Close Proximity of a Cytoplasmic Loop of Subunit awith c Subunits of the ATP Synthase fromEscherichia coli *

Interactions between subunit a and the c subunits of the Escherichia coli ATP synthase are thought to control proton translocation through the Fo sector. In this study cysteine substitution mutagenesis was used to define the cytoplasmic ends of the first three transmembrane spans of subunit a, as judged by accessibility to 3-N-maleimidyl-propionyl biocytin. The cytoplasmic end of the fourth transmembrane span could not be defined in this way because of the limited extent of labeling of all residues between 186 and 206. In contrast, most of the preceding residues in that region, closer to transmembrane span 3, were labeled readily. The proximity of this region to other subunits in Fo was tested by reacting mono-cysteine mutants with a photoactivated cross-linker. Residues 165, 169, 173, 174, 177, 178, and 182–184 could all be cross-linked to subunit c, but no sites were cross-linked to b subunits. Attempts using double mutants of subunita to generate simultaneous cross-links to two differentc subunits were unsuccessful. These results indicate that the cytoplasmic loop between transmembrane spans 3 and 4 of subunita is in close proximity to at least one csubunit. It is likely that the more highly conserved, carboxyl-terminal region of this loop has limited surface accessibility due to protein-protein interactions. A model is presented for the interaction of subunit a with subunit c, and its implications for the mechanism of proton translocation are discussed.

The ATP synthase from Escherichia coli is typical of the enzymes found in mitochondria, chloroplasts, and many other bacteria that synthesize ATP (for a recent review see Ref. 1). It is composed of two subcomplexes: an F 1 sector with subunits that contain the catalytic sites and a membrane-bound F o sector with subunits that conduct protons across the membrane. In the E. coli enzyme five different subunits are found in F 1 : ␣, ␤, ␥, ␦, and ⑀ with a stoichiometry of 3:3:1:1:1. In the F o sector, three different subunits are found: a, b, and c with a stoichiometry of 1:2:9 -12 (2).
Evidence for rotation of subunits through 360 degrees in response to ATP hydrolysis has been provided by direct observation of fluorescently labeled actin filaments attached to ␥ or ⑀ (3)(4)(5)(6). Therefore, it is thought that the enzyme functions as a rotary motor. Other evidence has been provided that a ring of c subunits is also part of the rotor (7)(8)(9)(10)(11)(12). Subunits that make up the stator include ␦, a, and b in addition to ␣ and ␤, which house the catalytic sites. The mechanism by which the proton motive force across the membrane drives rotation of the c oligomer of F o along with subunits ␥ and ⑀ is not known. High resolution structures of F 1 , primarily from bovine mitochondria, have provided details about the catalytic sites and their conformational changes (13)(14)(15), but less is known about the structure of the F o subunits.
Subunit b is thought to be embedded in the membrane via a hydrophobic region near its N terminus. Studies of an aminoterminal fragment of subunit b using NMR and detection of disulfide formation have shown it to be ␣-helical and dimeric (16). Other studies have shown that a truncated, soluble form of b is extended and dimeric (17)(18)(19)(20)(21). The amino terminus of subunit b is thought to interact with subunit a, and its carboxyl terminus is thought to interact with ␦ (22).
Subunit c is a hydrophobic protein with a conserved aspartic or glutamic acid that is thought to participate in proton translocation steps. NMR studies of the monomeric c subunit have shown it to be an ␣-helical hairpin with two transmembrane spans connected by a short polar loop (23)(24)(25), with the conserved Asp-61 residue near the center of the second helix. The conformation of subunit c appears to be pH-dependent, as indicated by NMR studies (26). The number of c subunits that make up the oligomer in F o from E. coli is still uncertain but is likely to be 10 (27).
Subunit a has been analyzed by surface labeling of unique, engineered cysteine residues. Such studies have established the number of transmembrane spans (28 -30) and have characterized the first cytoplasmic loop (31) and the first periplasmic loop (32). These results are summarized in Fig. 1, in which five transmembrane spans are shown. The first cytoplasmic loop is drawn to reflect the limited accessibility of its central region to the reagent MPB. 1 Residues important for function (33)(34)(35)(36) have been shown to reside in transmembrane spans 4 (Arg-210 and Glu-219) and 5 (His-245). In addition, Glu-196, appears to reside in the cytoplasmic loop preceding transmembrane span 4. It is likely that Arg-210 of subunit a interacts with the essential residue Asp-61 of subunit c during coupled proton translocation since disulfide cross-linking studies have shown that transmembrane span 4 of subunit a, between residues 207 and 225, appears to be in contact with subunit c (37).
