Synaptophysin chaperones the assembly of 12 SNAREpins under each ready-release vesicle

To test for the formation of oligomeric complexes of Syp and VAMP2 in the native environment of a phospholipid bilayer, we labeled these proteins with different colored dyes, reconstituted them into a common supported lipid bilayer, and determined whether they bound each other and if so in what numerical combinations. By counting single molecules at a very low surface density (~10 Syp molecules per ~1 µm²), we could exclude associations by coincidence despite the limits of optical resolution. We chose supported bilayers because the lateral diffusion of proteins is very slow due to absorption to the glass (40, 41), facilitating cluster analysis.

Therefore, recombinant Halo-tagged Syp was expressed in HEK293 cells, purified from detergent extracts using its 12-his-tag (SI Appendix, Fig. S1C), and labeled with Alexa488 chloroalkane. Recombinant VAMP2 (mutated from wild-type by S28C and C103A to allow for selective labeling with Alexa647-maleimide), was expressed in E. coli and purified according to its 6-his tag (29, 30, 34). These proteins were then co-reconstituted (SI Appendix, Fig. S1C) in a mole ratio of ~4 VAMP2 per Syp into small unilamellar vesicles (SUVs) containing phospholipids with 20 mol% cholesterol, included because cholesterol promotes VAMP-Syp association (21, 42, 43). The SUVs were then applied to a clean glass coverslip with 5 mM MgCl₂ to form supported bilayers (44, 45) and then imaged by TIRF microscopy sequentially with 488-nm (Syp) and 647-nm (VAMP2) lasers. We quantified cluster sizes from both the intensity values and step photobleaching curves.

Syp considered by itself (Fig. 1 B, Top) existed in clusters containing 6 ± 2 copies per cluster. When Syp molecules co-clustering with VAMP2 were considered separately, these Syp clusters (Fig. 1 B, Middle) contained an indistinguishable 6 ± 1 copies of Syp per cluster and now also contained 11 ± 3 copies of VAMP2 (Fig. 1 B, Bottom). Under these conditions ~50% of VAMP2 is co-clustered with Syp, the balance being unbound and free.

From the previous experiment, we can conclude that Syp efficiently forms hexamers in phospholipid bilayers, each of which efficiently sequesters up to ~12 copies of VAMP2, even when these proteins are very sparse. For comparison, the average surface concentrations of VAMP2 and Syp are about 100 and 10 monomers per µm², respectively, as compared to the ~9,000 and ~4,000 molecules/µm² that can be calculated for native SVs from published data (46).

These considerations suggest that complexes between VAMP2 and Syp may also exist in a dilute detergent solution to allow independent single-molecule measurements of stoichiometry following adsorption to cover slips from a dilute solution. Using 1 µM Alexa488-Halo-Syp resulted in clusters containing an average of 3 ± 1 copies (Fig. 2 A, Top). Incubation with unlabeled VAMP2 (4 µM final concentration) produced larger Syp oligomers, now averaging 5 ± 2 copies each (Fig. 2 A, Middle).

This difference suggested that hexamers of Syp not only bind VAMP2 but are also stabilized by VAMP2 binding. If so, sequestering VAMP2 should destabilize the Syp oligomers. This can be achieved by adding soluble t-SNAREs (the 1:1 complex of the cytoplasmic domain of Syntaxin1 and SNAP25) to form SNARE complexes with the VAMP2, and these can be further stabilized by Complexin. This is exactly what happens (Fig. 2 A, Bottom). Now Syp is almost completely dissociated to 2 ± 1 copies, i.e., to monomers.

To independently determine the stoichiometry of VAMP2 binding to Syp in a detergent solution supplemented with 0.1% cholesterol, we used even lower concentrations of Syp-Halo488 (100 nM) to favor dissociation into monomers and also added VAMP2 (400 nM, now labeled with Alexa647). We monitored both molecules coming together at the same spot simultaneously with 100 ms time resolution (Fig. 2B). Under these conditions, we can not only rule out coincidental binding but also determine the stoichiometry of this complex directly with a small number of photobleaching steps.

