Interactions within the yeast t-SNARE Sso1p that control SNARE complex assembly
October 2000 Volume 7 Number 10 pp 894 - 902
Interactions within the yeast t-SNARE Sso1p that control SNARE complex assembly
Mary Munson1, 2, Xin Chen1, 2, Amy E. Cocina2, Susan M. Schultz2 & Frederick M. Hughson2
1. These authors contributed equally to this work.
2. Department of Molecular Biology, Princeton University, Princeton, New Jersey 08544, USA.
Correspondence should be addressed to F M Hughson. e-mail: hughson@...
In the eukaryotic secretory and endocytic pathways, transport vesicles shuttle cargo among intracellular organelles and to and from the plasma membrane. Cargo delivery entails fusion of the transport vesicle with its target, a process thought to be mediated by membrane bridging SNARE protein complexes. Temporal and spatial control of intracellular trafficking depends in part on regulating the assembly of these complexes. In vitro, SNARE assembly is inhibited by the closed conformation adopted by the syntaxin family of SNAREs. To visualize this closed conformation directly, the X-ray crystal structure of a yeast syntaxin, Sso1p, has been determined and refined to 2.1 Å resolution. Mutants designed to destabilize the closed conformation exhibit accelerated rates of SNARE assembly. Our results provide insight into the mechanism of SNARE assembly and its intramolecular and intermolecular regulation.
The fidelity with which intracellular transport vesicles target to and fuse with the appropriate membranes is crucial for the maintenance of cellular integrity, normal growth and intercellular signaling. Vesicle docking and fusion require members of a conserved protein family called SNAREs (soluble NSF attachment protein receptors)1. During docking, SNAREs anchored in the vesicle membrane (v-SNAREs) form stable, membrane bridging complexes with SNAREs anchored in the 'target' membrane (t-SNAREs). Critical to the formation of these complexes is the 'SNARE motif' present in both v-SNARES and t-SNAREs, a sequence of 65 amino acids characterized by heptad repeats of hydrophobic residues generally located adjacent to the C-terminal transmembrane anchor2. SNARE motifs from v-SNARES and t-SNAREs assemble to form parallel -helical bundles3-6 thought to facilitate membrane fusion, at least in part, by drawing their adjacent transmembrane domains, and thus the two membranes, into close apposition7.
Members of the syntaxin family of t-SNAREs2, 8 have, in addition to a transmembrane anchor and an adjacent SNARE motif, an N-terminal domain termed Habc9 (Fig. 1a). In at least some cases, this domain is important for binding to other proteins such as neuronal Sec1 (nSec1) that modulate syntaxin function10. In addition, Habc interacts intramolecularly with the syntaxin SNARE motif, termed H3, to generate a 'closed' conformation. This closed conformation inhibits syntaxin from forming complexes with other SNAREs11-15; indeed, the Habc domain slows the entry of the yeast syntaxin Sso1p (ref. 16) into SNARE complexes by almost 3,000-fold compared to truncated derivatives in vitro12. Kinetic control of SNARE assembly could provide a mechanism for limiting exocytosis to specific sites in vivo; proteins that accelerate assembly and are localized at sites of secretion could control the rate-limiting step for SNARE complex assembly and thus membrane fusion. Genetic evidence points to several appropriately localized factors that appear to act upstream of SNARE assembly and to interact with syntaxin family members (ref. 17 and references therein). The kinetic bottleneck caused by slow assembly may also, in conjunction with accessory assembly factors, ensure the fidelity of SNARE assembly more generally in the secretory and endocytic pathways18.
