Aqueous access channels in subunit a of rotary ATP synthase.

The role of subunit a in proton translocation by the Escherichia coli F(1)F(o) ATP synthase is poorly understood. In the membrane-bound F(o) sector of the enzyme, H(+) binding and release occurs at Asp(61) in the middle of the second transmembrane helix (TMH) of subunit c. Protons are thought to reach Asp(61) via an aqueous access pathway formed at least in part by one or more of the five TMHs of subunit a. In this report, we have substituted Cys into a 19-residue span of the fourth TMH of subunit a and used chemical modification to obtain information about the aqueous accessibility of residues along this helix. Residues 206, 210, and 214 are N-ethylmaleimide-accessible from the cytoplasmic side of the membrane and may lie on the H(+) transport route. Residues 215 and 218 on TMH4, as well as residue 245 on TMH5, are Ag(+)-accessible but N-ethylmaleimide-inaccessible and may form part of an aqueous pocket extending from Asp(61) of subunit c to the periplasmic surface.

H ϩ -transporting F 1 F o ATP synthases consist of two structurally and functionally distinct sectors termed F 1 and F o (1). In the intact enzyme, ATP synthesis or hydrolysis takes place in the F 1 sector and is coupled to active H ϩ transport through the F o sector. Structurally similar F 1 F o ATP synthases are present in mitochondria, chloroplasts, and most eubacteria (1). The F 1 sector lies at the surface of the membrane and in Escherichia coli consists of five subunits in an ␣ 3 ␤ 3 ␥ 1 ␦ 1 ⑀ 1 stoichiometry. The F o sector spans the membrane and in E. coli consists of three subunits in an a 1 b 2 c 10 stoichiometry (2). The structures of several types of F 1 have been solved by x-ray crystallography (3)(4)(5)(6)(7)(8). In the case of the bovine mitochondrial enzyme crystallized in the presence of Mg 2ϩ and adenosine phosphates, the three ␣ and three ␤ subunits alternate around the central ␥ subunit, with subunit ␥ interacting asymmetrically with the three catalytic sites formed at the ␣␤ interface (3)(4)(5). In the widely accepted binding change mechanism for ATP synthesis, the alternate tight binding of ADP and P i and subsequent release of product ATP are mediated by ␥ subunit rotation between the alternating catalytic sites (9 -11). Rotation of the ␥ subunit during ATP hydrolysis was demonstrated by attaching an actin filament to an immobilized ␣ 3 ␤ 3 complex (12,13). In the complete membranous enzyme, the rotation of subunit ␥ is proposed to be driven by H ϩ transport-coupled rotation of a connected ring of c subunits in the F o sector of the enzyme, which extend through the lipid bilayer and maintain a fixed linkage with the ␥ subunit. Rotation of the c ring was also demonstrated using the filament rotation assay (14,15). The structure of monomeric subunit c has been solved by NMR in a membrane mimetic solvent mixture (16), and the structure of the oligomeric c 10 ring was predicted from this structure and cross-linking constraints (17,18). The proposed subunit packing is now supported by a 3.9 Å x-ray diffraction map of an F 1 c 10 subcomplex from yeast mitochondria (19). The c subunit spans the membrane as a hairpin of two ␣-helices and in the case of E. coli contains the essential aspartyl 61 residue at the center of the second TMH. 1 Asp 61 is thought to undergo protonation and deprotonation as each subunit of the oligomeric ring moves past a stationary subunit a. Subunit a is believed to provide access channels to the proton-binding Asp 61 residue, but the actual proton translocation pathway remains to be defined (20 -23).