As anticipated, the vast majority of Syp molecules (90 ± 5%) were now only monomers and only 10 ± 5% of these were in a complex with VAMP2. Each such monomer of Syp contained either one (Fig. 2 B, Top) or two copies of VAMP2 (Fig. 2 B, Bottom) but never more than 2 copies. The ratio of Syp₁Vamp₁: Syp₁Vamp₂ in this population of complexes was ~2:1. Given the concentrations of VAMP2 and Syp in this experiment (400 nM and 100 nM), respectively, and assuming a simple two-step independent binding model where $\begin{array}{cc}\left[Syp\right]+2\left[VAMP\right]\hfill & \to \end{array}$ $\begin{array}{ccccc}\left[Sy{p}_{1}VAM{P}_1\right]\hfill & +& \left[VAMP\right]& \to & \left[Sy{p}_{1}VAM{P}_2\right]\end{array}$ , then the unitary binding constant will be in the range of 1 to 100 µM in the detergent solution.

We conclude that each subunit of Syp can separately bind up to 2 copies of VAMP2 and that a hexamer of Syp in cholesterol-containing lipid bilayers should bind up to 12 copies of VAMP2, in agreement with the measurements using lipid bilayers.

Native Syp-VAMP2 complex was purified from rat brains using a previously described protocol (21). The purity of this complex was established with sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) (SI Appendix, Fig. S3A) and its components were confirmed by western blotting (SI Appendix, Fig. S3B). The ratio of VAMP2 to Syp was 1.7 ± 0.2 moles/mole based on the amount of Coomassie staining relative to subunit molecular masses (SI Appendix, Fig. S3A). The freshly purified samples were applied to grids and were frozen before being imaged by 200-kV Glacios (SI Appendix, Fig. S3C). A total of 731 movies were collected during a single session and the images were analyzed using RELION 3.1 (47). A template-free autopicking procedure was used for autopicking, which yielded a dataset of 635,143 particles. Several rounds of reference-free two-dimensional (2D) classifications were carried out to remove bad particles. Individual 2D classes of the final set of homogenous particles frequently revealed molecules with clear evidence of 6 subunits (SI Appendix, Fig. S3D) as implied by the biochemical analysis (Figs. 1 and 2) and as previously suggested by negative-stained images (19, 20). We believe that these 2D classes represent a Syp hexamer containing unit that can template 12 SNAREpins. Further processing of these selected set of particles (~70,000 particles) yielded 3D density maps with very low resolution which we decided not to pursue. The poor quality of the density maps is possibly due to a combination of reasons such as the presence of a strong detergent micelle of Triton X-100 covering the protein complex, the difficulty in aligning the particles due to the absence of strong cytoplasmic densities for alignment, preferential orientation (lack of side views) as wells as the innate conformational heterogeneity of the purified Syp-VAMP2 complex in solution. Extensive studies of the protein complex reconstituted in nondetergent platforms are required for elucidating a high-resolution structure and revealing the interaction between Syp and VAMP2.

To measure the number of SNAREpins associated with each docked vesicle, we take advantage of the fact that Complexin binds tightly with very low nM affinity to the SNARE complex in 1:1 stoichiometry via its central helix (48–51). In particular, the central helix binds to partially zippered SNARE complexes containing only VAMP2 residues 29 to 60 (52–54). Therefore, Complexin is expected to bind to half-zippered SNAREpins (52). As with ready-release vesicles in vivo, cell-free reconstitutions are clamped beyond this residue (55, 56); thus, we can count clamped SNAREpins simply by counting the number of bound Complexin molecules.

To fluorescently label Complexin1 (SI Appendix, Fig. S2 A and B), we utilize its sole cysteine (residue number 105) which is near the C terminus domain and is uninvolved in binding to VAMP2 and Syntaxin in the SNARE complex (48, 54). This Cys is labeled with maleimide-conjugated Alexa647, a measured efficiency ranging from ~50 to 85% (moles/mole) over numerous preparations (Materials and Methods).

When the labeling efficiency is high (~85%) and there are small numbers (<7) of SNAREpins on a vesicle, discrete steps of quantal photobleaching are reliably observed and counted (see Fig. 3B for examples). On the other hand, when larger numbers of (>6) SNAREpins are present, discrete step-wise photobleaching is rarely observed (Fig. 3C). In this case, the number of fluorescent dye molecules can be measured indirectly and less precisely from the prefactor of the exponential decay curve fitting, and considering average unit fluorophore intensity as measured from single bleaching steps observed in vesicles with low copy numbers in the vicinity (Fig. 3C, see Materials and Methods for details).

