The Structure of the Yeast Plasma Membrane SNARE Complex Reveals Destabilizing Water-filled Cavities*

SNARE proteins form a complex that leads to membrane fusion between vesicles, organelles, and plasma membrane in all eukaryotic cells. We report the 1.7Å resolution structure of the SNARE complex that mediates exocytosis at the plasma membrane in the yeast Saccharomyces cerevisiae. Similar to its neuronal and endosomal homologues, the S. cerevisiae SNARE complex forms a parallel four-helix bundle in the center of which is an ionic layer. The S. cerevisiae SNARE complex exhibits increased helix bending near the ionic layer, contains water-filled cavities in the complex core, and exhibits reduced thermal stability relative to mammalian SNARE complexes. Mutagenesis experiments suggest that the water-filled cavities contribute to the lower stability of the S. cerevisiae complex.

SNARE proteins form a complex that leads to membrane fusion between vesicles, organelles, and plasma membrane in all eukaryotic cells. We report the 1.7 Å resolution structure of the SNARE complex that mediates exocytosis at the plasma membrane in the yeast Saccharomyces cerevisiae. Similar to its neuronal and endosomal homologues, the S. cerevisiae SNARE complex forms a parallel four-helix bundle in the center of which is an ionic layer. The S. cerevisiae SNARE complex exhibits increased helix bending near the ionic layer, contains waterfilled cavities in the complex core, and exhibits reduced thermal stability relative to mammalian SNARE complexes. Mutagenesis experiments suggest that the water-filled cavities contribute to the lower stability of the S. cerevisiae complex.
Membrane fusion is a fundamental and highly regulated process that is required for the transport of proteins, lipids, and metabolites in all eukaryotes. Highly conserved SNARE 2 (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) proteins play a key role in the fusion of a transport vesicle with its target membrane (1,2). Specific sets of SNAREs located on vesicle and target membranes form a complex that draws together SNARE transmembrane domains, leading to juxtaposition and fusion of the two lipid membranes. SNARE-mediated fusion processes are either constitutive or triggered, and they require additional SNARE-interacting factors, such as Munc13, Sec1/Munc 18-like proteins, synaptotagmins, and complexins (3,4).
Neuronal SNAREs play a key role in the fusion of synaptic vesicles with the plasma membrane, a process that is critical for neurotransmission (5). Synaptic vesicles dock at the plasma membrane and upon cell depolarization and Ca 2ϩ entry fuse with the plasma membrane, thus releasing neurotransmitters into the synaptic cleft. The neuronal SNARE complex is composed of synaptobrevin 2, which is primarily localized to synaptic vesicles, and syntaxin 1A and SNAP-25, which are primarily associated with the plasma membrane.
The family of yeast SNAREs mediates constitutive fusion between transport vesicles and intracellular organelles or the plasma membrane (6,7). Yeast has been a valuable system for studying membrane fusion and vesicular transport because of the ease of genetic and biochemical manipulations. Yeast is also one of the first organisms that evolved to utilize intracellular membrane fusion and therefore provides a valuable snapshot of the fusion machinery in lower eukaryotes. The Saccharomyces cerevisiae SNARE proteins involved in exocytosis at the plasma membrane (the synaptobrevin homologue Snc1p, the syntaxin homologue Sso1p, and the SNAP-25 homologue Sec9p) show assembly properties similar to their neuronal and endosomal homologues. Relative to the neuronal SNARE complex the S. cerevisiae SNARE complex exhibits decreased thermal stability, and prior to SNARE complex formation, Sso1p interacts with Sec9p with 1:1 stoichiometry (as opposed to a mixture of 1:1 and 1:2 for the neuronal SNAREs) (8,9).
Several structures of neuronal SNARE complexes and endosomal SNARE complexes have been solved from higher eukaryotes (10 -13); however, no structural information is available for lower eukaryotes. To dissect the differences between S. cerevisiae, neuronal, and endosomal SNARE complexes we determined the structure of the S. cerevisiae plasma membrane SNARE complex consisting of the cytoplasmic SNARE core domains at 1.7 Å resolution. Surprisingly, we found a larger number of buried water molecules compared with the neuronal SNARE complex. We then performed mutagenesis experiments to determine the influence of these water molecules on the stability of the complex.
