Membrane Topology of Subunit a of the F1F0 ATP Synthase as Determined by Labeling of Unique Cysteine Residues*

The membrane topology of the a subunit of the F1F0 ATP synthase from Escherichia coli has been probed by surface labeling using 3-(N-maleimidylpropionyl) biocytin. Subunit a has no naturally occurring cysteine residues, allowing unique cysteines to be introduced at the following positions: 8, 24, 27, 69, 89, 128, 131, 172, 176, 196, 238, 241, and 277 (following the COOH-terminal 271 and a hexahistidine tag). None of the single mutations affected the function of the enzyme, as judged by growth on succinate minimal medium. Membrane vesicles with an exposed cytoplasmic surface were prepared using a French pressure cell. Before labeling, the membranes were incubated with or without a highly charged sulfhydryl reagent, 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid. After labeling with the less polar biotin maleimide, the samples were solubilized with octyl glucoside/cholate and the subunit a was purified via the oligohistidine at its COOH terminus using immobilized nickel chromatography. The purified samples were electrophoresed and transferred to nitrocellulose for detection by avidin conjugated to alkaline phosphatase. Results indicated cytoplasmic accessibility for residues 69, 172, 176, and 277 and periplasmic accessibility for residues 8, 24, 27, and 131. On the basis of these and earlier results, a transmembrane topology for the subunit a is proposed.

The membrane topology of the a subunit of the F 1 F 0 ATP synthase from Escherichia coli has been probed by surface labeling using 3-(N-maleimidylpropionyl) biocytin. Subunit a has no naturally occurring cysteine residues, allowing unique cysteines to be introduced at the following positions: 8, 24, 27, 69, 89, 128, 131, 172, 176, 196, 238, 241, and 277 (following the COOH-terminal 271 and a hexahistidine tag). None of the single mutations affected the function of the enzyme, as judged by growth on succinate minimal medium. Membrane vesicles with an exposed cytoplasmic surface were prepared using a French pressure cell. Before labeling, the membranes were incubated with or without a highly charged sulfhydryl reagent, 4-acetamido-4-maleimidylstilbene-2,2disulfonic acid. After labeling with the less polar biotin maleimide, the samples were solubilized with octyl glucoside/cholate and the subunit a was purified via the oligohistidine at its COOH terminus using immobilized nickel chromatography. The purified samples were electrophoresed and transferred to nitrocellulose for detection by avidin conjugated to alkaline phosphatase. Results indicated cytoplasmic accessibility for residues 69, 172, 176, and 277 and periplasmic accessibility for residues 8, 24, 27, and 131. On the basis of these and earlier results, a transmembrane topology for the subunit a is proposed.
The F 1 F 0 ATP synthase from Escherichia coli is typical of the ATP synthases found in mitochondria, chloroplasts, and many other bacteria (for recent reviews, see Refs. [1][2][3][4]. It comprises an F 1 complex, which contains the nucleotide-binding subunits involved in catalysis, and an F 0 complex, which conducts protons across the membrane. The enzyme from E. coli appears to be a minimal form of the ATP synthase, with eight essential subunits. Five different subunits are found in F 1 : ␣, ␤, ␥, ␦, and ⑀, in a stoichiometry of 3:3:1:1:1. Three different subunits, named a, b, and c form F 0 with a likely stoichiometry of 1:2: 9 -12 (5).
The mechanism by which an electrochemical proton gradient across the membrane drives ATP synthesis is slowly emerging (6), due in large part to success in recent years in obtaining structural information about the subunits of the enzyme. The crystallization of F 1 from bovine mitochondria (7) led to a high resolution structure of the ␣ 3 ␤ 3 hexamer, plus parts of ␥ in the central core. Electron cryomicroscopic images have also con-tributed to an understanding of subunit arrangement (8,9) and motion in F 1 subunits (10). Subsequently, the hypothesis of rotation of ␥ relative to ␣ 3 ␤ 3 (11) has been supported by studies involving engineered disulfide-cross-linking of ␤ to ␥ (12), a fluorescence technique termed "polarized absorption recovery after photobleaching" (13), and direct visualization of rotation of fluorescently labeled actin filaments covalently attached to ␥ (14). The three-dimensional structures of two other F 1 subunits from E. coli have been determined: ␦ (15) and ⑀ (16,17). Both subunits are small, two-domain proteins. The ⑀ subunit binds to ␥ through its NH 2 -terminal domain at the "base" of F 1 (18,19) and interacts with ␣ and ␤ subunits through its COOHterminal domain (20 -22). Both ␥ and ⑀ subunits have been cross-linked to c subunits (23,24). The ␦ subunit interacts with ␣ and ␤ subunits at the "top" of F 1 through its NH 2 -terminal domain (25,26), and probably with b subunits through its COOH-terminal domain (27,28). The three-dimensional structure of a monomeric subunit c has also been determined by NMR (29,30).
