s-Syntaxin (residues 2 to 175, additional N-terminal sequence: MRGSPR) comprising the N-terminal Habc domain and the linker region was produced in E. coli and originally tested whether it is sufficient for interaction with s-Sec1, different from the analogous closed nSec1/syntaxin 1A conformation 25. However, no interaction between s-Sec1 and s-syntaxin(2–175) could be detected by conventional methods including affinity chromatography pull-down, co-purification and native gel electrophoresis (data not shown). Given the fact that nSec1 plays a positive regulatory role during membrane fusion 27, we speculated that a potential low-affinity binding might be sufficient for its action as part of a postulated multi-protein complex at the site of fusion. We therefore tried to co-crystallize s-Sec1 and s-syntaxin in order to trap such potential low-affinity interactions, which however resulted only in crystals containing s-syntaxin. Recently, it has been shown that other SM proteins such as Sly1p and Vps45p interact with an N-terminal helical peptide of their cognate syntaxins mediated by contacts with domain I of the respective SM protein different from the nSec1 closed conformation syntaxin 1A interaction 36373839. Although s-syntaxin contains this potential peptide region, which has been predicted to have a propensity for α-helix formation 22, no binding in vitro and no co-crystallization could be detected. We can, however, not exclude the possibility that the extra sequences at the N-terminus of s-syntaxin (MRGSPR) might have interfered with binding to s-Sec1.

The crystals of s-syntaxin belong to space group P4₃2₁2 and diffracted X-rays to 3.3 Å resolution using synchrotron radiation. The asymmetric unit contains two molecules of s-syntaxin (molecules A and B, Figure 1A), which both fold into three-helical bundle structures with an approximate length of 45 Å. By convention, the helices are named Ha, Hb and Hc from N- to C-terminus. The N-termini, which include the proposed helical segment, the C-terminal linker to the SNARE motif and all but one of the loop regions connecting the helices were disordered in both molecules. The final models start at residues 31 or 32 and their C-terminal ends are defined by residues 151 (molecule A) and 156 (molecule B). Cα-Atoms of residues 32 to 142 from the two molecules can be superposed with an r.m.s. (root mean square) deviation of 0.555 Å. However, the C-terminal ends diverge substantially (Figure 1B). In molecule A, the third helix Hc is almost four helical turns shorter and the corresponding part, residues 143 to 151, which is highly conserved among syntaxins (Figure 3) forms an open coil structure which binds to a hydrophobic cleft opening between helices Ha and Hc of molecule B (Figure 1A). Although, the C-terminal coil region in molecule A (residues 143 to 151) and the corresponding helical region in molecule B have elevated B-factors compared to the overall model (A, 67.0 Ų, and B, 66.1 Ų, versus all atoms, 42.3 Ų), the electron density in this region is well interpretable (Figure 2). The extended conformation is stabilized by an intermolecular disulfide bond, Cys148A-Cys148B, and three hydrogen bonds (Figure 2). It should be noted, that if both syntaxin molecules assumed a conformation like in molecule B severe steric clashes would result. In molecule B, Cys148 lies rather protected at the interface of helices Hb and Hc. This indicates that the C-terminal end of the Habc domain can adopt different conformations, which may be important when interacting with different regulatory partners thus adding to the structural plasticity observed for neuronal syntaxins 40. In accordance, a recent time-resolved FRET (fluorescence resonance energy transfer) study indicates that rat neuronal syntaxin 1A undergoes fast conformational conversion with only a minority population in the closed conformation 41. Structural flexibility at or close to the linker region of syntaxin might promote rapid interconversion of conformations. Therefore, this additional syntaxin conformation might be of physiological relevance as it is not obvious how local unfolding would have been favored by the mild crystallization conditions (see Methods).

Molecule B and the core of molecule A are however strikingly similar to mammalian syntaxin 1A. The secondary structure of the three-helical bundle domain of rat neuronal syntaxin was initially determined by NMR (pdb code 1BR0 17) and confirmed by X-ray crystallography (pdb code 1EZ3 18). In addition, the linker region and the SNARE motif of neuronal syntaxin fold back onto the three-helical bundle in the closed conformation, as observed in the crystal structure of the complex with nSec1 (pdb code 1DN1 25), which leaves the N-terminal domain virtually unaltered.

