Arrangement of the Multicopy H 1 -translocating Subunit c in the Membrane Sector of the Escherichia coli F 1 F 0 ATP Synthase*

The multicopy subunit c of the H 1 -transporting F 1 F 0 ATP synthase of Escherichia coli is thought to fold across the membrane as a hairpin of two hydrophobic a -helices. The conserved Asp 61 , centered in the second transmembrane helix, is essential for H 1 transport. In this study, we have made sequential Cys substitutions across both transmembrane helices and used disulfide cross-link formation to determine the oligomeric arrangement of the c subunits. Cross-link formation between single Cys substitutions in helix 1 provided initial limitations on how the subunits could be arranged. Double Cys substitutions at positions 14/16, 16/18, and 21/23 in helix 1 and 70/72 in helix 2 led to the formation of cross-linked multimers upon oxidation. Double Cys substitutions in helix 1 and helix 2, at residues 14/72, 21/65, and 20/66, respectively, also formed cross-linked multimers. These results indicate that at least 10 and prob-ably 12 subunits c interact in a front-to-back fashion to form a ring-like arrangement in F 0 . Helix 1 packs at the interior and helix 2 at the periphery of the ring. The model indicates that the Asp 61 carboxylate is centered between the helical faces of adjacent subunit c at the center of a four-helix bundle. F the formation of utilizing the energy of a transmembrane H 1 electrochemical gradient, generated by electron transport complexes and other ion pump-ing systems. Closely related ATP synthases are found in the plasma membrane of eubacteria, the inner membrane of mitochondria, and the thylakoid membrane of chloroplasts. with Pst I and Ava I and ligated into the respective sites of plasmic pNOC. Primers designed for substitution of site. Amplification to 2172–2189, Hpa site cloning of into the BsrG 1 Hpa I sites of plasmid pNOC. and samples were incubated for 1 h at room temperature. Products not reduced by these conditions could be reduced by treatment with 2 3 sample buffer containing 10% b -mercap- toethanol and 8 M urea for 1–12 h at 22–24 °C. membrane by acrylamide gel electrophoresis with the Tris-Tricine buffer system of Scha¨gger and (28). After electrophoresis, proteins were transferred from the gel electrophoretically onto a polyvinylidene difluoride membrane (29). Rabbit antisera specific to subunit c (30) was pretreated as described (31) and diluted 1:10,000 prior to use. Immu-nostaining was carried out using the ECL system (Amersham Pharma- cia Biotech), and multiple exposures were scanned within a linear range of intensity to estimate the extent of cross-link product formation.

F 1 F 0 ATP synthases catalyze the formation of ATP utilizing the energy of a transmembrane H ϩ electrochemical gradient, generated by electron transport complexes and other ion pumping systems. Closely related ATP synthases are found in the plasma membrane of eubacteria, the inner membrane of mitochondria, and the thylakoid membrane of chloroplasts. The enzyme is a multisubunit complex with distinct extramembranous and transmembrane domains, termed F 1 and F 0 , respectively. Ion movement through F 0 is coupled to ATP synthesis/ hydrolysis at sites in F 1 (1,2). The simplest F 1 sectors, as found in Escherichia coli, consist of five subunits in an ␣ 3 ␤ 3 ␥␦⑀ stoichiometry. Homologous subunits are found in mitochondria and chloroplasts. A high resolution structure of a substantial part of bovine F 1 shows the three ␣and three ␤-subunits to alternate around a central core through which subunit ␥ extends and protrudes (3). The structure fits well with the binding change mechanism proposed by Boyer and co-workers (4), where each of the three ␤-subunits alternates between the loose binding of ADP plus P i , tight ADP plus P i binding and ATP synthesis, and ATP release during catalytic turnover.
Several recent studies now show that subunit ␥ rotates within the core of the hexagonally arranged ␣ 3 ␤ 3 complex to presumably drive the binding changes in each ␤-subunit (5-7). Both subunit ␥ and ⑀ appear to rotate as a unit (8). The mechanism of coupling H ϩ translocation through F 0 to ␥ subunit rotation is unknown.