The structure of subunit a and its role in H ϩ translocation are poorly defined. Subunit a is known to fold with five TMHs (24 -26) with aTMH4 packing in parallel to cTMH2, i.e. the helix to which Asp 61 is anchored (27). The interaction of the conserved Arg 210 residue in aTMH4 with cTMH2 is thought to be critical during the deprotonation-protonation cycle of cAsp 61 (23, 28 -32). The predicted aTMH4/cTMH2 interactions are in accord with second site revertant analysis (33), and cross-link analysis has confirmed closest neighbor proximity of cTMH2 with aTMH4 over a span of 19 amino acid residues (27). Both modeling and cross-linking experiments indicate that helix 2 of subunit c should be packed on the outside of the ring (17,34). Electron microscopic studies support the positioning of subunit a and the two b subunits at the periphery of the c ring (35)(36)(37).
The chemical labeling of cysteine side chains introduced by site-directed mutagenesis has been used as a means of mapping aqueous accessible regions on several membrane proteins (38 -51). Several reagents have been used to modify the genetically introduced Cys to determine accessibility, including NEM (38,(41)(42)(43)(44), MTS reagents (40, 48 -50), and Ag ϩ (45)(46)(47)(48). Modification of Cys by these reagents depends upon ionization of the Cys sulfhydryl to its thiolate form (52)(53)(54)(55), and this is expected to occur preferentially in an aqueous environment (38,39,41). In this report, a span of 19 residues in aTMH4 were replaced with Cys and tested for accessibility to water-soluble reagents NEM and Ag ϩ . We found that residues 206, 210, and 214 are NEM-accessible from the cytoplasmic side of the membrane. In contrast, residues 215 and 218 on helix 4 and residue 245 on helix 5 form an Ag ϩ -accessible but NEM-inaccessible pocket bridging aTMH4 and aTMH5. This pocket may form part of the aqueous access pathway extending from Asp 61 to the periplasmic surface.

EXPERIMENTAL PROCEDURES
Construction of Cys-substituted Mutants-Cysteine substitutions were introduced by a two-step PCR method using a synthetic oligonucleotide that contained the codon change and two wild type primers (56). Most substitutions had already been generated in this lab (27) and were transferred to a plasmid containing the entire unc (atp) operon, in which all endogenous Cys had been substituted by Ala (57), and a hexahistidine tag on the C terminus of subunit a (24), using two unique BamHI sites that lie within the subunit a gene (nucleotide positions 1110 and 1727) (58). The presence of the mutation was confirmed by sequencing the cloned fragment through the ligation junctions. For these studies, the plasmids were transformed into JWP292 (2), a strain with a chromosomal deletion of the entire unc operon. The aH245C/ D119H mutant described in Fig. 12 was constructed in a pDF163-like plasmid (59). Plasmid pDF163 contains the HindIII (870) to SphI (3216) fragment of unc DNA cloned between these sites in plasmid pBR322 and encodes subunits a, b, c, and ␥. For this experiment, the plasmid was transformed into strain JWP109 (27), a strain with a chromosomal deletion of unc genes coding subunits a, b, c, and ␥.
Comparative Growth Studies-The Cys-carrying strains were plated on glucose-containing minimal medium with 0.1 mg/ml ampicillin. Single colonies were then tested for growth on succinate-containing minimal plates with scoring for growth at 72 h, as well as growth yield in liquid minimal medium containing 0.04% glucose.
Membrane Preparation-Plasmid transformants of strain JWP292 were grown in M63 minimal medium containing 0.6% glucose, 2 mg/ liter thiamine, 0.2 mM uracil, 0.2 mM L-arginine, 0.02 mM dihydroxybenzoic acid, and 0.1 mg/ml ampicillin, supplemented with 10% LB medium, and harvested in the late exponential phase of growth (2). The cells were suspended in TMG-acetate buffer (50 mM Tris acetate, 5 mM magnesium acetate, 10% glycerol, pH 7.5) containing 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, and 0.1 mg/ml of DNase I and disrupted by passage through a French press at 124 MPa and membranes prepared as described (60). Following differential centrifugation and washing, the final membrane preparation was suspended in TMG-acetate buffer. The protein concentrations were determined using a modified Lowry assay (61).
