Cysteine-Disulfide Cross-linking to Monitor SNARE Complex Assembly during Endoplasmic Reticulum-Golgi Transport*

Assembly of cognate SNARE proteins into SNARE complexes is required for many intracellular membrane fusion reactions. However, the mechanisms that govern SNARE complex assembly and disassembly during fusion are not well understood. We have devised a new in vitro cross-linking assay to monitor SNARE complex assembly during fusion of endoplasmic reticulum (ER)-derived vesicles with Golgi-acceptor membranes. In Saccharomyces cerevisiae, anterograde ER-Golgi transport requires four SNARE proteins: Sec22p, Bos1p, Bet1p, and Sed5p. After tethering of ER-derived vesicles to Golgi-acceptor membranes, SNARE proteins are thought to assemble into a four-helix coiled-coil bundle analogous to the structurally characterized neuronal and endosomal SNARE complexes. Molecular modeling was used to generate a structure of the four-helix ER-Golgi SNARE complex. Based on this structure, cysteine residues were introduced into adjacent SNARE proteins such that disulfide bonds would form if assembled into a SNARE complex. Our initial studies focused on disulfide bond formation between the SNARE motifs of Bet1p and Sec22p. Expression of SNARE cysteine derivatives in the same strain produced a cross-linked heterodimer of Bet1p and Sec22p under oxidizing conditions. Moreover, this Bet1p-Sec22p heterodimer formed during in vitro transport reactions when ER-derived vesicles containing the Bet1p derivative fused with Golgi membranes containing the Sec22p derivative. Using this disulfide cross-linking assay, we show that inhibition of transport with anti-Sly1p antibodies blocked formation of the Bet1p-Sec22p heterodimer. In contrast, chelation of divalent cations did not inhibit formation of the Bet1p-Sec22p heterodimer during in vitro transport but potently inhibited Golgi-specific carbohydrate modification of glyco-pro-α factor. This data suggests that Ca2+ is not directly required for membrane fusion between ER-derived vesicles and Golgi-acceptor membranes.

In eukaryotic cells, transport between different organelles occurs through vesicular intermediates that originate from one membrane compartment and fuse selectively with another. These intracellular fusion events are mediated by a family of proteins termed soluble N-ethylmaleimide-sensitive factor attachment receptors (SNAREs) 2 (1). SNAREs are membrane-associated proteins that contain a characteris-tic heptad repeat region called the SNARE motif. SNARE motifs are ϳ60 amino acids in length and participate in forming oligomeric complexes with other SNAREs (2)(3)(4). Structural information from neuronal and endosomal SNARE core complexes indicate that the SNARE motifs assemble into parallel, coiled-coil four-helix bundles (3,4).
During intracellular fusion, SNAREs from one membrane compartment interact with cognate SNAREs on another membrane compartment. The assembly of trans-SNARE complexes from two opposing membranes is thought to provide the driving force for bilayer fusion (5,6). Indeed, purified SNARE proteins reconstituted into liposomes are sufficient for fusion events between proteoliposomes when physiologically relevant SNARE combinations are used (7)(8)(9). Despite the importance of SNAREs for membrane fusion, the precise mechanism by which SNARE proteins catalyze bilayer fusion or how SNARE complex assembly is regulated remains unclear. Current models describe pathways where trans-SNARE protein pairs produce membrane stalk structures that then lead to hemifusion and fusion pore intermediates (10 -12). Further speculation centers on the nature of the fusion pore and if it is a lipid intermediate or if the transmembrane segments of SNAREs or other proteins form a pore structure (13)(14)(15). Upstream factors, such as Rab GTPases, tethering complexes, and SM proteins, are also required for membrane targeting and SNARE-dependent fusion (16 -18). How these upstream components regulate SNARE complex assembly and disassembly are not known.
To explore these questions, we have taken a cysteine-disulfide crosslinking approach to monitor SNARE protein contacts during a round of membrane fusion in a model fusion assay. In Saccharomyces cerevisiae, fusion of ER-derived transport vesicles with Golgi-acceptor membranes has been recapitulated in vitro with washed membranes and purified cytosolic factors (19,20). Genetic and biochemical studies indicate that the SNAREs Sec22p, Bet1p, Sed5p, and Bos1p are required for anterograde trafficking between the ER and Golgi compartments (21)(22)(23). These four proteins, when mixed together, form a stable quaternary complex with their SNARE motifs predicted to be structurally arranged as observed in the neuronal and endosomal SNARE four-helix bundles (24,25). Previously, disulfide cross-linking has been used to map the spatial and dynamic arrangements of oligomeric membrane receptors such as the aspartate receptor (26) and the dopamine D2 receptor (27). In this study, we engineered unique cysteine residues into the SNARE motifs of Bet1p and Sec22p such that a disulfide-cross-linked heterodimer formed under appropriate conditions. This Bet1p-Sec22p heterodimer was present not only in membrane preparations expressing both SNARE cysteine derivatives, but more importantly, this cross-linked heterodimeric SNARE species was generated in vitro through the fusion of topologically distinct membrane compartments. By combining an established cell-free ER-Golgi transport assay with the disulfide crosslinking approach, we report on the kinetics and requirements for the assembly of nascent Bet1p-Sec22p heterodimers.

