Fusion of docked membranes requires the armadillo repeat protein Vac8p.

The discovery of molecules required for membrane fusion has revealed a remarkably conserved mechanism that centers upon the formation of a complex of SNARE proteins. However, whether the SNARE proteins or other components catalyze the final steps of membrane fusion in vivo remains unclear. Understanding this last step depends on the identification of molecules that act late in the fusion process. Here we demonstrate that in Saccharomyces cerevisiae, Vac8p, a myristoylated and palmitoylated armadillo repeat protein, is required for homotypic vacuole fusion. Vac8p is palmitoylated during the fusion reaction, and the ability of Vac8p to be palmitoylated appears to be necessary for its function in fusion. Both in vivo and in vitro analyses show that Vac8p functions after both Rab-dependent vacuole docking and the formation of trans-SNARE pairs. We propose that Vac8p may bind the fusion machinery through its armadillo repeats and that palmitoylation brings this machinery to a specialized lipid domain that facilitates bilayer mixing.

Homotypic organelle fusion is essential for the maintenance of organelle structure, copy number, and function. In Saccharomyces cerevisiae, lysosome (vacuole) homotypic fusion is involved in three distinct aspects of vacuole biogenesis. First, during cell division the mother cell vacuoles form vesicular/ tubular segregation structures that migrate into the daughter cells and fuse to establish a new vacuole (1)(2)(3)(4)(5). Second, during zygote formation the segregation structures from each mother cell fuse, generating the bud vacuole and exchanging vacuole material (2). Third, yeast vacuoles undergo regulated rapid fission and fusion events. This allows yeast cells to rapidly change the volume of the vacuole in response to changes in osmolarity of their environment.
Development of an in vitro vacuole fusion assay (6 -8) has revealed that most of the molecules known to function in homotypic vacuole fusion are the same or direct homologues of those required for heterotypic fusion (9). Vacuole fusion requires NSF (Sec18p), ␣-SNAP (␣-soluble NSF attachment protein) (Sec17p) (7), the trafficking factor LMA1 (10, 11), a Rab GTPase (Ypt7p) (12), a large protein complex containing the Class B vps proteins Vam2/Vps41p and Vam6/Vps39p (13,14) and Class C vps proteins (15), t-SNAREs 1 (Vam3p and Vam7p), v-SNAREs (Nyv1p) (16,17) (Vti1p and Ykt6p) (18) (see Fukuda et al. for a discussion of the assignment of the SNARE types (19)), phosphatidylinositol 4,5-bisphosphate (20), calmodulin (21), and protein phosphatase I (22). Extensive kinetic analysis performed by Wickner and co-workers (9) reveal three distinct events in the vacuole fusion pathway: priming, docking, and fusion. Several molecules required for priming and docking have been identified. However, details of the molecular basis of fusion are controversial. Several lines of evidence suggest that formation of the v-SNARE⅐t-SNARE complex catalyzes membrane fusion (23)(24)(25). However, equally compelling functional studies suggest that the formation of trans-SNARE pairs is required for docking of membranes but does not catalyze fusion (26 -28). Moreover, some molecules required for fusion have been shown to function after SNARE pair formation (21,22). Thus, to distinguish between these models, we need to know more about the identity and function of molecules that act after membrane docking.
Vac8p is an armadillo (ARM) repeat-containing protein (29 -31). ARM repeat proteins contain tandem arrays of multiple imperfect repeats. These repeats often have multiple binding partners, and a single ARM protein often has multiple functions (32). Yeast Vac8p contains 11 ARM repeats and, although not a functional homologue, it is closely related by sequence to plakoglobin and ␤-catenin. Like plakoglobin and ␤-catenin, Vac8p has multiple binding partners and interacts with its known partners through its ARM repeats (33,34). Vac8p is localized on the vacuole membrane and is enriched at vacuole-vacuole contact sites. Vac8p is both myristoylated and palmitoylated.
It had been shown that palmitoyl-CoA, by transfer to a target protein(s), stimulates vacuole fusion (7). Palmitoylation may be a universal requirement for fusion since palmitoyl-CoA also stimulates Golgi vesicle fusion with the Golgi (35). However, the role of palmitoyl-CoA had not been well defined.