The E. coli F 0 is the simplest type found in nature. It consists of three subunits with a stoichiometry of a 1 b 2 c 10 ϩ 1 (9). Electron microscopic studies suggest that the a and b subunits pack at the periphery of a complex of subunit c (10). Subunit b, with a single transmembrane helix and larger cytoplasmic domain, is proposed to associate with subunit ␦ to make up a stator that binds and fixes the F 1 ␣ 3 ␤ 3 catalytic head group to F 0 (11,12). Subunit a is thought to fold through the membrane with five or six transmembrane helices and play a key role in the H ϩ transport (2,13). Structural and genetic studies indicate that subunit c folds in the membrane as a hairpin of two hydrophobic ␣-helices connected by a polar loop on the F 1 binding side of the membrane (2,14). A conserved Asp or Glu (Asp 61 in E. coli) centered in the second transmembrane helix is essential for H ϩ transport (2,15). The most compelling evidence for a direct role of the side chain carboxyl in cation binding has come from work on the related Na ϩ -translocating enzyme of Propiogenium modestum (16,17) and on a modified E. coli enzyme that binds Li ϩ (18). The coupling of proton movements to binding changes in F 1 appears to occur by an interaction between the loop region of subunit c with subunits ⑀ and ␥ (15,19,20). Several models have been proposed whereby ATP synthesis in the F 1 domain is coupled to proton movements through F 0 via movements of subunit c relative to the multicopy subunit a (21)(22)(23).
Information on the structural organization of F 0 subunits is essential if we are to understand how H ϩ transport is coupled to ATP synthesis. In this study, Cys was substituted into subunit c in order to determine the arrangement of subunits by disulfide cross-linking. The cross-linking data support the structural model of monomeric subunit c as determined by NMR (24). 1 A ringlike arrangement of 12 subunits c with helix 1 in the center and helix 2 at the periphery is indicated. The subunits interact with the front face of one subunit packed against the back face of the next subunit with the Asp 61 carboxylate centered within a four-helix bundle between adjacent subunits c. The model provides insights and limitations into how H ϩ translocation can be coupled to the rotation and synthesis of ATP within F 1 .

Oligonucleotide-directed Mutagenesis and Plasmid Construction-
The plasmids used in this study are derivatives of plasmid pDF163, which contains the wild type uncBEFH genes (bases 870 -3216) 2 (25). * This work was supported by National Institutes of Health Grant GM23105 and a grant from the Human Frontiers Science Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Plasmid pNOC, a derivative of plasmid pDF163 with a C21S substitution in subunit b, was used as a template for the introduction of all the Cys substitutions described. All three of the F 0 subunits coded by plasmid pNOC lack Cys. Plasmid pNOC was constructed by a rapid site-directed mutagenesis procedure (26). An antisense oligonucleotide, 5Ј-GGCGGCCATACGTACTTCATGGAGAACAG-3Ј, corresponding to positions Leu 19 -Pro 27 of subunit b was synthesized to incorporate a single base change (underlined) to create a Ser codon at position 21 and to overlap the nearby SnaB1 restriction enzyme site (italics). The polymerase chain reaction (PCR) 3 was then performed using this primer and a sense oligonucleotide primer designed to the coding strand (bases 1540 -1560), upstream of the PstI restriction enzyme site (1561-1566), using plasmid pDF163 DNA as the template. The PCR product was then digested with PstI and SnaB1 restriction enzymes and ligated to the equivalent sites of plasmid pDF163 to generate plasmid pNOC. The substitution was verified by sequencing the entire fragment.