EXPERIMENTAL PROCEDURES
Plasmids and Plasmid Construction-To generate plasmid pRS315-BET1, the BET1 coding sequence with ϳ300 bp of flanking upstream and downstream sequence was amplified from genomic DNA using primers containing XbaI and XhoI restriction sites, respectively. The PCR product was ligated into XbaI-and XhoI-digested pRS315 (28). The plasmid containing the SEC22 gene (pRS313-SEC22) has been described (29). Cysteine residues were introduced into the SNARE motifs of Sec22p and Bet1p by site-directed mutagenesis using the QuikChange kit (Stratagene). The sequences of oligonucleotide primers used in the construction of these cysteine-containing derivatives are available upon request. All of the constructs were sequence verified by automated fluorescent sequencing (Dartmouth Molecular Biology Core Facility).
Homology Modeling of Yeast ER-Golgi SNARE Four-helix Bundle-The structure of the yeast ER-Golgi SNARE complex was constructed by homology modeling using the program SWISS-PDB Viewer version 3.7 (36,37). The template used for modeling was the mammalian endosomal SNARE complex (Protein Data Bank accession number 1GL2) (3). The template and target sequences were aligned based on a previous alignment strategy of the core SNARE motifs (38). Briefly, the SNARE motifs of Sec22p, Sed5p, Bos1p, and Bet1p were aligned with and manually threaded onto the SNARE motifs of endobrevin, syntaxin 7, vti1b, and syntaxin 8, respectively. The initial model was further refined using the energy minimization tools provided with the SWISS-PDB program. The quality of the model was evaluated further by using the WHAT_ CHECK verification routines (39) from the program WHAT_IF (40). Following model refinement and verification, the program MODIP (41,42) was used to identify potential sites for introduction of disulfide bonds between adjacent SNARE motifs of the yeast ER-Golgi SNARE complex.
In Vitro Budding and Transport Assays-Semi-intact cells from wildtype and SNARE mutant strains were prepared as previously described (20). Microsomes were isolated from CBY1676 (43) and used to generate ER-derived vesicles containing Bet1p(I83C). Vesicle budding and transport assays following [ 35 S]glyco-pro-␣-factor (gp␣f) have been previously described (19,33). Briefly, two-stage transport reactions were performed in which COPII-generated vesicles isolated from CBY1676 microsomes were incubated with CBY1584 semi-intact cells (i.e. acceptor membranes) in the presence (or absence) of fusion factors (Uso1p, LMA1) for the indicated times and temperatures. For experiments examining both the transport of gp␣f and the extent of formation of a Bet1p-Sec22p heterodimeric species, three parallel transport reactions (30 l) were set-up in which two of the reactions were processed and averaged for ␣-1,6-mannose modification (19), whereas the third reaction underwent oxidative cross-linking as described below to catalyze disulfide bond formation between adjacent Bet1p and Sec22p proteins. When plotting the percentage of [ 35 S]gp␣f transport and amount of heterodimer formation within the same graph, the data were scaled to maximal transport and maximal heterodimer levels, respectively.
Oxidative Cross-linking of Cysteine-containing SNAREs-Disulfide cross-linking between adjacent cysteine residues was induced by the addition of Cu(1,10-phenanthroline) 2 SO 4 (Cu 2ϩ /Phen) (26,44). For initial cross-linking experiments (Figs. 2 and 3), an equivalent amount of membranes from the indicated strains were washed three times with buffer 88 (20 mM HEPES, pH 7.0, 150 mM potassium acetate, 250 mM sorbitol, and 5 mM magnesium acetate) to remove cytosol. Each wash was followed by a brief centrifugation at 20,000 ϫ g to pellet washed semi-intact cells. After the last wash, the semi-intact cells were resuspended in buffer 88, and a freshly prepared Cu 2ϩ /Phen solution was added to a final concentration of 0.2 mM. The stock Cu 2ϩ /Phen solution was made by adding equal volumes of freshly prepared 50 mM CuS0 4 (dissolved in buffer 88) and 200 mM 1,10-phenanthroline (dissolved in ethanol) into buffer 88. The reactions were incubated on ice for 15 min. Following oxidative cross-linking, the reactions were mixed with a onehalf volume of 5ϫ SDS-PAGE non-reducing sample buffer that contained 150 mM N-ethylmaleimide to quench free sulfhydryls. The reactions were then heated at 70°C for 5 min, followed by a brief 20,000 ϫ g centrifugation to pellet insoluble material. After centrifugation, a portion of the sample was removed and resolved by non-reducing SDS-PAGE, transferred onto nitrocellulose, and probed with the indicated antibodies.
