The dynamin-like GTPase Sey1p mediates homotypic ER fusion in S. cerevisiae

The ER forms a highly dynamic network of sheets and tubules that extends throughout the cytoplasm. Homotypic ER fusion is required to maintain the ER as one continuous network. In metazoans, the fusion of ER membranes appears to be mediated by atlastins (ATLs; Orso et al., 2009), which are membrane-bound GTPases that belong to the dynamin family. The ATLs consist of a GTPase domain, a three-helix bundle, two closely spaced trans-membrane segments, and a C-terminal cytoplasmic tail (Moss et al., 2011). They localize to ER tubules and interact with proteins that are implicated in shaping ER tubules, the reticulons and DP1 (Hu et al., 2009). A role for the ATLs in ER fusion is supported by the observations that depletion of the ATLs or the overexpression of dominant-negative forms reduced the branching of ER tubules in tissue-culture cells. Furthermore, the addition of antibodies inhibited network formation in vitro, and proteoliposomes containing purified ATL underwent GTP-dependent fusion (Hu et al., 2009; Orso et al., 2009). Recent work suggested that fusion begins when two ATL molecules localized in different membranes bind GTP and form a dimer, thereby tethering the membranes together (Bian et al., 2011; Byrnes and Sondermann, 2011). After GTP hydrolysis and P_i release, the cytoplasmic domains of both ATL molecules undergo a major conformational change, pulling the apposing membranes together, resulting in their fusion. Fusion is also facilitated by the C-terminal tail, which follows the trans-membrane segments (Bian et al., 2011; Pendin et al., 2011). An alternative model of the role of GTP hydrolysis in ATL-mediated ER fusion has also been proposed (Morin-Leisk et al., 2011). Homotypic ER fusion by ATLs is distinct from fusion mediated by SNAREs (Wickner and Schekman, 2008). Here, three t-SNARE proteins in one membrane and a v-SNARE partner in another membrane zipper up to form a four-helix bundle, thereby forcing the fusion of the membranes. Subsequently, the helix bundle is disassembled by the NSF ATPase to allow the next round of fusion. Thus, in contrast to the ATL-mediated process, fusion, by itself, is independent of nucleotide hydrolysis.

Many organisms, including yeast and plants, lack ATL homologues, raising the question of how they manage to fuse ER tubules into a network. However, these species have an analogous GTPase, called Sey1p in Saccharomyces cerevisiae and RHD3 in Arabidopsis thaliana. No organism seems to have both ATL and Sey1p homologues. Like ATL, Sey1p consists of a dynamin-like GTPase with characteristic signature motifs, a helical bundle domain (which is significantly longer than that of ATL), two closely spaced trans-membrane segments, and a C-terminal tail. Sey1p localizes to the tubular ER and interacts physically with the tubule-shaping proteins Rtn1p and Yop1p, which are homologues of the reticulons and DP1, respectively (Hu et al., 2009). Sey1p also interacts genetically with Yop1p (hence the name synthetic enhancer of Yop1p; Brands and Ho, 2002). We previously showed that deletion of SEY1 alone had no obvious ER morphology defect, but deletion of both SEY1 and YOP1 or SEY1 and RTN1 resulted in the conversion of the tubular, cortical ER into sheets (Hu et al., 2009). Because cells missing only YOP1 or RTN1 have normal ER structure, the finding that cells missing Sey1p and either Rtn1p or Yop1 have abnormal ER morphology indicates that Sey1p plays a role in shaping the ER. Mutations in the plant ATL homologue RHD3 cause ER morphology defects similar to those seen with ATL mutants (Zheng et al., 2004; Yuen et al., 2005; Chen et al., 2011; Stefano et al., 2012). Despite these similarities, it was not clear whether Sey1p mediated ER fusion, particularly because there is little sequence similarity between Sey1p and ATLs, raising the question of whether Sey1p is truly a functional orthologue of the ATLs. Here, we show that Sey1p plays a role in homotypic ER fusion and can mediate GTP-dependent fusion of liposomes. Furthermore, we demonstrate that yeast contains a Sey1p-independent fusion mechanism that requires the ER SNARE Ufe1p.

