The stalk region of the Escherichia coli ATP synthase. Tyrosine 205 of the gamma subunit is in the interface between the F1 and F0 parts and can interact with both the epsilon and c oligomer.

The soluble portion of the Escherichia coli F1F0 ATP synthase (ECF1) and E. coli F1F0 ATP synthase (ECF1F0) have been isolated from a novel mutant γY205C. ECF1 isolated from this mutant had an ATPase activity 3.5-fold higher than that of wild-type enzyme and could be activated further by maleimide modification of the introduced cysteine. This effect was not seen in ECF1F0. The mutation partly disrupts the F1 to F0 interaction, as indicated by a reduced efficiency of proton pumping. ECF1 containing the mutation γY205C was bound to the membrane-bound portion of the E. coli F1F0 ATP synthase (ECF0) isolated from mutants cA39C, cQ42C, cP43C, and cD44C to reconstitute hybrid enzymes. Cu2+ treatment or reaction with 5,5′-dithio-bis(2-nitro-benzoic acid) induced disulfide bond formation between the Cys at γ position 205 and a Cys residue at positions 42, 43, or 44 in the c subunit but not at position 39. Using Cu2+ treatment, this covalent cross-linking was obtained in yields as high as 95% in the hybrid ECF1 γY205C/cQ42C and in ECF1F0 isolated from the double mutant of the same composition. The covalent linkage of the γ to a c subunit had little effect on ATPase activity. However, ATP hydrolysis-linked proton translocation was lost, by modification of both γ Cys-205 and c Cys-42 by bulky reagents such as 5,5′-dithio-bis (2-nitro-benzoic acid) or benzophenone-4-maleimide. In both ECF1 and ECF1F0 containing a Cys at γ 205 and a Cys in the ε subunit (at position 38 or 43), cross-linking of the γ to the ε subunit was induced in high yield by Cu2+. No cross-linking was observed in hybrid enzymes in which the Cys was at position 10, 65, or 108 of the ε subunit. Cross-linking of γ to ε had only a minimal effect on ATP hydrolysis. The reactivity of the Cys at γ 205 showed a nucleotide dependence of reactivity to maleimides in both ECF1 and ECF1F0, which was lost in ECF1 when the ε subunit was removed. Our results show that there is close interaction of the γ and ε subunits for the full-length of the stalk region in ECF1F0. We argue that this interaction controls the coupling between nucleotide binding sites and the proton channel in ECF1F0.

F 1 F 0 type ATP synthases play a key role in oxidative phosphorylation and photophosphorylation. The F 1 , which can be detached from the F 0 and studied separately, is a complex of five different types of subunits called ␣, ␤, ␥, ␦, and ⑀ that are present in the molar ratio 3:3:1:1:1. The F 0 part in the Esche-richia coli enzyme is composed of three different subunits: a, b, and c in the molar ratio 1:2:10 -12 (1)(2)(3)(4). Electron microscopy first showed that the ␣ and ␤ subunits are arranged hexagonally and alternate around a central cavity in which the ␥ subunit is located (5)(6)(7). Biochemical studies place the ⑀ subunit (nomenclature for the E. coli enzyme) at the bottom of the ␣ 3 ␤ 3 ␥ core complex (4,8) in the stalk region, which is a 40 -45-Å-long structure that links the F 1 to the F 0 part (9,10).
The recently published high resolution structure of a major part of the beef heart F 1 molecule confirms the above-described arrangement of the ␣, ␤, and ␥ subunits and adds important details (11). In particular, it shows the ␥ subunit arranged with a long C-terminal ␣-helix extending from the top of the ␣ and ␤ subunits into the stalk region. A shorter N-terminal ␣-helix is also present, running from the catalytic site region into the stalk region. These two ␣ helices form a coiled coil. A third short ␣-helix of the ␥ subunit (residues 83-99 in the E. coli sequence) is inclined at about 45°to the two larger helices at the bottom of the F 1 as it becomes the stalk. Approximately half of the ␥ subunit is unresolved in the structure, presumably because it is disordered in the crystal form.
Recently, we observed cross-linking between a Cys introduced at position 44 of the polar loop of the c subunits and a site or sites on the ␥ subunit in the region between residues 202 and 230 (12). This result implies that the ␥ subunit extends the full length of the stalk to interact with the F 0 part in the intact ATP synthase. Tentatively, we concluded that this cross-linking involved a Cys to Tyr covalent bond (12). There are four Tyr residues in the ␥ subunit between residues 202 and 230, i.e. Tyr-205, Tyr-207, Tyr-222, and Tyr-228.