For cross-linking of two-stage transport reactions, at the indicated times, the reactions were placed on ice for 2 min, followed by the addition of Cu 2ϩ /Phen to a final concentration of 0.2 mM. The reactions were incubated on ice for 15 min and then mixed with an equal volume of quenching solution (buffer 88 containing 100 mM N-ethylmaleimide) for an additional 15 min. The reactions were then centrifuged at ϳ148,000 ϫ g in a TLA 100.3 rotor (Beckman Coulter) for 10 min at 4°C to concentrate membranes. After centrifugation, the resulting membrane pellets were solubilized in 15 l of 2ϫ non-reducing SDS-PAGE sample buffer. The samples were heated at 70°C for 5 min and then centrifuged briefly at 20,000 ϫ g. The samples were resolved by nonreducing SDS-PAGE, transferred onto nitrocellulose, and probed with the indicated antibodies. For densitometric analysis, bands on immunoblots were quantified using the Labworks software package (UVP).

RESULTS
Rationale-Transport between the ER and Golgi involves a number of distinct steps that include tethering and fusion of vesicles to Golgiacceptor membranes. During vesicle fusion, SNARE complex assembly between donor and acceptor membranes is an essential feature of this intracellular transport event. ER-Golgi transport assays in yeast have primarily used [ 35 S]gp␣f glycosylation to monitor the fusion stage and to investigate mechanisms involved in this process. However, [ 35 S]gp␣f transport assays report indirectly on membrane fusion events because outer chain glycosylation relies on lumenal content mixing between vesicles loaded with [ 35 S]gp␣f and Golgi membranes containing ␣1,6mannosyltransferase activity. To further dissect subreactions involved in vesicle fusion, we developed a novel in vitro cross-linking assay to measure SNARE-SNARE interactions that occur between COPII-generated vesicles and Golgi-acceptor membranes. In this assay, single cysteine residues within the SNARE motifs of two ER-Golgi SNARE proteins were positioned such that when assembled into a SNARE complex, oxidizing conditions would produce a disulfide-cross-linked species that could be resolved by SDS-PAGE and detected by immunoblot. To ensure that the cross-linked heterodimeric SNAREs arise from different membrane compartments, isolated ER-derived vesicles expressing one of the unique cysteine-containing SNAREs were combined with Golgiacceptor membranes expressing the cognate cysteine-containing SNARE. The presence and detection of a heterodimeric SNARE species allows us to investigate SNARE interactions and complex assembly both pre-and post-fusion. This new cross-linking approach should provide additional insights into membrane targeting and fusion mechanisms by monitoring a subreaction within the vesicle fusion stage.
Homology Modeling and Prediction of Disulfide Pairs in Yeast ER-Golgi SNARE Complex-Based on the crystal structure of the endosomal SNARE complex (3), we generated a homology model of the yeast ER-Golgi SNARE complex by first aligning SNARE motifs of the ER-Golgi SNAREs with SNARE motifs of their endosomal counterparts ( Fig. 1A) (38,45). As is characteristic of coiled-coil domains, the sequences of the ER-Golgi SNAREs contained the hallmark heptad repeat unit in which hydrophobic amino acids are found at intervals of four and three amino acids. After the sequence alignments were completed, the ER-Golgi SNARE sequences were mapped onto the modeling template to generate a preliminary structure that required additional refinement using energy minimization tools provided by the SWISS-PDB software program. The yeast ER-Golgi SNARE complex model ( Fig. 1B) closely resembled the structures of both the endosomal template and the crystal structure of the neuronal SNARE complex (3,4). As in the latter structures, the ER-Golgi SNARE complex model consisted of a parallel four-helix coiled-coil bundle in which the hydrophobic layers and the central ionic layer contribute to core interactions between different SNARE helices.
Once the ER-Golgi SNARE complex model was generated, the disulfide bond prediction program MODIP was used to identify pairs of residues that were capable of disulfide bond formation if mutated to cysteines. From this ER-Golgi SNARE complex model, 21 potential disulfide bond pairs were identified by the MODIP algorithm, and then subsequently verified for disulfide bond formation in silico by converting potential residues to cysteines in the computer-generated structure. All predicted pairs of residues occur between SNAREs that are adjacent to each other in the ER-Golgi SNARE complex model, whereas no pairs were predicted involving residues from opposing SNARE helices (i.e. no Sec22p-Bos1p or Sed5p-Bet1p).
