Sec17/Sec18 can support membrane fusion without help from completion of SNARE zippering

Membrane fusion requires R-, Qa-, Qb-, and Qc-family SNAREs that zipper into RQaQbQc coiled coils, driven by the sequestration of apolar amino acids. Zippering has been thought to provide all the force driving fusion. Sec17/αSNAP can form an oligomeric assembly with SNAREs with the Sec17 C-terminus bound to Sec18/NSF, the central region bound to SNAREs, and a crucial apolar loop near the N-terminus poised to insert into membranes. We now report that Sec17 and Sec18 can drive robust fusion without requiring zippering completion. Zippering-driven fusion is blocked by deleting the C-terminal quarter of any Q-SNARE domain or by replacing the apolar amino acids of the Qa-SNARE that face the center of the 4-SNARE coiled coils with polar residues. These blocks, singly or combined, are bypassed by Sec17 and Sec18, and SNARE-dependent fusion is restored without help from completing zippering.


Introduction
Membrane fusion requires Rab-family GTPases and SNARE proteins. SNAREs constitute four families, termed R, Qa, Qb, and Qc . Each of them has an N-domain, an a-helical SNARE domain of 50-60 aminoacyl residues with heptad-repeat apolar residues, and often a C-terminal membrane anchor. Each SNARE a-helical turn is termed a 'layer.' The central '0-layer' of each fully-assembled SNARE complex has inwardly-oriented arginyl (for R-SNAREs) or glutaminyl (for Qa, Qb, and Qc SNAREs) residues, forming a polar center to the otherwise hydrophobic core of the 4helical SNARE bundle . The SNARE domain layers are numbered from the 0layer, in the positive direction toward the SNARE C-termini and in the negative direction toward the N-domains. Prior to 4-SNARE assembly, individual SNARE domains are random coil (Fasshauer et al., 1997;Hazzard et al., 1999). Sec1/Munc18 (SM) family proteins catalyze the N-to C-directional assembly of SNAREs anchored to each tethered membrane (Fiebig et al., 1999;Sørensen et al., 2006;Baker et al., 2015;Orr et al., 2017;Jiao et al., 2018). Each SNARE domain transitions from random coil to a-helix as the heptad-repeat apolar amino acyl residues become sequestered into the interior of the coiled coils (Fasshauer et al., 1997;Sutton et al., 1998). This hydrophobic collapse relies on the exclusion of water and is the driving force for SNARE assembly (Sørensen et al., 2006). Completion of SNARE zippering can release up to 40 kBT per SNARE complex (Gao et al., 2012;Min et al., 2013;Zhang, 2017) to overcome the 40-90 kBT hydration barrier for membrane stalk formation, the dominant energy barrier for fusion (Aeffner et al., 2012). Upon fusion, the trans-SNARE complex becomes a cis-complex, anchored to the fused membrane bilayer. Sec17 (aSNAP) and SNAREs are receptors for the Sec18 (NSF) AAA ATPase (Clary et al., 1990;Winter et al., 2009;Zick et al., 2015). Sec18 uses the energy from ATP binding and hydrolysis to SNARE complexes (Ma et al., 2016, but see Ryu et al., 2015). However, the mechanism by which HOPS-dependent fusion is stimulated by Sec17/Sec18 has been unclear.
Complete SNARE zippering is considered essential for SNARE-dependent fusion (Sørensen et al., 2006). We now report that Sec17 has a second mode of promoting fusion, which can compensate for incomplete zippering. Fusion with SNAREs and HOPS is completely arrested when several heptad repeats in the C-terminal region of the SNARE core complex are removed from any one of the three Q-SNAREs. With any such C-terminally truncated Q-SNARE domain, or even with C-terminal truncations to both Qb and Qc, blocked fusion is restored by Sec17 and Sec18 without ATP hydrolysis. Association between the C-terminal heptads of the R and the single remaining full-length Qa-SNARE, each anchored to one of the docked membranes, would not yield the same assembly energy as with wild-type SNAREs  and would contribute far less force toward the bilayer rearrangements of fusion. The N-terminal apolar loop of Sec17 is particularly important for this function of Sec17, and it may stabilize SNARE bundles or trigger fusion by insertion into lipid bilayers. Zippering-driven fusion is also arrested with full-length SNARE domains when the apolar, inward-facing residues of the Qa SNARE layers +4 to +8 are replaced by Ala, Ser, or Gly, but in each case fusion is restored by Sec17, Sec18, and non-hydrolyzable ATPgS. Strikingly, even fusion that is blocked by the concurrent replacement of apolar residues from the +4 to +8 layers of Qa and the deletion of the +4 to +8 layers of both Qb and Qc, removing all capacity for hydrophobic collapse between the +4 and +8 layers of the R and Q SNAREs, is fully restored by Sec17 and Sec18. We propose that Sec17 either creates a favorable folding environment for the assembly of the remaining full-length SNARE domains or directly promotes bilayer remodeling through insertion of the apolar loops of several SNARE-bound Sec17s or acts by a combination of these two mechanisms.