Lack of information about the tertiary and quaternary structure of F 0 subunits has limited progress in understanding how F 0 translocates protons and how it might drive rotation of ␥ and ⑀ subunits in F 1 . The b subunits seem to be embedded in the membrane via a span of hydrophobic amino acids at the NH 2 terminus. A truncated, soluble form of subunit b has been shown to be extended and dimeric (31). NMR studies of subunit c have confirmed the ␣-helical hairpin structure of the two predicted transmembrane spans (29,30) and also details of the essential residue Asp 61 and its local environment (32,33). Questions remain about the oligomeric structure of subunit c and how it interacts with subunits a and b. Mutagenesis has revealed that in addition to Asp 61 of subunit c, three residues in subunit a seem to be important in proton translocation: Arg 210 , Glu 219 , and His 245 (34 -38). Knowledge of the relative location of these four residues could provide much insight into proton translocation and subunit movements.
Subunit a is an extremely hydrophobic protein of 271 amino acids. It cannot be expressed at high levels in E. coli (39) and has only been purified in the presence of trichloroacetate (40). Hydropathy analysis (41) according to von Heijne (42) indicates five "certain" transmembrane spans and one "tentative" span. Because the important residues in subunit a reside in the last two predicted transmembrane spans, and the uncertain span immediately precedes this region, at least two plausible arrangements exist. Such questions of membrane topology can be addressed by gene fusion experiments (43). In the case of subunit a, two groups (44,45) have used phoA fusions to determine its membrane topology but failed to reach agreement. A more recent study used peptide-directed antibodies against polar regions of subunit a to determine membrane topology (46), but only three of the antibodies provided information.
To address the issues of how many transmembrane spans are in subunit a and how they are oriented, we have generated a collection of subunit a mutants, each with a single unique cysteine residue. These mutants can be probed with sulfhydryl reagents in oriented membrane preparations to determine the surface accessibility of different residues.

EXPERIMENTAL PROCEDURES
Materials-Restriction endonucleases and T4 DNA ligase were obtained from New England Biolabs. Materials for silver sequencing and plasmid minipreps were obtained from Promega Corp. Synthetic oligonucleotides were obtained from Operon Technologies or National Biosciences. Urea was from International Biotechnologies, Inc. MPB 1 and AMS were obtained from Molecular Probes. Nickel-nitrilotriacetic acid resin was obtained from Qiagen. Octyl glucoside was obtained from Sigma. Anti-a antibodies were provided by Dr. Karlheinz Altendorf (Osnabrü ck). Immunoblotting reagents were obtained from Bio-Rad.