To capture disulfide-linked SNARE species from bona fide fusion events, one cysteine-containing SNARE from ER-derived vesicles must interact with a cysteine-containing SNARE located on Golgi-acceptor membranes. Ideally, the two cysteine-containing SNARE derivatives would originate on the membrane compartment in which they are functionally required during transport. In previous studies using thermosensitive alleles of SNAREs, we determined that both Bet1p and Bos1p were functionally required on ER-derived vesicles (33), whereas Sec22p was functionally required on both vesicles and Golgi membranes during transport (32). Given the specific compartmental requirements for ER-Golgi SNARE functionality, we focused our attention on three predicted disulfide pairs involving Bet1p-Sec22p (Fig. 1B).
Disulfide Cross-linking of Bet1p and Sec22p Expressed within the Same Yeast Strain-We performed initial experiments to validate our experimental design and determine which predicted SNARE pairs, if any, could generate a disulfide-cross-linked species. Yeast strains were constructed in which cysteine residues were introduced near the Ϫ3, Ϫ1, and ϩ3 hydrophobic layers in the SNARE motifs of Sec22p and Bet1p (Fig. 1). These cysteine-containing SNARE derivatives were expressed under their endogenous promoters from CEN-based plasmids, and semi-intact cell membranes were prepared from strains coexpressing both cysteine-containing SNAREs. After incubation of membranes with Cu 2ϩ /Phen to promote an oxidizing environment, cross-linking reactions were quenched with N-ethylmaleimide, and proteins were resolved by non-reducing SDS-PAGE. Disulfide crosslinked proteins were detected by immunoblotting to identify species recognized by both anti-Bet1p and anti-Sec22p antibodies and having an electrophoretic mobility near the theoretical size of a Bet1p-Sec22p heterodimer (ϳ41 kDa).
For the initial evaluation of Bet1p-Sec22p pairs (Fig. 2), cysteinecontaining SNARE derivatives were expressed from CEN-based plasmids in a sec22⌬ strain (CBY773). Therefore, in these membranes, a single copy of SEC22 was provided by the plasmid, whereas two copies of BET1 (endogenous and cysteine-containing) were expressed. Because the primary objective at this stage was to test which pairs of SNARE proteins were capable of forming disulfide-cross-linked products, the presence of endogenous wild-type Bet1p should not prevent the identification of such pairs. Three of the predicted Bet1p-Sec22p pairs were tested and two of the pairs formed disulfide-cross-linked species (Fig. 2, solid arrowhead) that were both Cu 2ϩ /Phen-dependent and recognized by anti-Bet1p and anti-Sec22p antibodies. Although the Bet1p(S76C)-Sec22p(I146C) pair is predicted to form a disulfide bond by MODIP, this particular pair failed to generate a detectable heterodimeric cross-linked species. We also tested other SNARE pair combinations (Sec22p-Sed5p and Sed5p-Bos1p) for disulfide bond formation, but were unable to unambiguously detect corresponding heterodimers from these predicted pairs (data not shown). At this point we cannot explain why these predicted SNARE pairs failed to produce cross-links but speculate that either our SNARE model contains local inaccuracies or that the arrangement of SNARE proteins in the context of cellular membranes is different from the crystal structure.
Of the two positive pairs identified from Fig. 2, we chose the SNARE pair Bet1p(I83C)-Sec22p(D153C) for our in vitro transport experiments described below. However, before proceeding with any transport experiments, we further characterized these derivatives as well as generated a strain expressing only the Bet1p-cysteine derivative. To create the Bet1p(I83C) strain, a heterozygous diploid (BET1/bet1⌬) strain was obtained and transformed with a CEN-based plasmid containing the mutant BET1 gene. After sporulation and tetrad analysis, we recovered a spore expressing only the cysteine-containing version of Bet1p (CBY1676). Dilution series analysis of strains expressing only the cysteine mutants of Bet1p or Sec22p were conducted at 30 and 37°C, and no growth defects were observed (data not shown). This result indicates that the cysteine-containing derivatives Bet1p(I83C) and Sec22p(D153C) are functional in vivo.
Membranes containing wild-type or only one of the two cysteinecontaining SNARE derivatives did not generate a Cu 2ϩ /Phen-dependent disulfide-cross-linked heterodimer recognized by both anti-Bet1p and anti-Sec22p on immunoblots (Fig. 3) when compared with oxidized membranes expressing both cysteine-containing derivatives. These results provided further evidence that the formation of a Bet1p-Sec22p disulfide-cross-linked heterodimer required expression of both cysteine derivatives within the same strain. Approximately 3-5% of the total amount of Sec22p was cross-linked to Bet1p when both cysteine-containing derivatives were expressed in the same strain. This cross-linked percentage is in accord with the amount of Sec22p associated with Bet1p when measured by native co-immunoprecipitation (29).