Here, we demonstrate that Vac8p plays a direct role in vacuole fusion both in vivo and in vitro. Vac8p is not necessary for priming and docking but acts after most of the previously identified fusion components and is likely required for fusion per se. Moreover, we show that Vac8p is the target for palmitoyl-CoA and is palmitoylated during the fusion reaction.
Vacuole Docking and Fusion-Vacuole (55) and cytosol (55) preparation, vacuole fusion (55), and palmitoyl-CoA stimulation (7) assays were performed as described. 1 unit of fusion activity is defined as 1 mol of p-nitrophenol produced/min/g of BJ3505 vacuoles. For experiments involving vac8 -15 ts , cells were grown in synthetic complete medium minus uracil supplemented with 0.5% casamino acids before vacuole preparation. The microscopic assay for vacuole docking was performed as described (38), but the calcium chelator, BAPTA inhibitor (10 mM), was substituted for Microcystin LR. The use of BAPTA provided the ability to wash out the inhibitor and measure subsequent fusion. For visualization of docking, aliquots of the reaction were labeled with FM4-64 (80 M FM4-64 in fusion buffer with 0.6% agarose, kept liquid at 34°C), transferred to a slide, incubated for 5 min on ice, and then viewed by fluorescence microscopy. For measurement of fusion after release from the BAPTA block, the remaining reaction was washed with PS buffer (10 mM PIPES-KOH, pH 6.8, 200 mM sorbitol) to remove the inhibitor. Vacuoles were then incubated in fusion conditions for 90 min and examined by fluorescence microscope.
Identification of vac8 -15 ts -Random mutagenesis of VAC8 was used to generate vac8 temperature-sensitive mutants. VAC8 from pYW13 was mutagenized by polymerase chain reaction amplification with Taq polymerase (Roche Molecular Biochemicals) following the manufacturer's instructions and adding 0.2 nM manganese to the reaction. Primers used were Primer 1 (5Ј-CCCGGGTAATGAGCCCTTAAGG-3Ј) and Primer 2 (5Ј-GGTGAATCTAGAGCCAGTTCC-3Ј). The polymerase chain reaction product was purified using QIAquick gel extraction kit (Qiagen Inc.) after agarose gel electrophoresis. pYW2 was linearized with SphI and BclII and phosphatase-treated. Linearized pYW2 was cotransformed with the polymerase chain reaction-mutagenized VAC8 into LWY2887 cells. Transformants were replica-plated to yeast extract/peptone/dextrose containing 0.2% caffeine. ⌬vac8 cannot grow on 0.2% caffeine at any temperature. Colonies were selected for growth at 24°C and no growth at 33°C. Candidates were then labeled with 80 M FM4-64 and screened with a microscope for normal vacuole morphology at 24°C and a fragmented vac8-like morphology at 33 and 37°C. Candidates of interest were screened further using the in vitro vacuole fusion assay. Vacuole and cytosol preparation and vacuole fusion were performed as described. Cells were grown in synthetic complete medium minus uracil supplemented with 0.5% casamino acids before vacuole preparation. Fusion reactions were carried out at 24 and 33°C for 90 min. The value of the control sample at 0°C was subtracted from all the samples.
In Vivo Vacuole Fusion Assay-⌬vac8 LWY2887 cells containing the indicated plasmids were grown overnight at 24°C to an optical density of 0.3 to 0.7 A 600 in synthetic complete minus uracil medium with 0.5% casamino acids. 1 ml of the culture was harvested and resuspended in 220 l of fresh medium with 20 mM PIPES, pH 6.8, and 80 M FM4-64. The mixture was incubated with aeration at 24°C for 1 h. The cells were then harvested, washed twice with YEPD medium, and resuspended in 5 ml of YEPD. The cells were allowed to grow for 1 h at 24°C. Cells were harvested in a microcentrifuge at 10,000 rpm for 1 min.
Flow cells were prepared by coating slides and coverslips with concanavalin-A (1 mg/ml in 50 mM HEPES, pH 7.5, 20 mM calcium acetate, 1 mM MnSO 4 ) and allowed to air-dry. Coverslips were attached to the slides with the concanavalin-A sides facing each other using doublesided tape.
