SNARE Complex Zero Layer Residues Are Not Critical for N-Ethylmaleimide-sensitive Factor-mediated Disassembly*

Membrane-anchored SNAREs assemble into SNARE complexes that bring membranes together to promote fusion. SNARE complexes are parallel four-helix bundles stabilized in part by hydrophobic interactions within their core. At the center of SNARE complexes is a distinctive zero layer that consists of one arginine and three glutamines. This zero layer is thought to play a special role in the biology of the SNARE complex. One proposal is that the polar residues of the zero layer enable N-ethylmaleimide-sensitive factor (NSF)-mediated SNARE complex disassembly. Here, we studied the effects of manipulating the zero layer of the well studied synaptic SNARE complex in vitro and in vivo. Using a fluorescence-based assay to follow SNARE complex disassembly in real time, we found that the maximal rate at which NSF disassembles complexes was unaffected by mutations in the zero layer, including single replacement of the syntaxin glutamine with arginine as well as multiple replacement of all four layer residues with non-polar amino acids. To determine whether syntaxin with arginine instead of glutamine in its zero layer can support SNARE function in vivo, we introduced it as a transgene into a Caenorhabditis elegans syntaxin-null strain. Mutant syntaxin rescued viability and locomotory defects similarly to wild-type syntaxin, demonstrating that SNARE complexes with two glutamines and two arginines in the zero layer can support neurotransmission. These findings show that residues of the zero layer do not play an essential role in NSF-mediated disassembly.

Intracellular membrane trafficking depends on the ordered formation and consumption of transport intermediates and requires that membranes fuse with each other in a tightly regulated and highly specific manner. Membrane-anchored helical proteins known as SNAREs 2 have emerged as central players in membrane fusion. SNAREs assemble into coiled coil-like complexes that are thought to pull membranes together and thereby promote fusion (1)(2)(3)(4). After fusion, SNARE complexes must be disassembled to regenerate individual SNAREs for use in subsequent fusion reactions. Proteins responsible for disassembly are the AAAϩ (ATPase associated with a variety of cellular activities) protein NSF and its adaptor protein ␣-SNAP (5)(6)(7). Because of the critical role that SNARE complexes play in membrane fusion, understanding how their assembly and disassembly are controlled is critical to describing how cellular membrane trafficking is regulated.
SNAREs are defined by the presence of one or more copies of a conserved 60 -70-amino acid "SNARE motif" (3,8,9). SNARE motifs are usually adjacent to membrane-anchoring sequences and spontaneously assemble into coiled coil-like structures containing four helices aligned in parallel with each other (10 -12). Functional SNARE complexes consist of one each of four distinct groups of SNAREs defined by sequencebased profiling (9). Crystal structures of two SNARE complexes reveal characteristic packing in the mostly hydrophobic core of the complex (11,12). The core consists of 16 layers of interacting residues. At the midpoint is a unique layer containing polar rather than hydrophobic residues, typically one arginine and three glutamines. This layer is referred to as the zero layer. Its residues are hydrogen-bonded to each other and sequestered from solvent by hydrophobic layers on either side. The zero layer is the most distinctive feature common to all SNARE complexes and has therefore attracted much attention. It provides the basis for naming the four groups of SNARE motifs as Q A -, Q B -, Q C -, and R-SNAREs to reflect both the residue that each SNARE contributes to the zero layer and the position that it occupies in an assembled complex (8,9,13). The founding members of each of these groups are the SNAREs of the synaptic SNARE complex, with syntaxin-1 providing the Q A helix; the N-and C-terminal ends of SNAP-25 providing the Q B and Q C helices, respectively; and synaptobrevin (vesicle-associated membrane protein) providing the R helix.
Various ideas for how polar zero layer residues might impact SNARE function have been put forward. One attractive possibility is that favorable interactions among the polar residues within the layer ensure that SNARE motifs assemble in register with respect to each other and the membranes in which they are anchored (11,14). The zero layer may also help to specify the oligomeric state of the complex. It could provide a "half-way" point during complex assembly, separating formation of "trans-SNARE complexes" between membranes into two phases (15). Finally, the zero layer might serve as a starting point for NSF/␣-SNAPdriven complex disassembly (11,14,16). Whether one or several of these explain the high degree of conservation of the layer remains unknown.
A number of mutagenesis studies have investigated how manipulating zero layer residues affects SNARE-dependent processes in vivo using the yeast Saccharomyces cerevisiae as a model system (17)(18)(19)(20)(21). A general conclusion from these experiments is that trying to accommodate a second arginine in the zero layer perturbs SNARE complex function, whereas eliminating all arginines is less disruptive. The disruption caused by a second arginine can be overcome by changing the original arginine to glutamine (17)(18)(19). This demonstrates that the number (one versus two) rather than the absolute position of the arginine is what matters for complex function. Whether the overall presence of a polar zero layer is functionally important has not been examined in vivo because all mutant complexes studied to date in yeast retain at least a partially polar zero layer.