While conducting these disulfide-cross-linking experiments, we detected a Cu 2ϩ /Phen-dependent species appearing in membranes expressing only Sec22p-cysteine derivatives. In experiments that go beyond the scope of this paper, we determined that this species is an ϳ50-kDa Sec22p homodimer (not shown on gels) and are currently examining its functional significance in ER-Golgi transport. In Fig. 2 and other anti-Sec22p immunoblots, the protein band indicated by an asterisk is a breakdown product of the ϳ50-kDa Sec22p homodimeric species.
Bet1p-Sec22p Cross-linked Dimers Are Formed during in Vitro Twostage Transport Assays-Having identified pairs of SNARE residues capable of disulfide bond formation when expressed in the same strain, we next investigated whether a Bet1p-Sec22p cross-linked species could be generated in cell-free transport assays. In two-stage transport reactions, COPII vesicles generated from Bet1p(I83C) microsomes were incubated with Sec22p(D153C)-acceptor membranes in the presence of purified Uso1p and LMA1. At the end of transport, parallel reactions were processed to assess Golgi-specific [ 35 S]gp␣f modification or were oxidized with Cu 2ϩ /Phen to promote disulfide bond formation between any Bet1p(I83C)-Sec22p(D153C) SNARE complexes. Inclusion of [ 35 S]gp␣f in the transport experiments not only provided an internal standard for vesicle fusion, but also allowed us to compare the functional requirements for [ 35 S]gp␣f transport and Bet1p-Sec22p disulfide bond formation.
As shown in Fig. 4A, the addition of Uso1p and LMA1 to the transport reactions stimulated [ 35 S]gp␣f modification ϳ3.0-fold (black bars). The addition of Uso1p and LMA1 to the transport reactions also produced a ϳ41-kDa species (solid arrow) recognized by both anti-Sec22p and anti-Bet1p (Fig. 4B). From the immunoblots, the relative amounts of the Bet1p-Sec22p heterodimer were measured by densitometry and graphically represented in Fig. 4A (gray bars). The graph clearly shows that addition of Uso1p and LMA1 to transport reactions stimulated formation of a Bet1p-Sec22p disulfide-cross-linked product to a similar extent as [ 35 S]gp␣f transport (ϳ2.7-fold from anti-Sec22p panel and ϳ3.0-fold from anti-Bet1p panel).
Additional controls confirmed that the Bet1p-Sec22p cross-linked species was a direct result of transport. First, in the absence of Uso1p and LMA1, some [ 35 S]gp␣f transport (14.8%) occurred probably because of residual transport factors remaining associated with acceptor membranes even after washing (Fig. 4A). A similarly low level of cross-linked Bet1p-Sec22p was detected under this condition (Fig. 4B). However, if transport reactions were incubated on ice, [ 35 S]gp␣f transport was negligible, and subsequently, no cross-linked Bet1p-Sec22p was detected (Fig. 4). Second, in the absence of Sec22p(D153C)-acceptor membranes, a small amount of [ 35 S]gp␣f transport occurred because of trace amounts of Bet1p(I83C)-Golgi membranes present in the vesicle fraction (Fig. 4A). However, because the Golgi were from Bet1p(I83C) membranes, no Bet1p-Sec22p cross-linked species was formed (Fig. 4B). As a negative control, we observed that fusion reactions using the non-predicted SNARE pair of Bet1p(S76C) vesicles with Sec22p(D153)-acceptor membranes produced no detectable disulfide-  linked heterodimer (supplemental materials Fig. 1). Taken together, these results demonstrate that a specific Bet1p-Sec22p disulfide-crosslinked species forms when Bet1p(I83C) vesicles fuse with Sec22p(D153C)-acceptor membranes. The formation of this Bet1p-Sec22p heterodimeric species requires the same conditions as transport of [ 35 S]gp␣f.
Time Course of ␣-Factor Transport and Bet1p-Sec22p Heterodimer Formation Is Similar-During the course of ER-Golgi transport, vesicle SNAREs engage with acceptor membrane SNAREs to catalyze membrane fusion. Because the previous experiment only compared the extent of ␣-factor modification and Bet1p-Sec22p heterodimer formation at the end of a 60-min transport reaction, we investigated the kinetics of ␣-factor transport and Bet1p-Sec22p heterodimer formation using the two-stage fusion assay. Again, parallel reactions were set-up in which Bet1p(I83C) vesicles were incubated with Sec22p(D153C)-acceptor membranes in the presence of Uso1p and LMA1. At specific time intervals, the reactions were either processed for Golgi-specific ␣1,6mannose modification or were incubated with Cu 2ϩ /Phen to promote disulfide bond formation. As shown in Fig. 5, within the first few minutes of the reaction, a small amount of transport (ϳ10% of maximum) occurred as measured by both ␣-factor modification and Bet1p-Sec22p heterodimer formation. In addition, throughout the 60-min reaction, the kinetics of ␣-factor transport and Bet1p-Sec22p disulfide bond formation were very similar, suggesting that the mixture of lumenal compartments and SNARE complex assembly occur on a similar time scale and report on the same fusion event.