To add precision when measuring larger numbers of SNAREpins, we utilized defined mixtures of extensively labeled (~85%) and label-free Complexins produced to achieve partial (25% or 50%) labeling. To illustrate why this is useful, consider a case where there are 12 SNAREpins. This vesicle will bind 12 Complexins. If only 25% bear the dye, then on average every such vesicle will have 3 dye molecules, a small number that can be reliably counted by photobleaching (Fig. 4 D, Left panel). However, the actual number will differ from vesicle to vesicle. Some individual vesicles will have bound (at random) 4 dyes and 8 dye-free Complexins; others 2 dyes and 10 dye-free Complexins, and so on. The distribution of labeled complexin follows a binomial distribution with a probability equal to the labeled fraction. Hence, the raw discrete step data for 25%, 50%, or undiluted samples (Fig. 3 D and E) can be fitted by a binomial distribution to calculate the average number of SNAREpins for the entire vesicle population more accurately than would have been possible from total intensity analysis. The variation of the mean for the fraction of labeled Complexin follows a straight line with an intercept at 0 and a slope corresponding to 12 ± 1 SNAREpin per vesicle (Fig. 3F).

We performed several controls to further establish the specificity of the Complexin signal for counting SNAREpins. As expected, the Complexin fluorescence signal is only detected on the suspended bilayer portion of the chip where it is associated with VAMP2-containing SUVs docked to t-SNARE-containing suspended bilayers (SI Appendix, Fig. S2C) and is not observed when vesicles lack the v-SNARE and are instead docked by Synaptotagmin1 (SI Appendix, Fig. S2D). This also indirectly rules out the possibility that the labeled Complexin binds to docked vesicles via its curvature-sensing C-terminal amphipathic helix (57, 58) to a detectable extent when diluted for single-molecule studies under our conditions. This was directly established by deleting the amphipathic helix (Complexin1 residues 26 to 83; SI Appendix, Fig. S5). Using this probe and the linear fitting of the mean values (SI Appendix, Fig. S5D) we measured 11 ± 2 SNAREpins per vesicle, indistinguishable from the 12 ± 1 measured under the same conditions using the standard full-length Complexin probe (Fig. 4 B, Top). These controls, together with the nature of the results obtained, validate the sensitivity and specificity of the methods we employ to count SNAREpins.

Based on the robust observation of a VAMP₁₂-Syp₆ complex in lipid bilayers (Fig. 1) and the frequent observation of discrete VAMP_1-2-Syp₁ subcomplexes in dilute solution (Fig. 2), we hypothesize that the 12 copies of VAMP2 bound to Syp in vesicles will be kinetically advantaged in forming SNAREpins with t-SNARE-containing bilayers. Such a structure, involving 12 SNAREpins, a Syp hexamer, and potentially numerous copies of Syt, would occupy most or all of the interface between the vesicle and bilayer, excluding other SNAREpins from forming. Therefore, our hypothesis predicts that in the absence of Syp, there will be a variable number of SNAREpins, but adding Syp will standardize the number to (theoretically) 12.

We first measured the number of SNAREpins in vesicles lacking Syp (Fig. 4A) but containing variable (5 to 60) average numbers of externally oriented VAMP2 and ~22 (on average) copies of externally oriented Syt and labeled Complexin mixtures. In all of these compositions, vesicles are known from previous work to be efficiently clamped and synchronously released (<13 ms) when Ca⁺⁺ is added (29, 34). With low (5 or 10) copies of external VAMP2, the ready-release vesicles had 6 ± 1 and 7 ± 1 SNAREpins on average, respectively (Fig. 4 A, Top two panels). With high (25 or 60) copies of external VAMP2, the ready-release vesicles had now had 19 ± 6 (Mean ± SEM) and 23 ± 4 SNAREpins on average, respectively (Fig. 4 A, Bottom two panels), but the numbers in individual cases are highly variable and can even exceed ~30 SNAREpins per vesicle. These cases seem mainly to involve larger vesicles in the population (SI Appendix, Fig. S7) and likely correspond to cases where there are extensive contact surfaces (as distinct from a discrete point of contact) between docked vesicles observed by Cryo-electron microscopy (cryo-EM) (6, 7). For comparison, native SVs contain about 70 copies of VAMP2, ~30 copies of Syp (46), and ~15 to 20 copies of Syt1 (46, 59).