Plasmids-Plasmids used for mutagenesis and expression are shown in Fig. 1. Plasmids designated "His" code for 5 additional histidine residues following the natural carboxyl-terminal His 271 of subunit a. Silent mutations have been introduced into various regions of uncB to facilitate cassette mutagenesis. While this project was under way, it was discovered that several plasmids had very poor expression: pSBV16 and its derivatives (47). It is thought that this was due to the loss of the tet promoter originally found in pACYC184. Therefore, many of the mutations were subcloned into new plasmids that contained the promoter. Plasmid pSW18His was constructed from pSBV18 (47) using the oligonucleotides that also introduced Cys 277 (see Fig. 2). Plasmid pSW19His was constructed from the 1575-bp BsaHI-EcoRI fragment of pSW18His, containing most of uncB, and the 1633-bp BsaHI-EcoRI fragment of pBJA1018 (48). Plasmid pLN6His was constructed in the following way: pLN4 was constructed from pSBV4 (49) by replacing the 222-bp BglI-DraIII region in uncB with synthetic DNA, including new restriction sites BglII and BstEII. pLN6His was then constructed from the 815-bp PvuI-EcoRI fragment of pSW18His, containing the oligohistidine coding region, and the 2475-bp PvuI-EcoRI fragment of pLN4, containing most of uncB. Plasmid pLN46His was constructed from pLN6His using the DraIII-PstI synthetic cassette, as for pBJA46 (48). Plasmid pTW1HisHA was constructed in the following way: pLN6HisHA was constructed from pLN6His by inserting the HA-oligonucleotides ( Fig. 2) in the SapI site of uncB, near the carboxyl-terminal coding region. pTW1HisHA was constructed from the 760-bp BsaHI--PvuI fragment of pSBV18, containing most of uncB, and the 2560-bp BsaHI--PvuI fragment of pLN6HisHA.
Mutagenesis-Mutations were introduced into uncB using the oligo-nucleotides shown in Fig. 2. The mutation at position 69 was made by Dr. Brian Cain (University of Florida), using M13 phage and provided to us as replicative form DNA. The others were constructed by cassette mutagenesis, using the indicated restriction sites. Mutations 24 and 27 were originally constructed in pSW18His and were transferred to pSW19His using BsaHI and EcoRI, as described above. Mutation 8 was constructed in pSW19His. Mutation 131 was constructed in pLN6His.
Mutations 169, 172, and 176 were constructed in pLN46His. Mutations 196 and 238 were constructed in pTW1HisHA, and mutations 89 and 128 were constructed in pLN6HisHA. These two plasmids produce a subunit a that is 15 amino acids longer than wild type. Growth and Expression-For expression, RH305 (uncB205, recA56, srl::Tn10, bglR, thi-1, rel-1, Hfr PO1) was used as the background strain (50). It produces an a subunit that is truncated near Pro 240 (51) and is complemented by plasmids containing a wild type uncB gene. Cultures were grown at 37°C in LB or in minimal A medium supplemented with succinate (0.2%) (52). Media were also supplemented with chloramphenicol (34 mg/liter) or tetracycline (12.5 mg/liter) as appropriate.
Preparation of Oriented Vesicles-Inside out membrane vesicles were made from a 250-ml culture in LB medium grown to A 600 ϭ 1.0. Cells were resuspended in 5 ml of 50 mM Tris-HCl containing 10 mM MgSO 4 (pH 8.0) and passed through the French press at 14,000 p.s.i. After a low speed spin to remove unbroken cells (7000 ϫ g), the supernatant was centrifuged at 50,000 rpm for 1 h in a Beckman Ti-70 rotor. To prepare stripped membrane vesicles (loss of F 1 ), the vesicles were resuspended in stripping buffer (1 mM Tris-HCl, pH 8.0, 0.5 EDTA, and 10% glycerol) and agitated at 4°C overnight. The next day, the centrifugation was repeated, and the vesicles were resuspended in 200 mM Tris-HCl (pH 8.0). For nonstripped vesicles, the samples were resuspended directly in 200 mM Tris-HCl (pH 8.0).
Chemical Labeling and Blocking-The membrane vesicles were labeled in 200 mM Tris-HCl (pH 8.0), with MPB. The reaction was stopped by adding ␤-mercaptoethanol to a final concentration of 20 mM. The vesicles were then centrifuged at 50,000 rpm for 45 min and resuspended in the original buffer. Blocking with AMS was done in the same buffer, followed by centrifugation at 50,000 rpm for 45 min. The vesicles were resuspended in the same buffer, and labeled with MPB as described above.