Previous studies have characterized several important features of SNARE structure. Proteolysis of neuronal SNARE complexes trims away the N-terminal two-thirds of syntaxin, leaving a resistant core comprising a four-helix bundle in which H3 forms a single continuous -helix4, 5. Syntaxin's N-terminal Habc domain has also been studied by NMR9 and X-ray crystallography19, revealing a three-helix bundle with a conserved hydrophobic groove. Interactions between this groove and H3 have been proposed to stabilize the closed conformation of the intact protein9, 13, 14, 19. Support for such a model comes from the recent X-ray structure of the cytoplasmic domain of syntaxin in complex with nSec1 (ref. 10). In this structure, the N-terminal portion of H3 forms an irregular helix that occupies part of the Habc groove, while the region close to the C-terminus of H3 veers out of the groove to interact directly with nSec1. It is unclear whether syntaxin in this complex is fully closed or partially open; biological roles for nSec1 have been proposed that would be compatible with either possibility20.
Here, we characterize the yeast syntaxin Sso1p using genetic, biochemical and structural methods. We find that the Habc domain is required in vivo, which together with earlier results17 suggests that unregulated SNARE assembly may be deleterious. To characterize the interactions between Habc and the remainder of Sso1p that kinetically block SNARE assembly in vitro, we have determined the X-ray structure of Sso1p in its closed conformation. This structure and complementary mutagenesis data provide a comprehensive analysis of interactions that control SNARE assembly by stabilizing this conformation. Taking advantage of the sequence and structural homology between Sso1p and neuronal syntaxin, we also examine how nSec1 binding may alter the closed conformation of syntaxin. This analysis suggests that syntaxin bound to nSec1 might be specifically primed for SNARE assembly.
The N-terminal domain is required in vivo
Either Sso1p or its close homolog Sso2p is required for yeast viability16. To determine whether the regulatory Habc domain is essential, we replaced the endogenous SSO1 gene with copies lacking Habc using a plasmid shuffle strategy (see Methods). Deletion mutants coding for Sso1p residues 192–288 (Sso1N1) and 146–288 (Sso1N2) (Fig. 1a) were shuffled into an sso1 sso2 yeast strain on 2 plasmids. The level of Sso1N2 protein in these cells is similar to that of the full length protein (Fig. 1b); nonetheless, these cells were unable to grow in the absence of the wild type SSO1 gene (Fig. 1c). The level of Sso1N1 protein was lower for unknown reasons (Fig. 1b). These results demonstrate that Sso1p residues 146–288 cannot suffice as the sole source of Sso protein.
The truncated Sso1 proteins should be fully capable of forming SNARE complexes, because analogous truncations have little effect on the stability or structure of ternary SNARE complexes formed in vitro12. Furthermore, a similarly truncated syntaxin has an enhanced ability to mediate membrane fusion in a reconstituted system21. It is possible that the Habc domain is required for Sso1p to interact with some as yet unidentified factor(s). However, while many proteins have been shown to bind to the C-terminal SNARE motif of Sso1p or its homologs, very few have been shown to interact with Habc. Of these, neuronal Munc-13, which binds a region within the Habc domain of syntaxin22, has no clear yeast homolog. Binding of another protein, Sec1p, to Sso1p might be disrupted by N-terminal deletions. However, Sec1p, unlike its neuronal homolog nSec1 (see below), binds to the exocytotic SNARE complexes rather than to Sso1p alone17. Furthermore, the binding of Sec1p to SNARE complexes appears to be independent of Habc (C.M. Carr & M.M., unpublished results). We hypothesize, therefore, that the critical function of Habc is to regulate SNARE assembly.
The closed conformation blocks complex assembly
Assembly of yeast exocytotic SNAREs in vitro proceeds via an ordered pathway12, 23, 24. The initial, slow step is the formation of binary complexes between the t-SNAREs Ssop and Sec9p. These t-SNARE complexes then bind efficiently to the v-SNARE Sncp to form ternary complexes. In vivo, the individual t-SNAREs Ssop and Sec9p are broadly distributed over the inner surface of the plasma membrane25; however, exocytosis is restricted to the yeast bud tips for polarized growth26. These observations suggest that Ssop–Sec9p assembly might be under kinetic control in vivo, selectively accelerated by proteins localized to bud tips. Consistent with this idea, only low levels of Ssop–Sec9p complexes appear to be present at steady state in vivo (P. Brennwald, pers. comm.).
