Molecular Interactions and Cellular Itinerary of the Yeast RAVE (Regulator of the H+-ATPase of Vacuolar and Endosomal Membranes) Complex*

Background: The RAVE complex is required for glucose-sensitive V-ATPase assembly, but its mechanism is unknown. Results: Newly defined RAVE-V-ATPase interactions suggest assembly mechanisms. Conclusion: RAVE shows glucose-sensitive interactions with vacuolar membranes and may help orient V-ATPase subunit C during assembly. Significance: Mammalian rabconnectins likely require similar V-ATPase interactions to support V-ATPase assembly and may be reversibly recruited to membranes. The RAVE complex (regulator of the H+-ATPase of vacuolar and endosomal membranes) is required for biosynthetic assembly and glucose-stimulated reassembly of the yeast vacuolar H+-ATPase (V-ATPase). Yeast RAVE contains three subunits: Rav1, Rav2, and Skp1. Rav1 is the largest subunit, and it binds Rav2 and Skp1 of RAVE; the E, G, and C subunits of the V-ATPase peripheral V1 sector; and Vph1 of the membrane Vo sector. We identified Rav1 regions required for interaction with its binding partners through deletion analysis, co-immunoprecipitation, two-hybrid assay, and pulldown assays with expressed proteins. We find that Skp1 binding requires sequences near the C terminus of Rav1, V1 subunits E and C bind to a conserved region in the C-terminal half of Rav1, and the cytosolic domain of Vph1 binds near the junction of the Rav1 N- and C-terminal halves. In contrast, Rav2 binds to the N-terminal domain of Rav1, which can be modeled as a double β-propeller. Only the V1 C subunit binds to both Rav1 and Rav2. Using GFP-tagged RAVE subunits in vivo, we demonstrate glucose-dependent association of RAVE with the vacuolar membrane, consistent with its role in glucose-dependent V-ATPase assembly. It is known that V1 subunit C localizes to the V1-Vo interface in assembled V-ATPase complexes and is important in regulated disassembly of V-ATPases. We propose that RAVE cycles between cytosol and vacuolar membrane in a glucose-dependent manner, positioning V1 and V0 subcomplexes and orienting the V1 C subunit to promote assembly.

peripheral complex containing sites of ATP hydrolysis, called V 1 , and an integral membrane complex containing the proton pore, called V o . The V 1 complex contains eight subunits designated A to H, and the V o complex contains six subunits designated a, c, cЈ, cЉ, d, and e. In yeast, the V-ATPase subunits are encoded by single-copy VMA genes, with the exception of the V o a subunit, which has two isoforms encoded by the VPH1 and STV1 genes (1). ATP-driven proton transport requires stable association of the V 1 and V o complexes; free V 1 is inactive as a Mg 2ϩ -dependent ATPase, and free V o sectors are closed to H ϩ transport (3)(4)(5). Inactivation of the disassembled sectors is physiologically significant, because V-ATPases are regulated by a reversible disassembly mechanism through which V 1 sectors are partially detached from V o in response to glucose deprivation and reassembled upon glucose readdition (6 -8).
The yeast RAVE (regulator of the H ϩ -ATPase of vacuoles and endosomes) complex is required for both the initial biosynthetic assembly of the V-ATPase and for reassembly of the complex after glucose deprivation and readdition (9,10). RAVE contains three subunits: Rav1, Rav2, and Skp1 (9). Initial experiments established that Rav2 and Skp1 bind to Rav1, but not to each other, and that RAVE exists as a stable complex with Rav1 as the central component (9,10). Skp1 participates in multiple cellular complexes, including the SCF (Skp1-cullin-F-box) ubiquitin ligases, and loss of SKP1 is lethal (11,12). In contrast, rav1⌬ and rav2⌬ mutants exhibit a Vma Ϫ growth phenotype, characterized by poor growth at high pH and/or elevated calcium concentrations, but unlike other vma mutants only display this phenotype at elevated temperature (13).
Although the composition of the RAVE complex is quite well established, neither the basis of its activity in promoting V-ATPase assembly nor the sites of interaction between Rav1 and its partners are understood. Under conditions of V-ATPase disassembly, RAVE binds to V 1 sectors in cytosolic fractions via the E and/or G subunits of V 1 , but this interaction is not intrinsically glucose-sensitive (10,11). Rav1 also has a binding site for the cytosolic N-terminal domain of V o subunit Vph1 (13). The V 1 C subunit is present at the interface of the V 1 and V o sectors and is thus positioned to play a critical role in reversible disas-sembly of the V-ATPase (14 -16). Interestingly, this subunit is released from both the V 1 and V o sectors when the V-ATPase disassembles (6,17). Subunit C binds to Rav1 independently of the V 1 subcomplex and is the only subunit that has shown any binding to Rav2 (18). Binding of RAVE to detached V 1 sectors, V 1 subunit C, and V o subunit Vph1 is intriguing, because it suggests that RAVE might act by bringing these subunits together to establish or re-establish V 1 -V o assembly. However, without any information about structure or binding sites on RAVE, this proposed activity remains speculative.
RAVE function in V-ATPase assembly appears to be conserved in higher eukaryotes. Higher eukaryotes contain larger rabconnectin-3 complexes consisting of two subunits, rabconnectin-3␣ (also called DMXL2 in humans) and rabconnectin-3␤ (also called WDR7 in humans) (19). Mutations in either subunit have been shown to result in organelle acidification defects (20 -22), and in zebrafish hair cells, lower levels of V 1 -V o association were observed in a rabconnectin loss of function mutant (22). V-ATPase subunits co-immunoprecipitate with DMXL2 (23). Although other fungi contain RAV2 homologues, there is no sequence homology between the rabconnectin subunits and RAV2. In contrast, a highly conserved region has been identified in fungal RAV1 homologues and in mammalian rabconnectin-3␣ subunits and designated the "Rav1-C superfamily" (24). This region extends from amino acids 571 to 1191 of yeast RAV1, but the RAV1 sequence between amino acids 840 and 1125 is the most highly conserved across all species (25). In addition, the N-terminal 240 amino acids of yeast RAV1 also align with the N termini of rabconnectin-3␣ homologues of higher organisms. Based on sequence comparisons with both other fungi and higher eukaryotes, we identified six regions of Rav1 that are conserved to differing degrees (25).
