SNAP-25 with mutations in the zero layer supports normal membrane fusion kinetics

Intracellular membrane fusion involves a highly conserved machinery that functions in essentially all vesicular traffic steps and is conserved throughout evolution (Bennett and Scheller, 1993; Ferro-Novick and Jahn, 1994; Hanson et al., 1997; Rothman, 1994). The core of this machinery is formed by a complex of the SNARE (soluble NSF (N-ethylmaleimide-sensitive factor) attachment protein receptor) proteins. In regulated exocytosis in neurons, and certain other cell types, this SNARE complex is formed from syntaxin 1, SNAP-25 and VAMP (synaptobrevin) (Sollner et al., 1993b). These proteins can form a highly stable complex in vitro (Hayashi et al., 1994; Hayashi et al., 1995) due to the formation of a four-helix bundle with one helix donated each by syntaxin 1 and by VAMP and two helices donated by SNAP-25 (Poirier et al., 1998; Sutton et al., 1998). So far it has been shown that a SNARE complex of similar organisation consisting of related members of the SNARE family also functions in constitutive exocytosis in yeast (Katz et al., 1998; Rossi et al., 1997) and in endosome-endosome fusion (Antonin et al., 2000). Complexes of this type are likely, however, to have general functions in membrane fusion as SNARE homologues appear to be essential for all traffic steps (Hay and Scheller, 1997; McNew et al., 2000a; Pelham, 1999; Rothman, 1994). It has been suggested that a stable SNARE complex, similar to that seen in the crystal structure, represents a conserved intermediate in membrane fusion reactions (Hanson et al., 1997; Sutton et al., 1998; Weber et al., 1998). While others have suggested that it is not involved in fusion itself but in a preceeding step (Tahara et al., 1998; Ungermann et al., 1998), accumulating evidence is more consistent with the SNAREs alone or with other proteins playing a major role in bilayer fusion (Chen et al., 1999; Grote et al., 2000; Hanson et al., 1997; McNew et al., 2000b; Weber et al., 2000).

A current model for SNARE action in membrane fusion (Bock and Scheller, 1999; Brunger, 2000; Chen et al., 1999; Xu et al., 1998; Xu et al., 1999b) suggests that the interaction between vesicular and target SNAREs leads to the formation of an initial ‘loose complex’ (sensitive to the action of Clostridial neurotoxins) that brings the two opposing bilayers together. Membrane fusion would then be triggered by the progressive assembly (Fiebig et al., 1999) or zipping up of the SNAREs into a neurotoxin-insensitive ‘tight’ complex believed to be similar to the four-helix bundle visualised in the crystal structure (Sutton et al., 1998). While this is a plausible explanation for membrane fusion it leaves unexplained the observed biophysical aspects of fusion detected during exocytosis. In particular, an accumulation of data supports the idea that the initial fusion occurs through the formation of a fusion pore (Albillos et al., 1997; Ales et al., 1999; Alvarez de Toledo et al., 1993; Breckenridge and Almers, 1987; Fernandez et al., 1984; Lindau and Almers, 1995). This is a reversible structure that can return to the prefusion state even after a phase of pore expansion (Ales et al., 1999; Rosenboom and Lindau, 1994) that may be under physiological regulation (Burgoyne and Alvaraz de Toledo, 2000; Fernandez-Chacon and Alvarez de Toledo, 1995; Fisher et al., 2001; Hartmann and Lindau, 1995; Scepek et al., 1998). In synaptic vesicle exocytosis, this reversibility could enable a kiss-and-run type fusion (Fesce et al., 1994; Stevens and Williams, 2000) to occur with release of vesicle content through the pore on a sub-millisecond time-scale. The high stability of the SNARE structure formed in vitro from recombinant proteins or present in detergent extracts and the fact that it can only be disassembled by the action of the chaperones α-SNAP and NSF (Sollner et al., 1993a), is difficult to reconcile with a rapidly reversible fusion process (Burgoyne and Alvaraz de Toledo, 2000). It is not clear, therefore, if the stable structure seen in vitro is the structure that drives fusion or if it is a post-fusion ground state of the complex. In the case of influenza haemaglutinin-mediated fusion it appears that the stable low-pH structure originally thought to represent the fusogenic state of the protein may instead be an inactive conformation existing after the completion of fusion (Lentz et al., 2000).

