Cysteine-mediated Cross-linking Indicates That Subunit C of the V-ATPase Is in Close Proximity to Subunits E and G of the V1 Domain and Subunit a of the V0 Domain*

The vacuolar (H+)-ATPases (V-ATPases) are multisubunit complexes responsible for ATP-dependent proton transport across both intracellular and plasma membranes. The V-ATPases are composed of a peripheral domain (V1) that hydrolyzes ATP and an integral domain (V0) that conducts protons. Dissociation of V1 and V0 is an important mechanism of controlling V-ATPase activity in vivo. The crystal structure of subunit C of the V-ATPase reveals two globular domains connected by a flexible linker (Drory, O., Frolow, F., and Nelson, N. (2004) EMBO Rep. 5, 1-5). Subunit C is unique in being released from both V1 and V0 upon in vivo dissociation. To localize subunit C within the V-ATPase complex, unique cysteine residues were introduced into 25 structurally defined sites within the yeast C subunit and used as sites of attachment of the photoactivated sulfhydryl reagent 4-(N-maleimido)benzophenone (MBP). Analysis of photocross-linked products by Western blot reveals that subunit E (part of V1) is in close proximity to both the head domain (residues 166-263) and foot domain (residues 1-151 and 287-392) of subunit C. By contrast, subunit G (also part of V1) shows cross-linking to only the head domain whereas subunit a (part of V0) shows cross-linking to only the foot domain. The localization of subunit C to the interface of the V1 and V0 domains is consistent with a role for this subunit in controlling assembly of the V-ATPase complex.

The V-ATPases are composed of a peripheral V 1 domain responsible for ATP hydrolysis and an integral V 0 domain that carries out proton translocation (1,2). V 1 is a 600 -650-kDa complex with the subunit composition A 3 B 3 C 1 D 1 E 1 F 1 G 2 H 1-2 (1,2). The nucleotide binding sites are located at the interface of the A and B subunits (which form an alternating hexamer), with the catalytic sites located mainly on the A subunit and "non-catalytic" sites located mainly on the B subunit (12)(13)(14). The V 0 domain is a 250-kDa complex with the subunit composition a 1 d 1 e x c 4 cЈ 1 cЉ 1 (1,15). The three proteolipid subunits (c, cЈ, and cЉ) each contain a buried glutamic acid residue that is essential for proton transport (16,17) and together they form a hexameric ring. The V-ATPases thus resemble the F-ATPases (or ATP synthases), which are involved in ATP synthesis (18 -20). High resolution crystal structures have been obtained for both F 1 (21,22) and F 1 bound to a proteolipid ring (23).
The F-ATPases operate by a rotary mechanism (20). ATP hydrolysis in F 1 causes rotation of a central stalk composed of the ␥ and ⑀ subunits (23)(24)(25)(26), which in turn causes rotation of the ring of proteolipid c subunits (27)(28)(29). A peripheral stalk (composed of the ␦ subunit and the soluble regions of subunit b, Ref. 30) functions as a stator, holding subunit a fixed relative to the ␣ 3 ␤ 3 head during ATP-driven rotation of the proteolipid ring. It is rotation of the ring of proteolipid c subunits relative to subunit a that is believed to drive proton transport (19,31,32). Comparison of amino acid sequences of the V-and F-ATPases reveals clear sequence homology both between the nucleotide binding subunits and between the proteolipid subunits of the two classes (33,34,16). Because there is virtually no sequence similarity for the remaining subunits, it has not been possible to deduce the function or localization of the remaining V-ATPase subunits from their primary sequence.
Electron microscopy studies indicate that like the F-ATPases (35), the V-ATPases also contain multiple stalks connecting the peripheral and integral domains (36 -38). To determine the composition of the central and peripheral stalks of the V-ATPase, we have previously employed cysteine mutagenesis and photochemical cross-linking using MBP to determine the arrangement of subunits relative to subunit B (39,40). These studies suggest that subunit D is located in the central stalk whereas subunits E and G form part of the peripheral stalk connecting V 1 and V 0 . These assignments have been confirmed by rotation experiments demonstrating that subunits D and F are present in the central rotor (41) whereas subunit G is located in the peripheral stator (42).
