Stringent 3Q·1R Composition of the SNARE 0-Layer Can Be Bypassed for Fusion by Compensatory SNARE Mutation or by Lipid Bilayer Modification*

SNARE proteins form bundles of four α-helical SNARE domains with conserved polar amino acids, 3Q and 1R, at the “0-layer” of the bundle. Previous studies have confirmed the importance of 3Q·1R for fusion but have not shown whether it regulates SNARE complex assembly or the downstream functions of assembled SNAREs. Yeast vacuole fusion requires regulatory lipids (ergosterol, phosphoinositides, and diacylglycerol), the Rab Ypt7p, the Rab-effector complex HOPS, and 4 SNAREs: the Q-SNAREs Vti1p, Vam3p, and Vam7p and the R-SNARE Nyv1p. We now report that alterations in the 0-layer Gln or Arg residues of Vam7p or Nyv1p, respectively, strongly inhibit fusion. Vacuoles with wild-type Nyv1p show exquisite discrimination for the wild-type Vam7p over Vam7Q283R, yet Vam7Q283R is preferred by vacuoles with Nyv1R191Q. Rotation of the position of the arginine in the 0-layer increases the Km for Vam7p but does not affect the maximal rate of fusion. Vam7Q283R forms stable 2Q·2R complexes that do not promote fusion. However, fusion is restored by the lipophilic amphiphile chlorpromazine or by the phospholipase C inhibitor U73122, perturbants of the lipid phase of the membrane. Thus, SNARE function as regulated by the 0-layer is intimately coupled to the lipids, which must rearrange for fusion.

SNARE 4 proteins (1) are vital for membrane fusion. Initially discovered in neuronal tissues, SNAREs are found on all organelles of the exocytic and endocytic pathways, from yeast to humans. Their characteristic feature is a heptad-repeat "SNARE domain," flanked by varied N-domains and by C-terminal membrane anchors, either a single membrane-spanning apolar polypeptide or an acyl anchor. SNAREs form 4-helical complexes through their SNARE domains (2). Although the residues in each SNARE that face the others in a 4-helical complex are generally apolar, 3 glutamine (Q) and 1 arginine (R) near the center of the 4-complexed SNARE domains form a conserved and polar "0-layer" (3). Recombinant neuronal SNAREs spontaneously assemble into stable bundles, yet SNARE complex assembly in vivo requires SNARE-binding proteins of the Sec1-Munc18 "SM" family. SNARE complexes are disassembled by two chaperones: Sec17p/␣-SNAP, which binds directly to SNARE complexes, and Sec18p/NSF, which binds to Sec17p/␣-SNAP and couples the energy of ATP binding and hydrolysis to SNARE complex disassembly (4). SNARE complexes form in cis, with each SNARE anchored to the same membrane, or in trans, with SNAREs anchored to apposed, "tethered" membranes. SNARE function has been studied in vivo, on isolated organelles, and through the reconstitution of recombinant SNAREs into proteoliposomes. When several SNAREs, which are found in vivo on a "target" membrane, are co-reconstituted into one population of liposomes, they form a t-SNARE complex. These liposomes can selectively interact with other liposomes bearing a reconstituted "vesicle," or v-, SNARE to allow selective lipid mixing (5), vesicle content mixing (6), or lysis (7,8). This reconstituted reaction shows impressive specificity for cognate SNARE pairs (9) and can be directly promoted by other fusion-regulatory factors such as synaptotagmin and calcium (10) or Sec1p (11) under conditions that minimize lysis (12).