To determine whether Sey1p and human ATL1 have similar functions, we tested whether human ATL1 can replace Sey1p in S. cerevisiae. We have previously shown that cells missing the ER-shaping proteins Yop1p and Sey1p have abnormal ER morphology: they have a significant reduction in ER tubules and an increased amount of sheets (Fig. 1; Hu et al., 2009). The tubular ER network can be restored in a sey1Δ yop1Δ mutant by expressing wild-type Sey1p from a plasmid but not with a Sey1p mutant that is defective in GTP binding (Hu et al., 2009). To determine whether human ATL1 could replace Sey1p, we expressed ATL1 in sey1Δ yop1Δ cells that also express the ER-resident protein Sec63-GFP to visualize the ER. We found that wild-type ATL1 restored the ER tubular network in most sey1Δ yop1Δ cells, whereas a GTP-binding mutant of ATL1-K80A was inactive (Fig. 1). These data indicate that ATL1 can functionally replace Sey1p in maintaining ER morphology.

To investigate whether Sey1p functions in the fusion of ER tubules, we developed an in vivo assay that is similar to that used to study nuclear fusion (karyogamy) and mitochondrial fusion during the mating of yeast cells (Nunnari et al., 1997; Melloy et al., 2009; Tartakoff and Jaiswal, 2009). Haploid yeast cells expressing cytosolic GFP were mated with haploid cells expressing RFP in the ER lumen; RFP was targeted to the ER lumen by attaching a signal sequence (ss) to its N terminus and an ER retention signal to the C terminus (ss-RFP-HDEL). The GFP- and RFP-labeled cells were mixed, placed on an agarose pad, and imaged at 23°C every minute to follow the movement of the fluorescent proteins from one cell to the other. When cell fusion occurred, the cytosolic marker rapidly equilibrated between both cells, giving a starting point for the equilibration of the ER marker between the two cells. After two wild-type cells were mated, the ER marker began to equilibrate within a minute after cell fusion occurred (Fig. 2, a and b; and Video 1). Quantification of the fluorescence in both cells indicated that the decrease of the RFP signal in one cell can be accounted for by the increase of the RFP signal in the neighboring cell. Thus, the changes are not caused by newly synthesized ss-RFP-HDEL but rather by redistribution of protein from one cell to the other (Fig. 2 b). We found that with wild-type cells, the ER marker equilibrated ∼4 min after cell fusion (Fig. 2 e). When sey1Δ cells were mated with one another, the time necessary for equilibration of the ER marker increased about sixfold (∼25 min; Fig. 2, c–e; and Video 2).

One possibility is that the ER fusion that is ultimately observed is caused by nuclear envelope fusion. We tested this possibility by using the karyogamy mutant kar1-15Δ, in which the nuclei of the mating cells fail to fuse (Vallen et al., 1992). The results show that ER fusion in kar1-15Δ cells was indistinguishable from wild-type cells. Similarly, ER fusion was the same in kar1-15Δ sey1Δ and sey1Δ cells (Fig. 2 e). These results argue against a role of karyogamy in ER fusion. When SEY1 kar1-15Δ cells were mated with sey1Δ kar1-15Δ cells, ER fusion was as slow as when both cells lacked Sey1p (Fig. 2 e), indicating that Sey1p needs to be present in both cells to allow efficient ER fusion. Human ATL1 expressed in sey1Δ cells significantly improved ER fusion (Fig. 2 e); in many of the cells, the ER actually fused at a wild-type rate, but some cells showed delayed fusion. The cause of this variability is not known, but it may indicate that the expression levels of the ATL construct vary between cells or that ATL1 is only partially functional in yeast. However, a GTP-binding mutant of ATL1 (K80A) was inactive (Fig. 2 e), indicating that ATL activity is required for the rescue of the sey1Δ phenotype. The absence of the tubule-shaping proteins Rtn1p, Rtn2p, and Yop1p did not affect ER fusion (Fig. 2 e), indicating that aberrant ER morphology is not the reason for the observed fusion defects. Collectively, these data indicate that Sey1p indeed has a role in ER fusion and likely functions similarly to the ATLs.

