Interaction between the yeast RAVE complex and Vph1-containing Vo sectors is a central glucose-sensitive interaction required for V-ATPase reassembly

The yeast vacuolar H+-ATPase (V-ATPase) of budding yeast (Saccharomyces cerevisiae) is regulated by reversible disassembly. Disassembly inhibits V-ATPase activity under low-glucose conditions by releasing peripheral V1 subcomplexes from membrane-bound Vo subcomplexes. V-ATPase reassembly and reactivation requires intervention of the conserved regulator of H+-ATPase of vacuoles and endosomes (RAVE) complex, which binds to cytosolic V1 subcomplexes and assists reassembly with integral membrane Vo complexes. Consistent with its role, the RAVE complex itself is reversibly recruited to the vacuolar membrane by glucose, but the requirements for its recruitment are not understood. We demonstrate here that RAVE recruitment to the membrane does not require an interaction with V1. Glucose-dependent RAVE localization to the vacuolar membrane required only intact Vo complexes containing the Vph1 subunit, suggesting that the RAVE-Vo interaction is glucose-dependent. We identified a short conserved sequence in the center of the RAVE subunit Rav1 that is essential for the interaction with Vph1 in vivo and in vitro. Mutations in this region resulted in the temperature- and pH-dependent growth phenotype characteristic of ravΔ mutants. However, this region did not account for glucose sensitivity of the Rav1-Vph1 interaction. We quantitated glucose-dependent localization of a GFP-tagged RAVE subunit to the vacuolar membrane in several mutants previously implicated in altering V-ATPase assembly state or glucose-induced assembly. RAVE localization did not correlate with V-ATPase assembly levels reported previously in these mutants, highlighting both the catalytic nature of RAVE's role in V-ATPase assembly and the likelihood of glucose signaling to RAVE independently of V1.

Tight maintenance of organelle and cytosolic pH is required for cellular homeostasis, and perturbed pH homeostasis is associated with disease states ranging from metabolic acidosis (1) to neurodegeneration (2) and aging (3,4). V-ATPases 2 are highly conserved proton pumps that are responsible for organelle acidification in eukaryotic cells. They are multisubunit complexes composed of two major subcomplexes, a peripheral V 1 complex oriented toward the cytosol that contains the sites for ATP hydrolysis and an integral membrane V o domain that contains the proton pore. Mammalian genomes encode multiple isoforms for several subunits in both V 1 and V o (5). In yeast, however, all subunits except the V o a-subunit are encoded by single-copy VMA genes. In contrast, the yeast a-subunit consists of Vph1 and Stv1 isoforms that target V-ATPases to the vacuole or Golgi, respectively, and also endow their respective complexes with distinct regulatory properties (6 -8).
Reversible disassembly is a conserved, post-translational form of V-ATPase regulation that is able to tune V-ATPase activity to multiple different stimuli (9 -12). In yeast, acute glucose deprivation results in separation of the V 1 subcomplex and V 1 subunit C from the V o subcomplex, and glucose readdition induces rapid reassembly of both V 1 and V 1 subunit C with V o . After disassembly, free V 1 subcomplexes lack Mg-ATPase activity and free V o subdomains are closed to H ϩ translocation, and activity is restored by reassembly (13)(14)(15). The yeast RAVE complex is a V-ATPase-specific chaperone required both for initial biosynthetic assembly of the V-ATPase and for efficient glucose-dependent reassembly of V-ATPases (16,17). Upon glucose readdition, RAVE catalyzes recruitment of cytosolic V 1 and V 1 subunit C to the vacuolar membrane and assembly of functional V-ATPases. Yeast RAVE is a heterotrimeric complex comprised of three subunits, Rav1, Rav2, and Skp1 (18). Rav1 is the central component of RAVE and contains N-terminal and C-terminal binding sites for Rav2 and Skp1, respectively (19). Rav1 interacts independently with 1) V 1 subcomplexes via one or more of the EG peripheral stalks, 2) V 1 subunit C, and 3) the cytosolic N-terminal domain of Vph1 (Vph1NT) of the V o subcomplex (19). Rav2 interacts with V 1 subunit C, making V 1 subunit C the only V-ATPase subunit that contains multiple RAVE-binding sites (19). Skp1, a component of many essential cellular complexes, including SCF (Skp1-cullin-F-Box) ubiquitin ligases, does not appear to interact directly with V-ATPase subunits (16). The rabconnectin-3 (Rbcn3) complex is the structural and functional homologue of RAVE in higher eukaryotes, but it appears to contain only two larger subunits and no Skp1 (20). The Rbcn3 complex was shown to associate with V-ATPase subunits in Drosophila (21), and loss of function compromises organelle acidification in multiple organisms (21)(22)(23). The Rbcn3␣ subunit has regions of homology to the N-and C-terminal domains of Rav1 but is almost twice as large.
It was proposed that RAVE's role in glucose-dependent V-ATPase reassembly is to shuttle cytosolic V 1 and V 1 subunit C back to V o at the vacuolar membrane and provide a template for complex assembly, thus improving the efficiency of V-AT-Pase reassembly (19). However, the cellular levels of Rav1 and Rav2 are only 5-10% of the levels of V 1 subunits (24), suggesting that RAVE's role must be catalytic, requiring multiple cycles of binding and release of V 1 to achieve V-ATPase reassembly. We recently found that a significant portion of RAVE is present at the vacuolar membrane of glucose-replete cells, released from the membrane by glucose deprivation, and recruited back to the membrane upon glucose readdition (19). This recruitment and release of RAVE is entirely consistent with its role as a glucosedependent chaperone and assembly factor. However, neither the molecular basis of the glucose-sensitive recruitment of RAVE nor the signaling mechanisms involved are known. The interaction between RAVE and V 1 is not intrinsically glucosesensitive. Rather, RAVE can bind to the V 1 subdomain whether it is in the cytosol as a result of glucose deprivation or as a result of V o mutations that prevent membrane association (17).
In the current study, we find that only Vph1-containing V o subcomplexes are required for glucose-dependent recruitment of RAVE to the vacuolar membrane. Previous binding data indicated that a fragment consisting of amino acids 679 -898 of Rav1 (Rav1 679 -898) binds directly to Vph1NT. We identified a six amino acid conserved motif in this region that is required for interaction of Rav1 679 -898 with Vph1NT in vitro and for recruitment of RAVE to the vacuolar membrane in vivo. Although this interaction is not intrinsically glucose-sensitive, we tested whether mutations previously associated with altered assembly or reversible disassembly of the V-ATPase would affect steady-state levels of RAVE at the membrane in the presence or absence of glucose. We find that mutations in several glucose-sensing pathways significantly alter RAVE localization. The results indicate that the RAVE complex itself can be recruited to the vacuolar membrane without V 1 and that its steady-state localization does not strictly follow that of V 1 subcomplexes.

