The nature of the Syntaxin4 C-terminus affects Munc18c-supported SNARE assembly

Vesicular transport of cellular cargo requires targeted membrane fusion and formation of a SNARE protein complex that draws the two apposing fusing membranes together. Insulin-regulated delivery and fusion of glucose transporter-4 storage vesicles at the cell surface is dependent on two key proteins: the SNARE integral membrane protein Syntaxin4 (Sx4) and the soluble regulatory protein Munc18c. Many reported in vitro studies of Munc18c:Sx4 interactions and of SNARE complex formation have used soluble Sx4 constructs lacking the native transmembrane domain. As a consequence, the importance of the Sx4 C-terminal anchor remains poorly understood. Here we show that soluble C-terminally truncated Sx4 dissociates more rapidly from Munc18c than Sx4 where the C-terminal transmembrane domain is replaced with a T4-lysozyme fusion. We also show that Munc18c appears to inhibit SNARE complex formation when soluble C-terminally truncated Sx4 is used but does not inhibit SNARE complex formation when Sx4 is C-terminally anchored (by a C-terminal His-tag bound to resin, by a C-terminal T4L fusion or by the native C-terminal transmembrane domain in detergent micelles). We conclude that the C-terminus of Sx4 is critical for its interaction with Munc18c, and that the reported inhibitory role of Munc18c may be an artifact of experimental design. These results support the notion that a primary role of Munc18c is to support SNARE complex formation and membrane fusion.


Introduction
Vesicular trafficking in eukaryotic cells depends on targeted fusion reactions between vesicles and their specific target membranes. Two universally required components of the intracellular a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 membrane fusion machinery are soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins and Sec1/Munc18 (SM) proteins. Target membrane (t-) SNAREs, anchored to one membrane, form a molecular zipper with cognate vesicle (v-) SNAREs on a vesicle membrane generating an α-helical bundle that pulls the two membranes together and drives fusion [1][2][3].
Vesicular transport is dependent on SNARE complex assembly, which is regulated by SM proteins [4,5]. Thus, delivery of the glucose transporter-4 (GLUT4) to the cell membrane in response to insulin signaling requires the SM protein Munc18c as well as two t-SNAREs (Syn-taxin4 (Sx4), SNAP23) [6,7] and the v-SNARE (VAMP2/syntaptobrevin) [8]. Knockouts and overexpression of SM proteins have shown both positive and negative effects on SNARE complex assembly and vesicle fusion [9,10], and there is little agreement on the precise role of SM proteins.
In the cell, Sxs are anchored by a C-terminal transmembrane helix to the plasma membrane, providing an anchor for the adjoining SNARE helix (Fig 1A). Furthermore, a crystal structure of the neuronal SNARE complex shows that the C-terminal transmembrane domains of Sx1a and VAMP2 form continuous α-helices with their SNARE motifs [30]. This finding suggests that the transmembrane domain is not only an anchor but may also play an integral part in the protein structure. However, in vitro experiments designed to study Munc18:Sx interactions, both in the neuronal and GLUT4 systems, have generally used truncated Sxs, including soluble Sx lacking its C-terminal membrane anchor [15,18], soluble truncated Sx immobilized at its N-terminus (with the SNARE helix untethered) [31] or soluble Sx immobilized at its C-terminus [17,25,28]. Other reports have used full-length Sxs incorporated into proteoliposomes through native C-terminal membrane anchors [23,24,27,[32][33][34][35][36]. Based on the conclusions derived from these experiments we postulated that the nature of the Sx4 C-terminus affects