For purification, cells were resuspended in lysis buffer (100 mM Tris, pH 8.0, 6 M GdHCl) and lysed by sonication. Lysate was incubated in the lysis buffer for 1 h and then cleared by ultracentrifugation before incubation with nickel-nitrilotriacetic acid resin for 4 h. The resin was washed with 250 ml of lysis buffer, and proteins were eluted in 50 mM Tris, pH 8.0, 150 mM NaCl, and 200 mM GdHCl. Ssop1 protein was prone to aggregation and thus the buffer was supplemented with 1.5 M urea. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The atomic coordinates and structure factors (code 3B5N) have been deposited in His 6 tags were removed by thrombin cleavage at room temperature for 2 h, and cleavage was verified by mass spectrometry.
A polyalanine model of the neuronal SNARE complex (Protein Data Bank code 1N7S) was used for a molecular replacement search in PHASER (15). Three copies of the S. cerevisiae SNARE complex were located in the asymmetric unit. Alternate cycles of model building using the program COOT (16), positional, TLS, and individual restrained thermal factor refinement in the Crystallography and NMR System 1.2 (CNS) (17,27), and PHENIX (18) and addition of water molecules reduced the R and free R values to 20.7 and 24.7%, respectively, for all of the reflections (1.6 Å resolution). A total of 99.8% of the residues in S. cerevisiae SNARE complex are in the most favored regions of the Ramachandran plot, and 0.2% in the additional allowed regions as calculated with PROCHECK (19). The final model contains 6656 atoms (6064 protein and 592 waters). The coordinates for the structure have been deposited in the Protein Data Bank.
Mutagenesis-The removal of the first cavity was accomplished by a single mutation in Sec9p (T615M). The cavity 2 mutant consists of two mutations, T39V in Snc1p and S605I in Sec9p. The cavity 3 mutant has a single mutation in Sec9p (N486M). The removal of all three cavities was achieved by four combined mutations (T39V in Snc1p, and N486M, S605I, and T615M in Sec9p).
CD Analysis-CD data were collected on an Aviv 202-01 spectrometer equipped with a thermoelectric unit using a 1-mm-path length cell. Protein samples were 50 M in buffer containing 25 mM sodium phosphate, pH 7.5, and 150 mM KCl. Protein concentrations were determined by UV spectrophotometry at 280 nm and BCA assay (Pierce). Thermal melts were monitored at 222 nm. Data were collected every 2°C with an equilibration time of 2 min and an averaging time of 15 s. All

RESULTS AND DISCUSSION
Although the level of sequence identity between neuronal and S. cerevisiae SNAREs is not exceptionally high (Snc1p and synaptobrevin 2 (36%), Sso1p and syntaxin 1A (50%), and Sec9p and SNAP-25A (32%)), the overall structure of the S. cerevisiae SNARE complex is very similar to that of neuronal and endosomal SNARE complexes (Fig. 1). All structures consist of a four-helix bundle with all four helices arranged in parallel, i.e. their N and C termini are aligned and there is an ionic layer composed of one arginine and three glutamine residues at the center of the complex. Interestingly, the structures of SNARE complexes that are involved in plasma membrane fusion (the neuronal and S. cerevisiae SNARE complexes) are more similar to one another than they are to the endosomal SNARE complexes; the root mean square deviations between S. cerevisiae, and neuronal, early endosomal, and late endosomal structures are 0.9, 2.3, and 1.4 Å, respectively.
A striking difference between the neuronal and S. cerevisiae SNARE complexes is a more pronounced helical bend near the ionic layer ( Fig. 2A). This increased bend is likely caused by the glycine to glutamine (Gln-225) substitution in Sso1p. The larger side chain in the S. cerevisiae SNARE complex would clash with Gln-468 (one of the three ionic layer glutamines) if it had the conformation of the neuronal complex. Therefore, the backbone of Sso1p near residue 225 is pushed 1.0 -1.7 Å away from the ionic layer ( Fig. 2B; note the hydrogen bonding network involving Gln-225 in Fig. 2C), causing the more pronounced bending of the Sso1p helix. The degree of bending varies between the three independent S. cerevisiae SNARE complexes in the asymmetric unit, suggesting that this region of the S. cerevisiae SNARE complex is flexible in solution.
To investigate the contribution of the Sso1p bending to the stability of the S. cerevisiae SNARE complex, we created a Q225G mutant of Sso1p. The Q225G mutant should allow the S. cerevisiae SNARE complex to form a complex with a less pronounced bend at the ionic layer. However, the mutation decreased the thermal stability of the complex by 2°C (not shown), suggesting that the increased bending angle of Sso1p is not destabilizing.