SNAREs are required for the homotypic fusion of yeast vacuoles (13). Vacuoles fission and fuse during cell division and organelle inheritance (14) and in response to growth medium osmolarity (15). Purified vacuoles undergo a multistep fusion pathway, which can be monitored by content-mixing assays (16). In the initial "priming" step, Sec18p/Sec17p disassemble cis-SNARE complexes (17). The following tethering stage of "docking" requires the Rab-family GTPase Ypt7p (18) and its effector complex, termed HOPS (homotypic fusion and vacuole protein sorting) (19), which has Vps11p, Vps16p, Vps18p, Vps33p, Vps39p, and Vps41p as subunits (20,21). Tethered vacuoles are drawn against each other at closely apposed discshaped microdomains, termed "boundary membranes" to distinguish them from the "outside membranes," which are not in contact (22). The proteins (Ypt7p, HOPS, and SNAREs) and lipids (phosphoinositides, ergosterol, and diacylglycerol), which are required for fusion undergo striking, interdependent enrichment at a ring-shaped microdomain, termed the "vertex ring," which surrounds the boundary membrane (22)(23)(24)(25). Vacuole SNAREs form trans-complexes between the apposed organelles (17). Ypt7p and HOPS are needed for SNARE complex assembly, and HOPS is an integral part of the assembled SNARE complex (26). Fusion occurs around the vertex ring, joining the outside membrane from each vacuole to form the larger, fused organelle and joining the two apposed boundary membrane discs to yield a luminal vesicle (22,27).
Vacuoles have four SNAREs; their roles have been studied in the organelle-based fusion reaction (26,28) and in liposomebased studies (29). Vam3p, Vam7p, and Vit1p are Q-SNAREs, which constitute the t-SNARE, and the R-SNARE Nyv1p can serve as the v-SNARE. Three of these have C-terminal transmembrane anchors, but Vam7p has no apolar membrane anchor. Whereas most purified SNAREs require detergent for their solubility, and thus cannot be directly added to purified organelles without causing lysis, recombinant Vam7p is watersoluble without detergent and supports vacuole fusion (30).
Although SNAREs are central to fusion, it remains unclear how they act. SNAREs may apply torque to membranes (31,32), destabilize bilayers through poorly fitting TM domains (33,34), gather fusogenic lipids (e.g. diacylglycerol) into fusion microdomains (25), form the walls of a fusion pore (35), or act by other means. Studies of integral membrane SNAREs in the context of their native membrane and organelle are limited to producing and characterizing mutants, which may have significant effects on organelle composition and cell growth. The vacuolar Vam7p SNARE affords a unique window into SNARE function, because its cloned recombinant form is functional for fusion in vitro, it can readily be expressed and purified in soluble recombinant form, and many subreaction assays of vacuole fusion are available. Vam7p is a Q-SNARE with an N-terminal PX (Phox homology) domain, which directly binds phosphatidylinositol 3-phosphate (36) and HOPS (19), and a C-terminal SNARE domain. We now compare the functions of wild-type Vam7p with Vam7p bearing a Q283R mutation, changing the 0-layer Gln to Arg, in the context of purified vacuoles with the wildtype R-SNARE Nyv1p or with its 0-layer Arg changed to Gln. We find that the 3Q⅐1R rule (37) is important but not inviolate. Substantial fusion is seen with SNARE complexes having "rotated" positions of the 3Q⅐1R or with 4Q complexes, each showing dramatically elevated K m for Vam7p but similar maximal rates of fusion. 2Q⅐2R SNAREs form SNARE complex without fusion, but substantial fusion can be restored by the addition of chlorpromazine or U73122. Either of these agents modifies the lipid bilayer, suggesting that 0-layer function is directly coupled to the lipid bilayer properties.

RESULTS
Because Vam7p has no apolar membrane anchor, it can be produced in bacteria, purified in the absence of detergent, and added to in vitro fusion assays. To assess the role of the Vam7p C-terminal SNARE domain, we mutated its zero-layer glutamine to arginine (Q283R) to disrupt the 3Q⅐1R ratio. As a first test of this mutant Vam7p, we mixed bacterially expressed  (43). Although this inhibition requires all three of the Q-SNARE-soluble domains, it is simpler than the physiological SNARE complex assembly, because it bypasses regulation by Ypt7p (43). The inhibitory effect of the mixed soluble domains of the t-SNAREs was undiminished by violation of the 3Q⅐1R rule or by another mutation, which we characterize in detail elsewhere, 5 that disrupts the affinity of Vam7p for phosphatidylinositol 3-phosphate (Fig. 1). Because Vam7p Q283R can still interact with other SNAREs to regulate fusion, it warranted further characterization.