Addition of Uso1p Alone Does Not Produce Maximum Levels of Bet1p-Sec22p Heterodimer-Vesicle transport between ER and Golgi proceeds through defined stages of vesicle budding, vesicle tethering, and membrane fusion. We have demonstrated that addition of Uso1p to cell-free transport assays produces a dilution-resistant intermediate consisting of ER-derived vesicles tethered to Golgi-acceptor membranes (19). Because vesicle-and Golgi-SNARE associations can now be monitored using this disulfide-cross-linking technique, two-stage transport reactions were performed to assess the extent of Bet1p-Sec22p heterodimer formation when Uso1p-tethered intermediates were formed. In experiments similar to those described in Fig. 4, addition of Uso1p alone caused a 2.2-fold increase in [ 35 S]gp␣f transport when compared with reactions conducted in the presence or absence (Fig. 6A, black bars) of Uso1p and LMA1. Moreover, Uso1p addition decreased the levels of freely diffusible vesicles in our assay by ϳ30% (data not shown), a reduction that reflects vesicle tethering and fusion with Golgi membranes. Interestingly, when Bet1p-Sec22p heterodimer formation was assessed in these reactions, the stimulation of disulfidecross-linked products (Fig. 6, A, gray bars, and B) closely paralleled the stimulation of [ 35 S]gp␣f transport. Taken together, these results indicate that in the presence of Uso1p, vesicles tether to Golgi membranes, yet the amount of Bet1p-Sec22p heterodimer formation mirrored the amount of vesicle fusion.

Anti-Sly1p Antibodies Inhibit [ 35 S]gp␣f Transport and Bet1p-Sec22p
Heterodimer Formation-Although SNAREs are central components of membrane fusion, additional proteins are required to ensure proper regulation and fidelity of vesicle transport and SNARE complex assembly. One such regulatory protein, Sly1p, is a member of the Sec1 family of proteins (10). Sly1p is peripherally associated with membranes through its interactions with Sed5p and is required for ER-Golgi transport in vivo and in vitro (46 -49). Previously, we used thermosensitive alleles to show that Sed5p and Sly1p are not required for Uso1p-dependent vesicle tethering, but rather their activity is needed for vesicle fusion (50). Given the requirement of Sly1p for fusion, we performed two-stage transport reactions in the presence of inhibitory antibodies specific for Sly1p. Affinity purified anti-Sly1p antibodies were titrated into transport reactions to determine the optimal amount of antibody to use during the studies (data not shown). As seen in Fig. 7, two-stage transport reactions were sensitive to anti-Sly1p antibodies, whereas preimmune IgGs at comparable concentrations did not inhibit transport (black bars). Furthermore, when the transport reactions were processed for Bet1p-Sec22p heterodimer formation, it was clear that Sly1p antibodies decreased the amount of disulfide-cross-linked product (Fig. 7,  immunoblot). In addition, the stimulation of heterodimer by Uso1p and  LMA1 paralleled that of [ 35 S]gp␣f transport as in the experiments shown above. In summary, inhibition by affinity purified anti-Sly1p antibodies indicates that Sly1p is required for membrane fusion and Bet1p-Sec22p heterodimer formation.
Formation of Bet1p-Sec22p Heterodimer Is Not Inhibited by Metal Chelators-Previous studies examining ER-Golgi transport in mammalian and yeast systems have implicated Ca 2ϩ in a late-stage requirement for vesicle fusion (51,52). Additionally, these studies suggested that Ca 2ϩ chelation causes the accumulation of ER-derived vesicles firmly docked to Golgi membranes (53). As a means to block transport and further characterize the Bet1p-Sec22p heterodimer in the fusion path-way, metal chelators were used to inhibit transport. Two-stage transport reactions were performed in the absence or presence of the metal chelators, EGTA and BAPTA. When either EGTA (3 mM) or BAPTA (0.5 mM) was included in the reactions, the overall amount of [ 35 S]gp␣f transport was significantly reduced (ϳ83-89%; Fig. 8A). From the [ 35 S]gp␣f transport data alone, it appeared that EGTA or BAPTA inhibited ER-Golgi transport, consistent with previous reports (51,52). Strikingly, when Bet1p-Sec22p heterodimer formation was measured in fusion reactions that contained EGTA or BAPTA, no reduction was detected (Fig. 8B). Moreover, the amount of Bet1p-Sec22p heterodimer and the extent of Uso1p and LMA1 stimulation were similar under all three reaction conditions (Fig. 8B).