the observed in vitro activity of Munc18c [37]. Here, we test the hypothesis that experimental design-specifically focusing on the Sx4 C-terminus-influences the observed effect of Munc18c on SNARE assembly. We show that when the Sx4 C-terminus is anchored, SNARE assembly occurs in the presence of Munc18c yet when the Sx4 C-terminus is not anchored SNARE assembly is inhibited in the presence of Munc18c. C-terminal fusion protein T4-lysozyme (Sx4 1-275 -T4L); and the full-length protein complete with C-terminal transmembrane domain helix (Sx4 1-298 -TMD). We hypothesized that C-terminal anchoring of the Sx4 constructs may mimic the in vivo situation where Sx4 is inserted into the plasma membrane by its C-terminal transmembrane helix. The Sx4 1-275 -T4L protein fusion was included as a way of reducing the Sx4 C-terminal flexibility that is likely to occur when the transmembrane domain is removed, and potentially structure the otherwise disordered C-terminal region in Sx4 1-275 [17]. A similar strategy has been used successfully to decrease the flexibility of loops and termini in GPCR research [38]. A range of Sx4 constructs were engineered with either a C-terminal His 6 -tag for purification and immobilization on metal affinity resin, or a TEV-cleavable N-terminal His 6 -tag for the production of de-tagged Sx4 proteins. The Sx4, Munc18c, SNAP23 and VAMP2 constructs used in this work are shown schematically in Fig 1B. C-terminally anchored Sx4 with a T4L fusion forms a complex with Munc18c We undertook pulldown experiments to investigate if the Sx4 1-275 -T4L construct behaved similarly to Sx4 1-275 . Sx4 constructs were immobilized via their C-terminal His 6 -tag onto metal affinity resin, so that all Sx4 constructs were similarly bound by their C-terminus. The results of these experiments clearly show that Munc18c binds to both C-terminally immobilized Sx4 proteins Sx4 1-275 -His and Sx4 1-275 -T4L-His (Fig 2). Importantly, only a very small amount of Munc18c was pulled down in the negative control (labeled Munc18c), confirming that the binding of Munc18c was specific to Sx4. Moreover, no Munc18c above the negative control levels was pulled down by Sx4 30-275 -T4L-His, consistent with previous observations that binding of Munc18c to Sx4 requires the Sx4 N-terminal residues (the Syntaxin N-peptide) [17,28].
The importance of the Sx4 N-peptide for binding to Munc18c was confirmed once again using ITC, where binding was detected for HMunc18c (Munc18c with an N-terminal His 6tag) and Sx4 1-275 -T4L-His, but not for HMunc18c and Sx4 30-275 -T4L-His (S1 Table). The ITC determined binding affinity, K d , of 100 nM for the HMunc18c:Sx4 1-275 -T4L-His interaction (S1 Table, [17,39]. These results indicate that the T4L fusion does not affect the binding affinity of Sx4 for Munc18c.

C-terminal anchoring of Sx4 slows dissociation from Munc18c
To assess more quantitatively the effect of anchoring the C-terminus of Sx4, we assayed Munc18c interactions with Sx4 constructs using bio-layer interferometry. When Sx4 constructs were immobilized via their C-terminal His 6 -tag onto Ni-NTA biosensors, both Sx4 variants-Sx4 1-275 -His and Sx4 1-275 -T4L-His-had similar association rate constants (k on , M -1 s -1 ), dissociation rate constants (off-rates, k off , s -1 ) and binding affinities ( Table 1, S2 Fig) for detagged Munc18c. In the reverse experiment, HLMunc18c was immobilized. In this case, Sx4 1-275 -T4L (K d = 31 nM) bound to HLMunc18c with a higher affinity than Sx4 1-275 (K d = 72 nM). Further analysis showed that the two pairs of proteins had similar association rate constants, but that the dissociation rate constant for HLMunc18c:Sx4 1-275 was~2.5 times faster than for HLMunc18c:Sx4 1-275 -T4L demonstrating a weaker association for the former.
