SNAP-25 disease variants affect synaptic transmission by destabilizing SNARE-complexes within a multimeric SNARE-ring

Summary

SNAREopathies are neurodevelopmental syndromes arising from de novo mutations in the synaptic machinery. Two variants are SNAP-25 I192N and I192T, which cause early lethality and moderate developmental delay, respectively. Both variants led to destabilization of the C-terminal end of the SNARE-complex and reduced binding to the Munc18–1:VAMP2:syntaxin-1 acceptor complex, while reducing spontaneous and evoked synaptic release in autaptic glutamatergic neurons (severity: I192N>I192T). I192N, but not I192T, negatively affected neuronal survival and dendritic branching. I192N was strongly dominant-negative in presence of wildtype SNAP-25, leading to impaired survival and reduced synaptic release. The dominant-negative effect was stronger for spontaneous than for evoked release and mild for sustained release. We considered how release is affected upon random incorporation of variant and wildtype protein into a ring of SNARE-complexes. We suggest that only a subset (approximately three) of the pre-assembled SNAREs (approximately six) mediates evoked release, rendering evoked release less susceptible to variants.

Introduction

The etiology of neurodevelopmental disorders has gained significant attention in recent years. In chemical synapses, variants affecting the neuronal SNARE-machinery responsible for neurotransmitter release give rise to a distinct class of diseases, “SNAREopathies”, characterized by epilepsy, developmental delay, motor problems and in some cases autism ¹. While various SNAREs mediate membrane-to-membrane fusion events throughout the cell, the neuronal SNAREs are uniquely suited to synchronize vesicle-to-membrane fusion with action potential arrival on the sub-millisecond time scale, which accounts for the severe consequences of their mutation ¹.

SNAP-25 and its two primary interaction partners, syntaxin-1 and VAMP2, forms the SNARE complex², the smallest functional unit of the machinery required for fusion of membranes in vitro ³. The SNARE complex is a helix bundle composed of four α-helical SNARE domains, two from SNAP-25, and one each from VAMP2 and syntaxin-1 ². These helices are arranged in parallel and stabilized by 16 interaction layers, numbered from −7 to +8 (N- to C-terminus). The layers form (or ‘zip’) in discrete steps: the N-terminal layers form during vesicle priming, while C-terminal assembly, triggered by Ca²⁺ influx as the go-signal, provides the energy required for vesicle fusion ^4,5. Several proteins control assembly of the SNARE-complex as well as coupling its formation to Ca²⁺ influx. Munc18–1 and Munc13–1 chaperone formation of the SNARE complex ^6–12 while simultaneously protecting against untimely dismantling by N-ethylmaleimide Sensitive Factor (NSF) ^13–15. The fast calcium sensor synaptotagmin-1 ensures that fusion only occurs once Ca²⁺ enters the presynaptic terminal ^16–18. In the presence of Munc18–1, a trimeric syntaxin-1:VAMP2:Munc18–1 complex forms first, which effectively serves as a dynamic template complex for subsequent SNAP-25 binding ^7,11,19–22. Munc13–1 facilitates the formation of the template complex by changing the configuration of the initial Munc18–1:syntaxin-1 complex ^6,23, and by membrane bridging ²⁴. It may also assist SNAP-25 recruitment to the acceptor complex ²⁵. In contrast, tomosyn competes with SNAP-25 for binding to the template complex, thereby inhibiting SNARE complex assembly and neurotransmitter release ²⁶. Therefore, the template complex is a key intermediate for SNARE assembly in vivo ²⁷.

How SNAREopathy variants lead to synaptic defects is an active field of inquiry. Their effects on synaptic transmission appear to be strikingly diverse. Variants within the C2B-domain of synaptotagmin-1 weaken Ca²⁺-dependent membrane binding, typically leading to dominant-negative effects ²⁸. In SNAP-25, missense variants cluster within the SNARE-domains ²⁹, with those at the primary SNARE:synaptotagmin-1 interface ³⁰ strongly increasing spontaneous and asynchronous evoked release ^31,32. Two such variants acquired gain-of-function features, allowing them to integrate into the SNARE-complex while bypassing fusion control ³². Conversely, destabilizing SNAP-25 variants within the SNARE-bundle led to loss-of-function features and concurrent reductions in spontaneous and evoked release ^31–33. In VAMP2, disease variants also primarily map onto its SNARE-domain ³⁴ and affect membrane fusion ³⁴. Similarly, disease variants in syntaxin 1B (STX1B) localize to its SNARE-domain and the H_abc-domain ^35,36. Beyond SNARE proteins, a Munc13–1 gain-of-function variant was found to be associated with dyskinesis, developmental delay and autism ³⁷. Thus, both reductions and increases in synaptic transmission may be associated with disease. The most frequent cause of SNAREopathy is de novo heterozygous mutation in STXBP1, the gene encoding Munc18–1 ¹, which leads to cortical hyperexcitability ³⁸.

Most people affected by SNAREopathies are heterozygous for the disease variant. In the case of STXBP1, variant protein is usually unstable, leading to haploinsufficiency ^1,39. However, in other SNAREopathies, most missense variants are stable, and therefore co-expressed with wildtype protein in affected individuals. Because synaptic vesicle priming results in the formation of a multimeric ring of partially formed SNARE-complexes between vesicle and plasma membrane ^40–43, coexpression results in a situation where both wildtype and variant protein might be incorporated into the same multimeric fusion machine. The implication for synaptic transmission is currently unclear.

We here present a study of two SNAP-25 missense variants of residue I192 with very different clinical severity, I192N and I192T. Residue I192 localizes to layer +5 within the C-terminal half of the SNARE-complex, which generates the force necessary for membrane fusion. We investigated the functional impact of I192T and I192N using single-molecule force measurements, single-cell physiology, and mathematical modeling of the multimeric SNARE-ring involved in fusion.

Results

The two SNAP25 missense variants I192T and I192N were identified in a 14-year-old female with moderate developmental delay ²⁹, and a severely affected boy who died within 1 year of age, respectively ^31,44. Both individuals were heterozygous for the variant. We aimed to characterize the molecular and cellular defects with the goal of understanding the consequences of their incorporation into the multimeric SNARE-ring.

I192T and I192N mutants affect synaptic release and short-term plasticity

We introduced the I192T and I192N variants into neurons using a lentiviral EGFP-fused SNAP-25 construct, which we used previously for rescuing synaptic release in Snap-25 KO neurons ^32,45–47. These constructs resulted in similar expression levels (Fig. 1I) when expressed in wildtype (WT) neurons, like other SNAP-25 disease variants so far investigated ^31–33. We investigated the phenotype of each variant in autaptic hippocampal glutamatergic neurons from Snap25 knockout (KO) embryos ⁴⁸ seeded on pre-grown glial islands ⁴⁹ and transduced on DIV (day in vitro) 1 with one of the variants or WT EGFP-SNAP-25. Since SNAP-25 is essential for neuronal survival and arborization ^48,50,51, we examined neuronal morphology of autaptic neurons by staining against the dendritic marker MAP2 and the excitatory synapse marker vGLUT1 (Fig. 1A). Glutamatergic neurons expressing I192T or I192N exhibited no changes in the number of synapses per dendritic length (Fig. 1C) or per neuron (Fig. 1D), or in synaptic size (Fig. 1F). However, I192N neurons did show a tendency for shorter total dendritic length (Fig. 1G, p = 0.124). Consistently, in a Sholl analysis they had fewer crossings as well as a smaller dendritic arborization (Fig. 1H, repeated measures ANOVA: interaction between variant and distance: p <0.0001). When counting the number of neurons per glial islet, we found that I192T supported an unchanged number of neurons compared to WT (Fig. 1B), whereas I192N reduced survival (Fig. 1E). When co-expressing WT and variant at 1:1 ratio ³², I192N still significantly depressed survival compared to the WT (Fig. 1E), whereas WT and I192T co-expressing neurons displayed unchanged survival (Fig. 1B). Snap25 KO neurons essentially do not survive (<1% survival rate) on astrocyte islands ⁵⁰; I192N therefore does support survival to some extent, but not at WT levels.

Figure 1. Neuronal survival and morphology in neurons expressing SNAP-25 I192T and I192N variants.

A: Autaptic Snap25 KO neurons expressing WT EGFP-SNAP-25 (left), I192T (middle) or I192N (right) stained for MAP2 and vGLUT1. Neuronal morphology is traced in black using the MAP2 staining. The inserts show MAP2 staining in red and vGLUT1 staining in cyan. The scale bars represent 50 μm.

B, E: Number of neurons per glial micro island. Expression of I192T in the presence or absence of WT EGFP-SNAP-25 did not affect neuronal survival, whereas expression of I192N reduced survival of the neurons.

C, D, F: Synapse density per μm dendrite length (C), total number of synapses per neuron (D) and synapse size (F).

G: Total neurite length per neuron. There is a slight tendency of lower dendritic length in I192N (p = 0.124). Same cells as in C, D and F.

H: Number of intersections between neurites and concentric circles of different diameters. I192N neurons had a more contracted dendritic tree: p<0.0001 for the interaction between SNAP25 variant and distance. Same cells as in C, D and F.

I. Quantification of SNAP-25 expression (EGFP-SNAP-25 divided by endogenous SNAP-25) shows similar expression levels of all three constructs. VCP was used as a loading control. The expressed EGFP-SNAP-25 is distinguished by its higher molecular weight.

Data are represented as individual points and mean ± SEM is displayed. In panels C and E, the groups are compared using a Brown-Forsythe ANOVA with Dunnett’s multiple comparisons test. In B, D, F and G, groups are compared using a one-way ANOVA with a Dunnett’s multiple comparisons test. In panel H, a repeated-measures ANOVA was used. ***p<0.001, ****p<0.0001. See also Table S2.