In the present study, we have used a similar photochemical cross-linking approach to localize subunit C within the V-ATPase complex. This subunit is of particular interest because of its putative role in regulating dissociation of the V-ATPase complex in vivo (43). Dissociation of V 1 and V 0 represents an important mechanism of controlling V-ATPase activity in cells (44,45) and occurs with release of subunit C from both the V 1 and V 0 domains. Subunit C has therefore been suggested to function in triggering dissociation of the V-ATPase in vivo (43). The availability of a recently published high resolution crystal structure of subunit C (46) has allowed us to introduce unique cysteine residues into structurally defined sites within this protein. The results have identified specific contacts between subunit C and other subunits present in the peripheral stalk of the V-ATPase and are consistent with a role for subunit C in controlling the stability of the interactions between the V 1 and V 0 domains.

EXPERIMENTAL PROCEDURES
Materials and Strains-Zymolyase 100T was purchased from Seikagaku America, Inc. Concanamycin A was obtained from Fluka Chemical Corp. 9-Amino-6-chloro-2-methoxyacridine (ACMA) was purchased from Molecular Probes, Inc. SDS, nitrocellulose membranes (0.45-m pore size), Tween 20, horseradish peroxidase-conjugated goat antimouse IgG, and horseradish peroxidase-conjugated goat anti-rabbit IgG were from Bio-Rad. 4-(N-maleimido)benzophenone (MBP) and most common chemicals were obtained from Sigma. The yeast strain lacking subunit C (Vma5p) was constructed from YPH500 by replacing the VMA5 gene with the TRP1 gene (47). The strain was first selected on tryptophan-minus plates and the growth phenotype was then assessed on YEPD plates buffered with 50 mM KH 2 PO 4 or 50 mM succinic acid to either pH 7.5 or pH 5.5.
Antibodies-The monoclonal antibodies 3F10 (against the HA antigen) with and without conjugation to horseradish peroxidase were purchased from Roche Applied Sciences. The monoclonal antibody 10D7 against the 100 kDa a subunit was from Molecular Probes. The polyclonal antibody against subunit E (Vma4p) was a gift from Dr. Daniel Klionsky (University of Michigan). The polyclonal antibody against subunit G (Vma10p) was a gift from Dr. Tom Stevens (University of Oregon).
Cloning of the VMA5 Gene Encoding Subunit C-The VMA5 gene was amplified from genomic DNA isolated from the yeast strain YPH500 using the following primers with unique restriction enzyme sites, SacI and KpnI (underlined): forward, 5Ј-GAGCTCGATATCATG-TGCTTTCAAGCAGGATTTGTCACAGTG-3Ј; reverse, 5Ј-GGTACCAC-TAGTGCTAAATCCTGTAGATTTCCGTGCCAATTT-3Ј. The resulting PCR product was then cloned into the TA cloning vector pCR2.1 (Invitrogen). This construct was then digested with SacI and KpnI, and the VMA5 gene ligated into pRS316. An HA epitope tag was introduced at the N terminus of Vma5p using a PCR-based recombination method (48).
Construction of Mutants-Site-directed mutants of Vma5p were constructed using the Altered Sites II in vitro mutagenesis system (Promega) following the manufacturer's protocol. The VMA5 gene containing the HA tag was cloned into the pALTER-1 vector using the SacI and KpnI restriction sites. The Cys-less form of VMA5 was made by substituting the single endogenous cysteine residue at position 340 with serine using the oligonucleotide: 5Ј-CTCTCTAAATCCAAGTCTGAA-3Ј.
Single cysteine mutants were then constructed starting with the Cys-less form of the VMA5 gene (C340S) in pALTER-1. The oligonucleotides for the single cysteine mutants constructed are as follows: After the mutagenesis reaction, the SacI/KpnI fragment was cleaved and purified by agarose gel electrophoresis. The fragment was used to generate mutant forms of VMA5 in pRS316 encoding single cysteinecontaining subunit C proteins. The sequence of all mutants was confirmed by DNA sequencing. No other mutations were detected in the final products.
Transformation and Selection-Yeast cells lacking functional endogenous Vma5p were transformed using the lithium acetate method (49). The transformants were selected on uracil-minus plates, and the growth phenotypes were assessed on YEPD plates buffered with 50 mM KH 2 PO 4 or 50 mM succinic acid to either pH 7.5 or pH 5.5.