The other subunit known to extend from the ␣ 3 ␤ 3 domain to the F 0 subunits is the ⑀ subunit. We have previously obtained a structure for the ⑀ subunit by NMR (13), which shows this stalk-forming subunit arranged as two domains, a 10-stranded ␤-sandwich structure formed by the N-terminal 85 residues and an ␣-helix-loop-␣-helix structure of the C-terminal approximately 50 residues. In ECF 1 F 0 , the C-terminal domain interacts with the ␣ and ␤ subunits (8, 14 -16). The N-terminal domain is positioned below this, attached to the first of the two ␣ helices of the C-terminal domain at its top (13) and to the c subunits of the F 0 part via the bottom of the ␤ sandwich in an interaction that involves residues Glu-31 (17,18) and possibly His-38 (19). Our recent cross-linking and protease protection studies have established that one face of the ␤-sheet sandwich of the ⑀ subunit interacts with the ␥ subunit (16). In one set of experiments, cross-linking was obtained from Cys-10 of ⑀ to a region of the ␥ subunit around residue 228. There was also cross-linking from a Cys placed at either residue position 38 or 43 of ⑀ to a region of the ␥ subunit between residues 202 and approximately 230. This is the same region of ␥ that was identified in preliminary experiments as involved in binding to the c subunits (12).
To explore the interactions of the region of the ␥ subunit between residues 202 and 230 in more detail, we have now converted ␥ Tyr-202 to a Cys. This introduced Cys is shown to cross-link by disulfide bond formation with Cys residues in both the ⑀ subunit and in the polar loop of the c subunits. Functional consequences of such cross-links are described.

EXPERIMENTAL PROCEDURES
Strains and Plasmid Construction-The M13mp18 template used consisted of a 1.4-kilobase uncG-containing SmaI-EcoRI fragment as described by Aggeler and Capaldi (20). The oligonucleotide used for the mutagenesis was ATAAATCCTGGGATTGCCTCTACGAACCCGATC where the first underlined G changes a TAC codon (Tyr) to TGC (Cys), creating the ␥Y205C mutation, and the CTC codon replaces CTG, retaining Leu at position ␥206. A RsaI site is eliminated at position 8393 of the unc operon (numbering according to Walker et al. (21)), which is diagnostic of the presence of the mutation. The mutant phage was identified by RsaI digestion and transferred into pRA100 (14). The final plasmid containing the mutation ␥Y205C was called pCT100. Mutants cA39C, cQ42C, cP43C, and cD44C were obtained from Dr. Robert Fillingame and are described elsewhere (18). For construction of the double mutants ␥Y205C/cA39C and ␥Y205C/cQ42C, plasmids pYZ203 and pYZ217 were isolated from the strains expressing mutants cA39C and cQ42C, respectively. The 5.8-kilobase pair XhoI/NsiI fragment of plasmid pCT100 containing the uncG gene with the ␥Y205C mutation was used to replace the corresponding region of plasmids pYZ203 and pYZ217. These new plasmids are called pSW101 and pSW102. Strain AN888 (uncB ϩ Mu::416, argH, entA, nalA, recA) (14) was transformed with plasmids pCT100, pSW101, pSW102, and pYZ217 for expression of mutant enzyme. ⑀ subunit mutants were obtained, and the ⑀ subunit was isolated as described elsewhere (13,14,16).