Contributions of individual core residues, including those of the zero layer, to the function of the synaptic SNARE complex have been extensively examined using in vitro assays of PC12 cell secretory granule exocytosis (22)(23)(24) as well as in more limited ways in a variety of living preparations (25)(26)(27)(28)(29). For in vitro exocytosis studies, analysis has been limited to changes in the SNAP-25 Q B and Q C helices because these do not need to be anchored to the membrane to function. Not surprisingly, mutations of hydrophobic core residues in either the Q B or Q C helix disrupt the apparent stability of the SNARE complex as well as its ability to promote exocytosis (22)(23)(24). When examined in living preparations, mutations of hydrophobic layer residues in syntaxin or synaptobrevin have been similarly disruptive (25)(26)(27). Somewhat surprisingly, changes in the residue(s) that SNAP-25 brings to the zero layer have relatively minor effects on in vitro secretory granule exocytosis, particularly when compared with the effects of mutations in adjacent hydrophobic layers that have similar effects on apparent complex stability (22)(23)(24). SNAP-25 zero layer mutants introduced into living cells again have little or no effect on fusion per se (28,29), although expressing SNAP-25 with both its glutamines changed to leucine slows the kinetics of vesicle pool refilling in adrenal chromaffin cells (29).
Because of the minimal effects of synaptic SNARE complex zero layer mutations on secretory granule exocytosis, Scheller and co-workers (16) investigated whether these changes might instead affect a SNARE reaction not directly involved in fusion such as NSF-mediated complex disassembly. They found that disassembly appeared to be inhibited by some zero layer mutations, with the most dramatic effects seen when the Q A (syntaxin) zero layer glutamine was changed to any other amino acid (16). Neher and co-workers (29) considered slowing of disassembly as a possible explanation for the delay in pool replenishment caused by expressing the SNAP-25 zero layer leucine mutant in adrenal chromaffin cells, but dismissed it based on unpublished data showing that a SNARE complex with four leucines in its zero layer was disassembled in vitro as fast or faster than the wild-type complex. Disassembly of SNARE complexes in yeast expressing zero layer mutants has not been directly examined, but the range of phenotypes observed (17)(18)(19)(20)(21) and, in particular, the ability to restore function by rotating the arginine within the layer indicate that a general requirement for a Q A glutamine in disassembly is unlikely.
To understand more specifically how NSF disassembles SNARE complexes, we were interested in further exploring the proposed role of the zero layer in disassembly of the synaptic SNARE complex. Here, we revisited the question of how manipulating the zero layer affects complex disassembly in vitro and also investigated how a zero layer mutation in syntaxin affects the normal function of this Q A -SNARE in vivo in Caenorhabditis elegans. We found that changes in the zero layer, either single replacement of the syntaxin glutamine with arginine or multiple replacement of all four layer residues with non-polar amino acids, did not significantly affect NSF-mediated disassembly in vitro. Furthermore, replacing the syntaxin glutamine with arginine did not impair syntaxin function in C. elegans. This suggests that, at least in the case of the synaptic SNARE complex, there is no special role for polar zero layer residues and, in particular, for the syntaxin glutamine in NSF-mediated disassembly. Our results confirm that the function of the synaptic SNARE complex is remarkably tolerant to changes in its zero layer and open the way to further studies aimed at defining the contribution of this layer to the normal functioning of SNARE complexes.

EXPERIMENTAL PROCEDURES
Protein Expression Vectors-pET28-CFP and pET28-YFP vectors were created by moving enhanced CFP or YFP (Invitrogen) from pECFP-N1 or pEYFP-N1, respectively, as a BamHI-NotI fragment into pET28a. His 6 -syntaxin-CFP was created by introducing residues 1-265 of rat syntaxin-1a as a NdeI-AgeI fragment into pET28a-CFP. The resulting construct contains a His 6 tag, syntaxin-1a, and CFP with residues VPVAT between syntaxin and CFP. The syntaxin also contains a C145S mutation. His 6 -synaptobrevin-YFP was similarly constructed by introducing residues 1-96 of rat synaptobrevin II into the NdeI-BamHI fragment of pET28a-YFP with residues VDPPVAT between synaptobrevin and YFP. All fragments were generated by PCR using oligonucleotide primers containing the indicated restriction sites. The sequences of all constructs were confirmed by ABI PRISM BigDye terminator cycle sequencing (Applied Biosystems) at the Protein and Nucleic Acid Chemistry Laboratory of the Washington University School of Medicine. SNAP-25b in pHO4d, syntaxin-(1-265) in pHO4d, and synaptobrevin-(1-96) in pET15b were as described previously (10). Mutations in the SNARE zero layers (syntaxin(Q226R), syntaxin(Q226L), SNAP-25(Q53A/Q174L), synaptobrevin(R56Q), and synaptobrevin(R56A)) were introduced by QuikChange site-directed mutagenesis (Stratagene) into the indicated plasmids. For expression in C. elegans, a Q227R mutation was similarly introduced into pTX12, an unc-64 syntaxin minigene expressed under the control of the unc-64 promoter (25).