ATP-driven Quenching of ACMA Fluorescence-The membranes were suspended in 3.2 ml of the HMK buffers described below, containing either chloride or nitrate as the counter ion. ACMA was added to 0.3 g/ml final concentration, and 30 l of 0.1 M ATP, pH 7.0 was added to initiate quenching of fluorescence. The reaction was terminated by addition of 8 l of 288 M nigericin (final concentration, 0.5 g/ml). The level of fluorescence obtained after the addition of nigericin was normalized to 100% in calculating the percentage of quenching caused by ATP-driven proton pumping. For NEM treatment, 80 l of membranes at 10 mg/ml in TMG-acetate were incubated at room temperature with 5 mM NEM for 15 min prior to dilution into HMK-chloride buffer (10 mM Hepes-KOH, 5 mM MgCl 2 , and 300 mM KCl, pH 7.5). For AgNO 3 treatment, 160 l of membranes at 10 mg/ml were suspended in HMKnitrate buffer (10 mM Hepes-KOH, 1 mM Mg(NO 3 ) 2 , and 10 mM KNO 3 , pH 7.5) containing 40 M AgNO 3 and incubated at room temperature for 15 min before carrying out the quenching assay. Identical results were observed using silver acetate instead of AgNO 3 . As a control, the membranes were also pretreated at 10 mg/ml in TMG-acetate buffer with 5 mM NEM and then diluted into HMK-nitrate buffer before carrying out the same quenching assay.
NADH-driven Quenching of Quinacrine Fluorescence in Stripped Membrane Vesicles-Membranes in TMG-acetate were centrifuged and resuspended in TEG buffer (1 mM Tris acetate, 0.5 mM EDTA, 10% glycerol, pH 8.0) and incubated for 30 min at 30°C. The membranes were centrifuged again, washed once with TEG buffer, and then resuspended in TMG-acetate buffer at 10 mg/ml. The stripped membrane vesicle suspension (480 g) was added to 3.2 ml of HMK-chloride buffer, and following addition of quinacrine to 1.875 g/ml (final concentration), 16 l of 10 mM NADH was added to initiate quenching of fluorescence. The reaction was terminated by addition of 8 l of 288 M nigericin (final concentration, 0.5 g/ml). For NEM treatment, 480 g of stripped membranes at 10 mg/ml were incubated at room temperature with 5 mM NEM for 15 min prior to dilution in HMK-chloride buffer. For dicyclohexylcarbodiimide treatment, diluted stripped membranes at 150 g/ml in HMK-chloride buffer were treated with 18.75 M dicyclohexylcarbodiimide at room temperature for 15 min prior to the quenching measurement.
[ 14 C]N-Ethylmaleimide Labeling Studies-[ 14 C]NEM (PerkinElmer Life Sciences; 34.2 mCi/mmol) in pentane was added to TMG-acetate, and the pentane was evaporated by blowing a stream of argon over the solution. The resulting aqueous solution of NEM was then added to 3 mg of membrane vesicles in TMG-acetate such that the final concentration of membranes was 10 mg/ml and the final concentration of [ 14 C]NEM was 1 mM. The mixture was incubated at room temperature for 1 h, after which the reaction was stopped by the addition of ␤-mercaptoethanol to 10 mM. SDS was added to 1%, and unlabeled NEM was added to 20 mM. His-tagged subunit a was purified from the SDSsolubilized membranes as follows. After 10 min at room temperature, 660 l of binding buffer (50 mM Tris-HCl, 0.3 M NaCl, 0.1% SDS, pH 8) and 40 l of nickel-nitrilotriacetic acid-agarose (Qiagen) were added, and the solution was incubated at room temperature with mixing for 1 h. The beads were washed twice with 1 ml of wash buffer (50 mM Tris-HCl, 0.3 M NaCl, 1% SDS, pH 8), then 100 l of elution buffer (62.5 mM Tris-HCl, 20 mM EDTA, 10% glycerol, 2% SDS, 0.01% bromphenol blue, 2.5% ␤-mercaptoethanol, pH 6.75) was added, and the sample was boiled for 3 min prior to loading 30 l on a 12% Tris-Tricine gel (62). The resulting gel was stained with Coomassie Blue and dried. The dried gel was exposed to a storage phosphor screen and scanned with a Phospho-rImager (Molecular Dynamics) to quantitate radioactivity incorporated into subunit a. To determine the amount of protein in each band, the stained, dried gel was scanned using a flatbed scanner linked to a Macintosh computer using Epson Twain 5 software. The scan was quantitated using the public domain NIH Image program. 2 Alternatively, the band containing subunit a was excised from the gel following Coomassie Blue staining, and the gel slice was incubated with 12% H 2 O 2 at 90°C for 5 h to dissolve the gel. The amount of radioactivity incorporated into subunit a was determined by scintillation counting. The scintillation counting experiments confirmed the general accuracy of the PhosphorImager analysis.