Labeling ECF 1 and ECF 1 F 0 ATPases with Maleimides-ECF 1 1 and ECF 1 F 0 were isolated from various mutants and from AN1460 as a control for biochemical studies, as described in Wise et al. (22) and Aggeler et al. (23), respectively. For reaction of ECF 1 with maleimides, enzyme was passed through two consecutive centrifuge columns (Sephadex G-50, fine, 0.5 ϫ 6 cm equilibrated with 50 mM MOPS, pH 7.0, 0.5 mM EDTA, 10% glycerol). ECF 1  ECF 1 F 0 was reconstituted in vesicles of egg lecithin by mixing detergent-dissolved ECF 1 F 0 with egg lecithin solution in a ratio of 1:2 (w/w ECF 1 F 0 :lecithin) and passing the mixture through a 1.0 ϫ 26-cm Sephadex G-50 (medium) column equilibrated with a buffer containing 50 mM MOPS, pH 7.5, 5 mM MgSO 4 and 10% glycerol. The ECF 1 F 0 ATPase (0.5-0.7 mg/ml) was labeled with NEM, TFPAM-3, or CM as for ECF 1 (above) but using 10 mM DTT to quench excess maleimide. The time course of labeling of ECF 1 F 0 with CM was conducted by first incubating the enzyme with 2 mM ATP or 2 mM AMP-PNP in column buffer at room temperature for 30 min and subsequent addition of 25 M CM. At various time points, 10 mM DTT was added to the corresponding aliquots to stop the reaction.
Hybrid and Double Mutant F 1 F 0 Preparation-ECF 0 from wild-type and c subunit mutant strains was prepared by KSCN extraction of purified ECF 1 F 0 reconstituted into vesicles of egg lecithin as described previously (19). These ECF 0 -containing membranes were centrifuged at 150,000 ϫ g for 30 min at 4°C in a Beckman TLA100.2 rotor and resuspended at 1 mg/ml in Buffer A (50 mM MOPS, pH 7.0, 10% glycerol, 2 mM MgC1 2 ) with 1 mM DTT and 5 mM ATP. ECF 1 from the mutant ␥Y205C (1 mg/ml) was added, and samples were incubated for 10 h at 22°C. Excess unbound ECF 1 and reducing agent were removed by washing the membranes twice by centrifugation at 150,000 ϫ g for 30 min at 4°C. Samples were resuspended in Buffer A to a final concentration of 1 mg/ml with or without DTT depending on the experiment to be performed. ECF 1 F 0 containing a Cys at ␥ 205 and one of the ⑀ subunit mutations reconstituted as above was pelleted and resuspended in a buffer containing 10 mM Hepes, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , and 10% glycerol (Buffer B).
CuSO 4 /1,10-Phenanthroline and DTNB-induced Cross-linking-Hybrid enzyme complexes, along with ECF 1 F 0 isolated from double mutants, that contained the subunit c mutations were suspended at 1 mg/ml in Buffer A and then reacted with CuSO 4 /1,10-phenanthroline (30 M) for 1 h at 22°C. Hybrid forms of ECF 1 were cross-linked by incubating ECF 1 * (ECF 1 from which both the ␦ and ⑀ have been removed) (16) from the mutant ␥Y205C with a 10-fold excess of pure ⑀ subunit in 50 mM MOPS, pH 7.0, 5 mM MgC1 2 , 10% glycerol, before adding CuCl 2 to a final concentration of 200 M. Hybrid ECF 1 F 0 complexes containing ⑀ subunit mutations were suspended at 1 mg/ml in the Buffer B and reacted with 200 M CuCl 2 . The reaction was quenched in each case by the addition of 10 mM EDTA. The reaction mixtures were assayed immediately for ATPase activity or frozen at Ϫ20°C for subsequent SDS-polyacrylamide gel electrophoresis (PAGE). DTNB cross-linking was carried out by incubating ECF 1 F 0 at 1 mg/ml in Buffer A with 200 M DTNB for 1 h at 22°C. The DTNB-treated samples were divided into two aliquots, and 20 mM DTT was added to one, before incubating both for 30 min at 22°C. Samples were then assayed for ATPase activity and ACMA fluorescence quenching and subjected to SDS-PAGE. ECF 1 F 0 was reacted with 200 M BM following the same procedure used for DTNB. The reaction was stopped after 1 h by a 30-min incubation with 20 mM DTT at room temperature.
ACMA Fluorescence Quenching-Assays of ACMA fluorescence quenching were performed essentially as described in Aggeler et al. (19). ECF 1 F 0 was reconstituted into egg lecithin vesicles as described before (12), except that 0.75% sodium deoxycholate was used in the column buffer, and 2 mg/ml of egg lecithin was used in the reconstitution step. These vesicles were collected by centrifugation and resuspended to a protein concentration of 20 g/ml in 10 mM Hepes, pH 7.5, 100 mM KCl and 2 mM MgCl 2 , and then valinomycin (3.6 M), ACMA (1 M), ATP (1 mM), and nigericin (3.6 M) were added sequentially. ACMA fluorescence was measured at 480 nm with an excitation wavelength of 410 nm in an SLM8000 fluorometer.