Results
Vacuole SNAREs ( Figure 1A) have an N-domain and a SNARE domain. Several of them have a TM anchor, but Qc lacks a hydrophobic membrane anchor, and instead associates with membranes through its affinities for the other SNAREs, for HOPS (Stroupe et al., 2006), and for phosphatidylinositol 3-phosphate through its N-terminal PX domain (Cheever et al., 2001). SNARE domain layers are numbered from the central 0-layer, as shown. Fusion requires that R-and at least one Q-SNARE be anchored to docked membranes . When they are, soluble forms of the other Q-SNAREs without membrane anchors, termed sQ ( Figure 1A), will support Ypt7/HOPSdependent fusion. Vacuolar SNAREs with C-terminal truncations, corresponding to partial zippering, form stable complexes (Figure 1-figure supplement 1), supporting their use in fusion studies. Based on the single-particle cryo-EM (Electron Microscope) structure of the homologous human neuronal NSF/aSNAP/SNARE complex (Zhao et al., 2015), we modeled the associations of Sec17 (aSNAP) ( Figure 1B) and Sec18 (NSF) with vacuolar SNAREs, viewed in profile or in an end-on view from the membrane ( Figure 1C). In this model, we assume that four Sec17 monomers ( Figure 1B) assemble together, surrounding the 4-SNARE coiled coil ( Figure 1C). In contrast, only two aSNAP molecules have been observed in EM structures of the NSF/aSNAP/V7-SNARE complex (Zhao et al., 2015) and the NSF/aSNAP/SNARE complex that included the linker between the two SNAP-25 SNARE domains . The presence of the SNAP-25 linker in these two complexes may interfere with the binding of the other two aSNAP molecules. Since the vacuolar SNARE complex does not contain a linker between SNARE domains, it is reasonable to postulate that four Sec17 molecules bind to the vacuolar SNARE complex, but the precise number of Sec17 molecules is yet to be determined for the Sec18/Sec17/vacuolar SNARE complex. Rapid fusion needs Sec17 and Sec18 in addition to HOPS and SNAREs . To understand how they work together, we exploited direct assays of SNARE associations to show that incompletely zippered SNAREs can associate stably and that HOPS allows Sec17 to promote the completion of zippering and SNARE complex stability. We then examined their functional relationships to show that SNARE-bound Sec17 and Sec18 can promote rapid fusion when energy from zippering is greatly reduced or lost.  (Rice and Brunger, 1999, PDB ID 1QQE). Residues mutated in certain experiments are highlighted as in Schwartz et al., 2017. (C) Modeling of the vacuolar Sec18/Sec17/SNARE complex. MODELLER (Webb and Sali, 2016) was used to create individual homology models of the vacuolar SNARE complex (Nyv1, Vam3, Vti1, Vam7) and of Sec18 starting from the coordinates of synaptobrevin-2, SNAP-25, syntaxin-1A, and NSF in the cryo-EM structure of the neuronal 20S complex (PDB ID 3J96) (Zhao et al., 2015). These homology models, together with the crystal structure of Sec17 (PDB ID 1QQE) , were fit into the cryo-EM structure of the neuronal 20S complex (PDB ID 3J96) (Zhao et al., 2015). We used PDB ID 3J96 because this cryo-EM structure did not include the SNAP-25 linker and the H abc domain of syntaxin-1A. The vacuolar SNARE complex (Nyv1, Vam3, Vti1, Vam7) also does not include a linker between any of the SNARE motifs; in all structures of NSF/aSNAP/ternary SNARE complexes determined thus far, four aSNAP molecules are observed for SNARE complexes that do not contain a linker connecting two of the SNARE domains , and we therefore included four Sec17 molecules in our model of the vacuolar 20S complex. In the PDB coordinate file supplied as Source Data File of the homology model of the Sec18/Sec17/vacuolar SNARE complex, Sec18 molecules, chains A-F; Sec17, chains G-J; Nyv1, chain K; Vam3, chain L; Vti1, chain M; Vam7, chain N are included. Colors: cyan: Nyv1; magenta: Vam3; yellow: Vti1; salmon: Vam7; gray, orange, green, slate: Sec17; yellow, magenta, gray, blue, salmon, green: Sec18. Cartoon representations are shown. Two views related by a 90-degree rotation are included (left: side view; right: membrane-end view). Ó 2017, Schwartz et al. Figure 1B

Sec17 alters SNARE complex conformation
The capacity of vacuolar SNAREs to form stable partially zippered structures raised the question of whether these SNAREs can zipper efficiently, especially when anchored to membranes or associated with other fusion proteins such as HOPS. To study the kinetics of SNARE interactions, we employed an ensemble fluorescence resonance energy transfer (FRET) assay of vacuolar SNARE associations (Torng et al., 2020). The Qc-SNARE was prepared with a unique cysteinyl residue in any of three positions (Figure 2A), either the native Cys208, which is upstream (U) of the SNARE domain or, after substitution of serine for this cysteine, with Met250Cys at the N-terminal end of the SNARE domain or with Ala316Cys at the C-terminal end of the SNARE domain. Each was derivatized with Oregon Green 488. Fusion proteins were also prepared with a unique cysteine either immediately N-terminal, or C-terminal, to the Qb SNARE domain, and each was derivatized with Alexa Fluor 568. As a model for exploring the effects of HOPS and Sec17 on SNARE dynamics, these fluorescent proteins were co-incubated with proteoliposomes bearing Ypt7, R, and Qa. cis-SNARE complex assembly can occur spontaneously on these proteoliposomes, but assembly is stimulated by HOPS, allowing direct comparison of HOPS-dependent and HOPS-independent SNARE assembly (Torng et al., 2020). The average FRET efficiency in these studies is modest because they include, in bulk reactions, all the fluorescent Qb and Qc, many of which do not enter SNARE complexes. SNARE complex assembly with HOPS gave a high average FRET efficiency when fluorophores were at the N-terminus of the Qb SNARE domain and at or near the N-terminal end of the Qc SNARE domain ( Figure 2B, red and yellow curves). There was a lower FRET efficiency when the fluorophores were at opposite ends of the Qc and Qb SNARE domains ( Figure 2B; blue, green, and indigo). When both fluorophores were at the C-terminal ends of the Qb and Qc SNARE domains, a low signal was seen ( Figure 2B, purple), similar to the average FRET efficiency when fluorophores were at opposite ends of the SNARE domains, thus suggesting incomplete zippering. After 1 hr, Sec17 was added to each reaction. Strikingly, Sec17 only enhanced the average FRET efficiency between C-terminal fluorophores, rapidly rising to the level seen when the fluorophores were together at the N-terminii of the SNARE domains ( Figure 2B, purple; Figure 2-figure supplement 1), indicating a Sec17induced change at the C-terminal end of the SNARE complex. Sec17/aSNAP may promote the zippering of isolated SNARE domains (Ma et al., 2016) but is now seen to act in the context of membranes and HOPS.