Interestingly, residue Gly-227 in syntaxin 1A (equivalent to Gln-225 in S. cerevisiae Sso1p), which does not induce bending of the neuronal SNARE complex, is almost entirely conserved among syntaxins involved in plasma membrane fusion and in endosome/lysosome fusion (Syntaxins 1A, 2, 3, 4, 7, 12, and 16). It is, however, variable in Golgi, endoplasmic reticulum, and yeast syntaxins, where its identity varies. It is difficult to directly compare the differences in bending between S. cerevisiae and endosomal SNARE complexes due to increasing differences in helix position, especially toward the SNARE complex termini. The endosomal SNARE complexes, like the neuronal SNARE complex, have a glycine at residue 227 (residue 225 in neuronal syntaxin 1A). Inspection of the structures suggests that the degree to which the syntaxin helix is pushed out is governed by the size of side chains in the ϩ1 layer.
Apart from the increased bending of the Sso1p helix at the ionic layer, the ionic layer of S. cerevisiae SNAREs is very similar to that of the neuronal SNARE complex (Fig. 3, A and B), where Arg-53 forms a hydrogen bond network with Gln-224, Gln-468, and Gln-622 of Sso1p and Sec9p (Fig. 3A). The side chain of Arg-53 adopts two different rotamers in the three independent S. cerevisiae SNARE complexes in the asymmetric unit. Alternative rotamers of the equivalent arginine residue in the ionic layer have also been observed in neuronal SNAREs (10).
One remarkable property of the neuronal SNARE complex is its high stability: The neuronal SNARE complex undergoes cooperative unfolding with a melting temperature of ϳ90°C ( Fig. 4) and resists SDS denaturation up to 60°C (10). The S. cerevisiae SNARE complex is less stable and melts at T m ϭ 55°C (Fig. 4), nearly 35°C lower than its neuronal homologue. In addition to its lower T m , the S. cerevisiae SNARE complex is not SDS-resistant. The early and late endosomal SNARE complexes have melting temperatures of 87 (12) and 78°C, respectively (20).
The melting temperature of the S. cerevisiae SNARE complex increases with the length of the constructs from 55 to 64°C when an additional eight residues are included at the N terminus of the complex (data not shown). Similar behavior was observed for the neuronal complex, where the T m increases from 90 to 95°C (10). The longer, more thermally stable S. cerevisiae SNARE complex, however, diffracted only to ϳ3.5 Å resolution and was not pursued further. Similarly, the longer neuronal SNARE complex was solved at 2.4 Å resolution (11), whereas the shorter complex diffracted to 1.45 Å (10). The effect of SNARE complex length is further exemplified by our previous report (9). The melting temperature of the S. cerevisiae SNARE complex consisting of Sso1p N-terminal domain (residues 1-265), Snc1p (residues 1-93), and Sec9p (residues 401-651) is 71°C (9), whereas the S. cerevisiae SNARE complex described here (Snc1p (residues 28 -92), Sso1p (residues 189 -257), Sec9p (residues 429 -500), and Sec9p (residues 589 -651)) has a reduced melting temperature of 55°C. Therefore, the N terminus of Sso1p, length of the S. cerevisiae SNARE complex, or the linker between the two Sec9p helices can have a large effect on complex thermal stability. For these reasons, the comparison of the S. cerevisiae and neuronal SNARE complexes was carried out with the shorter neuronal SNARE complex that has nearly identical length (except for the SN1 fragment, which has eleven more residues at the N terminus). The early endosomal complex is also of similar length to the S. cerevisiae SNARE complex, whereas the late endosomal complex is shorter at both N and C termini.
The core of the S. cerevisiae SNARE complex contains several unusual layers that may contribute to its lower stability. For example, the neuronal SNARE complex has a methionine in the Ϫ2 layer (Met-167 in SNAP-25), whereas the S. cerevisiae Sec9p has a threonine (Fig. 3, C and D). The smaller side chain of Thr-615 allows water to penetrate the center of the SNARE complex, creating an internal water-filled cavity. Indeed, three ordered water molecules are present in this layer (Fig. 3C). These water molecules form hydrogen bonds to each other and to Thr-461 and Thr-615 of Sec9p.