When the native Syp-VAMP2 complex purified from rat brains (~30 copies of Syp per vesicle and ~60 copies of VAMP2) was added during reconstitution, the results were dramatic. Now only 12 ± 1 SNAREpins were assembled (Fig. 4 B, Top), compared with the exceptionally broad distribution ranging from 5 to 35 in the absence of Syp (Fig. 4 A, Bottom). Syp limited the number of SNAREpins to 12 ± 2 even in the absence of Syt (Fig. 4B, 2nd from top). We also introduced recombinantly purified Syp (~30 copies per vesicle) from HEK 293 cells and supplemented with recombinant VAMP2 (from E.coli; ~60 copies per vesicle). We observed similar results with a slightly broader distribution.

Importantly, when Syt1 was introduced, the result was a very narrow distribution that closely fits the predicted binomial distribution for 12 SNAREpins considering the experimental dye labeling efficiency of 85% (SI Appendix, Fig. S4). When Syt1 was present, but could not bind SNAREpins (due to mutations in its primary binding site; Fig. 4B, 3rd panel, Syt1Q), or could not assemble ring oligomers (due to an F349A mutation; Fig. 4 B, Bottom panel), SNAREpin templating by Syp was largely unaffected (12 ± 2 and 13 ± 1 SNAREpins per vesicle, respectively); however, distributions were broadened significantly. In control experiments, we confirmed that including Syp (30 copies) in the vesicles along with VAMP2 (60 copies) and Syt (22 copies) did not compromise clamping by Complexin (SI Appendix, Fig. S6A) or synchronous release (SI Appendix, Fig. S6 C and D). Interestingly, Complexin-dependent clamping was observed even in the absence of Syt as long as Syp was present (SI Appendix, Fig. S6A). However, Ca⁺⁺-dependent release from this (Complexin plus Syp) clamped state did not occur in the absence of the Ca⁺⁺-sensor Syt.

To study the kinetics of assembly of the 12 SNAREpins templated by Syp we continuously monitored the appearance of immobilized docked vesicles (Atto465) and SNAREpin formation (Complexin-647). Vesicles can be docked by Syt binding to PIP₂ in the suspended bilayer and/or by the first SNAREpin to form. Therefore, the vesicle always appears before or together with the Complexin fluorescence. Unexpectedly, we observed that the 12 SNAREpins form in two very distinct phases (Fig. 5A). In the first phase 6 ± 1 SNAREpins are rapidly assembled. Then, following a variable lag phase, a second rapid wave of assembly occurs, resulting in an additional 5 ± 2 SNAREpins. In other words, half the SNAREpins assemble right away, and the other half assemble after a prolonged delay, which averages a highly variable 0.5 ± 1 s (SI Appendix, Fig. S9D). Unlike the variable lag phase, the SNAREpins assemble at indistinguishable and rapid rates in the first (34 ± 21 SNAREpins/s) and second phases (30 ± 21 SNAREpins/s), corresponding to about 30 ms per SNAREpin.

The basis of the two phases of SNAREpin assembly seems to be in Synaptotagmin, and in particular to binding of SNAREpins to the simultaneously assembling ring of Synaptotagmin. Omitting Syt but retaining Syp produces ~12 SNAREpins, but they now assemble rapidly in a single phase (Fig. 5B). Replacing Syt with a mutant that does not bind SNAREpins (Syt1Q) has the same effect (Fig. 5C). Abolishing the ring organization but not SNAREpin binding (Syt1F349A) also eliminates the lag phase but now also greatly slows the rate of SNAREpin assembly (Fig. 5D). Introducing tripartite mutant in Syt1 has a little effect in the total number of SNAREpins; however, it produced wide variations in the SNAREpin assembly kinetics (SI Appendix, Fig. S10). Finally, the two phases of SNARE assembly are even observed in the absence of Syp, but now templating to a limit of 12 SNAREpins is lost (SI Appendix, Figs. S8 and S9). In the absence of Syp, the first phase produces an average of 10 ± 5 SNAREpins, and the second phase has an average of 10 ± 6 SNAREpins for a total of 21 ± 9 SNAREpins (SI Appendix, Fig. S9D).