We performed whole-cell patch-clamp on autaptic glutamatergic Snap25 KO neurons transduced with WT, variant (I192N or I192T), or WT and variant at approximately 1:1 ratio ³² (Fig. 2). Variant-expressing neurons were recorded alongside WT-expressing neurons from the same hippocampal preparations. We first recorded miniature excitatory post synaptic potentials (mEPSCs) while holding the cell body at −70 mV. These measurements showed that I192T significantly reduced the mEPSC frequency, whereas the WT and I192T co-expressing neurons displayed a mild non-significant reduction in mEPSC frequency (Fig. 2A, C). In contrast, for neurons expressing I192N alone, mEPSCs were essentially absent, while co-expression of I192N and WT resulted in a strong and significant decrease in mEPSC frequency (Fig. 2G, I). Neither variant affected the mEPSC amplitude (Fig. 2B, H), indicating that although SNAP-25 is also found postsynaptically⁵², the primary effect of these variants is presynaptic. The autaptic culture allows quantitative measurement of synaptic transmission without reverberating activity, which complicates measurements of glutamatergic neurotransmission in larger cultures. Thus, to examine evoked release we elicited an unclamped action potential (AP) in the axon. These measurements showed a reduction in evoked excitatory post synaptic potentials (eEPSC) amplitude for I192T expressed alone but not in the presence of WT (Fig. 2D). eEPSCs were virtually absent when I192N was expressed alone and strongly reduced in amplitude when I192N was co-expressed with WT SNAP-25 (Fig. 2J).

Figure 2. I192T and I192N mutants affect synaptic release and short-term plasticity.

A, G: I192T (A) and I192N (G) reduced mEPSC frequencies when expressed in Snap25 KO neurons alone or together with WT EGFP-SNAP-25.

B, H: mEPSC amplitudes were not affected by either variant.

C, I: mEPSCs recorded from neurons expressing WT (black), WT+I192T (light red) or I192T alone (dark red) (C) or WT (black), WT+I192N (light green) or I192N alone (dark green) (I).

D, J: I192T (D) reduced the amplitude of eEPSC, whereas I192N (J) abolished evoked release when expressed alone and reduced the amplitude when co-expressed with WT.

E, K: eEPSC amplitudes (mean ± SEM) during a 40Hz train in neurons expressing I192T (E) or I192N (K). Expression of I192T or WT+I192N induced facilitation of release. For the I192N, a non-responsive and a responsive neuron are shown. F, L: Paired-pulse ratios. Neurons expressing I192N often did not respond to the first stimulation and are therefore not included. Data are represented as individual points and mean ± SEM. Three-group comparisons were carried with one-way ANOVA or Brown-Forsythe ANOVA with a Dunnett’s multiple comparisons test for pairwise comparisons. PPRs in L were compared using an unpaired Welch’s t-test. See also Table S2 and Fig. S1.

We next stimulated the neurons repetitively to probe short-term plasticity. I192T-expressing neurons displayed a shift towards facilitation, indicated by an increase in paired-pulse ratio (PPR), whereas those expressing both I192T and WT showed no clear phenotype (Fig. 2E, F). The first few APs within a train were often unable to induce any release in most I192N neurons, only later did small EPSCs appear (Fig. 2K). Neurons co-expressing I192N and WT exhibited facilitation during train stimulation (Fig. 2K), and higher PPR values (Fig. 2L). We separated release during the train into synchronous and asynchronous release components, where the former is the release component triggered by each AP, and the latter is the continuous ‘baseline’ release upon which synchronous responses are sitting ^53,54. Interestingly, this revealed that in Snap-25 KO neurons expressing only I192N, the synchronous component towards the end of the train was unchanged compared to SNAP-25 WT, whereas asynchronous release was reduced (Fig. S1A). Since neither release component is seen during train stimulation in Snap25 KO neurons ⁵⁰ , this shows that although I192N is unable to support any EPSC upon single stimulation (Fig. 2G), its ability to support synchronous release at high stimulation intensities remains unchanged.

Application of a 500 mM hyperosmotic sucrose solution is often used to assess the Readily Releasable Pool (RRP) of vesicles ⁵⁵, whereas the ratio of the pools released by 250 mM and 500 mM sucrose concentration is used to probe the apparent energy barrier for fusion ⁵⁶. Upon application of 500 mM sucrose, I192T did not affect the RRP size (denoted RRP₅₀₀) when expressed alone or with WT (Fig. 3A, C). The RRP₅₀₀ measured in I192N neurons, on the other hand, was significantly reduced (Fig. 3H, J). The pool released by 250 mM sucrose (denoted RRP₂₅₀) was reduced in neurons expressing I192T only (Fig. 3B), which implies a reduction in the RRP₂₅₀/RRP₅₀₀ ratio (Fig. 3D) indicative of an increase in the apparent energy barrier for fusion ⁵⁶. A reduction in the RRP₂₅₀/RRP₅₀₀ ratio was also observed in I192N-expressing neurons (Fig. 3I, K); in those neurons, the RRP₅₀₀ itself was also reduced (Fig. 3H). Reductions in RRP₅₀₀ may be attributable to a defect in upstream priming reactions or to a severe increase in the apparent fusion barrier, which makes it impossible to empty the primed vesicle pool with this 500 mM stimulus ⁵⁶. We therefore measured the pool released by 0.75 M sucrose (RRP₇₅₀). However, even at this concentration the pool was reduced (Fig. 3M). It should be noted that regardless of why the RRP₅₀₀ is reduced, the reduction in RRP₂₅₀/RRP₅₀₀ ratio indicates an increase in the energy barrier for fusion. This is because a lower fusion rate leads to an incomplete depletion of the RRP during sucrose application, and this effect remains even if the stronger stimulus (RRP₅₀₀) does not fully deplete the RRP ⁵⁶ (Fig. 4 in ⁵⁶). Given that the fusion barrier is obviously strongly increased with this variant, it remains unclear whether priming is affected. The release probability was found to be reduced in I192T, I192N and I192N+WT neurons (Fig. 3E, L), consistent with the shift towards short-term facilitation. We also performed back-extrapolation of the 40 Hz train responses ⁵⁷ (Fig. S2) because in glutamatergic neurons the RRP for evoked release (RRP_ev) forms a subset of the RRP₅₀₀ ⁵⁸. This supported the change in RRP and the reduced evoked release probability in I192T, I192N+WT neurons and I192N neurons seen in sucrose experiments (Fig. 3F, G, N, O).

Figure 3. SNAP-25 I192T and I192N increase the apparent energy barrier for vesicle fusion.

A, H: Release upon application of 500mM sucrose for I192T (A) or I192N (H), which probes the RRP₅₀₀ size.

B, I: Release upon application of 250mM sucrose (B: I192T; I: I192N). Both mutations reduce RRP₂₅₀ size when expressed alone.

C, J: Traces upon supplication of 250mM (grey traces) and 500mM sucrose.

D, K: The ratio of 250 mm and 500mM sucrose pools. For I192T (D) and I192N (K) expressed alone, the ratio was significantly smaller, reflecting a higher energy barrier.

E, L: The release probability calculated by dividing the charge of a single eEPSC by the RRP₅₀₀.

F, N: Pool size as estimated from the 40Hz trains, denoted the evoked RRP (RRP_ev). The evoked RRP for I192N neurons could not be estimated due to facilitation.

G, O: Release probability estimated from the 40Hz trains.

M: RRP for I192N expressing neurons probed by 750 mM sucrose.

Data are represented as individual points and mean ± SEM. Statistical testing was carried out by ANOVA or Brown-Forsythe ANOVA, as appropriate. Both tests were followed by Dunnett’s multiple comparisons test. See also Table S2 and Fig. S2.

Overall, these measurements show that both variants reduce spontaneous and evoked release and cause a shift towards facilitation, due to an increase in the apparent energy barrier associated with vesicle fusion. However, I192N is significantly more deleterious than I192T and exhibits a clear phenotype in the presence of WT protein and, in addition, I192N also affects the neuronal arborization and survival.

SNAP-25 I192 variants destabilize the SNARE C-Terminal domain and attenuate template complex mediated SNARE assembly

To investigate the molecular consequences of the I192N and I192T variants, we used a previously developed single-molecule assay ^5,7,21,59 where preformed single SNARE-complexes are tethered at the C-termini of syntaxin-1 and VAMP2 to two polystyrene beads held in two optical traps (Fig. 4A). In addition, syntaxin-1 and VAMP2 are crosslinked at their N-termini via a disulfide bond to enable reversible SNARE complex assembly and disassembly. By controlling the distance between the two optical traps, a specific force was applied to the tethered SNARE complex. The end-to-end distance or extension of the protein-DNA tether was measured by monitoring bead displacements from the trap centers.

Figure 4. SNAP-25 I192 variants destabilize the SNARE C-Terminal domain.

A: Experimental setup to pull a single SNARE complex (PDB IDs: 3HD7 for the four-helix bundle and 1BR0 for the Habc domain) using optical tweezers.

B: Force-Extension-Curves (FECs) obtained by pulling (grey) or relaxing (black) single SNARE complexes. Different SNARE folding states are marked according to panel C. The dashed magenta oval marks reversible folding and unfolding transitions of the SNARE four-helix bundle. The dashed green line indicates the extension of different intermediate state.

C: Diagrams of different SNARE folding states. The folding states of the WT SNARE complex include the fully assembled SNARE state (1), the LD-unfolded four-helix bundle state (2), the half unzippered state (3), the unzipped t-SNARE state (4), and the fully unfolded state (5). Folding of both variant SNARE complexes I192T and I192N bypasses the state 2.