Interactions with the V 1 subcomplex are not required for glucose-dependent recruitment of RAVE to the membrane
Levels of RAVE subunits in isolated cytosolic fractions do not change with glucose levels (17), and very little RAVE is isolated with vacuolar membranes containing the V-ATPase. However, using GFP-tagged RAVE subunits, we previously demonstrated that RAVE is reversibly released from the vacuolar membrane upon glucose deprivation and restored to the membrane upon glucose readdition (19). GFP-tagged Rav1 and Rav2 comple-ment the growth phenotypes of the corresponding deletion mutants. We asked whether either subunit could be recruited to membranes in the absence of the other. As shown in Fig. 1, both Rav1-GFP and Rav2-GFP are present at the membrane, as well as the cytosol, in glucose-grown WT cells. In contrast, Rav2-GFP is almost completely cytosolic in rav1⌬ cells, and Rav1-GFP is almost completely cytosolic in rav2⌬ cells. We conclude that both Rav1 and Rav2 subunits must be present for efficient recruitment of the complex to the vacuolar membrane, and in subsequent experiments, we use Rav2-GFP to track localization of the RAVE complex.
Although V 1 subunit C is released from the vacuolar membrane upon glucose deprivation and is proposed to be a key player in reversible disassembly of the yeast V-ATPase, we previously found that the RAVE complex can be recruited to the vacuolar membrane in glucose-grown vma5⌬ cells, which lack V 1 subunit C. In addition, Rav2-GFP also cycles off and on the membrane with glucose deprivation and readdition in a vma5⌬ mutant (19), indicating that V 1 subunit C is not required for glucose-sensitive recruitment of RAVE to the vacuolar membrane. V 1 assembly is compromised (25), and RAVE-V 1 binding is abolished (17), in a vma4⌬ mutant, which lacks the V 1 subunit E, so we compared Rav2-GFP localization in vma4⌬ and WT strains. As shown in Fig. 2A, Rav2-GFP localized to the vacuolar membrane in both the vma4⌬ and WT strains in the presence of glucose, suggesting that binding to the V 1 subcomplex is not required for vacuolar localization of RAVE. We next tested whether Rav2-GFP was released from the membrane into the cytosol upon glucose deprivation. As shown in both images and line scans in Fig. 2A, vacuolar localization is lost in both vma4⌬ and WT cells. However, glucose readdition restores vacuolar localization in both strains. In line scans for high glucose conditions in both strains, there are maxima of fluorescence intensity corresponding to the vacuolar membrane. To compare localization of Rav2-GFP across a larger population of the cells, we performed line scan measurements across 10 -11 cells under each condition. The maximum fluorescence intensity for each scan was normalized to the average intensity of the cells maintained in glucose, and the results are plotted in Fig. 2B. For both WT and vma4⌬ cells, there is a significant decrease in maximal intensity in the absence of glucose, but no significant difference after glucose readdition. Top, Rav1 was tagged with GFP and expressed in WT and rav2⌬ strain. Bottom, Rav2 was tagged with GFP and expressed in WT and rav1⌬ strain. All cells were maintained in glucose. DIC images are shown on the right in each set; the vacuoles are identified as a single or several distinct indentations depending on conditions. Both GFP-tagged subunits were also present in small cytosolic puncta in some cells. The nature of these puncta is not known. Scale bar, 5 m.