Sec17-induced conformational change requires HOPS
Proteoliposomes with R, Qa, and Ypt7 (where indicated) were incubated with Qb-SNARE domain and Qc. Both SNARE domains were either fluorescently labeled at their N-termini ( Figure 2C, bars 1-5) or at their C-termini ( Figure 2C, bars 6-10). Incubations were in the presence or absence of HOPS. Sec17 addition after 1 hr did not enhance the average FRET efficiency between N-terminally disposed fluorophores in the presence or absence of HOPS ( Figure 2C, bars 1-5), but stimulated the average FRET efficiency between SNARE domain C-terminal fluorophores in a HOPS-dependent manner ( Figure 2C, bars 6-10), since the enhanced FRET between the SNAREs in the presence of HOPS and Sec17 (bar 9) is not seen without HOPS (bar 8) or without Sec17 (bar 10). This indicates a HOPS-dependent and Sec17-induced SNARE conformational change. This was diminished when zippering was inhibited by the absence of the +4 to +8 layers of sQa ( Figure 2D, bar 1 vs. 3) or by the conversion of each inward-facing apolar residue of the full-length Qa SNARE domain layers +4 to +8 to Gly ( Figure 2E, bar 1 vs. 3). The F22SM23S mutation of Sec17 (FSMS hereafter), diminishing the hydrophobicity of its N-domain loop , reduced the Sec17 capacity for inducing HOPS-dependent conformational change ( Figure 2F). We also examined the effects of other mutants of Sec17. The K159E,K163E mutation (KEKE hereafter) diminishes Sec17:SNARE association (Marz et al., 2003); one of these residues (Sec17 K159) is in a pair (aSNAP K122, K163) that abolishes disassembly of the neuronal SNARE complex by NSF/aSNAP (Zhao et al., 2015). The Source data 1. Source data file (PDB) for Figure 1C.  (A) Schematic of fluorescently labeled SNARE constructs used in this study. SNAREs were derivatized as described previously (Torng et al., 2020). Wild-type Qc contains a single Cys residue at 208 at the upstream (U) position. (N)-and (C)-labeled constructs replace residues 250 and 316 with Cys, while also replacing Cys208 with Ser. Each Qc construct was derivatized with Oregon Green 488 as described (Torng et al., 2020). A fusion of maltosebinding protein (MBP), a TEV site, and the Qb SNARE domain (residues 133-187) was expressed with an added cysteinyl residue immediately upstream or downstream of the SNARE domain. Each Qb construct was derivatized with Alexa Fluor 568. Qc and Qb labeled with a fluorescent probe at any position are written as *Qc and *Qb. (B-G) Bar graphs are reported as the mean of the relative ensemble fluorescence resonance energy transfer (FRET) change (%) per trial with propagated standard deviation for n = 3 trials. The relative change was calculated by averaging 10 data points, each from just before Sec17 addition and from the end of the measurement period 20 min later, except where indicated. Specific time points used as well as the statistics for the propagation of uncertainty are shown in Supplementary Data. A bar graph representation for (B) and the kinetic curves for (C-G) are provided in Figure 2-figure supplement 1 . (B) Sec17 modifies the conformation of the C-terminus of the SNARE complex. Ypt7/RQa proteoliposomes were incubated with pairs of *Qb and *Qc labeled at the N, C, or upstream (U) locations as indicated in (A). Curves are averages of n = 3 trials, and the shaded regions behind each curve show the standard deviation per time point. (C) Sec17-promoted zippering requires HOPS. RQa proteoliposomes were incubated with either the N-labeled *Qb/*Qc pair (left) or the C-labeled *Qb/*Qc pair (right), with Ypt7 and HOPS as indicated. A reaction with a buffer (RB150) added instead of Sec17 serves as a negative control. (D) Sec17 does not promote C-terminal zippering if the SNARE domain of Qa is truncated. Ypt7/R proteoliposomes were incubated with C-terminally labeled *Qb and *Qc and either soluble Qa or Qa3D, and the relative FRET change was calculated over 10 min after Sec17 or buffer addition. (E) Sec17-induced zippering requires the apolar heptad-repeat amino acyl residues in Qa SNARE domain layers +4 to +8. Proteoliposomes bearing Ypt7, R-SNARE, and either wild-type Qa or Qa with the +4 to +8 layers inwardly-oriented apolar residues mutated to Gly were incubated with C-terminally labeled *Qb and *Qc, and then Sec17 or its mutants were added at t = 60 min. (F) Sec17-driven SNARE conformational change is stunted by the F22SM23S mutation of Sec17 (FSMS). Ypt/R proteoliposomes were incubated with sQa and C-terminally labeled *Qb and *Qc, and then Sec17 or mutants as indicated were added at t = 60 min. (G) Sec17 stabilizes incompletely zippered SNARE complexes. Ypt7/R proteoliposomes were incubated with sQa and C-terminally labeled *Qb and *Qc. Sec17 or the indicated mutants were added at t = 60 min. Non-fluorescent Qc (8.5 mM) was added at t = 80 min, and the loss of FRET over 35 min was measured starting immediately after the addition of non-fluorescent Qc. Figure 2 continued on next page C-terminal L291A,L292A mutation of Sec17 (LALA hereafter) interferes with its cooperation with Sec18 for SNARE complex disassembly (Barnard et al., 1997;Zick et al., 2015), and 6AtoN is the conversion of six inward-facing acidic residues of Sec17, which face basic SNARE residues in the 20S structure ( Figure 1C), to neutral residues. Neither KEKE, LALA, nor 6AtoN had a large effect on the capacity of Sec17 to promote HOPS-dependent conformational change ( Figure 2F).