D: Extension-time trajectories at the indicated constant forces (F) showing folding and unfolding transitions of SNARE complexes. Red curves represent the idealized transitions derived from hidden Markov modeling. Green dashed lines mark the average extensions of the associated states. Double-Gaussian fits (blue curves) of the extension probability density distributions reveal transitions between the two discrete states indicated by their corresponding state numbers. See also Fig. S3 and S4.

We first pulled a single pre-assembled WT SNARE complex by moving one optical trap away from the other at a speed of 10 nm/s, yielding the force-extension curve (FEC; Fig. 4B, gray curve). At forces below 10 pN, the SNARE complex was fully assembled (state 1, Fig. 4C). As forces increased above 10 pN, discrete extension increases or decreases are seen in the FEC in different force regions, which indicate SNARE unfolding or refolding, respectively. The SNARE complex sequentially unzipped or unfolded in the linker domain (LD), the C-terminal domain (CTD), the N-terminal domain (NTD), and finally the binary syntaxin:VAMP2 t-SNARE complex. The derived intermediates and the pathway of SNARE assembly and disassembly are depicted in Fig. 4C ^7,22,60. Due to its small size (~2.6 nm) and high transition rate, the LD transition is observed in roughly half of the FECs ⁷. Once the SNARE complex was fully disassembled with SNAP-25 dissociation at forces exceeding ~20 pN, we began to relax the remaining syntaxin-VAMP2 conjugate. No SNARE reassembly was observed in the absence of SNAP-25, as expected ²⁷.

The SNAP-25 variants I192T and I192N exhibit reduced CTD stability, evidenced by a decreased force range for the CTD transition compared to the WT SNARE complex (Fig. 4B, FEC regions in dashed ovals). To quantify this, we measured the equilibrium force of the two-state transition, defined as the force where the folded and unfolded state probabilities are equal (Fig. 4D and Fig. S3). The WT SNARE complex shows an equilibrium force of 16.5 ± 0.8 pN for the CTD transition, which decreases to 13.0 (±0.8) for I192T and 10.5 (±1.1) pN for I192N. Consequently, the distinct CTD and LD transitions observed in WT SNARE complexes merge into a single LD/CTD in the mutants, as shown by Double-Gaussian fits of the extension probability density distributions (purple curves, Fig. 4D). The average extension change of the combined LD/CTD domain at the equilibrium force exceeds that of the individual CTD (Fig. 4D, compare the average extensions of the middle and bottom traces to that of the top one), supporting this interpretation. Relative to the equilibrium force of ~12 pN for the LD transition in WT, the LD/CTD equilibrium force is slightly lower for I192N and higher for I192T, suggesting a long-range effect of the CTD mutation on both domains. This altered SNARE assembly pathway due to CTD mutations aligns with prior observations in SNARE complexes containing SNAP-25 I67T, I67N, and VAMP2 L70A (+4 layer) ^59,61, indicating a general phenomenon.

Under our experimental conditions, the NTD unfolds irreversibly, preventing standard equilibrium analysis. To evaluate NTD stability, we compared NTD unfolding forces across WT and mutant complexes. The average NTD unfolding forces were nearly identical: 17.6 ± 0.6 pN for WT, 17.5 ± 1.2 pN for I192T, and 17.6 ± 1.4 pN for I192N (Fig. S4). This indicates that NTD mechanical stability is unaffected by I192 mutations, confirming that the observed effects are specific to the LD/CTD region.

We derived the energy of the LD/CTD transition by measuring the two-state transitions of each SNAP-25 variant at different constant force (Fig. 4D) and analyzing the resultant extension trajectories using hidden Markov modeling (HMM)^59,62. These derivations were then fit to a nonlinear model for force-induced protein folding and unfolding transitions to estimate the unfolding free energies at zero force ⁶². The LD/CTD unfolding energy was 24 ±3 (mean ± SD) k_BT for the I192T variant and 19 ±3 k_BT for the I92N variant, both lower than the combined unfolding energy of CTD and LD of 33 ±3 k_BT for the WT SNARE complex ⁵⁹.

Thus, the two SNAP-25 I192 variants result in destabilization of the C-terminal half of the SNARE-complex and reduced LD/CTD folding energy available for membrane fusion, with I192N being more severe than I192T. This molecular phenotype aligns very well with the lowered release probability, reduced mEPSC frequency and increase in PPR observed in autaptic neurons, since less energy is available to fuse membranes both in the presence and absence of an AP. To understand the consequences of I192 variants on the assembly of the SNARE-ring, we next asked whether the variants change SNARE-complex formation.

The regulated assembly pathway involves SNAP-25 binding to the preformed Munc18–1:VAMP2:Syntaxin-1 trimeric template complex, leading to the formation of the SNARE complex ^7,8,11,22. To assess how the I192 variants affect SNAP-25 association with the template complex, we tethered a single performed WT SNARE complex between two polystyrene beads as before but added 1 μM Munc18–1 and 60 nM free WT or mutant SNAP-25 to the solution (Fig. 5A). After disassembly of the SNARE complex at high force (Fig. 5C, from state 1 to state 5; Table S1), we relaxed the syntaxin-VAMP2 conjugate and detected template complex formation and SNARE assembly at lower force (Fig. 5B, FECs in black). A fast transition occurred around 10–16 pN (Fig. 5B, blue dashed rectangle), indicating Munc18–1 binding to syntaxin-1 (Figs. 5C, state 6). This was followed by another transition around 3–7 pN resulting from formation of the template complex (Figs. 5B, 5C, state 7). To study SNAP-25 binding, we held the template complex around its equilibrium force (between states 6 and 7) and monitored SNAP-25 binding events for 120 s (Fig. 5B, red region; Fig. 5D). SNAP-25 association was manifested by a sudden extension drop from the extension of the template complex state (Fig. 5B,C,D, from state 7 to state 1 indicated by red arrows). Subsequent pulling confirmed the status of SNAP-25 association, either SNAP-25 bound (Fig. 5B, gray curves in FEC#3, 4, and 8) or unbound (grey curves in FEC#2, 6, and 7). The probability of SNAP-25 binding and SNARE assembly was scored from more than 50 such folding experiments (Fig. 5E). The SNAP-25 binding probabilities for I192T and I192N were 0.18 ±0.03 (mean ± SEM) and 0.13 ±0.01, respectively, and were lower than that of WT (0.27 ± 0.04; Fig. 5E).

Figure 5. SNAP-25 I192 variants attenuate Munc18–1-chaperoned SNARE complex assembly.

A: Experimental setup to investigate Munc18–1-chaperoned SNARE assembly.

B: FECs obtained by pulling (grey) and relaxing (black) a single syntaxin-VAMP2 conjugate or holding at a constant trap separation (red) in the presence of Munc18–1 and SNAP-25 variants in the solution for four consecutive rounds. The states associated with different FEC regions are labeled (see C) and the SNAP-25 binding events are indicated by red arrows. Because the preassembled SNARE complex initially contained WT SNAP-25, the Munc18–1-chaperoned SNARE assembly was accompanied by a switch of SNAP-25 molecule from the WT to its variant. Therefore, the second pulling round exhibits lower LD/CTD unzipping force than the first round.

C: Schematic diagrams of different states in Munc18–1-chaperoned SNARE assembly. These states include the fully assembled SNARE complex with Munc18–1 bound to the Habc domain of syntaxin (1), the half-zippered SNARE complex (3), the unzipped SNARE complex (4), the fully unfolded SNARE motifs (5), the Munc18–1 bound open syntaxin (6), and the template complex (7).

D: Extension-time trajectories at the indicated constant forces (F) showing folding and unfolding transitions of the template complex and SNAP-25 binding (indicated by red arrows).

E: Probabilities of WT and mutant SNAP-25 binding to template complex. * p < 0.05, Fisher’s exact test.

See also Table S1 and S2.

Thus, both variants numerically decreased the probability of SNAP-25 association to the template complex, which would be expected to impair Munc18–1-chaperoned SNARE assembly.

The I192N variant is dominant-negative in the presence of WT SNAP-25

The I192N variants affect the properties of the assembled SNARE-complex, while at the same time reducing the probability of associating to the acceptor complex. In a mixed situation (presence of both variant and wildtype protein), as it occurs in affected individuals²⁹, the variant would be less likely to enter the multimeric SNARE-ring than wildtype protein. We therefore asked whether I192 variants would be dominant in the presence of WT SNAP-25.

We isolated neurons from WT littermates of Snap25 KO embryos, created by crossing heterozygotes, and transduced them with increasing volumes of I192T or I192N lentivirus. Electrophysiological measurements were carried out in six different groups according to viral load: 0, 5, 15, 30, 45, and 60 μl lentivirus, respectively. Western blotting in Fig. S5 showed overall higher expression with higher viral loads. However, for I192N the expression was variable between blots, and the endogenous SNAP-25 also seemed to increase expression at higher viral loads. This is probably because I192N negatively affects neuronal survival, even when expressed in WT neurons (see below), which especially affects neurons with high I192N/endogenous ratios, rendering the surviving neurons unrepresentative for the population. For this reason, the titration series of I192T and I192N cannot be directly compared.