Isolation of Vacuolar Membrane Vesicles-Vacuolar membrane vesicles were isolated using a modification of the protocol described by Uchida et al. (50). Yeast cells were grown overnight at 30°C to 5 ϫ 10 7 cells/ml in 1 liter of selective medium. Cells were pelleted, washed twice with water, and resuspended in 50 ml of 100 mM Tris-HCl, pH 9.4, containing 10 mM dithiothreitol. After incubation at 30°C for 20 min, cells were pelleted again, washed once 25 ml of YEPD medium containing 0.7 M sorbitol, 2 mM dithiothreitol, and 100 mM MES-Tris (pH 7.5), resuspended in 25 ml of YEPD medium containing 0.7 M sorbitol, 2 mM dithiothreitol, 100 mM MES-Tris, pH 7.5, and 2 mg of Zymolase 100T, and incubated at 30°C for 90 min with gentle shaking. The resulting spheroplasts were osmotically lysed, and the vacuoles were isolated by flotation on two consecutive Ficoll gradients and diluted in transport buffer (15 mM MES-Tris, pH 7.0, 4.8% glycerol).
Photochemical Cross-linking of the V-ATPase with MBP-Vacuolar membranes (200 g of protein) were washed three times with phosphate-buffered saline (pH 7.2) containing 2 mM EDTA, 2 g/ml aprotinin, 0.7 g/ml pepstatin, 5 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride, and resuspended in 100 l of the same buffer. 1 l of 100 mM MBP dissolved in dimethylformamide or 1 l of dimethylformamide (negative control) were added to the washed vacuoles. After incubation at 23°C for 30 min in the dark, the unreacted MBP was quenched by addition of 10 mM dithiothreitol. The vacuole membranes were pelleted, washed twice with phosphate-buffered saline, pH 7.2, containing 2 mM EDTA, 2 g/ml aprotinin, 0.7 g/ml pepstatin, 5 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride, resuspended in 100 l of the same buffer, and irradiated with a long wavelength (366 nm) ultraviolet lamp at 4°C for 5 min. The vacuolar membranes were then solubilized with 2% C 12 E 9 and the V-ATPase complex immunoprecipitated using the anti-HA antibody (3F10), and protein G-agarose.
Analysis of Cross-linked Products-Immunoprecipitated V-ATPase complexes were separated by SDS-PAGE using 7.5% acrylamide gels (51), and the proteins were electrophoretically transferred to nitrocellulose membranes. Western blotting was then performed using the horseradish peroxidase-conjugated monoclonal antibody 3F10 against the HA epitope tag attached to subunit C and antibodies specific for other V-ATPase subunits, as previously described (39,40).
Other Procedures-Protein concentrations were determined by the Lowry method (52). ATPase activity was measured using a coupled spectrophotometric assay in the presence or absence of 1 M concanamycin A as described previously (14). ATP-dependent proton transport was measured using the fluorescence probe ACMA in the presence or absence of 1 M concanamycin A as described previously (14).

Construction of Single Cysteine-containing Mutants of
Vma5p and the Effect of Mutations on Growth Phenotype, Assembly, ATPase Activity, and Proton Transport Activity-To determine the arrangement of subunits in the V-ATPase complex relative to subunit C, we have employed a strategy involving cysteine-mediated cross-linking using the photoactivated cross-linking reagent MBP (39,40). Subunit C was chosen for this study in part because of the recently published high resolution crystal structure of this protein (46). This structure facilitated the introduction of unique cysteine residues into structurally defined sites on the surface of subunit C (Fig. 1).
A vma5⌬ strain was constructed from YPH500 by replacing the VMA5 gene with the TRP1 gene. The wild-type VMA5 gene was cloned from YPH500 genomic DNA and inserted into the pRS316 vector, which contains the URA3 selectable marker. An HA epitope tag was introduced at the N terminus of Vma5p to facilitate immunoprecipitation of the V-ATPase complex and to permit detection of subunit C and its cross-linked products by Western blot. Transformation of the vma5⌬ strain with this epitope-tagged construct gave a wild-type growth phenotype (see below) and gave rise to a V-ATPase complex possessing wild-type assembly and activity properties (data not shown). Thus wild type in this study refers to cells transformed with this epitope-tagged construct of Vma5p.