Fusion Activity of Vam7 Q283R Protein-Although the other, integral-membrane SNAREs are present on isolated vacuoles in the unpaired state as well as in SNARE complexes, all the Vam7p on purified vacuoles appears to be in cis-SNARE complexes (26). In standard fusion reactions with ATP, the disassembly of these complexes by Sec18p and Sec17p provides the Vam7p for trans-SNARE complex formation. For this reason, the requirement for Sec18p/Sec17p and for ATP can be bypassed by added Vam7p (30). We first tested the effects of adding recombinant wild-type and mutant Vam7p to our standard fusion reaction, which does not required exogenously added Vam7p. The addition of either wild-type Vam7p, Vam7p Q283R , or other mutants had little effect on fusion ( Fig.  2A). Recombinant wild-type Vam7p can "bypass" the need for priming, the disassembly of cis-SNARE complexes, by forming complexes with the unpaired vacuolar SNAREs (30). In bypass fusion assays, priming is blocked by either the absence of ATP or, in the presence of ATP, by antibody to Sec17p. Although wild-type Vam7p supported bypass fusion (circles, Fig. 2, B and C), Vam7p Q283R did not support the bypass fusion of wild-type vacuoles (triangles, Fig. 2, B and C). These assays allow direct determination of whether Vam7p Q283R can compete with wildtype Vam7p. Even when Vam7p and Vam7 Q283R were premixed prior to addition to vacuoles, a 100-fold excess of Vam7p Q283R only gave modest inhibition of Vam7p-supported fusion (Fig. 3). In these studies, Vam7p Q283R was added at up to micromolar levels, a large excess over the endogenous Vam7p (30). Because Vam7p Q283R binds to vacuoles as well as wildtype Vam7p (data not shown) but does not support fusion (

2, B and C), vacuoles show exquisite discrimination for Vam7p
with the normal Gln residue at the 0-layer position.
Vam7p Q283R Stimulates the Fusion of Vacuoles Bearing Nyv1p R192Q -To test whether Vam7p Q283R is merely denatured or could under some circumstances form functional SNARE complexes, we replaced the 0-layer arginine of Nyv1p with a glutamine and purified vacuoles from yeast harboring Nyv1p R192Q . In a standard fusion reaction with ATP but without added Vam7p, vacuoles bearing Nyv1p R192Q were unable to fuse (Fig. 2D), suggesting that four Q-SNAREs are not readily primed (44). In standard fusion reactions, the same Vam7p Q283R that was inert with vacuoles bearing the wild-type Nyv1p supported the fusion of Nyv1 R192Q vacuoles with a K m of ϳ50 nM (Fig. 2D). Surprisingly, wild-type Vam7p also promoted fusion of Nyv1 R192Q with a similar K m , albeit only ϳ60% as well as Vam7 Q283R . Thus, as seen in other studies (45), the 3Q⅐1R rule is not inviolable; fusion can occur with four Q-SNAREs. When priming was blocked in the presence of ATP by the addition of anti-Sec17p, Vam7p Q283R or wild-type Vam7p supported fusion (Fig. 2E), as seen under standard fusion conditions (Fig. 2D).
In the absence of ATP (Fig. 2F), Vam7 Q283R promoted fusion of Nyv1p R192Q vacuoles whereas wild type Vam7p did not, just as vacuoles bearing wild type Nyv1p could only undergo bypass fusion with wild-type Vam7p (Fig. 2C). However, the concentration of Vam7 Q283R needed for fusion of Nyv1p R192Q vacuoles was ϳ100 times greater than with wild-type Vam7p and Nyv1p (Fig. 2, F versus C). It may take far more energy to form SNARE bundles containing both Vam7p Q283R and Nyv1p R192Q than with wild-type Vam7p and Nyv1p, even though both have 3Q⅐1R 0-layers. This is consistent with energy being needed to distort the SNARE ␣-helical backbone at the 0-layer to accommodate the altered positions of bulky arginine versus the less bulky glutamine, or with "proofreading" by a SNARE-bound factor such as HOPS. Once formed, these 3Q⅐1R complexes have the same capacity to promote fusion. A detailed examination of the specific protein and lipid requirements for fusion supported by each form of Vam7p, with wildtype vacuoles or Nyv1 R192Q vacuoles, is presented in the supplemental Fig. S1.