The Cα atoms of squid (molecule B) and rat neuronal syntaxin (X-ray structure, chains A to C 18) can be superposed with r.m.s. deviations between 0.711 and 0.761 Å for residues 32 to 149 (not shown). These values are very close to the deviation between the two copies in the asymmetric unit of squid syntaxin itself and therefore the structures of rat and squid neuronal syntaxin can be considered virtually identical despite the considerable evolutionary distance. The deviation from the all-helical fold at the C-terminal region of the Habc domain in molecule A, however, is novel albeit its functional significance has yet to be proven.

Within animals, neuronal syntaxin is remarkably well conserved. The homologues of rat, H. medicinalis, C. elegans and D. melanogaster share 47.5 % sequence identity (80.3 % homology) with the sequence of the squid L. pealei (Figure 3). In comparison to the SNARE motif, which is highly conserved (residues 183 to 255, 63 % identity, 86.3 % homology), the three-helical bundle domains share somewhat lower sequence identity (residues 32 to 154, 50.8 % identity, 75.6 % homology). The surface of the Habc domain is highly conserved however (Figures 4B and 4E), underlining their functional conservation. While the N-terminal region preceding the three-helical bundle is generally the least conserved region in neuronal syntaxin, the ultimate N-terminus, which comprises the SM protein binding signature region, residues 1 to 18, in the syntaxin homologues Sed5p and Tlg2p exhibits remarkable conservation 363738. This is even more pronounced among the invertebrate sequences. However, in vitro binding of this region of neuronal syntaxin to nSec1 could not be shown experimentally for both the rat and the squid proteins 3942 (unpublished observations).

Surface regions of high amino acid sequence conservation within neuronal syntaxins match very well with the sites of intramolecular interactions in the closed conformation. Indeed all residues involved in the formation of the closed conformation as well as those involved in contacting nSec1 in the complex structure 25 are strictly conserved between squid and rat syntaxin (Figures 3, 4). Therefore, we predict that squid neuronal syntaxin can assume the same closed conformation as its rat homologue and that it will bind s-Sec1 in the same way 16, although the degree of conservation between squid and rat neuronal SM proteins is somewhat lower (66 % vs. 83 % identity).

The syntaxin binding partner Munc13 (murine Unc-13) was found to be essential for neurotransmitter release in organisms ranging from C. elegans to mammals 313243, a conservation which also postulates a squid homologue of Munc13. It has been proposed that Munc13 facilitates the transition from the closed syntaxin conformation in complex with nSec1 to the open state participating in SNARE complex assembly 4344. This is in accordance with biochemical data showing that Munc13 interacts with the three-helical bundle domain of neuronal syntaxin in a yeast two-hybrid screen 35. Specifically, the binding region was located between residues 53 and 79 in rat neuronal syntaxin, which includes part of Ha, the linker and part of Hb (Figure 3). The proposed binding site partly overlaps with the binding of the SNARE motif in the closed conformation, which might render nSec1 and Munc13 interactions exclusive.

Interestingly, the surface corresponding to this region also contains highly conserved residues, which are involved in neither the intramolecular closed conformation nor the intermolecular nSec1 interactions (arrow, Figure 4). In addition, the extended conformation of the C-terminal end of Habc locates next to this putative Munc13 interaction site. Therefore it is possible that the structural changes might be either necessary for Munc13 binding or might be a consequence of Munc13 binding or of some other potential ligand in vivo.

Another possible interaction partner of syntaxin that is capable of dissociating Munc18 from syntaxin-1 is tomosyn 45, which might imply interaction with the Habc domain. On the other hand, a C-terminal SNARE motif in tomosyn competes for SNARE complex assembly 46.

Three-helical bundle Habc domains appear to be a general feature of syntaxin-type SNARE proteins, e.g. the paralogous yeast syntaxins Sso1p, Sed5p, Tlg2p and Vam3p have all been shown to contain homologous domains 20213637. Furthermore, Q-SNARES such as syntaxin 6, syntaxin 7 and Vti1p, which substitute SNAP-25 in the endosomal SNARE complex, contain structurally similar three-helical bundle domains 1922.