Purification of Subunit a-After labeling, membrane vesicles were resuspended in 100 mM Tris-HCl (pH 8.0), 1.5% octyl glucoside, 0.1% deoxycholate, 0.5% cholate, 10 mM ␤-mercaptoethanol, 10 mM imidazole, and 1% Tween 20 (53). The samples were incubated with agitation for 1 h at 4°C and centrifuged at 14,000 rpm (16,000 ϫ g) for 10 min in a microcentrifuge (1.5 ml). The supernatant was added to 0.4 ml of nickel-nitrilotriacetic acid resin that had been previously equilibrated in the extraction buffer. The mixture was incubated with agitation for 45 min at room temperature, and centrifuged for 30 s at 14,000 rpm. The resin was washed three times with 1 ml of wash buffer, consisting of equal volumes of extraction buffer and 200 mM Tris-HCl (pH 8.0). Subunit a was eluted by adding 0.25 ml of elution buffer (extraction buffer containing 1 M imidazole). The mixture was incubated at room temperature for 5 min and centrifuged (16,000 ϫ g) for 30 s. The supernatant containing the purified subunit a was collected and stored at Ϫ20°C.
MPB Detection and Immunoblotting-Samples of purified subunit a were subjected to SDS-polyacrylamide gel electrophoresis (13% acrylamide) and transferred to nitrocellulose membrane (0.2 m) using a Trans-Blot apparatus (Bio-Rad) overnight at 16 V. The nitrocellulose membrane was blocked with 5% low fat powdered milk in 20 mM Tris-HCl, 500 mM NaCl, pH 7.5 (TBS) for 1 h and rinsed with TBS/ Tween 20 (0.05% Tween 20) three times. For MPB detection, the blocked membrane was incubated with avidin-conjugated alkaline phosphatase for 2 h, rinsed three times in TBS/Tween 20, and developed with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium according to the manufacturer's instructions. For subunit a detection, the blocked nitrocellulose membrane was incubated at room temperature for 2 h with rabbit anti-a serum, diluted 1:5000. After washing five times with TBS/Tween 20, it was incubated with goat, anti-rabbit IgG-alkaline phosphatase conjugate at a dilution of 1:1000 for 2 h. After another five washings with TBS/Tween 20, color was developed as described above.

RESULTS
Unique cysteine residues were introduced into subunit a at the following positions: 8, 24, 27, 69, 89, 128, 131, 172, 176, 196, 238, and 277 (COOH terminus). The resulting plasmids were transformed into RH305 (uncB205) and tested for growth on succinate minimal medium. All grew as well as wild type, but three other mutants grew poorly and were not further analyzed: 169 and two double mutants, 24/27 and 238/241. Everted membrane vesicles were prepared using a French pressure cell from cultures of E. coli cells bearing the unique cysteine mutations. Membranes were reacted with various concentrations of MPB from 10 M to 1 mM, at temperatures from 0 to 25°C, and for various lengths of time from 10 min to 1 h. The standard conditions chosen were the mildest under which significant labeling was detectable, to minimize the penetration of the membrane by MPB. The labeling patterns were the same whether or not the membranes were stripped of F 1 . The standard conditions included reaction at 25°C for 15 min with 120 M MPB, either with or without a pretreatment with the blocking reagent AMS under the same conditions. MPB (Fig.  3A) is somewhat polar but can penetrate membranes at higher concentrations. AMS (Fig. 3B) is highly charged and is much less able to cross biological membranes. Labeling by MPB that can be blocked completely by pretreatment with AMS is indicative of a cytoplasmic facing residue. Labeling by MPB that cannot be blocked by such AMS treatment is indicative of a residue that is not facing the cytoplasm (54,55).
To detect labeling, the subunit a was extracted from membranes by solubilization with a mixture of detergents (cholate, deoxycholate, and octyl glucoside). The subunit a was partially purified via its oligohistidine using nickel affinity chromatography. The samples were electrophoresed, transferred to nitrocellulose, and probed with avidin-conjugated alkaline phosphatase. Results are presented in Fig. 4 for each of the cysteine mutants. These are representative results of experiments performed at least four times. The labeling experiments were controlled for quantity of subunit a present by loading a second panel of samples and transferring to the same sheet of nitrocellulose. The nitrocellulose was cut in half, and the second panel was probed with anti-a antibody. These results are presented in Fig. 5. Subunit a could be detected for all of the mutants except 277C, which was known to be expressed at low levels. Nevertheless, labeling of 277 was relatively high.