We then investigated these same pairs of interactions by fluorescence anisotropy where neither protein is immobilized. The Sx4 proteins (Sx4 1-275 -His, Sx4 1-275 -T4L-His) were labeled with the Alexa Fluor1 488 maleimide probe (see Materials and Methods) and kinetic titrations of the labeled Sx4 proteins (each at 50 nM) were measured with increasing concentrations of HMunc18c. Addition of HMunc18c to the Alexa488-labeled Sx4 1-275 -His or Sx4 1-275 -T4L-His resulted in an increase in fluorescence anisotropy indicating formation of Munc18c complexes with both Sx4 variants (Fig 3A and 3B). The T4L fusion did not greatly affect the association rate constant. In contrast, the dissociation rate constant (k off , s -1 ) of HMunc18c: Sx4 1-275 -His was almost an order of magnitude faster than that of HMunc18c:Sx4 1-275 -T4L-His ( Table 1). These results are consistent with the outcomes from the bio-layer interferometry experiments and support the notion that the nature of the Sx4 C-terminus affects dissociation kinetics with Munc18c.
In combination, these results show that anchoring the C-terminus of Sx4 (either by using a C-terminal His-tag anchored to affinity resin or by using a C-terminal T4L fusion) slows dissociation of Sx4 from Munc18c, although the mechanism for this is unclear. Sx4 C-terminal anchoring promotes SNARE assembly in the presence of Munc18c We next investigated the effect of Sx4 C-terminal anchoring on the impact of Munc18c-mediated SNARE complex assembly. Using a pull-down assay, we explored the ability of Sx4 variants (Sx4 1-275 -His and Sx4 1-275 -T4L-His) to bind SNARE partners. Both Sx4 constructs, when immobilized on metal affinity resin by their C-terminal His 6 -tags in a pre-formed complex with Munc18c, were able to pull down SNAP23 and VAMP2 to form SNARE complexes after overnight incubation (Fig 4).
The importance of Sx4 C-terminal anchoring on SNARE complex assembly was then examined by using a pre-formed binary complex between immobilized HL-Munc18c and de-tagged Sx4 constructs in SNARE pull-down experiments (Fig 5). HL-Munc18c was first immobilized onto beads, then Sx4 proteins were pulled down. For this critical experiment, we used detagged Sx4 1-275 , Sx4 1-275 -T4L and the full-length Sx4 1-298-TMD (in detergent micelles). After extensive Table 1. Summary of kinetic data for the Munc18c-Syntaxin4 interaction. Values shown for the bio-layer interferometry experiments are mean ± s.d. where each experiment was repeated at least three times. Values and standard errors are shown for the fluorescence anisotropy experiments and these were obtained from a global fit to multiple kinetic data sets with differing concentrations of Munc18c. Bold text indicates dissociation rate constants (k off ) for Munc18c-Sx4 1-275 where Sx4 1-275 is free in solution.

Sx4 construct
Experimental design k on , M -1 s -1 k off , s -1 K d , nM

Sx4 Munc18c
Bio-layer interferometry Fluorescence Anisotropy  washing to remove unbound Sx4 proteins, the beads were incubated with SNARE partner proteins (SNAP23 and VAMP2). We found that when bound to HL-Munc18c, both Sx4 1-275 -T4L and Sx4 1-298 -TMD were able to assemble ternary complex by pulling down SNAP23 and VAMP2, whereas Sx4 1-275 did not (Fig 5). The experiments were repeated four times, including the use of different batches of each purified protein. In each experiment, the results were the same. This result shows that SNAP23 and VAMP2 are pulled down only when Munc18c is in complex with a Sx4 construct that has a C-terminal anchor. Pulldown experiments using T4L-His immobilized on beads showed negligible binding of SNAP23 or VAMP2 (S3 Fig) indicating that the capture of these two proteins (in the experiment presented in Fig 5) is due to interaction with Sx4 and not with T4L. Munc18c only inhibits SNARE assembly when Sx4 C-terminus is not anchored To assess the role of Munc18c in SNARE complex formation when each cognate SNARE and SM protein is free in solution, we made use of a fluorescence anisotropy assay that has been used previously to assess neuronal SNARE assembly [15,18]. In the previously reported work, SNARE complex formation using the soluble cytoplasmic form of neuronal Sx1a (Sx1a 1-262 ) was inhibited in the presence of Munc18a. To gain insight into the importance of the Sx4 Cterminus, our experiments were designed to evaluate whether the same result occurs for Sx4  (analogous to the cytoplasmic Sx1a 1-262 construct) and whether this result changes when the C-terminus is fused to T4L (Sx4 1-275 -T4L). SNARE complex formation was monitored by the change in fluorescence anisotropy of Alexa488 labeled VAMP2 in the presence or absence of Munc18c. In the absence of Munc18c, SNARE ternary complex was formed in the presence of both Sx4 1-275 -His and Sx4 1-275 -T4L-His at almost the same rates ( Table 2). In contrast, the rate of SNARE complex formation for Sx4 1-275 -His in the presence of HMunc18c is 7-fold lower than for Sx4 1-275 -His alone (Table 2, Fig 6A). These findings for Munc18c are in agreement with the results of previous studies on neuronal counterparts [15,16,18] that showed Munc18a inhibited assembly of Sx1a 1-262 SNARE complexes in solution.