I192N expression led to a strong, dose-dependent reduction in mEPSC frequency (down to 8% of WT values) without changing mEPSC amplitudes (Fig. 6A–C). Survival of the neurons on glial islands was impaired as I192N expression increased (Fig. 6D). eEPSCs likewise suffered a dose-dependent reduction, although milder than for mEPSC frequencies (reduced to 28% of WT values, Fig. 6E). Repetitive stimulation revealed a shift toward short-term facilitation with higher virus load (Fig. 6F, N). Sucrose experiments demonstrated a gradual decrease in pool size, especially in RRP₂₅₀ (Fig. 6G, J, K), resulting in a progressive reduction in RRP₂₅₀/RRP₅₀₀ ratio (Fig. 6L). Release probability also decreased with higher I192N expression (Fig. 6H, I, M). Together, these data indicate a strong dominant-negative property of the I192N mutant on synaptic release.

Figure 6. The I192N variant is dominant-negative in the presence of WT SNAP-25.

A, B, E, F, N: I192N decreased mEPSC frequency (A) and the amplitude of eEPSC (E) in a dose-dependent manner without affecting mEPSC amplitude (B). Short-term plasticity shifted towards facilitation (F, N) as I192N expression increased.

C: mEPSC traces. Darker green colors indicate higher I192N expression levels.

D: Reduced neuronal survival as the amount of I192N increases.

G: Sucrose traces. RRP₂₅₀ measurements are shown in grey and RRP₅₀₀ are shown in green colors, with darker green indicating higher I192N expression.

H, I: RRP_ev (H) and release probability (I) from the 40Hz train. Release probability decreased as the amount of I192N increases.

J, K, L, M: RRP₂₅₀ decreased as the expression levels of I192N increased. The reduced RRP₂₅₀/RRP₅₀₀ ratio reflects an increased energy barrier for release (L), which is also reflected in the reduced release probabilities (M) calculated by dividing the eEPSC charge with the 500mM sucrose pool.

Data are represented as individual points and mean ± SEM. The groups were compared using one-way ANOVA, or Brown-Forsythe ANOVA, with Dunnett’s multiple comparisons test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See also Table S2 and Fig. S5 and S6.

In contrast, overexpression of I192T at increasing viral load had a mild effect. While mEPSC frequencies were reduced (Fig. S6A–C), eEPSC amplitudes were only mildly affected (Fig. S6E). There were no significant changes upon high frequency train stimulation (Fig. S6F, H, I) or after stimulation with 250 mM or 500 mM sucrose (Fig. S6G, J–K), except for a significant decrease in RRP₂₅₀/RRP₅₀₀ ratio for the highest viral load (Fig S6L). RRP_ev and RRP_0.5 as well as release probability were unchanged (Fig. S6H, I, K, M) and PPRs were also not significantly different (Fig. S6N). Overall, I192T had only mild effects on synaptic transmission in the presence of WT SNAP-25.

Consequences of the co-existence of mutated and wildtype SNAP-25 within the SNARE-ring

We next considered the titration series of I192N in WT cells together with the expression in KO cells. Normalizing titration data to non-transduced WT data and expression in Snap25 KO to the expression of the WT protein, we plotted the amplitudes of single EPSCs, the frequency of spontaneous (mEPSC) release, and the amplitude of asynchronous release towards the end of trains as a function of viral load (Fig. 7A, points). Strikingly, spontaneous release was most severely affected, while the effect on single EPSCs was milder. Asynchronous release was only mildly affected and persisted in the absence of WT protein. Fig. S6O shows the same plot for I192T. A similar plot of the synchronous release component towards the end of trains showed no effect of the I192N variant (Fig. S1C), indicating that this component remains fully supported by this variant.

Figure 7. Modeling the effect of a mutant species within the SNARE-ring.

A. Data (points) and model fit (lines) of asynchronous (green), evoked (blue) and spontaneous (red) release, under the assumption that these are supported by 1, 3 and 6 SNARE-complexes, respectively, as shown in panel B. Data for Mutant alone were obtained from expression in KO neurons (Fig. 2).

B: A vesicle seen from above, with 6 SNARE-complexes assembled, k of which are WT and n-k of which are mutants, together with the resulting vesicular fusion rate according to eq. (1). See also Fig. S7.

We considered the principal consequences of random incorporation of variant and WT SNAP-25 into the multimeric SNARE-rosette or ring formed between vesicle and plasma membrane during priming ^40–42. We derived a model for this situation, by assuming that $n$ SNARE-complexes are involved in fusion, and that each SNARE-complex contributes additively to the lowering of the energy barrier for fusion, but WT and Mutant SNARE-complexes are not equally effective. With this assumption we derived an equation for the fusion rate of a vesicle with any combination of WT and Mutant SNARE-complexes (STAR Methods, eq.(1); fig. 7B). The total fusion rate from the RRP can then be calculated by considering that vesicles with all combinations of WT and Mutant will co-exist within the RRP according to a binomial distribution (STAR Methods, eq. (2)):

$$k^{total,RRP}=\sum _{k=0}^n\left(\genfrac{}{}{0pt}{}{n}{k}\right){\cdot p}^k\cdot {\left(1-p\right)}^{n-k}\cdot {k^{WT}\left(\frac{k^{Mut}}{k^{WT}}\right)}^{\frac{n-k}{n}}$$

where $p$ is the probability of incorporating a WT protein (and $1-p$ is the probability of incorporating a Mutant), and $k_{WT}$ and $k_{Mut}$ are the fusion rates supported by the WT and Mutant proteins. We next need a way of estimating $p$; $p$ depends on the relative expression levels of WT and Mutant protein at the level of the vesicle, but it may also be affected by the propensities to enter the SNARE-complex, which can vary between WT and Mutant protein (Fig. 5E). Here, we calculated $p$ as:

$$p=\frac{1}{q\cdot Viralload+1}$$

where $Viralload$ is the amount of virus added to WT neurons in μl. For simplicity, we assume that $q$ is a constant, although other possibilities exist (discussed in STAR Methods). Note that our model only modifies the fusion rate of primed vesicles; any defects in priming that would lead to fewer vesicles being present are not included.

Within this model, each mode of release (spontaneous, evoked and asynchronous) could vary in terms of the $n$ (number of SNARE-complexes involved), $k_{Mut}$ (ability of the mutant protein to sustain the release mode), and $q$. To fully determine the model, we would need to determine nine parameters ($n$, k_Mut, and $q$ for the three release modes). With the titration data we have at hand (Fig. 7A), this is not possible. We therefore must constrain certain parameters. Recent structural data describing a six-fold symmetry underneath docked synaptic vesicles led to the suggestion that a ring of 6 SNARE-complexes applies force to the membranes at rest ^63,64. If this is the case, $n$ should be equal to 6 for spontaneous release, which happens without a synchronizing signal and therefore will be affected by all 6 SNARE-complexes (Fig. 7B). Functional and modeling data indicate that full assembly of two-three SNAREs are sufficient to drive fast exocytosis ^65–69, indicating n = 3 for evoked release (Fig. 7B). Finally, it has been suggested that release during sustained high-Ca²⁺ conditions is rate-limited by a single SNARE-complex ⁶⁵, indicating that n = 1 for asynchronous release (Fig. 7B). We further assumed that $q$ is the same for all three release modes. We then fitted the data set with different k_Mut for spontaneous, evoked and sustained release (using normalized data in Fig. 7A, we determined the k_Mut/k_WT ratios). This resulted in the fit shown in Fig. 7A, obtained with q = 0.0198, k_Mut/k_WT (spontaneous) = 1.16 × 10⁻⁷, k_Mut/k_WT (evoked) = 3.68 × 10⁻⁴, k_Mut/k_WT (asynchronous) = 0.311. In this fit, the stronger effect of I192N variant on spontaneous than evoked release is partly due to the higher stoichiometry of spontaneous release (n = 6 vs. 3), partly due to a reduced ability to support spontaneous release, whereas the mutation is relatively mild for asynchronous release. Accordingly, a fit with n = 6 for all release phases led to smaller difference between spontaneous and evoked release (Fig. S7). Likewise, we fitted eq. (2) to titration data from I192T. This fit reproduces the stronger effect of the variant on spontaneous than evoked release (Fig. S6O). As discussed in STAR Methods, different combinations of $n$, k_Mut, and $q$, might also fit the data under different assumptions.

Discussion

We investigated two SNAP-25 disease variants of the isoleucine-192 residue to gain a comprehensive understanding of their molecular defects. Both I192T and I192N resulted in synaptic phenotypes characterized by reduced mEPSC frequency, decreased eEPSC amplitude, lower release probability, and a shift toward facilitation. The effect of these variants, along with other variants and mutations studied within the SNARE layers ^31–33,46, can be summarized as increasing the apparent energy barrier for release ⁵⁶. However, rather than an increase in the energy barrier itself, single-molecule measurements revealed that the variants reduce the energy available to overcome it. In addition, both I192 variants cause the LD and CTD to unfold and refold as a single unit, unlike WT protein, where they unfold and refold separately ⁵⁹. Adding up the folding energies for the CTD and LD revealed a clear reduction in folding energy, which was stronger for the I192N than the I192T variant, which agrees with the more severe clinical and cell physiological phenotype of the former. The variants of I192 also reduce association of SNAP-25 to the Munc18–1:VAMP2:syntaxin-1 template complex, which drives SNARE-complex formation. SNAP-25 binds to the template complex either spontaneously (as in our in vitro experiments) or guided by Munc13-binding to the SNAP-25 linker connecting the two SNARE motifs (in the cell) ²⁵. The I192 residue might assist in template binding via transient interactions, or by impacting conformational flexibility, although full assembly of layer +5 is unlikely to occur until membrane fusion.