The neuronal SNARE complex is stabilised by hydrophobic interactions between the four helices in a series of layers (–1 to –7 and +1 to +8) (Sutton et al., 1998). Mutations in these layers disrupt SNARE-SNARE interactions, compromise SNARE complex stability and are functionally disruptive (Chen et al., 1999; Fasshauer et al., 1998; Washbourne et al., 1999). At the 0 layer are ionic interactions between three conserved glutamines, two from SNAP-25 and one from syntaxin, and a conserved arginine from VAMP. These residues are absolutely conserved thoughout evolution leading to a suggested classification of the SNAREs as the ‘Q’ and ‘R’ SNAREs (Fasshauer et al., 1998). The conserved Q and R residues were suggested to be important either in stabilising the SNARE complex or for its disassembly by α-SNAP and NSF (Sutton et al., 1998). Mutagenesis of these residues in yeast in the exocytotic SNAREs has shown that they are biologically relevant and confirmed their crucial importance for growth and secretion (Katz and Brennwald, 2000; Ossig et al., 2000). These studies in yeast do not, however, provide information on when and where these residues are of importance. Mutation of conserved Q residues in SNAP-25 or the yeast homologue Sec9 reduces the affinity of SNARE-SNARE interactions and reduces the thermal stability of the SNARE complex assembled in vitro (Chen et al., 1999; Katz and Brennwald, 2000; Ossig et al., 2000; Wei et al., 2000). Surprisingly mutation of one or other of the conserved glutamines of SNAP-25 did not modify the ability of individual helices to reconstitute exocytosis in populations of PC12 cells in which endogenous SNAP-25 was cleaved with botulinum neurotoxin E (BoNT/E) (Chen et al., 1999; Scales et al., 2000). This toxin cleaves SNAP-25 within the C-terminal helix to release a 22-residue fragment and inactivate the endogenous protein (Binz et al., 1994; Schiavo et al., 1993). It has been pointed out, however, that the assay used with PC12 cells would not reveal subtle effects of these mutations such as on the kinetics of membrane fusion (Katz and Brennwald, 2000; Ossig et al., 2000). In another study, a SNAP-25 (Q174L) mutant was expressed in adrenal chromaffin cells (Wei et al., 2000). This mutant reduced the extent but not the initial overall kinetics of exocytosis from the releasable pool. The interpretation of the data from this study is complicated as the cells still retained their endogenous wild-type SNAP-25 intact and so this could contribute to the apparently normal exocytosis kinetics.

In this study we have examined the importance of the conserved Q residues in SNAP-25 in cells that express BoNT/E-resistant constructs, in which endogenous SNAP-25 has been inactivated by BoNT/E treatment. This has allowed us to examine, in the absence of a wild-type background, the consequences of mutations in the Q residues not only on the overall level of exocytosis but also on the kinetics of single granule release events measured using carbon-fibre amperometry (Wightman et al., 1991). We show that disruptive mutations in the 0 layer that impair SNARE-SNARE interactions and SNARE complex stability do not prevent exocytosis, and do not affect the kinetics of single release events. These data show that SNARE complex function in membrane fusion is tolerant of significant modifications in the 0 layer.

The plasmid pEGFP-C1 was obtained from Clontech (Basingstoke, UK). The plasmid encoding BoNT/E light chain as an EGFP fusion construct (pEGFP-BoNT/E) was described previously (Graham et al., 2000). The BoNT/E-resistant construct of SNAP-25 (Em2) was described previously (Washbourne et al., 1999). Specific mutations were introduced into Em2 using site directed mutagenesis with a Quickchange site-directed mutagenesis kit (Stratagene Europe, Amsterdam, The Netherlands). The primers used were as follows with the changed codons underlined:

5′-GTTATGTTGGATGAGGAAGGCGAACAACTG-3′ (Q53E sense); 5′-CAGTTGTTCGCCTTCCTCATCCAACATAAC-3′ (Q53E antisense); 5′-GAGATTGACACCGAGAATCGCCAGATTGAC-3′ (Q174E sense); 5′-GTCAATCTGGCGATTCTCGGTGTCAATCTC-3′ (Q174E antisense); 5′-TTATGTTGGATGAGCGAGGCGAACAACTG-3′ (Q53R sense); 5′-CAGTTGTTCGCCTCGCTCATCCAACATAAC-3′ (Q53R antisense); 5′-GAGATTGACACCCGGAATCGCCAGATTGAC-3′ (Q174R sense); 5′-GTCAATCTGGCGATTCCGGGTGTCAATCTC-3′ (Q174R antisense). The SNAP-25 delta35-44 mutant was created by the megaprimer PCR method using the primer 5′-CGTCGCATGCTGCTGCAAAGGACTTTGGTTATG-3′, which loops out bases 113-132 of the SNAP-25 coding sequence and introduces a silent point mutation (G112A), eliminating the PstI site for screening purposes. This primer was used in conjunction with the primers SBf and SBr described previously (Washbourne et al., 1999) and the product inserted into pCDNA/HA1. All constructs were checked by automated sequencing.

The assay of release from transfected PC12 cells used a modification of the growth hormone (GH) release assay (Wick et al., 1993). PC12 cells were maintained in culture in collagen-coated 24 well trays and transiently co-transfected with 4 μg growth hormone plasmid pXGH5, along with 2 μg of each test plasmid (control vector, Em2 or Em2 containing additional mutations) as previously described (Graham et al., 1997) using lipofectamine (Gibco BRL). GH secretion was used as a marker for transfected cells. Cells were permeabilized 72 hours after transfection with 20 μM digitonin, incubated with or without bacterially expressed 14 nM BoNT/E-His₆ light chain (Glenn and Burgoyne, 1996) for 40 minutes and then challenged by the addition of buffer containing no Ca²⁺ or with 10 μM Ca²⁺. Buffer samples and the cells were processed and assayed for GH levels using an enzyme-linked immunosorbent kit according to the manufacturer’s instructions (Boehringer-Mannheim, Indianapolis, IN). In all experiments, the amount of GH release is expressed as a percentage of total cellular content of GH. To verify expression levels of the transfected Em2 constructs, cells were solubilised, separated by SDS polyacrylamide gel electrophoresis and samples probed by immunoblotting with mouse monoclonal anti-HA at 1:1000 dilution.

Freshly isolated bovine adrenal chromaffin cells (Burgoyne, 1992) were plated on non-tissue culture-treated 10 cm Petri dishes at a density of 1×10⁶/ml. The following day non-attached cells were pelleted by centrifugation and re-suspended in growth medium at a density of 1.5×10⁷/ml. The cells (1 ml) were mixed with 22.5 μg of pEGFP, 7.5 μg of pEGFP-BoNT/E and 30 μg of the Em2 construct to be tested. 1 ml of cells and plasmids were electroporated at 250 V and 975 μF for one pulse, using a Bio-Rad Gene Pulser II (Bio-Rad, CA) and 4 mm cuvettes. Cells were then diluted as rapidly as possible to 1×10⁶/ml with fresh growth media and 1×10⁶ cells grown on 35 mm Petri dishes in a final volume of 3 ml of growth medium for a further 3-5 days.

