Assembly of intermediates for rapid membrane fusion

Membrane fusion is essential for intracellular protein sorting, cell growth, hormone secretion, and neurotransmission. Rapid membrane fusion requires tethering and Sec1-Munc18 (SM) function to catalyze R-, Qa-, Qb-, and Qc-SNARE complex assembly in trans, as well as SNARE engagement by the SNARE-binding chaperone Sec17/αSNAP. The hexameric vacuolar HOPS (homotypic fusion and vacuole protein sorting) complex in the yeast Saccharomyces cerevisiae tethers membranes through its affinities for the membrane Rab GTPase Ypt7. HOPS also has specific affinities for the vacuolar SNAREs and catalyzes SNARE complex assembly, but the order of their assembly into a 4-SNARE complex is unclear. We now report defined assembly intermediates on the path to membrane fusion. We found that a prefusion intermediate will assemble with HOPS and the R, Qa, and Qc SNAREs, and that this assembly undergoes rapid fusion upon addition of Qb and Sec17. HOPS-tethered membranes and all four vacuolar SNAREs formed a complex that underwent an even more dramatic burst of fusion upon Sec17p addition. These findings provide initial insights into an ordered fusion pathway consisting of the following intermediates and events: 1) Rab- and HOPS-tethered membranes, 2) a HOPS:R:Qa:Qc trans-complex, 3) a HOPS:4-SNARE trans-complex, 4) an engagement with Sec17, and 5) the rapid lipid rearrangements during fusion. In conclusion, our results indicate that the R:Qa:Qc complex forms in the context of membrane, Ypt7, HOPS, and trans-SNARE assembly and serves as a functional intermediate for rapid fusion after addition of the Qb-SNARE and Sec17 proteins.

The docking and fusion of biological membranes is mediated by conserved families of proteins which assemble into functional, fusion-competent complexes. Rab-family GTPases mark each organelle, binding tethering effector complexes to stabilize membrane associations (1). Fusion requires SNARE (soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein receptor) 3 proteins, characterized by their heptad-repeat SNARE domains (2). SNARE domains have an arginyl (R) or glutaminyl (Q) residue at their center, and constitute conserved R, Qa, Qb, and Qc subfamilies (3). SNAREs assemble into four-helical coiled coils of R, Qa, Qb, Qc composition (4). Sec1-Munc18 family proteins catalyze the assembly of SNARE proteins anchored to each membrane into trans-SNARE complexes (5,6). High SNARE levels can reconstitute slow fusion, often accompanied by lysis (7,8), but physiological SNARE levels require additional proteins for rapid fusion (9). In the face of this complexity, it is important to map the affinities between fusion proteins and to seek functional intermediates in the pathway to fusion.
We study fusion with vacuoles from Saccharomyces cerevisiae. Vacuoles undergo constant fission and fusion in the cell. Vacuole fusion is essential for cell viability under certain growth conditions, but not on common rich media. Wada and colleagues (10) discovered vam (vacuole morphology) mutants which selectively block vacuole fusion, defining nine genes that were later shown to encode the vacuolar Rab Ypt7, the Qa-SNARE Vam3 and the Qc-SNARE Vam7, and all six subunits of HOPS complex (11,12). Two other SNAREs are required for vacuole fusion, the R-SNARE Nyv1 (which is partially redundant with Ykt6 (13), thus escaping Wada's screen) and the Qb-SNARE Vti1, which is required for other trafficking reactions and thus essential for cell growth (14). For simplicity, we refer hereafter to the four vacuole SNAREs Nyv1, Vam3, Vti1, and Vam7 as simply R, Qa, Qb, and Qc. Two other proteins are required for all trafficking reactions and thus were missed in the vam screen, the ATPase Sec18/NSF which disassembles 4-SNARE complexes, and the SNARE-binding protein Sec17/ ␣SNAP. Sec17 and Sec18 act twice in the catalytic cycle of fusion, promoting the fusion event per se without requiring energy input from ATP (15) and then coupling the energy of ATP hydrolysis to the disassembly of postfusion cis-SNARE complexes (16), freeing the SNAREs to again engage in trans.
Vacuole fusion has been studied extensively (17,18), both in vivo, in vitro with the purified organelle, and in recent years with defined lipids and recombinant proteins. Reconstitution of rapid fusion at physiological protein levels requires vacuolar lipids (19 -21) with a physiologically fluid fatty acyl phase (9), the Rab Ypt7 activated by binding GTP (9,22), the four SNAREs, and HOPS, stimulated by Sec17 and Sec18 (15). Whereas early reconstitutions employed SNAREs at a 1:1000 molar ratio to lipids, careful quantitation (23) showed that SNAREs are present at far lower levels on the isolated organelle. We found that lipids of physiological fluidity (9) allow proteoliposomes bearing physiological levels of SNAREs (23) to fuse with kinetics that are similar to the fusion kinetics of native vacuoles, indicating that they were largely reconstituted in functional form. This chemically defined reconstitution is not an end point of study, but rather an opportunity to define the binding relationships and intermediates which lead to rapid fusion.
We now report defined intermediates on a functional pathway to rapid fusion. Prior studies have shown that HOPS and Ypt7 are necessary and sufficient (24) for membrane tethering. HOPS catalyzes the functional assembly of SNAREs into transcomplexes, with the R-SNARE anchored in one bilayer and one or more Q-SNARE anchored in the apposed bilayer (5,6). Physical and functional data now indicate that the Qb-SNARE may be the last to enter the functional trans-SNARE complex. Sec17 associates with HOPS:trans-SNARE complex to trigger an extremely rapid fusion (15,25).