Many questions remain about RAVE/rabconnectin structure and function. Although multiple interactions between the V-ATPase and yeast RAVE have been identified, these interactions have not been mapped to specific regions of the RAVE complex. Mapping these interactions could provide insights into the mechanism of RAVE-promoted assembly. In addition, although RAVE participates in glucose-sensitive reversible disassembly of the V-ATPase, it has not been determined which interactions with RAVE might be glucose-sensitive. Given the high level of sequence conservation among all eukaryotic V-ATPases and the evidence that RAVE and rabconnectin complexes both regulate V-ATPase activity at the level of V 1 -V o assembly, we hypothesized that functionally critical V-ATPase subunit binding sites might be located in the conserved regions of Rav1. In this work, we test this hypothesis by generating a series of deletions and fragments of Rav1 and evaluating their binding to V-ATPase subunit partners. In addition, we demonstrate that the RAVE complex shows glucose-dependent association with the vacuolar membrane in vivo, consistent with its role in glucose-dependent V-ATPase assembly.

Experimental Procedures
Materials and Growth Media-Oligonucleotides were purchased from MWG Operon. Anti-FLAG-M2 resin, mouse anti-FLAG-M2 antibody, and FLAG peptide were purchased from Sigma. Amylose resin was purchased from New England Biolabs, and glutathione-Sepharose was from GenScript. TALON resin, for purification of His 6 -tagged protein, was obtained from Clontech. Anti-myc monoclonal antibody (9E10) was purchased from Roche Applied Science. Media for growth of yeast and Escherichia coli were from Fisher Scientific.
Yeast cells were maintained in rich medium, YEPD (yeast extract, peptone, 2% dextrose) medium buffered to pH 5 with 50 mM potassium phosphate and 50 mM potassium succinate, or in fully supplemented minimal medium (SC) with individual nutrients omitted for selection as indicated. The Vma Ϫ phenotype was tested by comparing growth of cells on YEPD plates buffered to pH 5 to growth on YEPD plates buffered to pH 7.5 with 50 mM MES and 50 mM MOPS, to which 60 mM CaCl 2 was added.
Bacterial Expression Constructs and Protein Purification-The MBP (maltose-binding protein)-Rav1NT-His 6 expression plasmid was constructed by amplifying the sequence corresponding to amino acids 2-729 of RAV1 by PCR using oligonucleotides Rav1 NTerm BamHI and Rav1 NTerm SalI (which introduces the His 6 tag just before the stop codon) and a combination of LA-Taq (TaKaRa) and Pfu (New England Biolabs) polymerases to enhance fidelity. The resulting PCR fragment was cloned into the pGEM T-Easy vector (Promega) after A-tailing with Taq polymerase (Fisher). The insert was verified by sequencing at the Upstate DNA sequencing facility. Digestion with BamHI and SalI removed the insert, which was ligated to the pMAL-pAse vector (31) that had been digested with BamHI and SalI. The MBP-RAV2-FLAG construct was constructed by PCR amplifying the entire RAV2 sequence using oligonucleotides Rav2BamHI forward and Rav2SalI reverse (which introduces a single FLAG tag immediately before the stop codon). The amplified fragment was cloned into the  pGEM T-Easy and pMAL-pAse vectors as described above. The MBP-RAV1(679 -898)-His 6 expression plasmid was constructed in two steps. First, the RAV1 fragment corresponding to amino acids 679 -898 was amplified by PCR, using oligonucleotides (Rav1679RI and Rav1898-SalI), and this fragment was cloned into the pMAL-pAse vector as described above. The His 6 tag was subsequently added by inverse PCR mutagenesis using oligonucleotides (Rav(679 -898)-HisF and Rav(679 -898)-HisR) to generate a mutagenized PCR product, followed by digestion of the unmutagenized template with DpnI and transformation of E. coli DH5␣ cells. Incorporation of the tag was confirmed by sequencing. Construction of GST-tagged Rav1(840 -1125) (13), MBP-Vph1NT (31), and MBP-V 1 C (15) expression vectors have been described. All of the above expression plasmids were introduced into E. coli expression strain BL21, with the exception of MBP-V 1 C, which was expressed in the Rosetta2 strain. Cells containing the expression plasmids were grown in rich broth (Luria broth supplemented with 2% glucose and 125 mg/ml ampicillin) to a density of 0.5-0.6 A 600 /ml, and then 0.5 mM isopropyl ␤-D-thiogalactopyranoside was added to induce expression. Induction conditions were as follows: 6 h at 30°C for MBP-Rav1NT-His 6 , 2 h at 37°C for MBP-Rav2-FLAG and MBP-Rav1(679 -898)-His6, and 16 h at 19°C for MBP-Vph1NT(1-372). After inductions, cell pellets were frozen at Ϫ80°C and then lysed by sonication after addition of 50 ml of amylose column buffer (20 mM Tris-HCl, 200 mM NaCl, 1 mM EDTA, pH 7.2) per liter of cell pellet. PMSF and ␤-mercaptoethanol were added to 1 and 5 mM, respectively, immediately before lysis. The MBP-tagged proteins were then affinity-purified by binding to amylose columns, washing with 12 column volumes of amylose column buffer, and elution with 10 mM maltose in column buffer. In experiments where MBP was removed, the peak fractions containing MBP fusion protein were combined and incubated with 10 g of Prescission protease overnight at 4°C. FLAG-tagged proteins were further purified on a anti-FLAG-M2 affinity column using FLAG column buffer (20 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, pH 7.2) for loading and washes. Tagged protein was eluted in the same buffer containing 100 g/ml FLAG peptide. For purification on TALON resin, His 6 -tagged proteins were dialyzed into 50 mM sodium phosphate, pH 7.5, 150 mM NaCl containing 5 mM ␤-mercaptoethanol. The His 6tagged protein was mixed with 0.5 ml of TALON resin, either in the presence or absence of its interacting partner, and incubated with mixing for 1 h. at 4°C. The resin was then poured into a column for washing with the same buffer containing 5 mM imidazole, followed by elution with 50 mM phosphate, pH 7.5, 300 mM NaCl, and 250 mM imidazole. 0.5-ml fractions were collected during the elution, and peak fractions, based on absorbance at 280 nm, were combined and precipitated with 10% trichloroacetic acid. Purification of GST-Rav1(840 -1125) was described previously (13).