Since this Sec17-induced SNARE conformational change is seen with C-terminal fluors but not with N-terminal fluors, requires HOPS as the SNARE assembly catalyst, and requires SNAREs that can zipper, a major part of the conformational change may be zippering itself. This might be spontaneous after Sec17-induced release of HOPS from SNAREs (Collins et al., 2005;Schwartz et al., 2017) or be promoted by Sec17 binding along the SNAREs ( Figure 1C), reflecting its affinity for the C-terminal region of SNARE bundles (Ma et al., 2016). HOPS bound to SNAREs may inhibit C-terminal zippering, even though it helps to assemble the N-terminus of the four-helix bundle. Sec17 could then increase FRET because it displaces HOPS. Such C-terminal zippering would be favored by Sec17 binding, but could also be spontaneous in the absence of these factors.
Sec17 also interacts with partially zippered SNARE complex to promote SNARE complex stability. SNARE complex was assembled by HOPS on Ypt7/R proteoliposomes with soluble Qa, with the Qb SNARE domain labeled with a fluorophore at a cysteinyl residue upstream of the SNARE domain, and with Qc of full length (w.t.) or with the 3D C-terminal truncation, each bearing a fluorophore at its native cysteinyl residue upstream of the SNARE domain. When the complex of proteoliposomes with these fluorescent Qb and Qc had full-length SNARE domains, it was stable whether or not it included Sec17, as there was no loss of average FRET efficiency after addition of excess non-

Figure 2 continued
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Source data file (Excel) for Figure 2B.   Source data 1. Source data file (Excel) for Figure 3A,B and C. In contrast, fluorescent Qc3D was 'chased' by exchange with non-fluorescent Qc (bar 3), but Sec17 stabilized this SNARE complex, blocking the chase (bar 4). Thus, in the absence of Sec17, the assembly of Qc3D into partially zippered SNARE complex is reversible. Each domain of Sec17 helps to stabilize Qc3D against exchange (bars 5-8), especially the Sec17 N-terminal apolar loop (bar 5). The HOPS-dependent functions of Sec17, such as promotion of zippering, may be aided by the direct affinity between these proteins (Figure 2-figure supplement 2).

Generality of Sec17 and Sec18 bypass of arrested zippering
Sec17 can restore fusion when Qc has truncations at the C-terminal end of its SNARE domain , stimulated by Sec18 (Schwartz et al., 2017). We asked whether Sec17 can restore fusion with similar deletions of residues after the +3 layer in the other Q-SNAREs. Proteoliposomes bearing Ypt7 and R-SNARE were assayed for fusion with proteoliposomes bearing this Rab and any two anchored Q-SNAREs. Proteoliposome mixtures were incubated with HOPS, and the soluble form of the remaining Q-SNARE was deprived of its membrane anchor and a C-terminal portion of its SNARE domain (Figure 3; A, Qc3D; B, sQb3D; C, sQa3D). As reported by Schwartz et al., 2017, there was no fusion with Qc3D unless 600 nM Sec17, 300 nM Sec18, and ATP or ATPgS were present ( Figure 3A); these concentrations are in the physiological concentration ranges of Sec17 (150-1100 nM) and Sec18 (250-760 nM) (Ho et al., 2018). While ATP and its nonhydrolyzable analog ATPgS support comparable fusion with wild-type SNAREs , hydrolyzable ATP inhibits fusion through Sec17/Sec18-mediated disassembly of SNARE complexes when defective SNAREs such as Qc3D are present, a proofreading function. The same pattern was seen for fusion with sQb3D ( Figure 3B) and sQa3D ( Figure 3C). The unique spatial disposition of the Sec17/aSNAP molecules with respect to each SNARE ( Figure 1C and Zhao et al., 2015) makes it unlikely that Sec17 could somehow fill each of the gaps left by each of these deletions to shield apolar residues within the SNARE bundle and thereby continue to drive zippering, or that Sec17 binding could induce the remaining R and two Q +4 to +8 layers to somehow rotate to form a hydrophobic two-or three-layered core.