The Habc domain of neuronal syntaxin interferes with SNARE complex formation by promoting the closed conformation and serves as a scaffold for the interaction with nSec1. However, only the syntaxin homologue Sso1p was shown to form a stable closed conformation 21. Sec1p, the corresponding SM protein and the closest homologue to nSec1 in yeast, binds exclusively to the assembled SNARE complex 47. Furthermore, short N-terminal signature sequences in the yeast syntaxin homologues Sed5p and Tlg2p and their mammalian counterparts are sufficient for recruitment of the respective SM proteins Sly1p and Vps45p 363738.

The structural conservation of the autonomously folded N-terminal domain of Qa SNAREs suggest that the Habc domains might recruit not only SM proteins but also other regulatory proteins involved in vesicle trafficking. Accordingly, constitutively open full-length but not N-terminally truncated yeast Sso1p can complement a ΔSSO1 ΔSSO2 yeast strain 2148. In addition, the formation of a closed conformation as observed in the neuronal system seems to be rather an exception among diverse syntaxins and might be necessary to ensure a tighter control of the syntaxin conformation and thus SNARE complex assembly than in other vesicle transport and fusion systems.

A fragment of Loligo pealei syntaxin comprising residues 2 to 175 was cloned into a modified pQE30 vector (Qiagen) missing the region coding for a N-terminal His6-tag (additional N-terminal sequence, MRGSPR). The protein was expressed in E. coli Bl21 pUBS cells.

After 3 h of induction at 37°C, the cells were lysed by ultrasonication in 50 mM HEPES KOH pH 7.4. The clarified supernatant was applied to a Sepharose Q column (Amersham Biotech) at 4°C and eluted with a linear gradient to 500 mM KCl. Subsequently, 4 M ammonium sulphate was added until the pooled fractions had a final concentration of 0.8 M. The solution was applied to a Phenyl Sepharose 6 FF column (Amersham Biotech), and the unbound fraction collected. Size exclusion chromatography on Superdex200 (Amersham Biotech) equilibrated with 20 mM HEPES KOH pH 7.4 and 100 mM KCl served as final purification step.

Large octahedral crystals were obtained by the hanging drop method after mixing equal volumes of 26 mg/ml s-syntaxin with reservoir buffer containing either 30 % PEG-400, 0.1 M Tris HCl pH 8.5 and 0.2 M Na-citrate or 18 – 20 % PEG-550MME, 0.1 M Tris HCl pH 8.5 and 0.2 M Na-citrate. For cryo-protection, the crystals were incubated with 30 % PEG-400, 10 % glycerol, 0.1 M Tris HCl pH 8.5 and 0.2 M Na-citrate over night. The crystals of space group P4₃2₁2 with lattice dimensions of a, b = 73.056 Å and c = 224.434 Å diffracted to 3.3 Å resolution at ESRF beamline ID29.

Diffraction data were processed and scaled with Denzo and Scalepack 49. The structure was solved by molecular replacement using the program AMoRe with rat syntaxin (chain A in pdb entry 1ez3 18) as a search model 50. The asymmetric unit contains two molecules. A Matthews coefficient of 3.74 ų Da^-1 and a solvent content of 66.9 % were calculated 51. The model was refined using simulated annealing, gradient minimization and individual B-factor refinement as implemented in CNS alternated by rebuilding cycles using the program O 5253. In a final step, the tensor elements describing the anisotropic displacement of the individual helices of syntaxin were refined using the TLS option in Refmac5 followed by 5 cycles of restrained minimisation 5455. The final R-factor is 0.309 with a Rfree of 0.375.

The final model contains only 227 out of 360 residues including residues 31 to 66, 74 to 107 and 113 to 151 of chain A and residues 32 to 105 and 113 to 156 of chain B. All of the residues are in allowed regions of the Ramachandran plot as defined in Procheck 56. Due to missing electron density for side chains, 66 solvent exposed residues were modeled as alanine, which, together with the disordered regions (> 1/3 of the protein crystal content) contribute to the somewhat high R factors.

The model coordinates were deposited at the pdb database under accession code 1s94.

Figures were generated with the programs Molscript 57, Bobscript 58, Raster-3D 59, ESPript 60 and Grasp 61.