PCR mutagenesis procedures were used to generate other Cys substitutions. Where possible a one-step PCR strategy was employed by taking advantage of restriction enzyme sites in the vicinity of the desired substitution. The majority of helix 1 substitutions were constructed in this fashion using BsrG1 (bases 1911-1916, overlapping amino acid positions 9 -11) and AvaI (bases 1976 -1981, overlapping amino acid positions 31 and 32) sites. Briefly, if the substitution was within 10 amino acid residues of the restriction site, a primer was designed so as to incorporate the substitution and the restriction site. Residues 20 -30 were substituted using antisense primers incorporating the Cys codon and AvaI site. Amplification was carried out using a sense primer 5Ј to the PstI site (bases 1561-1566). The PCR product was then digested with PstI and AvaI and ligated into the respective sites of plasmic pNOC. Primers designed for substitution of residues 12-17 incorporated the BsrG1 site. Amplification was achieved with an antisense primer to bases 2172-2189, 3Ј to the nearby HpaI site (bases 2162-2167). This enabled cloning of the product into the BsrG1 and HpaI sites of plasmid pNOC.
For the remaining substitutions and those in helix 2, a two-stage PCR mutagenesis procedure was used (26). This procedure requires a specific mutagenic primer and two wild type primers, 5Ј and 3Ј to the region of interest. In this case, the sense primer 5Ј to the region was designed to bases 1540 -1560 so that the PstI site was incorporated into the PCR product. The antisense primer 3Ј to the region was designed to bases 2303-2319, so that the HpaI and SnaB1 (bases 2256 -2262) sites were incorporated into the PCR product. The first PCR step involves use of the mutagenic primer with one of the wild type primers. This first product then serves as a megaprimer for the second round of PCR with the second wild type primer. The product was then digested with PstI and HpaI or SnaB1 and ligated into the respective sites of plasmid pNOC.
Double Cys substitutions were introduced in combination in helix 1 and helix 2 by subcloning. The PstI/AvaI fragment from a plasmid carrying a single Cys substitution in helix 1 was ligated into the respective sites of a pNOC plasmid derivative carrying a helix 2 Cys substitution. Correct subcloning was confirmed by DNA sequencing. To generate double Cys substitutions in helix 1, or double Cys substitutions in helix 2, the PCR mutagenesis procedures described above were employed using one of the single Cys mutant DNA as the template with a primer generated to create the second Cys substitution. All substitutions were verified by sequencing the entire subcloned fragment.
Preparation of membrane vesicles, protein and ATPase assays, and analysis of cross-linked products by gel electrophoresis and immunblotting were carried out as described previously (19). Cross-linking studies involved the addition of 1.5 mM CuP as an oxidant to catalyze disulfide bond formation. Membrane vesicles at 5 mg/ml (protein concentration) in TMG buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , and 10% (v/v) glycerol) were treated with 1.5 mM CuP for 1 h at room temperature (22-24°C). The reaction was terminated with 50 mM EDTA and 25 mM N-ethylmaleimide and incubated for an additional 10 min. The addition of EDTA and N-ethylmaleimide to the reaction mix for 10 min prior to oxidant addition prevented cross-link formation. This was also true for membrane vesicles solubilized in sample buffer. N-Ethylmaleimide alone was not sufficient. The sample was then mixed with an equal volume of 2ϫ SDS sample buffer (125 mM Tris-HCl, pH 6.8, 4% SDS, 20% (v/v) glycerol, and 0.02% bromphenol blue) containing 20 mM EDTA. To reduce disulfide cross-links, 2ϫ sample buffer was made with 5% (v/v) in ␤-mercaptoethanol, and samples were incubated for 1 h at room temperature. Products not reduced by these conditions could be reduced by treatment with 2ϫ sample buffer containing 10% ␤-mercaptoethanol and 8 M urea for 1-12 h at 22-24°C.