Other Methods-Samples for SDS-PAGE were supplemented with 20 mM NEM and incubated for 30 min at 22°C prior to addition of one-half volume of dissociation buffer (10% SDS, 100 mM Tris, pH 6.8, 30% glycerol, and 0.03% bromphenol blue) with or without reducing agent as indicated. Subunits were separated by electrophoresis on 10 -22% polyacrylamide gradient gels (24). Protein bands on SDS-PAGE were visualized by staining with Coomassie Brilliant Blue R according to Downer et al. (25). Proteins were transferred from the gel to polyvinylidene difluoride (Millipore Corp.) as described previously (12). The mouse monoclonal antibody to ␥, anti-␥ III , was characterized previously (26). The rabbit antiserum to subunit c was described by Girvin et al. (27). Immunodetection was by the alkaline phosphatase method. ATPase activities were performed according to Lötscher et al. (28). 500 M dicyclohexylcarbodiimide was used in inhibitor studies because of the large excess of phospholipid. Protein concentrations were determined by the BCA protein assay from Pierce. Quantitation of [ 14 C]NEM incorporated into ␥ Cys-205 was carried out as described previously (e.g. Ref. 15). 1 and ECF 1 F 0 Isolated from the Mutant ␥Y205C-ECF 1 isolated from the mutant ␥Y205C contained ␣, ␤, ␥, ␦, and ⑀ subunits in the same relative amounts as wild-type strain, whereas ECF 1 F 0 isolated from the mutant had the same subunit composition as wild type based on SDSpolyacrylamide gel electrophoresis (see later).

Characterization of ECF
The ATPase activity of ECF 1 isolated from the mutant ␥Y205C was around 3.5-fold higher than for wild-type enzyme (Table I). This is due to an altered binding of the inhibitory ⑀ subunit, as demonstrated by the concentration dependence of the inhibition of ECF 1 * (16) by purified ⑀ subunit. ECF 1 * from the mutant ␥Y205C showed half-maximal inhibition at a concentration of 16 nM of added pure ⑀ subunit, compared with only 9 nM for ECF 1 * from wild type (result not shown). ATPase activity was also measured in ECF 1 F 0 isolated from the mutant ␥Y205C. In the intact ATP synthase, the rate of ATP hydrolysis was essentially the same as that of wild-type enzyme (Table I).
ECF 1 F 0 isolated from the mutant ␥Y205C showed ATP hydrolysis linked to proton translocation, but this was somewhat less efficient than observed for wild-type enzyme under equivalent conditions (see below), implying that the mutation ␥Y205C caused some disruption of the F 1 -F 0 interface. ECF 1 F 0 from the mutant also showed a significantly lower sensitivity to dicyclohexylcarbodiimide than wild-type enzyme ( Table I).
Reaction of ␥ Cys-205 with Maleimides under Different Nucleotide Conditions-The Cys introduced at residue 205 of the ␥ subunit was reactive to a variety of maleimides in both ECF 1 and ECF 1 F 0 isolated from the mutant ␥Y205C. As shown in Table I, reaction of ␥ Cys-205 in ECF 1 with CM or TFPAM-3 (without photolysis) resulted in an increase in activity such that with these bulky maleimides, the modified enzyme had the same activity as ⑀-free ECF 1 (from either the mutant ␥Y205C or wild-type enzyme). After reaction of ECF 1 * from the mutant ␥Y205C with CM to modify the introduced Cys, there was essentially no inhibition of ATPase activity on addition of a large excess of purified ⑀ subunit. The Cys at residue 205 was also reactive to maleimides in ECF 1 F 0 , but in this case, the modification did not activate the enzyme significantly (Table I).
