used to estimate the final number of molecules per cluster if steps are not discernible, typically >5 to 7. Histograms were generated with a bin size of 1.

Single-vesicle fusion assays

Single-vesicle fusion assays on the confocal microscope were performed as described previously (16). First, 500 nM DHG was added to the Munc18/Syntaxin and SNAP25 bilayer and incubated for 5 min. Vesicles (~100 nM lipids) and soluble proteins (200 nM Munc13 and 2 μM Complexin 1) were premixed and added from the top using a pipette and allowed to interact with the bilayer for 3 min. For preassembled t-SNAREs, experiments were performed as described previously. Subsequently, excess vesicles were removed by buffer exchange with 1× buffer supplemented with 1 mM MgCl₂. A total of 100 μM CaCl₂ was added from the top to monitor calcium-evoked fusion at a 147-ms frame rate. Vesicle docking, spontaneous fusion, and Ca²⁺-evoked fusion were calculated as described previously.

nMS analysis

All bacterially purified VAMP2^WT/Mutants proteins were buffer exchanged to 200 mM ammonium acetate, 2 mM dithiothreitol, 0.2% Triton X-100, and 0.1% cholesterol buffer using Zeba spin desalting columns (Thermo Fisher Scientific). The protein concentration ranged between 1 and 5 μM. Stable electrospray ionization was achieved with in-house nanoemitter capillaries, formed by pulling borosilicate glass capillaries (outer diameter: 1.2 mm; inner diameter: 0.69 mm; length: 10 cm; Sutter Instruments) using a Flaming/Brown micropipette puller (Model P-1000, Sutter Instruments). These nanoemitters were used for all nMS analyses performed on a Q Exactive UHMR (Thermo Fisher Scientific). For nMS of proteins, the nanoemitter capillary was filled with proteins exchanged in ammonium acetate and installed into the Nanospray Flex ion source (Thermo Fisher Scientific) with a platinum wire. Mass spectrometry parameters were optimized for each sample: spray voltage (1.0 to 1.2 kV), capillary temperature (270°C), resolving power (3125 to 6250 for MS analysis), ultrahigh vacuum pressure (5.51 × 10 to 10 to 6.68 × 10 to 10 mbar), and in-source trapping range (10 to 200 V). The HCD voltage was optimized for each sample, ranging from 0 to 21 V. Data were collected using Thermo Tune software (Q Exactive UHMR 2.13 Build 3159) and analyzed with Thermo Freestyle software. UniDec (75, 76) was used for the final deconvoluted mass calculation and assembled into figures using Adobe Illustrator.

Human gene mutational analyses

To understand the spectrum of genetic variation in SYP and assess its potential involvement in human disease, as previously suggested by the identification of four hemizygous variants in males with intellectual disability in a prior study (77), we systematically screened several large-scale sequencing datasets. These included two population-based resources consisting predominantly of unaffected individuals (gnomAD and the UK Biobank), as well as the Queen Square Genomics Database, which contains exome data from more than 40,000 individuals with diverse neurological disorders across a wide range of ancestries (78). Through this analysis, we identified a previously unidentified, evolutionarily conserved homozygous missense variant in SYP(NM_003179.3):c.95 T > C p.(Leu32Pro) (fig. S10) in a young adult woman with unexplained ataxia, born to consanguineous parents of Middle Eastern origin. While this variant represents a compelling candidate based on its conservation and rarity, its pathogenicity and direct causal relationship to the clinical phenotype remain uncertain and require further functional investigation.

The proband carrying the VAMP2^A104T variant is a 4-year-old boy who presented with global developmental delay, autism spectrum disorder, bilateral congenital talipes equinovarus (clubfoot), hypotonia, minimal verbal language, and stereotypic behaviors. Whole-exome sequencing identified a heterozygous variant of uncertain significance in VAMP2 (NM_014232.3:c.310G > A; p.Ala104Thr), not previously reported in population or clinical databases. The same variant was inherited from his father, who retrospectively reported childhood social delays, balance difficulties, and auditory processing challenges. This intrafamilial segregation, together with functional evaluations presented in this study, supports the pathogenicity of the VAMP2^A104T variant in a spectrum of neurodevelopmental phenotypes, consistent with incomplete penetrance and variable expressivity.

Supplementary Materials

This PDF file includes:

Figs. S1 to S15

Legends for movies S1 to S6

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S6