The solubilized membrane proteins were separated by SDS-polyacrylamide gel electrophoresis with the Tris-Tricine buffer system of Schä gger and von Jagow (28). After electrophoresis, proteins were transferred from the gel electrophoretically onto a polyvinylidene difluoride membrane (29). Rabbit antisera specific to subunit c (30) was pretreated as described (31) and diluted 1:10,000 prior to use. Immunostaining was carried out using the ECL system (Amersham Pharmacia Biotech), and multiple exposures were scanned within a linear range of intensity to estimate the extent of cross-link product formation.  Table I). The Cys substitutions coded in plasmid pNOC, which carries the uncBEFH genes encoding subunits a, c, b, and ␦ of F 1 F 0 respectively, were transformed into the ⌬uncBEFH strain JWP109. Growth of transformants was tested on succinate minimal medium, where growth depends on a functional oxidative phosphoryla-

Effect of Single Cys Substitutions on Function-Single
a Amino acid substitution in subunit c. WT corresponds to wild type uncBEFH genes in plasmid pDF163. Plasmid pNOC harbors the bC21S substitution in pDF163.
b Colony size after 72 h of incubation at 37°C on minimal plates containing 22 mM succinate.
Disulfide Cross-link Formation by Single Cys Substitutions-Membrane vesicles were treated with the oxidant CuP and analyzed by gel electrophoresis and immunoblotting. Residues within the N-terminal region (positions 8 -11) and Cterminal region (positions 73-75 and 78) formed high yield, disulfide cross-linked homodimers in the presence of oxidant ( Fig. 1 and Fig. 2). The F53C protein also formed a high yield cross-linked homodimer. Significant spontaneous homodimer formation occurred at positions L8C, L9C, and V78C. These positions are predicted to be close to where the helices emerge from the hydrophobic core of the membrane, where they may be more susceptible to spontaneous oxidation. The consecutive stretch of cross-links from residues 8 -12 and 73-75 are consistent with these regions being very flexible, with a greater susceptibility to thermal collision, perhaps because they lie at the N and C termini of the protein.
High yield cross-link formation (Ͼ25%) also occurred at positions 15, 26, and 30 within the predicted transmembrane region of helix 1. Each of these residues falls on the same face of helix 1 in the NMR-derived structure (see Fig. 3A). The G29C substitution would also fall on this face, but the significance of cross-linking here can be discounted due to lack of function and possible structural perturbations. Two other proteins, A20C and G27C, formed disulfide homodimers in lower yield. The positions of these residues in the final model are discussed below.
Formation of cross-linked homodimers within the transmembrane segment of helix 2 occurred more extensively but in generally lower yield. The region around Asp 61 was particularly prone to cross-linking and may reflect structural flexibility that relates to proton translocating function. These crosslinks are considered further under "Discussion." Cross-link Formation between Helix 1 Double Cys Substitutions-Double Cys substitutions were introduced into helix 1 to see if cross-linked multimers of subunit c could form (Table II). As discussed above, single Cys substitutions at positions 15, 26, and 30 result in formation of cross-linked homodimers in high yield, and these residues fall on the same face of helix 1 in the NMR structure. We reasoned that Cys substitution at two of these positions should provide information on (i) whether homodimer formation in the single Cys substitutions was due to two helices 1 coming together face to face along this surface (Fig. 3B) or (ii) whether these faces of the two helices neighbor each other as would be the case in a ring type arrangement (Fig. 3C). If the helices are arranged face-to-face (Fig. 3B), then only homodimer formation would occur, whereas in a ring type arrangement cross-linked multimers, up to the number of subunits c in the F 0 , should be detected. All combinations of double Cys substitutions at positions 15, 26, and 30 led to an extensive ladder of subunit c cross-linked multimers ( Fig. 4; Table II).
The G18C mutant was unable to grow on succinate and was present at relatively low levels. Gly is also found at the position corresponding to position 18 in a number of other bacteria (32). In a ring type arrangement of helices, Gly 18 falls next to Leu 19 of a neighboring helix 1 (Fig. 3C). Leu at position 19 is also conserved in these bacteria. One exception is in Streptococcus faecalis, where position 18 is a Met and position 19 a Gly. We reasoned that if helix 1 is arranged as depicted in Fig. 3B or Fig. 3C, then changing Leu 19 to an amino acid with a smaller side chain might revert the phenotype of the mutant G18C. Membrane vesicles were treated with (ϩ) and without (Ϫ) CuP, and 20 g of protein was separated on a 15% polyacrylamide gel and electroblotted to polyvinylidene difluoride membrane. The blot was probed with antiserum specific to subunit c. Cys substitutions are indicated according to their position within subunit c. CuP-treated membranes from the L8C mutant were incubated with sample buffer containing 5% ␤-mercaptoethanol for 1 h prior to loading (ϩ/R). Membranes from the uncBEFH-deleted strain JWP109 (c DEL), plasmid pDF163 in strain JWP109 (WT), and plasmid pNOC (bC21S) in strain JWP109 were also analyzed. The positions of the monomer (c) and homodimer (c 2 ) of subunit c are shown by arrows.