The following cDNA constructs were used in this study: full-length VAMP2 (VAMP2-His⁶, residues 1 to 116); full-length VAMP2^4× (VAMP2-His⁶, residues 1 to 116 with L70D, A74R, A81D, L84D mutations), full-length t-SNARE complex (mouse His⁶-SNAP25B, residues 1 to 206 and rat Syntaxin1A, residues 1 to 288); Synaptotagmin (rat Synaptotagmin1-His⁵, residues 57 to 421); and Complexin (mouse His⁶-Complexin 1, residues 1 to 134). All mutants including Syt1^F349A (F349A); Syt1^Q (R281A/E295A/Y338W/R398A/R399A); were generated as described previously (24, 29, 34). Mouse complexin 26 to 83 was modified with a C-terminal Cys using site-directed mutagenesis. 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). ATTO647N-DOPE and ATTO465 DOPE were purchased from ATTO-TEC, GmbH (Siegen, Germany). Alexa maleimide 647 C2 and Alexa Fluor^R 488 Halo ligand were purchased from Thermo Fisher Scientific (Waltham, MA) and Promega Madison, USA), respectively.

v- and t-SNAREs, Cpx, Syt1 recombinant proteins were expressed and purified as described previously (24, 29, 69). The mouse Syp gene was cloned as a BamHI/NotI fragment into a modified pCMV6-AN-His vector (Origene, USA) that contains an N-terminal 12×His tag followed by a PreScission cleavage site. The plasmid was used to transfect Expi293F cells (Thermo Fisher, USA) following the manufacturer’s protocol. Briefly, 1 µg of DNA per 1 mL of cell culture was mixed in a 1:2.5 ratio of DNA: ExpiFectamine reagent for 20 min at RT and then added to approximately 2.5 × 10⁶ cells with >95% viability. Cells were grown at 37 °C under 8% CO2 with orbital shaking for 2 d without enhancers. The cells were centrifuged at 200 × g for 10 min, washed with Phosphate-buffered saline (PBS), and flash frozen. Harvested pellets were stored at –80 °C until needed. Expi293F cell pellets (from 50 mL of culture) were thawed and resuspended in 8 mL of 0.25 × 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 100,000 × g for 45 min. The pellet was resuspended in 5 mM NaH₂PO₄ (pH 6.8), 0.25% cholesterol (Avanti; Ovine Catalog # 700000P), and protease inhibitors. Triton X-100 was added to a final volume of 0.5% (v/v) and incubated on ice for 2 h with occasional inversion. The lysate was centrifuged at 45,000 × g for 30 min to pellet the insoluble fraction, and solubilized protein was incubated with 0.125 µL of HisPur Ni-NTA resin (Thermo Fisher, USA) at 4 °C overnight on a rotating wheel. The next day, the resin was washed multiple times with 20 mM NaH₂PO₄ (pH 6.8), 0.5% Triton X-100, 0.05% cholesterol, and 40 mM sucrose. Protein was eluted using the same buffer with 500 mM Imidazole (pH 6.8).

SVs were purified from Sprague Dawley rat brains as described previously (70). Briefly, four rat brains were homogenized in an ice-cold buffer solution (4 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) (HEPES), pH 7.4, and 320 mM sucrose). The homogenate was centrifuged at 1,000g for 10 min at 4 °C. The pellet was removed, and the resulting supernatant was centrifuged at 15,000g for 15 min at 4 °C. The supernatant was collected and stored at 4 °C while the pellet containing synaptosomes was resuspended in the homogenization buffer and was diluted tenfold with ice-cold ddH₂O and homogenized thoroughly. The lysate was centrifuged at 17,000 g for 15 min at 4 °C. The pellet was discarded, and the resulting supernatant was combined with the supernatant from the previous step and centrifuged at 48,000 g for 25 min at 4 °C. The supernatant was further centrifuged at 300,000 g for 2 h at 4 °C, and the pellet containing SVs was collected. The SVs were immediately resuspended in 3 mL of 5 mM sodium phosphate (pH = 6.8), supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 µg/mL Pepstatin A and snap frozen in liquid nitrogen and stored in –80 °C until further use.