All of the expressed protein-protein interactions were observed at least three times, except the Rav1NT-Rav2 interaction experiment, which was done twice. The gels and immunoblots in Figs. 3-5 show representative experiments.
Immunoprecipitations and Immunoblots-For co-immunoprecipitations of myc-tagged Rav1 deletions with V 1 B and Skp1, yeast cytosolic fractions were prepared from each rav1 deletion strain as previously described (18). Protein concentrations of the cytosolic fractions were measured by Lowry assay (32); 0.4 mg of protein was precipitated with trichloroacetic acid (input), and 4.0 mg of cytosol were combined with 100 l of anti-V 1 B monoclonal antibody (13D11) or 6 l of anti-myc monoclonal antibody (9E10) followed by the addition of 60 l of a 50% (v/v) suspension of protein A-Sepharose CL4B (Sigma). The input and immunoprecipitates were solubilized at 75°C in cracking buffer (50 mM Tris-Cl, pH 6.8, 8 M urea, 5% SDS, 5% ␤-mercaptoethanol) for analysis on SDS-PAGE. Western blots were probed with mouse monoclonals 9E10 (anti-myc) or 13D11 (anti-yeast Vma2 (33)) or with rabbit polyclonal anti-Skp1 antibody (a generous gift from Ray Deshaies (Caltech)).
For co-immunoprecipitation of FLAG-V 1 C with Rav2myc 13 , whole cell lysates were prepared as previously described (10). Briefly cells (30 A 600 of each strain) were resuspended in TBS buffer (50 mM Tris-HCl, 150 mM NaCl, pH 7.4) and lysed by agitation with glass beads. Lysates were incubated with 20 l of anti-FLAG M2 resin at 4°C and washed with TBS. The FLAG-V 1 C fusion protein was eluted with 100 g/ml FLAG peptide in TBS. Eluted proteins were analyzed on SDS-PAGE, and Western blots were probed with anti-myc and anti-FLAG antibodies.
Two-hybrid Analysis-The two-hybrid plasmids pAS-RAV1, pACT-RAV2, pACT-VMA4, and pACT-VMA5 and the plasmid pACT-VPH1-NT containing the first 406 N-terminal amino acids of VPH1 have been described previously (18,31). pAS2 plasmids (34) containing RAV1 fragments were constructed as follows. A RAV1 fragment encoding amino acids 2-240 was amplified from yeast genomic DNA using oligonucleotides RAV1BamHI and RAV1-NT2H-3. The PCR product was then cloned into the pGEM-T Easy and sequenced as described above. The 716-bp BamHI/SalI fragment was excised and cloned into the pAS2 vector. Rav1 fragments containing amino acids 679 -840 and 840 -1125 were cloned into the pAS2 vector in a similar manner after PCR amplification with oligonucleotide pairs RAV1-2H-679 and RAV1-2H-840 and oligonucleotide pairs RAV1-CT2H-5 and RAV1-CT2H-3, respectively. Both PCR products were cloned into the pGEM-T Easy vector and then into the pAS2 vector as described above.
The pAS2 plasmids containing the RAV1 C-terminal deletion mutants, ⌬840-stop and ⌬940-stop were constructed as follows. Yeast strains containing the corresponding C-terminal deletions in Rav1 described above were used as templates and the oligonucleotide pairs YJR10 and RAV1M6 for ⌬840-stop and YJR10 and RAV1M7 for ⌬940-stop. The PCR products generated were then cloned in the pGEM-T Easy cloning vector that was then cleaved with EcoRI-SmaI to release the inserted C-terminal RAV1 fragments. The Rav1 fragments containing the C-terminal RAV1 deletions were cloned into the pAS-RAV1 plasmid and sequenced for accuracy. For internal RAV1 deletions ⌬840 -1125 and ⌬580 -839, the following method was used. A RAV1 PCR fragment was generated using wild type genomic DNA as template and the oligonucleotide pair YJR10 and RAV1M2. The ⌬840 -1125 mutation was introduced using the pAltered Sites protocol (Promega) and oligonucleotides R1M14F and R1M14R. The deletion mutation was sequenced for accuracy, and the mutant RAV1 was cloned into the pAS-RAV1 plasmid that had been cleaved with Cla1-Sal1 to remove the unmutagenized Cla1-Sal1 fragment. The ⌬580 -839 was constructed in a similar manner using oligonucleotides R1M11F and R1M11R. The pAS2 plasmids containing the N-terminal deletions ⌬2-240 and ⌬98 -240 were constructed by amplifying the desired fragments by PCR using the original pAS-RAV1 plasmid as template and the oligonucleotides RAV1M8-2Hyb2 and YJR11 for ⌬2-240 and RAV1M9-2Hyb and YJR11 for ⌬98 -240. The resulting PCR products were cloned into the pGEM-T Easy vector, sequenced for accuracy and subsequently excised with Sma1 and Cla1. The Sma1-Cla1 fragment containing the ⌬2-240 and the ⌬98 -240 mutations were cloned into the pAS-RAV1 plasmid that had been cleaved with Sma1-Cla1 to remove the original RAV1 N-terminal fulllength fragment.
pAS and pACT plasmid constructs were transformed into the two-hybrid reporter strains PJ69-4A (MATa) and PJ69-4␣ (MAT␣), respectively. MATa strains containing pAS plasmids were crossed to MAT␣ strains containing pACT plasmids, and diploids were selected by growth on SC medium lacking leucine and tryptophan. The resulting diploids were tested for expression of two-hybrid reporter genes by plating onto plates lacking adenine and histidine, as well as leucine and tryptophan (35).
