TM2 but Not TM4 of Subunit c (cid:1) Interacts with TM7 of Subunit a of the Yeast V-ATPase as Defined by Disulfide-mediated Cross-linking*

The vacuolar (H (cid:1) )-ATPase (or V-ATPase) is an ATP-de-pendent proton pump which couples the energy re-leased upon ATP hydrolysis to rotational movement of a ring of proteolipid subunits (c, c (cid:2) , and c (cid:1) ) relative to the integral subunit a. The proteolipid subunits each contain a single buried acidic residue that is essential for proton transport, with this residue located in TM4 of subunits c and c (cid:2) and TM2 of subunit c (cid:1) . Subunit c (cid:1) contains an additional buried acidic residue in TM4 that is not required for proton transport. The buried acidic residues of the proteolipid subunits are believed to interact with an essential arginine residue (Arg 735 ) in TM7 of subunit a during proton translocation. We have previously shown that the helical face of TM7 of subunit a containing Arg 735 interacts with the helical face of TM4 of subunit c (cid:2) bordered by Glu 145 and Leu 147 (Kawasaki-Nishi et al. (2003) J. Biol. Chem . 278, 41908–41913). We have now analyzed interaction of subunits a and c (cid:1) using disulfide-mediated cross-linking. The results indicate that the helical face of TM7 of subunit a containing Arg 735 interacts with the helical face of TM2 of subunit c (cid:1) centered on EDTA and N -ethylmaleimide to a final concentration of 15 m M and 20 m M , respectively. Samples were then subjected to SDS- PAGE on 4–15% acrylamide gels and transferred to nitrocellulose membranes. The blots were probed with horseradish peroxidase-conju- gated monoclonal antibody 3F10 against the HA epitope tag (36). Blots were developed using the Supersignal ULTRA chemiluminescent sys- tem (Pierce). Other Procedures— ATPase activity was measured using a coupled spectrophotometric assay and ATP-dependent proton transport was measured by fluorescence quenching using the fluorescence probe ACMA as described previously (43). All assays were carried out in the presence or absence of 1 (cid:1) M concanamycin A, a specific inhibitor of the V-ATPase (44). SDS-polyacrylamide gel electrophoresis was carried out as described by Laemmli (45).

The vacuolar (H ؉ )-ATPase (or V-ATPase) is an ATP-dependent proton pump which couples the energy released upon ATP hydrolysis to rotational movement of a ring of proteolipid subunits (c, c, and c؆) relative to the integral subunit a. The proteolipid subunits each contain a single buried acidic residue that is essential for proton transport, with this residue located in TM4 of subunits c and c and TM2 of subunit c؆. Subunit c؆ contains an additional buried acidic residue in TM4 that is not required for proton transport. The buried acidic residues of the proteolipid subunits are believed to interact with an essential arginine residue (Arg 735 ) in TM7 of subunit a during proton translocation. We have previously shown that the helical face of TM7 of subunit a containing Arg 735  The vacuolar (H ϩ )-ATPases (or V-ATPases) 1 are a family of ATP-dependent proton pumps that function in both intracellular compartments and the plasma membrane (1)(2)(3)(4)(5)(6)(7)(8). Within intracellular compartments such as lysosomes, endosomes, the Golgi and secretory vesicles, V-ATPases function in a variety of processes, including protein degradation, receptor-mediated endocytosis, viral entry, intracellular membrane traffic, protein processing, and neurotransmitter uptake (1). Plasma membrane V-ATPases have been shown to function in acid secretion in the kidney, bone degradation by osteoclasts, pH homeostasis in macrophages and neutrophils, sperm maturation in the vas deferens, K ϩ transport by insect goblet cells, and invasion by tumor cells (9 -13).
Subunits of the V-ATPase are organized into a peripheral domain (V 1 ) responsible for ATP hydrolysis and an integral domain (V 0 ) that carries out proton translocation (1)(2)(3)(4)(5)(6)(7)(8). ATP hydrolysis in the V 1 domain occurs at nucleotide binding sites located at the interface of the A and B subunits (14,15), which form the hexameric structure observed in electron microscopic images of the V-ATPase (16). ATP hydrolysis has been shown to drive rotation of a central stalk, composed of the D and F subunits (17,18), which in turn drives rotation of a ring of proteolipid subunits (c, cЈ, and cЉ) relative to subunit a (19,20). Subunit a is an integral membrane protein possessing an Nterminal hydrophilic domain located on the cytoplasmic side of the membrane and a hydrophobic C-terminal domain containing nine transmembrane segments (21). Subunit a is held fixed relative to the hydrolytic head of V 1 by a peripheral stalk composed of subunits C, E, G, H, and the hydrophilic domain of subunit a (18,22,23).