To confirm directly that the closed conformation of Sso1p blocks Sec9p binding, we sought to establish a direct connection between the stability of the closed conformation and the rate of t-SNARE assembly. The stability of the closed conformation is sensitive to pH, decreasing significantly as the pH is raised over the range from 7.0 to 8.8 (Fig. 2a). The stability of the isolated Habc domain remains nearly constant over the same range (Fig. 2a). The second order rate constant for Sso1p–Sec9p complex formation was measured as a function of pH using either gel filtration for slow assembly reactions or circular dichroism (CD) for faster assembly reactions12 (Fig. 2b). The rate of complex assembly increased at higher pH values as the stability of the closed state decreased, as predicted if the closed conformation blocks binary complex assembly.
Sso1p X-ray structure
Based on NMR studies by Fiebig et al.13 showing that 30–40 residues at each end of the cytoplasmic domain of Sso1p are disordered in solution, we engineered a recombinant Sso1p fragment comprising the entire ordered region (residues 31–225) plus an N-terminal Met residue. This fragment, including the Habc domain, a linker region of 35 residues, and more than half of H3 (residues 185–257), crystallized readily (see Methods). The present structural model, derived from multiwavelength anomalous diffraction (MAD) experiments, is refined to 2.1 Å resolution and includes 179 of 185 amino acid residues (Fig. 3 and Table 1). Two unidentified metal ions apparently facilitate crystallization by coordinating surface side chains on adjacent molecules within the crystal lattice (see Methods).
The main body of the Sso1p structure is an antiparallel four-helix bundle (Fig. 3b). As anticipated based on previous structures9, 10, 19, the N-terminal Habc domain (residues 31–152) contributes three of these helices. A linker region connects Habc to a C-terminal -helix (residues 185–219) that completes the four-helix bundle. This linker region (residues 153–178) forms two short -helices, which we term HL1 and HL2. HL1 (residues 156–164) is nearly perpendicular to the helix bundle, while HL2 (residues 165–178) is roughly parallel. The short loop (residues 179–184) connecting HL2 to H3 is apparently disordered and is not included in the model. At the C-terminus, residues 220–225 are extended.
Comparison with neuronal syntaxin
Despite limited amino acid sequence identity (26%), the Habc three-helix bundle in Sso1p is very similar to that seen previously in the X-ray structures of the Habc domain of syntaxin alone19 (root mean square (r.m.s.) deviation of 1.2 Å for 103 helical C pairs) and in the syntaxin–nSec1 complex10 (r.m.s. deviation of 1.1 Å for 111 helical C pairs). The main difference is in the region around the Hb-Hc loop, where syntaxin has 10 additional residues that extend each helix and form a longer loop (Fig. 4a–c). Significant structural differences between Sso1p and the syntaxin–nSec1 complex are evident in the linker (23% amino acid identity) and H3 (47% amino acid identity) regions. Although some of these differences may reflect sequence divergence between Sso1p and syntaxin, others are likely to be attributable to conformational changes in syntaxin upon binding nSec1.
To analyze structural differences between Sso1p and syntaxin in complex with nSec1, the two structures were aligned using their structurally conserved Habc motifs (Fig. 4a,b). This structure-based alignment (Fig. 4c) differs from previously reported sequence-based alignments9, 16 by four residues, or about one helical turn, in Hb. The alignment reveals striking differences in the linker region (Fig. 4d). The HL2 helix is absent from the syntaxin–nSec1 structure. The syntaxin HL1 helix (earlier simply called the linker helix10) is displaced 6 Å down the helix bundle relative to the HL1 helix of Sso1p (Fig. 4d). This displacement may be attributed to the reported attractive polar interactions between HL1 and nSec1 (ref. 10). These structural differences in the linker region may have significant consequences for SNARE assembly, as discussed in detail below.