When we conducted the same experiment using Sx4 1-275 -T4L-His the rate of SNARE complex formation is unchanged by the addition of HMunc18c (Table 2, Fig 6B). This outcome is in line with the results reported above from pull-down assays where Sx4 1-275 -T4L or Sx4 1-298 -TMD bound to immobilized HLMunc18c were able to pull down SNARE partners, whereas Sx4 1-275 bound to Munc18c failed to form SNARE complexes unless the Sx4 1-275 was C-terminally immobilized on beads.
Removing the Sx4 N-peptide releases the effect of Munc18c on SNARE assembly To assess whether the N-terminal peptide of Sx4 plays a role in SNARE assembly, we performed fluorescence anisotropy experiments analogous to those reported previously for Munc18a:Sx1a [18]. We studied complex formation for Sx4  in the presence or absence of Munc18c (Table 2, Fig 6C). We also analyzed SNARE complex formation using Sx4 30-275 -T4L  Fig 6C). When the C-terminal Sx4 30-275 -T4L fusion protein is used, there is again no difference observed in the rate of SNARE assembly in the presence of Munc18c (Table 2, Fig 6D). Indeed, for Sx4s lacking Nterminal residues, SNARE assembly in solution proceeds at very similar rates whether or not the Sx4 C-terminus is fused to T4L and whether or not Munc18c is present. We note that Sx4 30-275 binds very weakly, if at all, to Munc18c [17] (Fig 3, S1 Table), so that SNARE assembly for this construct probably does not involve Munc18c. Thus the "release" of Munc18c inhibition observed when the N-peptide of Sx4 is removed, may simply be a consequence of Munc18c not participating in SNARE assembly.

Discussion
Previously reported in vitro evidence for the role of Munc18c has been used to support both positive [27,28] and negative [32,40] regulatory roles on SNARE assembly. This situation is mirrored for the neuronal counterpart Munc18a [15, 16, 18, 19, 21-27, 33, 34, 41]. We questioned whether these contradictory conclusions for the role of Munc18c -derived from in vitro experiments-could be a consequence of experimental design.
Removing the Sx4 C-terminus would have two immediately apparent effects: first, Sx4 would not be associated with a membrane; and, second, the Sx4 SNARE region would not be anchored at its C-terminus and its C-terminal residues would be less ordered/more flexible. Indeed, a crystal structure of the C-terminally truncated Sx1 in complex with Munc18a revealed that the C-terminal region of Sx1 was disordered [42]. Similarly, the SNARE region of a soluble C-terminally truncated Sx1 construct was flexible [17,43]. On the other hand, a crystal structure of the neuronal SNARE complex that includes the Sx1 and VAMP2 C-terminal transmembrane domains showed a continuous helical structure for the SNARE motifs and TMD domains [30]. Moreover, NMR analysis of Sx1 including the transmembrane domain in micelles showed two well-ordered helices in the SNARE domain and a well-ordered helix in the transmembrane region [44]. Together this evidence supports the conjecture that removing the transmembrane anchor of Sxs could affect the C-terminal SNARE structure. Here we investigated whether this effect might impact on the outcome of in vitro SNARE assembly experiments.