The mechanism by which subunit a contributes to the proton conducting path of the ATP synthase is not clear, but it is likely to play such a role. Models have been presented in which * Support for this study was provided by National Institutes of Health Grant GM40508 and The Welch Foundation Grant N-1378. 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  subunit a contributes amino acids that make up two halfchannels, one opening to the periplasm and one to the cytoplasm, that allow access to the proton binding site on subunit c (38,39). Previously identified residues in subunit a (33)(34)(35)(36), thought to be important in proton access to subunit c, are found within transmembrane spans, near the periplasmic surface (32). So far, it is not clear how proton access might be controlled at the cytoplasmic surface. These studies were undertaken to examine the structure of the cytoplasmic loop between transmembrane spans 3 and 4 and to determine which other F o subunits those residues were near.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes were obtained from New England Biolabs. Synthetic oligonucleotides were obtained from Operon Technologies. DNA sequencing was done by Lone Star Labs. MPB and TFPAM-3 were obtained from Molecular Probes. Nickel-nitrilotriacetic acid (Ni-NTA) resin and DNA miniprep kits were obtained from Qiagen. Mouse anti-HA antibody was obtained from Roche Molecular Biochemicals. Goat anti-mouse IgG-alkaline phosphatase conjugate, goat antirabbit IgG-alkaline phosphatase conjugate, avidin-conjugated alkaline phosphatase, 5-bromo-4-chloro-3-indoylphosphate p-toluidine salt, pnitro blue tetrazolium chloride, SDS-PAGE, polyvinylidene difluoride membranes, and low and broad range protein molecular weight standards were obtained from Bio-Rad. N-Octyl-␤-D-glucoside was purchased from Anatrace. All other chemicals were purchased from Sigma or Fisher. The UV lamp was purchased from UVP, model UVL-56, with a wavelength of 365 nm. Anti-b antibodies and anti-c antibodies were gifts of Dr. R. Capaldi, University of Oregon and Dr. K. Altendorf, University of Osnabrü ck, Osnabrü ck, Germany, respectively.
Plasmids, Mutagenesis, Growth, and Expression-The plasmids, pLN6HisHA (28), pLN7HisHA (31), pLN46HisHA (32), pTW1HisHA (28), pARP2HisHA (39), and pDP1018HisHA were used for the construction of mutants. Plasmid pDP1018HisHA was constructed by ligation of the 770-base pair BsaHI-AflIII fragment from pBJA1018 (39) and the 2592-base pair BsaHI-AflIII fragment from pLN7HisHA. These plasmids produce subunit a that includes an HA epitope (YPYDVP-DYA), derived from the hemagglutinin protein of human influenza virus, and a His 6 tag at the carboxyl terminus of the protein. These tags have no effect on function. These plasmids differ primarily in the placement of unique restriction sites that are necessary for cassette mutagenesis. Mutagenesis and growth of cultures were carried out as described previously (32).
Preparation of Inside-out Membrane Vesicles-Inside-out membrane vesicles were made from a 250-ml culture (per experimental sample) in LB medium grown to A 600 ϭ 1.0. Cells were resuspended in 10 ml of 50 mM Tris-HCl, 10 mM MgSO 4 , pH 7.5, and passed through a French press at 14,000 p.s.i. Cell debris and unbroken cells were removed by a low speed spin at 8000 rpm for 15 min. The supernatant was then centrifuged at 50,000 rpm for 1 h at 4°C in a Beckman Ti-70 rotor. The pellet was resuspended and used in the experiments or stored at Ϫ80°C.
Labeling of Membrane Vesicles-The inside-out membrane vesicles were labeled in 200 mM Tris-HCl (pH 8.0), with 120 M MPB at room temperature for 15 min. 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 subunit a was purified by Ni-NTA resin as described below.
Cross-linking of Subunit a by TFPAM-3-This cross-linker is expected to span 10 -15 Å. The cross-linking was carried out by the methods described previously (31). The membrane vesicles were suspended in 50 mM MOPS (pH 7.0), 5 mM EDTA, and 10% glycerol. It was incubated with 200 M TFPAM-3 for 60 min at room temperature, and the reaction was terminated by addition of 15 mM cysteine. After addition of 5 mM ATP, cross-linking was activated by UV light. The reactions were terminated after 2 h at room temperature.