Cells (40 l) were introduced into the flow cell, and the flow cell was incubated upside down for 5 min to allow the cells to attach to the coverslip. Slides were focused on a Zeiss Axioskop using only bright field illumination to prevent bleaching of the FM4-64, then the zerotime fluorescence image was acquired. The excitation was kept to a minimum by setting the lamp power to 5 watts and using a 75% neutral density filter. These precautions were taken to prevent perturbation of vacuole morphology due to the excitation light. A hypotonic solution of 25% yeast extract/peptone/dextrose, 75% H 2 O was flushed through the flow cell, and images were taken every 15 s using a SPOT RT monochrome camera (Diagnostic Instruments). For the temperature shift experiments the flow cells were heated to 37°C on a heat block before loading and were kept heated while the cells settled. The hypotonic solution was also heated to 37°C before use.
Miscellaneous-In vivo labeling of vacuoles with FM4-64 was performed as described (56). Conditions for immunoprecipitation of Vam3p and analysis of proteins that co-immunoprecipitate with Vam3p were performed as described (42). To detect incorporation of palmitate into Vac8p in vitro, vacuoles were washed and resuspended in standard fusion reactions containing 2 g/ml Sec18p, 4 M [ 3 H]palmitoyl-CoA (specific activity 60 Ci/mmol), and no cytosol. Fusion samples were incubated for 90 min at 27°C. Vacuoles were pelleted and solubilized in fusion buffer containing 1% (w/v) Triton X-100, and Vac8p was immunoprecipitated as described (29).  (36). A requirement for Vac8p in homotypic vacuole fusion was confirmed by the demonstration that ⌬vac8 vacuoles cannot fuse in vitro or in vivo. The in vitro assay measures the extent of content mixing of two sets of vacuoles. Vacuoles with proalkaline phosphatase that lack the activating proteinases A and B are mixed with vacuoles that contain these proteases but that lack proalkaline phosphatase (8,37). The amount of the resulting activated alkaline phosphatase reflects the extent of vacuole fusion.

Vac8p Is Required for Vacuole Fusion-Preliminary
Wild-type levels of vacuole fusion require Vac8p on both partner vacuoles (Fig. 1a). This requirement for fusion-related molecules on both vacuole types has been observed in several instances (12,38). The lowered levels of alkaline phosphatase activation with vacuoles from the ⌬vac8 mutant result from defective vacuole fusion, as levels of both the activating proteases and proalkaline phosphatase are normal (Fig. 1b).
The ⌬vac8 fusion defect is not due to the absence of a previously described molecule(s) that is directly required for fusion. As shown in Fig. 1c, Sec18p, Sec17p (7), Ypt7p (12), Vam3p, Nyv1p (16), and calmodulin (21) are all present on ⌬vac8 vacuoles at the same levels found on the wild-type organelle.
Vac8p Functions Directly in Fusion-To further test for a direct role for Vac8p in fusion, we identified a temperaturesensitive vac8 mutant, vac8 -15 ts , that is defective in vacuole fusion at elevated temperatures. This vac8 mutant has normal vacuole morphology at 24°C but has fragmented vacuoles after a short incubation period at 37°C. Isolated vacuoles from the vac8 -15 ts mutant fuse with near wild-type efficiency at 24°C, but at 33°C, fusion occurs at about 50% that seen with wildtype ( Fig. 1d) (at temperatures above 33°C, vacuole fusion in vitro is inhibited in both the wild-type and mutant vacuoles).
To test the immediate outcome of loss of Vac8p function, we developed an in vivo assay to look at the fusion of vacuoles over time. It had been observed that transferring Schizosaccharomyces pombe to hypotonic media induces vacuolar fusion (39). Thus we tested the effect of hypotonic media on S. cerevisiae. We found that a majority of the vacuoles from both the wildtype and the vac8 -15 ts strain fused within 45 s at 24°C (84 and 79%, respectively), whereas few of the vacuoles in the vac8⌬ strain fused (10%). After a 5-min shift to 37°C, a majority of the wild-type vacuoles still fused (80%), but few ⌬vac8 and vac8 -15 ts vacuoles fused (11 and 14%, respectively ( Fig. 2 and Table I)). We attribute the greater block in fusion seen in vac8 -15 ts in the in vivo assay to be due to the ability to perform this assay at 37°C rather than 33°C. That a defect in fusion of vac8 -15 ts vacuoles can be detected after just 5 min at the non-permissive temperature strongly suggests that the role of Vac8p in fusion is direct.