The finding that ER fusion still occurs in S. cerevisiae in the absence of Sey1p indicates that another fusion mechanism must exist. Previous work has suggested that homotypic ER fusion requires Ufe1p, an essential ER SNARE (Patel et al., 1998). ER structure is disrupted in cells with a temperature-sensitive ufe1-1 allele when they are shifted to a nonpermissive temperature (Prinz et al., 2000). We found a strong genetic interaction between UFE1 and SEY1; although ufe1-1 cells grow about as well as wild-type cells at a permissive temperature, ufe1-1 sey1Δ cells grow very poorly (Fig. 3 a). A similar genetic interaction was also found between SEY1 and temperature-sensitive alleles of two other essential ER SNAREs, USE1 and SEC20 (Fig. S1, a and b). No genetic interaction was detected between the nonessential ER SNARE SEC22 and SEY1 (Fig. S1 c). We found that both Sey1p and Ufe1p are required for normal ER morphology. ufe1-1 sey1Δ cells have severely disrupted peripheral ER structure at a permissive temperature, whereas the single mutants have normal ER (Fig. 3 b; Hu et al., 2009).

We used the in vivo ER fusion assay to determine whether the rate of fusion decreases in ufe1-1 cells. When ufe1-1 cells were mated at permissive temperature (23°C), ER fusion occurred ∼5 min after cell fusion (Fig. 3 c), similar to what was found for wild-type cells (Fig. 2 e). It was not possible to perform this assay with cells shifted to a nonpermissive temperature (37°C) immediately before mating because yeast will not mate at this temperature (Grote, 2010). Instead, we shifted cells to 32°C just before mating. At this temperature, ufe1-1 strains were viable but grew much more slowly than wild-type cells (unpublished data), suggesting that Ufe1p function is substantially reduced at this temperature. When ufe1-1 cells were mated at 32°C, ER fusion occurred a mean of ∼10 min after cell fusion (Fig. 3 c), indicating that these cells have a modest fusion defect. At this temperature, cells missing only Sey1p fused their ER ∼16 min after cell fusion (Fig. 3 c), somewhat faster than at 23°C (Fig. 2 e). However, when both Ufe1p and Sey1p were absent, fusion was dramatically affected; ER fusion occurred a mean of 37 min after cell fusion (Fig. 3 c). It is likely that ER fusion is not entirely abolished because some Ufe1p activity is retained at the semipermissive temperature used. Together, these findings suggest that both Ufe1p and Sey1p are involved in homotypic ER fusion in S. cerevisiae. Because ER fusion is less affected in cells missing only one of these two proteins than in the double mutant, Ufe1p and Sey1p may function in separate fusion pathways. If there are two pathways, overexpression of Sey1p might compensate for the decrease in the rate of homotypic ER fusion found in ufe1-1 cells. We found that overexpression of Sey1p in ufe1-1 cells at 32°C reduced the median time of ER–ER fusion in these cells from 10 to 6 min (Fig. 3 c), suggesting that Sey1p and Ufe1p indeed function in two separate pathways of ER–ER fusion.

It has been reported that the absence of ATLs in Drosophila melanogaster cells leads to fragmented ER, as determined by fluorescence loss in photobleaching experiments (Orso et al., 2009). This finding suggests that one of the primary functions of ATLs is to maintain the ER as one continuous network. We therefore tested whether the absence of Sey1p was also required to prevent ER fragmentation. Sec63p-GFP was expressed in sey1Δ cells, a portion of the ER was continuously bleached, and the cells were imaged over time (Fig. 3 d). The results show that the entire ER gradually loses fluorescence, indicating that the ER remains contiguous as in wild-type cells (Fig. 3 d). Similar results were obtained with sey1Δ yop1Δ cells (Fig. 3 d). Interestingly, sey1Δ ufe1-1 cells also did not show fragmented ER, not even at the nonpermissive temperature (Fig. 3 d), at which ER morphology is severely disturbed (Fig. 3 b). Assuming that no ER fusion occurs in these cells, these findings suggest that there is little ER fission in yeast cells. It is also possible that there is a third fusion pathway in yeast that does not require Sey1p or Ufe1p.

