The γε-c Subunit Interface in the ATP Synthase ofEscherichia coli

Mutants with a cysteine residue in the γ subunit at position 207 and the ε subunit at position 31 were expressed in combination with a c-dimer construct, which contains a single cysteine at position 42 of the second csubunit. These mutants are called γY207C/cc′Q42C and εE31C/cc′Q42C, respectively. Cross-linking of ε to thec subunit ring was obtained almost to completion without significant effect on any enzyme function, i.e. ATP hydrolysis, ATP synthesis, and ATP hydrolysis-driven proton translocation were all close to that of wild type. The γ subunit could also be linked to the c subunit ring in more than 90% yield, but this affected coupling. Thus, ATP hydrolysis was increased 2.5-fold, ATP synthesis was dramatically decreased, and ATP hydrolysis-driven proton translocation was abolished, as measured by the 9-amino-6-chloro-2-methoxyacridinequenching method. These results for εE31C/cc′Q42C indicate that the c subunit ring rotates with the central stalk element. That the γ-ε cross-linked enzyme retains ATPase activity also argues for a γε-c subunit rotor. However, the uncoupling induced by cross-linking of γ to the c subunit ring points to important conformational changes taking place in the γε-c subunit interface during this. Blocking these structural changes by cross-linking leads to a proton leak within the F0.

A proton-translocating F 1 F 0 -type ATP synthase can be found in the periplasmic membrane of bacteria, the thylakoid membrane of chloroplasts, and the cristae membranes of mitochondria. This enzyme can use a proton gradient to synthesize ATP, a process that is reversible in bacteria, where the hydrolysis of ATP is used to generate a proton motive force for substrate and ion transport (1,2). The best-characterized F 1 F 0 -type ATP synthase, that from Escherichia coli, is composed of two parts: a membrane-embedded F 0 part containing three different subunits (a, b 2 , c 12 ) (3-5) and a water-soluble ECF 1 part with five different subunits (␣ 3 , ␤ 3 , ␥, ␦, ⑀). The structures of the ␣ and ␤ subunits are known in detail from x-ray crystallography (6,7). The F 1 portion contains the three catalytic sites, each predominantly on a ␤ subunit, but with the contribution of residues of the ␣ subunits. Recent electron microscopy studies show that the F 1 and F 0 parts are connected by two stalks (8,9). The more central stalk has been shown to include the ␥ and ⑀ subunits (10,11), whereas the outer stalk is made by the b and ␦ subunits (12,13). The available evidence indicates that the central stalk is rotating inside the ␣-␤ subunits, with the more peripheral stalk acting to hold the ␣-␤ hexagon in position relative to the a subunit (14 -16).
Various models of the catalytic mechanism have been proposed where the ring of c subunits is a part of the rotor coupling proton translocation across the membrane to conformational changes that lead to the formation of ATP (17,18). However, the direct evidence that the c subunit ring rotates with the ␥-⑀ subunits is lacking. One approach to establishing that ␥, ⑀, and c subunits work together as the rotor would be to show that cross-linking of these subunits does not block energy-coupling within the complex. A positive result would be strong evidence that the linked subunits move in unison. A negative result would not rule out co-rotation but would require a molecular explanation, which might give important insight into the coupling mechanism.
The sites of interaction between ␥ and ⑀ with the c subunit have been partly mapped by genetic and chemical cross-linking studies. The ␥ subunit interacts with the polar loop of c subunits via a region including residues Tyr-205, Tyr-207, and Glu-208 (19,20). The interaction of the ⑀ subunit with the polar loop of c subunit is via a loop provided by residues Glu-31 to His-38 (21). Although cross-linking of both ␥ and ⑀ to the c subunit ring have been obtained already, the effects of these cross-links on function have been hard to assess. Introducing Cys residues into each of the polar loop regions of c causes dimer formation of this subunit, and the effects of these are difficult to differentiate from ␥-c or ⑀-c interactions. To avoid such problems, we have now generated mutants in which Cys residues are present only in the polar loop of every second c subunit. This was accomplished by generating subunit c fusions genetically, where the linkage is from the C terminus of the first to the N terminus of the second copy of c. The mutant is constructed so that a Cys is present only in the second c subunit of each pair. As a result, there is very little crosslinking between c subunits, allowing the functional effects of cross-linking of ␥ or ⑀ to the c subunit ring to be assessed unambiguously.