The rates of incorporation of [ 14 C]NEM and the fluorescent maleimide CM into ␥ Cys-205 were compared under different nucleotide conditions. In ECF 1 and ECF 1 F 0 , the rate of modification of this Cys by both reagents was significantly faster when ATP (as AMP-PNP ϩ Mg 2ϩ ) was bound in catalytic sites than with ADP present (generated on the enzyme by adding ATP ϩ Mg 2ϩ and allowing turnover for 30 min). With [ 14 C]NEM, there was twice as much reagent incorporated into ␥ Cys-205 within 2 min in AMP-PNP ϩ Mg 2ϩ compared with in ADP ϩ P i ϩ Mg 2ϩ (result not shown). Fig. 1A shows the time course of reaction with CM in AMP-PNPϩMg 2ϩ (left side) compared with ADP ϩ P i ϩMg 2ϩ (right side), whereas Fig. 1B summarizes the reactivity of ␥ Cys-205 in ECF 1 , ECF 1 * and ECF 1 F 0 . In ECF 1 *, there was no nucleotide dependence in the reactivity of ␥ Cys-205 to CM. With ADP ϩ P i ϩMg 2ϩ bound, the rate of modification was as fast as in AMP-PNPϩ Mg 2ϩ .
Studies with Reconstituted ECF 1 F 0 Establish the Proximity of ␥ Cys-205 to the Polar Loop Region of the c Subunits-F 1 purified from the ␥ subunit mutant Y205C was reconstituted with membrane vesicles containing F 0 isolated from the c subunit mutants cA39C, cQ42C, cP43C, and cD44C (see "Experimental Procedures"). Cross-linking within the reconstituted ECF 1 F 0 complexes was induced by the addition of Cu 2ϩ (1,10phenanthroline). Fig. 2 shows the results obtained using a hybrid enzyme ECF 1 (␥Y205C)F 0 (cQ42C). The oxidizing reagent generated a cross-link between the ␥ subunit and c subunit of apparent M r 38,000 that was readily observed in Coomassie Blue-stained gels. Based on the disappearance of the ␥ subunit, the yield of cross-linked product was 95%. No crosslinked product involving ␥ and c was observed in ECF 1 F 0 from the single mutant ␥Y205C or in ECF 1 F 0 from the single mutant cQ42C after Cu 2ϩ treatment. The M r 38,000 product was lost and the ␥ subunit (monomer) reappeared in the gel profile when Cu 2ϩ -treated samples were incubated with DTT prior to electrophoresis (result not shown), confirming that the crosslink is a disulfide bond between the introduced Cys in ␥ and that in the c subunit.
As shown by the Western blot in Fig. 2B, there was also Cu 2ϩ -induced generation of the M r 38,000 cross-linked product of ␥ and c subunits in the hybrid mutants ECF 1 (␥Y205C)-F 0 (cP43C) and ECF 1 (␥Y205C)F 0 (cD44C). This product was lost on the addition of DTT with both mutants. No cross-linking between the ␥ and c subunits was obtained using the hybrid enzyme ECF 1 (␥Y205C)F 0 (cA39C) (result not shown). There were additional cross-linked products in small amounts con-  taining ␥ and c, particularly in the mutant ECF 1 (␥Y205C)F 0 -(cD44C) (see Fig. 2B), some of which did not disappear with DTT treatment. These probably involve the Cys on the c subunit with remaining Tyr residues in the ␥ subunit region 205-228, particularly Tyr-207. One band migrating just above the main ␥-c product could be a cross-link between ␥ and two c subunits, one via a Cys-Cys linkage involving Cys-205, the second a Cys-Tyr linkage to Tyr-207. The Western blotting data in Fig. 2B show that with all three hybrid mutants, Cu 2ϩ also induced considerable subunit c dimer formation. Dimers of subunit c have been observed in the studies of Zhang and Fillingame (18), as well as in our previous studies using mutants with Cys residues in the polar loop region of the c subunit (12).
ATPase activities were measured before and after crosslinking. For the mutant ECF 1 (␥Y205C)F 0 (cQ42C), a 95% crosslinking of ␥ to c caused at most a 30% inhibition of ATP hydrolysis. This inhibition is most likely partly due to subunit c dimer formation, because there was a similar loss of ATPase activity when ECF 1 F 0 from the single mutant cQ42C was reacted with Cu 2ϩ (result not shown).