The syntaxin H3 helix also seems to be altered by its interaction with nSec1. In the syntaxin–nSec1 complex, H3 is fragmented into several curved segments10 (Fig. 4c), whereas the corresponding helical regions seen in the Sso1p structure are straight and track along the conserved hydrophobic groove formed by Hb and Hc (Fig. 4a,b). Beyond residue 212, syntaxin interacts extensively with nSec1 (ref. 10). It seems likely that, in uncomplexed syntaxin, the H3 region might relax into a more Sso1p-like conformation.
Interactions stabilizing the closed conformation
The H3 helix of Sso1p interacts extensively with Habc (Fig. 5a). Many salt bridges and hydrogen bonds distributed along the entire length of H3 stabilize this interaction (not shown). Toward its C-terminus, the H3 helix participates with the Habc domain in forming a hydrophobic core centered around Phe 122 that includes residues from Hb, Hc and H3 (Fig. 5b). While many of these residues are hydrophobic, polar residues also contribute the aliphatic portions of their side chains. Nearer the N-terminal end of H3, there are fewer hydrophobic interactions between H3 and Habc, due in part to a large number of buried waters (Fig. 5a). The linker region flanks H3 at its N-terminal end (Fig. 5a), apparently stabilized by a small hydrophobic core involving Hc, HL1 and HL2 (Fig. 5c).
To visualize directly how the closed conformation may block SNARE assembly, we compared the Sso1p structure with the SNARE complex structure. Because the structure of a yeast SNARE complex has not yet been determined, and because the yeast and neuronal SNAREs share structural and functional homology6, 23, we used the X-ray structure of the neuronal SNARE core complex5 for this comparison (Fig. 6). The crystals of the neuronal SNARE complex5 were derived from proteolytically digested SNARE complexes27 and are composed exclusively of SNARE motifs: the entire H3 region of syntaxin, both SNARE motifs from the Sec9p homolog SNAP-25 (designated SNAP-25N and SNAP-25C), and the SNARE motif from the v-SNARE synaptobrevin. Because the H3 regions of Sso1p and syntaxin can be unambiguously aligned based on the underlying heptad repeat2 and the high overall sequence identity (see Fig. 4c), we were able to superimpose the Sso1p and neuronal core complex structures by overlaying the corresponding H3 helical regions (Fig. 6).
We found to our surprise that the other three non-syntaxin helices in the SNARE complex (blue) superimpose poorly with the three helices of the Habc domain (red). Thus, the conserved hydrophobic groove in Habc does not mimic the interaction surface presented by the other SNAREs in the core complex. Conversely, SNAP-25 and synaptobrevin do not simply take the place of the Habc helices during SNARE complex formation. Indeed, several hydrophobic residues (green in Fig. 6) are exposed on the surface of the closed Sso1p, although they occupy a positions in the heptad repeats (ref. 2) and are consequently buried in the hydrophobic core of the SNARE complex5.
Strikingly, HL2 occludes a significant portion of the SNAP-25N 'binding site' on H3 (Fig. 6). Except for HL2 and a few other linker residues, the position occupied by SNAP-25N in the core complex is essentially unobstructed in the closed conformation. SNAP-25N is of particular interest because it can bind to syntaxin independently of SNAP-25C or synaptobrevin28, indicating that SNAP-25N binding might be an early step in SNARE assembly. To substantiate the roles of this and other intramolecular interactions in the kinetic control of SNARE assembly, we made a large number of point mutations and measured their rates of binary complex formation.
Mutational analysis of SNARE assembly
Prior to the determination of the Sso1p structure, a number of Ala substitutions were made in the region of Sso1p judged (based on homology modeling) to be the conserved hydrophobic groove between Hb and Hc. These mutations were designed to interfere specifically with the inhibitory function of Habc by weakening the intramolecular interaction between Habc and H3. Only positions on the surface of the Habc domain were mutated to avoid destabilizing Habc itself. Furthermore, to circumvent complications that might result from changing the stability of the Sso1p–Sec9p complex, we avoided mutating the SNARE motif. We found that mutations in the Hb-Hc groove often had substantial effects on the kinetics of SNARE complex assembly, whereas mutations in either of the other two Habc grooves had no effect (see below). In light of the crystal structure, additional mutations were made in the linker helix region; these mutations had unexpectedly large effects on SNARE assembly rates.