Requirements for vacuolar localization of yeast RAVE
From these data, we conclude that the association of RAVE with V 1 is not required for glucose-dependent recruitment of RAVE to the vacuolar membrane.
The results above were somewhat surprising, given that RAVE has been shown to interact with both V 1 (17) and V 1 subunit C (19,26) in the cytosol of glucose-deprived cells and to be released from the membrane under conditions of V-ATPase disassembly (19). We hypothesized that recruitment of RAVE to the membrane might solely depend on interaction with V o subcomplexes and thus would be disrupted in a vma11⌬ mutant, which lacks an essential subunit of the proteolipid ring in V o and fails to assemble or transport V o subcomplexes to the vacuole (27,28). Indeed, Rav2-GFP remained cytosolic even in the presence of glucose in a vma11⌬ strain (Fig. 3A), indicating that interactions with the V o subcomplex are required for membrane recruitment of RAVE. We next asked whether V o complexes needed to be free of V 1 for RAVE to bind. A mutation of a critical glutamate (E145L) in VMA11 allows V-ATPase assembly, but abolishes both ATPase activity and proton transport (27). The vma11 E145L mutant was shown to have increased levels of V 1 subunits assembled at the vacuole, apparently as a result of assembly of virtually all V o subcomplexes with V 1 (27). We introduced Rav2-GFP into this strain to assess whether RAVE was able to bind to fully assembled, but inactive, V-ATPases. Interestingly, Rav2-GFP was still able to bind to the membrane in presence of glucose and to be reversibly released upon glucose deprivation and readdition (Fig. 3B). These results suggest that the site of RAVE interaction with the V o subcomplex must be available when the V-ATPase is assembled and suggest that release and recruitment of RAVE do not require catalytic cycling of the V-ATPase.
There are two isoforms of the V o a-subunit, Vph1 and Stv1, but only V-ATPases containing the Vph1 isoform require RAVE for their assembly (29). We also found that the cytosolic domain of Vph1 (Vph1NT) binds directly to Rav1 in vitro (19). Consistent with these observations, we found that Rav2 failed to localize to the vacuolar membrane in a vph1⌬ strain, even in the presence of glucose (Fig. 3C). This suggests that Vph1 is likely to be the molecular target for RAVE recruitment to the membrane. In fact, the Rav1-Vph1 interaction appears to be the

Requirements for vacuolar localization of yeast RAVE
central glucose-dependent interaction between the RAVE and V-ATPase complexes.