The cells were washed three times with Krebs-Ringer buffer (145 mM NaCl, 5 mM KCl, 1.3 mM MgCl₂, 1.2 mM NaH2PO4, 10 mM glucose and 20 mM Hepes, pH 7.4), incubated in bath buffer (139 mM potassium glutamate, 20 mM Pipes, 0.2 mM EGTA, 2 mM ATP, 2 mM MgCl₂, pH 6.5) and viewed using a Nikon TE300 inverted microscope. Transfected cells were identified as those expressing EGFP. A pre-cut 5 μm diameter carbon fibre electrode (NPI, Germany), was placed in direct contact with the surface of a cell. For stimulation of the cells, a digitonin-permeabilisation protocol (Jankowski et al., 1992) was used. A micropipette was filled with cell permeabilisation buffer (139 mM potassium glutamate, 20 mM Pipes, 5 mM EGTA, 2 mM ATP, 2 mM MgCl₂, 20 μM digitonin and 10 μM free Ca²⁺, pH 6.5) and positioned on the opposite side of the cell from the carbon fibre and buffer pressure ejected on to the cell for 20 seconds. A holding voltage of +700 mV was applied between the carbon fibre tip and the Ag/AgCl reference electrode in the bath. Amperometric responses were monitored with a VA-10 amplifier (NPI Electronic, Tamm, Germany), collected at 4 kHz, digitised with a Digidata 1200B acquisition system and monitored online with the AxoScope 7.0 program (Axon Instruments, CA). Data were subsequently analysed using an automated peak detection and analysis protocol within the technical graphics program Origin (Microcal, MA). Spikes were only analysed in detail if they had a base width greater than 6 milliseconds and an amplitude greater than 40 pA. This amplitude was chosen so that analyses were confined to spikes arising immediately beneath the carbon fibre and to limit effects on the data of diffusion times from exocytotic sites distant from the carbon fibre.

We have previously developed a mutated form of SNAP-25b (Em2) that is resistant to cleavage by BoNT/E and can reconstitute exocytosis following transfection in BoNT/E-treated PC12 cells (Washbourne et al., 1999; Washbourne et al., 2001). This construct allowed the testing, in the present study, of the effect of additional mutations in SNAP-25 on exocytosis. The Em2 mutant (Washbourne et al., 1999) contains three mutations including a change of the conserved isoleucine in the +2 layer to glutamate (Fig. 1). The mutations in Em2 do not affect its ability to become incorporated into a SNARE complex but its interactions with other SNAREs are compromised as it cannot form a binary complex with VAMP suggesting that this mutation could already impair SNARE complex formation and stability in vivo (Washbourne et al., 1999). The intention in this study was to introduce further disruptive mutations in the 0 layer residues and to examine the consequences for the extent and kinetics of exocytotic fusion events supported by the constructs. A major advantage of this assay is that it is not confounded by potentially disruptive effects of introduced inhibitory constructs. In addition, reconstitution will only occur if constructs are correctly post-translationally modified and targeted.

As shown in Fig. 2, treatment of control-transfected PC12 cells with BoNT/E light chain after permeabilisation essentially abolished Ca²⁺-induced exocytosis (in cells transfected with pcDNA3) measured as growth hormone (GH) release from co-transfected cells. Prior transfection with the Em2 construct completely reconstituted exocytosis as described previously (Washbourne et al., 1999). It has been shown that the C-terminal helix of SNAP-25 alone can, at least partially, restore exocytosis in BoNT/E-treated PC12 cells (Chen et al., 1999), although co-addition with the N-terminal helix is more effective (Scales et al., 2000). To determine whether the reconstitution by full-length SNAP-25 in our studies required a functional N-terminal helix or if the C-terminal helix of the protein was sufficient, a disruptive deletion (Δ35-44) was introduced within the N-terminal helix of Em2 removing residues from layers –3 to –5. The Δ35-44 construct was expressed to similar levels as Em2 (Fig. 2A) but gave almost no reconstitution (Fig. 2B). This shows that the reconstitution of exocytosis in our assay involved contributions from both the N- and C-terminal helices of the expressed full-length protein and would, therefore, allow testing of the effect of mutations in either helix. The result with the Δ35-44 construct also rules out a trivial explanation for the maintenance of exocytosis by the toxin-resistant mutant due to binding and sequestration of the BoNT/E toxin.