Physical Interactions of Vam7p with SNAREs and HOPS-Because Vam7p mutant proteins bind to vacuoles in a manner indistinguishable from wild-type Vam7p yet differentially support fusion, we examined their capacities to form protein complexes with SNAREs and HOPS. For this study, vacuole fusion was blocked by antibody to Sec17p for 15min, secondary inhibitors were added, and, after 5 min, the priming block was rescued by the addition of Vam7p (Fig.  4A). As reported (26), wild-type Vam7p relieves the anti-Sec17p block to allow fusion (Fig. 4B, filled bars) and enters into complexes with SNARES and HOPS (Fig. 4C, left panel). These interactions were inhibited by antibodies to Vam3p or Vps33p. Strikingly, Vam7p Q283R , although it does not support the fusion of wild-type vacuoles (Figs. 2B,  2C, and 4B, open symbols), forms stable and isolable HOPS⅐SNARE complexes (Fig. 4C) which contained Nyv1p and are thus 2Q⅐2R. The formation of this complex is also inhibited by antibody to Vam3p or to the HOPS subunit Vps33p, the SM protein of the vacuole.
Bypass of the 2Q⅐2R Block-Because Vam7p Q283R binds to vacuoles and enters SNARE complexes as readily as wild-type Vam7p, but with little or no consequent fusion, we sought conditions that might activate fusion in this 2Q⅐2R system when priming was blocked by either antibody to Sec17p or by the omission of ATP. Bypass fusion with Vam7p Q283R in the absence of ATP was restored by the addition of the phospholipase C inhibitor U73122, but restoration was not seen in the presence of ATP (Fig. 5A). Strikingly, bypass fusion with Vam7p Q283R in the presence of ATP, when priming was blocked by antibody to Sec17p, was restored by the addition of chlorpromazine (Fig. 5B), and this restoration required ATP. In each case, the restored fusion required the normal fusion pathway, because it was blocked by characterized inhibitors that target SNAREs, Ypt7p, and HOPS. Even though bypass fusion with wild-type Vam7p is unaffected by ATP (30), ATP directly regulates the ability of U73122 or chlorpromazine to restore 2Q⅐2R fusion. To further examine this bypass, we characterized the effects of chlorpromazine on both the physical associations of added Vam7p Q283R as well as on the functional restoration of fusion for the same samples (Fig. 6). Chlorpromazine had no measurable effect on the binding of Vam7p Q283R to vacuoles (Fig. 6B, top panel). Chlorpromazine had only a small effect on the association of vacuolebound Vam7p Q283R with HOPS or with other SNAREs (Fig. 6B, lane 4 versus 5) but was absolutely required for fusion (Fig. 6C). In the presence of chlorpromazine, both the physical association of the Vam7p Q283R with HOPS and with SNAREs and the function of restored fusion were still dependent on the R-SNARE Nyv1p (Figs. 5, 6B (lane 8), and 6C), showing that fusion occurred using a 2Q⅐2R SNARE complex, and on Vam3p, Ypt7p, and the SM Vps33p subunit of HOPS (Fig. 6, B (lanes 7, 9 -11) and C).