To construct mutants of Vma5p containing single cysteine residues, it was first necessary to construct a Cys-less form of Vma5p in which the single endogenous cysteine residue at position 340 was replaced with serine. This was accomplished using site-directed mutagenesis (see "Experimental Procedures"). Using the Cys-less form of Vma5p as a starting point, 25 mutants containing single cysteine replacements were then constructed using the same mutagenesis protocol. Cysteine residues were introduced at exposed sites on each of the surfaces of the protein, as indicated from the x-ray structure of subunit C (Fig. 1). All mutants were selected on uracil-minus plates and tested for growth on YEPD plates buffered to either pH 7.5 or pH 5.5. Both the Cys-less Vma5p and the single cysteine-containing mutants were able to grow at both pH 7.5 and 5.5, indicating that they were able to complement the vma Ϫ phenotype of the vma5⌬ strain. Based upon previous studies (54,13), this suggests that the mutants possessed at least 20% of wild-type levels of activity. To assess the assembly competence of the Vma5p mutants, vacuolar membranes were isolated from each strain, and Western blotting was performed using an anti-HA antibody. It has previously been observed that loss of any V-ATPase subunit (with the exception of Vma13p) results in failure of V 1 subunits to associate with the vacuolar membrane (55). As shown in Fig. 2A, all mutants showed near wild-type levels of Vma5p present in isolated vacuolar membranes. As a test of the functional properties of the V-ATPase complexes containing the mutant forms of Vma5p, concanamycin-sensitive ATPase activity and ATP-dependent proton transport (as measured by fluorescence quenching of ACMA) were determined for vacuolar membranes isolated from each of the yeast strains. As shown in Fig. 2B, vacuolar membranes from all of the mutant strains displayed concanamycin-sensitive ATPase activity and proton transport that was at least 70% of that measured for wild-type vacuolar membranes. These results indicate that the mutations introduced into Vma5p do not significantly perturb the assembly or activity properties of the resultant V-ATPase complexes.
Photochemical Cross-linking of V-ATPase Complexes Containing Mutant Forms of Vma5p Using MBP-To determine the proximity of other V-ATPase subunits to each of the unique cysteine residues introduced into subunit C, photochemical cross-linking was carried out using the photoreactive sulfhydryl reagent MBP (39,40). This reagent reacts with sulfhydryl groups via the maleimide moiety and, upon irradiation with UV light, generates a reactive intermediate capable of insertion into proximal residues. As a result, cross-linking between the initial sulfhydryl-containing polypeptide and the protein that reacts with the photoactivated group occurs. The linker arm within MBP requires that the cross-linked proteins be separated by no more than ϳ10 Å (56).
Because V-ATPase complexes contain many subunits possessing endogenous cysteine residues, it is important to determine whether MBP reaction at these sites outside of subunit C might result in photoactivated cross-linking between subunit C and these other subunits in the V-ATPase complex.
To determine whether this is the case, vacuolar membranes were isolated from the yeast strain expressing the Cys-less form of Vma5p. These membranes were then reacted with 1 mM MBP for 30 min in the dark followed by quenching of the reagent with dithiothreitol, washing by sedimentation and irradiation with a long wavelength UV lamp. The membranes were solubilized using C 12 E 9 and the V-ATPase complexes immunoprecipitated using an anti-HA antibody. The proteins were separated by SDS-PAGE and Western blotting was performed using the anti-HA antibody. As shown in Fig. 3A, no HA-reactive bands of molecular weight greater than that of the C subunit monomer were observed in the presence of MBP. These results indicate that any cross-linked products obtained with the single cysteine-containing mutants of Vma5p do not arise as a result of MBP reaction with sulfhydryls present in other subunits.
Vacuolar membranes were next isolated from the yeast strains expressing each of the single cysteine-containing mutants of subunit C and photocross-linking by MBP was evaluated as described above. Of the twenty five mutants tested, nine showed the generation of higher molecular weight crosslinked products when Western blotting was performed using the anti-HA antibody, indicating cross-linking of subunit C to other proteins in the complex. The remaining sixteen mutants showed no evidence of cross-linking, although this could be attributed to several possible causes. First, a given cysteine residue, although exposed on the surface of subunit C, may nevertheless be prevented from reaction with MBP by being located at the interface of two subunits. Second, although a cysteine may react with MBP, the photoreactive group may not be sufficiently close or have the correct orientation to give rise to a cross-linked product. Thus, only a positive cross-linking result is interpretable.