Fluorescence Microscopy-Yeast strains expressing Rav1-GFP and Rav2-GFP were grown to log phase in SC medium and visualized directly using a Zeiss Imager.Z1 fluorescence microscope equipped with a Hamamatsu CCD camera and Axio-Vision software. Micrographs were assembled into figures using Adobe Photoshop CS4.

Rav1 Is Predicted to Have Structurally Distinct N-and C-terminal
Halves-There are no three-dimensional structures for Rav1, Rav2, or either of the rabconnectin subunits, although we did previously identify six regions of Rav1 with varied levels of sequence conservation (Fig. 1A (25)). Secondary structure prediction (Psipred (36)) indicates that the N-terminal 720 amino acids of Rav1 have a very high content of ␤-sheet, amino acids 746 -1194 have a high helical content, and the C-terminal 153 amino acids are likely to be very disordered (Fig. 1A). The amino acids between the ␤-sheet and ␣-helix rich regions (amino acids 721-745) are predominantly disordered, suggesting that the two halves of Rav1 might behave somewhat independently, with the most highly conserved Rav1-C superfamily sequences mapping predominantly to the C-terminal half rich in ␣-helices, and the conserved N terminus mapping to the region rich in ␤-sheet. We combined the information from sequence comparisons and secondary structure prediction of Rav1 to design deletion mutations and fragments that could be tested for interactions with Rav2, Skp1, and V-AT-Pase subunits.
The C-terminal Half of Rav1 Is Required for Efficient Binding to V 1 and Skp1-To begin to address the distribution of binding sites in the C-terminal half of Rav1, we used information from sequence comparisons to design myc-tagged C-terminal deletions through the most highly conserved region of yeast RAV1 and then inserted them at the RAV1 locus (Fig. 1A). Fig. 1B demonstrates that the shortest deletion, which removed only the very poorly conserved last 86 amino acids of Rav1, supported wild type growth on YEPD, pH 7.5 plates containing 60 mM CaCl 2 at 37°C, suggesting that the RAVE complex is functional. In contrast, all of the other truncations resulted in a rav1⌬ phenotype, characterized by growth on YEPD plates buffered to pH 5 but not on YEPD, pH 7.5 plates with added CaCl 2 , indicating that they impact RAVE function. The truncations produced stable mutant Rav1 proteins, as shown by Western blotting with anti-myc antibodies (Fig. 1C). We immuno-  (25), and deletions were designed in part based on the regions of differential homology. Deletions were integrated at the RAV1 genomic locus, and wild type and mutant proteins were tagged with the myc 13 epitope at their C termini as described under "Experimental Procedures." B, strains were tested for a rav1⌬ phenotype by growing strains to log phase in YEPD pH 5, diluting to an A 600 of 0.5 and serially diluting 10-fold in a microtiter plate. Cells were transferred to YEPD, pH 5 plates and grown for 3 days at 37°C and YEPD, pH 7.5 ϩ 60 mM CaCl 2 plates and grown for 5 days at 37°C. C, cytosolic fractions were prepared from each strain, and 0.4 mg of cytosol from each strain was directly trichloroacetic acid-precipitated and run as Input. Cytosolic proteins (4 mg) from each strain were immunoprecipitated separately with an antibody against V 1 subunit B (13D11) or a monoclonal antibody against the myc epitope (9E10). Immunoprecipitations were separated by SDS-PAGE and immunoblotted with anti-myc or polyclonal Skp1 antibody. The experiment shown is representative of three independent experiments. WB, Western blot. NOVEMBER 13, 2015 • VOLUME 290 • NUMBER 46 precipitated myc-tagged wild type and mutant Rav1 and assessed co-precipitation of Skp1 by immunoblotting. Similar levels of Skp1 were co-precipitated with full-length Rav1-myc 13 and the shortest truncation (⌬1126-stop), but all of the other truncations immunoprecipitated much less Skp1. In contrast, anti-V 1 B subunit antibodies co-precipitate the Rav1-myc 13 constructs ⌬1126-stop and ⌬941-stop mutations, as well as wild type Rav1-myc 13 . Only in the Rav1⌬841-stop did the V 1 interaction with Rav1 appear to be decreased. Although these data cannot prove direct binding between a deleted region and V 1 or Skp1, they do indicate a requirement for amino acids 1126 -1271 for Skp1 binding and amino acids 840 -941 for V 1 binding. In addition, they indicate that the final 86 amino acids of Rav1 are dispensable for RAVE function.

Interactions and Localization of the Yeast RAVE Complex
We previously identified a number of two-hybrid interactions of intact Rav1 with Rav2; V 1 subunits E, G, and C; and the N-terminal domain of V o subunit a isoform Vph1 (13,18). To identify regions of the Rav1 protein required for these interactions, we expressed fragments of Rav1 that contained fragments of the C-terminal conserved domain or deletions in this domain and then tested for interactions with the Rav1-binding partners in the two-hybrid assay (Fig. 2). In this version of the two-hybrid assay, interacting partners promote transcription of the ADE2 and HIS3 genes and enable diploid cells containing those partners to grow in medium lacking adenine and histidine. The Rav1 fragments were tested, and their interactions with Vma4 (V 1 E), Vma5 (V 1 C), Rav2, and Vph1NT (the N-terminal cytosolic domain of the Vph1 isoform of V o subunit a) are scored on the top of Fig. 2, and growth of the diploids on plates lacking adenine and histidine is shown at the bottom of Fig. 2. Consistent with immunoprecipitation results described above, the Rav1(840 -1125) fragment exhibited a strong two-hybrid interaction with V 1 E, one of the V-ATPase subunits shown to support the interaction of RAVE with the V 1 complex (10). No interactions with V 1 E were observed with the other two small fragments tested. We complemented the two-hybrid interactions with Rav1 fragments by testing two-hybrid interactions between Rav1 deletions and V 1 E. These interactions can be more complex but were consistent with the fragment interactions overall. A Rav1 deletion allele terminating at amino acid 840, as well as a deletion allele removing the highly conserved segment between amino acids 840 -1125, decreased the twohybrid interaction with V 1 E. However, the interaction with V 1 E remained strong in the deletion terminating at amino acid 940 and in a deletion removing amino acids 890 -940. This set of interactions highlights the importance of Rav1 amino acids 840 -890 for the Rav1-V 1 E interaction and also implicates this region in RAVE-V 1 binding, previously shown to depend on V 1 subunits E and G (10).