RESULTS
Cys Substitutions in TMH4 of Subunit a-In this study, Cys substitutions in the fourth TMH of subunit a were transferred into a His-tagged version of subunit a in a plasmid that coded the entire unc (atp) operon in which all the endogenous Cys in F 1 and F o had been substituted by Ala (57). These plasmids were transformed into JWP292, a strain with a chromosomal deletion of the entire unc operon. The growth of transformant strains was tested on glucose and succinate-containing minimal medium. The Cys substitutions in subunit a reported here have little effect on growth, with the exception of R210C, G213C, and E219C, which grew poorly or not at all on succinate and exhibited low growth yields with glucose as a carbon source (Table I).
Sulfhydryl-specific Reagents Inhibit Function of the aN214C Mutant-We wished to examine the aqueous accessibility of residue 214. Previously, Jiang and Fillingame (27) demonstrated that aN214C forms a cross-link with cM65C and cA62C, indicating that this residue may be in close proximity to the proton-binding Asp 61 residue of subunit c. The capacity of several sulfhydryl-specific reagents to inhibit ATP-driven quenching of ACMA fluorescence by aN214C inverted membrane vesicles was tested. NEM and nonpolar MTS reagents, i.e. ethylamino-MTS and carboxyethyl-MTS, inhibited quenching in aN214C membranes ( Fig. 1) but not in control membranes lacking the aN214C (data not shown). The larger, charged MTS reagents, i.e. trimethylammonium-MTS and sulfonatoethyl-MTS, were not inhibitory. NEM was found to inhibit quenching in aN214C membranes maximally at 5 mM but also quite significantly at 1 mM (Fig. 2). The [ 14 C]NEM labeling studies described below were carried out with 1 mM NEM. The inhibition of ATP-driven quenching observed here may be due to a direct block in proton transport through F o because NEM also blocked passive, F o -mediated proton translocation by stripped membrane vesicles (Fig. 3). This is indicated by the increase in NADH-driven quenching of quinacrine fluorescence after NEM treatment, which can be attributed to a decrease in the proton leakiness of the stripped membrane vesicles.

NEM Inhibition of Quenching with Other aTMH4
Cys Substitutions-The ⌬unc strain was transformed with the set of plasmids containing single Cys substitutions in aTMH4. Inside-out membrane vesicles were prepared from these strains. The membranes were treated with 5 mM NEM, and ATP-dependent quenching of ACMA fluorescence was tested for each substitution. Representative quenching traces for several mutants are presented in Fig. 4. NEM inhibited quenching most strikingly with two Cys substitutions, aS206C (Fig. 4) and aN214C (Fig. 1), whereas proton translocation activity remained largely unchanged in the other 17 mutants tested (Table II).