The membrane integral domain of subunit a contains a number of buried charged residues, including Glu 789 , His 743 , and Arg 799 , whose mutation results in partial inhibition of proton transport (24 -26). The only a subunit residue absolutely required for proton translocation is Arg 735 located in TM7 (26). Even conservative replacement of this residue with lysine results in complete loss of proton transport (26). Arg 735 has been postulated to function in displacement of protons bound to buried acidic residues on the ring of proteolipid subunits, analogous to the function of Arg 210 in proton transport by the F-ATPases (27)(28)(29).
The V-ATPases contain three different proteolipid subunits (c, cЈ, and cЉ) which are present in a stoichiometry of one copy each of cЈ and cЉ and 4 -5 copies of c (30,31). Subunits c and cЈ are composed of four transmembrane segments (32,33) whereas subunit cЉ contains four or five transmembrane segments (34,35). Each proteolipid subunit contains a single glutamic acid residue buried in the middle of one of these segments that is essential for proton transport (33). For subunits c and cЈ, the essential glutamic acid residue is present in the last transmembrane segment, whereas for subunit cЉ the essential glutamic acid residue is present near the middle of the molecule (32)(33)(34)(35). It is these essential glutamate residues that are believed to undergo reversible protonation and deprotonation during rotary catalysis and that are thought to interact with Arg 735 of subunit a to activate proton release. It is therefore important to define the helical interactions that occur between the proteolipid subunits and subunit a within the V 0 domain.
We have previously demonstrated by cysteine-mediated crosslinking that TM7 of subunit a is in close proximity to TM4 of subunit cЈ and have identified the helical faces of these subunits that interact (36). In the present study we have extended this analysis to the interaction of subunit a and subunit cЉ, whose topology and location of essential residues make it unique among the proteolipid subunits of both the V and F-ATPases.

EXPERIMENTAL PROCEDURES
Materials and Strains-Zymolyase 100T was obtained from Seikagaku America, Inc. Protease inhibitors and the monoclonal antibody 3F10 (directed against the HA antigen) that is conjugated with horseradish peroxidase were from Roche Applied Science. The monoclonal antibody 10D7 against the yeast 100 kDa a subunit Vph1p (37) was from Molecular Probes. Escherichia coli and yeast culture media were purchased from Difco Laboratories. Restriction endonucleases, T4 DNA ligase and other molecular biology reagents were from Fisher and New England Biolabs. Phenylmethylsulfonyl fluoride and most other chemicals were purchased from Sigma. Yeast strains lacking the VMA16, VPH1, and STV1 genes were constructed by replacing the entire coding region of VMA16 with the TRP gene and insertion of the LEU gene into the VPH1 gene and the LYS gene STV1 gene at the positions indicated by Manolson et al. (38) using the YPH500 strain (MAT alpha ade2, ura3, leu2, his3, trp1, lys2). The proteolipid Vma16p was tagged at the C terminus with the 9-amino acid epitope (YPYDVPDYA) from influenza hemagglutinin (HA) as described previously (34). YEPD buffered to pH 5.5 or pH 7.5 was used for selection of strains showing a vma Ϫ phenotype.
Transformation and Selection-Site-directed mutants of Vph1p and Vma16p were constructed using the Altered Sites II in vitro mutagenesis system (Promega), and the presence of the mutations was verified by sequencing the entire length of subcloned DNA. Plasmids carrying mutations of a and cЉ are co-transformed into yeast cells lacking functional endogenous Vph1p, Stv1p, and Vma16p by the lithium acetate method (39). The transformants were selected on histidine minus and uracil minus plates and growth phenotypes of the mutants 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.