A conserved motif within Rav1 is required for RAVE function, recruitment to vacuolar membranes, interaction with Vph1NT, and V-ATPase assembly
Through a combination of bioinformatics, molecular modeling, and interaction assays (19), we previously predicted that the N-terminal region of yeast Rav1 (Rav1 55-669) is rich in ␤-sheet structures, likely forming a double ␤-propeller, and showed that it contains a binding site for Rav2. The C-terminal half of Rav1 (Rav1 840 -1357) contains the most highly conserved portion of the protein (Rav1 840 -1125), in a region that is predicted to be predominantly ␣-helical. The C-terminal half of Rav1 contains binding sites for V 1 , V 1 subunit C, and Skp1. We identified a Vph1NT-binding site between the N-and C-terminal regions in a less conserved segment encompassing Rav1 679 -898 (19). Sequence analysis of this region revealed the presence of a conserved six amino acid motif ( 757 LPVYHP in yeast; Fig. 4A). To determine whether this region was essential for RAVE function, we generated point mutations targeting conserved residues (rav1 760 YH 761 mutated to AA (YH-AA) and rav1 K773A). We also deleted the entire six amino acid motif (rav1 6⌬). The mutants were then tested for a Rav1 Ϫ phenotype by comparing growth on plates buffered to low pH to growth on plates at high pH, with added calcium, at 37°C (16,17). rav1⌬ and rav2⌬ mutants are characterized by a temperature-sensitive Vma Ϫ growth phenotype and show WT growth at pH 5 but poor growth on YEPD, pH 7.5, or YEPD, pH 7.5, containing 60 mM CaCl 2 at 37°C (16,17). Interestingly, the rav1 YH-AA and rav1 6⌬ mutations resulted in a Rav Ϫ phenotype, whereas the K773A mutation allowed WT growth (Fig. 4B).
The data above suggest that the six amino acid conserved motif is essential for RAVE function, but it is possible that the mutation reduces Rav1p levels or causes overall misfolding of the protein. We tagged both WT Rav1 and the Rav1 6⌬ mutant protein with a Myc13 tag and assessed protein stability on immunoblots of whole-cell lysates (Fig. 4C). We observed only a slight decrease in mutant Rav1p levels compared with WT, suggesting that the Rav1 Ϫ phenotype is not due to loss of Rav1. We next assessed whether the rav1 6⌬ mutation altered interaction between Rav1 and V 1 . Amino acids 840 -940 of Rav1 were previously shown to be required for V 1 binding (19), so if the rav1 6⌬ mutation has a major effect on structure, V 1 binding might be lost. We immunoprecipitated the Myc13-tagged WT Rav1 and Rav1 6⌬ proteins and assessed co-precipitation of V 1 subunits. The rav1 6⌬ mutation did not alter the interaction with V 1 , as indicated by comparable levels of V 1 subunit A co-precipitated with Rav1-Myc13 (Fig. 4C). We next hypothesized that the six amino acid conserved motif might be required for RAVE's vacuolar localization. We compared Rav2-GFP localization in the rav1 6⌬ mutation to its localization in WT Rav1. As shown in Fig. 4D, Rav2-GFP is largely cytosolic in the rav1 6⌬ mutant, even in glucose-replete conditions, indicating that this motif is necessary for membrane recruitment of RAVE.
Because RAVE's vacuolar localization depends on both the presence of glucose and Vph1, we hypothesized that rav1 6⌬ mutant might prevent recruitment of RAVE to the vacuolar membrane as the result of a signaling defect, in which RAVE is no longer able to sense glucose, or as the result of a binding defect that prevents RAVE from binding to Vph1. To distinguish between these possibilities, we expressed Vph1NT and A, alignment of the indicated portion of yeast Rav1 with the following rabconnectin-3␣ homologues in higher eukaryotes: human, Homo sapiens DMXL2 (NCBI locus AAI44540); fly, Drosophila melanogaster rabconnectin-3␣ (NCBI locus NP_572302); and plant, Arabidopsis thaliana hypothetical protein (NCBI locus AAD20167). Numbers indicate the position of the first amino acid in the protein sequence shown. B, growth phenotype of strains containing WT and mutant RAV1. 10-Fold serial dilutions of log-phase liquid cultures of the indicated strains grown in YEPD, pH 5, were prepared in 96-well plates and then transferred by pinning to YEPD, pH 5, plates or YEPD, pH 7.5, plates containing 60 mM CaCl 2 . For each plate, the highest concentration of cells is shown on the left, with successive serial dilutions to the right. Growth at 30°C and 37°C as indicated. The Rav Ϫ mutant phenotype is characterized by sensitivity to high pH and calcium concentrations at 37°C. C, a C-terminal Myc13 tag was added to Rav1 in all of the strains except the no-Myc control samples (No myc). Cytosolic fractions were prepared from the indicated yeast strains as described under "Experimental procedures," and a portion was reserved as input. The remainder was combined with mouse anti-Myc mAb followed by Protein A-Sepharose to immunoprecipitate Myc-tagged Rav1 (Myc IP). Both input and immunoprecipitate samples from each strain were separated by SDS-PAGE and transferred to nitrocellulose. The blots were probed was anti-Myc to determine the levels of Rav1-Myc and with mAb 8B1F3 to determine the level of co-precipitated V 1 A subunit. D, cellular distribution of Rav2-GFP in WT cells and in the rav1 6⌬ mutant. DIC images are shown to the left of each set and GFP fluorescence to the right. Scale bar, 5 M.

Requirements for vacuolar localization of yeast RAVE
Rav1 679 -898, with or without the rav1 6⌬ mutation, in Escherichia coli and assessed their interaction in vitro. Both WT and mutant Rav1 fragments were tagged with a C-terminal His 6 tag. Vph1NT and WT or mutant Rav1 fragments were incubated with TALON resin, which binds to the His 6 tag, and co-elution of Vph1NT with the Rav1 fragments at high imidazole concentrations was assessed by immunoblotting. As shown in Fig. 5, whereas Vph1NT eluted with WT Rav1 679 -898 (Fig. 5A), Vph1NT showed very little binding to the Rav1 679 -898 fragment containing the rav1 6⌬ mutation (Fig. 5B) and no binding in the absence of the Rav1 fragment (Fig. 5C). These data indicate that the conserved motif is directly involved in the interaction between Rav1 679 -898 and Vph1NT, rather than indirectly affecting the interaction by contributing to glucose sensing.
We next assessed how the rav1 6⌬ mutation affects the assembly of the V-ATPase. We first compared localization of Vma5-GFP (V 1 subunit C) to the vacuolar membrane in WT and rav1 6⌬ mutant cells. As shown in Fig. 6A, Vma5-GFP is localized to the vacuolar membrane in WT cells but is largely cytosolic in the rav1 6⌬ cells, suggesting that V-ATPase assembly is compromised in the mutant. We isolated vacuolar vesicles from WT and rav1 6⌬ mutant cells and compared the levels of V-ATPase. As described previously, rav1⌬ cells have WT levels of V o subunit Vph1 in vacuolar vesicles, but reduced levels of V 1 subunits at the membrane (26), and the rav1 6⌬ exhibits a very similar assembly defect (Fig. 6B). Consistent with their assembly defect, vacuolar vesicles from the rav1 6⌬ mutants also have significantly reduced levels of concanamycin-sensitive ATPase activity (Fig. 6C). Taken together, these data indicate that loss of interaction with Vph1 in the rav1 6⌬ mutant