The Syp-VAMP2 complex was purified from SVs as described previously (21). Briefly, SVs purified from four rat brains were solubilized for 45 min at 4 °C (at 5 mg/mL) in 5 mM sodium phosphate (pH = 6.8) buffer supplemented with 1 mM PMSF, 1 µg/mL Pepstatin A, 0.2% Triton X-100, 1% cholesterol, and protease inhibitor cocktail (Roche). The solubilized Triton X-100 extract was centrifuged at 142,000 g for 45 min at 4 °C. The supernatant was applied to a dry hydroxyapatite/celite (2:1 w/w) column and was eluted with 20 mL of 5 mM sodium phosphate (pH = 6.8), 0.2% Triton X-100, 0.1% cholesterol, 1 mM PMSF, 1 µg/mL Pepstatin A, and protease inhibitor cocktail (Roche, USA). The first five fractions of the flow-through were collected as 1 mL fractions and the rest were collected as 2 mL fractions. A Coomassie gel was run to determine the purity of the Syp-VAMP2 complex in the collected fractions (SI Appendix, Fig. S1A). The Syp-VAMP2-containing fractions were then pooled and concentrated to 2 mg/mL at 4,200 g and 4 °C using 4-mL Amicon Ultra 100-kDa molecular weight cutoff centrifugal filters. The purified protein was stored at 4 °C and was immediately used for cryo-EM sample preparation and reconstitution into liposomes. The purified protein was analyzed by western blot using monoclonal Syp1 (Synaptic Systems, Germany) and monoclonal VAMP2 (Abcam, USA) antibodies to confirm the presence of the Syp-VAMP2 complex (Fig. 1A).

2.5 µL of freshly purified Syp-VAMP2 complex (1.5 mg/mL) was applied to C-flat-1.2/1.3-2C-T copper grids that had been glow discharged for 10 s in Argon/Oxygen atmosphere using a Solarus plasma cleaner (Gatan, Inc.). The grids were blotted for 2 s with a blot force of –1 at 4 °C under 100% humidity and were flash-frozen in liquid ethane using a FEI Vitrobot Mark IV (FEI). The grids were stored in liquid nitrogen until data acquisition.

The frozen-hydrated samples were loaded on the 200-kV Glacios (Thermo Fisher Scientific) with a K2 Summit direct electron detector (Gatan, Inc.). The micrographs were collected in super-resolution mode operated using SerialEM (Version 3) software (71). A total of 731 movies were acquired at a nominal magnification of 36,000× with a physical pixel size of 1.143 Å on the specimen level. The movies were dose fractionated into 40 frames of 0.25 s each at a dose rate of 5.8 electrons/Ų/s for a total dose of 58 e^–/ Ų over a defocus range between –0.9 µm and –2.0 µm. The image processing was done using RELION 3.1 (47). All the 731 micrographs were motion-corrected and dose-weighted using MotionCor2 (72) with a binning factor of 2 and divided into 5 × 5 patches. The contrast transfer function was calculated with CTFFIND-4.1 (73). A template-free autopicking procedure based on Laplacian-of-Gaussian filter was used for autopicking, which yielded a dataset of 635,143 particles. Several rounds of 2D classifications using a mask of 70 Å diameter were carried out to remove bad particles and obtain a final set of homogenous particles.