Deletion of SNAP-25, syntaxin-1A/B and Munc18–1 all cause neuronal death ^50,70, although the mechanism remains elusive, with recent data implicating disinhibition of a sodium leak channel in the death of syntaxin-depleted neurons ⁷¹. Very small amounts of Munc18–1 protein are sufficient to support survival in human neurons ⁷², whereas in rodent neurons some disease variants led to suboptimal neuronal survival ⁷³. Similarly, another SNAP-25 variant, I67N, did not lead to noticeable changes in survival or neuronal development in human neurons ³³. In contrast, here we report that the clinically very severe SNAP-25 variant, I192N, results in suboptimal rescue of neuronal survival, and even induces neuronal death upon expression in WT mouse neurons, indicating that this is a dominant feature that might constitute a pathogenic mechanism. In contrast, I192T only mildly affects synaptic transmission, and supports full neuronal survival.

Assembly of the C-terminal end of the SNARE-complex is widely assumed to provide the energy required for overcoming the fusion barrier. However, we found that different phases of synaptic release are affected very differently by the severe I192N variant. The variant exerts a strong dominant-negative effect on spontaneous fusion, whereas evoked release was less severely affected. Asynchronous release during a train was only mildly affected (Fig. 7A), and, surprisingly, residual synchronous release during trains was entirely unaffected by the I192N variant (Fig. S1C). We created a general framework for understanding such phenotypes, by considering the general implications of the co-existence of WT and variant SNAP-25 within the SNARE-ring, as is expected in heterozygous individuals. Assuming a binomial choice between WT and variant protein, we derived general formulas for the fusion rate of individual vesicles (eq. 1) and the RRP (eq. 2, STAR Methods). This general model identified three parameters that together account for the effect of a variant on a particular phase of release: the ability of the Mutant to support the fusion mode in question ($k_{Mut}$), the numbers of SNAREs involved ($n$), and the ratio of Mutant to WT protein, corrected by the relative propensities of Mutant and WT to enter the SNARE-complex (Fig. 5A) ($q$).

These parameters cannot all be determined from our data. However, we argue that the recent description of one or two rings of 6 SNARE-complexes that forms during vesicle priming ^63,64, governed by Munc13–1 ^{43,74 75}, indicates that 6 partially assembled SNARE-complexes exert force at rest. This is supported by the recent visualization of 6–7 SNAP-25 molecules undergoing conformational changes in the vesicle-plasma membrane contact zone during priming ⁷⁶. Direct identification of short tethers, assumed to be SNARE-complexes, by cryoelectron tomography (cryo-ET) ^77–79, yielded an average number of tethers in RRP vesicles of 5–6 ⁷⁹. Thus, the stoichiometry for spontaneous release might be 6, since 6 partially assembled SNARE-complexes participate in mildly lowering the energy barrier for release. On the other hand, both functional experiments, direct counting of SNAREs and biophysical modeling using very different approaches suggested 2–4 as the number of SNARE-complexes involved in triggered release ^{65,66,68,69,80–82}. We therefore assumed that n = 3 for evoked (mainly synchronous) release. Less is known about SNAREs assembling during conditions of sustained high-frequency stimulation. We followed previous suggestions ⁶⁵ that under these circumstances, formation of the first SNARE-complex becomes rate-limiting, yielding in effect n = 1 for asynchronous release. The resulting fit indicates $k_{Mut}/k_{WT}$ (asynchronous release) >> $k_{Mut}/k_{WT}$ (evoked release) > $k_{Mut}/k_{WT}$ (spontaneous release) (Fig. 7A). Moreover, comparison to a fit where the $n$ is assumed equal (n = 6) for all release phases shows that having a higher $n$ for spontaneous than for evoked release helps to reproduce the milder effect of the I192N variant on evoked release (comp. Fig. 7A and Fig. S7). Previous modelling work using different approaches showed that sequential assembly of SNAREs within a rosette of 6 complexes typically leads to fusion upon full assembly of ∼3 SNARE-complexes ^68,69,81 because at that point the fusion rate is so high that fusion ensues before assembly of the next SNARE-complex.

Our model assumes that SNARE-complexes influence each other solely by acting on the same energy barrier, with additive energy contributions translating into multiplicative effects on fusion rates via Arrhenius’ equation ⁵⁶. As such, the model represents the simplest conceivable framework for SNARE complex cooperation and does not assume any direct lateral interactions between the ring of 6 SNARE-complexes. Future models may need to incorporate such interactions to account for mutations at the interface between different SNARE complexes ^42,74,83,84.

The difference in ability of the variants to support the different release phases might – in addition to the differences in SNARE stoichiometry – be affected by the different local calcium concentrations and calcium sensors that are involved, most prominently synaptotagmin-1, which is responsible for synchronous release and synaptotagmin-7 and Doc2A/B, which are responsible for asynchronous and/or spontaneous release ^85–87, probably in a combination ⁸⁸. In turn, these Ca²⁺ sensing proteins might affect, or be responsible for, the different stoichiometries, by oligomerization and lateral interaction with the SNAREs ^89,90.

Even the very severe I192N variant supports substantial amounts of asynchronous release during trains and has no deleterious effect on residual synchronous release during trains. How can we account for this, given the importance of the C-terminal end of the SNAREs for membrane-membrane fusion? During an action potential train, the RRP becomes depleted, while Ca²⁺ accumulates, activating synaptotagmins and Munc13s. Under these conditions, the rate-limiting event for asynchronous fusion might be the formation of a single SNARE-complex, which is governed by N-terminal SNARE-interactions. With all inhibitory restraints lifted, this SNARE-complex will either progress to full zippering and fuse the vesicle single-handedly ^67,91, or rapidly recruit additional SNARE-complexes, making the formation of the initial SNARE-complex the rate-limiting step. In this situation, the I192N variant might have a mild effect on release, partly because the WT protein has a higher probability to associate with the acceptor complex (Fig. 5E) in combination with the low stoichiometry, and partly because C-terminal SNARE-interactions are no longer rate limiting. This allows synaptic transmission to ‘escape’ a deleterious variant at higher AP frequencies.

Overall, we conclude that a comprehensive understanding of the molecular and cellular consequences of SNAREopathy variants requires examination of their effects at the level of individual fusion complexes, while also considering the consequences upon incorporation into the SNARE-ring, and the distinct stoichiometries of SNARE-complexes that are rate-limiting for different modes of release.

Limitations of the study

The present investigation has certain limitations. We studied two variants using lentiviral expression of a EGFP-SNAP-25 fusion construct, because this construct has been demonstrated to rescue survival of SNAP-25 KO neurons ⁵⁰, and mutations have been shown to result in decreases or increases in spontaneous and evoked release similar to studies performed with non-fused SNAP-25 ^31,32,45,92. We cannot rule out that the EGFP-tag could affect the results obtained, for instance in titration experiments, where the EGFP-tagged construct competes with the non-tagged endogenous SNAP-25, and therefore might be at a relative disadvantage. We also note that survival of neurons in titration experiments (Fig. 6D, Fig. S6D) might have been affected by the presence of the lentivirus per se, although the lentiviral constructs successfully rescue survival in SNAP-25 KO neurons (Fig. 1B, 1E).

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Jakob B. Sørensen (jakobbs@sund.ku.dk).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement. There are restrictions on the availability of mouse lines because of a material transfer agreement.

Data and code availability

• All data reported in this paper is present in supplementary information or will be shared by the lead contact upon request.

• Custom code for Igor Pro is deposited at DOI 10.5281/zenodo.17776533.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANTS

Plasmids, protein expression, purification and SNARE preparation for molecular tweezers experiments

The amino acid sequences corresponding to WT SNARE and Munc18–1 and their purification were described elsewhere in detail⁷. Briefly, the genes containing the cytoplasmic domains of rat syntaxin-1A (a.a. 1–265 with mutations C145S and R198C), VAMP2 (a.a. 1–96 with N29C), and rat Munc18–1 were cloned into the pET-SUMO vector encoding 6xHis-tag followed by a SUMO tag at the N termini. The full-length cysteine-freer SNAP-25B (C85S, C88S, C90S, C92S) was cloned into pET-15b vector encoding 6xHis-tag at the N terminus. These proteins were expressed in BL21 E. coli cells and purified using Ni-NTA-agarose beads. Proteins were eluted with 300 mM imidazole and exchanged to buffer containing 25 mM HEPES (pH 7.4), 140 mM KCl, and 2 mM tris(2-carboxyethyl) phosphine (TCEP). Syntaxin-1 was then biotinylated at their C-terminal Avi-tags with the biotin ligase BirA as previously described. Ternary SNARE complexes were prepared and cross linked with DNA handles as was previously described⁷. Briefly, syntaxin-1, SNAP-25, and VAMP2 were mixed at a molar ratio of 0.8:1:1.2, incubated at 4 °C, and purified using the 6xHis tag on SNAP-25 and Ni-NTA-agarose. The eluted SNARE complexes were cross-linked with DTDP (2,2’-dithiodipyridine disulfide) treated DNA handles with a molar ratio of 50:1 in 100 mM phosphate buffer, 500 mM NaCl, pH 8.5.

Animals and cell lines

Snap25 knockout (KO) and wildtype (WT) embryos were generated by crossing C57BL/6J Snap25 +/− mice ⁴⁸. Embryos (WT or KO) of either sex were obtained from pregnant females by cesarean section on day E18. The mothers were sacrificed by cervical dislocation, embryos of either sex were obtained by cesarean section and euthanized by decapitation. KO embryos were identified by their small size and tucked position with rounded necks ⁴⁸. WT or KO genotype was subsequently verified by PCR in all cases. The sex of the embryos was not recorded, and the current investigation therefore does not report sex differences. Cortices from outbred CD1 mouse pups (P0-P2) were used for preparation of glial cells upon which the KO or WT hippocampal neurons were eventually plated. Hippocampal CD1 neurons were used for high-density cultures for western blot. Animals were kept and bred according to a permission obtained from the Danish Animal Experiments Inspectorate (2018–15-0202–00157; 2023–15-0202–00197) and followed institutional guidelines as overseen by the Institutional Animal Care and Use Committee (IACUC). The HEK293FT cells used for virus production were grown in DMEM with GlutaMAX I supplemented with 10% FBS, 0.2% penicillin/streptomycin (pen/strep; 20000 IU/20mg) and 0.5 mg/ml G418. They were passaged once a week and only used for virus production when they were between p11 to p25. HEK293FT cells were not authenticated; negative Mycoplasma status was confirmed using a commercial kit (Venor GeM OneStep, Minerva biolabs, Art.Nr. 11–8025).