Characterization of NEM Reaction with Residues aS206C
and aN214C-In the models of subunit a discussed elsewhere (24 -27, 32), residue 206 is located on the cytoplasmic face of the membrane, whereas residue 214 is located near the center of the lipid bilayer. Because NEM reacts preferentially with the ionized form of Cys (52,53), the pH dependence of NEM inhibition of the two residues was examined. When S206C membranes were treated with NEM at pH 7.5, inhibition of quenching was observed. However, NEM treatment at pH 7.0 resulted in no inhibition of quenching and indicated a higher level of protonation of the sulfhydryl side chain at this pH (Fig. 5A). In contrast, quenching with N214C membranes was inhibited by NEM treatment at either pH, indicating that this Cys residue is subject to ionization at pH 7.0 (Fig. 5B). The Cys 214 residue must therefore have an unusually low pK a and be in an unexpectedly hydrophilic environment for a residue centered in the middle of the membrane. The Cys 214 side chain may be close to the essential aR210 residue, which could provide charge neutralization, leading to the low pK a .
[ 14     aTMH4 mutants were treated with 1 mM [ 14 C]NEM for 1 h to determine the aqueous accessibility of each residue. The hexahistidine-tagged subunit a was then purified by Ni 2ϩ affinity chromatography, and the amount of label incorporated into subunit a was quantified by scintillation counting and Phos-phorImager analysis (Fig. 6). Cys substitutions at positions 206, 210, and 214 were strongly labeled with NEM, indicating that this hydrophilic reagent has access to these residues, possibly via the same aqueous channel by which protons move through the enzyme.
Silver as a Probe of Aqueous Accessibility to aTMH4 -Silver has been used to map aqueous pores of several membrane proteins (45)(46)(47)(48), and it was shown to inhibit the Na ϩ ,K ϩ -ATPase (63). Silver ion has an ionic radius of 1.26 Å, which is close to that of Na ϩ (0.97 Å) and H 3 O ϩ (1.54 Å) (64). It forms a covalent bond with sulfhydryl groups of Cys (45,55,65). In our initial experiments with aN214C membranes, Ag ϩ treatment caused an inhibition of ATP-driven quenching of ACMA fluorescence, the extent of inhibition varying with the amount of Ag ϩ used (Fig. 7), whereas membranes lacking the Cys substitution were unaffected by Ag ϩ treatment. The extent of inhibition also proved to depend upon the amount of membrane added to the cuvette (Fig. 8). The amount of AgNO 3 needed for a given extent of inhibition increased proportionally with the amount of membrane present and suggested that the silver

FIG. 4. Differing sensitivity of Cys substitutions in aTMH4 to NEM inhibition. Membranes from various aTMH4
Cys mutants were treated at 10 mg/ml in TMG-acetate with 5 mM NEM for 15 min prior to the quenching measurement as described in the legend to Fig. 1. A,  aS206C membranes; B, aR210C membranes; C, aM215C membranes; D, aE219C membranes. In all panels, trace 1 represents no treatment of the membranes, whereas trace 2 represents membranes that have been treated with 5 mM NEM.  may partition to the membrane phase rather than remaining soluble in the aqueous phase. Following AgNO 3 treatment, the sample was centrifuged, and atomic absorption analysis was performed on the membrane pellet and the buffer. We found that 80% of the AgNO 3 added to the solution was associated with the membrane pellet, further supporting the idea that silver complexes with the membrane.
Ag ϩ Inhibition of Quenching in aTMH4 Cys Substitutions-Cys-substituted aTMH4 membranes were treated with 40 M AgNO 3 in chloride-free assay buffer, and ATP-driven quenching of ACMA fluorescence was measured. As with NEM treatment, Ag ϩ treatment inhibited quenching by aS206C and aN214C membranes. Additionally, and in striking contrast, Ag ϩ treatment also resulted in dramatic inhibition with sev-eral mutants that were NEM insensitive. For example, the aM215C and aG218C mutants were very sensitive to inhibition by Ag ϩ but were NEM-insensitive (Fig. 9). Other Cys substitutions showing a lesser sensitivity to Ag ϩ but still Ͼ50% inhibition include 207, 213, 216, 217, 219, and 220 (Fig. 10). The Ag ϩ inhibition observed in several of these mutants, e.g. 213 and 219, may be due in part to the generally feeble quenching response (Table II).