To disrupt 0 layer interactions, the conserved glutamines Gln53 and Gln174 were mutated individually or together to glutamate to introduce additional charges into the 0 layer. Fig. 2B shows that these additional mutations did not affect the overall reconstituting effect of transfected toxin-resistant SNAP-25 in PC12 cells. Since this assay would not reveal subtle effects of the mutations, exocytosis from the transfected cells was looked at in more detail by comparing Em2 with the double Q/E mutant. The time course of evoked release from transfected BoNT/E-treated cells was followed but no differences in the time course of release was seen between the two constructs (Fig. 3A). Cleavage of SNAP-25 by botulinum neurotoxins has been shown to modify the Ca²⁺-sensitivity of exocytosis implying that SNAP-25 may interact with the Ca²⁺-sensor (Gansel et al., 1987) such as synaptotagmin (Schiavo et al., 1997). We therefore, examined the effect of the Q/E mutations on the Ca²⁺-dependency of exocytosis in PC12 cells. No difference in Ca²⁺-dependency was observed, however, between exocytosis supported by Em2 and the double Q/E mutant in BoNT/E-treated cells (Fig. 3B).

In contrast to the apparent lack of effect of Q mutations in SNAP-25 in PC12 cells seen here and elsewhere, a partial reduction in exocytosis was observed in chromaffin cells overexpressing a SNAP-25 (Q174L) construct. Therefore, we examined the effect of the double Q/E mutant in adrenal chromaffin cells. These experiments gave us the opportunity to analyse in detail the kinetics of single granule release events by the use of carbon-fibre amperometry (Schroeder et al., 1996; Wightman et al., 1991) and thereby analyse the importance of the SNAP-25 0 layer residues for the kinetics of membrane fusion. In these experiments the cells were stimulated by local application of digitonin and Ca²⁺ to permeabilise the cells and directly activate exocytosis.

To inactivate endogenous SNAP-25 in chromaffin cells, they were transfected with a GFP-BoNT/E light chain construct for 3-5 days to allow time for effective cleavage of endogenous SNAP-25 and so avoid complications due to the functioning of residual wild-type protein. Only three of nine GFP-BoNT/E transfected cells still responded to stimulation and there was an overall 90% reduction in the mean number of evoked amperometric spikes (Fig. 4A) similar to previous findings (Graham et al., 2000). Co-transfection of the SNAP-25 Em2 construct along with the GFP-BoNT/E light chain effectively reconstituted exocytosis in chromaffin cells so that the evoked amperometric spikes due to the function of expressed SNAP-25 mutants could be examined. From the examination of a large number of cells it was apparent that Em2 expressing cells produced, on average, as many release events as seen in control non-transfected cells. While the double Q/E mutant did reconstitute exocytosis, the number of amperometric spikes elicited in cells expressing this mutant was reduced by around 50% compared with non-transfected cells (Fig. 4B). Fig. 4C-E shows typical overall responses from control non-transfected cells (in the same dishes as transfected cells) and cells transfected to express GFP-BoNT/E along with Em2 or the double Q/E mutant. Apart from the reduction in number of spike events in cells expressing the double Q/E mutant, the overall responses looked similar. To determine whether the double Q/E mutations affected the dynamics of the individual release events, the kinetics of individual amperometric spikes were analysed. We concentrated on analysis of the rise times of the spikes as this reflects the time course of initial fusion pore expansion (Schroeder et al., 1996). We have previously shown this parameter to be modified under certain specific conditions including overexpression of cysteine string protein (Graham and Burgoyne, 2000; Graham et al., 2000) or a munc-18 mutant (Fisher et al., 2001), although it is unaffected by changes in vesicle catecholamine content. This parameter can therefore report modifications in the fusion machinery. Comparison of the rise-times of spikes from control non-transfected, Em2-expressing and double Q/E mutant expressing cells showed no significant differences in the mean values (Fig. 5A) or in the frequency distribution of the individual spike values for the rise-times (Fig. 5B-D).

The above data suggest that the 0 layer residues of SNAP-25 do not have a crucial role in determining the kinetics of SNARE complex assembly during exocytotic membrane fusion. To confirm this idea we also examined the effect of Q→R mutations recently shown to significantly impair the affinity of SNARE interactions and to be functionally disruptive in yeast. Single substitutions of R for Q in either helix of Sec9 is sufficient to impair growth in yeast (Katz and Brennwald, 2000). Responses were determined in cells co-transfected with GH and with Q/R mutants in SNAP-25. The Q174R mutant (data not shown) and a double Q/R mutant were able to functionally replace the BoNT/E-cleaved endogenous SNAP-25 in exocytosis in PC12 cells (Fig. 6A), even though the double Q/R mutant was less well expressed than Em2 (Fig. 6B).