The rapidity of vacuole fusion in response to osmotic stress, shown by the in vivo assay, suggests that specific regulatory mechanisms control fusion. This helps explain recent observations that, like regulated secretory events at the synapse, homotypic vacuole fusion is regulated by Ca 2ϩ (21). Thus, this assay should be useful both for testing molecules involved in fusion per se and also in testing for those molecules that play a regulatory role.
Palmitoylation of Vac8p Is Required for Fusion Activity-Fusion of Golgi transport vesicles with their acceptor compartment is stimulated by palmitoyl-CoA (35), as is vacuole-vacuole fusion (7) (Fig. 3a). This stimulation was proposed to be due to palmitoylation of a target protein(s). Palmitoylation is generally reversible and can serve a regulatory role (40).
Several lines of evidence suggest that Vac8p is a target for the palmitoyl-CoA stimulation of vacuole fusion. First, in contrast to wild-type, the addition of palmitoyl-CoA to ⌬vac8 vacuoles does not stimulate fusion (Fig. 3a). Second, the mutant vac8 -3 (C4G, C5T, C7S), which cannot be palmitoylated (29), is defective in fusion in vitro (Fig. 3b), and most importantly, there is no stimulation of fusion by palmitoyl-CoA (Fig. 3b). Moreover, in parallel with these in vitro defects, vac8 -3 is defective in vacuole fusion in vivo ( Fig. 2 and Table I). Although it is possible that the specific residue changes in vac8 -3 abolish Vac8p function independently of its ability to be palmitoylated, this is unlikely. vac8 -3 functions normally in the cytoplasm to vacuole targeting pathway, demonstrating that this mutant fully retains some of the functions of Vac8p (29). Third, when [ 3 H]palmitoyl-CoA is added during fusion of wild-type vacuoles, significant levels of [ 3 H]palmitate are incorporated into Vac8p (Fig. 3c). Vac8p is ϳ0.1-0.5% of the total vacuolar protein (not shown). In the reaction utilized, there were 0.12 pmol of [ 3 H]palmitate incorporated (as measured by counts associated with immunoprecipitated Vac8p) into 0.48 -2.4 total pmol of Vac8p (based on total vacuole protein per reaction). Therefore, 1 in every 4 -20 Vac8p molecules is modified by exogenous palmitate during fusion. Furthermore, modification of Vac8p might be concentrated to regions of vacuole-vacuole contact, where a much higher fraction of Vac8p molecules may be palmitoylated. Additional vacuolar proteins also are palmitoylated (Fig. 3c). Vac8p is the band at 64 kDa (absent in ⌬vac8 vacuoles). The four other palmitoylated polypeptides migrate with apparent molecular masses of ϳ46, 40, and 30 kDa; their identities remain to be determined.
Vac8p Functions after Vacuole Docking and Trans-SNARE Pair Formation-To determine the step where Vac8p functions in vacuole fusion, we examined each partial reaction. First, we examined priming. During priming, ATP hydrolysis by Sec18p releases Sec17p from the vacuolar membrane and disassembles cis-SNARE pairs (41,42). Priming was first measured by examining the ability of exogenous Sec18p to disrupt cis-SNARE pairs. The extent of SNARE pair disruption was the same for wild-type and ⌬vac8 vacuoles (Fig. 4a). We also examined priming in the presence of endogenous Sec18p. In this case a time course of Sec17p release from the SNARE complex was assayed. Vacuoles were incubated under fusion conditions and transferred to ice at the times indicated. The amount of Sec17p associated with the SNARE complex was measured by Western blot analysis after immunoprecipitation with anti-Vam3p antibody. Both wild-type and ⌬vac8 vacuoles show the same kinetics of Sec17p release (Fig. 4b), again indicating that Vac8p is not required for vacuole priming.