Analysis of Subunit Expression and V-ATPase Assembly-Yeast vacuolar membranes and whole cell lysates were prepared using the protocol described previously (40,41). Whole cell lysates and vacuolar membrane enriched fractions were separated by SDS-PAGE on 4 -15% gradient acrylamide gels. The presence of Vma16p on vacuolar membranes was detected by Western blotting using a horseradish peroxidase-conjugated monoclonal antibody 3F10 against HA, while Vph1p was detected using the monoclonal antibody 10D7 (Molecular Probes), followed by a horseradish peroxidase-conjugated secondary antibody (Bio-Rad) (36). Blots were developed using a chemiluminescence detection method obtained from Kirkegaard & Perry Laboratories (Gaithersburg, MD).
Cross-linking of Subunits a and cЉ by Cu(1,10-phenanthroline) 2 SO 4 (CuP)-Cross-linking between cysteine residues introduced into subunits a and cЉ was performed using the protocol described previously by Jiang and Fillingame (42). Vacuolar membrane vesicles (ϳ50g of protein) were washed in labeling buffer (5 mM Tris-Mes, pH 7.5, 0.25 mM MgCl 2 and 1.1 M glycerol). 2.5 mM Cu(1,10-phenanthroline) 2 SO 4 was added, and samples were incubated for 60 min at room temperature to catalyze disulfide bond formation. The reaction was terminated by addition of EDTA and N-ethylmaleimide to a final concentration of 15 mM and 20 mM, respectively. Samples were then subjected to SDS-PAGE on 4 -15% acrylamide gels and transferred to nitrocellulose membranes. The blots were probed with horseradish peroxidase-conjugated monoclonal antibody 3F10 against the HA epitope tag (36). Blots were developed using the Supersignal ULTRA chemiluminescent system (Pierce).
Other Procedures-ATPase activity was measured using a coupled spectrophotometric assay and ATP-dependent proton transport was measured by fluorescence quenching using the fluorescence probe ACMA as described previously (43). All assays were carried out in the presence or absence of 1 M concanamycin A, a specific inhibitor of the V-ATPase (44). SDS-polyacrylamide gel electrophoresis was carried out as described by Laemmli (45).

Growth Phenotype of Yeast Strains Expressing Single Cysteine-containing Mutant Forms of Vma16p and Vph1p-To
study the interaction between subunits a and cЉ, a yeast strain disrupted in the two genes that encode subunit a (VPH1 and STV1) as well as the gene encoding subunit cЉ (VMA16) was constructed as follows. Using the parental strain YPH500, the VMA16 gene was replaced with the TRP gene, and the LEU and LYS genes were inserted into the VPH1 and STV1 genes, respectively, as previously described (38,46). The VMA16 gene was expressed in this strain using the pRS413 plasmid containing the HIS marker whereas the VPH1 gene was expressed using the pRS316 plasmid containing the URA marker. A Cysless form of Vph1p has previously been shown to support wildtype growth at pH 7.5 (21), suggesting it forms a V-ATPase complex possessing at least 20% of wild-type levels of activity in vitro (47,15). Cells co-expressing Cys-less forms of both Vph1p and Vma16p also displayed wild-type growth at pH 7.5, indicating that none of the endogenous cysteine residues in either protein are essential for activity. TM7 of Vph1p contains a buried arginine residue (Arg 735 ) that is absolutely required for proton transport by the V-ATPase and that has been proposed to interact with the buried carboxyl groups on the proteolipid subunits (26). Unique cysteine residues were therefore introduced at nine different positions along TM7 of Vph1p, including sites predicted to reside on the same helical face as Arg 735 (Ser 728 , Ala 731 , Ser 732 , Ala 738 , Leu 739 , and Ala 742 ), sites adjoining this helical face (Leu 734 and Leu 736 ) and sites on the opposite helical face (Tyr 733 ). Unique cysteine residues were also introduced into the two transmembrane helices of Vma16p that contain buried acidic residues. Assuming a four transmembrane segment model of Vma16p (34), these correspond to TM2 and TM4 (see "Discussion"). In TM2 of Vma16p, which contains the essential glutamic acid residue Glu 108 , cysteine residues were introduced at Ser 103 , Ile 104 , Ile 105 , Phe 106 , Ser 107 , Glu 108 , Val 109 , Val 110 , Ala 