To directly test whether Sey1p mediates ER fusion, we performed in vitro experiments. Full-length, codon-optimized Sey1p was expressed as a GST fusion protein in Escherichia coli. The protein was purified in the detergent dodecylmaltoside (DDM), the GST tag was removed, and the protein was reconstituted into proteoliposomes. The protein ran as a single band in SDS gels (Fig. 4 a) and was effectively reconstituted into vesicles, as demonstrated by flotation experiments (Fig. 4 a). Donor and acceptor proteoliposomes containing equal concentrations of Sey1p were used for the fusion assay (Fig. 4 b). The donor vesicles contained lipids labeled with two fluorophores at quenching concentrations, so that after fusion with unlabeled vesicles, the fluorophores were diluted, and quenching was reduced. The results show that Sey1p mediates fusion of the vesicles in a concentration-dependent manner (Fig. 4 c). No fusion was observed in the absence of magnesium ions (Fig. 4 c) or when GTP was replaced by GDP or by the nonhydrolyzable analogue GTPγS (Fig. 4 c). A GTP-binding mutant of Sey1p-K50A (Hu et al., 2009) did not support fusion (Fig. 4 d). Fusion was also reduced when a mutation was introduced into the helical bundle of Sey1p (A592V; Fig. 4 d); the analogous mutation in the plant protein RHD3 causes ER morphology defects (Wang et al., 1997). We confirmed that Sey1p-K50A has negligible GTPase activity, whereas Sey1p-A592V has near wild-type GTPase activity (Fig. 4 e and Fig. S2 a). In addition, expression of the A592V mutant of Sey1p, or of only the GTPase domain of Sey1p (Sey1 1–498), in a sey1Δ yop1Δ yeast strain did not restore the tubular ER, indicating that the in vitro results are relevant to the situation in vivo (Fig. 4 f). Because we found that homotypic ER fusion in our in vivo fusion assay required that both cells express Sey1p (Fig. 2 e), we determined whether in vitro fusion had a similar requirement. No fusion was detected between proteoliposomes containing Sey1p and liposomes that did not contain Sey1p (Fig. S2 b). Interestingly, fusion was reduced but not completely abolished between proteoliposomes containing Sey1p and those with Sey1p-K50A; no fusion occurred when both proteoliposomes contained Sey1p-K50A. It may be that the low GTPase activity of Sey1p-K50A allows it to function, albeit inefficiently, when it interacts with wild-type Sey1p. Collectively, these results strongly support the notion that Sey1p mediates ER–ER fusion both in vitro and in cells.

To elucidate further the fusion mechanism, we investigated the oligomeric nature of Sey1p. We replaced the transmembrane domains of Sey1 (amino acids 681–727) with a 12–amino acid linker. The resulting protein, Sey1-ΔTM, was purified and subjected to sedimentation velocity analysis. Most of the protein ran as a monomer in the absence of nucleotide or in the presence of GDP but as a dimer in the presence of GDP and AlF_x as a mimic of the transition state of nucleotide hydrolysis (Fig. 5). These results are similar to those obtained for ATL (Bian et al., 2011; Byrnes and Sondermann, 2011).

Our in vivo and in vitro results strongly suggest that Sey1p functions in homotypic ER fusion in S. cerevisiae. It thus appears to be the functional orthologue of the ATLs in mammalian cells, for which in vitro data and ER morphology changes suggested a role in ER fusion. Now, the data for Sey1p provide strong evidence that the GTPases actually have a direct role in ER fusion in intact cells. As for the ATLs, the fusion reaction by Sey1p could begin with the GTP-dependent dimerization of GTPase domains sitting in apposing membranes (Bian et al., 2011). After GTP hydrolysis, a conformational change would occur, which pulls the membranes together and forces them to fuse. The predicted coiled-coil region of Sey1p is significantly longer than that of the ATLs, so the exact conformational change in Sey1p remains to be established.

In mammalian cells and in Drosophila, the ATLs may be the only ER fusogen, as their depletion or deletion results in long, unbranched ER tubules or fragmented ER tubules, respectively (Hu et al., 2009; Orso et al., 2009). The same may be true for Sey1p homologues in plants, as mutations in A. thaliana RHD3 cause drastic ER morphology defects (Zheng et al., 2004; Yuen et al., 2005; Chen et al., 2011; Stefano et al., 2012). However, in S. cerevisiae, there is clearly an additional fusion mechanism. The alternative pathway is not provided by nuclear envelope fusion, as demonstrated by the use of a karyogamy mutant, but is probably mediated by ER SNAREs. This hypothesis is supported by previous in vitro experiments (Patel et al., 1998) as well as by our finding that in vivo ER–ER fusion slows dramatically in ufe1-1 sey1Δ cells and by the strong negative genetic interactions of SEY1 and UFE1. A similar strong negative interaction was also found between SEY1 and genes encoding other essential ER SNAREs, SEC22 and USE1. Therefore, ER SNAREs probably play a direct role in Sey1p-independent ER–ER fusion. Because SNAREs have essential functions in vesicular trafficking, it is difficult to evaluate the relative contributions of Sey1p- and SNARE-mediated homotypic ER fusion. Although our findings suggest that yeast has two different ER–ER fusion pathways, they do not exclude the possibility that, in intact cells, Sey1p and ER SNAREs could also cooperate in the same pathway.