Constructs and viruses for expression in neurons

SNAP-25 isoform b was expressed as an N-terminal EGFP fusion protein from the pLenti vector driven by the CMV promoter ⁵⁰. DNA mutations resulting in the I192T and I192N substitutions were introduced using the QuickChange II XL kit (Agilent Technologies) and verified by sequencing. Infectious replication-deficient lentiviruses were packaged in HEK293FT cells¹⁷. Preliminary titration of viral preparations using transfection in HEK293-cells and the fluorescent EGFP-tag as a marker revealed similar titers of the three constructs (WT SNAP-25: 1.50 ×10⁷ transducing units/ml; I192T: 1.69 ×10⁷ transducing units/ml; I192N: 1.47×10⁷ transducing units/ml). Therefore, we transduced with identical volumes of the three constructs. Neurons were transduced with lentivirus on DIV0 or DIV1.

METHOD DETAILS

Dual-trap optical tweezers

The optical tweezers were home-built as described elsewhere^5,7,59. Briefly, a 1064 nm laser beam is expanded, collimated, and split into two orthogonally polarized beams, one of which is reflected by a mirror attached to a nano-positioning stage (Mad-city Labs, WI). The two beams are then combined, expanded again, and are then focused by a water-immersed 60x objective with a 1.2 numerical aperture (Olympus, PA) to form two optical traps in the sample plane in the central channel of a home built microfluidic flow chamber. One of the two traps is stationary; the other trap can be moved using the nano-positioning stage. The outgoing laser beams are collimated by a second water-immersed objective, split again by polarization, and projected onto two position-sensitive detectors (Pacific Silicone Sensor, CA). Displacements of the trapped beads are detected by back-focal plane interferometry. Optical tweezers are remotely operated through a computer interface written in LabVIEW (National Instruments, TX). The force constants of two optical traps are determined by the Brownian motion of the trapped beads before each experiment.

Single-molecule experiments and data analysis

An aliquot of the crosslinked protein-DNA sample was incubated with anti-digoxigenin coated polystyrene beads 2.17 μm in diameter (Spherotech, IL), diluted in buffer containing 25 mM HEPES (pH 7.4), 140 mM KCl and injected into the top channel of a microfluidic chamber. Streptavidin-coated polystyrene beads of 1.76 μm were injected into the bottom channel. Both top and bottom channels were connected to a central channel by capillary tubes, where both kinds of beads were trapped. A single SNARE complex was tethered between two beads by bringing them close. Data were recorded at 20 kHz, mean-filtered to 10 kHz. The single-molecule experiment was conducted in in buffer containing 25 mM HEPES (pH 7.4), 140 mM KCl at 23 (±1) °C. An oxygen scavenging system was added to prevent potential protein photo-damage by optical traps. The single protein- DNA tether was pulled or relaxed by increasing or decreasing trap separation at a speed of 10 nm/s. The methods of data analysis were described in detail elsewhere⁶². Briefly, by analyzing the extension trajectories using a two-state hidden-Markov modeling (HMM), the probability, extension, force, lifetime, and transition rates for each state were obtained. To relate the experimental measurements to the conformations and energy (or the energy landscape) of different SNARE states at zero force, we constructed structural models for 1 these states based on crystal structures of the SNARE four-helix bundle and the template complex. These states were characterized by the contour lengths of the unfolded polypeptides and free energy, which were chosen as fitting parameters. The extension and energy of the whole tethered dumbbell, including the DNA handle, were calculated using the Marko-Siggia formula. Then, we computed the probability of each state based on the Boltzmann distribution and transition rates based on the Kramers’ equation. Finally, we fit the calculated state extensions, forces, probabilities, and transition rates to the corresponding experimental measurements using the nonlinear least-squares fitting, which revealed the conformations and energies of different SNARE folding states as best-fit parameters.

Preparation of neuronal cultures

25mm coverslips were used to plate neurons for immunostainings and 30mm coverslips were used to plate neurons for electrophysiology and western blot. Coverslips were washed in 1M HCl overnight followed by washing in 1M NaOH for 1hr and then in plenty MilliQ water before being stored in 96% ethanol. For immunostainings and electrophysiology, coverslips were allowed to air-dry before being coated in 0.15% agarose. After UV-sterilization, the coverslips were stamped with a mixture of 1 part PDL (0.5mg/ml), 1 part rat tail collagen (1 mg/ml) and 3 parts glacial acetic acid (17mM) using a custom-made rubber stamp to provide substrate for the glial cells to make glial micro-islands. For high-density cultures, the entire coverslip was coated in 0.5 mg/ml PDL overnight and subsequently thoroughly washed with MilliQ water prior to plating neurons.

Glial cells were prepared from P0-P2 outbred CD1 pups. The pups were decapitated and the cortices dissected from the brain in HEPES-buffered HBSS. After removal of the meningeal tissue, the cortices were chopped and trypsinized in 0.25% trypsin + EDTA in pure DMEM for 15 min at 37°C. The trypsin was inactivated by addition of glial culture medium (DMEM with 10% FBS, 1% MEM non-essential amino acids and 0.2% pen/strep) containing 25mg bovine serum albumin and 25mg chicken egg trypsin inhibitor. The tissue was gently triturated using a P1000 pipette until a homogenous cell suspension was obtained. The glial cells were plated in T75 flasks and grown in glial culture medium. On DIV 14, the glial cultures were gently washed in pre-warmed HBSS-HEPES before being trypsinized in 0.25% trypsin + EDTA for 10 min at 37°C. Glial culture medium was added to terminate the trypsinization. The cells were counted and plated on the pre-stamped coverslips at a density of 35K/well or 40K/well in a 6-well plate. Hippocampal neurons were plated on the glial micro-islands after 2–5 days.

For immunostaining and electrophysiology, hippocampal glutamatergic Snap25 KO or WT littermates were grown as self-innervating (autaptic) cultures plated on glial micro-islands ⁴⁹. For western blot, CD1 hippocampal glutamatergic neurons were grown as high-density cultures. Embryos or pups were euthanized by decapitation, the meningeal tissue removed, and the hippocampi dissected in HEPES-buffered HBSS. The hippocampi were then trypsinized in 0.25% trypsin + EDTA for 20 min at 37°C. The tissue was washed and triturated until a smooth cell suspension was obtained. For electrophysiology and immunostaining, the neurons were plated on the glial micro-islands at a density of 7K/well or 7.5K/well whereas for western blot the CD1 neurons were plated at 600K/well in a 6-well plate.

Western blot

On DIV14, cultures for western blot were harvested in RIPA buffer supplemented with proteinase inhibitor cocktail (Sigma). The protein concentration in the resulting supernatant was quantified in a BCA assay (ThermoFisher Scientific). Equal amounts of protein from each sample were incubated for 5 min at 95°C in 4x Bolt^™ SDS Sample buffer containing Bolt^™ reducing agent (ThermoFisher Scientific) and subsequently run on a Novex 4–12% Bis-Tris Plus gel (ThermoFisher Scientific). The proteins were wet-transferred to an Amersham Hybond LFP 2.0 PVDF membrane (GE Healthcare) after which the membrane was blocked in blocking solution (0.1% TBS-Tween 20 with 5% skimmed milk powder). The membrane was incubated overnight at 4°C with primary antibody in blocking solution (mouse α-SNAP-25 1:10000 Synaptic Systems 111011; rabbit α-VCP antibody 1:5000 Abcam AB109240; mouse α-VCP 1:2000 Abcam AB11433). After primary antibody, the membrane was washed in 0.1% TBST and then incubated with HRP-conjugated secondary antibody in blocking solution (goat α-mouse-HRP 1:10000 for SNAP-25, 1:2000 for VCP; goat α-rabbit-HRP 1:5000 for VCP). The membrane was washed again and chemiluminescence visualized on a FluorChem E machine (ProteinSimple) after incubation with Pierce ECL HRP substrate (ThermoFisher Scientific). Quantification of Western blots was performed in ImageJ 1.54p using the integrated density of bands, while subtracting background intensity. To quantify the expression of EGFP-SNAP-25 (WT or variant) relative to endogenous SNAP-25, we divided the integrated density of the EGFP-SNAP-25 band with the integrated density of the endogenous SNAP-25 band from the same lane. A minor band was observed at slightly lower molecular size for EGFP-SNAP-25 (Fig. S1, S5); this band was included in the quantification of EGFP-SNAP-25. For quantification of endogenous SNAP-25 (Fig. S5), we normalized the integrated density of endogenous SNAP-25 to the intensity of the VCP-band, which was used as a loading control. This value was divided by the similarly normalized value of endogenous SNAP-25 in the lane without addition of lentivirus (marked “0 μl”).