FIG. 2. Immunoblot analysis of Cys substitutions in helix 2.
Membrane vesicles were treated with (ϩ) and without (Ϫ) CuP and (12.5 g of protein) separated on a 15% polyacrylamide gel, and electroblotted. The blot was probed with antiserum specific to subunit c. Cys substitutions are indicated according to their position within subunit c.
Further, the G18C/L19C mutant might form cross-linked multimers upon oxidation if the arrangement shown in Fig. 3C is correct, whereas it should only form dimers if the arrangement is like that shown in Fig. 3B. Leu 19 in the G18C mutant was changed to Ala and to Cys. The Leu 19 substitutions in the G18C/L19A and G18C/L19C pairs both worked as second site suppressors to allow growth on succinate (Table II). The G18C/ L19C mutant membranes also formed subunit c multimers in the presence of oxidant (Table II; Fig. 4).   To further investigate this arrangement, double Cys substitutions were generated at positions 14/16, 16/18, 16/18 with the L19A suppressor substitution, and 21/23. All of the mutants formed subunit c multimers upon oxidation (Fig. 5A). The M16C/G18C mutant grew poorly on succinate, whereas the M16C/G18C/L19A grew well. Cys substitutions at positions 16 and 18 gave the most defined ladder of subunit c multimers, as shown in Fig. 5, A and B. The ladder clearly extends to distinct multimers at the position of c 10 and lighter multimers at the position of c 12 .

Cross-link Formation between Helix 2 Double Cys Substitutions-The
Cys substitutions in helix 1 are consistent with subunit c being arranged in a ring, implying that a similar arrangement must exist for helix 2. A variety of double Cys substitutions were generated in helix 2, falling at different offsets around the helix (Table II; Fig. 3D). Double Cys substitutions between positions 67-72 were of particular significance, since this region was not subject to high yield dimer homodimer formation in the respective single Cys substitutions. Further, the region starting at Met 65 appears to be of a continuous and regular ␣-helix in the NMR structure (Fig. 3D). Of the eight mutants constructed, L70C/L72C formed multimers upon oxidation (Fig. 6). The propensity of L70C/L72C to form a cross-linked multimer can satisfy a number of arrangements of helix 2 relative to helix 1. However, an oligomeric arrangement of subunit c in a ring with the interacting faces of helix 1 and helix 2 packed as in the NMR model and with helix 1 on the inside and helix 2 on the outside fits well with the cross-linking data (Fig. 3E). This places the critical Asp 61 toward the interior, positioned within a four-helix bundle.
Cross-link Formation between Helix 1 and Helix 2 Double Cys Substitutions-Cysteines were introduced into both helix 1 and helix 2 to determine their orientation with respect to each other (Table III). We reasoned that a Cys on helix 1 of one subunit c should be able to form a "diagonal" cross-link with a Cys on helix 2 of a neighboring subunit c in the oligomeric ring structure and thus elicit the formation of cross-linked multimers.