The results presented in Fig. 4A show that the following residues can be both labeled by MPB and blocked completely by AMS, indicating a cytoplasm-facing location: 69, 172, 176, 277. In contrast, residues 8, 24, 27, and 131 can be labeled with MPB but cannot be blocked completely by pretreatment with AMS, indicating a location that does not face the cytoplasm. The restriction sites used in the cassette mutagenesis are also indicated. In some cases, an equal mixture of two bases was used at a single site to generate multiple mutations (M, A plus C; K, G plus T, W, A plus T; Y, C plus T; R, A plus G; S, G plus C). The oligohistidine construct was originally made with 50% two stop codons, and 50% cysteine preceding a single stop codon. The HA was inserted at a SapI site in pLN6His that precedes the His 6 . Along with the oligohistidine, it generates a sequence following Ser 268 of YPYDVPDYASEEHHHHHH.
Other residues, 89, 128, 196, and 238, were not labeled at all under these conditions (data not shown). In Fig. 4B, these residues were treated with higher concentrations of MPB, but little if any labeling was detected. In Fig. 5, it can be seen that of the residues that label very poorly, only 196C and 238C seem to be present at lower levels. DISCUSSION The membrane topology of subunit a of the F 1 F 0 ATP synthase from E. coli has been investigated previously, using several different experimental approaches (44 -46). In this study, unique cysteine residues were introduced into subunit a and tested for accessibility to two maleimide reagents of differing water-solubility (54 -56). A simple and reproducible method of preparation of oriented membrane vesicles was used, and multiple sites were probed in each region of subunit a.
The results indicated that residues 69, 172, 176, and 277 (COOH terminus) are accessible to the cytoplasmic face, because they can be labeled by MPB, and this labeling can be blocked completely by AMS. Among these sites, a great range of reactivity exists, indicating greater exposure of residues 69 and 277 than the region of 172 and 176. Also, nearby residues, 89 and 196, were even more resistant to labeling. The labeling pattern of residues 8, 24, 27, and 131 suggests a periplasmic location for these residues, because although they are labeled by MPB, blocking by AMS was not complete. Again, there is a wide range of reactivity from the highly labeled residue 8 to the much less labeled 131. Again, a nearby residue, 128, was much more resistant to labeling. Finally, residue 238 could not be labeled at all. Such a result provides no topological information and could occur because of shielding by protein, including subunit b or c. Such interactions are discussed in the accompanying paper (48).
Two previous studies of the membrane topology of subunit a using the alkaline phosphatase gene fusion technique have been reported (44,45). Such experiments are complementary to those described here, because high activity of alkaline phosphatase is indicative of a periplasmic location of the site of the fusion. Each of the two studies offered eight-transmembrane span models for subunit a but were largely incompatible with each other. With the results presented here, there is now more direct evidence that several regions are cytoplasmic, and it is possible to bring the models into agreement. One of the decisions to be made in analysis of fusion studies is to determine a threshold activity for assigning periplasmic location. If this value were to be set at 20% of the maximal activity in the results of Bjørbaek et al. (44) or at 25% in the results of Lewis et al. (45) using LB-glucose, the same two periplasmic regions would be established in each study: near residues 116 and 229 -241 in the first study (44) and residues 110 -132 and 230 -246 in the second study (45). Our results support the first periplasmic assignment and are not in conflict with the second. Recent MPB labeling experiments using permeabilized cells support this second periplasmic region. 2 A discrepancy does exist when considering the results of Lewis et al. (45), using an LB growth medium lacking glucose. Under those conditions, they found two additional sites of high activity: at residue 200 (28% of maximum) and at the COOH terminus (83% of maximum). Similar results have been seen in other proteins when the site of fusion occurs at a cytoplasmic region without any positively charged residues to anchor COOH-terminal regions of the fusion protein to the cytoplasm (43). In subunit a, lysine normally occurs at position 203, and at the COOH terminus there are no positively charged residues except the final residue, His 271 .