DISCUSSION
With vacuoles bearing the wild-type R-SNARE Nyv1p, Vam7p Q283R is inactive for fusion, yet forms an HOPS⅐SNARE complex of 2Q⅐2R composition. Vacuoles bearing Nyv1p R192Q can fuse when given either Vam7p Q283R , restoring the 3Q⅐1R composition of the 0-layer, or wild-type Vam7p, yielding a 4Q 0-layer. These findings are in accord with studies showing that 4Q SNARE complexes are functional for exocytosis (46) or endoplasmic reticulum to Golgi traffic (45), whereas 2Q⅐2R complexes are at least somewhat defective (45). Studies with recombinant neuronal SNAREs have shown that the syntaxin 0-layer Gln is essential for NSF factor and ␣-SNAP-mediated SNARE bundle disassembly (44), and this may account for some of the loss of fusion that accompanies mutation of the Vam3p 0-layer Gln to Arg in an earlier study (47). However, diminished disassembly of cis-SNARE complexes is not a factor in our current studies in which the action of Sec18p(NSF)/ Sec17p(␣-SNAP) is blocked, by antibody or by the absence of ATP.
What is the role of the 0-layer in vacuole fusion? Our studies show that the composition of glutamine and arginine residues in the 0-layer and their precise spatial distribution control the energy needed to assemble functional SNARE complexes, because normal bypass fusion has a K m for Vam7p of only 3-8 nM, whereas far higher concentrations of mutant or wild-type Vam7p are needed for 4Q fusion (Fig. 2). Once formed, SNARE complexes with 2Q⅐2R are stable but inactive, yet fusion can be restored when the membrane is perturbed by the intercalating amphiphile chlorpromazine or by the phospholipase C inhibitor U73122. These may modulate bilayer properties such that a "weakened" 2Q⅐2R complex is still capable of driving fusion.
Chlorpromazine is an amphipathic molecule that partitions into the negatively curved inner leaflets of membrane bilayers. Once intercalated into membranes, chlorpromazine alters the physical properties of bilayers. Upon insertion into inner leaflets, chlorpromazine deforms membranes to induce cupping (48) and can alter lateral diffusion and membrane tension (48). In doing so, chlorpromazine reduces relaxation times of membrane tethers by lowering thresholds for remodeling. This was seen in force measurement experiments using optical tweezers to pull membranes (49). Membrane remodeling by chlorpromazine may also alter phosphoinositide metabolism by activating both phosphoinositide kinases and phospholipase C (50). Each of these actions of chlorpromazine may lower the threshold for the bilayer rearrangements of fusion, permitting 2Q⅐2R SNARE complexes to function. Similarly, U73122 inhibits vacuolar phospholipase C activities (51) and thereby alters the ratio of two vacuolar lipids that are crucial for fusion, phosphatidylinositol 4,5-bisphosphate and diacylglycerol. Restoration of 2Q⅐2R fusion by chlorpromazine or U73122 is regulated by the presence or absence of ATP; one of the effects of ATP in this restoration may be to support phosphoinositide synthesis.
SNAREs may promote membrane fusion by several means: 1) They may provide physical stress on the bilayer, which could lower the activation energy for lipid rearrangements for fusion FIGURE 6. Chlorpromazine permits Vam7p Q283R to form normal complexes with SNAREs and HOPS. A, reaction scheme. B, fusion reactions were incubated with ␣-Sec17 to block priming. After 10 min, inhibitors were added and the mixture was incubated for 5 min before addition of 150 M chlorpromazine (CPZ). Chlorpromazine was allowed to act for 5 min before addition of 100 nM Vam7p Q283R . Reactions were incubated for an additional 70 min, and then assayed for Vam7p Q283R -associated proteins as described in Fig. 4. C, fusion was determined by alkaline phosphatase activity. Bars represent mean fusion Ϯ S.E. (n ϭ 3).
(5). 2) SNAREs are needed to gather lipids such as diacylglycerol, which promote the bilayer rearrangements of fusion, to the fusion site (25). 3) The trans-membrane domain of certain SNAREs will destabilize bilayers (33,52), and SNARE-driven bilayer stress and destabilization can lead to the lipid rearrangements of lysis as well as fusion (7,8). 4) SNAREs may also serve as a binding platform to localize and even activate other fusion factors, such as surface-active and calcium-triggered synaptotagmin or lipid-binding HOPS. Assays of each of these four separate SNARE contributions to fusion will be required to determine the importance of each means of SNARE action and the role of the 0-layer in their regulation. Our current findings suggest that SNARE 0-layer function is intimately tied to the lipid bilayer composition and physical properties.