111 , and Ile 112 . These residues correspond to approximately the central half of TM2 of Vma16p and represent all of the helical faces of this segment (33). In TM4 of Vma16p, which contains a buried glutamic acid reside (Glu 188 ) that is not essential for proton transport (33), cysteine residues were introduced at Ile 184 , Leu 185 , Val 186 , Ile 187 , Glu 188 , Ile 189 , Phe 190 , Gly 191 , Ser 192 , and Ile 193 . The yeast strain disrupted in VMA16, VPH1, and STV1 was then transformed with plasmids bearing the Cys-less form or one of the single cysteine-containing mutants of Vma16p as well as one of the single cysteine-containing FIG. 1. Assembly of cysteine-substituted mutants of subunits a and c؆. To assess the ability of the indicated mutant forms of subunit a (Vph1p) and subunit cЉ (Vma16p) to assemble into V 0 complexes, vacuolar membrane fractions were isolated from each strain, subjected to SDS-PAGE. Western blotting was performed using the monoclonal antibodies 10D7 (against Vph1p) or 3F10 (against the HA tag introduced at the C terminus of Vma16p) as described under "Experimental Procedures." mutants of Vph1p. The growth phenotype of the resultant 189 double-replacement strains at pH 7.5 is shown in Table I. As expected from previous results, strains bearing mutations at Glu 108 of Vma16p showed no growth at neutral pH whereas most strains expressing the E188C mutant of Vma16p displayed a wild-type growth phenotype (33). For mutations in the a subunit, the only mutation to seriously compromise growth at pH 7.5 was Y733C. The majority of the 189 double mutants tested showed normal (or near normal) growth at neutral pH, suggesting that the resultant V-ATPase complexes possessed significant activity in vivo.
Assembly of V 0 Complexes Containing Cysteine Mutations in Subunits a and cЉ-To assess the effects of the mutations in Vma16p and Vph1p on assembly of the V 0 domain, partially purified vacuolar membranes were subjected to SDS-PAGE and Western blot analysis was performed using antibodies against both Vph1p and the HA-tagged Vma16p. As can be seen in Fig. 1, vacuolar membranes from all of the double replacement strains tested showed wild-type levels of both Vph1p and Vma16p, suggesting normal assembly and targeting of the V 0 domain (37).
While proton transport and ATPase activities have not been measured for vacuoles from the 189 double mutant strains analyzed, it was felt that useful information about the arrangement of V 0 subunits could be obtained even from complexes lacking substantial activity, as for those displaying a vmaϪ phenotype (Table I). This is similar to previous studies of contacts between subunits a and cЈ of the V-ATPase (36) and subunits a and c of the F-ATPase (42), where many of the strongest cross-links were observed for complexes lacking activity as assessed by their growth phenotype.
ATPase and Proton Transport Activity of E108Q and E188Q Mutants of Subunit cЉ-It had previously been reported that the mutations E108Q in TM2 but not E188Q in TM4 of subunit cЉ led to a vmaϪ phenotype (33). In addition, the E108Q mutation resulted in loss of 99% of wild-type ATPase activity in isolated vacuolar membranes and qualitative loss of vacuolar acidification as assessed by quinacrine staining of vacuoles in vivo (33). By contrast, the effect of the E188Q mutation on ATPase activity was not determined and the effect of this mutation on proton transport was not quantitated. We therefore measured both proton transport and concanamycin-sensitive ATPase activity in vacuolar membranes isolated from the wild-type strain and strains expressing the E108Q or E188Q mutations in subunit cЉ as described under "Experimental Procedures." In agreement with the previous findings (33), vacuolar membranes from the strain expressing the E108Q mutant of subunit cЉ possessed no detectable ATP-dependent proton transport or concanamycin-sensitive ATPase activity. Interestingly, vacuolar membranes from the strain expressing the E188Q mutant had 87 Ϯ 7% of the wild-type ATPase activity but only 64 Ϯ 4% of wild-type proton transport activity. These results suggest a partial uncoupling of the E188Q mutant and a possible function of this residue in proton transport.