Cross-linking of the ␥ subunit to a c subunit could also be induced by adding DTNB instead of Cu 2ϩ . Fig. 3 shows data obtained with the double mutant ␥Y205C:cQ42C. In the experiment in Fig. 3A, the yield of ␥-c cross-linked product was 50% based on the disappearance of ␥ (lane 2), compared with a 95% cross-linking yield in the double mutant when Cu 2ϩ was used (lane 3). Note that a small amount of cross-linking of ␥ to c occurs in the absence of DTNB, probably due to oxidation reactions during sample preparation and incubations. Western blotting results in Fig. 3B confirm the cross-linking and show the greater selectivity of DTNB in generating the ␥-c crosslinked product over the subunit c dimer even under conditions where the ␥-c product obtained with Cu 2ϩ is in nearly the same yield as with DTNB. We assume that at the levels of DTNB used, Cys residues in the c subunit are modified more rapidly than the Cys at position 205 in the ␥ subunit. With most or all of the Cys of c modified, disulfide bond formation between these sites is prevented. One of the DTNB modified c subunit monomers then reacts with unmodified ␥ Cys-205 to generate the ␥-c cross-linked product. Table II lists the ATPase activities of the mutant after reaction with DTNB. Reaction of the double mutant with DTNB to generate around a 50% yield of cross-linked product ␥-c caused an approximately 10 -15% reduction in ATP hydrolysis. This compares with an approximately 40% inhibition of ATPase activity for ECF 1 F 0 from the single mutant cQ42C but no inhibition of ECF 1 F 0 from the mutant ␥Y205C after the identical DTNB treatment. Apparently, in the double mutant, the modification of both ␥ Cys-205 and c Cys-42 in some enzyme molecules compensates for the inhibition due to DTNB reaction with c Cys-42 alone.
A Cys at Position 205 of the ␥ Subunit Forms Disulfide Bonds with Cys Residues Introduced at Positions 38 or 43 of the ⑀ Subunit-In the course of our recent studies, we have made several mutants containing a Cys in the ⑀ subunit (14,16). These include ⑀S10C, ⑀H38C, ⑀T43C, ⑀S65C, and ⑀S108C. ECF 1 * (the ␣ 3 ␤ 3 ␥ complex) isolated from the mutant ␥Y205C was combined with each of the above-listed ⑀ subunit mutants to examine the proximity of Tyr-205 to the ⑀ subunit. In each case, the mutant ⑀ subunit was added in a 10-fold molar excess to ensure full occupancy of the ⑀ binding site. Samples were treated with 200 M Cu 2ϩ to induce disulfide bond formation. This led to cross-linking of ⑀ to the ␥ subunit in the hybrids made with ⑀H38C and ⑀T43C but not with ⑀S10C, ⑀S65C, or ⑀S108C (result not shown). This cross-linking of ␥ to ⑀ had little effect on the rates of ATP hydrolysis.
Cross-linking studies were also conducted in ECF 1 F 0 . For these experiments, ECF 1 F 0 from wild type was stripped of F 1 and these F 0 -containing membranes were then reconstituted with the ␣ 3 ␤ 3 ␥ complex from the mutant ␥Y205C, along with purified ␦ subunit and one of the different ⑀ subunit mutants. After the reconstitution step, membranes were separated from excess ␦ and ⑀ subunits prior to addition of Cu 2ϩ . There was Cu 2ϩ -induced cross-linking of the ␥ to the ⑀ subunit in hybrids containing the mutations ⑀H38C (maximum yields 50%) and ⑀T43C (yields as high as 90%) but not with the other ⑀ subunit mutants. As in ECF 1 , the cross-linked product ␥-⑀ ran very close to the ␤ subunit in SDS-PAGE and was best observed in Western blotting experiments, such as in Fig. 4 (A and B). The monoclonal antibody anti-␥ III was used for analysis (Fig. 4A). This mAb, like the others we have obtained to the ␥ subunit, reacted close to Tyr-205 (see Ref. 29), giving only a weak reaction with the ␥-⑀ cross-linked product. As a result, high concentrations of the mAb had to be used and mAb ␥ III then cross-reacted with the ␣ subunit. Nevertheless, Fig. 4 (A and B) shows clearly the disappearance of ␥ and ⑀ and appearance of the ␥-⑀ cross-linked in ECF 1 F 0 containing ␥ Cys-205 and ⑀ Cys-38 (lanes 3) or ⑀ Cys-43 (lanes 4). The yield of cross-linking was estimated from Coomassie Brilliant Blue-stained gels. Fig.   FIG. 2. Cross-linking between the ␥ subunit of ECF 1 ␥Y205C  and the c subunit of ECF 0 mutants cQ42C, cP43C and cD44C 4C shows data for Cu 2ϩ -induced cross-linking of ECF 1 F 0 containing the ␥Y205C and ⑀T43C mutations. The yield of crosslinking in this experiment, calculated from the change in area of the ␥ subunit band using the a subunit band as a control, was 55%. This was accompanied by a 20% loss of ATPase activity.