Fusion assays were also performed with Ypt7/R proteoliposomes and each of the three Ypt7/single-anchored Q-SNARE proteoliposomes in the presence of HOPS and the other two soluble Q-SNAREs ( Figure 4). With membrane-anchored Qa and with sQb and Qc having complete SNARE domains, there was HOPS-dependent fusion without further addition ( Figure 4A, black line), though Sec17 and Sec18 with AMP-PNP, ATPgS, or ATP did stimulate (compare black curves, A-D). Deletion of the +4 to +8 layers from either soluble Qb or Qc abolished fusion ( Figure 4A), which was restored by Sec17 and Sec18 with either AMP-PNP, ATPgS, or ATP (B-D, red and blue curves). There was no fusion when both soluble Q-SNAREs had truncated SNARE domains ( Figure 4A, orange), but, strikingly, the fusion was partially restored by Sec17 and Sec18 with ATP ( Figure 4D, orange) and more fully restored when the adenine nucleotide was resistant to hydrolysis ( Figure 4, B and C, orange). With two Q-SNAREs lacking the C-terminal portion of their SNARE domains, the apolar amino acyl residues of the remaining two SNAREs would not be as effectively shielded from water if they continued zippering together. Fusion could not be restored by Sec17 and Sec18 if either soluble SNARE was omitted instead of truncated (green and purple). This fusion with sQb3D and Qc3D occurs in stages, initially sensitive to antibody to either the HOPS SM subunit Vps33 or Sec18, then acquiring resistance to Vps33 antibody while remaining sensitive to the Sec18 ligand (Figure 4-figure supplement 2). When only the Qb-SNARE was membrane anchored, there was little fusion without Sec17 and Sec18 ( Figure 4E, black curve). When the SNARE domain of sQa or Qc had been truncated, fusion was strictly dependent on non-hydrolyzable ATP analogs, and little fusion was seen with dual SNARE domain truncation. Similar patterns were seen with anchored Qc (Figure 4, I-L). We term the fusion induced by Sec17 and Sec18 in the presence of 3D SNARE domain truncations 'zippering bypass fusion'.
To explore the capacity of Sec17 and Sec18 to compensate for partial SNARE zippering, we assayed the fusion of Ypt7/R and Ypt7/Qa proteoliposomes with sQb3D, Qc3D, and HOPS, using various concentrations of wild-type or mutant Sec17, and with or without Sec18 and ATPgS (Figure 5). Without Sec17, fusion is not supported by Sec18 (A and B, blue curves). Limited fusion is possible with 1 or 2 mM wild-type Sec17 alone (A, black and red), but Sec18 allows faster fusion and with less Sec17 (A and B; tan). Fusion requires the apolar loop near the N-terminus of Sec17, as the F21S,M22S mutation (FSMS) blocks fusion entirely (C and D). The K159E,K163E mutation (KEKE) diminishes Sec17:SNARE association (Marz et al., 2003). The KEKE mutation prevents fusion without Sec18 (E), but a slow and limited fusion with KEKE-Sec17 is restored by Sec18 (F). The C-terminal L291A,L292A mutation of Sec17 (LALA), which interferes with its cooperation with Sec18 for SNARE complex disassembly (Barnard et al., 1997;Zick et al., 2015), diminishes zippering bypass fusion (A vs. G, red curves), and a limited fusion is restored through the addition of Sec18 (H). These data suggest that Sec17 action directly requires its apolar loop domain, since the loss of this apolar region is not bypassed by Sec18. Sec18 may stimulate fusion by modulating the conformation of Sec17 associations with trans-SNARE complexes, but Sec18 is not simply promoting Sec17 binding, since it is still needed for fusion when Sec17 is joined to an integral N-terminal membrane anchor ( Figure 5-figure supplement 2). Basic residues in the +3 to +8 layers near the C-terminus of the R and Qa SNARE domains are near acidic residues on the interior of the Sec17 assembly ( Figure 5I). To determine whether Sec17 might rely on these ionic interactions to support      fusion, we converted these acidic residues of Sec17 to alanine or serine, creating the mutant Sec17-E34S,E35S,D38S,E73A,D74A,E75A (termed Sec17 6 Acidic to Neutral or Sec17-6AtoN), but this mutant Sec17 still supports fusion ( Figure 5, J and K). Interestingly, mutation of acidic residues of aSNAP near the C-terminal end of the neuronal SNARE complex also only had a modest effect on disassembly activity of NSF/aSNAP (Zhao et al., 2015).