A second water-filled cavity is formed between the Ϫ4 and Ϫ5 layers in the S. cerevisiae SNARE complex (Fig. 3, E and F). The presence of a water molecule at this position is made pos-sible by a hydrophobic to hydrophilic change of two residues compared to the neuronal SNARE complex: Val-36 in synaptobrevin 2 and Ile-157 in SNAP-25A versus Thr-39 in Snc1p and Ser-605 in Sec9p (Fig. 3F). The water molecule is held in place by forming hydrogen bonds to both Thr-39 in Snc1p and Ser-605 in Sec9p (Fig. 3E).
A third cavity is present between the ϩ5 and ϩ6 layers. The cavity is created by a change of methionine in SNAP-25 (Met-71) to an asparagine residue (Asn-486) in the S. cerevisiae SNARE complex. The water molecule in this cavity is not present in all three complexes in the asymmetric unit and most likely is only partially occupied. In addition, two distinct conformations of Asn-486 can be seen in one of the three complexes, suggesting loose packing and conformational variability in the ϩ5 and ϩ6 layers.
Comparison of all SNARE complex structures determined at high resolution (2 Å or higher) showed that the S. cerevisiae SNARE complex has the most water-filled cavities. To estimate the destabilizing effect of water-filled cavities in the S. cerevisiae SNARE complex, we mutated residues in these layers to match those found in the neuronal complex, which contain no waterfilled cavities. Separate removal of cavities 1 and 3 stabilizes the S. cerevisiae SNARE complex by 1.9 and 3.8°C, respectively, whereas removal of cavity 2 destabilizes the SNARE complex by   2°C. The combined removal of all three cavities stabilizes the S. cerevisiae SNARE complex by 6.7°C (Fig. 4).
The electrostatic potentials of all SNARE complexes solved to date are very negative in the center region (11) (Fig. 5). The overall charge of the neuronal SNARE complex (PDB code 1N7S) is Ϫ19, whereas the overall charge of the S. cerevisiae SNARE complex presented here is Ϫ11. The early and late endosomal SNARE complexes (PDB codes 2NPS and 1GL2) have overall charges of Ϫ9 and Ϫ18, respectively.
The negative electrostatic potential of the center region of the neuronal SNARE complex is thought to be important for interaction with the Ca 2ϩ sensor synaptotagmin I (21). Synaptotagmin I has a positively charged surface, especially in the presence of Ca 2ϩ (22). The interaction between synaptotagmin I and SNARE complex can be disrupted by increasing the salt concentrations above physiological levels (4), further supporting the importance of electrostatic interactions in the formation of the SNARE-synaptotagmin complex. It is unknown whether the negative surfaces present in the S. cerevisiae and endosomal SNARE complexes are structural features of SNARE proteins or if they are also involved in protein-protein interactions with other regulatory partners such as calmodulin (23).
The ϩ7 layer of the SNARE complex has been proposed to be one of the key differences between SNARE complexes that fuse constitutively and those that undergo triggered fusion (24). For Drosophila SNAREs, the T254I mutation enhances both constitutive and evoked neurotransmitter release, suggesting that the tighter packing of isoleucine promotes SNARE formation and fusion, whereas the presence of threonine prevents good packing (Fig. 6B) and halts full SNARE complex formation. However, the plasma membrane yeast SNARE Sso1p contradicts this simple rule because it is involved in constitutive fusion and yet it has a threonine in the equivalent position (Thr-249). The structure of the S. cerevisiae SNARE complex shows that the ϩ7 layer is well packed despite the presence of threonine (Fig. 6A): The Sec9p valine (Val-493) and leucine (Leu-647), Snc1p Ala-78, and Sso1p Thr249 form a well-packed layer (Fig.  6C). Thus, rather than a particular residue, it is the combination of residues that determines the packing in the ϩ7 layer and perhaps contributes to the difference between constitutive and triggered fusion.
SNARE proteins play an important role in vesicle trafficking, which is a fundamental process in all eukaryotic cells for the transfer of proteins, lipids, and metabolites between compartments. Because of the importance of these processes, the SNARE complex has remained conserved throughout evolution. Here we have shown that the S. cerevisiae SNARE complex is structurally similar to SNARE complexes from higher eukaryotes but exhibits more pronounced bending near the ionic layer and contains water-filled cavities. These cavities are in part contributing to the decreased stability of the S. cerevisiae plasma membrane SNARE complex.