Characterization of Ag ϩ Reaction with Residues aS206C and aN214C-The ability of DTT to reverse the inhibition of quenching caused by treatment of membranes with AgNO 3 was also explored (Fig. 11). We found that the addition of DTT rapidly reversed AgNO 3 inhibition with aS206C membranes. In contrast, inhibition was very slowly reversed with aN214C membranes, indicating that the hydrophilic DTT may have more direct access to the cytoplasmically located S206C than the membrane-buried Cys 214 residue.
Involvement of aTMH5 in Proton Conductance-In the models of subunit a suggested previously (20, 22, 24 -26, 32), Gly 218 in TMH4 is thought to be adjacent to His 245 in TMH5. This model is supported by several second site suppressor pairs between TMHs, including the aG218D mutation, which was corrected by a second mutation, aH245G (66). We wondered whether the Ag ϩ -accessible region of aTMH4 was also bounded by aTMH5. The aH245C substitution, located in aTMH5, has been characterized as a nonfunctional mutant. However, partial function can be restored by the introduction of a second mutation, aD119H (24). Membranes carrying the aH245C/ D119H double mutant were treated with silver, and ATPdriven quenching of ACMA fluorescence was measured (Fig.  12). Ag ϩ inhibits proton pumping in this mutant, suggesting that several helices of subunit a may be involved in the formation of an Ag ϩ -accessible cavity extending into the interior of subunit a. The boundaries of this cavity will be investigated further in the future. DISCUSSION In this paper, we report on the aqueous accessibility of cysteine residues substituted into the fourth transmembrane helix in subunit a of the rotary ATP synthase. aTMH4 is thought to interact with Asp 61 of subunit c during proton translocation and may play a part in forming an aqueous access channel to cAsp 61 from one or both sides of the membrane. The proximity

FIG. 8. Concentration of AgNO 3 required for maximal inhibition of ATP-driven quenching increases in proportion to the membrane concentration.
Varying volumes of N214C membranes (80, 160, or 320 l) at 10 mg/ml in TMG-acetate were diluted into 3.2 ml HMK-nitrate buffer (pH 7.5), and AgNO 3 was added to the concentration indicated. Samples were incubated for 15 min at room temperature prior to the quenching measurement. The relative inhibition brought about by AgNO 3 was plotted as a function of AgNO 3 concentration for each set. ࡗ, 0.8 mg of membranes; f, 1.6 mg of membranes; OE, 3.2 mg of membranes. of aTMH4 to cTMH2 was initially postulated on the basis of second site suppressor analysis (33) and is now supported by intermolecular Cys-Cys cross-links between cTMH2 and aTMH4 (27). Aqueous access to Asp 61 of subunit c had been postulated previously based upon the pH-sensitive function of the cA24D mutant (67) and is further supported by the discovery that simultaneous mutation of three residues surrounding Asp 61 makes the enzyme sensitive to inhibition by Li ϩ (68). FIG. 9. ATP-driven ACMA quenching by Cys 215 and Cys 218 membranes is inhibited by Ag ؉ but insensitive to inhibition by NEM. M215C (A) and G218C (B) membranes at 10 mg/ml in TMGacetate were diluted into 3.2 ml of HMK-nitrate buffer, and AgNO 3 was added to 40 M, or NEM was added at 5 mM to membranes in TMGacetate buffer prior to dilution into HMK-nitrate buffer. Following incubation for 15 min at room temperature, the quenching measurement was made as described in the legend to A 160-l aliquot of membranes at 10 mg/ml in TMGacetate was diluted into 3.2 ml of HMKnitrate buffer, and AgNO 3 was added to 40 M for 15 min, or NEM was added to 5 mM for 15 min prior to dilution into HMKnitrate, as described in the legend to Fig.  9. The results are presented as the ratios of quenching in the presence of Ag ϩ or NEM to the quenching in the absence of a reagent. The gray bars represent the quenching ratio Ϯ Ag ϩ treatment, whereas the black bars represent the quenching ratio Ϯ NEM treatment. Each bar represents the average ratio from n Ն 2 determinations Ϯ S.D. Additionally, the analogous enzyme from Propiogenium modestum, with a very homologous subunit c, alternatively transports Na ϩ , Li ϩ , or H ϩ (69,70). Thus, it seems likely that these various ions gain access to the membrane-embedded carboxyl of subunit c via a water-filled channel. We have substituted cysteine over the length of aTMH4 and tested the susceptibility of each substitution to modification with water-soluble, thiolmodifying reagents. The approach has been used previously to define surfaces of membrane proteins with aqueous accessibility (38 -51).