An expression vector for NSF was created by amplifying full-length Chinese hamster NSF from pQE9-NSF (5) and inserting it as a BamHI-XhoI fragment into pET28a. The resulting protein has N-terminal His 6 and T7 tags. Bovine ␣-SNAP in pET28a was as described (30).
Protein Purification and Complex Preparation-Proteins were purified essentially as described (10,30). NSF, syntaxin, SNAP-25, and synaptobrevin were quantified using Bradford reagent (Bio-Rad) with bovine serum albumin as a standard. ␣-SNAP was quantified by measuring absorbance at 280 nm using ⑀ ϭ 40,200 (30). All proteins and complexes were flash-frozen in liquid nitrogen and stored at Ϫ80°C.
A SNARE complex was assembled by combining syntaxin-CFP, SNAP-25, and synaptobrevin-YFP (8 -10 M final concentration of each); dialyzing overnight into 20 mM Tris (pH 7.4), 100 mM NaCl, and 1 mM dithiothreitol; and then separating the mixture on a Mono Q anion exchange column. Fractions containing the highest concentrations of SDS-resistant complex relative to unincorporated individual proteins were pooled, aliquoted, and stored as described above. At least three independent preparations of each complex were purified and studied.
Quick-freeze Deep-etch Electron Microscopy-SNARE complexes adsorbed to mica chips were replicated and viewed by transmission electron microscopy as described previously (10).
Temperature-dependent Dissociation in SDS-The SNARE complex (30 g) was brought to a final volume of 150 l in 1ϫ Laemmli buffer (50 mM Tris (pH 6.8), 1% SDS, 10% glycerol, and 0.1% bromphenol blue). Aliquots (15 l) were incubated at the indicated temperatures for 5 min, returned to room temperature, and loaded onto a 10% SDS-polyacrylamide gel. Proteins in the gel were stained with SYPRO Red and visualized and quantified using a STORM 860 imaging system (30).
SNARE Complex Disassembly Reactions-Disassembly reactions were performed in 30 mM HEPES (pH 7.6), 100 mM potassium glutamate, 10 mM MgCl 2 , 2 mM ATP, and 1 mg/ml bovine serum albumin supplemented where indicated with an ATP-regenerating system of 1 mM creatine phosphate and 20 g/ml creatine phosphokinase. Bovine serum albumin was omitted from disassembly reactions analyzed by gel electrophoresis. Unless indicated otherwise, all reactions were performed at 30°C.
For FRET analysis, disassembly reactions were carried out in a final volume of 500 l. Buffer, SNARE complex (100 nM unless indicated otherwise), and ␣-SNAP (2 M or as indicated) were premixed and transferred into a 10 ϫ 2-mm quartz cuvette. An initial emission scan was acquired ( ex ϭ 434 nm and em ϭ 450 -600 nm with slits at 3 nm and measurements every 1 nm), followed by two kinetic scans ( ex ϭ 434 nm and em ϭ 473 and 523 nm in separate photomultiplier tubes). The first scan was used to confirm a stable fluorescence base line. In the second scan, NSF 6 (1-10 nM or 0.25-2.5 g) was added after 60 s, and the reaction was followed for another 500 s. Fluorescence measurements were performed using a Spex FluoroLog-3 operating in T-format.
For monitoring disassembly of SDS-resistant complexes by gel electrophoresis, ␣-SNAP was premixed with the SNARE complex (100 -800 nM, 150-l reaction volume) prior to initiating disassembly with NSF. Aliquots were removed at the indicated times and quenched by addition of 5 mM EDTA and Laemmli buffer. Samples were immediately separated on a 10% SDS-polyacrylamide gel. Protein was visualized and quantified by staining with SYPRO Red as described above.
Data Analysis-All scans were corrected for buffer background. Apparent FRET efficiency was calculated from corrected CFP emission (477 nm) as 1 Ϫ (F DA /F D ), where F DA is the measured CFP fluorescence of the complex at a given time and F D is the CFP fluorescence of the fully disassembled complex. The FRET signal measured for each complex was taken to represent 100% assembled complex, whereas that for the fully disassembled complex was set to 0%. The initial rate of complex disassembly was calculated from the slope of a line fit to the first 20 s of data collected after addition of NSF (data rejected if the line had R 2 Ͻ 0.98), taking into account the amount of SNARE complex (in pmol) and NSF (in g) in the reaction. Rate data for disassembly at different ␣-SNAP concentrations (see Fig. 6) were fit directly to a modified Hill equation , assuming equivalent binding sites for ␣-SNAP on the SNARE complex. Curve fitting and plotting were carried out using KaleidaGraph (Synergy Software, Reading, PA) and GraphPad Prism (GraphPad Software, San Diego, CA).