Effect of Chemical Modification of ␥ Cys-205 and Crosslinking of ␥ to ⑀ or c Subunits on Proton Pumping-ECF 1 F 0 from the ␥Y205C mutant or the mutant cQ42C showed a reduced proton pumping activity compared with wild-type enzyme, when measured by the ACMA fluorescence quenching assay (compare Fig. 5, traces A and B). Modification of the introduced Cys-205 with several maleimides (result not shown) or by DTNB reduced the levels of proton pumping by the mutant further. With DTNB reaction, proton pumping activity was partly recovered on adding DTT (Fig. 5A). Maleimide or DTNB modification of ECF 1 F 0 isolated from the mutant cQ42C also had a small effect on the proton pumping activity (Fig. 5B), whereas DTNB modification of the double mutant ␥Y205C: cQ42C caused essentially full loss of ATP-driven proton pumping (Fig. 5C), which was also mostly recovered on DTT addition. This essentially full inhibition occurred with around a 50% yield of cross-linking of ␥ to c, suggesting that it was the chemical modification of both sites in the complex rather than the cross-linking that was responsible. As shown by Fig. 5 (trace D), reaction of the mutant ␥Y205C:cQ42C with benzophenone maleimide likewise caused full loss of proton pumping activity. This reagent reacts with the introduced Cys residues but does not induce disulfide bond formation. The proton pumping activity of benzophenone maleimide-modified enzyme was not recovered upon the addition of DTT.
Attempts were made to measure the proton pumping activity after Cu 2ϩ treatment of various double mutants including both the Cys at ␥ 205 and mutations in the ⑀ or c subunits described above. The combination of mutations ␥Y205C and ⑀H38C was found to abolish ATP-driven proton pumping, whereas each mutation individually retained this activity. Cu 2ϩ treatment to induce cross-linking greatly reduced the proton pumping activity of all mutants tested under conditions where ATPase activity was retained. However, this treatment inhibited proton pumping activity to 80% even in wild-type enzyme, an inhibition that was only partly regained by incubation with DTT. As a consequence, the effect of cross-linking on proton pumping was not pursued further. DISCUSSION The studies described here extend our original finding (12) that oxidizing conditions generate a covalent cross-link between a Cys introduced at position 44 of the c subunit of the F 0 and the ␥ subunit of the F 1 part of ECF 1 F 0. Here Tyr-205, one of four Tyr residues in the short stretch of the ␥ subunit identified as involved in the cross-linking reaction with the c subunits, has been replaced by a Cys. ECF 1 from the mutant,  4. Cross-linking between ␥Y205C and sites on the ⑀ subunit in ECF 1 F 0 . A, Western blot analyzed with mAb anti-␥ III using a 12% SDS-polyacrylamide mini gel to separate the ␥-⑀ product from ␣ and ␤ subunits. This mAb also had significant reaction with the ␣ subunit under the conditions used. The reactivity with the ␥ subunit is diminished in the cross-linked product, probably because of steric effects, and so the Western blot was heavily over-exposed. B, Western blot analyzed with the anti-⑀ antibody after identical Cu 2ϩ treatments. Lane 1, wild-type enzyme; lane 2, ⑀S10C; lane 3, ⑀H38C; lane 4, ⑀T43C. C, density scanning profile of ECF 1 F 0 reconstituted with the ␥Y205C, ␣ 3 ␤ 3 ␥ complex, and ⑀T43C. The dotted line shows untreated enzyme. The solid line is after treatment with Cu 2ϩ . The region of the SDS-polyacrylamide gel (10 -22%) between the ␥ and a subunit is shown to demonstrate the disappearing gradient of the ␥ subunit by cross-linking to the ⑀ subunit.
␥Y205C, had an increased ATPase activity due to an altered affinity of the ␣ 3 ␤ 3 ␥ core complex for the inhibitory ⑀ subunit, but the ATPase activity of the intact ECF 1 F 0 was not altered compared with wild type by the mutation. The Tyr 3 Cys change also caused a small reduction in efficiency of ATPdriven proton translocation when measured by the ACMA quenching assay, but the ECF 1 F 0 clearly retained this coupling function.