Binding of Rav2 to the N-terminal Conserved Domain of Rav1-The N-terminal 240 amino acids of Rav1 constitute the second most highly conserved region among Rav1 homologues (25). When this region was expressed separately in the twohybrid vector, it showed no interaction with V 1 E but did exhibit a two-hybrid interaction with V 1 C, Vph1NT, and Rav2 (Fig. 2). Deletion of the N-terminal 240 amino acids of the full Rav1 compromised the two-hybrid interactions of V 1 C, Vph1NT, and Rav2, as expected by the interaction of these proteins with the Rav1 N-terminal fragment. However, deletion of only amino acids 98 -240 of Rav1 largely restored the two-hybrid interaction with Rav2. These results suggest that Rav2 requires the first 98 amino acids for binding to Rav1.
We confirmed the interaction of Rav2 in the first half of Rav1 through expression of these two proteins in E. coli and testing for interaction in vitro. The first 729 amino acids of Rav1 were expressed with an N-terminal MBP tag and a C-terminal His 6 tag, and the full-length Rav2 was expressed with an N-terminal MBP and a C-terminal FLAG tag. Both proteins were purified via the MBP tag on amylose columns, and the MBP tag was cleaved from MBP-Rav2-FLAG (Fig. 3). When Rav2-FLAG was applied to the anti-FLAG affinity column, it eluted as a single band of the expected molecular mass (Fig. 3A). MBP-Rav1NT-His 6 did not bind to the FLAG resin without bound Rav2-FLAG (Fig. 3B). However, when MBP-Rav1-His 6 , which had been further affinity-purified via the His 6 tag, was added to the anti-FLAG affinity column to which Rav2-FLAG had been bound, it co-eluted with Rav2 (Fig. 3C). These data confirm that Rav2 can bind to the N-terminal region of Rav1, the region that is predicted to be rich in ␤-sheet.
Interactions of Rav1 with V o Subunit Vph1-We previously determined that V-ATPase complexes containing the Vph1 isoform of the V o a subunit require RAVE for assembly and function (13). We also found that Vph1NT bound to full-length Rav1 in the two-hybrid assay and determined that purified Vph1NT could bind to a fragment of Rav1 containing amino acids 840 -1125 (13). To further explore binding of Vph1NT on Rav1, we again examined interactions with fragments and deletions in RAV1 by two-hybrid assay. We expressed a small segment (Rav1 679 -840) extending from the end of the N-terminal domain through the beginning of the highly conserved C-terminal domain. This region displayed a strong two-hybrid interaction with Vph1NT, but not with V 1 subunits or Rav2 (Fig. 2). Vph1NT also showed a strong two-hybrid interaction with the fragment corresponding to the N-terminal 240 amino acids of Rav1. Interestingly, Vph1NT did not interact with any of the Rav1 deletion constructs, except for a weak interaction with the ⌬840 to stop construct, even though a number of these constructs contain the 1-240 fragment, 679 -840 fragment, or both of these fragments. Even the two-hybrid construct lacking only the functionally dispensable last 87 amino acids failed to interact with Vph1NT, although it interacted with all of the other partners tested. These results suggest that the Rav1(1-240) and (679 -840) fragments may contain binding sites for the Vph1NT but that these fragments either assume a different structure in the larger Rav1 constructs or contain binding sites that are masked in the larger constructs.
In light of the strong two-hybrid interaction between the Rav1(679 -840) fragment and Vph1NT, we expressed a fragment of Rav1 containing amino acids 679 -898 as an N-terminal MBP and C-terminal His 6 -tagged construct in E. coli and then purified the Rav1(679 -898)-His 6 protein and tested for binding of Vph1NT(1-372). As shown in Fig. 4, Vph1NT(1-372) binds to this fragment, consistent with a direct binding site in this region of Rav1. There was no binding of Vph1NT(1-372) to the TALON resin in the absence of the Rav1 fragment. Regions of Rav1 required for binding to Rav1 binding partners. Rav1 fragments and deletions were cloned into the pAS two-hybrid plasmid as described in "Experimental Procedures." All pAS plasmids containing Rav1 fragments or deletions were transformed into the MATa PJ694A two-hybrid reporter strain. Each of these strains was crossed separately to each MAT␣ PJ694␣ strain containing either pACT-VMA4, pACT-VMA5, pACT-RAV2, or pACT-VPH1-NT, and diploids containing both plasmids were selected on SC medium lacking leucine and tryptophan. Diploids were then tested for expression of the two-hybrid reporter genes by plating on plates lacking adenine and histidine. The results shown are representative of at least three biological replicates. NOVEMBER 13, 2015 • VOLUME 290 • NUMBER 46

Interactions and Localization of the Yeast RAVE Complex