We next examined docking. First, we used a staged assay where vacuoles are arrested at docking with BAPTA, a calcium chelator. This assay is based on studies showing that the addition of BAPTA prevents vacuole fusion and that BAPTA inhibits a post-docking step (21,43). Importantly, the BAPTA inhibition is reversible. To visualize docking, we used a previously developed morphological assay of isolated vacuoles (38). Both wild-type and ⌬vac8 vacuoles form extensive clusters in the docking assay (Fig. 5, b and e). To show that the ⌬vac8 vacuoles are specifically docked and not simply aggregated, ATP was omitted from the incubation; this prevents priming and subsequent docking (Fig. 5, a and d). Moreover, vacuole docking does not occur when ⌬ypt7 vacuoles are utilized (Fig.  5h). The lack of docking of ⌬ypt7 vacuoles is consistent with previous studies showing that anti-Ypt7p antibody blocks vacuole docking (38). The BAPTA was subsequently removed, and vacuoles were incubated under fusion conditions for 90 min. Fusion occurred for docked, wild-type vacuoles but not for Vacuoles (0.5 g of protein) from each strain were separated by SDSpolyacrylamide gel electrophoresis. Indicated proteins were detected by Western blot analysis. C, molecules required for vacuole fusion are present on ⌬vac8 vacuoles at normal levels. Isolated vacuoles (10 g for calmodulin, 1.0 g for all others) were run on SDS-polyacrylamide electrophoresis gels, and Western analysis was performed. D, vacuoles from the temperature-sensitive mutant, vac8 -15 ts , fused at 24°C but were defective in fusion at 33°C. Vacuoles were isolated from DKY6281 ⌬vac8 (LWY4502) and BJ3505 ⌬vac8 (LWY4798), each containing a wild-type VAC8 plasmid (pYW8), a vector control (pRS416), or vac8 -15 ts (pEK50). Strains were grown in synthetic complete medium minus uracil supplemented with 0.5% casamino acids to maintain the plasmid. Standard fusion reactions were performed at 24 and 33°C for 90 min. Background fusion activity was subtracted. The results from three experiments were averaged. Error bars indicate the S.E. of the mean.

Fusion of Docked Membranes Requires Vac8p
docked ⌬vac8 vacuoles (Fig. 5, c and f). Thus, ⌬vac8 vacuoles dock as well as wild-type vacuoles but are defective at a subsequent step in fusion.
A further test of the ability of ⌬vac8 vacuoles to dock can be seen in Fig. 4a (0 time), where it can be seen that the t-SNARE, Vam3p, complexes normally with the v-SNARE, Nyv1p. SNARE pairing was measured by assaying the degree of Nyv1p co-immunoprecipitation with Vam3p. The majority of SNARE pairs assayed are cis pairs, meaning that the v-and t-SNAREs are on the same vacuole (27). The extent of pairing was the same for wild-type and ⌬vac8 vacuoles.
The above experiments demonstrate that SNARE pairs both form and can be disrupted in ⌬vac8. However, because wildtype vacuoles contain the same v-and t-SNARES, it remained to be shown whether the v-SNAREs from a single vacuole will complex with t-SNARES on another vacuole, forming a trans-SNARE pair. As described below, we find that trans-SNARE pairing is normal as well. Nichols et al. (16) previously demonstrated that one could assay solely for trans-SNARE formation by measuring the Nyv1p⅐Vam3p complex formed when vacu-   Table I.

Fusion of Docked Membranes Requires Vac8p
oles lacking Nyv1p were mixed with vacuoles lacking Vam3p. Thus, to examine trans-SNARE pairing in the absence of Vac8p, two new yeast strains were generated, BJ3505, ⌬vac8, ⌬nyv1, and DK6281, ⌬vac8, ⌬vam3. BJ3505, ⌬nyv1, and DK6281, ⌬vam3, served as the corresponding VAC8ϩ controls. Trans-SNARE pairing was measured by incubating the vacuole pairs under conditions where docking occurs but fusion is inhibited and assaying for Nyv1p co-immunoprecipitation with Vam3p. The extent of trans-SNARE pairing was similar for wild-type and ⌬vac8 vacuoles (Fig. 4c). Although it is not possible in this type of assay to determine whether a SNARE complex is potentially functional or non-functional (44,45), note that the only opportunity for trans-SNARE pairs to form in this experiment is during the 30-min incubation. Also note that trans-SNARE pairs did not form when the samples were kept at 0°C (Fig. 4c).
Differences in the in vivo phenotypes of ⌬ypt7 and ⌬vam3 compared with ⌬vac8 further support a model whereby Ypt7p

FIG. 3. Palmitoylation of Vac8p is important for vacuole fusion.