As an initial screen, the rate of binary complex assembly was evaluated for all the mutant proteins using a gel mobility shift assay (Fig. 7a and data not shown). A number of mutants bound Sec9p significantly faster; for these, the rate was measured quantitatively using either gel filtration or circular dichroism assays (Fig. 7b,c; Table 2). These mutations presumably accelerate binary complex assembly by destabilizing the closed conformation. Six mutant Sso1 proteins (K95A, K99A, R119A, L123A, A144K and Y148A; red in Fig. 7d) bind Sec9p at least 10-fold more rapidly than wild type. The first four of these mutations localize to the large hydrophobic cluster stabilizing the C-terminal portion of H3 (see Fig. 5b). The three positively charged residues, Lys 95, Lys 99 and Arg 119, serve double duties. For each of these residues, aliphatic side chain atoms contribute to the hydrophobic cluster whereas polar atoms form a salt bridge (Lys 95), a hydrogen bond (Lys 99), or both (Arg 119) with H3. A number of additional residues clustered around the conserved hydrophobic Hb-Hc groove significantly stabilize the closed conformation (Table 2; Fig. 7d).
To confirm that the effect of mutations on SNARE assembly was due to their effect on interactions between Habc and H3, we made use of the observation that the isolated Habc domain (residues 1–145) binds weakly to a C-terminal fragment (residues 192–265) in trans12. This interaction is stronger in the intact protein because of the higher local concentrations provided by covalent attachment. Nonetheless, the relative strength of the bimolecular interaction yields information about the relative strength of intramolecular Habc–H3 interaction. The Sso1p Habc domain with the single K95A mutation had a reduced ability to bind to Sso1p residues 192–265, while no binding was detected for the double K95A/L123A mutant (Fig. 7e). Together, these results provide strong support for the conclusion that interactions between Habc and H3 are energetically important for maintaining the closed conformation and kinetically inhibiting SNARE assembly.
Assembly is also accelerated markedly by two other mutations, A144K and Y148A (Fig. 7d). These mutations lie far from the large hydrophobic cluster and were designed to destabilize the linker helices (Figs 4d, 5c). Together with the five-fold effect of G169E in the same region, these results indicate that the linker region plays an important role in regulating SNARE assembly by blocking access to Sec9p, as suggested above on the basis of structural alignment (Fig. 6). This conclusion is also consistent with a study suggesting that syntaxin is less closed when multiple mutations are introduced into HL1 (ref. 14).
Combining mutations in the hydrophobic core and linker region led to a triple mutant (V84E, K95E and Y148A) that assembles 1,100 times more rapidly than wild type Sso1p. This rate approaches that achieved by deleting Habc and the linker region altogether, suggesting that the main interactions stabilizing the closed conformation have been identified.
nSec1 may prime syntaxin for assembly
As shown here, interactions between H3 and Habc play an important role in slowing the rate of yeast SNARE assembly in vitro. Based on our structural and mutational data, we also propose a novel role for the linker region, and more specifically for HL2, in regulating SNARE assembly. Since HL2 occupies the position that the N-terminal SNAP-25/Sec9p helix is expected to occupy in the SNARE complex, it is well positioned to block an early step in SNARE assembly.
Syntaxin in complex with nSec1 lacks an HL2 helix (Fig. 4d); instead, this region was difficult to visualize due to weak electron density and was modeled in an irregular conformation10. The average B-factor for this region was 120 Å2 (compared to 62 Å2 overall)10. Several lines of evidence suggest, however, that this region may be better ordered when syntaxin is not in a complex, potentially forming an HL2 helix as seen in Sso1p. NMR studies on mobile regions of syntaxin do not identify residues within the linker region14. Furthermore, a site within the HL2 region (asterisk in Fig. 4c) is relatively protease-resistant in monomeric syntaxin19. Taken together with the presence of an HL2 helix in Sso1p (average B factor 45 Å2 compared to 36 Å2 overall), these observations suggest that flexibility in the HL2 region of syntaxin may be induced by nSec1 binding.