Fusion despite triply-crippled SNARE zippering
Since SNARE zippering is driven by the sequestration of apolar residues into the interior of the 4-SNARE bundle, we examined the effect of converting the apolar residues of the Qa SNARE domain +4 to +8 layers to Ala, Ser, or Gly. Fusion between Ypt7/R and Ypt7/Qa proteoliposomes in the presence of HOPS, sQb, and Qc, but without Sec17 or Sec18 ( Figure 6A, black curve), was diminished by replacing each of the +4 to +8 layer apolar residues of Qa with Ala ( Figure 6A, green curve) and was abolished when they were replaced by Ser (red curve) or by Gly (blue curve). The persistence of some fusion with the Ala substitutions may reflect that two of the residues were already Ala, that Ala has the greatest propensity among the amino acids to form a-helices, Gly the least, and Ser is in-between (Pace and Scholtz, 1998), and that alanine itself is the least hydrophilic of these three amino acids. When these same incubations were performed with Sec17, Sec18, and ATPgS, rapid and comparable fusion was seen in each case ( Figure 6B). When hydrolyzable ATP was Figure 6. Sec17, Sec18, and ATPgS restore fusion to Ypt7/R and Ypt7/Qa proteoliposomes, which were triply-crippled from completion of SNARE domain zippering by deletion of the +4 to +8 layers of the sQb and Qc SNAREs and by substitution of the apolar residues of the +4 to +8 layers of Qa, substituting Ala, Ser, or Gly for L238, M242, A245, L249, and A252.  used instead of ATPgS, there was little effect on the fusion kinetics with wild-type SNARE domain sequences ( Figure 6, B vs. C, black curves). In contrast, hydrolyzable ATP led to fusion inhibition when SNARE packing stability was reduced by substitution of apolar residues by Ala, Ser, or Gly ( Figure 6C). Though the apolar residues are not required for fusion aided by Sec17 and Sec18 ( Figure 6B), they apparently stabilize the SNAREs against ATP-driven proofreading disassembly ( Figure 6C). To triply weaken the completion of zippering, the same proteoliposomes with wild-type Qa or the Qa with +4 to +8 layers having small side-chain residues instead of apolar residues were incubated with HOPS, sQb3D, and Qc3D, either without Sec17 or Sec18 ( Figure 6D) or with Sec17, Sec18, and ATPgS ( Figure 6E) or ATP ( Figure 6F). Fusion was optimally supported by Sec17, Sec18, and ATPgS ( Figure 6E). The independence of this fusion from energy derived by zippering is underscored by the similar fusion rates in all incubations with Sec17, Sec18, and ATPgS, whether with apolar or polar Qa +4 to +8 residues or with full-length or +4to +8-truncated sQb and Qc ( Figure 6, B vs. E). With the Qb and Qc SNARE domains truncated, and the Qa lacking apolar inward-facing amino acyl side chains, little or no energy could be gained from the completion of R and mutant-Qa zippering. Thus, Sec17 acts in three ways: triggering a zippering-dependent SNARE conformational change in the presence of HOPS and full-length SNARE domains (Figure 2), acting with Sec18 to promote fusion independent of energy from zippering (Figure 6), and supporting the disassembly of post-fusion cis-SNARE complexes by Sec18.
Fusion can be supported by either complete 4-SNARE zippering without Sec17 or Sec18, or by Sec17 and Sec18 association with only partially zippered SNAREs, but the most rapid fusion requires both . While Sec17 and Sec18 are known to bypass the fusion blockade by Qc C-terminal truncation alone Schwartz et al., 2017), the generality of this bypass with respect to any one Q-SNARE or even two Q-SNAREs (Figures 3 and 4) shows that it is not specific to Qc alone. Moreover, when zippering with full-length SNARE domains is weakened by the substitution of small amino acyl residues for large apolar residues in the +4 to +8 layers of Qa, Sec17 and Sec18 will also restore fusion ( Figure 6A,B). Strikingly, Sec17 and Sec18 drive fusion despite a triple blockade to complete zippering, namely the absence of two C-terminal Q-SNARE domains and the lack of apolarity in the third ( Figure 6E). In the NSF/aSNAP/SNARE complex, the 4-SNAREs wrap around each other in a left-handed super helix, and the Sec17s wrap around them in a right-handed fashion, yet they form a specific structure (Zhao et al., 2015;White et al., 2018). It seems unlikely that residues from one or more Sec17 could substitute for the missing residues when two heptads are removed from the C-termini of one or even two Q-SNAREs and the bulky apolar residues are removed from the third.
From the earliest reconstitutions of HOPS-dependent fusion  and in subsequent studies (e.g., Zick et al., 2015), Sec17 and Sec18 gave strong stimulation. In contrast, Sec17 and Sec18 inhibit SNARE-only fusion, or fusion with non-physiological tethers Zick and Wickner, 2014;Schwartz et al., 2017;Song and Wickner, 2019). HOPS not only binds each SNARE, but also has direct affinity for Sec17 (Figure 2-figure supplement 2). In model studies with cis-SNARE complexes, HOPS is necessary for Sec17 to enhance zippering per se ( Figure 2C). HOPS:Sec17 binding may underlie their synergy for fusion, but further studies are needed to test this idea.
Earlier studies and our current work suggest a working model of vacuole membrane fusion, encompassing findings here and elsewhere (Figure 7). HOPS exploits the affinity of its Vps39 and Vps41 subunits for the Rab Ypt7 (Brett et al., 2008) on each fusion partner membrane ( Figure 7A) to mediate (Step 1) tethering (Hickey and Wickner, 2010). Tethered membranes ( Figure 7B) are a prerequisite for SNARE assembly in an active conformation, likely a common N to C SNARE domain orientation (Song and Wickner, 2019). HOPS has direct affinity for each of the four vacuolar SNAREs  and is allosterically activated by vacuolar lipids and Ypt7:GTP (Torng et al., 2020) as a catalyst of SNARE assembly (Step 2). SNAREs begin to zipper ( Figure 7C) from the N-toward the C-terminal end of their SNARE domain. When any one Q-SNARE is omitted, fusion intermediates assemble, which undergo very rapid fusion when the missing Q-SNARE is supplied , suggesting that fusion without Sec17/Sec18 is rate-limited by the completion of SNARE zippering and/or the