Requirements for vacuolar localization of yeast RAVE
results in loss of RAVE-dependent V-ATPase assembly and function.

Cellular factors that contribute to glucose-dependent localization of RAVE
As described above, although RAVE is recruited to the vacuolar membrane by its interaction with Vph1, this interaction is reversed by glucose deprivation. This suggests that either the binding site becomes unavailable, or some other cellular pathway alters the interaction between RAVE and Vph1-containing V o sectors, directly or indirectly. Several pathways have been associated with signaling the V-ATPase assembly state in yeast, and we hypothesized that one or more of these pathways might affect glucose-dependent association of RAVE with the vacuolar membrane. We tested effects of three mutations previously reported to affect reversible disassembly of the V-ATPase on glucose-dependent RAVE recruitment to the membrane. Deletion of phosphofructokinase subunit Pfk2 (pfk2⌬) was reported to reduce V-ATPase reassembly after glucose readdition and enhance the RAVE-V 1 association in cytosolic fractions (30). Deletion of the Ras inhibitor protein Ira2 (ira2⌬) prevented disassembly of the yeast V-ATPase upon glucose deprivation, prompting further investigations that suggest reduced levels of Ras and protein kinase A activity are required to trigger disassembly upon glucose deprivation (31). In contrast, deletion of the master metabolic regulator Sch9 (sch9⌬) increases V-AT-Pase assembly in both the presence and absence of glucose (32). Rav2 was labeled with GFP in each of the mutant strains, and co-localization of Rav2-GFP with the vacuolar membrane dye FM4-64 was quantitated. This quantitation was necessary because many of the signaling pathways implicated in V-AT-Pase reversible disassembly give partial effects, so we anticipated that simple visualization of membrane association as described above might be inadequate to show effects.
As shown for the WT cells in Fig. 7, release of Rav2-GFP from the vacuolar membrane upon glucose deprivation results in a corresponding reduction in Pearson coefficient for colocalization with FM4-64, and the Pearson coefficient increases upon glucose readdition. We repeated this analysis for the pfk2⌬, ira2⌬, and sch9⌬ mutants, analyzing Rav2-GFP localization in the presence of glucose, 15 min after glucose removal and 15 Error bars, S.D. For each condition, the four strains were compared by one-way analysis of variance, and significant differences from the WT strain are indicated: ***, p Ͻ 0.0005; **, p Ͻ 0.005; *, p Ͻ 0.05; ns, p Ͼ 0.05.

Requirements for vacuolar localization of yeast RAVE
min after glucose readdition. As shown in Fig. 7, there is significantly less RAVE at the membrane under all conditions in the ira2⌬ strain. In contrast, Rav2-GFP localization in the pfk2⌬ mutant is significantly different from WT cells only under conditions of glucose deprivation, where pfk2⌬ maintained significantly higher levels of Rav2-GFP at the vacuole. The sch9⌬ mutant had lower levels of vacuolar Rav2-GFP than WT cells in the presence of glucose (both before and after glucose deprivation) but similar levels of vacuolar Rav2-GFP during glucose deprivation. Interestingly, these results suggest that steadystate localization of RAVE does not always reflect the previously observed localization in V 1 in the mutants.