Purification and Detection of Subunits and Molecular Models-After reaction, membrane vesicles were resuspended in extraction buffer (200 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). Subunit a was purified using Ni-NTA as described previously (32). Samples of purified subunit a were subjected to SDS-PAGE and transferred to a polyvinylidene difluoride membrane as described previously (31). For detection of subunit a, previously published procedures were followed (29). For detection of subunit b, the blocked membrane was incubated at room temperature for 2 h with b antibody at a dilution of 1:1000. After washing three times with TBS/Tween 20, it was incubated with goat anti-mouse IgG-alkaline phosphatase conjugate at a dilution of 1:1000 for 1 h. After another three washings with TBS/Tween 20, color was developed as described above. For detection of subunit c, the blocked membrane was incubated at room temperature for 2 h with c antibody at a dilution of 1:5000. After washing three times with TBS/ Tween 20, it was incubated with goat anti-rabbit IgG-alkaline phosphatase conjugate at a dilution of 1:1000 for 1 h. After another three washings with TBS/Tween 20, color was developed. The molecular models shown were created from the Protein Data Bank file 1c99, model 1, using RasMol (40).

RESULTS
A series of 46 monosubstituted cysteine mutants of subunit a that are thought to be near the cytoplasmic surface were constructed. All mutants grew in succinate minimal medium, indicating the ability to carry out oxidative phosphorylation. Inverted inner membrane vesicles were prepared from these mutants and were then labeled with MPB to test the surface accessibility of each residue. The results for mutations between residues 167 and 174 are shown in Fig. 2, and the results for mutations between residues 177 and 184 are shown in Fig. 3. The results of all mutants are summarized in Table I. For comparative purposes, each labeling experiment contained G70C, a residue shown previously (29) to label strongly in membrane vesicles, i.e. from the cytoplasmic surface, and E131C, a residue that was shown to label poorly in membrane vesicles (28) but strongly from the periplasmic surface (29). In Fig. 2A, the results of MPB labeling in membrane vesicles followed by Ni-NTA purification of subunit a are shown. In B an immunoblot of the corresponding samples is shown to confirm that the level of protein is similar. The cytoplasmic loop between residues 64 and 100 is drawn to indicate that the central region has limited accessibility to the reagent MPB, but the segments nearest the membrane are highly exposed. The periplasmic loop between residues 124 -146 has been drawn to reflect its partial exposure throughout this region. results of H185C, W186C, A187C, F188C, I189C, V191C, and N192C are shown in Table I. None of them show significant labeling, expect for H185C and W186C. Other residues between 193 and 206 show only a trace of labeling, at most.

The Proposed Cytoplasmic Loop between Transmembrane Spans 3 and 4 -In
The Cytoplasmic Ends of Transmembrane Spans 1, 2, and 3-Residues Phe-60, Arg-61, Val-63, and Ala-64 near the end of transmembrane span 1 were analyzed, and the results are presented in Table I. In this group, only A64C, lane 7, is labeled significantly when compared with the controls, G70C and E131C. The results of labeling I101C, A102C, and P103C, three residues near the cytoplasmic end of transmembrane span 2, are shown in Table I. Only I101C is labeled above a background level. The results of labeling residues near the cytoplasmic surface of transmembrane span 3, L160C, I161C, L162C, F163C, Y164C, and S165C, are also presented in Table I. None of these residues were labeled to a significant extent.
Cross-linking Analysis-A series of cross-linking experiments was conducted to determine if any of the residues in the cytoplasmic loop between transmembrane spans 3 and 4 are near subunits b or c. Monosubstituted cysteine mutants at the following positions were reacted with the photoactivated crosslinker TFPAM-3: 165, 167-171, 173-175, 177-179, and 182-184. After UV activation, nine residues showed evidence of cross-linking to subunit c, and the results are presented in Fig.  4. In A, samples S165C, K169C, G173C, and F174C are probed with anti-HA for detection of subunit a, on the left side, and the same samples are probed with anti-c antibodies on the right side. The a-c cross-link can be seen in both blots and is dependent upon UV activation, designated by the "ϩ" sign. Similarly, as shown in B, samples E177C, L178C, P182C, and F183C are probed with anti-HA for detection of subunit a on the left side and with anti-c antibodies on the right side. Residue N184C was also found to cross-link to subunit c in a similar manner (results not shown).