Evidence for Binding Sites for the V 1 C Subunit on both Rav1 and Rav2-The V 1 C subunit is likely to be a critical player in reassembly of the V-ATPase. This subunit binds at the interface of V 1 and V o and is released from both sectors upon glucose deprivation (6,14,17). The two-hybrid experiments in Fig. 2 did not reveal a clear binding domain for subunit C. Previous twohybrid experiments indicated interactions of subunit C with both intact Rav1 and Rav2. There was considerable overlap between regions of Rav1 that bound to Rav2 in the two-hybrid assay shown in Fig. 2 and those that showed interactions with subunit C. This is suggestive of bridging interactions in the two-hybrid assay and could indicate that V 1 C is binding to Rav1 only through Rav2. However, there are also fragments, such as the Rav1⌬2-240 fragment, that exhibited a two-hybrid interaction with the C subunit but not to Rav2 (Fig. 2). To test whether there was direct binding between Rav1 and subunit C, we tested for binding of expressed and purified V 1 C to a previously described fusion of GST to Rav1(840 -1125) expressed in E. coli (13). As shown in Fig. 5A, subunit C co-eluted from glutathione-Sepharose with GST-Rav1(840 -1125), and the identity of subunit C was confirmed by immunoblotting. (No expressed subunit C bound to glutathione-Sepharose in the absence of GST-tagged Rav1.) These data support a direct interaction of subunit C with Rav1, even though the Rav(840 -1125) two-hybrid construct did not show binding to subunit C in Fig. 2. We next tested whether there was a separate site for subunit C binding on Rav2 by testing for interactions of the proteins expressed and purified from E. coli (Fig. 5B). Although purified V 1 C showed low level binding to anti-FLAG resin in the absence of Rav2-FLAG, there was much better binding in the presence of Rav2-FLAG, indicating a direct interaction. To confirm that the Rav2-V 1 C subunit interaction also occurred in yeast cells, we replaced the endogenous copy of the VMA5 gene with a FLAG-tagged version, in wild type and rav1⌬ mutant cells containing a C-terminally tagged Rav2p (Rav2myc 13 ). As shown in Fig. 5C, Rav2-myc 13 was co-immunoprecipitated by FLAG-tagged V 1 C from both wild type and rav1⌬ cells but was not bound to the anti-FLAG resin in the absence of a FLAG tag on subunit C. These results indicate that Rav2 and Rav1 can each bind independently to V 1 C. This might suggest two binding sites for subunit C in the RAVE complex, but in the absence of additional structural  RAVE Shows Glucose-dependent Interactions with the Vacuolar Membrane-RAVE functions in promoting glucose-dependent assembly of V 1 and V o sectors of the V-ATPase. In the absence of glucose, the RAVE complex can be isolated from soluble, cytosolic fractions in combination with V 1 subunits. However, Rav1 also binds to V o subunit Vph1, suggesting that it may localize to membranes under certain conditions. Previous efforts to localize RAVE have given conflicting information (9,37). We introduced a C-terminal GFP tag to both RAV1 and RAV2 and confirmed that the tagged strains were functional, because neither exhibited a rav1⌬ growth phenotype. We then assessed the localization of Rav1-GFP and Rav2-GFP constructs in the presence of glucose and during glucose deprivation and readdition. As shown in Fig. 6, Rav2-GFP and Rav1-GFP both distribute between the vacuolar membrane, the cytosol, and several cytosolic puncta in cells maintained in glucose. (Yeast vacuoles appear as indentations under differential interference contrast microscopy.) Notably, under these conditions, V 1 and V o sectors are predominantly assembled (6). We next asked whether localization would change when cells are deprived of glucose and disassembly of the V-ATPase occurs. Rav1-GFP and Rav2-GFP are largely released into the cytosol after 15 min of glucose deprivation. However, within 15 min after glucose readdition, both GFP-tagged RAVE subunits rebind to the vacuolar membrane. Previous results have established that RAVE is a stable complex both in the presence and absence of glucose (10,11), so the very similar movements of Rav1-GFP and Rav2-GFP in response to glucose deprivation were expected. Importantly, these results are entirely consistent with the role of RAVE in V 1 -V o assembly. V 1 sectors and subunit C lose tight association with Vph1-containing V o sectors at the vacuolar membrane upon glucose deprivation and reassemble with V o upon glucose readdition (6,9,17).
Glucose-dependent Interaction of RAVE with the Membrane Does Not Require the C Subunit-Given the key role of the V 1 C subunit in V 1 -V o assembly, we hypothesized that the V 1 C subunit might be required for membrane localization of the RAVE complex. In a vma5⌬ mutant that lacks V 1 C, V-ATPase activity in the vacuoles is lost (30). V 1 and the V o sectors can partially assemble, but the V 1 -V o interaction appears to be destabilized. In support of this, V 1 sectors are lost during vacuole isolation from vma5⌬ cells (30), but V 1 is distributed between the vacuole and cytosol as visualized in vma5⌬ cells by immunofluorescence microscopy (18). Fig. 7 demonstrates that Rav2-GFP showed a wild type distribution, including staining at the vacuolar membrane and cytosolic puncta in a vma5⌬ mutant maintained in glucose. We next asked whether Rav2-GFP localization would respond to glucose deprivation. As shown in Fig. 7, Rav2-GFP was released from the membrane by glucose deprivation and recruited back to the membrane by glucose readdition. Thus V 1 C is not required for glucose-dependent recruitment and release of RAVE from the membrane.