The formation of cross-linked Bet1p-Sec22p in the presence of EGTA or BAPTA could reflect SNARE complexes that have arrested "in trans" prior to bilayer fusion (14). Alternatively, bilayer fusion could proceed normally under these conditions and the apparent inhibition of [ 35 S]gp␣f transport could be because of inhibition of outer chain carbohydrate addition. The mannosyltransferase enzymes responsible for Golgi-specific carbohydrate modifications are known to require Mn 2ϩ (54,55). To address divalent cation requirements for in vitro outer chain modification of [ 35 S]gp␣f, two-stage transport reactions were initially incubated in the presence of inhibiting amounts of EGTA and then   supplemented with Ca 2ϩ , Mn 2ϩ , or both, in an attempt to reverse the inhibition. As shown in Fig. 8C, addition of Ca 2ϩ alone failed to restore EGTA inhibition even when an excess of Ca 2ϩ (5 mM) over chelator was provided. In contrast, transport reactions supplemented with a high concentration of Mn 2ϩ (5 mM) completely reversed the inhibition, and similar levels of outer chain-modified [ 35 S]gp␣f were observed as in a standard transport reaction without EGTA. EGTA inhibition was not relieved when a lower concentration (0.1 mM) of Mn 2ϩ was provided. This was probably because of a 30-fold excess of EGTA to Mn 2ϩ in the reaction. Interestingly, when transport reactions were supplemented with low concentrations of both Mn 2ϩ (0.1 mM) and Ca 2ϩ (0.5 mM), a modest restoration of outer chain modification was observed as previously reported (51). Although this low concentration of Mn 2ϩ could not rescue the inhibition alone, the additional presence of Ca 2ϩ may compete for binding to EGTA and allow free Mn 2ϩ to participate in outer chain modification reactions. Taken together, the maximal formation of Bet1p-Sec22p heterodimers in the presence of metal chelators and the reversal of Golgi-specific outer chain modification by free Mn 2ϩ suggest that ER-Golgi transport may not directly require Ca 2ϩ .

DISCUSSION
The centrality of SNARE proteins in catalyzing fusion of intracellular membranes is well appreciated (10). However, the mechanisms that govern SNARE complex assembly and disassembly in successive rounds of membrane fusion are not well understood. In this report, we establish a new disulfide cross-linking approach to monitor the status of two ER-Golgi SNAREs during fusion of ER-derived vesicles with Golgi-acceptor membranes. This cross-linking approach relies on the placement of single cysteine residues within the SNARE motifs of Bet1p and Sec22p such that when assembled into a SNARE complex, oxidizing conditions produce an intermolecular disulfide bond.
As predicted from a molecular model of the four-helix ER-Golgi SNARE bundle, we found that cysteine pairs near the Ϫ1 or ϩ3 interaction layers in Bet1p and Sec22p produced disulfide-cross-linked heterodimers when co-expressed in yeast cells. Expression of the Ϫ1 pair in separate yeast strains allowed us to prepare ER-derived vesicles containing the Bet1p(I83C) derivative and Golgi-acceptor membranes containing the Sec22p(D153C) derivative. When these membranes were incubated in cell-free transport assays, a disulfide-cross-linked Bet1p-Sec22p adduct was detected under conditions of membrane fusion. Formation of the cross-linked heterodimer was temperature and time dependent and required the same fusion components (e.g. Uso1p, LMA1, and Sly1p) known to regulate fusion of ER-derived vesicles with Golgi-acceptor membranes (19,50). Moreover, the rate of Bet1p-Sec22p heterodimer formation mirrored the rate of Golgi-specific outer chain modification of gp␣f used in cell-free fusion assays to assess lumenal content mixing (20). Based on these results, we conclude that the formation of the cross-linked Bet1p-Sec22p species is an authentic reporter for fusion of ER-derived vesicles with Golgi membranes.
Once the validity of the SNARE cross-linking approach was established, we used the assay to examine the role of Ca 2ϩ and other divalent cations in formation of the Bet1p-Sec22p heterodimer during fusion of ER-derived vesicles with Golgi-acceptor membranes. Ca 2ϩ performs a well established role in regulated fusion of synaptic vesicles with presynaptic membranes. Membrane depolarization in neuronal cells causes a rapid Ca 2ϩ influx and is thought to trigger synaptic vesicle exocytosis through binding to synaptotagmin. Upon influx, calciumbound synaptotagmin undergoes conformational changes that alter its affinity for lipids and SNARE proteins at the neural synapse and stimulate fusion. In addition to synaptotagmin, there are likely other targets of Ca 2ϩ involved in synaptic vesicle exocyotosis that remain to be characterized (10). Ca 2ϩ has also been proposed to function in several other intracellular membrane fusion events in the early secretory pathway (51,52,56) and between endocytic organelles (57,58). In contrast to neuronal exocytosis, the role for specific Ca 2ϩ -binding proteins during these intracellular fusion reactions remains unclear. For in vitro ER-Golgi transport assays in yeast, the divalent cation chelators EGTA (51,53) and BAPTA (46) are reported to inhibit a late fusion stage of transport and are reversed by the addition of Ca 2ϩ and Mn 2ϩ .