t-SNAREs and VAMP2 + Syt1 were reconstituted into SUVs using rapid detergent (1% Octyl glucoside) dilution and dialysis method as described previously (24, 29, 34, 35). The lipid composition was ~82 (mole)% DOPC, 15% DOPS, 3% PIP₂, and 0.1% ATTO465-PE for t-SNARE SUV and 88% DOPC, 10% PS, and 2% ATTO647-PE for VAMP2 and Syt1 SUVs. We used several different lipid: protein ratios in order to reconstitute to desired copy numbers (outward facing) of individual proteins. The copy number of outside-facing proteins was determined using our previously well-established method (29, 34, 35). Recombinant Syp and VAMP2 (± Syt1) were reconstituted into 50 µL of 3 mM lipids [68% (mole) DOPC, 10% DOPS, 20% cholesterol, and 2% ATTO465] that had previously been dried down to a film in a glass test tube using nitrogen gas and then stored under a vacuum for 1 h to remove any residual chloroform. Lipids were resuspended in protein and buffer (25 mM HEPES, pH 7.4, 400 mM KCl, 1 mM DTT, and 0.5% Triton X-100) up to a final volume of 100 µL and then passed through a detergent removal spin column (Pierce, catalog #87777) following the manufacturer’s protocol. Volume was increased to 150 µL and dialyzed against 25 mM HEPES, 120 mM KCl, and 1 mM DTT overnight to remove any residual detergent. The next day, 150 µL of 80% Nycodenz (in the same buffer) was mixed with the sample and then loaded into a 0.8-mL tube (Beckman Coulter, catalog #344090), overlaid with 250 µL of 30% Nycodenz, followed by 50 µL of HEPES buffer. Tubes were centrifuged in an SW55 rotor at 48,000 rpm for 4 h. Proteoliposomes were collected from the gradient near the buffer/30% Nycodenz interface. Native Syp-VAMP2 complex was also reconstituted following the same protocol except no extra VAMP2 was supplied. Protein reconstitution efficiency was routinely checked with SDS-PAGE and Coomassie staining (SI Appendix, Fig. S1). Based on the densitometry analysis of Coomassie-stained SDS gels, the copy number of each protein per vesicle was estimated from trypsin digestion and Coomassie blue staining as described previously.

To measure the stoichiometry of the Syp-VAMP2 complex on a supported bilayer, 1 mM (total) SUVs [60 (mole)% DOPC, 15% DOPS, 5% DOPE-PEG 2000, and 20% cholesterol] were dried with N₂ gas and kept under vacuum for another 2 to 4 h. Dried lipid was dissolved with 25 mM HEPES, pH 7.4, 150 mM KCl, 0.2 mM tris(2-carboxyethyl)phosphine (TCEP), and 2% Triton X-100. Prelabeled and purified Syp-Halo (Alexa 488) and VAMP2 (Alexa 647) proteins were added into the mixture with a protein: lipid ratio of 1:20,000 and 1:5,000, respectively. 100 mg prewashed Bio-beads SM2 (Bio-Rad, Hercules, CA) were incubated with that mixture for another 30 min at room temperature with gentle shaking. The liposomes were dialyzed overnight with a 6 to 8 kD cutoff against a detergent-free HEPES buffer. The liposomes were floated up with a discontinuous gradient of Opti prep. Samples were collected from the top carefully avoiding any Opti prep contamination and further dialyzed for another 2 h at 4 °C. The supported bilayer was created by Mg²⁺ (5 mM) induced bursting of the liposomes in ibidi glass-bottom chambers (ibidi GmbH, Germany). The bilayer was extensively washed with HEPES buffer. The bilayer was imaged with a TIRF (Nikon) microscope using 488- and 633-nm lasers. Photobleaching of the protein cluster was analyzed with Image J.

Recombinantly purified and labeled Syp-halo 488 (1 µM) was mixed with unlabeled VAMP2 at a 1:4 molar ratio overnight at 4 °C in a buffer containing 5 mM NaH₂PO₄, pH 6.8, 50 mM KCl, 0.2% Triton X-100, 0.2 mM TCEP, and 0.1% cholesterol. A 5-µL sample was spotted very briefly onto a freshly BSA-coated glass bottom dish (MatTek Corporation, Ashland, MA) and then quickly washed with excess buffer. Syp-halo 488 was imaged with TIRF (Nikon) using a 488-nm laser. The same overnight mixture was further supplemented with 2 µM preassembled t-SNAREs and Complexin. After 30-min incubation at room temperature, a similar method was applied to count Syp again. For dual monitoring, both Syp-Halo 488 (1 µM) and VAMP2 Alexa 647 (4 µM) labeled proteins were incubated overnight at 4 °C in a buffer consisting of 5 mM Na₂HPO₄, pH 6.8, 50 mM KCl, 0.2% Triton X-100, 0.2 mM TCEP, and 0.1% cholesterol. The mixture was diluted (10×) with the same buffer and quickly spotted on the glass bottom dish (BSA coated). Both proteins were monitored using 488- and 647-nm lasers with dual viewer 2 equipped in Nikon TIRF.