Cysteine-mediated Cross-linking of Subunits a and cЉ Using Cupric Phenanthroline-To determine the proximity of cysteine residues introduced into TM7 of subunit a to cysteine residues introduced into TM2 and TM4 of subunit cЉ, vacuolar membranes were isolated from each of the strains expressing single cysteinecontaining forms of Vph1p and Vma16p followed by cross-linking using cupric phenanthroline, as described under "Experimental Procedures." Samples were separated by SDS-PAGE and Western blotting was performed using the horseradish peroxidaseconjugated monoclonal antibody 3F10 directed against the HA epitope tag introduced at the C terminus of Vma16p. Crosslinking was tested for all 189 strains expressing single cysteinecontaining forms of Vph1p and Vma16p, but results are shown in Fig. 2 for only those mutants in which cross-linked products were detected. As can be seen, strong cross-linked products of molecular mass ϳ125 kDa (corresponding to a heterodimer of subunits a and cЉ) were observed for the pairs aS728C/cЉE108C, aA731C/ cЉI112C and aA738C/cЉI105C. Bands of intermediate intensity were observed for the pairs aS728C/cЉF106C and aS728C/ cЉV109C and very faint cross-linked products were observed for the pairs aS728C/cЉI104C, aS732C/cЉV109C and aA742C/ cЉI112C. All of these products involved cross-linking between TM7 of subunit a and TM2 of subunit cЉ. Of the ninety strains expressing single cysteine mutations in TM4 of subunit cЉ, only a single very faint cross-linked product was observed for the pair aS728C/cЉI189C (Fig. 2). These results are summarized in Table  II and the strong and intermediate cross-linked products are depicted in Fig. 3. The left panel of Fig. 3 is based on a four transmembrane helix model of subunit cЉ (34), which places the N terminus of the helix containing Glu 108 on the luminal side of the membrane. The right panel of Fig. 3 is based on a five transmembrane helix model (35), which places the N terminus of this helix on the cytoplasmic side of the membrane (see "Discussion").
It should be noted that in all single cysteine-containing mutants of subunit cЉ (but not in the Cys-less form), a strongly reactive band of molecular mass ϳ40 kDa was observed (Fig. 2). Because subunits c and cЈ contain their endogenous cysteine residues in these strains (six in subunit c and five in subunit cЈ), and because this 40 kDa species was abolished following reduction of samples with 20% 2-mercaptoethanol (data not shown), it is likely that this 40 kDa species corresponds to a cross-linked heterodimer of either subunits c and cЉ or subunits cЈ and cЉ. DISCUSSION Subunit cЉ has a number of properties that make it unique among the proteolipid subunits of the V and F-ATPases (33-35). First, it contains an N-terminal hydrophobic region that was originally postulated to correspond to the first of five  transmembrane segments, giving it one more than subunits c and cЈ (33). Initial topological studies suggested that both the N and C terminus of subunit cЉ are exposed on the cytoplasmic side of the membrane, and that the N-terminal hydrophobic region is dispensable for function (34). Thus, subunit cЉ, as for the other two proteolipid subunits of the V-ATPase, would contain four transmembrane segments instead of five. The orientation of these segments relative to the cytosolic and luminal sides of the membrane in this model would be opposite to that of subunits c and cЈ (34). Recent studies using protease sensitivity of epitope tags introduced at the N and C termini of subunit cЉ, however, suggest that the N and C termini are on opposite sides of the membrane, consistent with a five transmembrane helix model (35). Subunit cЉ is also unique in the location of the essential buried acidic residue (33). Unlike subunits c and cЈ, which contain an essential glutamic acid residue in the last transmembrane segment, the essential glutamate in subunit cЉ is located in TM2 (Glu 108 ) (assuming a four TM model) or TM3 (assuming a 5 TM model). A glutamic acid residue present in the last transmembrane segment of subunit cЉ (Glu 188 ) has been shown not to be required for function (33). Although the stoichiometry of subunits in V 0 is a 1 d 1 c 4 -5 cЈ 1 cЉ 1 (30,31), the arrangement of these subunits has not been defined.
Because proton translocation through the V 0 domain is thought to occur at the interface of the a subunit and the ring of proteolipid subunits, it is important to define the helical interactions which occur at this interface. We have previously demonstrated that TM7 of subunit a containing the essential arginine residue Arg 735 is in close proximity to TM4 of subunit cЈ containing the critical glutamic acid residue Glu 145 (36). Moreover, these two helices do not appear to adopt a single relative orientation, but rather to expose somewhat different helical faces to each other. Thus, strong cross-linked products were observed for both aL739C/cЈL147C and aA731C/cЈE145C (36) (see Fig. 4), which requires an ϳ90 o rotation of TM7 of subunit a and a greater than 180 o rotation of TM4 of subunit cЈ. These results suggest a rotation of these helices relative to one another and are consistent with evidence for helical rotation within F 0 (29,48). Recent studies characterizing the bafilomycin and concanamycin binding sites on the V-ATPase indicate that these inhibitors bind at the interface of helices of the c subunit and may block activity by preventing helical rotation within or between c subunits (49,50).