Three of the double Cys mutants, A14C/L72C, A20C/I66C, and A21C/M65C, formed extensive cross-linked multimers upon oxidation (Fig. 7). M17C/V68C and M17C/L72C also formed cross-linked multimers but at a lower yield (data not shown). The diagonal cross-linked products fit a model where each subunit c monomer has a structure similar to that proposed in the NMR structure, with each subunit c arranged in a ring as depicted in Fig. 3E. The A14C/L72C and A21C/M65C results are the most compelling, since the A14C and A21C mutants do not form homodimers, and the L72C mutant forms very low yield homodimers. In these cases, the cross-linked multimers of subunit c must be due to disulfide bond formation between helix 1 and helix 2. In the case of A20C/I66C, the single Cys substitution in either position does give rise to a noticeable c-c homodimer. Hence, in this case, multimer formation could result from cross-linked products of Cys 20 -Cys 20 and Cys 66 -Cys 66 , as well as helix 1/helix 2 cross-links. In some mutants, such as A14C/V68C and M16C/G69C, a cross-linked product corresponding to a subunit c trimer was identified that satisfied the arrangement. The ability of A20C/V68C to form a cross-linked band up to c 5 (results not shown), can be rationalized if the larger side chain of the A20C substitution is forced to occupy a position on the opposing side to where it normally packs. This must also be the case for c-c dimer formation to occur in the A20C and G27C mutants as well as several of the helix 2 single Cys mutants in this vicinity. DISCUSSION Singly and doubly Cys-substituted mutants of subunit c were generated here to determine by disulfide cross-link formation the arrangement of subunits in F 0 . The results confirm that subunit c is folded in a hairpin-like structure with two transmembrane helices. Double Cys substitutions that formed crosslinked multimers on oxidation but showed little or no homodimer formation as single Cys substitutions provide the most compelling evidence for a ringlike arrangement of subunits, as shown in Fig. 3E. The defining double Cys substitutions include the A14C/M16C, M16C/G18C, and G18C/L19C FIG. 5. Immunoblot analysis of double Cys substitutions in helix 1. Membrane vesicles were treated with (ϩ) and without (Ϫ) CuP, and 25 g of protein was separated on a 10 -15% polyacrylamide gradient gel and electroblotted. The blot was probed with antiserum specific to subunit c. The blots are presented and marked as described for Fig. 2.   FIG. 6. Immunoblot analysis of double Cys substitutions in helix 2. Membrane vesicles were treated with (ϩ) and without (Ϫ) CuP and 25 g of protein was separated on a 12-15% polyacrylamide gradient gel and electroblotted. The blot was probed with antiserum specific to subunit c. The blot is presented and marked as described for Fig. 2. pairs in helix 1 and the L70C/L72C pair in helix 2. Multimers formed by "diagonal" cross-linking between helix 1 and helix 2 define the orientation of helices with respect to each other and provide further evidence for a ringlike arrangement. The key double Cys substitutions that fall into this category are A14C/ L72C and A21C/M65C. The cross-linking data fit remarkably well with the proposed NMR model of monomeric subunit c (24). 1 In fact, the NMR model was used as a basis for predicting residues that were likely to cross-link in the experiments described above.
The formation of cross-linked homodimers from single Cys substitutions in helix 2 conflict with the model shown in Fig.  3E. Notably, the stretch of cross-linking is centered around Asp 61 . Cross-link formation may reflect structural flexibility in this region that is necessary to function, possibly a swiveling of helix 2 relative to helix 1 in events related to proton transport. As mentioned under "Results," it also seems possible that Cys replacement of Gly or Ala residues, which are normally packed at the interface between helices in the monomer, may result in packing of the larger side chain at either side of the interface. If there is some breathing in the packing of helices, the position of the sulfhydryl may change from side-to-side by a dynamic equilibrium. Cys side chains packed on opposite sides of the interface would obviously be susceptible to disulfide bond formation. This could account for the cross-linked homodimers seen in the G58C and A62C mutants. We have also carried out cross-linking studies in the presence of aminoxid WS35, a detergent that was previously used to solubilize and reconstitute intact F 1 F 0 (33). The addition of detergent resulted in high yield homodimer formation with all mutant membranes exhibiting even minor cross-linking with oxidant alone and also cross-linking with membranes in some mutants where crosslinking was not observed in the absence of detergent. The structure of solubilized F 0 must be somewhat different, i.e. more flexible than the structure in the lipid bilayer.