spontaneous release of bulky HOPS. Sec17 has direct affinity for SNAREs (Sö llner et al., 1993), HOPS (Figure 2-figure supplement 2), lipids (Clary et al., 1990;Zick et al., 2015), and Sec18 (Sö llner et al., 1993). Sec17 displaces HOPS from the SNAREs (Figure 7, Step 3), as shown by earlier studies: (1) vacuolar SNAREs are found in complex with Sec17 or HOPS, but not with both, and Sec17 can displace HOPS from SNAREs (Collins et al., 2005); (2) when vacuolar HOPS-dependent reconstituted fusion is arrested by Qc3D, HOPS is bound to the incompletely zippered SNARE complex until it is displaced by Sec17 (Schwartz et al., 2017); and (3) trans-SNARE complexes, which assemble between isolated vacuoles, are largely associated with Sec17 (Xu et al., 2010). The Sec17 association with partially zippered trans-SNARE complex ( Figure 7D) promotes a conformational change (Figure 2) leading to complete zippering. The Sec17: 4-SNARE complex will bind Sec18 (Step 4) through its affinities for both SNAREs  and for Sec17. Sec18 regulates, in some unknown fashion, the Sec17/a SNAP assembly into an oligomeric structure surrounding the SNARE complex intermediate (Zhao et al., 2015), a trans-anchored Sec18/Sec17/SNARE pre-fusion complex ( Figure 7E). Some ATP hydrolysis-dependent disassembly can occur, which may represent proofreading of incorrect and unstable trans-SNARE complexes . Sec17 oligomerization may also be stabilized or guided by the insertion of its N-terminal loop into the membranes. While SNAREs can slowly complete zippering and support fusion without Sec17 or Sec18 , and slow fusion can occur with Sec17 and Sec18 when sQb and Qc lack the C-terminal portion of their zippering domain ( Figure 4D), optimal fusion requires four complete SNARE domains, Sec17/aSNAP, and Sec18/NSF. The energy for fusion (Step 5) derives from multiple sources: the completion of SNARE zippering, the binding energies which create the Sec17 structure surrounding the SNAREs, and the energy of bilayer distortion through Sec17 apolar loop insertion. Cis-SNARE complexes ( Figure 7F) formed by fusion (Sö llner et al., 1993;Mayer et al., 1996) are disassembled by Sec17, Sec18, and ATP (Step 6), freeing SNAREs from each other ( Figure 7A) for later assembly in trans. Each of these components -the four SNAREs, the Rab Ypt7, HOPS, Sec17, and Sec18 -is part of this ordered pathway of associations and disassembly, constituting a holoenzyme for vacuole membrane fusion.
Our current studies reveal a general capacity of Sec17/aSNAP and Sec18/NSF to support fusion, even when little or no energy would be derived from completion of zippering. Sec17 provides a cage-like environment (Chakraborty et al., 2010), albeit with side fenestrations (Schwartz et al., 2017), which may favor SNARE zippering or, where zippering is blocked, allow the remaining SNARE domains to attain positions and conformations that approximate zippering. The apolar N-domains of the four Sec17s are clearly essential for fusion ( Figure 5 and Figure 5-figure supplement 2), whether through positioning the Sec17s, facilitating their assembly, or directly inserting as a 'membrane wedge' and thereby contributing to bilayer disruption at the fusion site. The continued need for this apolar loop when Sec17 is integrally membrane-anchored ( Figure 5-figure supplement 2) shows that it does not simply contribute to Sec17 membrane association. Elements of 'Sec17 cage' and 'membrane wedge' action are not mutually exclusive. Further work will be needed to determine the relative energies of Sec17 and Sec18 binding to the assembling 4-SNARE trans-complex, each of the four Sec17s inserting its apolar loops into the membrane, Sec17 forming lateral associations with other Sec17 molecules in the cage around the SNAREs, and complete SNARE zippering, as each of these helps to achieve the bilayer distortions of fusion.
Our current studies further our understanding of how Sec17(aSNAP), Sec18 (NSF), and SNAREs cooperate to promote membrane fusion. Sec17 has multiple functions: (a) It supports Sec18 association with SNARE complexes for their ATP-driven disassembly for subsequent rounds of fusion (Ungermann et al., 1998;Choi et al., 2018). Sec17/aSNAP and Sec18/NSF are also part of a quality control system that, for vacuoles, includes HOPS (Starai et al., 2008) and, in the neuronal system, includes Munc18 and Munc13 (Ma et al., 2013;Lai et al., 2017). These disassembly reactions require ATP hydrolysis and can be interrupted by the Sec17 LALA mutation. (b) Sec17 promotes a conformational change that leads to complete SNARE zippering. While neuronal SNAREs, which are properly assembled (i.e., involving the Munc18/Munc13 pathway), will fully zipper with a high energy yield, this may not be true for all non-neuronal SNAREs. (c) Sec17 promotes fusion even when SNARE zippering is incomplete. Sec17 uses the partially zippered SNAREs as a platform to bind to the fusion site and its apolar N-domain loop to trigger fusion. Sec18 also has multiple functions: (a) Sec18/NSF drives ATP-driven SNARE disassembly (Zhao et al., 2015), blocked by the Sec17 LALA mutation. (b) Sec18 supports Sec17 for direct promotion of fusion, built on a platform of partially zippered SNAREs. Note that the loading of SNARE N-terminal residues into the pore of the D1 ring of NSF does not require hydrolysis , so initial assembly of the NSF/a SNAP/SNARE complex occurs in the absence of hydrolysis. This does not need ATP hydrolysis and is insensitive to the LALA mutation. Thirdly, SNAREs also have multiple functions: (a) They can completely zipper, thereby stressing the bilayer and triggering fusion (Weber et al., 1998;Sutton et al., 1998). As previously reported , HOPS alone, without Sec17 or Sec18, will support slow SNARE-dependent fusion. (b) SNAREs support the assembly of a microdomain in which multiple fusion proteins and lipids become highly enriched (Fratti et al., 2004). (c) As shown here, SNAREs form an essential platform for Sec17 and Sec18 to contribute to fusion independently of completion of zippering. With physiological levels of wild-type vacuolar SNAREs, rapid fusion requires Rab-activated HOPS, Sec17, Sec18, and ATP.