A key finding of the present study is that the Cys in ␥ at position 205 can be cross-linked to Cys residues introduced in the c subunits at positions 42, 43, or 44 in essentially full yield by Cu 2ϩ treatment and in good yield by reaction with DTNB. This establishes that the ␥ subunit extends the full 40 -45 Å of the stalk. The other subunit now known to extend the entire length of the stalk is the ⑀ subunit (13). In addition to covalent linkage to the c subunits, ␥ Cys-205 could also be cross-linked to the ⑀ subunit via Cys residues in this subunit at positions 38 and 43 (but not at positions 65 or 108). Residues 38 and 43 are at one end of the 10-stranded ␤ sandwich structure constituted by the N-terminal domain of the ⑀ subunit (13). This end has been shown to interact with the c subunit oligomer by a combination of genetic and cross-linking experiments (17,18).
The fact that ␥ Cys-205 can interact with Cys introduced into either the c subunits or the ⑀ subunit not only indicates the proximity of these sites but also emphasizes the dynamic character of the complex. The ATP synthase presumably switches between several conformations during energy coupling. In this regard, the observed nucleotide dependence of the reactivity of ␥ Cys-205 to maleimides is interesting. This site reacted significantly faster with maleimides in both isolated ECF 1 and in ECF 1 F 0 in the ATP state (with Mg AMP-PNP bound) than in the ADP (Mg 2ϩ ) state. We have previously observed nucleotidedependent conformational changes in the ␥ subunit near catalytic sites (by cross-linking and fluorescence changes of probes attached at Cys introduced at position 8 (30)) at Cys-87 (by changes in reactivity to maleimides (31)) and at position 106 (from fluorescence changes of CM bound to a Cys at this site (30)). The results for ␥ Cys-205 now establish that the conformational change occurring with ATP hydrolysis is propagated by the ␥ subunit from the catalytic sites to the interface of the F 1 with F 0 .
As with the conformational changes detected at other parts of the ␥ subunit (30,31), the nucleotide dependence of the reaction of Cys-205 was lost on removal of the ⑀ subunit. ECF 1 from which the ⑀ subunit has been removed is a highly active ATPase, and therefore, as we have discussed before (30,31), the conformational changes observed at residues 8, 87, 106, and now 205 cannot be necessary for ATP hydrolysis. Rather, they likely represent structural changes that are part of the energy coupling within the complex that are in addition to rotation of the ␥ and ⑀ subunits that occurs during the cooperative functioning of catalytic sites (32)(33)(34).
The covalent linkage of ␥ via Cys-205 to the c subunits at positions 42, 43, or 44 proved to have very little effect on ATPase activity. A pattern of effects of covalent cross-linking of subunits to one another on enzyme function is emerging. Crosslinking of ␥ or ⑀ to ␣ or ␤ subunits in our studies has in all cases fully inhibited ATPase activity (14 -16, 35) (also see Ref. 36). In contrast, cross-linking of ␥ to ⑀ (14, 16) and now of ␥ to c subunits did not greatly reduce ATP hydrolysis. The implication is that the rotations or translocations of subunits that are a part of the cooperativity of catalytic sites must involve the ␣ 3 ␤ 3 domain moving relative to ␥ and ⑀ plus the c subunit oligomer. This could occur by the ␣ 3 ␤ 3 domain rotating relative to a fixed unit of the rest of the ATP synthase or the ␥, ⑀, and c subunit oligomers rotating within a scaffold of the ␣ 3 ␤ 3 domain plus the ␦, a, and b subunits.
It is not clear at present whether covalent cross-linking of ␥ to the c subunit oligomer alters proton pumping. It was not possible to assess this issue when Cu 2ϩ oxidation was used to generate cross-links. After DTNB-induced cross-linking, there was complete loss of proton pumping without concomitant loss of ATPase activity, implying an uncoupling of the two functions. However, it was the maleimide modification of both ␥ Cys-205 and the adjacent Cys in the c subunits that caused the effect, not the cross-linking reaction per se.  D). A-C, untreated control (thin line), DTNB treated (thick line), and DTNB-treated followed by reduction with DTT (dotted line). In D, benzophenone maleimide was used as the modifying reagent. Traces represent the results from a single experiment, but in each case equivalent results were obtained in several experiments.