In this study we have exploited the use of a BoNT/E-resistant mutant of SNAP-25 to allow expression of engineered constructs in the absence of functional endogenous SNAP-25. As shown previously (Washbourne et al., 1999), the toxin-resistant form of SNAP-25 (Em2) fully supported exocytosis in permeabilised PC12 cells treated with BoNT/E light chain. In addition, the Em2 construct supported exocytosis in chromaffin cells co-transfected to express GFP-BoNT/E. In these cells the single granule release events supported by the BoNT/E-resistant mutant had identical kinetic properties to those in control non-transfected cells when assayed by amperometry. We examined the effect of additional mutations in the 0 layer residues by introducing them into the Em2 construct. In these assays support of exocytosis required a functional N-terminal helix as well as the C-terminal helix needed to replace that of the endogenous protein cleaved by BoNT/E. This differs from the situation in a previous study using permeabilised PC12 cells where addition of the C-terminal helix alone was stimulatory, although this was enhanced by co-addition of the N-terminal helix (Chen et al., 1999; Scales et al., 2000). The difference is likely to be due to our study being based on the use of a full length SNAP-25 protein rather than free helices. In our study, mutation of either or both conserved glutamines to glutamate had no effect on the reconstitution in PC12 cells in terms of the extent, time course or Ca²⁺-dependency of release. We have previously demonstrated that the rise-time of amperometric spikes, a correlate of fusion pore expansion, can be modified by overexpression of Csp (Graham and Burgoyne, 2000) or a Munc18 mutant (Fisher et al., 2001). Expression in chromaffin cells of the double Q/E mutant not only supported exocytosis but the single release events were not modified compared with control cells. Disruptive mutations within the 0 layer of one or both glutamines to arginine in the double Q/R mutant did not prevent the reconstitution of exocytosis in PC12 cells. Some differences are apparent in this and in previous studies (Chen et al., 1999; Scales et al., 2000; Wei et al., 2000) between PC12 and chromaffin cells. Mutations in Q residues reduced the overall extent of exocytosis in chromaffin but not in PC12 cells in this study and a Q174L mutation was shown to reduce sustained secretion in chromaffin cells (Wei et al., 2000). One possible explanation for these differences between cell types is that much of the exocytosis triggered in PC12 cells comes from docked granules. while sustained secretion in chromaffin cells requires recruitment of granules from a non-docked pool. Since α-SNAP and NSF appear to be important for recruitment (priming) and sustained secretion in chromaffin cells (Chamberlain et al., 1995; Xu et al., 1999a) these effects of mutations in SNAP-25 may be consistent with the Q residues of SNAP-25 being important in the disassembly of cis-SNARE complexes during priming.

The lack of effect of the Q→E and Q→R mutations and those previously reported (Chen et al., 1999; Scales et al., 2000; Wei et al., 2000) suggest a surprising level of tolerance to mutation in the 0 layer considering the evolutionary invariance of these residues and their conservation in all SNAREs acting in distinct vesicle fusion steps. Previous mutations examined in the 0 layer in assays for exocytosis have involved substitution of glutamine with uncharged residues (alanine, isoleucine or leucine) that may not have markedly altered the packing of the SNARE helices. We chose to substitute each glutamine with a glutamate so that the introduction of two negative charges in the 0 layer of the double Q/E mutant would cause electrostatic repulsion and disruption in this layer and potential weakening of the interactions between helices. Surprisingly even these mutations were tolerated, although the overall level of exocytosis was partially reduced in the chromaffin cells but not PC12 cells. In the case of the double Q/R mutant the presence of two bulky positively charged arginines would result in both electrostatic and physical distortion within the 0 layer, modifying the packing of the SNARE helices (such changes have been predicted for only a single arginine substitution within the 0 layer) (Ossig et al., 2000). The double Q/R mutant was, nevertheless, still functional in exocytosis. Overall, the results suggest that conservation of the 0 layer residues is not required for membrane fusion to proceed with normal kinetics.