A surprising difference between our findings and work on Drosophila ATL (Orso et al., 2009) is that we found no ER fragmentation in sey1Δ ufe1-1 cells at the nonpermissive temperature. One explanation could be that there is less ER fragmentation in yeast than in higher eukaryotes. In contrast to S. cerevisiae, in higher eukaryotes, the nuclear envelope breaks down, and there are dramatic alterations in ER morphology during mitosis, changes that may cause ER fragmentation. Thus, yeast may have little requirement for homotypic ER fusion during normal growth conditions.

Given that ER SNAREs associate and collaborate with COPI, the coat normally involved in Golgi–ER transport, it is interesting that a mutant in COPI causes ER morphology defects (Prinz et al., 2000). Recent results suggest that COPI plays a role in the generation of the cortical ER, independent from its role in retrograde vesicle transport (Lavieu et al., 2010). Thus, one may speculate that there is an ER SNARE- and COPI- dependent pathway of cortical ER fusion. The reason why this pathway may be sufficient to maintain ER morphology in yeast, but not in other organisms, could be that, in S. cerevisiae, the cortical ER represents a much larger percentage of the total ER (West et al., 2011).

Strains used in this study are listed in Table S1. Strain RSY275 was obtained from R. Scheckman (University of California, Berkeley, Berkeley, CA), strain ACY85 was obtained from G. Fischer von Mollard (Universität Bielefeld, Bielefeld, Germany), and strains WPY804 and WPY805 were a gift from R. Rothstein (Columbia University, New York, NY).

Cells were imaged live in growth medium at room temperature using a microscope (BX61; Olympus), U Plan Apochromat 100×/1.35 NA lens, and a camera (Retiga EX; QImaging). IPLab version 3.6.3 software (Scanalytics) was used for image acquisition and analysis.

Cells were imaged live in growth medium at room temperature using a laser-scanning inverted microscope (LSM 510 NLO; Carl Zeiss) using a Plan Neofluar 100×/1.3 NA oil objective with argon laser line of 488 nm (optical slices <1.1 µm). LSM 510 software version 3.2 (Carl Zeiss) was used for image acquisition and analysis. Magnification, laser power, and detector gains were identical across samples. Images were captured at 2-s intervals, and cells were photobleached every 2 s.

Sey1 (Sey1-ΔTM) in which the two transmembrane domains of Sey1 (amino acids 681–727) were replaced with a 12–amino acid linker was used for analytical ultracentrifugation. A maltose-binding protein–Sey1-ΔTM fusion was expressed in E. coli, bound to amylose resin (New England Biolabs, Inc.), and eluted with 10 mM maltose. After overnight cleavage with Tobacco etch virus protease, Sey1-ΔTM was purified by size exclusion chromatography in 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 4 mM MgCl₂, 2 mM EGTA, and 2.5 mM 2-mercaptoethanol (BME; buffer A), in buffer A containing 2 mM GDP (buffer B), or in buffer B containing 2 mM AlCl₃ and 20 mM NaF (buffer C). Sedimentation velocity experiments were conducted at 20.0°C in an analytical ultracentrifuge (ProteomeLab XL-I; Beckman Coulter). 400 µl of 8.2-µM Sey1 GTPase each in buffers A, B, and C was loaded in the sample chamber of two-channel centerpiece cells, along with an equal volume of the matching buffer in the reference chamber. 45 scans, collected at a rotor speed of 50 krpm and 7.5-min intervals using the Rayleigh interference detection system (Beckman Coulter), were analyzed in SEDFIT 12.1b (Schuck, 2003) in terms of a continuous c(s) distribution of Lamm equation solutions covering an s_20,w range of 1.0–15.0 S with a resolution of 140 and a confidence level of 0.68. We found that the Sey1 GTPase in buffer C had a slight excess of salt in the sample (corresponding to ∼0.8 fringes). To ensure that the best-fit frictional ratio required for the estimation of the molecular mass reflects only the protein contribution, data were analyzed in terms of a continuous c(s) distribution with one discrete component. The discrete component having a best-fit sedimentation coefficient of 0.6 S accounts for the contribution of the excess salt in the sample. We also found that the reference matching buffer B had a slightly larger refractive index (corresponding to ∼0.4 fringes) than that of the sample indicative of excess buffer components. To account for the signal offset arising from the unmatched sedimentation of solvent components, we modeled the contribution of the excess salt in the reference. This correction to the c(s) distribution is implemented in the buffer mismatch model of SEDFIT (Zhao et al., 2010). Excellent fits were obtained with root mean square deviation values ranging from 0.0030 to 0.0053 fringes. Solution densities (ρ) were measured at 20°C on a density meter (DE51; Mettler Toledo), and solution viscosities (η) were measured at 20°C using a rolling ball viscometer (AMVn; Anton Paar). The partial specific volume (v) of Sey1 GTPase was calculated in SEDNTERP 1.09 (Cole et al., 2008), and sedimentation coefficients (s) were corrected to s_20,w.