The c subunit dimer was created using a two-stage polymerase chain reaction strategy (26). The oligonucleotide 5Ј-CGTGATGTTCGCT-GTCGCGGGTACCACTAGTTAAGCGTTGCTTTTATTTAAAGAGC-3Ј incorporates consecutive in-frame KpnI (bold) and SpeI (bold, underlined) restriction enzyme sites at the 3Ј end of the uncE (subunit c) gene before the stop codon. This oligonucleotide was used in a polymerase chain reaction in conjunction with an antisense primer (TCTGGCG-CAAGCGCGC), which anneals downstream of the uncE region of the wild-type template pRA100 (described in Aggeler et al. (27)). The purified "megaprimer" product of this reaction was used in a second polymerase chain reaction with a sense primer (CTGTCGGAATGGACGA), * This work was supported by National Institutes of Health Grant HL 24526. 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 U.S.C. Section 1734 solely to indicate this fact. which anneals upstream of the uncE region of the template plasmid pRA100. The product, which encodes subunit c with a C-terminal 4-residue GTTS extension, was excised and integrated into the plasmid pRA100 with BssHII/BsrGI, creating the plasmid pJM1.
In a separate polymerase chain reaction, sense and antisense oligonucleotides were employed to amplify the uncE region. The sense primer, 5Ј-CGGGTACCAGTGCTAGCAACGGCGCGAGTAGCGCGAT-GGAAAACCTGAAT-3Ј, incorporates a 5Ј KpnI restriction site (bold) preceding a 27-base region (italic, underlined) encoding the sequence SASNGASSA, a random sequence predicted to be flexible. The corresponding antisense primer, 5Ј-TTAACTAGTCGCGACAGCGAACATC-3Ј, incorporates a 3Ј SpeI restriction site (bold, underlined). Using the plasmid pBS106, which was created by inserting a 7.3-kilobase BsrGI/ NsiI fragment of pYZ215 into pRA100, as template DNA, the amplification product resulted in an uncE region with the mutation Q42C. This amplification product is inserted into pJM1 using restriction sites KpnI and SpeI, creating pRA197, which encodes two c subunits linked by the 11-residue loop GTSASNGASSA with a cysteine residue in position 42 and threonine-serine at the C terminus of the second c subunit only (ccЈQ42C). The double mutant pRA198, containing the c-dimer and a cysteine in position 31 of the ⑀ subunit (ccЈQ42C/⑀E31C), was generated by ligating the 7.2-kilobase XhoI/NsiI fragment of pRA197 with the 5.8-kilobase fragment of pBS106.
Mutant pRA230 containing a cysteine in position 207 of the ␥ subunit (␥Y207C) was created by site-directed mutagenesis, using the oligonucleotide 5Ј-GGGATTACCTGTGCGAACCCGATCC-3Ј, and subcloning of the mutated ␥ gene into pRA100 was performed as described in Aggeler and Capaldi (28). pRA214 with the c-dimer and a cysteine in position 207 of the ␥ subunit (ccЈQ42C/␥Y207C) was obtained from pRA197 and pRA230.
Other Methods-Inner membranes were isolated from the strain RA1 (11), which was transformed with mutant plasmid as described by Foster and Fillingame (29). ATP synthase was purified according to Foster and Fillingame (29), modified by Aggeler et al. (30). Reconstitution of F 1 F 0 in egg lecithin vesicles and CuCl 2 -induced cross-linking at 0.5 mg of protein/ml in 50 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 , and 10% glycerol was carried out as described by Aggeler et al. (31). ATP hydrolysis activity was measured with a regenerating system (32). ATP synthesis was determined with a luciferin/luciferase detection system (Sigma), and NADH and ATP-dependent proton translocation by following the quenching of ACMA 1 (Molecular Probes) with an SLM 8000 fluorometer. 2 Protein concentrations were determined according to Sedmak and Grossberg (34). Cross-linked products were analyzed by SDS gels containing 10 -18% polyacrylamide (35) followed by staining with Coomassie Brilliant Blue R (36) or blotting on nitrocellulose membranes for identification with a pair of monoclonal antibodies against ␥ and ⑀ subunit, respectively, as described by Mendel-Hartvig and Capaldi (37).