Immunostaining and neuronal survival

On DIV14, autaptic cultures were fixed in 4% paraformaldehyde (PFA), washed and permeabilized in 0.5% Triton-X 100. The coverslips were then transferred to blocking solution (0.1% Triton-X 100 with 4% goat serum) and subsequently incubated face-down in primary antibody solution (chicken α-MAP2 1:1000; guinea pig α-vGLUT1 1:1000). After washing, they were incubated in secondary antibody solution in a similar manner in the dark. The coverslips were washed and mounted on glass slides using FluorSave^™ reagent and allowed to dry overnight in the dark before the edges were sealed with clear nail polish. The neurons were imaged at the Core Facility for Integrated Microscopy at the Department of Biosciences at Panum, UCPH. Specifically, they were imaged on a Zeiss CellObserver spinning disc confocal microscope (Zeiss Instruments) connected to an ORCA fusion detector (Hamamatsu Photonics) using a 40x objective with water-based immersive (NA=1.2). Image processing included tile stitching and maximum intensity projections in the Z-plane using Zen Blue software (Zeiss Instruments). The total number of synapses as well as their average size, the number of synapses per μm and the total length of neurites were quantified based on the immunostainings using the SynD (version 2022) plugin for MATLAB ⁹³. The MAP2 staining was used to define the soma as well as the neurites. Synapses were identified using the vGLUT1 staining. Only synapses within 1-μm of the dendritic mask were included in the analysis. XY resolution was 0.163μm per pixel, as dictated by the confocal images.

The number of neurons on each glial micro-island was counted between DIV11-DIV15 using a Leica DMIL LED using phase contrast. Only neurons looking shiny and alive were counted. On each coverslip, 2×6 islands were counted on two or three random locations.

Electrophysiology and analysis

Autaptic cultures were used for patch clamp electrophysiology from DIV10-DIV14, and only glial micro-islands harboring one neuron were selected for experiments. All recordings were performed at RT in extracellular solution containing 140mM NaCl, 2.4mM KCl, 10mM HEPES, 14mM glucose, 2mM CaCl₂ and 2 mM MgCl₂. Solution pH was adjusted to 7.4 using NaOH and osmolality measured to be around 300mOsm. The patch pipette contained 15.3 mM KCl, 2mM HEPES, 1 mM EGTA, 15mM creatine phosphate, 2mM Na-ATP, 50 U creatine phosphokinase and 4.6 mM MgCl₂ with pH adjusted to 7.4 using KOH and an osmolality measured to be around 295 mOsm. Cells were visualized on an Axiovert 200 inverted microscope (Zeiss Instruments) and patched in whole-cell configuration using an EPC10 amplifier (HEKA electronics). Data acquisition was performed at 20kHz and lowpass filtered with a 2.9kHz Bessel filter using the Patchmaster software version 2×73.5 (HEKA electronics). Patch pipettes were freshly pulled on a P-1000 puller (Sutter Instruments) from borosilicate capillaries with filament (GB150F-8P 150×0.86, Science Products) and only pipettes with resistance between 3 and 5MΩ were used. Series resistances were always compensated to 70%. Cells were clamped at a holding potential of −70mV and spontaneous activity recorded for 60 seconds at this potential. Single EPSCs were evoked by a 2ms depolarization of the cell body to +0mV. High frequency train stimulations were performed by sequential depolarizations to +0mV for 2ms at 40Hz. Paired pulse responses were measured by depolarizing the cell to +30mV for 1ms each with an interpulse interval of 50ms or 200ms. Paired pulses with 25ms between were calculated from the two first responses in the 40Hz train response. Sucrose was dissolved in extracellular solution and applied in 250mM and 500mM or 375mM and 750mM concentrations to estimate the size of the RRP as well as vesicular release probability. At least one minute recovery time was allowed between protocols. For sucrose, the recovery time was 2 min. All extracellular solutions were applied using a custom-made barrel system coupled to an SF-77B perfusion fast step device (Warner Instruments), which was operated directly from Patchmaster.

All electrophysiological data were analyzed using custom-written scripts in Igor Pro version 6.21, except the sucrose data, which were analyzed using Igor Pro version 6.37 (Wavemetrics) and spontaneous activity, which was analyzed in MiniAnalysis version 6.0.7 (Synaptosoft). The RRP and vesicular release probability were estimated from the 40Hz train responses by integrating the charge contained in each train response, summating them and plotting them as a function of the number of stimulations. The RRP is estimated as the y-intercept when extrapolating from the linear part of the plot and the gradient of this line is the forward priming rate. The vesicular release probability is calculated as the ratio between the charge contained in the first EPSC and the charge contained in the RRP (see Fig. S2).

Combinatorics of WT and mutant complexes and consequences for synaptic transmission

Assume that we have two different SNAP-25 molecules, WT and Mutant, which assemble into SNARE-complexes randomly. If each vesicle forms $n$ SNARE-complexes, we want to arrive at a reasonable equation describing the fusion rate of a vesicle having k WT-complexes and n-k Mutant complexes. Afterwards, we want to derive the fusion rate of the entire fusing pool of vesicles (i.e. the RRP), which will consist of a mix of vesicles with different numbers of WT and Mutant complexes.

Let us assume that we have access to the fusion rate, $k^{WT}$, of a vesicle which has formed only WT complexes, as well as the fusion rate, $k^{Mut}$, of a vesicle which has formed only Mutant complexes. These fusion rates can be determined by expressing either WT or Mutant protein in the knockout background. The fusion rates are related to the respective energy barriers for spontaneous fusion via Arrhenius’ equation ⁵⁶. We will assume that each SNARE-complex and the associated proteins help lowering an energy barrier E₀ for fusion. Then it is lowered by the quantity E_WT in case it contains WT protein, and E_WT + ∆E_Mut in case it contains Mut protein, and if we further assume that all SNARE-complexes act additively on the energy barrier, then

$$k^{WT}=A\cdot exp\left(-\frac{{E_0+n\cdot E}_{WT}}{k_{B}T}\right)$$

and

$$k^{Mut}=A\cdot exp\left(-\frac{E_0+n\cdot \left(E_{WT}+{∆E}_{Mut}\right)}{k_{B}T}\right)$$

where ${\mathrm{E}}_{\mathrm{W}\mathrm{T}}<0$, such that the assembly of a WT complex lowers the energy barrier for release and increases the fusion rate. For a loss-of-function mutation (such as those studied here), ${∆\mathrm{E}}_{\mathrm{M}\mathrm{u}\mathrm{t}}>0$, whereas for a (hypothetic) gain-of-function mutation, ${∆\mathrm{E}}_{\mathrm{M}\mathrm{u}\mathrm{t}}<0$.

We will assume that the pre-exponential factor, A, is the same for both situations, which is reasonable for a preassembled complex, where transitions do not depend on collision rates. If a vesicle subpopulation has formed $k$ WT-complexes and $n-k$ Mut complexes, then the fusion rate and associated energy barrier are related by:

$$k^{n,k}=A\cdot exp\left(-\frac{E_0+k\cdot E_{WT}+\left(n-k\right)\left(E_{WT}+{∆E}_{Mut}\right)}{k_{B}T}\right)$$

$$=A\cdot exp\left(-\frac{E_0+n\cdot E_{WT}+\left(n-k\right){\cdot ∆E}_{Mut}}{k_{B}T}\right)$$

$$=A\cdot exp\left(-\frac{E_0+n\cdot E_{WT}}{k_{B}T}\right)\cdot exp\left(-\frac{\left(n-k\right){\cdot ∆E}_{Mut}}{k_{B}T}\right)$$

where the first factor is the fusion rate for all-WT complexes, $k^{WT}$, whereas the second factor can be calculated from the ratio of $k^{Mut}$ and $k^{WT}$:

$$\frac{k^{Mut}}{k^{WT}}=exp\left(-\frac{n\cdot {∆E}_{Mut}}{k_{B}T}\right)$$

which means that

$${\left(\frac{k^{Mut}}{k^{WT}}\right)}^{\frac{n-k}{n}}=exp\left(-\frac{\left(n-k\right)\cdot {∆E}_{Mut}}{k_{B}T}\right)$$

Substituting in the above:

$$k^{n,k}={k^{WT}\left(\frac{k^{Mut}}{k^{WT}}\right)}^{\frac{n-k}{n}}$$ (eq. 1)

which is the equation describing the fusion rate of a vesicle with $k$ WT and $n-k$ Mutant complexes. If $k=n$, this equation yields the fusion rate of the WT situation, whereas if $k=0$ it yields the fusion rate of the Mutant situation. Note that this equation could be derived separately for spontaneous, evoked and sustained/asynchronous release, yielding three different versions of the equation.

Moving to the population level, each vesicle is going to select the WT form with probability $p$ (and the Mut form with probability $1-p$), so that the probability of having $k$ WT-complexes formed for a given vesicle is given by the binomial distribution:

$$f\left(k,n,p\right)=\left(\genfrac{}{}{0pt}{}{n}{k}\right)p^k{\left(1-p\right)}^{n-k}$$

The RRP will consist of subpopulations with separate fusion rates, such that the fraction $f\left(k,n,p\right)$ of the RRP has the fusion rate $k^{n,k}$. The total normalized fusion rate will be

$$\begin{gathered} k^{total,RRP}=\sum _{k=0}^{n}f\left(k,n,p\right)\cdot k^{n,k} \\ =\sum _{\begin{array}{c}k=0\\ \end{array}}^n\left(\genfrac{}{}{0pt}{}{n}{k}\right){\cdot p}^k\cdot {\left(1-p\right)}^{n-k}\cdot {k^{WT}\left(\frac{k^{Mut}}{k^{WT}}\right)}^{\frac{n-k}{n}} \end{gathered}$$ (eq. 2)

The equation (2) is valid for spontaneous release, but evoked release is caused by a synchronizing signal, i.e. Ca²⁺, which causes SNARE-complexes to fuse the membranes on a millisecond timescale. Different modelling approaches of this process have shown that even if six (or more) fusion complexes are preassembled when the action potential arrives, only ~3 complexes will have time to finalize assembly and trigger release ^69,81. The three SNARE-complexes that zip to fuse the vesicle will be sampled among the six that have preassembled. Therefore, evoked release implies two selection processes: selection (between WT and Mutant) of the initial six complexes during vesicle priming, and selection of the three out of the six that will fuse the vesicle. However, despite this double selection process, evoked release rate from the RRP (e.g. when summing over all vesicles in the RRP) follow equation 2 as well, which is shown next:

For a given vesicle that fuses with the help of $j$ complexes, of which $i$ are WT, ($i\le j)$, the fusion rate will be - in analogy with eq. (1):

$$k_{ev}^{j,i}={k_{ev}^{WT}\left(\frac{k_{ev}^{Mut}}{k_{ev}^{WT}}\right)}^{\frac{j-i}{j}}$$

where the subscript $ev$ indicates evoked rates.