Using multiple lines of evidence, we found a consistent relationship. In essence, different experimental designs give rise to contradictory conclusions on the role of Munc18c in SNARE assembly. Strategies in which the Sx4 C-terminus is attached to beads in in vitro pull-down assays show that Munc18c does not inhibit SNARE assembly. Conversely, strategies in which Sx4 is not immobilized by its C-terminus or is free in solution show that SNARE assembly is inhibited when C-terminally truncated Sx4 is used, but not when using Sx4 that includes its TMD, or when fused to T4L. The in vitro experimental design using a C-terminal anchor (T4L or TMD in a detergent micelle) does not include a membrane bilayer, but nevertheless the outcomes closely resemble those reported for Sx4 reconstituted in liposomes [27].
Unlike neuronal Munc18a, Munc18c is thought to only bind Sx4 in an open conformation [17,27]. The inclusion of a Sx4 C-terminal anchor in in vitro experiments may enable a conformation of Sx4 in the Munc18c:Sx4 complex that can also accommodate SNAP23 and VAMP2 binding. This finding is of critical importance in the field, and explains conflicting conclusions from studies of the role of Munc18c in promoting SNARE assembly. In its native form Sx4 is membrane-embedded. Our results thus provide compelling evidence that Munc18c does not inhibit SNARE assembly in vivo. We reason that the negative regulatory role ascribed to Munc18c is artifactual and is a consequence of using soluble truncated Sx4. However, further experiments will now be required to define the precise role of the Sx4 C-terminal region in modulating Munc18c/SNARE interactions.
More broadly, our findings suggest that experimental design and interpretation of in vitro data should perhaps be revisited for other vesicle fusion systems. We note, for example, that in vivo studies using Sx1 with a lipidic anchor in place of the transmembrane helix fully supported fusion, but that distancing the SNARE motif from the membrane inhibits membrane fusion [45], in support of our model. Our findings suggest that the most appropriate Sx construct to use for in vitro experiments with Munc18 partners is the membrane-anchored form or, failing that, Sx with a C-terminal tag attached to affinity beads, or a C-terminal fusion such as T4L.
In summary, our data show that Munc18c only has an inhibitory effect on SNARE assembly when the Sx4 C-terminus is removed. Because native Sx4 is anchored by its C-terminus to the plasma membrane, we conclude that Munc18c almost certainly does not block native Sx4 SNARE assembly in vivo. Other membrane fusion components may exert complex regulatory control in these exquisitely organized trafficking systems, including a wide range of positive and negative regulation of Munc18 proteins by a wide range of mechanisms. Finally, our findings call into question the common practice of removing transmembrane anchors to study protein interactions when the interacting region of the protein adjoins the transmembrane helix.