Reactivity of Cys Substitutions with NEM, Ag ϩ , and MTS Reagents-The ionized sulfhydryl group of cysteine is the form that preferentially reacts with the thiol-specific reagents used here. For example, the reactivity of MTS reagents with the thiolate is preferred by a factor of 10 9 over reaction with a nonionized thiol group (54). Similarly, NEM (52,53) and Ag ϩ (45,55) react preferentially with ionized thiolates rather than with neutral thiols. The differential reactivity of substituted cysteines can thus provide information about the ionization state of different residues, which in most cases should be related to aqueous accessibility, i.e. the interpretation given in similar studies of other membrane proteins (38,39,41). The means by which NEM penetrates the membrane to react with Cys residues is not certain. It may be sufficiently lipid-soluble to gain access to transmembrane Cys via the hydrophobic phase of the membrane. However, modification should only be observed with those residues subject to ionization. It is also conceivable that uncharged MTS reagents could access reactive Cys residues via the lipid phase, although reactivity would again depend upon the ionization state of the sulfhydryl group. It is of interest that Ag ϩ reacts with transmembrane residues that are NEM-insensitive. This may indicate that NEM-sensitive residues need to be bounded by larger aqueous cavities, sufficient in size to accommodate the bulkier NEM molecule.
NEM Reactive TMH4 Substitutions-Residues 206, 210, and 214 on aTMH4 are preferentially modified with NEM, as indicated by direct labeling with [ 14 C]NEM and also by inhibition of ATP-driven quenching in the case of the 206 and 214 mutants. The radioactive labeling studies were carried out to address the possibility that residues might be modified with NEM without effect on function. Additionally, we were able to examine the labeling of a Cys at position 210 in a mutant that is nonfunctional in proton transport. These three residues fall on one face of an ␣-helix (Fig. 13), and it seems likely that NEM FIG. 13. Location of NEM-and Ag ؉ -sensitive residues in a hypothetical model of subunit a. Subunit a modeled by Rastogi and Girvin (Ref. 22; Protein Data Bank code 1c17). Positions of NEMreactive Cys residues are shown in red; the Arg 210 side chain is also indicated. Ag ϩ -sensitive, NEM-insensitive residues are shown in yellow. The model was constructed by standard NMR structure calculation methods using ␣-helical backbone constraints for TMHs 2-5 and five helix-helix contacts suggested from second site suppressor analysis. In topological models (24, 32) the loops at the top of the structure are thought to extend into the cytoplasm, and loop 4,5 is thought to extend into the periplasm, placing residue 214 in aTMH4 at the center of the membrane. The figure is drawn from the program MOLMOL (77). a In most cases ATP-driven proton pumping was also measured, and the decrease in activity correlates with the growth on succinate parameter. modification results in a block to proton transport through this region of the protein. Inhibition of proton translocation by attachment of the ethylmaleimide group is consistent with evidence from mutagenesis studies that bulky substitutions are not as well tolerated as smaller ones at these positions (Table III). The fact that the three residues found to label with NEM line one face of aTMH4 suggests that NEM could be gaining access to these residues via an aqueous pathway formed at this face of an ␣-helix.