Results
Reconstituted proteoliposomes, bearing fluorescent lumenal marker proteins to allow rigorous assay of fusion (8), were prepared with vacuolar lipids, prenyl-anchored Ypt7, and either the R-or Qa-SNARE (Fig. 1A). Fusion incubations were initiated by the addition at t ϭ 0 of HOPS, Qb lacking its transmembrane anchor (termed soluble Qb, or sQb), and Qc, which has no integral membrane anchor (Fig. 1B). Sec17 was added either from the start of the incubation (Fig. 1B, blue) or at later times (black). Proteoliposomes with physiological concentrations of membrane-anchored Qa and Qb (15), or with substantially higher concentrations of membrane-anchored Qa and sQb (25), have been shown to undergo a slow HOPS-dependent fusion which is barely stimulated by Sec17. However, with physiological levels of sQb the fusion is very slow without Sec17 (arrow). Upon Sec17 addition, there is a burst of fusion which then meets or exceeds the fusion seen in reactions with Sec17 present from the start (Fig. 1B, blue curve). There is greater fusion when Sec17 is only added at later times. This likely reflects the inhibitory activity of Sec17 when it binds individual SNAREs prior to SNARE complex assembly (25). The fusion rate achieved during the Sec17-triggered reaction rises over time (Fig. 1C), suggesting the accumulation of a prefusion intermediate which awaits Sec17 to trigger the fusion event.
The observation that the burst of fusion upon Sec17 addition reaches an equivalent extent of fusion to that seen when Sec17 had been present from the start suggests that the fusion intermediate is stable. Sec17 could not restore fusion if it either bore the F21S/M22S mutation, which removes the apolar character of its N-proximal loop domain (26), or if it bore the K159E/ K163E mutation (15,27) which impairs its interaction with the SNARE bundle (Fig. 2). As a control, we found that the Sec17 L291A/L292A mutation which selectively inhibits interactions of Sec17 with Sec18 (15,28,29) had no effect, as expected in the absence of Sec18.
Earlier studies had shown that Sec18 and ATP␥S are needed for fusion of proteoliposomes with wildtype SNAREs at limiting Sec17 levels (15). With the sQb employed in our current studies, the rate of fusion increases as the Sec17 concentration increases from 100 nM to 600 nM (Fig. 3A), whereas maximal fusion rates are seen with only 100 nM when Sec18 and ATP␥S are also present (Fig. 3B). Similarly, fusion increases with Sec17