In the structure of the syntaxin–nSec1 complex10, the HL1 helix is displaced 6 Å relative to its position in Sso1p (Fig. 4d). This movement might unravel HL2, thereby priming syntaxin for SNARE assembly. Misura et al.10 also noted that displacement of the linker would be important for SNAP-25 binding. Confirmation of these hypotheses will require further structural and biochemical studies of neuronal SNARE assembly.
The role of Sec1 family members in trafficking has been controversial (see refs 17, 29, and references therein); in particular, both inhibitory and stimulatory roles for nSec1 in neurotransmitter release have been proposed. The structural comparisons made possible by the Sso1p structure, and the marked effect of linker region mutants on assembly kinetics, suggest that nSec1 binding may cause structural changes in syntaxin that are important for SNAP-25 binding, subsequent ternary complex formation, and membrane fusion. In yeast, it is unclear whether Sec1p plays a similar role in priming Sso1p. Arguing against this notion, earlier results suggested that Sec1p binds with high affinity only to SNARE complexes17. Alternatively, Sec1p may indeed play an inhibitory and/or priming role in vivo in a manner not yet faithfully reproduced in vitro29.
Regardless of the exact role of Sec1p, it is likely that the closed conformation of Sso1p is modified by upstream factors to promote SNARE assembly. The potential generality of such a model is supported by the observation that a number of t-SNAREs have N-terminal domains that appear to adopt helical bundle structures resembling those of Sso1p and syntaxin (ref. 30; L.F. Cavanaugh & F.M.H., unpublished observations). Upstream factors, then, would act to achieve spatial and temporal restriction of membrane fusion through controlling the receptivity of the t-SNARE to complex assembly. Candidates for upstream factors include, in addition to Sec1p, the Rab protein Sec4p, and the multisubunit Exocyst complex16, 17, 31-33. Notably, both Sec1p and the Exocyst complex are localized to sites of exocytosis in yeast17, 33. It remains as an outstanding question to delineate precisely how these or other upstream factors function, individually or in concert, to ensure the topological specificity of SNARE function.
Another role for the closed t-SNARE conformation is suggested by recent evidence indicating that SNARE assembly is promiscuous in vitro — noncognate complexes form readily with stabilities comparable to those of cognate complexes30, 34, 35. Thus, beyond spatial and temporal restriction, the built-in 'off' switch provided by the Habc and linker domains may protect against the formation of fusogenic noncognate SNARE complexes. Indeed, recent evidence suggests that SNARE assembly in vivo may be more specific than suggested by in vitro experiments15. Finally, the closed conformation may prevent SNARE complexes disassembled by Sec18p/NSF from immediately reassembling3. Our characterization of the intramolecular interactions stabilizing the closed conformation should aid in the design of experiments to test these models in vivo.
Yeast methods. All strains were grown at 30 °C on YPD medium or SC medium minus the appropriate amino acid. Strains 1060 (MATa his31 leu20 met150 ura30 sso1:kanMX4) and 10768 (MAT his31 leu20 lys20 ura30 sso2:kanMX4) as described37 were purchased from Research Genetics (Huntsville, Alabama). Wild type SSO1 plasmids were created in CEN-URA3 (pRS316), CEN-LEU2 (pRS315) and 2–LEU2 (pRS425) by subcloning SSO1 (residues 1–288) plus 0.5 kb of upstream and downstream flanking genomic DNA into the BamHI and XbaI (or NotI for pRS425) sites of these plasmids38, 39. SSO1-CEN-URA3 (pMM250) was introduced into 1060 by PEG-lithium acetate transformation. This strain was mated to 10768, and the resultant Met+ Lys+ diploid was sporulated and subjected to tetrad analysis. The haploid strain FHY102 (MATa his31 leu20 ura30 sso1:kanMX4 sso2:kanMX4 pMM250) was selected based on its inability to grow on 5-
(Message over 64 KB, truncated)