The high yield, cross-linked homodimers that form in the single Cys substitutions of helix 1, V15C, I26C, and I30C, fall on one face of an ␣-helix in the NMR structure, i.e. on the face that lines the inner core of the ring shown in Fig. 3E. The predominance of Gly and Ala in residues to either side of this face reduces the cross-sectional diameter throughout helix 1. For this reason, we depict helix 1 as being smaller in diameter than helix 2 in the schematic model (Fig. 3E). In a ring type arrangement, this is expected from from geometric packing constraints. The side chains of the face protruding into the central core are in close proximity with those in the neighboring helix 1, and preliminary modeling indicates C ␣ -C ␣ distances of 9 -10 Å, i.e. short enough to allow disulfide cross-link formation with only minor movements of helices. The opposing face of helix 2, i.e. the one positioned to the outside of the ring, is also hydrophobic with multiple branched chains. Residues in these two opposing faces are hydrophobic but not highly conserved (32), which is consistent with both surfaces being exposed to the lipid milieu of the membrane. In other membrane proteins, the surfaces involved in protein-protein contact show the greatest sequence conservation (34). Met 11 , Val 15 , and Leu 19 on the interior helical face are selectively labeled by 3-(trifluoromethyl)-3-(iodophenyl)diazirine, in addition to other selectively labeled residues elsewhere in the protein, which led to the suggestion that these residues were on one face of an ␣-helix that was exposed to the fatty acyl phase of the lipid bilayer (35). In the ring of subunit c suggested here, the diameter of the space in the central core is Ն25 Å, and it seems likely that lipid molecules will be present and account for the 3-(trifluoromethyl)-3-(iodophenyl)diazirine accessibility. Similar situations with centrally located lipid are seen with the bacteriorhodopsin trimer of the purple membrane (36) and presumed for the homooligomeric ring of the light-harvesting complex LH2 (37). The arrangement proposed here correlates well with recent scanning force microscopy studies, where F 0 appears as a ringlike structure surrounding a central dimple when viewed from the cytoplasmic side (38,39).
A variety of experiments support the idea that Asp 61 in E. coli and equivalent residues in other species bind and release protons in the transport step coupled to ATP synthesis (15,40). The essential Asp 61 in helix 2 of the E. coli protein can be moved to position 24 in helix 1 with retention of function (41). The NMR model of monomeric subunit c shows Ala 24 of helix 1 in close proximity to Asp 61 of helix 2, but the side chains point to opposite surfaces of the packed helices. 1 How then is function retained in the Asp interchange mutant? In the model depicted here, the Ala 24 and Asp 61 side chains are positioned within the center of a four-helix bundle formed by the front and back face of two adjacent monomers (Fig. 3E). The interchange of the essential carboxyl from one position to another can be c Extent of cross-link formation was as follows: M, high yield multimers (a ladder of subunit c cross-links as seen on the immunoblot); m, relatively low yield multimers; t, low yield trimer; D, dimer; d, low yield dimer. See Fig. 7.   FIG. 7. Immunoblot analysis of double Cys substitutions in helix 1 and helix 2. Membrane vesicles were treated with (ϩ) and without (Ϫ) CuP, and 25 g of protein was separated on a 12-15% polyacrylamide gradient gel and electroblotted. The blot was probed with antiserum specific to subunit c. The blot is presented and marked as described for Fig. 2. rationalized in such an arrangement. The model can also be used to rationalize the position of essential liganding residues in subunit c of the Na ϩ -translocating F 1 F 0 ATPases and a Li ϩ binding variant of the E. coli F 1 F 0 ATPase. Kaim et al. (17) have shown that residues Gln 32 , Glu 65 , and Ser 66 are essential for Na ϩ binding in the P. modestum enzyme, i.e. residues at positions equivalent to residues 28, 61, and 62 in E. coli. The conserved residues in P. modestum (Gln, Glu, and Thr) have been identified at equivalent positions within the transmembrane helices of subunit c in the Na ϩ -F 1 F 0 ATPase of Acetobacterium woodii (42). In both enzymes, Pro is found at a position equivalent to Ala 24 in E. coli. In the model depicted here, each of these side chains points toward the center of the four-helix bundle formed by neighboring subunits. The model also explains why VD 61 AI 3 AE 61 S(G/A) substitutions in E. coli subunit c enables Li ϩ binding (18); i.e. the Ser 62 hydroxyl can serve as a liganded group opposite the Glu 61 carboxyl.