Of the nine mutants showing evidence of cross-linking, two gave HA-reactive species of molecular mass ϳ70 kDa, including G27C and H324C. Because this mass corresponds to the sum of the molecular masses of subunit C (43 kDa) and subunit E (27 kDa), Western blotting was performed using an antibody specific for subunit E. As shown in Fig. 3A, the 70-kDa bands recognized by the anti-HA antibody were also recognized by the anti-E subunit antibody for both G27C and H324C, indicating that MBP-mediated formation of an E/C heterodimer can occur at both of these positions. Importantly, probing the blot of the Cys-less mutant using the anti-E subunit antibody revealed no such 70 kDa species (Fig. 3A). When blots of the other single cysteine-containing mutants of subunit C were probed with the anti-E subunit antibody, two others showed cross-linked products that were recognized by both the anti-HA and anti-E subunit antibodies: N216C and A220C (Fig. 3B). The apparent molecular mass of the cross-linked products formed for these mutants (ϳ90 kDa) was higher than that expected for an E/C heterodimer. We have previously observed such aberrant migration of cross-linked products upon MBP-mediated photocross-linking to subunit B (40). This aberrant migration may be caused by the formation of cross-linked products in which the polypeptides are joined near the middle of one or both sequences, giving rise to a non-linear species. Because subunit E does not contain endogenous cysteine residues, this aberrant migration cannot be because of the presence of other subunits that have become cross-linked as a result of MBP modification of subunit E in the initial reaction.
Of the four cysteine mutants showing cross-linking to subunit E, the A220C mutant showed a second cross-linked product recognized by the anti-HA antibody of molecular mass ϳ60 kDa (Fig. 3B). Because this is close to the sum of the masses of subunits C and G (43 kDa ϩ 13 kDa ϭ 56 kDa), the blots were probed with an antibody against subunit G. As can be seen in Fig. 4, this species was recognized by both the anti-HA and anti-G subunit antibodies, suggesting that it corresponds to a C/G heterodimer.
Finally, for two subunit C mutants (S57C, S69C), a crosslinked product of ϳ180 kDa was observed (Fig. 5). This product was recognized by both anti-HA and anti-a subunit antibodies, although the molecular mass was considerably larger than that FIG. 2. Effects of Vma5p mutations on assembly and activity of the V-ATPase complex. A, V-ATPase assembly was assessed by measurement of the levels of Vma5p associated with vacuolar membranes. Vacuolar membranes (10 g of protein) were prepared from the vma5⌬ strain expressing the wild-type VMA5 gene (WT), the pRS316 vector alone (Vector), the Cys-less mutant of VMA5 (C340S-Cys-less), or the double mutants of VMA5 containing the C340S mutation and the indicated mutations. All mutants and wild type contained an N-terminal HA epitope tag and were expressed using pRS316. The proteins were separated by SDS-PAGE on a 7.5% acrylamide gel and transferred to nitrocellulose membranes. Western blotting was then performed using the antibody 3F10 against the HA epitope tag as described under "Experimental Procedures." B, effects of Vma5p mutations on ATPase activity and proton transport by the V-ATPase complex present in isolated vacuolar membranes. Vacuolar membranes were prepared from the same strains described above and assayed for concanamycin A-sensitive ATPase activity (3 g of protein, solid bars) and concanamycin A-sensitive, ATP-dependent proton transport (3 g of protein, open bars) as described under "Experimental Procedures." Activities are expressed relative to vacuolar membranes isolated from the vma5⌬ strain expressing the HA-tagged wild-type Vma5p in pRS316 (defined as 100%), which had a specific activity of 0.41 M ATP/min/mg protein. Proton transport was estimated from the initial rate of ATP-dependent fluorescence quenching using the dye ACMA as described.