Construction of Syntaxin(Q227R) Transgenic Strains-C. elegans strains were transformed using a standard microinjection protocol (32). Plasmid pPD118.33 (a gift of Andy Fire) expressing green fluorescent protein in the pharynx under the control of the myo-2 promoter was used as a cotransformation marker. Transgenic strains containing unc-64(js115) in the homozygous state were created by injecting unc-64(js115)/bli-5(e518) heterozygotes and screened for fluorescent transgenic animals that failed to segregate the bli-5 chromosome. Progeny of the resulting animals consisted only of green fluorescent protein-positive viable animals and green fluorescent protein-negative Unc-64 L1-arrested paralyzed animals. Three independent lines were analyzed for each transgene.

Real-time Assay of NSF-mediated SNARE Complex Disassembly-
To study the assembly and disassembly of SNARE complexes in vitro, we created a reporter system in which fluorescent proteins (enhanced CFP and YFP) are fused to the C termini of the cytoplasmic domains of rat syntaxin-1a and synaptobrevin II in place of the proteins' usual transmembrane domains (Fig. 1A). When the SNAREs assemble into a SNARE complex, they do so in a parallel orientation and thus bring the fluorescent protein tags close enough together for FRET between CFP and YFP. Advantages to using genetically fused fluorescent protein tags to follow SNARE complex dynamics are that (i) all SNAREs are uniformly tagged; (ii) the distance between the tags in assembled SNARE complexes is likely to be optimal for FRET; (iii) fusing fluorescent protein tags to SNARE proteins improves the yield of recombinant protein obtained from Escherichia coli; and (iv) identical tags can be fused to different SNARE proteins, making it possible to directly compare the in vitro behavior of SNARE complexes. Two early studies using blue fluorescent protein and YFP fused to syntaxin and synaptobrevin to test their relative orientation within a SNARE complex demonstrated that, in principle, this approach will work (37, 38). To generate a tagged SNARE complex, we combined equimolar amounts of syntaxin-CFP, synaptobrevin-YFP, and non-fluorescent SNAP-25. Following overnight incubation, we separated the SNARE complex from unincorporated component proteins by ion exchange chromatography. The resulting complex contained the three SNAREs in an ϳ1:1:1 molar ratio and was resistant to dissociation by SDS unless boiled (Fig. 1B), as is typical of the synaptic SNARE complex (39). To confirm that the complexes were properly assembled, we carried out quick-freeze deep-etch electron microscopy and looked at the morphology of individual particles (Fig. 1C). Similar to what we reported in a previous study of SNARE complexes tagged with maltose-binding protein (10), two protein tags (in this case, CFP or YFP, which appeared as ϳ4-nm diameter balls) were present where expected on one end of each ϳ12-14-nm rod-like SNARE complex.
We next compared the fluorescence emission spectra of an assembled SNARE complex ( Fig. 2A, solid line), a SNARE complex disassembled by NSF/␣-SNAP (dashed line), and an unassembled mock SNARE complex prepared without SNAP-25 (dotted line). CFP emission ( max ϭ 477 nm) was reduced by ϳ50% in SNARE complex samples compared with samples containing either disassembled or unassembled SNAREs, whereas YFP emission ( max ϭ 527 nm) was correspondingly increased. These changes are consistent with significant FRET between CFP and YFP when the SNAREs to which they are attached are assembled in complexes. The apparent FRET efficiency in SNARE complexes ranged from 45 to 52% (see "Experimental Procedures" for calculation). Brief treatment with trypsin had the same effect as disassembling the complex with NSF, confirming that FRET arises from the constrained proximity of CFP and YFP (data not shown). No FRET was seen when complexes containing only one fluorescent protein each were mixed (SNARE complex with only syntaxin-CFP combined with SNARE complex with only synaptobrevin-YFP), further demonstrating that FRET was due to interactions within but not between complexes (data not shown).