The following double cysteine mutants were constructed and tested for cross-linking using TFPAM-3: S165C/ Labeling is scored as indicated in Figs. 2 and 3. "ϩ" indicates a level of labeling that is not significantly higher than background. "ϩϩϩϩϩ" indicates the highest level of labeling seen, that of G70C. FIG. 2. Labeling of subunit a from the cytoplasmic surface:  residues 167, 168, 169, 170, 171, 173 and 174 P182C, K169C/P182C, G173C/P182C, F174C/P182C, E177C/ P182C, L178C/P182C, S165C/F183C, K169C/F183C, G173C/ F183C, F174C/F183C, E177C/F183C, L178C/F183C, S165C/ N184C, K169C/N184C, G173C/N184C, F174C/N184C, E177-C/N184C, and L178C/N184C. In each case only a single crosslinked product was seen with no indication of an a-c 2 product (results not shown). DISCUSSION The labeling results presented here provide further information about the junctions of the first three transmembrane spans of subunit a, and this information is summarized in Fig. 5. This model differs from that shown in Fig. 1 in two respects. The junctions of the transmembrane spans at the cytoplasmic surfaces have been adjusted to reflect the results of labeling, and the periplasmic loop between spans 4 and 5 has been adjusted, as discussed below. Previous work (29) had identified residue T37C, at the amino-terminal end of transmembrane span 1, as accessible by MPB labeling, but residues W39C and D44C were not. At the carboxyl-terminal end of transmembrane span 1, residues 67-70 labeled strongly, while residue 64 labeled weakly (29). In this work, residues F60C, R61C, and V63C were shown not to label, relative to the weak labeling of residue A64C. The results indicate a core transmembrane segment of ϳ26 residues from 38 to 63. This matches well the calculated hydropathy peak centered near residues 50 -53 (41,42). The carboxyl-terminal end of transmembrane span 2 was analyzed in previous work (32). Residues D119C, L120C, and P122C were inaccessible to MPB labeling, while residues D124C and P127C were labeled. Near the amino-terminal end of transmembrane span 2, residues 92-98 were all labeled by MPB (31). Here it was shown that residue I101C was labeled to a small degree relative to residues A102C and P103C. These results indicate a core transmembrane segment of ϳ22 residues from 102 to 123, which again matches closely the calculated hydropathy peak centered near residues 109 -111 (41,42). The amino-terminal end of transmembrane span 3 was analyzed in previous work (32), and residues V142C, S144C, and D146C were all shown to be labeled with MPB. At the Four cysteine mutants at the indicated positions were analyzed in preparations of membrane vesicles. Samples were reacted with TFPAM-3 and then exposed to UV radiation. After incubation, subunit a was extracted with detergent and purified. The resulting blots were probed with anti-a antibodies on the left or with anti-c antibodies on the right. The plus (ϩ) sign indicates UV exposure, while the minus (Ϫ) sign indicates identical handling without the UV exposure. The expected sizes of the a, c, and a-c cross-linked subunits are indicated. B, four additional cysteine mutants were treated in an identical fashion. Note that uncross-linked subunit c also co-purified with His-tagged subunit a and that the c antibodies recognized a second band that might be a noncovalent c dimer co-migrating with standard 28.8 kDa.
FIG. 5. Transmembrane spans of subunit a. Only residues analyzed in this study are colored. Residues that were labeled with MPB in membrane vesicles are colored dark blue, while those with little or no labeling are colored light blue. Residues that were cross-linked using TF-PAM-3 to subunit c are identified with a red asterisk. Transmembrane spans were determined as described under "Discussion." carboxyl-terminal end of transmembrane span 3 residue K169C was shown to be labeled with fluorescein-maleimide (30) and G172C was shown to be labeled with MPB (28). Here, residues 160 -165 and K167C were shown to be inaccessible to MPB, while M168C was labeled. Therefore, a core transmembrane segment of ϳ21 residues exists from 147 to 167, which again matches well the calculated hydropathy peak centered near residues 156 -158 (41,42).
In contrast, the results presented here give no indication of the amino-terminal end of transmembrane span 4. Previous studies had indicated that residue E196C could be labeled from the cytoplasmic surface, but those results occurred under slightly more strenuous labeling conditions (30), or the labeling was rather weak (28). In this study the labeling of all residues tested between 185 and 206 was seen to be very weak, and only a few showed a level of labeling that could be considered above background. While it is likely that absolute levels of labeling would differ under different conditions and with different maleimido-reagents, the lack of relative differences seen here makes it impossible to recognize the junction of transmembrane span 4 with the cytoplasmic surface. This also applies to the periplasmic ends of both transmembrane spans 4 and 5, where only three residues were found to be labeled by MPB (29). In studies of lac permease, single amino acid insertions (43) or deletions (44,45) have been used as indicators of the ends of transmembrane spans. In previous work reported by this laboratory (39,41), insertions of alanine after residues 202 and 225 did not seriously impair function of the ATP synthase, while those after positions 212, 217, and 222 did. These results are consistent with a hydrophobic core segment in transmembrane span 4 approximately between residues 202-225. Likewise, consideration of insertions of alanine (39) and MPB labeling (29) identifies a core segment in transmembrane span 5 between residues 238 and 259. The loop now identified between residues 226 and 236 is consistent with the marginal effects of alanine insertions after residues 225, 229, and 233. This is in contrast to the loss of function seen with insertions after residues 217, 222, 238 and 243 (37,39).