During two-stage transport reactions, we observed that the level of Bet1p-Sec22p heterodimer formed was equivalent in the presence or absence of EGTA and BAPTA even though Golgi-specific outer chain modification of [ 35 S]gp␣f was completely inhibited by both chelators. Importantly, Mn 2ϩ , not Ca 2ϩ , was most effective in reversing the EGTA block with regard to [ 35 S]gp␣f modification. Both EGTA and BAPTA have a high affinity (K d ϭ 10 Ϫ8 -10 Ϫ9 M) for Mn 2ϩ (59). Many glycosylation reactions require divalent cations for their activity and, more specifically, the Golgi-resident mannosyltransferases are known to require Mn 2ϩ for the addition of outer chain carbohydrate (54,55,60), a reaction that is critical for monitoring in vitro transport of [ 35 S]gp␣f to Golgi compartments. Previous work has shown that both Ca 2ϩ (0.5 mM) and Mn 2ϩ (0.1 mM) were needed to reverse EGTA inhibition of the [ 35 S]gp␣f-based ER-Golgi transport assay (51). Under similar conditions, we observed a partial restoration of EGTA inhibition of outer chain-modified [ 35 S]gp␣f (Fig. 8C). However, a higher concentration of Mn 2ϩ alone (5 mM) was sufficient to completely reverse EGTA inhibition, whereas an equivalent concentration of Ca 2ϩ alone did not rescue the block. Based on these observations, we hypothesize that EGTA chelation does not interfere with SNARE complex assembly and the subsequent fusion of ER-derived vesicles with Golgi membranes. The observed inhibition of outer chain-modified [ 35 S]gp␣f by EGTA is probably because of inhibition of the Mn 2ϩ -dependent mannosyltransferase reaction. We speculate in previous studies (51,53), that the apparent late-stage fusion was stimulated by Ca 2ϩ because under suboptimal Mn 2ϩ concentrations added Ca 2ϩ displaced sufficient Mn 2ϩ from EGTA to support an increased level of outer chain modification.
What exactly is the molecular arrangement of the Bet1p-Sec22p heterodimer we detect by cross-linking? In cells, SNARE proteins undergo dynamic cycles during membrane fusion reactions and exist in a variety of disassembled and assembled states. Monomeric and oligomeric forms of SNARE proteins in topologically distinct compartments are thought to assemble into four-helix trans-SNARE intermediates that convert to cis-SNARE complexes when opposed membrane bilayers fuse (10). We speculate that most of the Bet1p-Sec22p heterodimer detected with the cysteine-disulfide cross-linking approach reflects Bet1p and Sec22p assembled into a four-helix cis-SNARE complex produced from membrane fusion. In support of this idea, it should be noted that our molecular model to predict suitably positioned cysteine pairs was based on the structure of a stable four-helix SNARE bundle. Furthermore, we observed a strict correlation between the kinetics of Bet1p-Sec22p cross-linking with lumenal content mixing, an indication that the Bet1p-Sec22p heterodimer measured was a product of membrane fusion. In addition, Bet1p-Sec22p cross-linked adducts were detected in native Bos1p immunoprecipitates from detergent-solubilized membranes after immunoprecipitates were exposed to oxidant (data not shown). This result indicates that at least some of Bet1p-Sec22p heterodimer exists in an ER-Golgi SNARE complex with Bos1p. However, whereas we favor the interpretation that our assay measures cis-SNARE complexes, we cannot exclude the possibility that a fraction of the Bet1p-Sec22p cross-linked species detected originates from trans-SNARE complexes or binary and ternary SNARE complexes. Further experimentation will be required to determine the precise molecular arrangement of Bet1p-Sec22p heterodimers detected in this assay.
The ability to cross-link cognate SNAREs from two distinct membrane compartments during a fusion reaction provides a new approach to decipher the mechanisms of SNARE complex assembly and dynamics. The MODIP program used to predict sites for introduction of disulfide cross-links identified 21 potential pairs in our ER-Golgi SNARE complex model. At this time only a few of the pairs have been characterized in cell-free fusion reactions and further analyses should provide insight into other molecular contacts during SNARE-mediated fusion. It may be possible to identify conformation specific pairs that form during distinct stages of the SNARE cycle and distinguish trans-SNARE from cis-SNARE complexes. This cross-linking method could also be used to follow the fate of cis-SNARE complexes as they are disassembled or captured into vesicles by coat protein complexes. Finally, this approach may be generally applicable to dissect other SNARE-dependent membrane fusion reactions and to establish assays for organellar fusion events for which in vitro techniques are not currently available.