A, palmitoyl-CoA has no effect on ⌬vac8 vacuole fusion. Vacuoles were isolated from wild-type BJ3505 and DKY6281 and the corresponding ⌬vac8 strains and preincubated with fusion buffer (55) at 30°C for 10 min. Vacuoles were re-isolated and used in standard reactions without cytosol but containing 2 g/ml Sec18p (7). Palmitoyl-CoA (4 M) was added as indicated. Background fusion activity (0.173-0.200 units (U)) was subtracted. WT, wild type; ALP, alkaline phosphatase. B, vacuoles from strains LWY4798 (⌬pep4, ⌬vac8) and LWY4502 (⌬pho8, ⌬vac8), each carrying either a low copy wild-type VAC8 plasmid (WT CEN) or high copy palmitoylation minus vac8 -3 plasmid (vac8 -3 (2)), were isolated. Vacuole fusion reactions were performed as in A, except that strains were grown in synthetic medium minus uracil to maintain the plasmid. The data presented are the average of three experiments. Note that the levels of fusion obtained from vacuoles from cells grown in minimal media are significantly lower than vacuoles from cells grown in rich medium. Error bars indicate the S.E. of the mean. Background activity (0.190 -0.321 units) was subtracted. C, vacuole fusion reactions were performed as in A, but [ 3 H]palmitoyl-CoA (4 M) was used. At the end of the reaction, total vacuoles from the 2.4ϫ standard reaction were subjected to SDS-polyacrylamide gel electrophoresis and fluorography (left panel). In addition, Vac8p from 6ϫ standard reactions was immunoprecipitated and then subjected to SDS-polyacrylamide gel electrophoresis and fluorography (right panel). ⌬vac8 mutant. A, assay for total SNARE pairs present on isolated vacuoles and the ability of the pairs to be disrupted. The vacuole fusion reactions were composed of 18ϫ standard reactions and were incubated either at 0°C (0 time, lanes 2 and 4, total SNARE pairs on untreated vacuoles) or at 30°C with 35 g/ml Sec18p for 15 min (lanes 3 and 5, SNARE pairs remaining after Sec18p treatment). Immunoprecipitation (IP) with anti-Vam3p antibody was performed, and the presence of Vam3p and Nyv1p was assessed by Western analysis. WT, wild type. B, assay for Sec17p release. Each vacuole fusion reaction was composed of 18ϫ standard reactions and was incubated at 30°C. At the indicated times, reactions were transferred to 0°C and incubated for a total of 40 min. Immunoprecipitation with anti-Vam3p antibody was performed, and the presence of Sec17p was determined by Western analysis. C, measurement of trans-SNARE pairs during docking. Vacuoles (120 g each) were incubated together on ice (lanes 1 and 3) or at 27°C (lanes 2 and 4) with cytosol, ATP, and nodularin (a phosphatase inhibitor that is similar to Microcystin LR (57)) (16 M) for 30 min. Lanes 1 and 2 contain vacuoles from BJ3505, ⌬nyv1, and DK2681, ⌬vam3. Lanes 3 and 4 contain vacuoles from BJ3505, ⌬nyv1, ⌬vac8, and DK2681, ⌬vam3, ⌬vac8. and Vam3p function sequentially in docking, and Vac8p functions after trans-SNARE pair formation. Both ⌬ypt7 and ⌬vam3 mutants have numerous small vacuoles dispersed throughout the cytoplasm (Refs. 12, 16, 46, and 47 and Fig. 6, Table II), a morphology consistent with their respective defects in the early and late stages of vacuole docking. In contrast, the ⌬vac8 vacuoles are tightly clustered (Fig. 6, Table II), which suggests that the small vacuoles can dock but not fuse. To determine the prevalence of clustered versus dispersed vacuoles, a minimum of 200 random cells was scored for each strain. In some of the strains, vacuoles filled the entire cytoplasm. Thus, we also added this third category so that all cells in a given field would be included in the analysis.
If Ypt7p and Vam3p act before Vac8p, then the corresponding double mutants should display the phenotype of the mutant blocked at the earlier step in the pathway. We constructed the double-mutant strains ⌬ypt7/⌬vac8 and ⌬vam3/⌬vac8 and examined their vacuole morphology. As expected, both double mutants have small, dispersed vacuoles, as seen in the ⌬ypt7 and ⌬vam3 single-mutant strains (Fig. 6, Table II). In further support of this ordered pathway, neither overexpression of Ypt7p nor Vam3p suppressed the vacuole fragmentation defect seen in ⌬vac8 (not shown). DISCUSSION Our discovery that Vac8p is required for homotypic vacuole fusion addresses three questions. 1) Is the act of SNARE pair formation sufficient for membrane fusion? 2) Is membrane fusion coordinated with other membrane-related processes? 3) Is a specialized lipid domain required to promote bilayer mixing, and is this domain brought together with proteins that comprise the fusion machinery?