Materials and methods Constructs
Plasmids encoding rat Sx4 (C141S, amino acids 1-275), with a C-terminal hexahistidine (His 6 ) tag, and the glutathione S-transferase (GST) fusion proteins, mouse SNAP23 (amino acids 1-210), rat VAMP2 (amino acids 1-96) were generated as described previously [17,28]. Mouse HMunc18c (N-terminal His 6 -tag), HTMunc18c (N-terminal His 6 -tag, TEV cleavage site, referred to as detagged Munc18c after the His tag is removed) and HL-Munc18c (N-terminal His 6 -tag, with a 53 amino acid linker) constructs were also prepared as described by Rehman et al. [39]. The 53 amino acid linker has the following sequence: MSPIDPMGHHHHHHGRRASV AAGILVPRGSPGLDGIYARGIQASMAAGFG. For production of T4L fusion constructs, a synthetic gene for T4L (amino acids 1-164) was purchased from GeneArt1 Gene Synthesis (Life Technologies, Carlsbad, CA) and cloned into the pET20b vector using the HindIII and XhoI restriction sites. The cytoplasmic Sx4 constructs (amino acids 1-275 and 30-275) were subsequently ligated into the pET20b-T4L vector between the NdeI and HindIII sites, resulting in the constructs designated Sx4 1-275 -T4L-His and Sx4 30-275 -T4L-His. Constructs containing a TEV cleavage site were generated by ligation independent cloning (LIC) [46]. TEV cleavage was required to produce de-tagged protein for the analysis of the interaction between unbound Sx4 and Munc18c, and unbound Sx4 and SNARE proteins. To generate constructs with an N-terminal cleavable His-tag, a ligation independent cloning (LIC) strategy was used [46]. For this, each fragment i.e. Sx4 1-275 , Sx4 1-275 -T4L, Sx4 1-298 -TMD or Munc18c was PCR amplified using specific forward (5'-CAGGGCGCCATGCGCGACAGGACCCATGAGTTGAGGC-3') and reverse (5'-GACCCGACGCGGTAAGTGAGCTCCAGGTTTTTATACGCAT-3') primers containing LIC overhangs (in bold). The PCR product was then treated with T4-polymerase. T4 polymerase treated DNA was ligated to a LIC vector (pMCSG7 vector digested with Ssp1 restriction enzyme). The LIC vector pMCSG7 encodes an N-terminal leader sequence containing a His 6 fusion tag and a TEV protease site (EXXYXQG/S). The plasmid was isolated and the sequence confirmed (Australian Genome Research Facility, University of Queensland).
To generate HT-Sx4-TMD a synthetic gene encoding full length Sx4-TMD (mouse, amino acids 1-298) was purchased from GenScript1 (Piscataway, NJ) and cloned into a LIC vector with an N-terminal His 6 affinity tag as described above.

Production of soluble Sx, SNAP23, VAMP2 and Munc18 constructs
Munc18c proteins were expressed in E. coli and purified as described previously [39], with the following changes. Tris was replaced with 50 mM phosphate (pH 8.0) in all purification buffers. Cell pellet from 1 L of cells (OD~18) was lysed using 500 mg lysozyme in lysis buffer (300 mL) at room temperature. The soluble forms of Sx4 proteins and its T4L fusions were expressed in E. coli and purified using the methods described for Sx4 previously [17,28]. To remove the His 6 -tag, proteins eluted after immobilized metal ion affinity chromatography (IMAC) with TALON™ Co 2+ resin were mixed with TEV protease (A 280 ratio of 1:100, protease: protein) and incubated overnight at 4˚C. The following day, the protein was subjected to a reverse IMAC step and cleaved protein collected in the flow-through was further purified on a size exclusion chromatography column (SEC) Superdex S200 16/60 using an Ä KTA FPLC system (GE Healthcare, Little Chalfont, UK). An additional serine residue remained at the N-terminus of Sx4 and Munc18c proteins after TEV protease treatment (Fig  1B). GST-cleaved SNAP23 and VAMP2 were produced and purified as described previously [28]. Briefly proteins were lysed in lysis buffer (25 mM TrisCl, pH7.5, 150 mM NaCl, 0.5% Triton-X 100, 2 mM βME). The lysate was subsequently cleared by centrifugation and incubated with GSH-agarose resin (Thermo-Fisher Scientific, Massachusetts, USA) for 2 hrs. The beads were then washed with wash buffer (25 mM TrisCl, pH7.5, 150 mM NaCl, 2 mM βME) prior to treatment with thrombin to cleave the GST affinity tag. Cleaved protein was further purified by anion exchange chromatography on a MonoQ HR 5/5 column (GE Healthcare, Little Chalfont, UK) for SNAP23 and by cation exchange chromatography on a MonoS HR 5/5 column (GE Healthcare, Little Chalfont, UK) for VAMP2 and VAMP2 E78C.