Membrane fusion intermediates
concentration over the 100 to 600 nM range when Sec17 is only added after 30 min (Fig. 3C), whereas 100 nM Sec17 suffices for the fusion burst when Sec18 and ATP␥S are also present ( Fig.  3D). Thus fusion with sQb is also stimulated by Sec18 and ATP␥S when Sec17 is limiting. For the sake of chemical simplicity, further analysis was done with 600 nM Sec17 alone.
Can fusion intermediates form in the absence of fusion catalytic proteins other than Sec17? Fusion incubations were performed with a mixture of Ypt7/R and Ypt7/Qa proteoliposomes plus HOPS, Sec17, sQb, and Qc (Fig. 4, lane 1), or with single component omissions (lanes [2][3][4][5] or with omission of Sec17 and either sQb, Qc, or HOPS (lanes 6 -8). Although fusionburst competent intermediate accumulated best when Sec17 alone was withheld (lane 2), rapid fusion was also restored when sQb was initially withheld and then restored, either alone or in addition to Sec17 (lanes 3 and 6). In contrast, there was less increase in the rate of fusion when Qc or HOPS was omitted initially compared with a complete reaction at t ϭ 0 (lanes 4 and 5 and 7 and 8). Do the enhanced fusion rates seen when Sec17 and/or Qb are initially omitted and then added back correspond to an accumulating complex which includes either the four SNAREs or three SNAREs other than Qb?
To physically demonstrate these functional intermediates, fusion incubations were initiated for 30 min, then solubilized with CHAPS. Immunoprecipitation with antibody to Qa allowed assay by immunoblot of other proteins which were bound to Qa. Comparable amounts of R were Qa-bound in complete reactions or in the absence of Sec17 (Fig. 5, A, lanes 1 and  2, and B, black bars), and approximately half the R was bound in the absence of Sec17 and Qb (Fig. 5A, lane 3), in accord with the functional assays of accumulated fusion intermediates (Figs. 1B and 4). However, in the absence of Sec17 and either Qc or HOPS, there was no detectable association of R or Qc with Qa (Fig. 5A,  lanes 4 and 5). Thus physical and functional data indicate that HOPS and the Qc-SNARE are needed to form a rapid-fusion intermediate, but the Qb-SNARE is not required.

Discussion
In a pioneering study, Fukuda and colleagues (19) evaluated the capacity of recombinant vacuolar SNAREs to form stable subcomplexes of the canonical RQaQbQc quaternary SNARE complex in detergent mixed micellar solution. They found that the RQaQc subcomplex was stable in the absence of Qb. We now report that a trans-RQaQc complex forms in the context of membranes ( Vacuolar fusion needs trans-SNARE complexes (30) anchored to apposed bilayers (31), fusion-permissive lipids with small head groups (23) and fluid fatty acyl chains (9), an intermembrane tether (21), and Sec17 (25,29,32,33). Although SNAREs at high concentrations can spontaneously assemble into RQaQbQc 4-SNARE complexes (19), Sec1-Munc18 proteins catalyze SNARE complex assembly at physiological SNARE concentrations (5). Surprisingly, although assembled trans-SNARE complexes suffice to keep membranes stably associated, they do not suffice for fusion; fusion also requires tethers (21). A working model to explain this might be that the energy of SNARE zippering in trans is opposed by the energy needed to distort the apposed bilayers as they are drawn together, and thus SNAREs which would fully zip as membrane-bound cis-complexes or in detergent micellar solution are only partially zippered in trans (31,34). Tethering may provide additional energy to draw the bilayers close together, facilitating the completion of SNARE zippering. Similar considerations may at least partially explain the function of Sec17 in triggering fusion. Several Sec17 proteins can bind along the length of a SNARE complex (27). This association may occur by Sec17 binding its C-terminal domain to SNARE-bound Sec18 (15,28,29), through direct interactions of basic residues in the center of Sec17 with SNAREs (27), and through insertion of the N-terminal Sec17 apolar loops into the apposed membranes (26). This may have two effects. It may stabilize the completion of SNARE zippering, as indicated both by atomic force microscopy studies (34) and by the observation that artificial addition of a transmembrane anchor on Sec17 obviated the need for the apolarity of its N-terminal loop with wildtype SNAREs which can fully zipper (15). The Sec17 apolar loop may also exert a direct effect on the bilayer; when full zippering is blocked by deletion within one of the SNARE domains, the apolarity of the loop is still required for fusion when the Sec17 is anchored by an N-terminal transmembrane domain (15).