predicted for a C/a heterodimer (43 kDa ϩ 100 kDa ϭ 143 kDa). This aberrant migration may be due either to the unusual migration behavior of the heterodimer or to the presence of other subunits that have been cross-linked as a result of MBP reaction with cysteine residues in subunit a. This latter explanation is made unlikely by the absence of higher molecular weight species recognized by the anti-a subunit antibody in the strain expressing the Cys-less mutant of subunit C (Fig. 5). DISCUSSION Subunit C is unique among the V-ATPase subunits in being released from both the V 1 and V 0 domains upon dissociation The samples were then irradiated with a long wavelength ultraviolet lamp (366 nm) for 5 min at 4°C, solubilized with 2% C 12 E 9 , and immunoprecipitated with an anti-HA antibody. The immunopurified proteins were separated by SDS-PAGE using a 7.5% acrylamide gel and transferred to nitrocellulose membranes. Western blotting was then performed using an antibody against the HA epitope tag or an antibody against subunit E (Vma4p). The position of the monomeric forms of subunits C and E as well as the heavy chain of the IgG used for immunoprecipitation (HC) are indicated with solid arrowheads, whereas the migration positions of molecular mass markers are indicated by the numbers shown to the left of each panel. B, the experiment described in A was repeated using vacuolar membranes isolated from the vma5⌬ strain expressing the double mutants C340S/N216C (upper panel) or C340S/A220C (lower panel). of the complex in vivo (44). For reversible dissociation to be used as a mechanism of regulating V-ATPase activity in vivo, it is necessary that ATPase activity of the free V 1 domain and passive proton conduction by free V 0 be silenced following dissociation, and these predictions have been confirmed experimentally (57,58). Interestingly, overproduction of subunit C in yeast has a more lethal phenotype than its disruption, leading to the suggestion that binding of subunit C to free V 1 may activate a futile ATPase activity (59). Thus, release of subunit C may function both in controlling in vivo dissociation and in preventing potentially lethal ATP hydrolysis by the released V 1 domain. Subunit C has also been identified as one of two potential binding sites for actin within the V-ATPase complex (60,61) and has been shown to display actin bundling activity (62).
The recently published high resolution crystal structure of subunit C (46) reveals a protein possessing two globular domains, termed the "head" and "foot" domains, connected by pair of ␣-helical segments. The larger foot domain contains both the N terminus (residues 1-151) and the C terminus (residues 287-392) of the protein, whereas the head domain is composed of residues 166 -263 located in the central region of the primary sequence. To localize subunit C within the V-ATPase complex, we have made use of this high resolution structure to select specific surface residues for mutagenesis to cysteine. These cysteine residues were then used as sites of attachment of the photo-activated maleimide reagent MBP. Photocross-linking followed by analysis of cross-linked products by Western blot thus allows us to place other V-ATPase subunits with respect to specific sites on the surface of the C subunit. A similar approach using photoactivated cross-linking to subunit B has allowed the localization of subunits E and G to the peripheral stalk and subunit D to the central stalk connecting V 1 and V 0 (39,40). In addition, localization of subunit H to the interface of V 1 and V 0 using this method has been confirmed by electron microscopy (38).
The results in the present study indicate that subunit C binds to the V-ATPase complex with the foot domain oriented toward the membrane, interacting with the a subunit of V 0 . Although the site within the a subunit that is in proximity to subunit C has not been determined, the cytoplasmic localization of the N-terminal hydrophilic domain of subunit a (63) makes this region a likely candidate. By contrast, the head domain of subunit C is in close proximity to subunit G. Subunit G is a 13-kDa protein present in two copies per complex (64) that appears to form a dimer (65), analogous to the coiled-coil structure observed for the b subunit in the peripheral stalk of F 1 F 0 (66). In fact subunit G and subunit b show partial sequence homology along one helical face (67) and both are tolerant to short deletions (68,69). Previous cross-linking results indicate that at least a portion of subunit G is present near the top of the V 1 complex farthest from the membrane (40), but this subunit may extend some length along the outer surface of the V 1 domain.
Such an extended disposition perpendicular to the mem- brane is clearly indicated for subunit E. Thus not only does subunit E cross-link to residues in both the head and foot domains of subunit C (a distance of nearly 100 Å), but also to residues near the top and bottom of subunit B (a distance of over 70 Å) (39,40). An interaction between subunits C and E is consistent with previous cross-linking results (64) as well as the formation of a C/E heterodimer on dissociation of the V 1 domain of the bovine-coated vesicle V-ATPase (70). Similarly, interaction between subunits E and G has been suggested from both cross-linking (64) and the formation of an E/G heterodimer upon disruption of normal V 1 assembly in yeast (71,72).
Recently, the N-terminal region of subunit E has been shown to play an important role in interaction with both subunits C and G (73). Thus, deletion of the first 19 amino acids of subunit E significantly perturbs binding to subunit C whereas deletion of the next 19 amino acids causes loss of binding to subunit G. Point mutations in subunit E that perturb binding to subunit C do not prevent subunit C from assembling with the remainder of the V-ATPase complex, leading the authors to suggest that subunit C makes contact with V-ATPase subunits in addition to subunit E (73). Results from the current study confirm this prediction and identify subunits G and a as additional contacts for subunit C. Thus, subunit C interacts with three different subunits (E, G, and a) of the peripheral stalk (Fig. 6).