While intracellular fusion reactions share many requirements, such as for SNAREs and SM protein, there are major differences as well. Synaptic vesicle fusion and other calcium-dependent secretion systems require calcium, synaptotagmin, Munc13, and complexin, none of which have their obvious counterparts in calcium-independent systems. Vacuole and endosome fusion have their SM protein as an integral subunit of the tethering complex, which is not seen in other organelles. The similarities and differences in fusion pathways at each organelle will be clarified as each is more thoroughly studied. PI3P was from Echelon (Salt Lake City, Utah), ergosterol from Sigma (St. Louis, MO), fluorescent lipids from Thermo Fisher (Waltham, MA), and other lipids were from Avanti (Alabaster, AL). Biobeads SM2 were from BioRad, Cy5-Streptavidin from SeraCare (Milford, MA), biotinylated phycoerythrin from Invitrogen (Eugene, OR), and underivatized streptavidin from Thermo Fisher. Spectrapor six dialysis tubing (7.5 mm diameter, 25 kDa cutoff) was from Spectrum Labs (Las Vegas, NV). Octylb-D-glucopyranoside was purchased from Anatrace (Maumee, OH).
For inverse PCR of pParallel1 GST and Qa 1-227, Vam7 mutants with cysteines inserted at the N-and C-termini of the SNARE domain were generated by inverse PCR with the Vam7 intein vector ) and Phusion high-fidelity DNA polymerase (NEB). First, the native cysteine was removed by mutating it to serine (C208S) using the mutant primer set below. Vam7 mutants with cysteines inserted near the N-and C-termini of the SNARE domain, M250C and A316C, respectively, were generated from the cysteine-lacking plasmid in the same fashion.

Proteoliposome fusion
Proteoliposomes were prepared by detergent dialysis from b-octylglucoside-mixed micelles for fusion assays Song and Wickner, 2019) and SNARE assembly assays (Torng et al., 2020) as described. Briefly, for the fusion assay, proteoliposomes (1 mM lipid) were prepared with membrane-bound Ypt7 and R at 1:8000 and 1:16,000 molar ratios to lipid, respectively, and with lumenal biotin-phycoerythrin. Proteoliposomes were also prepared with membranebound Ypt7 and the indicated Q-SNAREs at 1:8000 and 1:16,000 molar ratios to lipid, respectively, plus lumenal Cy5-streptavidin. These were incubated separately for 10 min at 27˚C with 1 mM GTP and 1 mM EDTA, and then MgCl 2 was added to 2 mM. After prewarming (27˚C for 10 min) the separately GTP-exchanged proteoliposomes in a 384-well plate, fusion reactions were initiated by mixing 5 ml of each proteoliposome preparation and supplementing with other fusion factors in volumes summing to 10 ml, continuing incubation at 27˚C in a Spectramax fluorescent plate reader. Fusion incubations (20 ml) in RB150 (20 mM HEPES/NaOH, pH 7.4, 150 mM NaCl, 10% glycerol) had proteoliposomes (0.5 mM total lipid concentration), 50 nM HOPS, the indicated concentrations of sSNAREs, 400 or 600 nM Sec17, 300 nM Sec18, 1 mM ATP or its analogs, and 3 mM MgCl 2 , as modified in each figure legend.

SNARE assembly assay
Assays were performed as described previously (Torng et al., 2020) with one addendum. In brief, reactions (20 ml) were performed at 27˚C in a SpectraMax Gemini XPS (Molecular Devices) plate reader. Standard reactions include HOPS (160 nM), proteoliposomes (0.5 mM lipid, with SNARE and Ypt7 at a molar ratio of either 1:2000 or 1:4000 for Ypt7/R proteoliposomes and Ypt7/ RQa proteoliposomes, respectively), and fluorescently labeled Qb and Qc (1 mM), and sQa (1 mM) as necessary. These were incubated for 60 min, and then Sec17 was added to 500 nM. Three fluorescence channels were read simultaneously at intervals of 30 s: the donor channel Oregon Green 488 from Qc (excitation [ex]: 497 nm; emission [em]: 527 nm; cutoff [c/o]: 515 nm), the acceptor channel Alexa Fluor 568 from Qb (ex: 568 nm; em: 605 nm; c/o: 590 nm), and the FRET channel (ex: 490 nm; em: 615 nm; c/o: 590 nm). For each time point, the bleed through-corrected FRET signal was obtained by subtracting the background signals coming from the donor and acceptor channels from the signal in the FRET channel as detailed in Torng et al., 2020. This was further corrected by dividing by the geometric mean of the donor and acceptor signals. The final corrected signal, reported as 'Average FRET efficiency,' is a combined measure of the proportion of fluorescent SNAREs undergoing FRET and their average FRET efficiency.

Molecular models
MODELLER (Webb and Sali, 2016) was used to create individual homology models of the vacuolar SNARE complex (Nyv1, Vam3, Vti1, Vam7), and of Sec18 starting from the coordinates of synaptobrevin-2, SNAP-25, syntaxin-1A, and NSF in the cryo-EM structure of the neuronal NSF/aSNAP/ SNARE complex (PDB ID 3J96) (Zhao et al., 2015). For Sec18, the linker between the N and D1 domains was deleted from the generated homology model since there was no information about these linkers in this structure (PDB ID 3J96) of the neuronal 20S complex.