The F 1 F 0 ATP synthases share many structural similarities with a family of so called vacuolar or V 1 V 0 ATPases. The V 1 V 0 ATPases function as primary proton pumps to acidify intracellular organelles and extracellular spaces (43,44). The subunit c of V 1 V 0 H ϩ -ATPases is approximately twice the size of its counterpart in the F 1 F 0 ATPases and appears to have evolved by duplication of a progenitor gene (45). Four transmembrane helices are predicted. Helices 1 and 3 and helices 2 and 4 of this larger subunit c show homology with helices 1 and 2 of the subunit c of the F 1 F 0 ATP synthases, respectively. Helices 1 and 3 of V-type subunit c show a striking enrichment and sequence conservation of Gly and Ala residues, similar to that seen in F-type subunit c. A model has been proposed for this multicopy subunit based on electron microscopy image analysis and chemical modification (46 -48). In the model, six subunits c, each folding as a bundle of four ␣-helices, come together to form a hexameric complex with a central pore. The model does differ in critical respects from the one presented here. However, the prediction that helix 1 lines the pore of the oligomer (48), based upon Cys substitution analysis and chemical modification, is borne out in E. coli subunit c model. It seems likely that the arrangement of the V-type subunit c will be like that proposed here, with helices 1 and 3 lining the inner ring and helices 2 and 4 lining the outer. The loss of the conserved carboxylate in the second helix of the V-type subunit c would effectively halve the number of proton binding sites. This loss would lower the H ϩ /ATP ratio, allowing ATP-driven H ϩ pumping to generate greater electrochemical gradients while preserving the overall structural features of the complex (49). Based on the precedent with the E. coli subunit c (41), attempts have been made to generate a functional enzyme after exchange of the essential carboxyl from helix 4 to helix 3, but they have failed. 5 The model shown here suggests that an interchange between helix 4 and helix 1 is more likely to work, with the proton binding site at the center of a four-helix bundle and helix 4 packing close to helix 1. Finally, the model would predict that in the Na ϩ -translocating V 1 V 0 ATPase of Enterococcus hirae (50), the Na ϩ -liganding groups should fall between helix 4 and helix 1. The Na ϩ -liganding residue Gln 32 in P. modestum appears to be replaced by Ser 30 in a conserved Glyand Ala-rich sequence (32).
The arrangement of subunit c presented here provides insights and limitations on how proton movement can be coupled to ATP synthesis/hydrolysis. The coupling takes place by rotation of the ␥-subunit within the core of the ␣ 3 ␤ 3 subunit (5-7). This has led to proposals that the oligomeric ring of subunit c rotates past a static, peripherally located subunit a in the membrane sector (5,22,23,51). In the models, the protontranslocating carboxyl is located on the outside of the ring, and upon protonation it is envisioned as moving into the hydrocarbon region of the lipid bilayer. In the ring arrangement shown here, the proton or ion would be coordinated in the center of the four-helix bundle rather than being exposed directly to the lipid milieu. Subunit a would in some way be expected to cause an opening and closing of the four-helix bundle with resultant ion uptake and release. We should note that it is still unproven that the subunit c oligomer rotates in the membrane. Alternatively, others have suggested that the membrane sector may remain fixed relative to the ␣ and ␤ subunits in F 1 (15,48). In such a model, only subunit ␥ and the associated subunit ⑀ are predicted to move at the polar loop surface of subunit c. Conformational changes would be relayed from the site of ion binding/release to the polar loop of subunit c, and conformational changes in the polar loop then drive movement of the ␥⑀ complex from loop to loop in a circular fashion. The key question obviously remaining is whether rotations take place within the membrane sector.