Because of the unique properties of subunit cЉ, we wished to determine whether the transmembrane helices of this subunit adopt a similar relative orientation to TM7 of subunit a. In particular, we wished to compare the helical interactions of the two transmembrane segments of subunit cЉ that contain buried glutamic acid residues (TM2 and TM4 in the four TM model). Fig. 4 compares the helical surfaces that can be cross-linked to subunit a for both subunits cЈ and cЉ of the V-ATPases and subunit c of the F-ATPases, assuming the four TM model of subunit cЉ. As can be seen, TM2 of subunit cЉ is capable of forming a number of strong and intermediate cross-links with TM7 of subunit a, suggesting that these two helices are in close proximity. Moreover, the radial distribution of cross-links suggests rotation of these helices relative to one another. It should be noted that a number of a subunit residues were included in this analysis that are predicted to be on the sides (Leu 734 , Leu 736 ) or back (Tyr 733 ) of TM7 relative to Arg 735 , and that no cross-linking to subunit cЉ was observed at any of these sites. The locations of the critical glutamic acid residues are shown, assuming that they are buried within a four helix bundle corresponding to each subunit when not in contact with subunit a. The helical surfaces of subunits cЈ and cЉ that can be cross-linked to TM7 of subunit a are shaded. Arrows indicate the proposed direction of helix rotation. The arrangement of proteolipid subunits within the ring is not known, but is shown for the purposes of this model with subunits cЈ and cЉ at adjacent positions. All helices are viewed from the cytoplasmic side of the membrane.
We have also analyzed five additional a subunit positions for cross-linking to subunit cЈ, including His 729 , Thr 730 , Ser 732 , Leu 734 , and Leu 736 (Fig. 4), and found no cross-linking of cysteines at these positions to any of the ten cysteine residues introduced into subunit cЈ (data not shown).
Helical rotation is postulated to bring the arginine residue in subunit a into close proximity to the acidic residue on the proteolipid subunit to facilitate proton release (29,48). As can be seen from Fig. 4, the minimal movement of TM2 of subunit cЉ required to bring Glu 108 into close proximity to FIG. 6. Helical wheel diagrams depicting the location of the major cross-linked products obtained between subunit a and either subunit c or subunit c؆. Major cross-links observed between subunits a and cЈ are shown in the left-hand panels while those between subunits a and cЉ are shown on the right. The critical arginine residue in subunit a and the critical glutamic acid residue in subunits cЈ or cЉ are indicated by filled circles. The cross-linked products have been arranged in an order related to a possible mechanism of proton translocation occurring at the interface of subunit a and the proteolipid ring (see text). Helices are viewed from the cytoplasmic side. A four TM model of subunit cЉ is assumed in this diagram. Arg 735 involves a counterclockwise rotation, whereas for TM4 of subunit cЈ and TM2 of the F-ATPase subunit c, a clockwise rotation is required. It should be noted, however, that for the five TM model of subunit cЉ, the helical rotation required to bring Glu 108 into contact with Arg 735 would be clockwise as predicted for the other proteolipids. Moreover, the orientation of the transmembrane segment containing Glu 108 relative to subunit a would be very similar to that of TM4 of subunit cЈ (compare Fig. 3, right panel and Fig. 4). It is thus easier to explain the cross-linking results in the present study based upon a model of subunit cЉ containing five rather than four transmembrane segments.