Two major lines of evidence support the prevailing view that formation of the v-SNARE/t-SNARE complex catalyzes membrane fusion. Liposomes containing a purified v-SNARE fuse with liposomes containing purified t-SNARES (23,25). Moreover, structural analysis of the SNARE pair complex suggests that SNARE pairing may be the fundamental event in membrane fusion. The complex forms four parallel helicies with structural similarities between the SNARE complex and viral fusion proteins (for discussion see, Skehel and Wiley (24)).
Although SNARE pairing as a catalyst of fusion is an attractive model, several studies indicate that SNARE pairing is not sufficient for fusion. First, functional studies with both yeast vacuoles and sea urchin cortical vesicles demonstrate that when the SNARE complex is disrupted, fusion still proceeds (26,27). Furthermore, in synaptosomes, N-ethylmaleimide treatment causes an increase in SNARE pair formation, yet it does not result in a concomitant increase in fusion (28). In addition, staged assays have revealed molecules that function after SNARE pairing, protein phosphatase I (22), and Ca 2ϩ / calmodulin (21).
To build an accurate model of how fusion occurs, the key players in this event need to be identified and characterized. Our studies present a new type of molecule required for fusion. Results from both in vivo and in vitro studies demonstrate that Vac8p is required for fusion and strongly suggest that Vac8p functions downstream of SNARE pairing. Most compelling is our finding that trans-SNARE pair formation appears to be normal in ⌬vac8 vacuoles. The SNARE pairs from ⌬vac8 vacuoles formed within 30 min and to the same extent as that seen with wild-type vacuoles. Although ⌬vac8 vacuoles are not competent for fusion, they are competent for the earlier steps of priming and docking.
Analysis of both the localization and the sequence of Vac8p suggests a model of how Vac8p may function in fusion. Vac8p accumulates at vacuole-vacuole junctions in vivo (30,31), a site consistent with a direct role in fusion. Vac8p has two types of binding motifs. One motif is composed of 11 ARM repeats of Vac8p. These repeats are found in a diverse set of proteins, have no known catalytic function, and are generally sites of protein-protein interaction. The other motif is composed of acyl chains at the Vac8p amino terminus, and as discussed below, this acylation may bring Vac8p to a specific lipid domain. Thus, an attractive model for Vac8p function is that through its ARM repeats, Vac8p binds the protein portion of the fusion machinery, and when Vac8p is acylated, it brings the protein complex to a specialized lipid domain.
In addition to its role in homotypic vacuole fusion presented here, Vac8p is required in at least two other distinct membrane processes, where it interacts with a unique protein partner specific for that process. Vac8p complexes with Apg13p to facilitate the closure of vesicles in the cytoplasm-to-vacuole-targeting pathway (34). Vac8p complexes with Nvj1p to form specific junctions between the nucleus and vacuole (33). The interactions of Vac8p with these proteins are mediated through the Vac8p ARM repeats. Interestingly, the functional role of each of the Vac8p partners above is restricted to a specific process, and none are required for homotypic vacuole fusion. 2 Thus, it is likely that there is a Vac8p binding partner dedicated to homotypic vacuole fusion. Discovery of this partner FIG. 5. ⌬vac8 vacuoles can dock but cannot fuse. Vacuoles from BJ3505 (wild type (WT)), LWY4798 (⌬vac8), and ⌬ypt7 were analyzed with a morphological docking assay. BAPTA was added to block fusion. In the no-ATP control, no docking occurs. At the end of the docking reaction, 12-l aliquots were labeled and viewed by fluorescence microscopy. The remaining docked vacuoles were washed with PS buffer to remove BAPTA. Vacuoles were then incubated in fusion conditions for 90 min at 27°C. Twelve microliters of each reaction was then labeled and examined. The wild type and ⌬vac8 strains were isogenic except at the VAC8 locus. The ⌬ypt7 parental strain is SEY6211. The large difference in the vacuole size of ⌬ypt7 vacuoles is accounted for in part by the phenotype of the mutant (fragmented vacuoles) and in part by the difference in strain background.
should provide further insight into the role that Vac8p plays in fusion.