Characterization of the Genetically Fused c Subunit Mutant
ccЈQ42C-It has recently been demonstrated that c subunits of the F 0 part of ECF 1 F 0 can be linked from the C terminus of one copy to the N terminus of a second copy to form fused dimers (4). Furthermore, the enzyme containing these genetically fused c subunits was shown to retain function. By introducing a Cys residue in the polar loop of only one of the two copies of such a c subunit dimer, we reasoned that it might be possible to prevent the cross-linking between polar loops of near neighbor c subunits, which has been problematic in previous studies that examined the effects of cross-linking ␥ or ⑀ subunits to the c subunit ring. To this end, the mutant shown in Fig. 1 was constructed.
This new mutant contained the c subunit as a dimer connected by a loop of 11 amino acids from the C terminus of the first to the N terminus of the second copy of the subunit. The second c subunit of each pair also had two additional amino acids at its C terminus, as a result of the cloning strategy used. Finally, the polar loop region of the second, but not the first c subunit of the construct, had a Cys replacing the Gln at position 42. If the c subunits were arranged as a ring of 12 (33), there would have been 6 Cys residues at alternating positions in the ring in ECF 1 F 0 from this mutant.
As shown in Fig. 2A, the ATPase activity of the inner membranes from the ccЈQ42C mutant was approximately one-half that of wild-type cell membranes. ECF 1 F 0 isolated from the mutant had an ATPase activity of 16.5 units/mg compared with around 24 units/mg for wild type, a reduction of around 30%. The lower activity appeared to be in part due to lower levels of assembled enzyme, but as shown by data for the purified enzyme, the mutant also had a lower turnover rate. Enzyme in inner membranes from the mutant ccЈQ42C retained a level of DCCD sensitivity of the ATPase activity close to that for wildtype, 70% compared with 80% using 20 M DCCD. Proton pumping was only marginally affected (compare plot C against plot B of wild type in Fig. 2). In this respect, the dimer construct prepared in this study was different from that described recently by Jones and Fillingame (4), which had a much reduced proton-pumping function. The efficient coupling of proton translocation to catalytic site activity was further evaluated by ATP synthesis measurements, studies not done by Jones and Fillingame (4). ATP synthesis activity by the mutant containing the fused c subunit dimer was not measured directly, but even in the mutant with the ⑀31C, there was no dramatic decrease in ATP synthesis (90% that of wild type). Thus, dimers linked by a loop on the periplasmic side, away from the F 1 , functioned well in energy coupling within the ECF 1 F 0 complex.
As hoped, there was no significant cross-linking between Cys residues in the ccЈ subunit ring at levels of Cu 2ϩ , which generated high yields of ␥-c 2 or ⑀-c 2 products. At higher Cu 2ϩ concentrations, tetramer formation was observed without concomitant loss of activity.
Covalent Linkage of ⑀ to the c Subunit Ring in the Mutant ⑀E31C/ccЈQ42C-To examine the effects of cross-linking of the ⑀ subunit to the c subunit ring, the mutant ⑀E31C/ccЈQ42C was constructed. As shown in Fig. 3, there was ready disulfide bond formation between ⑀ and the ccЈ in this mutant. At 150 -200 M Cu 2ϩ , the yield of ⑀-c 2 cross-linking was 80 -90% based on the disappearance of the ⑀ subunit in the Coomassie Blue-stained gels when purified ECF 1 F 0 from the mutant was reacted with Cu 2ϩ . A similar yield of cross-linking was calculated in inner membranes after treatment with the same Cu 2ϩ concentration, based on the disappearance of ⑀ subunit when determined by Western blotting.
The effects of cross-linking ⑀ to ccЈ on functioning of ECF 1 F 0 are summarized in Fig. 4. High cross-linking yields had no significant effect on ATP hydrolysis rates or on proton pumping function when measured in the ACMA quenching assay (Fig.  4B). The rate of ATP synthesis was only reduced 30% when 200 M Cu 2ϩ was used for cross-linking. However, a similar level of inhibition was obtained when wild-type enzyme was treated in the same way. Therefore, this effect could be due to crosslinking of other enzymes in the E. coli membrane that lead to some proton leakage or an effect on ECF 1 F 0 other than crosslinking of ⑀ to c 2 .
Covalent Linkage of ␥ to the c Subunit Ring in the Mutant ␥Y207C/ccЈQ42C-Our previous studies have shown that a Cys at positions 205 or 207 (19) can be cross-linked to the polar loop of the c subunit in high yield. The mutant ␥Y205C/cQ42C was examined in detail previously (19). Here, the mutant ␥Y207C/ccЈQ42C was studied because it grew like wild type on succinate, whereas the ␥Y205C/ccЈQ42C mutant did not.