RAVE/V-ATPase Binding Sites Suggest a Role for RAVE in
Orienting V 1 Peripheral Stalks, V 1 C and V o Vph1NT in V-ATPase Assembly-Consistent with RAVE functioning as a V-ATPase assembly factor, RAVE binds to all three compo-FIGURE 5. V 1 subunit C binds to both Rav1 and Rav2. A, the V 1 C subunit was expressed as an MBP fusion protein (MBP-V 1 C) and purified by amylose chromatography as described under "Experimental Procedures," and then the MBP tag was removed with Prescission protease (MBP-V 1 C cleaved). GST-Rav1(840 -1125) was expressed and purified on glutathione-Sepharose as described (13). The GST-Rav1(840 -1125) fragment was eluted directly from one column, and V 1 C was added to a second column, incubated with the GST-bound protein, and then eluted with glutathione. Top gel, Coomassie-stained gel; bottom gel, immunoblot with anti-C subunit antibody. Nonadjacent lanes from the same gels are shown; the vertical line represents the deleted lanes. B, V 1 C input was prepared as in A and applied directly either to an anti-FLAG M2 column followed by washing and elution with FLAG peptide (no Rav2) or to an anti-FLAG column to which Rav2-FLAG had been bound as in Fig. 3, followed by washing and elution (FLAG eluant ϩRav2). The top panel is an anti-FLAG immunoblot, and the bottom panel is an anti-V 1 C subunit immunoblot. Nonadjacent lanes from the same gels are shown in both panels; the vertical line represents the deleted lanes. C, A FLAG-tagged version of the subunit C gene (VMA5) was integrated at the VMA5 locus in wild type yeast cells and a rav1⌬ mutant. Both strains also contained an integrated myc 13 -tagged Rav2. Immunoblots of Rav2-myc 13 NOVEMBER 13, 2015 • VOLUME 290 • NUMBER 46 nents of the V-ATPase that are dissociated during glucose deprivation: the V 1 subcomplex via the E and G subunits (10), the V 1 C subunit (18), and V o subunit Vph1 (13). The overall affinity of the V 1 sector for the V o sector appears to be derived from distinct interactions of three V 1 EG-containing peripheral stalks at the V 1 -V o interface (14). We propose that RAVE is important in establishing key interactions with these stalks at the V 1 -V o interface. The V 1 C subunit interacts in a defined orientation with two of the three EG peripheral stalks at the V 1 -V o interface of the assembled complex (38). V 1 C is thus positioned to play a critical role in reversible disassembly, and consistent with this role, post-translational modifications of V 1 C, as well as cytoskeletal interactions, have been implicated in reversible disassembly in other systems (17,39). The third EG peripheral stalk interacts with the V 1 H subunit and maintains this interaction in both the assembled and disassembled complex (4,31,38). We have found no evidence of RAVE interactions with V 1 H (10), suggesting that this stalk is not directly affected by RAVE intervention in assembly.

Interactions and Localization of the Yeast RAVE Complex
We propose that the RAVE complex acts as an escort to bring the V 1 subcomplex, V 1 subunit C, and V o together. Rav1 is the largest subunit and the central component of the RAVE complex. Rav2 and Skp1 bind to Rav1 but not to each other (9). Brace et al. (37) identified a skp1 mutation that separated the essential Skp1 function in ubiquitin ligases from its role in RAVE and used this mutation to show that Skp1 is not required for V-ATPase assembly but may assist in release of RAVE from the membrane. This implicates Rav1 and Rav2 in promotion of V-ATPase assembly. Rav1 is responsible for the interaction between RAVE and V 1 (10) and also appears to contain the V o interaction site (13). As shown in Fig. 5 and described below, subunit C binds to both Rav1 and Rav2. Rav1 is thus positioned to bind to both V 1 and V o during assembly and to act with Rav2 to orient subunit C at the V 1 -V o interface.
To go beyond secondary structure prediction and Rav1 and/or rabconnectin sequence comparisons for interpreting the binding interactions we mapped across Rav1, we asked whether larger conserved structural elements could be identified in Rav1 using a structural bioinformatics approach. Submission of the entire RAV1 sequence to the Phyre 2 server (41) gave a strong predicted model (100% estimated confidence that a remote homology has been identified and modeled well) for amino acids 55-669 of Rav1, based on alignment with Caenorhabditis elegans Aip1 (model c1nr0a). Although the model is based on remote homology, this level of confidence strongly suggests that the homology is real, and the overall fold is likely to be valid (41). The model consists of a double sevenbladed ␤-propeller structure (Fig. 8A). Such ␤-propeller structures are comprised of a series of WD or WD-like repeats, each containing four antiparallel ␤-sheets. Motif-based predictions had identified up to eight rather weakly conserved WD and WD-like repeats in Rav1 (25), but none had predicted 14 WD repeats. Interestingly, motif predictions had identified 7-10 WD repeats in Aip1 prior to structure determination (42), indicating that these motif predictions can be incomplete. In fact, not only the highest scoring structure for Rav1 but also the next five structures all predict double ␤-propellers covering a similar region of Rav1. Because the hits from submission of the entire RAV1 sequence were overwhelmingly directed toward the N-terminal ␤-sheet-rich region, we submitted the sequence for amino acids 760 -1270, which contains the ␣-helical region, separately. The highest confidence model (97.5% confidence) aligned a small portion of this sequence, amino acids 967-1081 of RAV1, with a helical region of the ␤Ј subunit of the COP I coatomer (model c3mkqA). This model is also shown in Fig. 8A. Interestingly, this region of the ␤Ј subunit of COP I forms an ␣ solenoid structure and is adjacent to a double ␤-propeller structure at the N terminus of this protein (43).
In Fig. 8A, we map binding sites for Rav1 binding partners to regions within the N-and C-terminal halves of Rav1. The Rav1 FIGURE 6. RAVE shows glucose-dependent interactions with the vacuolar membrane. Log phase cells from strains containing an integrated C-terminally tagged Rav1-GFP or Rav2-GFP were visualized directly by fluorescence microscopy as described under "Experimental Procedures." Cells were first visualized in growth media containing glucose. Cells were then visualized after pelleting, washing, and resuspending in media lacking glucose. Glucose-deprived cells were again pelleted and resuspended in media containing glucose. The images shown are representative of those seen in three independent experiments. FIGURE 7. Glucose-dependent interaction of RAVE with the vacuolar membrane does not require the V 1 C subunit. A vma5⌬ mutation was introduced into the Rav2-GFP strain, and the cells were treated and visualized as in Fig. 6. The images shown are representative of those seen in three independent experiments. DIC, differential interference contrast.
C-terminal ␣-helical region contains the most conserved sequence among Rav1 homologues (amino acids 840 -1125) (25). Immunoprecipitations from C-terminal deletions of RAV1 suggest a requirement for amino acids 840 -940 for V 1 binding and for amino acids 1126 -1271 for Skp1 binding (Fig.  1). Interestingly, two Rav1 mutants, ⌬940-stop and ⌬1126-stop were able to bind V 1 at wild type levels but had compromised binding to Skp1; these mutations were not able to rescue the rav1⌬ phenotype. This result suggests that both V 1 and Skp1 binding in the C-terminal half of Rav1 are necessary for RAVE activity and that Skp1 plays an essential role in RAVE function despite evidence that it is not essential for assembly (37). Twohybrid data (Fig. 2) support the interaction of the highly conserved Rav1(840 -1125) region with V 1 subunit E and are consistent with the requirement for amino acids 840 -890 suggested by immunoprecipitations in Fig. 1.