To follow SNARE complex disassembly in real time, we monitored SNARE complex fluorescence as a function of time. Different amounts of NSF added to the SNARE complex (100 nM) in the presence of saturating concentrations of ␣-SNAP and MgATP had the effects shown in Fig. 2B. FRET decreased over time, eventually reaching a plateau near the point of full complex disassembly. At low complex concentrations (Ͻ500 nM), disassembled SNAREs did not reassemble, and the reaction was effectively irreversible. The initial rate of complex disassembly under these conditions was 18.6 Ϯ 2.1 pmol of SNARE complex⅐g of NSF Ϫ1 ⅐min Ϫ1 (n ϭ 25) at 30°C or ϳ0.2 SNARE complexes/s/NSF Ϫ1 hexamer. Leaving out NSF, ␣-SNAP, Mg 2ϩ , or ATP blocked changes in SNARE complex fluorescence, as expected ( Fig. 2B) (data not shown). The initial reaction rate was constant over a range of added NSF 6 from 1 to 10 nM (Fig. 2B) (data not shown). The rate of disassembly varied with temperature, increasing from 14.5 Ϯ 1.5 pmol of SNARE complex⅐g of NSF Ϫ1 ⅐min Ϫ1 at 25°C to 23.6 Ϯ 2.6 at 35°C. Although 100 nM SNARE complex may not represent a saturating concentration of substrate, the rates seen here are similar to those estimated previously for NSF-mediated disassembly of membrane-anchored cis-SNARE complexes (40).
Effect of a Two-arginine/Two-glutamine Zero Layer on NSF-mediated Disassembly-To understand how NSF and ␣-SNAP disassemble SNARE complexes, we need to define features of the SNARE complex important for its efficient disassembly. The proposal that interactions between the disassembling machinery and polar residues in the zero layer might be involved in early stages of the reaction was attractive (11,14,16,41), and we initially set out to further explore this role of the zero layer residues. We began by replacing the glutamine at position 226 in syntaxin-CFP with arginine to create a complex in which the zero layer contained two glutamines and two arginines. This mutation did not affect the yield of SDS-resistant complex (Fig. 3A). On the basis of the crystal structures of SNARE complexes (11,12,14) and molecular dynamics simulations of the effects of zero layer mutations on the structure of the synaptic SNARE complex (17), it is clear that the zero layer cannot accommodate two arginines without some distortion. Consistent with this, a previous study of zero layer mutations in the synaptic SNARE complex found that the syntaxin(Q226R) mutation decreased the apparent stability of the complex, with an ϳ20°C reduction in the melting temperature measured by circular dichroism and an ϳ10°C reduction in the temperature needed to dissociate the complex in 1% SDS (16). To confirm that the CFP/YFP-tagged complex was similarly affected by a syntaxin(Q226R) mutation, we incubated the wild-type or mutant complex in 1% SDS at varying temperatures prior to electrophoresis and measured the amount of remaining SDS-resistant complex (Fig. 3B). As expected, the complex containing syntaxin(Q226R)-CFP was half-dissociated at an ϳ10°C lower temperature compared with the wild-type complex. Because of the hysteresis observed in studies of SNARE unfolding and refolding (42), it is not clear whether this change in apparent stability reflects a change in the kinetics of complex dissociation or in the free energy of the complex itself, but the change confirms that a structural change has occurred.
On the basis of the previous study of synaptic SNARE complexes with zero layer mutations (16), we expected that changing the syntaxin glutamine to an arginine (Q226R), despite decreasing the apparent stability of the complex, would slow the rate at which NSF disassembles the complex. We again monitored complex disassembly using FRET between attached CFP and YFP tags. Surprisingly, FRET was completely lost within 10 min of addition of NSF just as for the wild-type B, resistance to dissociation in SDS as a function of temperature. SNARE complexes containing wild-type syntaxin-CFP (f) or syntaxin(Q226R)-CFP (‚) were incubated in 1% SDS at the indicated temperatures for 5 min prior to electrophoresis. Gels were stained with SYPRO Red, and the amount of complex remaining in the high molecular mass complex was quantified using a STORM 860 imaging system. The means Ϯ S.E. of three experiments for each complex are shown. The temperatures at which 50% of the complex was dissociated were 68°C for the wild-type complex and 58°C for the mutant complex. The arrow shows the point at which NSF (1 g) was added to SNAP/SNARE complex mixture. Apparent FRET efficiency was calculated from CFP emission as described under "Experimental Procedures." The initial rate of disassembly was 17.7 Ϯ 0.8 pmol of SNARE complex⅐g of NSF Ϫ1 ⅐min Ϫ1 . complex (Fig. 4A). Kinetic analysis of the loss in FRET revealed that the complex containing syntaxin(Q226R) was disassembled with an initial rate of 17.5 Ϯ 1.0 pmol of SNARE complex⅐g of NSF Ϫ1 ⅐min Ϫ1 (n ϭ 8) at 30°C, which is statistically indistinguishable from the rate at which the wild-type complex was disassembled. The ability of NSF to disassemble the synaptic SNARE complex thus does not appear to require special interactions with the syntaxin zero layer glutamine.