Discussion
The data above suggest that the information required for glucose-dependent recruitment of RAVE to the vacuolar membrane resides in the interaction of Rav1 with Vph1-containing V o sectors. This is an important insight, because it suggests that RAVE may play a more active role in V-ATPase reassembly than previously appreciated. Earlier work has established that RAVE can bind to both cytosolic V 1 and V 1 subunit C and has highlighted RAVE in a passenger role, accompanying V 1 subunits to the membrane upon glucose addition and potentially aligning them for reassembly. The results reported here indicate that RAVE can relocalize in response to glucose levels independent of its V 1 subunit C and V 1 subcomplex partners (Fig. 2) (19). RAVE can even be recruited reversibly to the membrane under conditions where the V-ATPase does not disassemble (the vma11 E145L mutant; Fig. 3). These and previous results support the Rav1-Vph1 interaction as a key, glucose-sensitive interaction and imply that Rav1 can recognize Vph1 even when V 1 is bound, consistent with the observed presence of RAVE at the vacuolar membrane in the presence of glucose.
A small, conserved, six amino acid motif located 88 amino acids upstream of the most conserved region of Rav1 is essential for RAVE function, and specifically for interaction of RAVE with Vph1 and recruitment to the vacuolar membrane. Rbcn3␣ subunits in higher eukaryotes share homology with Rav1 but are much larger. This motif is positioned 130 -208 amino acids upstream of the most conserved sequence (corresponding to amino acids 840 -1125 of yeast Rav1) in the higher eukaryotic Rbcn3␣ subunits shown in Fig. 3A, suggesting its position may be relevant for function. It is unlikely, however, that this region is the sole point of contact between Rav1 and Vph1, as we have previously shown that the most conserved fragment of Rav1 (Rav1(840 -1125)), which does not contain this motif, also binds to Vph1NT in vitro (29). Nevertheless, it is notable that the rav1 6⌬ mutation is sufficient to generate V-ATPase assembly and activity defects comparable with that seen in a complete RAV1 deletion mutant (Fig. 6).
We tested glucose-dependent localization of RAVE in three mutants involved in nutritional signaling, chosen because they have been reported to affect the level of V 1 bound to V o in the presence or absence of glucose (30 -32). However, the steadystate localization of GFP-tagged RAVE in these mutants did not parallel the effects of the mutations on V 1 localization. Specifically, the two mutants that were reported to induce increased V 1 -V o assembly, ira2⌬ and sch9⌬, actually had significantly lower levels of RAVE at the vacuolar membrane under most conditions. In contrast, the pfk2⌬ mutant, which was reported to slow reassembly of V 1 with V o after glucose deprivation, maintained higher levels of RAVE at the vacuolar membrane during glucose deprivation. Although somewhat unexpected, these results emphasize that RAVE can move independently of V 1 . This might be expected, because the very different levels of RAVE and V 1 subunits in yeast cells suggest that RAVE acts catalytically in V-ATPase assembly, requiring multiple cycles of recruitment from cytosol to vacuole to restore V-ATPase levels after glucose deprivation and readdition.
It is possible that localization of both the RAVE and V 1 complexes arises from an equilibrium of release and reassociation with the vacuolar membrane that can be pushed in the disassembly or reassembly direction in the presence or absence of glucose. In WT cells maintained in glucose, the membranebound forms of RAVE and V 1 predominate, whereas in the absence of glucose, cytosolic levels increase. However, the mutations perturb these equilibria, resulting in different steady-state localizations of RAVE and V 1 . There may also be states between fully membrane-bound and fully released for both complexes. Such intermediate states could account for the presence of more stably assembled V-ATPase complexes in the presence of high extracellular pH (33) and less stably assembled complexes in the presence of low phosphatidylinositol 3,5-bisphosphate (34). In addition, when viewed by fluorescence microscopy, a higher proportion of GFP-tagged V 1 subunits appear to be at or near the vacuolar membrane in the absence of glucose than would be expected from biochemical experiments with lysed cells or isolated vacuoles (35). These V 1 complexes may be in an intermediate disassembled state that is easily dissociated during lysis. Very little RAVE complex is present in isolated vacuolar membranes under any condition, also suggesting a fragile attachment. Additional studies of the spatial and temporal behavior of RAVE and V-ATPase subcomplexes, in mutants with varied nutrient signaling capacities, will be needed to address these questions.
Taken together, the data shown here indicate that RAVE can operate much more independently from the V-ATPase and its V 1 and V o domains than previously anticipated. The RAVE complex must also be considered a central player in glucose signaling of V-ATPase assembly state.

Materials and growth media
Oligonucleotides were purchased from MWG Operon. Sepharose A beads were purchased from GE Healthcare. Anti-Myc mAb (9E10) was purchased from Santa Cruz Biotechnology, Inc. Anti-FLAG M2 resin, mouse anti-FLAG antibody, and FLAG peptide were purchased from Sigma. Amylose and TALON resin were purchased from New England Biolabs and Clontech, respectively.
Media for yeast and Escherichia coli growth were purchased from Fisher. Yeast were maintained in either yeast extract-peptone-2% dextrose (YEPD) medium, buffered to pH 5.0 with 50 mM potassium phosphate and 50 mM potassium succinate, or in fully supplemented minimal synthetic complete (SC) medium

Requirements for vacuolar localization of yeast RAVE
with individual amino acids omitted for selection as indicated.
The Rav Ϫ phenotype was tested by comparing growth on YEPD plates buffered to pH 5.0 with growth on YEPD plates buffered to pH 7.5 with 50 mM MES and 50 mM MOPS containing 60 mM CaCl 2 at 30°C and 37°C. E. coli were maintained in either Luria broth (LB) or rich broth (LB supplemented with 2% glucose).