Examination of mutations in the SNAREs involved in constitutive exocytosis in yeast have demonstrated the importance of the conserved 0 layer residues for growth and secretion (Katz and Brennwald, 2000; Ossig et al., 2000). However, these assays would be dependent not only on effective SNARE function in exocytosis but also on SNARE recycling. It has been suggested that the 0 layer may be more important for SNARE disassembly and recycling by α-SNAP and NSF, although mutations in the 0 layer did not apparently affect NSF-mediated disassembly in vitro (Chen et al., 1999). Their effects on the dynamics of disassembly in vivo remain to be determined. It is also possible that the 0 layer residues are crucial for other protein-protein interactions occurring post-fusion.

The stability of the SNARE complex in vivo and the extent to which it exists in a stable confirmation, resembling that in the crystal structure, before or during exocytosis is not possible to assess. The only way to analyse SNARE complex formation is after detergent solubilisation, and it is possible that the complex could fall into the stable complex only after solubilisation. Discrimination between cis complexes on the same membrane and the trans complexes that mediate fusion is also not possible. It is unclear, therefore, whether the structure that drives bilayer fusion is equivalent to the stable complex that assembles in vitro, but we have attempted to address this question using high-resolution analysis of exocytosis. The toxin-resistant SNAP-25 (Em2) that we used already has mutations in conserved residues including substitution of glutamate for isoleucine in the conserved hydrophobic layer 2 (Fig. 1). These mutations impair the binary interaction of Em2 SNAP-25 with VAMP, although it can still associate into a complex with VAMP and syntaxin (Washbourne et al., 1999). Mutations within the hydrophobic layers of either the N- or C-terminal helices of SNAP-25 that are close to the 0 layer reduced thermal stability of the SNARE complex (Chen et al., 1999) and loss of function in the yeast homologue sec9 (Rossi et al., 1997). Despite a potential effect on SNARE interactions, and probably on stability, exocytosis supported by Em2 was indistinguishable from that due to endogenous wild-type SNAP-25. The fact that additional mutations in residues within the 0 layer, including potentially disruptive substitutions such as the introduction of two charged residues in the double Q/E mutant, can also be tolerated and that exocytosis proceeds with normal kinetics argues that the stable SNARE complex may not be the structure that drives fusion.

Various experimental data have suggested the existence of multiple states of SNARE complex assembly in addition to distinct cis- and trans-complexes (Fiebig et al., 1999; Hua and Charlton, 1999; Weber et al., 2000; Xu et al., 1998; Xu et al., 1999b). A comparison of the effects of two Clostridial neurotoxins, BoNT/D and tetanus toxin, that cleave VAMP has suggested the existence of a SNARE complex prior to fusion of synaptic vesicles in which the N-terminal but not the C-terminal domain of VAMP is shielded (Hua and Charlton, 1999). This partial neurotoxin-sensitive complex would be distinct from the neurotoxin-resistant tight complex postulated to occur in chromaffin cells (Xu et al., 1998; Xu et al., 1999b) leading to the possibility of the existence of a series of distinct quasi-stable SNARE complexes. Putting this information together with the data in the present paper leads us to suggest a model in which fusion involves the formation of a non-fully zipped SNARE complex that can be reversible (Fig. 7). In the in vitro liposome fusion assay the SNARE complex formed prior to fusion is resistant to α-SNAP and NSF (Weber et al., 2000). As suggested, this could be due to steric hindrance of α-SNAP and NSF binding or, alternatively, the SNARE complex that mediates fusion may be unable to recruit α-SNAP and NSF until functional 0 layer interactions occur to allow the formation of the stable and now NSF-sensitive complex. Our model shown in Fig. 7 has the merit that it would be consistent with reversible fusion pore opening and closure as demonstrated in biophysical studies of regulated exocytosis and the possibility that fusion pore dynamics is subject to physiological regulation.

This work was supported by grants from the Wellcome Trust to R.D.B.