Because of the discrepancy between these findings and the earlier work of Scales et al. (16), we carried out disassembly experiments using a gel-based assay of SDS-resistant complexes to follow the reaction (Fig.  5). We added NSF together with ␣-SNAP and MgATP to either the wildtype or Q226R mutant SNARE complex and measured the SDS-resistant SNARE complex remaining at different times during the course of a disassembly reaction. To do this, we removed aliquots of the complex where indicated, mixed them into SDS-containing sample buffer, and separated them without boiling on an SDS-polyacrylamide gel (Fig. 5). Once again, the wild-type (three-glutamine/one-arginine zero layer) and mutant (twoglutamine/two-arginine zero layer) complexes were both efficiently disassembled, confirming that SNARE complex disassembly is unaffected by eliminating glutamine from the syntaxin zero layer.

Two-arginine Zero Layer Causes a Small Increase in [␣-SNAP]
Needed for Disassembly-As described above, it is clear from both structural simulations and experimental data that the two-glutamine/two-arginine zero layer distorts SNARE complex structure (Fig. 3B) (16,17). To promote disassembly, up to three molecules of ␣-SNAP recognize and bind to a SNARE complex through interactions that depend on complementarity in both charge and shape (30,43). Because of this, we wondered if there might be differences in the amount of ␣-SNAP needed to disassemble wild-type versus zero layer mutant complexes. As shown in Fig. 6, requirements for ␣-SNAP in disassembly of the two complexes were similar, with doseresponse curves for both showing a small degree of positive cooperativity (apparent Hill coefficient of 1.3). More ␣-SNAP was needed to promote half-maximal disassembly of the mutant complex compared with the wildtype complex (wild-type complex K 0.5 ϭ 0.22 Ϯ 0.07 M and two-glutamine/two-arginine mutant complex K 0.5 ϭ 0.50 Ϯ 0.12 M), consistent with a small decrease in the affinity of the two-glutamine/two-arginine mutant complex for ␣-SNAP. This (together with Fig. 3B) confirms that the Q226R mutation affects the structure of the SNARE complex, but not its overall susceptibility to NSF-mediated disassembly. These findings also reinforce the essential role of ␣-SNAP in selecting SNARE complexes for disassembly.
Hydrophobic Repacking of the Zero Layer Does Not Affect Disassembly-If the syntaxin zero layer glutamine is not essential for efficient SNARE complex disassembly, we wondered whether the presence of a polar layer near the center of the complex might nonetheless be important for the reaction. We therefore repacked the zero layer of the synaptic SNARE complex with non-polar residues and studied disassembly of the resulting complex. The SNARE complex containing syntaxin(Q226L)-CFP, SNAP-25(Q53A/Q174L), and synaptobrevin(R56A)-YFP had a fluorescence emission profile similar to that of the wild-type SNARE complex (Fig. 7A). FRET was once again completely lost following a 10-min incubation with NSF. The progress of this reaction was similar to that for the wild-type complex (Fig. 7B).
Syntaxin Zero Layer Mutant Restores Function in a C. elegans Syntaxin-null Strain-To determine whether the three-glutamine/onearginine composition of the zero layer is important for function of the synaptic SNARE complex in neurotransmitter release, we took advantage of the high degree of conservation between rat and C. elegans SNAREs (25,26) and the ability to rescue mutant strains of C. elegans with appropriate transgenes. We investigated whether syntaxin (UNC-64) containing arginine in place of its usual zero layer glutamine (Q227R) rescues movement in a strain lacking UNC-64 syntaxin (25). Worms expressing comparable levels of wild-type or Q227R mutant UNC-64 from transgenes were generated (Fig. 8A). The overall movements of parental, wild-type, and mutant transgenic strains were indistinguishable from each other (Fig. 8B), showing that the overall func-  tion of the synaptic SNARE complex was restored by the UNC-64 syntaxin(Q227R) mutation. This implies that synaptic SNARE complexes containing two arginines can promote membrane fusion in vivo.
To look for more subtle impairment of neurotransmission in worms expressing this mutant syntaxin, we measured the sensitivity of various strains to the acetylcholinesterase inhibitor aldicarb. Mutant worms that release less acetylcholine than normal are typically resistant to aldicarb, whereas those that release more are hypersensitive (35, 44 -46). Disruptive mutations of hydrophobic layer residues in the SNARE motifs of syntaxin (unc-64) and synaptobrevin (snb-1) have been shown previously to generate worms resistant to aldicarb, whereas a syntaxin mutation that is "open" and thus more readily incorporated into SNARE complexes creates worms with a hypersensitive phenotype (25,26,47). Eliminating tomosyn, a proposed negative regulator of SNARE function, also leads to hypersensitivity (46,48). As expected, mutant syntaxin worms expressing wild-type UNC-64 responded to aldicarb similarly to the parental strain (N2), whereas known syntaxin-deficient worms were resistant (Fig. 8C). Surprisingly, mutant worms rescued with transgenic Q227R mutant UNC-64 were paralyzed sooner than those expressing wild-type syntaxin, indicating a hypersensitivity to aldicarb. Although the basis for this hypersensitivity merits further exploration, the fact that worms expressing the syntaxin zero layer mutant are not resistant to aldicarb confirms that synaptic SNARE complexes with a zero layer of two glutamines and two arginines are functional in C. elegans.