Cells of opposite mating types were grown to an OD (600 nm) of 0.1–0.4, mixed, and concentrated onto a filter (type HA, 25 mm, and 0.45-µm pore size; Millipore). The filters were placed cell side up on YPD (yeast extract/peptone/dextrose) plates and incubated at 30°C for ∼40–50 min. The cells were washed from the filters, pelleted for 2 min at 2,000 g, and resuspended in synthetic complete (SC) medium. 2 µl cell suspension was placed onto a 1-mm-thick pad made of 3% agarose and SC medium. The cells were covered with an 18-mm square coverslip, excess SC-agarose was cut off, and the edges were sealed with VALAP (equal parts Vaseline, lanolin, and paraffin). The cells were imaged using a laser-scanning microscope (LSM 5 Live; Carl Zeiss) and a Plan Apochromat 63×/1.4 NA oil differential interference contrast M27 objective with argon laser line at 488 and 532 nm (optical slices <1.7 µm). LSM 5 Live software version 4.2 SP1 was used for all image acquisitions and analysis. Magnification, laser transmission percentage, and detector gains were identical across samples. Images were captured at 60-s intervals with a scan time of 2.07 s.

Full-length S. cerevisiae SEY1 was PCR amplified from a codon-optimized, synthesized template (GenScript) using the oligonucleotides 5′-CTTCTGGATCCATGGCGGATCGTCCGGCC-3′ and 5′-CTGGACTCGA­GTCATTTTTCT­TTCTGTTCG-3′ as forward and reverse primers, respectively. The PCR product was digested with BamHI and XhoI and ligated into pGEX-4T-3 cut with the same enzymes. Recombinant wild-type and mutant GST-SEY1 was expressed in 4-liter cultures using BL21 cells and Luria broth medium containing 100 µg/ml ampicillin. Expression was started with 120 µM IPTG when cultures reached an OD₆₀₀ of 0.8 and grown overnight at 16°C. The cells were harvested, resuspended in A200 (25 mM Hepes-KOH, pH 7.4, 200 mM KCl, 2 mM EDTA, 10% glycerol, and 2 mM BME), and lysed in a microfluidizer. The membranes were sedimented by centrifugation for 45 min at 4°C in a rotor (45 Ti; Beckman Coulter) at 40 krpm. The pellet was solubilized in A200 and 1.5% DDM (Anatrace) and separated from insoluble material by centrifugation as before. The extracts were incubated for 3 h with 1 ml glutathione (GSH)-Sepharose (GE Healthcare) and then washed with 50 ml A100 (25 mM Hepes-KOH, pH 7.4, 100 mM KCl, 1 mM EDTA, 10% glycerol, and 2 mM BME) containing 0.1% DDM. The protein was eluted from GSH-Sepharose with A100 containing 0.1% DDM and 10 mM of reduced GSH (Sigma-Aldrich). The fractions containing protein were pooled, and the GST tag was cleaved off by incubation with 25 U thrombin (GE Healthcare)/mg protein at 4°C overnight. GSH was removed using a desalting column (PD10; GE Healthcare), and the protein was separated from the GST moiety by incubation with 300 µl GSH-Sepharose. The purified protein was concentrated to 1 mg/ml and reconstituted into preformed liposomes.