In general, one might expect to encounter some difficulties when applying the alkaline phosphatase fusion technique to subunit a. It is now known by immunoblot analysis that subunit a is not found in membranes when either subunit b or c is missing (57). More recently, subunit a was discovered to be a specific target of the ftsH protease (58). Finally, a variety of results has indicated that the COOH-terminal region of subunit a is important in interactions with the other F 0 subunits (41,48,59,60,61). Therefore, it is not likely that subunit a fusions missing the COOH terminus will be able to assemble into an F 0 complex, and the resulting topology might not be meaningful. The presence of the ftsH protease adds another variable in interpretations of alkaline phosphatase activity levels of fusions with subunit a.
Hydropathy analysis of the sequence of subunit a (41, 62) reveals five potential transmembrane spans that are scored as "certain" by the method of von Heijne (42) and one, residues 182-202, that is scored as "tentative." The labeling results presented here indicate that there should be an even number of transmembrane spans between residues 176 and the COOH terminus. If the region 230 -240 is periplasmic, as both alkaline phosphatase fusion studies indicated, then the "tentative" span is likely to be largely on the cytoplasmic surface.
The most amino-terminal of the predicted transmembrane spans consists of residues 40 -66. The labeling results indicate that residue 69 is cytoplasmic and that residues 8, 24, and 27 may be periplasmic, although seemingly significant levels of MPB labeling were blocked by AMS. The localization of the amino terminus to the periplasm is in disagreement with the results of studies using peptide-generated antibodies (Ref. 46 Fig. 4 was carried out in the same polyacrylamide gel and was probed with anti-a antibody, as described under "Experimental Procedures." Labeling is according to Fig. 4. may be useful to consider alternative interpretations. Lewis and Simoni (63) analyzed the effect of a series of small deletions in subunit a on assembly and function. They found that deletions at the amino terminus, residues 2-35, had a great effect on the level of subunit a found in the membrane and also on the alkaline phosphatase activities of fusions made at residue 125. They concluded that the amino terminus is important for targeting subunit a to the cytoplasmic membrane or for insertion of subunit a into the membrane. A sequence alignment of the amino-terminal portion of subunit a from E. coli and from several closely related organisms is shown in Fig. 6. A conserved region, residues 11-19, has been identified as capable of forming an amphipathic ␣-helix (62). Two possibilities might explain why the amino-terminal residues seem to show accessibility to both sides of the membrane. First, this region might have a function, such as to act as an "intramolecular chaperone" for the folding and membrane assembly of subunit a. As a consequence of this role, it might have increased accessibility on the cytoplasmic surface. Second, the physical properties of this region may lead to artifactual exposure on the cytoplasmic surface during membrane preparation or may in some way contribute to its labeling. Experiments currently under way in our laboratory have demonstrated the accessibility this region of subunit a to the periplasm. 2 Finally, a model of the membrane topology of subunit a is presented in Fig. 7, based on results presented here, and also on this reinterpretation of the alkaline phosphatase fusion studies. Transmembrane spans are all modeled as containing 28 residues. This is consistent with the difficulty of labeling many of the residues that have been tested and with hydropathy analysis (41,62). Furthermore, these long spans would better match the length of the extremely hydrophobic subunits c, which have been shown by NMR analysis to contain similarly long ␣-helices (30). Extramembranous segments that are modeled as helical are those that were predicted to be amphipathic, surface ␣-helices (62). The implications of this membrane topology model on the function of F 0 are considered in the accompanying paper (48). FIG. 7. Transmembrane topology model of the a subunit. Residues that were changed to cysteine are numbered in boldface type and are shown as thick circles. The carboxyl-terminal region is shown in the cytoplasm. The wild type protein ends with His 271 . Cys 277 is shown as the terminal residue, but it is not present in any of the other mutants. The amino-terminal region is shown in the periplasm. Residue Met 1 is shown, although the mature protein has been reported to begin with alanine (3). Other numbers indicate the first and last residues of the putative 28-residue spans.
FIG. 6. Amino acid sequence comparison of subunit a. The amino-terminal regions of all four sequences, E. coli (64), Haemophilus influenzae (65), Vibrio alginolyticus (66), and Pseudomonas putida (67), are shown. The regions aligned at the right represent the beginning of the first transmembrane span. These four proteins show high identity throughout the rest of the sequence. The conserved residues at the amino termini are shown in boldface type.