DISCUSSION
Although the SNARE complex conserved zero layer has attracted much attention, its actual function remains unknown. As part of an effort to understand how NSF disassembles SNARE complexes, we tested the hypothesis that polar zero layer residues and, in particular, the glutamine in the Q A -SNARE syntaxin are important for efficient SNARE complex disassembly (11,16). Surprisingly, we found that changing the composition of the zero layer had no significant effect on the rate at which synaptic SNARE complexes could be disassembled by NSF. Furthermore, syntaxin containing arginine instead of glutamine in its zero layer position rescued viability and movement in a C. elegans syntaxin-null strain, demonstrating in a living animal that the zero layer of this complex is more tolerant to change than are some of the surrounding hydrophobic layers (25,26). Together with previous studies, our findings suggest that the true function(s) of the zero layer remain to be discovered. Based on the literature and our in vivo findings, it seems likely that the importance of the zero layer will vary from complex to complex. The lack of involvement of the zero layer in NSFmediated disassembly draws attention to other features of the SNARE complex that NSF might utilize in initiating its disassembly reaction, as discussed further below.
Where did the idea that disassembly depends on polar residues in the core of the zero layer come from? It was first proposed as a possible explanation for the layer's high degree of conservation and for the seemingly unique sequestration of the layer's polar residues from surrounding solvent (11). It re-emerged in efforts to explain the weak effects of zero layer mutations in SNAP-25 SNARE motifs on exocytosis (22,23,28,29). Scales et al. (16) further developed the hypothesis in a study examining the effects of zero layer mutations on features of the SNARE complex cycle not directly linked to membrane fusion, including overall stability and NSF-mediated disassembly. Like us, they found that changes in the zero layer reduced the apparent stability of the synaptic SNARE complex. However, their primary conclusion that changes in the zero layer and especially in the glutamine of syntaxin severely impair NSF-mediated SNARE complex disassembly differs from our results. We found that SNARE complexes containing arginine instead of glutamine in the syntaxin zero layer position were readily disassembled under a variety of assay conditions using two distinct reporter systems. Furthermore, we were able to show that syntaxin with this potentially perturbative zero layer mutation remained functional in living worms. This latter result is consistent with the relatively benign effects previously associated with mutations of SNAP-25 zero layer residues in semiintact and intact cell systems (22,23,28,29) and seems to be inconsistent with a major change in the ability of NSF to disassemble SNARE complexes with mutant zero layers.
Although there are a number of minor differences in the recombinant proteins and assays used here versus those used by Scales et al. (16), none clearly explain the different results obtained. One way in which changes in the SNARE complex can inhibit NSF-mediated disassembly is to decrease ␣-SNAP binding. As shown in Fig. 6, the one difference we found in disassembly of wild-type and zero layer mutant SNARE complexes was a small shift (K 0.5 ϭ 0.2 M to K 0.5 ϭ 0.5 M) in the amount of ␣-SNAP needed to reach the half-maximal disassembly rate. This shift is readily explained by the likely effects of changes in the shape of the SNARE complex around the zero layer upon ␣-SNAP binding (17,30,43). If active ␣-SNAP was limiting in the assays of Scales et al. (16), then changes in the affinity of different mutant complexes for ␣-SNAP could explain some of the reported effects of zero layer mutations on disassembly. A more recent study in which arginine in the synaptobrevin zero layer was replaced with a highly disruptive proline residue found that cells expressing this mutant accumulated SNARE complexes in their plasma membrane (41). Although the effects of this mutation on ␣-SNAP binding were not examined, our studies of ␣-SNAP binding to the SNARE complex predict that such a disruptive mutation will impair ␣-SNAP binding and thus slow disassembly at physiologic concentrations of ␣-SNAP (30). Given the high degree of conservation of the zero layer and the common ability of NSF/␣-SNAP to disassemble all SNARE complexes, it seems likely that any essential role for residues of the zero layer in NSF-driven disassembly would apply to more than just the synaptic SNARE complex. Studies examining the effects of zero layer mutations in other SNARE complexes have not precisely addressed this question, but they are at the least inconsistent with an obligate role for the Q A glutamine in ongoing complex function. In particular, whereas yeast exocytic and endoplasmic reticulum/ Golgi complexes containing two-glutamine/two-arginine zero layers were found to be seriously impaired or nonfunctional in vivo, their function could be restored by replacing the arginine in the original R-SNARE with glutamine to create a rotated three-glutamine/one-arginine layer (17)(18)(19). Because none of the SNARE complexes so far studied in yeast have the convenient property of resisting dissociation by SDS, the precise quantita- A, shown are the results from immunoblot analysis of syntaxin, synaptobrevin (vesicle-associated membrane protein (VAMP)), and synaptotagmin in whole worm extracts from transgenic animals expressing UNC-64 wild-type (wt) syntaxin or syntaxin(Q227R) as well as control strains. Three independent syntaxin(Q227R) transgenic strains (strains 1-3) and one wild-type syntaxin transgenic strain all had comparable expression of syntaxin. The unc-64(md130) splice lesion reduces the levels of syntaxin by ϳ10 -20-fold (51); snb-1(md247) alters the mobility of SNB-1 synaptobrevin (26); and snt-1(md290) deletes most of the synaptotagmin coding region (34). All confirm antibody specificity. N2 is the parental strain. B, the thrashing behavior of transgenic animals expressing UNC-64 wild-type syntaxin and syntaxin(Q227R) (strain 1) was comparable with that of the control strain. Adults were used for analysis, except in the case of the null worms, which arrest as L1 larvae (n ϭ 10). C, shown is the effect of 0.5 mM aldicarb on transgenic animals expressing UNC-64 wild-type syntaxin and syntaxin(Q227R) and on appropriate control animals. unc-64(e246) expresses syntaxin with a lesion in the hydrophobic face of the H3 domain that reduces evoked transmitter release by ϳ20-fold (36); locomotion of these animals was disrupted much more slowly than that of control animals. In contrast, locomotion of worms expressing the Q227R transgene was disrupted faster than that of control worms. The means Ϯ S.E. of three trials each using 25 animals are shown.