All the single-vesicle fusion measurements were carried out with suspended lipid bilayers as previously described (29, 30, 35). Complexin at a final concentration of 2 μM was added before the addition of vesicles when required. Vesicles (100 nM lipids) were added from the top and monitored for 10 min using a 641-nm laser. We employed ATTO647-PE fluorescence to track vesicle clamping and fusion events. Fusion (lipid mixing) was attested by monitoring a burst and then a rapid decrease in fluorescence intensity as the lipid diffuses away. The fate of each vesicle was monitored and plotted as a “survival curve” which represents the fraction of vesicles that docked to the planar bilayer and proceed to spontaneous fusion or immobile docked. After the initial 10-min interaction phase, 100 μM CaCl₂ was added from the top to monitor the effect of Ca²⁺ on the docked vesicles. Again, the vesicle survivals of the immobile docked vesicles were analyzed, and fractions were binned with 147 ms time. Calcium was monitored as with Ca²⁺- sensor Calcium Green conjugated to a lipophilic 24-carbon alkyl chain (Calcium Green C24) introduced in the suspended bilayer to directly monitor the arrival of Ca²⁺. Details of this method were described in our earlier work. The fusion probability was calculated at the end of 2 min incubation. All experiments were performed at 37 °C using an inverted laser scanning confocal microscope (Leica-SP5) equipped with a multiwavelength argon laser including 488, diode lasers (532 and 641 nm), and a long-working distance 40X water immersion objective (NA 1.1).

Purified complexin proteins (Cpx1^1-134 and Cpx1^26-83-Cys) and VAMP2 (S28C, C103A) were incubated overnight with Alexa maleimide 647 C2 dye (Thermo Fisher Scientific Cat No: A20347) with a 1:5 molar ratio in the presence of 5% Dimethylsulfoxide (DMSO). Syp-Halo protein was incubated with 5 times molar excess halo ligand Alexa 488 (Promega, USA) overnight at 4 °C. Excess dye was removed twice using the dye removal column (Thermo Scientific, Cat No 22858), and then, protein was loaded into Superdex 200 and concentrated using Amicon Ultra-4 (3K cutoff, Merck). The labeling percentage was calculated according to the manufacturer's protocol. Protein concentration was checked with Bradford assay and dye concentration was measured from the absorbance (Alexa 647 at 651 nm and Alexa 488 at 494 nm), and labeling efficiency was calculated with three repeats.

All the complexin counting experiments underneath docked vesicles were performed with TIRF (Nikon) on a reconstituted planar suspended bilayer on the silicon surface. The method to create a suspended bilayer compatible with single-molecule imaging was described in detail elsewhere (39). Briefly, t-SNARE-containing giant unilamellar vesicles (~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, 140 mM KCl, and 1 mM DTT) 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, 2 µM total complexin (with corresponding labeled protein) was 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 monitored both vesicle and complexin simultaneously with 488- and 633-nm solid-state laser, respectively, using dual viewer 2. To generate more statistics on the vesicle and complexin counts, excess complexin was removed by washing three times with HEPES buffer. Both vesicle and complexin were imaged, and complexin signals were subjected to photobleaching using a 633-nm laser with 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. However, we could not precisely determine the number of steps when the number of molecules was higher than 5 to 7. The photobleaching curves were fitted with an exponential decay curve with Eq. 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.

Where i is the average unit intensity of a single fluorophore as determined from the small clusters. For rate of assembly, vesicle and corresponding complexin signal were plotted to monitor the SNAREpin formation. Realtime SNAREpin formation was calculated from the complexin signal using Eq. 3.

I_Cpx(t) represents the complexin intensity at time = t, B = background intensity, l_p =labeling percentage factor, and i, average intensity for one complexin molecule. All the complexin counting data were converted to histograms with 1 bin size. The prediction was calculated using a binomial distribution. We assume that there are exactly N_S SNAREpins that form in the contact between the vesicle (Syp³⁰ VAMP2⁶⁰ Syt1²²) and the membrane and that a single Cpx binds to each SNAREpin. Only a fraction f of the Cpx molecules is labeled. Since Cpx binds randomly to SNAREpins, the number of fluorescent dyes, n, follows a binomial distribution:

Using N_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.

M.B., A.R., S.R., F.P., and J.E.R. designed research; M.B., A.R., J.C., and R.V.K.S. performed research; M.B., A.R., S.R., and F.P. analyzed data; and M.B., A.R., J.C., F.P., and J.E.R. wrote the paper.

The authors declare no competing interest.