Yeast strain construction
All strains used this in this work were derived from either SF838-5A, BY4741, or BY4742 background. Strain genotypes and their sources are listed in Table 1. Oligonucleotides used to generate strains are listed in Table 2. Strains containing Rav1 or Rav2 C-terminally tagged with GFP or 13 copies of the Myc epitope were generated as described previously (19,36). Rav1 C-terminally tagged with 13 copies of the Myc epitope was generated as described previously (16,17).
Strains harboring the rav1 6⌬ and rav1 YH-AA mutations were constructed by first replacing the sequence for amino acids 730 -830 of RAV1 with the URA3 gene and then replacing the URA3 gene with a mutagenized RAV1 fragment. To generate the rav1 730 -830⌬::URA3 strain, the URA3 gene was amplified from pRS316 using oligonucleotides Rav1 730 -830 Ura For and Rav1 730 -830 Ura Rev ( Table 2). The RAV1 mutations were introduced by a variation of QuikChange mutagenesis (Agilent), into a fragment encoding Rav1 679 -898 cloned into pGEM T-Easy (Promega) (19). PCR amplification with mutagenic primers Rav1 6⌬ For and Rav1 6⌬ Rev, Rav1 YH-AA For and Rav1 YH-AA Rev, or Rav1 K773A For and Rav1 K773A Rev introduced the rav1 6⌬, rav1 YH-AA, and Rav1 K773A mutations, respectively. The product was treated with DpnI to remove the template plasmid before transformation in E. coli strain DH5␣. Plasmids were sequenced to confirm the presence of the desired mutation, digested with EcoRI, and then transformed into the rav1 730 -830⌬::URA3 strain. The transformation mixture was plated on YEPD, pH 5, and grown overnight at 30°C and then replica-plated onto SDϪUra plates containing 5-fluoro-orotic acid (37) to select for replacement of URA3. Integration of the desired mutant allele was verified with PCR and DNA sequencing. Deletion of RAVE or V-ATPase subunits was performed as described (19).

Bacterial expression constructs, protein purification, and protein pulldowns
MBP-Vph1NT and WT MBP-Rav1(679 -898)-His 6 constructs were generated as described previously (19). Rav1 679 -898 was mutagenized and verified as described above. Protein expression of MBP-Vph1NT (amino acids 1-372) and MBP-Rav1(679 -898)-His 6 was as described (19,38). Rav1 and Vph1 constructs were inserted into BL-21 or Rosetta cells, respectively, for expression. Protein expression was induced by the addition of 0.5 mM isopropyl ␤-D-thiogalactopyranoside cells at a density of 0.5-0.6 OD/ml, and growth was continued for 2 h at 37°C for Rav1 or 16 h at 18°C for Vph1. MBP-tagged proteins were affinity-purified on amylose columns, as described (19). Peak fractions were pooled and incubated with 10 g of Prescission protease overnight at 4°C to cleave off the MBP.
After overnight cleavage, proteins were dialyzed overnight in TALON equilibration buffer (50 mM sodium phosphate, 150 mM NaCl, 5 mM beta-mercaptoethanol, pH 7.4). Rav1 (679 -898)-His 6 was then mixed with 300 l of TALON resin in combination with its binding partner, at a ratio of 1:5 Rav1(679 -898)-His 6 with Vph1-NT, and gently rocked at 4°C for 2 h. For each sample, 30 l from a total volume of 1.2 ml was reserved as an input fraction. The reserved input was combined with 30 l of cracking buffer (50 mM Tris-HCl, pH 6.8, 8 M urea, 5% SDS, 1 mM EDTA), and 2.5 l of the mixture was loaded for the gels in Fig. 4. The resin-protein mixture was poured into a column and washed with 3 ml of TALON wash buffer (50 mM sodium phosphate, 150 mM NaCl, 5 mM imidazole, pH 7.4), and proteins were eluted with 50 mM sodium phosphate, 150 mM NaCl, 250

Requirements for vacuolar localization of yeast RAVE
mM imidazole, pH 7.4, into 500-l fractions. Elution fractions were TCA-precipitated and dissolved in 80 l of cracking buffer; 5 l from each sample was separated by SDS-PAGE and transferred to nitrocellulose. The resulting blots were cut, and the top was probed with the anti-Vph1 (10D7), and the bottom portion was probed with mouse anti-His 6 antibody (Gen-Script), followed by alkaline phosphatase-conjugated secondary antibody.