tion of the steady-state levels of the SNARE complex will require the development of better methodology.
If the zero layer is not conserved to facilitate disassembly of complexes by NSF, what role(s) does it play? Among the more obvious possibilities that have been considered are that (a) it helps SNAREs assemble with each other in the proper register and orientation to avoid off-pathway assembly intermediates; (b) it contributes to establishing the desired (four-helix) oligomeric state; and (c) it provides an intermediate "stopping point" at the stage at which complexes are half-way zippered up between membranes (partially assembled "trans-complexes"). This latter role would not be obvious in studies of solubilized SNAREs because half-assembled SNARE complexes are stable only when anchored in two membranes. Buried polar residues are known from studies of other coiled-coil systems to affect assembly kinetics as well as the orientation and oligomeric state of the resulting complexes, making all of these reasonable possibilities. Which of these or other roles the zero layer plays will require systematic study of SNARE complexes in which the zero layer polar residues are replaced with structurally appropriate hydrophobic residues.
Our finding that worms expressing the zero layer mutant syntaxin(Q227R) are hypersensitive to aldicarb provides evidence for a previously unappreciated role for the syntaxin zero layer glutamine in negatively regulating syntaxin function. Further studies will be needed to establish what this is, but three possibilities are that (a) the Q227R mutation could affect the ratio between open and "closed" syntaxin, thus promoting SNARE complex assembly; (b) a decrease in affinity for ␣-SNAP could prolong the lifetime of a fusion-promoting assembly intermediate; and (c) replacement of the syntaxin glutamine with arginine could selectively perturb interaction of syntaxin with possible negative regulators of exocytosis such as tomosyn (48,49). It will be important to consider effects such as these in future studies of how mutations in syntaxin affect its function.
If NSF and ␣-SNAP do not initiate disassembly from the zero layer, what do they disrupt to promote SNARE complex disassembly? The original goal of this study was to use the differences between disassembly of wild-type versus Q226R mutant complexes as a starting point for understanding how interactions between ␣-SNAP and SNAREs are modified by NSF to promote disassembly. With the finding that mutations in the zero layer have no effect on disassembly, we turn our attention to other features of the SNARE complex that ␣-SNAP and NSF might perturb. Our previous studies of the interaction between ␣-SNAP and the SNARE complex have defined a critical role for charged residues (30,50). Within the SNARE complex, interhelical interactions fall into two categories. First are the interactions within the core of the complex that are hydrophobic except for the hydrogen-bonded network of the zero layer. Second are a large number of interhelical hydrogen bonds and salt bridges between partially surface-exposed residues. These interactions extend the length of the complex and include approximately 24 interhelical salt bridges in the synaptic SNARE complex and somewhat fewer in other complexes (11,12). In two-stranded coiled coils, stability often depends on interhelical salt bridges. We propose that destabilizing salt bridges would be one way for NSF and ␣-SNAP to disassemble the SNARE complex. This could happen by using the charged SNARE-binding surface of ␣-SNAP to provide alternative salt bridge partners for residues of the SNARE complex involved in interhelical salt bridges. Conformational changes in NSF driven by ATP binding and/or hydrolysis could place ␣-SNAP in position to promote "salt bridge switching," leading to destabilization and disassembly of the SNARE complex. Insight into how efficiently this might happen will come from direct measurement of the amount of ATP needed to disas-semble a SNARE complex. Experiments to define the energy requirements of the reaction and to test this idea are currently in progress.