Membrane fusion intermediates
When and why is Sec17 sometimes essential and sometimes dispensable for fusion? When membrane lipids are deformable, through having small head group lipids (23) and fluid fatty acyl chains (9), and when the Qa-and Qb-SNAREs are integrally anchored, and when the SNAREs are abundant in the membrane, i.e. at a high molar ratio to lipid, then HOPS-mediated assembly of a 4-SNARE complex in trans suffices for a modest fusion rate (25). However, when one or several of these conditions are not met, then Sec17 becomes essential for fusion, either by driving the completion of zippering or through contribution to bilayer disruption through insertion of its apolar N-terminal loop. Furthermore, our current studies suggest a functional order of assembly of the fusion state. HOPS may act first, promoting tethering and beginning SNARE complex assembly through its affinities for R-and Qa-SNAREs. The additional presence of Qc allows a fusion-competent state to be formed. Analysis of stable SNARE association by detergent solubilization and co-immunoprecipitation with immobilized antibody to Qa-SNARE shows that a novel stable complex can even form in the absence of Qb. Our working model of assembly of the fusion state is that HOPS binds Ypt7 on each membrane, effecting tethering. It then engages with R-and Qa-SNAREs (5)

Membrane fusion intermediates
to initiate SNARE complex assembly. Exploiting its direct affinity for the Qc-SNARE (35), it then catalyzes 4-SNARE transcomplex assembly. SNARE zippering is opposed by the limited elasticity of the apposed bilayers, but when several Sec17 monomers bind to the trans-SNARE complex, they may promote zippering as well as contribute to bilayer productive disruption.

Proteoliposome preparation
Proteoliposomes were prepared as described (9). Chloroform-dissolved lipids were mixed in the presence of 50 mM ␤-octyl glucoside, and a 4-mM lipid solution in chloroform was prepared at the following lipid ratios: 44.   The lysates were incubated with 10 l protein A magnetic beads and 10 g affinity-purified ␣-Qa antibody for 2 h at 4°C. As controls the components present in a complete reaction were mixed in presence of 1% CHAPS to determine the amount of postlysis protein-protein interaction, or a complete reaction was incubated with control IgGs to exclude unspecific protein binding. The beads were washed with CHAPS containing buffer and proteins were eluted in 100 l SDS sample buffer without reducing agent. Samples were analyzed by SDS-PAGE and immunoblotting using the indicated antibodies. Serial two-fold dilutions of the starting material were also analyzed to allow the immunoprecipitated protein to be quantified. Initial amounts were 50% for Qa, 10% for Qc, and 5% for R, Qb, and Sec17. B, mean values of the Qa-associated proteins of three independently performed experiments. Quantifications were performed using ImageJ software. Values are presented relative to the amount of protein associated to Qa after a complete reaction. Black, R associated with Qa; red, Qc associated with Qa; yellow, Qb associated with Qa; blue, Sec17, associated with Qa. Error bars indicate S.D. UT). All lipid mixes contained either 0.2 mol % 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (Marina-Blue-PE) or 3 mol % nitrobenzoxadiazole (NBD)-PE (Thermo Fisher). Solvent was removed under a stream of nitrogen followed by a SpeedVac for 2 h. Lipids were hydrated in RB150ϩMg 2ϩ (20 mM Hepes-NaOH, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 10% glycerol (v/v)) and mixed with prenylated Ypt7 and R or Qa at a molar protein:lipid ratio of 1:8000 (Ypt7) or 1:32,000 (SNAREs), respectively. R-RPLs were formed in presence of biotinylated R-phycoerythrin (Thermo Fisher) whereas Qa-RPLs bore Cy5-derivatized streptavidin (SeraCare Life Sciences, Milford, MA). The GST-tag of Qa was removed by TEV protease (1 M final concentration) during dialysis. Proteoliposomes were formed by detergent dialysis using a 25-kDa cutoff membrane (Spectrum Laboratories, Rancho Dominguez, CA) against 250 ml RB150ϩMg 2ϩ containing 1 g BioBeads (Bio-Rad) for 18 h at 4°C with stirring in the dark. Proteoliposomes were separated from unincorporated proteins by centrifugation (1.5 h, 55,000 rpm, 4°C, SW 60 rotor) through a Histodenz step gradient (35,25, and 0% Histodenz in iso-osmotic Rb150 plus 1 mM MgCl 2 ). Proteoliposomes were harvested at the 25 and 0% Histodenz interface, the concentration was determined by measuring the lipid phosphorus and adjusted to 2 mM, and small aliquots were frozen in liquid nitrogen and stored at Ϫ80°C.