For donor liposome production, POPC, DOPS, rhodamine-DPPE, and NBD-DPPE (Avanti Polar Lipids, Inc.) were mixed in a glass vial at a molar ratio of 82:15:1.5:1.5, respectively (10 mM of total lipid). For acceptor liposome formation, POPC, DOPS, and dansyl-DOPE were mixed at a molar ratio of 84:15:1.0. The chloroform was evaporated under a stream of N₂ gas, and the dried lipid film was rehydrated in A100. To produce large unilamellar vesicles, the multilamellar vesicles formed during the rehydration step were subjected to 10 freeze–thaw cycles using liquid N₂ and a water bath at room temperature. To form 100 nm–sized unilamellar liposomes, the large unilamellar vesicles were extruded 21 times through a polycarbonate filter with a pore size of 100 nm (Avestin). The total lipid concentration of donor and acceptor liposomes was determined by measuring NBD or dansyl fluorescence against a standard series, respectively (wavelengths: NBD, 460/538 nm and cutoff 515 nm; dansyl, 336/517 nm and cutoff 495 nm).

Protein reconstitution was performed by detergent-assisted insertions as previously described by Rigaud and Lévy (2003). Purified Sey1p in 0.1% DDM was mixed with preformed liposomes at the indicated molar ratios and an effective detergent to lipid ratio (R_eff) of 1.3 and incubated for 30 min at room temperature. R_eff is defined by the equation R_eff = (D_total − D_water)/[total lipid]. D_total is the final detergent concentration; D_water is the critical micelle concentration of the detergent in the presence of lipid (0.3 for DDM; Rigaud and Lévy, 2003). Detergent was removed by four additions of SM-2 Bio-Beads (Bio-Rad Laboratories). Insoluble material was pelleted by centrifugation at 10 kg for 10 min and 4°C. Final proteoliposome concentration was determined by fluorescence as described for empty liposomes in the previous paragraph.

The reconstitution efficiency was determined by flotation of proteoliposomes on a sucrose step gradient. 5 µl proteoliposomes (protein/lipid, 1:200) was mixed with 55 µl of 1.9-M sucrose (Sigma-Aldrich) and overlaid with 140 µl of 1.25-M sucrose and 40 µl of 0.25-M sucrose (all sucrose solutions are buffered by 25 mM Hepes-KOH, pH 7.4). After centrifugation at 55 krpm for 2 h and 4°C in a rotor using appropriate adaptors (TLS-55; Beckman Coulter), the gradient was fractionated into four 60-µl fractions and analyzed by SDS-PAGE stained with Coomassie.

In vitro membrane fusion was performed as previously described (Orso et al., 2009; Bian et al., 2011), except that donor and acceptor proteoliposomes were mixed at a molar ratio of 1:3. In brief, labeled donor proteoliposomes (150 µM of total lipid) were mixed with unlabeled acceptor proteoliposomes (450 µM of total lipid) in the presence of 5 mM Mg²⁺ and A100 in a total volume of 49 µl per reaction. The reaction mixture was transferred into a black polystyrene 96-well plate (Corning) and incubated at 37°C for 10 min. The fusion reaction was started by the addition of 1 mM GTP (final concentration). NBD fluorescence was measured at 1-min intervals. After 60 min, 10 µl of 2.5% DDM was added to determine total fluorescence in the sample. Fusion is expressed as the percentage of total fluorescence.

GTPase activity was determined using the phosphate assay kit (EnzChek; Invitrogen) following the instructions of the manufacturer, except that the reaction buffer was replaced by A100. In brief, wild-type or mutant Sey1p was mixed with 0.5 mM GTP in a total volume of 200 µl per reaction. The samples were incubated for 15 min at 37°C, and the reaction was started by the addition of 5 mM Mg²⁺ (final concentration). Absorbance at 360 nm was measured at 1-min intervals for 30 min. The catalytic constant (k_cat) was determined using three different protein concentrations for wild type and each mutant and is expressed as P_i released per Sey1p molecule per minute ± standard.

Fig. S1 shows that there is a genetic interaction of SEY1 with the gene encoding essential ER SNAREs other than Ufe1p. Fig. S2 shows that fusion efficiency is dependent on Sey1p GTPase activity. Videos 1 and 2 show ER fusion after mating wild-type (Video 1) or sey1Δ (Video 2) cells. Table S1 lists the strains used in this study.

We thank G. Fischer von Mollard, R. Rothstein, and R. Schekman for strains.

J. Hu is supported by the National Basic Research Program of China (973 Program; grant 2010CB833702). W.A. Prinz, K. Anwar, and A. Condon are supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. T.A. Rapoport is a Howard Hughes Medical Institute Investigator.