That Vac8p is required for multiple vacuole membrane-related events yet acts through several partners suggests that there is a mechanism to prevent these diverse processes from occurring simultaneously. Perhaps each occurs at a similar region of the vacuolar membrane, creating the need to coordinate these events with each other.
Are There Functional Homologues of Vac8p?-Most of the proteins that have previously been described to play a role in homotypic vacuole fusion have been shown to have direct homologues in other organisms and, moreover, to have direct homologues that function in fusion of other membranes. Although Vac8p homologues are likely to be present in metazoans, none has yet been identified. The difficulty in searching for a Vac8p homologue by sequence alone is that there are at least 25 ARM repeat proteins in Caenorhabditis elegans, 30 in Drosophila melanogaster, and 93 in Arabidopsis thaliana. Moreover, there is not just one consensus sequence for palmitoyla-tion. Thus, it is likely that a metazoan Vac8p homologue will need to be identified by functional analysis.
Although Vac8p homologues may exist in other organisms, it appears that Vac8p functions in some but not all yeast membrane fusion pathways. The role of Vac8p in nuclear vacuole junction formation and in the cytoplasm-to-vacuole targeting pathway may be a specific function in fusion; however, other yeast membrane fusion events do not appear to require Vac8p. Some membrane fusion events are essential for yeast viability, yet VAC8 is not an essential gene. However, the fact that palmitoyl-CoA stimulates fusion of mammalian Golgi vesicles with the Golgi and also stimulates homotypic vacuole fusion strongly suggests that palmitoyl-CoA provides a general, fundamental role in membrane fusion.
What Is the Role of Palmitoyl-CoA in Fusion?-The lipid binding motif of Vac8p is composed of Gly2, which is myristoylated, and Cys4, Cys5, and Cys7, which are palmitoylated. Interestingly, the sites of myristoylation and palmitoylation on Vac8p are very similar to those found in the plasma membrane Src-family kinases (29). Moreover, these proteins share a conserved Ser at position 6 and a conserved Lys at position 9. These motifs are not common, and it is tempting to speculate that they serve a similar function in these diverse proteins. In the case of the Src-family kinases, palmitoylation drives them into cholesterol-rich lipid rafts (50). By analogy, palmitoylation of Vac8p may bring it (and the fusion machinery that is attached to the ARM repeats of Vac8p) to an ergosterol-rich membrane domain (ergosterol is the sterol in yeast that functions similarly to cholesterol).
Properties of cholesterol are consistent with it playing a major role in facilitating fusion. Cholesterol is intercalated in the cytoplasmic side of the phospholipid bilayer and stabilizes the region near the polar hydroxyl group. However, the region of the acyl chains that are most distal from the phospholipid head group are destabilized by the presence of cholesterol, precisely the situation that is expected to accompany bilayer mixing. In addition, cholesterol has been shown to have an important role in viral fusion (for examples, see Kielian and Helenius (51) and Chatterjee et al. (52)). Moreover, SNARE proteins concentrate at sites of exocytosis in cholesterol-rich domains (53). In addition, it has been proposed that the Vo subunit of the vacuolar ATPase plays a key role late in fusion (49). Interestingly, the Vo subunit of the v-ATPase in synaptic-

Vac8p acts after Ypt7p and Vam3p
Strains listed below were labeled with FM4 -64. Random fields of cells were photographed, and the vacuole morphology of each cell was scored. In some cases cells contained too many vacuoles to distinguish between a clustered vs. dispersed phenotype. These cells are reported in the last column. A minimum of 200 cells for each strain were included in the analysis.
FIG. 6. Vac8p functions downstream of Ypt7p and Vam3p. ⌬ypt7, ⌬vam3, ⌬vac8, and double mutants ⌬ypt7/⌬vac8, ⌬vam3/⌬vac8 were labeled with FM4-64, and vacuolar morphology was examined by fluorescence microscopy (56). Quantitation of vacuole morphology of random fields of cells for each strain is presented in Table II. like vesicles has been shown to be a major protein that crosslinks to cholesterol (54).
The observation that palmitoylation of Vac8p is required for fusion strongly suggests that this region of Vac8p attaches to a defined membrane patch that promotes bilayer mixing. Determination of the composition of this patch is likely to yield further insight into the final steps in fusion.