In a second set of experiments, HL-Munc18c (20 μg, 1.4 μM in a volume of 200 μL) was first immobilized on TALON™ Co 2+ beads. Beads were incubated separately with de-tagged Sx4 1-275 , Sx4 1-275 -T4L or Sx4 1-298 -TMD (50 μg, 21.7 μM) for 4 h in the binding buffer described above and washed with binding buffer. For quaternary complex formation, the beads (containing HL-Munc18c/Sx4 binary complexes or Sx4 proteins without HL-Munc18c for the control) were mixed with purified SNAP23 (100 μg, 21.7 μM)) and VAMP2 (80 μg, 31.7 μM) proteins and incubated overnight at 4˚C using each of the three de-tagged Sx4 proteins (Sx4 1-275 , Sx4 1-275 -T4L and Sx4 1-298 -TMD) in binding buffer. Unbound and non-specific proteins were removed by washing with binding buffer and the level of binding was analyzed by SDS-PAGE. Each binding assay was repeated at least three times to confirm the results. Protein concentrations were measured using the Bio-Rad protein assay with bovine serum albumin used as a standard. Nonspecific

Isothermal titration calorimetry (ITC)
ITC experiments were carried out at 298 K using an iTC200 (Malvern Instruments Ltd., Malvern, United Kingdom). The proteins were buffer exchanged in ITC buffer (25 mM HEPES pH 8.0, 200 mM NaCl, 10% (v/v) glycerol and 2 mM β-ME) by gel filtration prior to ITC experiments. Sx4 1-275 -T4L-His or Sx4 30-275 -T4L-His at a concentration of 200-250 μM was titrated into 20-40 μM of HMunc18c in the cell. Injection volumes of 2.5 μL were used for all titrations. The heat released was measured and integrated using the Microcal Origin 7.0 program with a single site binding model to calculate the equilibrium association constant K a (= 1/K d ), enthalpy of binding (ΔH) and the stoichiometry (n). The Gibbs free energy (ΔG) was calculated using the equation: Binding entropy (ΔS) was calculated by To determine the average standard error of the mean (SEM) values of binding affinities, four experiments were performed for each set of samples.

Bio-layer interferometry
The binding kinetics of Sx4 1-275 or Sx4 1-275 -T4L for their interaction with Munc18c was determined by bio-layer interferometry using a BLItz 1 system (ForteBio, Menlo Park, CA). In the first series of experiments C-terminally His-tagged Sx4 constructs (200-400 nM) were loaded onto Ni-NTA biosensors until a binding height of~1 nm was reached. Sx4 immobilized sensors were dipped into a dilution series of de-tagged Munc18c at concentrations of 25, 50, 100, 200 and 400 nM for Sx4 1-275 -His and 10, 25, 50, 100, 200 and 400 nM for Sx4 1-275 -T4L-His immobilized biosensors. The association reaction was allowed to proceed for 60-100 s and the dissociation reaction for 120-180 s. Experiments were conducted with a shaking speed of 2200 rpm in 25 mM Tris-Cl pH 7.5, 150 mM NaCl, 10 mM imidazole, 0.1% (v/v) TX-100, 10% (v/v) glycerol and 2 mM βME buffer. We were unable to assess the association of Munc18c with Sx4 1-298 -TMD using biolayer interferometry because of poor yield and purity of C-terminally His-tagged full-length Sx4 protein.
In a second series of experiments HL-Munc18c was immobilized onto Ni-NTA biosensors until a binding height of~1 nm was achieved. The biosensors were dipped into a concentration series of 25, 50, 100, 200 and 400 nM for detagged Sx4 1-275 and 25, 50, 100, 200 and 400 nM for detagged Sx4 1-275 -T4L. The association reaction was for 100 s and dissociation was followed for 120 s. Experiments were conducted three times using at least two replicates from independently prepared samples and the data were fitted to an interaction with 1:1 stoichiometry using BLItz 1 Pro Software.