Fusion assay
To load the Ypt7 with GTP, 1R-and 1Qa-RPLs (250 M lipid each) were incubated with 5 M streptavidin, 1 mM EDTA, and 100 M GTP in RB150 (20 mM Hepes-NaOH, pH 7.4, 150 mM NaCl, 10% glycerol (v/v)) for 10 min at 27°C, then mixed with MgCl 2 (2 mM). A mixture of the soluble components was prepared at the same time containing 50 nM each for Qc, sQb, and HOPS. If added, Sec18 and ATP␥S were present at 300 nM and 1 mM, respectively. Unless otherwise stated, Sec17 was present at 600 nM. The RPL suspension and the mixture of soluble components were transferred to wells of 384-well plates and incubated for 10 min at 27°C to prewarm all components. For each assay, 10 l of the RPL suspension was mixed with 10 l of the mixed soluble components and incubated at 27°C in a Spectra-Max Gemini XPS fluorescence plate reader (Molecular Devices, Sunnyvale, CA). Membrane fusion was determined as content mixing by measuring the phycoerythrin-Cy5-FRET (excitation at 565 nm; emission at 670 nm). To determine the relative amount of fusion, the maximal content mixing values were evaluated through complete lysis by addition of 0.2% (w/v) Thesit without nonfluorescent streptavidin. If individual components were omitted, only 7 l of the mixtures of soluble components were added to 10 l RPL suspension. After the indicated times, the missing components or RB150 was added back to yield a final volume of 20 l.

Assay of SNARE complexes by immunoprecipitation
The amount of SNARE complex formed during a fusion reaction was assayed by the co-isolation of R, Qc, and sQb after immunoprecipitation of Qa. Fusion reactions were performed for 30 min as described above but in a 320-l volume. A complete reaction contained 1R-and 1Qa-RPLs (250 nM each), Qc (50 nM), sQb (50 nM), HOPS (50 nM), and Sec17 (600 nM). When Sec17, sQb, Qc, and/or HOPS were omitted, the volume was adjusted by addition of RB150 prior to lysis. Reactions were stopped and membranes were lysed by the addition of immunoprecipitation (IP) buffer (20 mM Hepes-NaOH, pH 7.4, 150 mM NaCl, 0.2%(w/v) BSA) containing 2.78% CHAPS, 1 M GST-R, 1 M GST-Qc, 1 M MBP-sQb, and 10 g affinitypurified anti-Qa antibody. A 50-l portion was mixed with 50 l of nonreducing 2ϫ SDS sample buffer and incubated at 95°C for 5 min for immunoblot assay of the total starting sample. The rest of each sample was added to 10 l protein A magnetic beads (Thermo Scientific) which had been washed three times with 1 ml water followed by three times with 1 ml IP buffer containing 1% CHAPS. The bead/sample suspension was nutated for 2 h at 4°C. Unbound material was discarded, and the beads were washed three times with 1 ml of IP buffer, proteins were eluted in 100 l of nonreducing 1ϫ SDS sample buffer and incubated at 95°C for 5 min. 10 l of each sample were subjected to SDS-PAGE and analyzed by immunoblot using the indicated antibodies.