The N-terminal half of Rav1 supports the interaction with Rav2 (Figs. 2 and 3). The expressed Rav1(2-729) and Rav2 proteins proved to interact in vitro (Fig. 3). The second most conserved region of RAV1 is the first 240 amino acids of the N-terminal domain. Further two-hybrid experiments using deletions of Rav1 in this region suggest that Rav2 requires the first 98 amino acids of Rav1 for binding. This Rav1(1-240) conserved region also interacted with V 1 C and Vph1NT in a two-hybrid assay (Fig. 2). We have not yet confirmed these interactions in vitro, so it is possible that these two-hybrid interactions (particularly those between V 1 C and Rav1(1-240)) could be indirect, possibly caused by bridging by Rav2.
Previous two-hybrid data indicated an interaction of subunit C with both Rav1 and Rav2 (18). In Fig. 5, we present the first evidence for direct binding of subunit C to both Rav1(840 -1125) and Rav2. The possibility of two distinct binding sites is important because it suggests a role for RAVE in orienting subunit C, which must establish interactions to two different peripheral stalks, during assembly. As described above, the V 1 C subunit is localized to the interface of V 1 and V o and likely to be a critical player in V-ATPase assembly. In EM reconstructions and cryo-EM structures, subunit C binds to two of the three peripheral EG peripheral stalks (38, 44 -46). The crystal structure of subunit C shows three domains, a globular "head" (C head ), an elongated "neck," and another globular "foot" (C foot ) (47). Subsequent work with purified subunits showed that C head binds to EG with high affinity in vitro to form a stable EGC head complex (15,48). The lower affinity of C foot for the EG heterodimer precludes the formation of a complex in vitro, but both C foot and EG bind Vph1NT with low affinity, and it has been suggested that a high avidity quaternary complex forms, including C foot , EG, and Vph1NT (14,15). Data showing crosslinking between C foot and the N-terminal region of Vph1 support this possibility (49).
We had previously demonstrated that a purified Rav1(840 -1125) fragment could bind to Vph1NT (13). We provide evidence here that Rav1(679 -840), which lies just upstream of the most conserved region and contains the junction between the N-and C-terminal halves of Rav1, interacts with Vph1NT in two-hybrid assay. This interaction is further supported by in vitro binding of purified Vph1NT to expressed Rav1(679 -898)-His 6 (Fig. 4). Taken together, the arrangement of interaction sites for Vph1NT, V 1 (through the EG peripheral stalks) and V 1 subunit C on RAVE supports the proposition that RAVE could serve as a template for V 1 and V o assembly by helping to bring the two V-ATPase subdomains into proximity and properly orienting subunit C to promote the assembly process. A possible model for RAVE intervention in assembly is shown in Fig. 8B.
RAVE Is Recruited to the Vacuolar Membrane in a Glucosedependent Manner-Although the requirement for RAVE in V-ATPase reassembly had been known for some time, we present here the first evidence that the RAVE complex itself is reversibly recruited to the vacuolar membrane in the presence of glucose, conditions that promote V-ATPase reassembly. In Fig. 6, we show that RAVE is localized primarily to the vacuolar membrane with some staining in the cytosol and a few smaller puncta when cells are maintained in glucose. When cells are subsequently deprived of glucose, RAVE is released from the vacuolar membrane into the cytosol but is able to rebind to the membrane upon glucose readdition. This is entirely consistent with evidence for the assembly of V 1 and V 0 at the vacuolar membrane in the presence of glucose and the disassembly and release of V 1 and subunit C into the cytosol in the absence of glucose (6,7,40). It is more difficult to reconcile with recent data suggesting that GFP-tagged V 1 subunits other than subunit C remain in proximity to the membrane when cells are deprived of glucose (17). We previously found increased coimmunoprecipitation of RAVE subunits with V 1 in cytosolic fractions from glucose-deprived cells (10). This result supports release of V 1 sectors from the vacuolar membrane with glucose deprivation and implicates V 1 binding to RAVE as a mechanism for sequestering V 1 from V o and/or priming V 1 for reassembly upon glucose readdition (10). However, if V 1 sectors lacking subunit C remain near the membrane during glucose deprivation and RAVE subunits are released almost completely from the vacuolar membrane, it is hard to envision how RAVE could act on V 1 to promote assembly. V 1 subunit C can bind to RAVE independently of V 1 (18), so it is possible that RAVE-subunit C complexes are released to the cytosol upon glucose deprivation and that RAVE sequesters subunit C and/or primes it for rebinding upon glucose readdition. These questions require further investigation.
Interestingly, RAVE is able to localize to the vacuolar membrane in the absence of the V 1 C subunit, and its localization responds to glucose deprivation and readdition similarly to wild type cells (Fig. 7). This is consistent with previous results indicating partial assembly of inactive and unstable V 1 -V o complexes at the vacuolar membrane in the absence of subunit C (18,30). However, these results also have implications for the nature of glucose signaling to the V-ATPase and the RAVE complex. We previously showed that under conditions where V 1 was not able to assemble with V o , the cytosolic V 1 -RAVE interaction did not change with glucose deprivation, suggesting that this interaction is not intrinsically glucose-sensitive (10). As described above, work in other systems had implicated subunit C in reversible disassembly and suggested that it might be a direct recipient of glucose signals (39). However, glucoseresponsive recruitment of RAVE to the vacuolar membrane in the absence of subunit C indicates that there must be glucosedependent signals that do not depend on V 1 C. These signals could be focused on the RAVE-V o interaction, but further experiments will be necessary to identify the nature and location of this signal.