As shown in Fig. 5A, there were significant levels of crosslinking of ␥ to ccЈ in ECF 1 F 0 purified from the mutant at 200 M Cu 2ϩ , but maximal yields required levels up to 1 mM. In membranes, lower levels of Cu 2ϩ gave higher levels of ␥-c 2 product than in the isolated membrane, based on monoclonal antibody binding in Western blots (Fig. 5C), but cross-linking was still incomplete at 200 M. Fig. 5B shows the effects of the crosslinking on ATPase activity. At 200 M Cu 2ϩ , with an approximate 50% yield of cross-link, there was a 170% increase in ATPase activity. At 500 M Cu 2ϩ , the activation was over 2-fold. Fig. 5A shows that subsequent dithiothreitol treatment failed to break all of the ␥-c 2 cross-links or regenerate fully the free ␥ subunit. This is presumably because of covalent linking of ␥ to c 2 via the Cys of ccЈ and the Tyr at position 205 of ␥, based on previous work.
Functional studies on E. coli membranes were conducted at 200 M Cu 2ϩ , a concentration at which effects unrelated to cross-linking within ECF 1 F 0 were minimal. The cross-linking of ␥ to c 2 in membranes led to the same activation of ATPase activity seen with the purified enzyme. Concomitant with this was a major reduction in proton pumping and a 70% reduction in ATP synthesis rates (Fig. 6), even though the yield of crosslinked product was maximally 50%. These represent a direct effect of cross-linking of ␥ to c 2 and not a secondary result of Cu 2ϩ addition, as proton pumping was mostly recovered on addition of dithiothreitol, which breaks disulfide bonds (result not shown).

DISCUSSION
The cross-linking of ␥Y207C or ⑀E31C to cQ42C has been studied before (19). However, functional studies were complicated by the fact that the Cys introduced into the c subunit led to the formation of c dimers in addition to the ␥-c or ⑀-c product being examined. In our previous work (19), this dimer formation caused an inhibition of the enzyme function, making interpretation of the effects of cross-linking of the central stalk subunits to the c subunit ring difficult. To avoid this problem , we followed an approach developed by Jones and Fillingame (4) and created a c dimer in which the two copies were connected by a flexible loop at the periplasmic side. We then introduced 1 cysteine/dimer in the second subunit in the polar loop part at position 42. The effect was to place Cys at alternate c subunits around the ring on the F 1 -facing side. This proved to eliminate significant cross-linking between adjacent polar loops of the c subunits. Tetramers, i.e. cross-links of two ccЈ subunits, were obtained at higher Cu 2ϩ concentrations. As reported by Jones and Fillingame (4), E. coli with ECF 1 F 0 in which the c subunit is a dimer instead of monomer, grew well on succinate. With the linkage between the two monomers on the periplasmic side, ECF 1 F 0 retained ATP-coupled proton translocation close to that of wild-type enzyme.
In the mutant ⑀E31C/ccЈQ42C, there was ready cross-linking of ⑀ to the c subunit ring, an observation first made by Zhang and Fillingame (38) using a mutant incorporating a Cys at position 31 in ⑀ but with a Cys for Gln-42 in all 12 c subunits. In their studies, they compared enzyme function in membranes of their mutant that had been prepared with and without reducing conditions. This was possible because cross-linking between the Cys at position 31 in ⑀ and a Cys in the polar loop at positions 40, 42, or 43 was formed in significant yield without the addition of Cu 2ϩ . Zhang and Fillingame (38) report that the yield of cross-linking of ⑀ Cys-31 to c Cys-42 in their studies was 63-78%, based on staining in Western blots, and that this level of cross-linking reduced ATPase activity by around 30% and proton pumping by 50 -90%. Based on these findings, they concluded that cross-linking of ⑀ to the c subunits blocks alternation between catalytic sites in F 1 , with residual ATPase activity reflecting mutationally induced uncoupling of F 1 from F 0 . In their studies, Zhang and Fillingame (38) formed significant levels of c dimer due to cross-linking of proximal Cys-42 residues. They argued that c-c cross-linking was not responsible for the observed effects because the ACMA quenching of enzyme in the mutant cQ42C was normal.  5 and 6). 100-l samples were then kept for 1 h without (lanes 1-5) or with 15 mM dithiothreitol (lane 6). 