The labeling of the residues in the putative loop between transmembrane spans 3 and 4 showed a strikingly asymmetric pattern. Twelve of thirteen residues tested between 168 and 184 were found to be labeled by MPB, while none of the 20 residues tested between 185 and 206 showed significant labeling. In consideration of the bulky nature of the reagent MPB, it is likely that the amino-terminal half of this loop is highly exposed, given the extensive labeling pattern. The lack of labeling at the carboxyl-terminal end of the loop is consistent with significant protein-protein interactions. This asymmetry correlates with amino acid sequence conservation, in which conservation for residues 190 -225 in subunit a is high (46), while for residues prior to 190 conservation is low. Evidence for a surface of interaction between transmembrane span 4 of subunit a and the carboxyl-terminal transmembrane span of subunit c has been provided by studies of engineered disulfides by Jiang and Fillingame (37). In the structural model of subunit c (26) determined at pH 8, shown in Fig. 6, these residues form a curved surface, suggesting that transmembrane span 4 of subunit a wraps around that surface. If Glu-196 of subunit a were part of an ␣-helical extension of transmembrane span 4 then it would be found near the loop of subunit c (residues 41-43) since L207C of subunit a forms a disulfide with I55C of subunit c and residues 43-55 in subunit c are all ␣-helical in the structural model at pH 8. This possibility is illustrated in Fig. 6 in which residues 190 -225 of subunit a are modeled as an ␣-helix.
The labeling pattern of residues 168 -184 is consistent with a connection from transmembrane span 3 to the top of a ring of c subunits with high surface accessibility. The accessibility of these residues to MPB resembles that of the residues near the cytoplasmic surface of transmembrane spans 1 and 2 of subunit a. The results of cross-linking with TFPAM-3 indicate that this region is near subunit c, i.e. at a distance of 10 -15 Å. The range of sites that can be cross-linked indicates that both the end of transmembrane span 3 and the center of the second cytoplasmic loop are near subunit c. It is likely that the efficiency of cross-linking was too low to detect a doubly cross-linked product. Therefore, it remains uncertain if all positions cross-link to the same subunit c. Although this loop is not predicted to be entirely ␣-helical (42), the pattern of labeling and cross-linking from residues 173-179 (GFTKELT) is consistent with an ␣-helix.
The proposed interactions between subunit a and subunit c discussed above have important implications for the mechanism of proton translocation of which several proposals have been described recently (26,47). As shown in Fig. 6, a-Arg-210 (blue) would be near the c-Asp-61 (red). That would place a-Glu-219 pointing into the bundle of transmembrane spans of subunit a (30) and a-Glu-196 at the top near the loop of subunit c. The residues in subunit a that cross-link to subunit c are likely to be near the adjacent subunit c in the ring. It has been proposed that the proton motive force causes protonation of a network of residues in subunit a near the periplasmic surface (32), including a-Glu-219 and a-His-245. This could cause a twisting or bending of transmembrane span 4 of subunit a allowing protonation of c-Asp-61 from the periplasmic side. This would cause a-Arg-210 to move away from c-Asp-61, as subunit c changes to its protonated conformation (26) and the c-Asp-61 rotates to its new position. The conformational changes in the transmembrane span 4 of subunit a and in the loop region of subunit c, would be transmitted to the Glu-196 region of subunit a. This could promote the deprotonation of c-Asp-61 of the adjacent subunit c, perhaps by the exposure of a-Glu-196 and the introduction of water. At this point, the a-Arg-210 would now be in position to be attracted electrostatically to c-Asp-61, but since it would be to the adjacent subunit c, one step of rotation will have occurred.  6. Interacting surfaces of subunit a and subunit c. The structure of subunit c with deprotonated Asp-61 as determined by NMR (26) is shown on the right. Residues that formed disulfides with subunit a residues are colored cyan (37). In subunit c, residue Asp-61 is colored red, and all other residues are colored in the Corey-Pauling-Koltun scheme (oxygen, red; nitrogen, blue; carbon, gray; hydrogen, white; sulfur, yellow). Transmembrane span 4 of subunit a and its aminoterminal extension to residue 190 are shown on the left. Residues in subunit a that form disulfides to residues in subunit c are colored magenta (37). In subunit a, residue Arg-210 is colored blue, and other residues are colored in the Corey-Pauling-Koltun scheme. Expected membrane boundaries are marked by horizontal lines: subunit a (K203 and I225) and subunit c (T51 and V78).