Immunoprecipitations and vacuolar purification
Co-immunoprecipitation of RAVE and V 1 was based on methods described in Smardon et al. (17). For the experiments described here, cells containing WT or mutant Rav1 tagged with a Myc13 epitope were grown to log phase, and 50 OD 600 units were pelleted by centrifugation. The cells were then lysed in cold extraction buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 0.1 mM EDTA) by agitation with glass beads, and cell debris and beads were removed by centrifugation. Sepharose A beads were added to preclear the suspension and then removed by centrifugation. 6 l of anti-Myc mAb (9E10) was added to the supernatant, and the mixture was rocked at 4°C for 6 h. Sepharose A beads were then added to the antibody-lysate mixture, which was rocked for another 2 h at 4°C. After incubation, Sepharose beads were pelleted and washed three times with cold extraction buffer. 80 l of hot cracking buffer was added directly to the beads, and immunoprecipitated protein was eluted at 95°C for 10 min. Samples were separated by SDSpolyacrylamide gels, transferred to nitrocellulose, and then immunoblotted with anti-Myc (9E10) or anti-V 1 A (8B1F3) mAb followed by alkaline phosphatase-conjugated secondary antibody. Vacuolar vesicles were isolated and characterized as described previously (34).

Fluorescence microscopy
Yeast strains expressing Rav1-GFP or Rav2-GFP were grown to log phase in SC medium. Localization of GFP-tagged proteins was visualized by differential interference contrast (DIC) microscopy or fluorescence microscopy using a GFP filter set on a Zeiss Imager.Z1 fluorescence microscope equipped with a Hamamastu CCD camera and AxioVision software. The micrographs were assembled into figures using Adobe Photoshop CC. For all figures, the images shown are representative of at least 50 images obtained in at least two different experiments.
To observe the localization of GFP-tagged proteins under conditions of glucose deprivation and readdition, cells were grown overnight to log phase in SC medium. Cells from the overnight culture were imaged for the "ϩglucose" condition. For the "Ϫglucose" condition, 1 ml of culture was removed, washed, and resuspended in SC medium lacking glucose and incubated at 30°C for 15 min before imaging. For the "glucose readdition" condition, glucose was added back to the glucosedeprived cells to a concentration of 2% (w/v), and incubation was continued at 30°C for an additional 15 min. Cells were then imaged at room temperature for a maximum of 15 min. RAVE subunit fluorescence images in Figs. 1-3 were obtained by importing .zvi files from the microscope into ImageJ, manually setting the brightness and contrast range to 0 -3000, and then subtracting background using the "rolling ball radius of 50.0 pixels" setting. For comparison of glucose-dependent localization in WT and vma4⌬ cells (Fig. 2), lines were drawn across fluorescence images of cells with well-defined vacuoles, and the fluorescence intensities across the line were determined using the "plot profile" feature. The intensity data were imported into Excel and plotted directly for the single-cell scans shown in Fig.  2A. Maximum values for each line scan were determined and divided by the average value of the ϩglucose sample to obtain the normalized maximal intensities in Fig. 2B. Data in Fig. 2B were plotted using GraphPad Prism.
For co-localization of Rav2-GFP with the vacuolar membrane, cells were grown overnight to log phase in YEPD. A volume of cells corresponding to 3 OD 600 units was pelleted by centrifugation and resuspended in 1 ml of YEPD, and 8 M FM4-64 was added (39). Tubes were covered in foil and rocked at 30°C for 30 min. Cells were diluted to 15 ml with YEPD, spun down to remove free FM4-64, and then resuspended in 6 ml of YEPD and rocked at 30°C for 90 min. Cells were washed twice with 15 ml of SC medium. After washing, cells were resuspended in 1 ml of SC medium, allowed to recover for 15 min, and then imaged. Images were obtained under DIC, GFP, and Texas Red optics in succession, from cells with glucose present, absent, and readded as above. To correct for drift during the relatively long exposure times required for visualizing Rav2-GFP, the resulting micrographs were layered in Photoshop CC, using the vacuole as a guide; the GFP image was placed over the DIC image, where vacuoles are prominent indentations, and then the Texas Red image was placed over GFP.
Co-localization of Rav2-GFP with vacuolar FM4-64 was quantified using Just Another Colocalization Plugin (JACoP) (40) in ImageJ to calculate the Pearson coefficient. Briefly, images were cropped from aligned GFP and FM4-64 micrographs and imported into ImageJ and then JACoP. Pearson coefficients were calculated with the GFP channel serving as Image A and FM4-64 serving as Image B. The depicted Pearson coefficients in Fig. 6 represent an average of at least 24 cells for each strain. Data were plotted using GraphPad Prism, and sta-

Requirements for vacuolar localization of yeast RAVE
tistical significance was assessed using a one-way analysis of variance to compare WT and mutant strains under each condition.

Sequence analysis
Sequence analysis and identification of the conserved motif was performed using BLASTP on the NCBI BLAST server.