10 -18% polyacrylamide gels were run after the addition of 7 mM EDTA, 50 mM N-ethylmaleimide, and dissociation buffer without reducing agent. B, cross-link yield was determined by densitometrically scanning a Coomassie Brilliant Blue-stained gel and quantitating the disappearance of ␥ subunit (Ⅺ). ATP hydrolysis activity, expressed as percentage of 14.9 units/mg (mol of ATP hydrolyzed/min/mg) (E). C, identification of cross-link products of ␥ mutant ATP synthases. Inner membranes from ccЈQ42C (i), ccЈQ42C/␥Y207C (ii), and ␥Y207C (iii) were treated with 5 mM dithiothreitol (lanes 1), 0 (lanes 2), or 200 M CuCl 2 (lanes 3). 40-g samples were applied under nonreducing conditions on 10 -18% polyacrylamide gels and blotted onto nitrocellulose membranes, and bands were identified with two ␥-specific monoclonal antibodies. Here, we examined the effects of cross-linking of a Cys in ⑀ at position 31 and in the c subunit at position 42 in both purified ECF 1 F 0 from the mutant ⑀E31C/ccЈQ42C and in inner membranes from this mutant. Cross-linking had essentially no effect on ATPase activity in purified enzyme, where the yield of ⑀-c 2 cross-linking was very high, and yet generation of crosslinking between polar loops of ccЈ subunits was essentially zero. Cross-linking of ⑀ to the ccЈ subunit ring also had no significant effect on ATPase activity in inner membranes, and there was little or no effect on proton translocation or ATP synthesis because of covalent linkage of the two subunits. These results are consistent with the c subunit ring and the ⑀ subunit rotating together during coupling of catalytic site events and proton channeling in both ATP hydrolysis and ATP synthesis.
It is interesting to note that the yield of cross-link of the ⑀ to c 2 was close to 100% (and that of ␥ to the c 2 subunits greater than 80%), although only 50% of the 12 polar loops of the c subunit ring contain a Cys. The most likely explanation is that the polar loop regions of c are deformable enough that the Cys-42 in two (or more) c 2 subunits can reach Cys-31 of ⑀ or Cys-207 of ␥. This is supported by the fact that some tetramer formation was observed, requiring significant flexibility of the loop.
There is now considerable data to indicate that both ␥ and ⑀ subunits rotate during functioning of F 1 F 0 -type ATPases (14 -16). The observation that the two can be cross-linked together without effect on ATP hydrolysis or ATP synthesis is strong evidence that the two subunits move together during coupling (25,27). It was surprising, then, to find that cross-linking of ␥ to the c subunit ring induced an uncoupling of the F 1 from the F 0 part such that ATPase activity was increased by release from constraints of the F 0 , and that the F 0 part became leaky to protons. This leak through the F 0 not only affected ATP-driven proton translocation but also NADH-driven proton pumping (see Fig. 6). Accepting that ␥, ⑀, and c subunits all rotate in unison, the data for the mutant ␥Y207C/ccЈQ42C are best interpreted as indicating conformational changes of the ␥ subunit and/or c subunit ring during the rotation that are important for functioning. It is important to note that the mutant functions like wild type in the absence of cross-link formation. Only after disulfide bond formation, which would prohibit the conformational change, was there loss of coupling. It is interesting that the recent model of the c subunit ring provided by Fillingame and co-workers (33) has a key residue in the proton-translocating path, Asp-61, pointing away from the surface of the a subunit thought to complete the proton channel. These authors propose that there must be rearrangement of the c subunits as each is brought into interaction with the a subunit during the rotation process. It could be this reorientation of all or a part of the c subunits that is being blocked by the cross-linking described here. Clearly, the ␥Y207C/ccЈQ42C mutant will be interesting to study by physical methods such as video microscopy to explore the linkage of rotation of the ␥-⑀ rotor to proton translocation in more detail.
In summary, we provide the first strong evidence that the c subunit ring rotates along with the ␥Ϫ⑀ rotor, based on results with the mutant ⑀E31C/ccЈQ42C. It appears that the rotor is not a rigid entity but undergoes conformational rearrangements transmitted from ␥Ϫ⑀ to the c subunits during coupling steps based on studies of the ␥Y207C/ccЈQ42C mutant. Such conformational flexibility of the ␥Ϫ⑀ to the c subunit ring interface was already apparent in previous cross-linking studies in that a Cys at position 31 of ⑀ readily cross links at several sites on the polar loop of c and that a Cys at 205 of ␥ can link to several positions on this loop of the c subunit.