By contrast with TM2, only one very weak cross-link was observed between TM4 of subunit cЉ and TM7 of subunit a. Although such a negative result is not conclusive, it suggests that these two transmembrane segments are not typically in close proximity. From the crystal structure of the partial F-ATPase complex containing a ring of 10 protolipid subunits (51), it might be predicted that TM2 and TM4 of the V-ATPase proteolipids would occupy equivalent positions relative to subunit a. The cross-linking results obtained, however, suggest that for subunit cЉ the preferred point of contact with subunit a is TM2 rather than TM4. If the transmembrane segments of subunits c and cЈ containing the essential glutamate residues are likewise preferred points of contact with subunit a (as suggested by previous results (36)), the position in the proteolipid ring occupied by subunit cЉ may form a gap in which these preferred contacts are not present on every other helical segment (see Fig. 5, top panel). This assumes that the arrangement of the four transmembrane helices in the three proteolipid subunits is the same. Such a gap might constitute an energy barrier to rotation that prevents passive proton translocation through V 0 . In fact, the isolated V 0 domain, unlike the F 0 , domain, has been shown to not normally catalyze passive proton translocation (52). On the other hand, if the arrangement of the four transmembrane helices is different for subunit cЉ relative to subunits c and cЈ (Fig. 5, bottom panel), no gap in the proteolipid ring would occur at the position of subunit cЉ. For a 5 TM model of subunit cЉ, the first transmembrane segment could be accommodated within the proteolipid ring (see box in Fig. 5). As noted above, however, rotation of the TM containing the critical glutamic acid residue in subunit cЉ would occur in the opposite direction to that depicted in Fig. 5.
Based upon lipid accessibility and intrasubunit cross-linking of the V-ATPase c subunit, Harrison et al. (53,54) have proposed a significantly different model for the arrangement of transmembrane segments in the proteolipid ring as compared with the structure observed for F 0 (51). In the Harrison model, TM4 is present at the outer surface of the proteolipid ring, TM2 and TM3 are present closer to the center of the ring and TM1 forms the inner most boundary of the ring. Our data are not consistent with subunit cЉ adopting such a helical arrangement, since in that case TM4 rather than TM2 would show strong cross-linking to subunit a. It is possible, however, to arrange the transmembrane segments in such a way as to place TM2 of subunit cЉ on the outer most face of the proteolipid ring and TM3 on the inner most surface. While this would require a long cytoplasmic loop spanning TM2 and TM3, this is consistent with the available sequence data, in which TM2 and TM3 are connected by a 23 amino acid hydrophilic bridge (33,34).
The major cross-linked species observed between subunits a and cЈ or cЉ have been arranged in Fig. 6 to suggest a possible sequence of states the enzyme may occupy during rotational catalysis. Beginning with the upper left hand panel, Glu 145 of subunit cЈ is facing away from TM7 of subunit a and Arg 735 is above the point of contact with TM4 of subunit cЈ. In going to state 2 (the second panel from the top), TM4 of cЈ is postulated to undergo clockwise rotation of 120 o , bringing this residue into closer proximity to TM7 of subunit a. From state 2 to state 3, TM4 of subunit cЈ continues its clockwise rotation while TM7 of subunit a also undergoes clockwise rotation of 60°. These two events bring Glu 145 and Arg 735 into close proximity and thus cause Glu 145 to lose a proton, which exits the membrane via a luminal access channel in subunit a. From state 3 to states 4 and 5, the ring of proteolipid subunits undergoes rotation, replacing one proteolipid subunit with another having the same helical orientation as in state 1. In states 4 and 5, however, Arg 735 is below the point of contact with TM4 of the proteolipid subunit, and must undergo counterclockwise rotation of 90°to "reset" to state 1. Not shown is the resetting of TM4 of the original proteolipid subunit, which requires reprotonation from a cytoplasmically oriented hemichannel in subunit a. The transmembrane segment of subunit cЉ containing Glu 108 is proposed to undergo a similar set of rotational states, involving either an initial counterclockwise rotation for the four TM model (Fig. 6, right panels), or an initial clockwise rotation for the five TM model (not shown). It should be noted that the different cross-linked products observed may not correspond to specific substates in the proton translocation pathway but rather to the possible helical orientations that can be accessed through thermal movements. It will be of interest to determine whether inhibitors such as bafilomycin, which are proposed to block activity by inhibiting helical rotation in the proteolipid subunits (49,50), effect the pattern of cross-linking between these subunits and subunit a.
The mechanism depicted is similar to, although less detailed than, models proposed for proton transport through F 0 (27,29,48,55). Although much less is know concerning the a subunit residues lining the aqueous access channels of the V-ATPase than of the F-ATPase (55), it is interesting to note that His 729 and His 743 of Vph1p are present on TM7 of subunit a above and below Arg 735 . Both of these residues (whose mutation leads to partial inhibition of proton transport (24,25)) would be expected to move during rotation of TM7. It is possible that rotation of TM7 may move these histidine residues in such a way that either the luminal or cytoplasmically oriented hemichannels become open. Additional information concerning the arrangement of transmembrane helices in subunit a will be required to test this model.