Labeling proteins for fluorescence experiments
Cysteine variants were prepared using QuikChange Mutagenesis (Stratagene, San Diego, CA). These were (i) Sx4 1-275 -His and Sx4 1-275 -T4L (native C141 rather than C141S), to study Munc18c/Sx4 kinetics and (ii) GST cleaved VAMP2 E78C to study SNARE ternary complex formation. Site-specific labeling of these two proteins was performed using the thiol reactive Alexa488 C 5 Maleimide dye according to the manufacturer's instructions (Alexa Fluor1 488, Life Technologies, Carlsbad CA). Briefly, purified protein (200 μM) was mixed with dye to give a final dye concentration of 1 mM (in DMSO, final concentration 10% v/v) and incubated for 2-3 h in the dark at room temperature. Unbound dye was removed by sequentially passing the protein-dye mix through two equilibrated desalting PD10 columns (GE Healthcare, Little Chalfont, UK). The protein was collected and concentrated to the desired concentration. The absorbance of labeled protein at 280 nm (A 280 ) and 494 nm (A 494 ) was measured using a 1 cm cuvette in a nanodrop, and the degree of labeling was then calculated as recommended by the supplier: where the approximate molar extinction coefficient of Alexa 488 (ε Alexa-488 ) at 494 nm is 71000 M -1 cm -1 . The molar extinction coefficient of the protein (ε protein ) was calculated from each protein sequence using the Expasy Protparam server [47]. In all cases, the degree of labeling was calculated to be greater than 90%.

Fluorescence anisotropy measurements
Long-time base-fluorescence anisotropy measurements were performed at 25˚C in 1 mL quartz cuvettes (Hellma, Mullheim, Germany) without stirring on a Fluoromax-4 spectrofluorometer fitted with polarizers of L-geometry and slits set at 4 nm (Jobin Yvon Inc., Edison, NJ). The experiments were carried out in a buffer containing 25 mM HEPES pH 8.0, 200 mM NaCl, 1 mM TCEP and 2% (w/v) glycerol. The Alexa488 fluorophore was excited at 488 nm and emission was observed at 515 nm. Fluorescence anisotropy (r) was determined as: < r >¼ I VV À G I VH =I VV À 2GI VH Investigation of Syntaxin4 C-terminal anchoring where I represents the fluorescence intensity, and the first subscript letter indicates the direction of excited light and second subscript shows the emitted light. The intensities of vertically (V) and horizontally (H) polarized emission light after excitation were measured. The "G-factor" G was defined as: Primary data analysis was performed with the programs Graffit 5.0 (Erithacus software) and Fluorescence implementation (Jobin Yvon Inc., Edison, NJ) of the Origin7.0 software (Origin lab Corporation, Northampton, MA). Experiments investigating the kinetics of the interaction between Sx4 1-275 -His or Sx4 1-275 -T4L-His and HMunc18c were performed once at each concentration of HMunc18c. The experiments investigating the formation of the SNARE complex used Alexa488-labeled VAMP2 (GST-cleaved), GST cleaved SNAP23, and the same Sx4 and Munc18c proteins, and were each performed at least three times. All spectra were corrected for background fluorescence from buffer.
To confirm the specificity of binding and complex formation, all proteins (Sx4 1-275 -His, Sx4 1-275 -T4L-His, HMunc18c, GST-cleaved SNAP23) were tested individually for their ability to bind Alexa488-labeled VAMP2 (GST cleaved). The change in anisotropy was monitored and showed that none of these proteins interacted on their own with VAMP2-Alexa488 under the given experimental conditions. We were unable to assess the association of HMunc18c with the full length Sx4 1-298 -TMD-His using fluorescence anisotropy because the presence of detergent prevented a stable baseline under the same conditions. The global fit for the anisotropy data was calculated using the Dynafit 4.0 program. During the fitting procedure (1:1 complex), the concentration of proteins was allowed to vary and then checked against the anticipated value. The script used in global calculations of kinetic parameters for the interaction of Munc18c with Sx4 or Sx4-T4L is provided as supporting information (S1 Text).