The Cys 206 residue appears to titrate as the pH of the medium is lowered from 7.5 to 7.0, as indicated by the dramatic decrease in NEM reactivity, suggesting that this residue is accessible to the bulk solvent. The solvent accessibility of this residue is also supported by the rapid reversal of Ag ϩ inhibition by dithiothreitol. On the other hand, the reactivity of Cys 214 is unaffected by lowering the pH of the medium to 7.0, suggesting that this residue remains ionized at a lower pH than Cys 206 , even though topological analysis would place residue 214 in the middle of the membrane. Despite its low apparent pK a , Cys 214 does not appear to be generally accessible to the aqueous medium based upon the slow reversal of Ag ϩ inhibition by DTT. We suggest that the low pK a of Cys 214 at the center of the membrane may be due to salt bridge formation with the proximal Arg 210 residue.
Silver Used to Probe aTMH4 Cys Substitutions-Silver has been used previously as an irreversible covalent modifier of Cys introduced into several membrane proteins (45)(46)(47)(48). The irreversibility of the Ag ϩ inhibition under the conditions used here is indicated by an experiment where Ag ϩ pretreatment of membrane vesicles protected Cys 215 from becoming labeled with [ 14 C]NEM after solubilization of the membrane with SDS (experiment not shown). Further, if membranes are treated with silver and centrifuged to remove the assay buffer, inhibition is still observed on resuspension in Ag ϩ -free buffer with Cl Ϫ present to precipitate any free silver released to solution. As would be expected (65), the reaction and inhibition can in some cases be reversed by competing sulfhydryl reagents such as DTT, e.g. in the case of the aS206C substitution. It is not clear from our studies exactly how silver gains access to these sulfhydryls. In this experimental system, we have shown by atomic absorption analysis that 80% of silver added to membrane vesicles becomes complexed with the membrane. Conceivably, the silver may be transported to the interior of the membrane vesicles, where it gains access to subunit a. Alternatively, the silver may bind to other membrane proteins or sites of unsaturation on fatty acid tails (71,72). To test whether other proteins in the membrane were transporting silver or were necessary for silver absorption by the membrane, F 1 F o was purified and reconstituted into liposomes and tested for Ag ϩ inhibition of ATP-driven proton pumping by Ag ϩ . The Cys 214 and Cys 215 mutant enzymes still exhibited inhibition by Ag ϩ , suggesting that the silver may gain access to these cysteines without the aid of other membrane proteins and that the inhibitory Ag ϩ does not need to be transported to the interior of these vesicles by an inner membrane Ag ϩ transport system.
As seen in Fig. 12, aH245C/D119H shows inhibition of ATPdriven quenching when treated with silver. This is the first direct evidence that the aqueous access pathway to cAsp 61 may involve more than one helix of subunit a. Looking at the silversensitive residues highlighted on the model of subunit a in Fig.  13, it is certainly possible that these residues form a pocket that is accessed from an aqueous channel extending to the cytoplasmic side of the membrane. The aqueous pathway to the cytoplasm appears to include NEM-sensitive residues 206, 210, and 214 on one ␣-helical face of aTMH4. To the periplasmic side of aAsn 214 , residues 215-220 are characterized here as being NEM-inaccessible but Ag ϩ -sensitive, with residues 215 and 218 showing the greatest silver sensitivity. Residues 215 and 218 would fall on the ␣-helical face of aTMH4 opposite to residues 206, 210, and 214. If an aqueous pocket bridges these residues and residue 245 in aTMH5, then aTMH4 may have to swivel to gate access between the periplasmic pocket and Asp 61 of subunit c. As we have discussed elsewhere (76), such swiveling may be coupled mechanically to other helical movements that drive stepwise rotation of the c ring.