Interactions between subunit C and subunits of the peripheral stalk may explain the postulated stimulatory role of subunit C on ATPase activity of the free V 1 domain (59). ATP hydrolysis by free V 1 may be suppressed by interaction between the central and peripheral stalks, which would prevent ATP-dependent rotation of the central stalk. In fact, crosslinking between subunits of the central and peripheral stalk in dissociated V 1 has been observed (64). By binding to multiple sites on subunit E, subunit C may prevent the peripheral stalk from folding inwards toward the central stalk, thereby preventing the normal inhibition of ATP hydrolysis by V 1 .
Drory et al. (46) report two distinct crystal structures for subunit C. The second structure (solved at 2.9 Å) differs from the first in a ϳ12 o movement of the head domain relative to the foot domain, indicating considerable flexibility in the alpha helical neck connecting these two domains. Assuming subunit C remains stably bound to subunit E via both the head and foot domains during the transition between these two conformations, this would suggest that subunit E may need a comparable level of structural flexibility. This is consistent with the predicted high ␣-helical content of subunit E (74) where it existed in an elongated conformation. If, however, there is more than one copy of subunit E (as has previously been suggested (53,75)), it is possible that the two different conformations of subunit C actually bind to two different copies of subunit E. Alternatively, the head domain may bind to one copy of E and the foot domain may bind the other. Definitive information on the number of copies of subunit E per complex should help to resolve this question. It should be noted that the two conformations of subunit C also make it possible that the head domain binds to subunit E in one conformation and to subunit G in the other conformation.
Mutagenesis studies of Vma5p have identified three mutations (F260A, Y262A, and F385A) that decrease the stability of the V-ATPase complex but nevertheless significantly increase the maximal velocity for both proton transport and ATP hydrolysis (43). Two of these mutations (F260A and Y262A) are in the head domain near the surface that shows cross-linking to subunits E and G (Asn 216 and Ala 220 ) whereas the third (Phe 385 ) is in close proximity to Ser 69 in the foot domain that shows cross-linking to subunit a. The increased catalytic activity despite decreased stability observed upon mutation of the three aromatic residues is not readily interpretable in terms of subunit C simply playing a structural role in the peripheral stalk. Curtis et al. (43) have suggested that during the catalytic cycle the V-ATPase adopts a conformation that is less stable, and that the C subunit controls the ability of the complex to adopt such a conformation. By interacting with peripheral stalk subunits in both V 1 (E and G) and V 0 (a), subunit C is well positioned to play such a modulatory role.
Control of interactions between V 1 and V 0 is also crucial in regulating the assembly state of the V-ATPase in vivo, which in yeast is modulated by glucose levels in the media (44). If, as suggested above, subunit C is able to induce the complex to adopt a less stable conformation, glucose removal may trigger dissociation through changes brought about by subunit C. By interacting with key subunits of the peripheral stalk (including subunits E, G, and a), subunit C may be able to disrupt interactions between these subunits that would otherwise lead to a stably assembled peripheral stalk. In this regard subunit C plays a role that is unique to the V-ATPases. As previously pointed out (43), the peripheral stalk of the V-ATPases must be both sufficiently stable to survive the torque generated during rotary catalysis and sufficiently labile to lead to dissociation in the absence of glucose. Because there is no evidence for functional dissociation of the F-ATPase in vivo, it is not surprising that no homolog to subunit C exists for these enzymes.
In summary, we have identified novel interactions between specific domains of subunit C and both subunit G in the pe- Right panel, structural model of subunit C showing the location of residues whose mutation to cysteine results in MBP-mediated cross-linking to subunit E (Gly 27 , Asn 216 , Ala 220 , and His 324 ; shown in green), subunit G (Ala 220 , shown in orange) or subunit a (Ser 57 , Ser 69 , shown in blue). Residues in black are facing out of the page and those shown in gray are facing into the page. Left panel, placement of subunit C into the complete structural model of the V-ATPase complex, showing contacts identified between subunit C and subunits E and G in the V 1 domain and subunit a in the V 0 domain. ripheral domain and subunit a in the integral domain, and have further defined the sites of interaction between subunits C and E. By interaction with multiple subunits that bridge V 1 and V 0 , subunit C is able to effectively modulate the stability and properties of the V-ATPase complex.