If we consider a vesicle, which has formed $n$ SNARE-complexes, of which $k$ are WT, the vesicle must then select $\mathrm{j}$ complexes ($j\le n)$ for fusion, of which $i$ are WT and $j-i$ are mutant. The $i$ WT complexes must be chosen from the $k$ WT complexes, whereas the $j-i$ mutants must be chosen from the $n-k$ mutant complexes. In total, $j$ complexes must be chosen from $n$ complexes. Therefore, the probability of choosing $i$ WT complexes in this situation is given by a hypergeometric distribution:

$$P\left(n,k,j,i\right)=\left[\frac{\left(\genfrac{}{}{0pt}{}{k}{i}\right)\left(\genfrac{}{}{0pt}{}{n-k}{j-i}\right)}{\left(\genfrac{}{}{0pt}{}{n}{j}\right)}\right]$$

The probability that we will have $i$ WT complexes selected in the end is:

$$\begin{gathered} P(n,p,j,i)={\sum }_{k=0}^n\left(f(k,n,p)\cdot \left[\frac{\left(\genfrac{}{}{0pt}{}{k}{i}\right)\left(\genfrac{}{}{0pt}{}{n-k}{j-i}\right)}{(\genfrac{}{}{0pt}{}{n}{j})}\right]\right) \\ ={\sum }_{k=0}^n\left(\left(\genfrac{}{}{0pt}{}{n}{k}\right)p^k(1-p{)}^{n-k}\left[\frac{\left(\genfrac{}{}{0pt}{}{k}{i}\right)\left(\genfrac{}{}{0pt}{}{n-k}{j-i}\right)}{(\genfrac{}{}{0pt}{}{n}{j})}\right]\right) \\ ={\sum }_{k=0}^n\left(\genfrac{}{}{0pt}{}{j}{i}\right)\left(\genfrac{}{}{0pt}{}{n-j}{k-i}\right)p^k(1-p{)}^{n-k} \\ ={\sum }_{k=0}^n\left(\genfrac{}{}{0pt}{}{j}{i}\right)p^i(1-p{)}^{j-i}\left(\genfrac{}{}{0pt}{}{n-j}{k-i}\right)p^{k-i}(1-p{)}^{n-k-j+i} \\ =\left(\genfrac{}{}{0pt}{}{j}{i}\right)p^i(1-p{)}^{j-i}\left({\sum }_{k=0}^n\left(\genfrac{}{}{0pt}{}{n-j}{k-i}\right)p^{k-i}(1-p{)}^{n-k-j+i}\right) \\ =\left(\genfrac{}{}{0pt}{}{j}{i}\right)p^i(1-p{)}^{j-i}\left({\sum }_{m=0}^r\left(\genfrac{}{}{0pt}{}{r}{m}\right)p^m(1-p{)}^{r-m}\right) \\ =\left(\genfrac{}{}{0pt}{}{j}{i}\right)p^i(1-p{)}^{j-i}=f(i,j,p) \end{gathered}$$

The probability is therefore given by a binomial distribution with parameters $j$ and $p$. The parameters $n$ and $k$, and thereby the original choice, do not matter when we summarize over vesicles with all values of $k$.

The total normalized evoked rate is therefore given by an expression similar to eq. (2):

$$\begin{gathered} k_{ev}^{total,RRP}=\sum _{i=0}^{j}f\left(i,j,p\right)\cdot k_{ev}^{j,i} \\ =\sum _{i=0}^{j}f\left(i,j,p\right)\cdot {k_{ev}^{WT}\left(\frac{k_{ev}^{Mut}}{k_{ev}^{WT}}\right)}^{\frac{j-i}{j}} \end{gathered}$$

Titration with mutant protein.

To get from the expression level of WT and Mutant protein to the probability $p$ of incorporating WT protein, it is natural to assume that

$$p=\frac{\left[WT\right]}{\left[Mut\right]\cdot q+\left[WT\right]}$$

where $q$ is the preference for Mutant over WT protein assembly into the supercomplex. We can rearrange this into:

$$p=\frac{1}{q\cdot \frac{\left[Mut\right]}{\left[WT\right]}+1}$$ (eq. 3a)

This equation requires us to determine $\left[Mut\right]/\left[WT\right]$-ratios, which in principle could be done from western blots such as those in Fig. S5. However, this is complicated by the fact that one of the variants (I192N) causes neuronal death when expressed. We therefore cannot be sure that the western blots correctly represent the $\left[Mut\right]/\left[WT\right]$-ratios, since some neurons will have died and the surviving neurons might not be representable for the population. This could be why it appears that endogenous SNAP-25 becomes more prevalent as the viral load increases; this can be explained if neurons that died had a higher $\left[\mathrm{M}\mathrm{u}\mathrm{t}\right]/\left[\mathrm{W}\mathrm{T}\right]$-ratio than neurons that survived, combined with the fact that similar protein amounts were loaded into each lane. In contrast, for I192T, higher viral titers increased the amount of expressed EGFP-SNAP-25, while not affecting endogenous SNAP-25, as one would expect. A further complication is the fact that the western blots were carried out on mass cultures, whereas electrophysiology was carried out on autaptic neurons, where neuronal densities are much lower, which might affect survival. Because of these complications, we instead used the following equation to determine $p$:

$$p=\frac{1}{q\cdot Viralload+1}$$ (eq. 3b)

Here, $p$ is 1 for zero Viral load, and 1 when Viral load tends to infinity. q could in principle depend on the viral load (for instance if increased Viral load leads to a saturation of the amount of expressed EGFP-SNAP-25), but in the absence of reliable information on this, we consider $q$ a constant. By using (eq. 3b) instead of (eq. 3a), $q$ will combine information on the expression level induced by the virus, and the propensity of choosing Mutant and WT protein for assembly into the supercomplex.

Fitting of the model to titration data

When fitting this model to normalized data (Fig. 6 and Fig. S6O), we inserted eq (3b) into eq. (2), and assumed that k_WT = 1. Data from expression in Snap25 KO background were normalized to the expression of WT EGFP-SNAP-25 in KO neurons and included into the fit by assigning to it a very high value of viral load (6000, i.e. 100x the maximal experimental value), such that for this data set, essentially p = 0 (only the Mutant can be incorporated). To fit the full model to titration data for spontaneous, evoked and asynchronous release we would need to determine nine parameters ($n$, k_Mut, and $q$ for spontaneous, evoked and asynchronous release). With the titration data at hand (Fig. 7B), this is not possible. Instead, as described in the text, we have assumed – based on the literature – that $n$ for spontaneous, evoked and sustained release are 6, 3, and 1, respectively, and then fitted the data while assuming the same $q$ for all phases, but separate $k_{Mut}$’s (Fig. 7C, Fig. S6O). Fitting took place using the Global Fit routine in Igor 8.04 (Wavemetrics) and weighting was performed using the inverse size of the error bars. It should be noted that if increased viral loads result in a saturating amount of EGFP-SNAP-25, instead of a linear increase in amount, then the points at higher viral load would effectively move to the left (Fig. 7C), which might lead to a better fit of the model with the present parameters, but would also change the fitted parameters. Also note that the model can be fitted to data using other assumptions, for instance the same $n$ (number of SNARE-complexes) for all phases. We present an example in Fig. S7, where the data are fitted with n = 6, one shared value of $q$, and separate values of $k_{Mut}$. Because different assumptions lead to viable fits of the data, the model cannot be used to determine any specific SNARE-stoichiometry without making further assumptions.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analyses were performed in Prism version 10.1.1 (GraphPad) and significance was tested using α=0.05. A one-way ANOVA test was used to identify statistically significant differences between groups (>2) where data points were homoscedastic, which was assessed using the Brown-Forsythe test. Dunnett’s post-test was used to compare groups to WT. If data points were heteroscedastic, the Brown-Forsythe ANOVA test, which does not assume equal variances, was used instead. P-values are categorized in different significance levels according to the following: ****p<0.0001, ***p<0.005, **p<0.01, *p<0.05. No asterisks indicate no significant difference.

Supplementary Material

Supplementary figures

Table S1 Structural assignments of the seven defined states during SNARE assembly and disassembly, Related to Fig. 5C.

Table S2

Table S2 Quantitative data. A) Data for expression of variants and WT SNAP-25 in Snap25 KO neurons, related to Fig. 1, Fig. 2, Fig. 3, Fig. S1. B) Data for expression of variants and WT SNAP-25 in Snap25 WT neurons, related to Fig. 6, Fig. S6. C) Binding data, related to Fig. 5E, D) NTD folding forces, related to Fig. S4.

Document containing Figure S1–S7 and Table S1. Separate excel file with Table S2.