Cross-linking of the δ Subunit to One of the Three α Subunits Has No Effect on Functioning, as Expected if δ Is a Part of the Stator That Links the F1 and F0 Parts of the Escherichia coli ATP Synthase*

A mutant of the Escherichia coliF1F0-ATPase has been generated (αQ2C) in which the glutamine at position 2 of the α subunit has been replaced with a cysteine residue. Cu2+ treatment of ECF1from this mutant cross-linked an α subunit to the δ subunit in high yield. Two different sites of disulfide bond formation were involved,i.e. between Cys90 (or the closely spaced Cys47) of α with Cys140 of δ, and between Cys2 of α and Cys140 of δ. Small amounts of other cross-linked products, including α-α, δ internal, and α-α-δ were obtained. In ECF1F0, there was no cross-linking between the intrinsic Cys of α and Cys140. Instead, the product generated between Cys2 of α and Cys140 of δ was obtained at near 90% yield. Small amounts of α-α and δ internal were present, and under high Cu2+ concentrations, α-α-δ was also formed. The ATPase activity of ECF1 and ECF1F0 was not significantly affected by the presence of these cross-links. When Cys140 of δ was first modified with N-ethylmaleimide in ECF1F0, an α-δ cross-link was still produced, although in lower yield, between Cys64 of δ and Cys2 of α. ATP hydrolysis-linked proton pumping of inner membranes from the mutant α2QC was only marginally affected by cross-linking of the α to the δ subunit. These results indicate that Cys140 and Cys64 of the δ subunit and Cys2 of the α subunit are in close proximity. This places the δ subunit near the top of the α-β hexagon and not in the stalk region. As fixing the δ to the α by cross-linking does not greatly impair either the ATPase function of the enzyme, or coupled proton translocation, we argue that the δ subunit forms a portion of the stator linking F1 to F0.

A proton-translocating ATPase is found in the plasma membrane of bacteria, inner membrane of mitochondria, and thylakoid membrane of chloroplasts. This enzyme complex can catalyze both ATP synthesis coupled to an electrochemical gradient and ATP hydrolysis driven by proton translocation. The enzyme from Escherichia coli (ECF 1 F 0 ) consists of two parts, a water-soluble F 1 part, composed of five different subunits ␣, ␤, ␥, ␦, and ⑀ in a molar ratio of 3:3:1:1:1, and the membrane bilayer integrated F 0 part, which is made up of three different subunits a, b, and c in a molar ratio of 1:2:10 -12 (1)(2)(3).
Electron microscopy studies have shown that the F 1 portion of the enzyme is separated from the F 0 by a narrow stalk of around 45 Å in length (4,5). The recently published high resolution structure of the bovine heart enzyme (MF 1 ) shows the ␣ and ␤ subunits arranged alternately in a hexagon around a central cavity (6). Within this cavity, the long C-terminal ␣-helix of the ␥ subunit extends the full length of the ␣ 3 ␤ 3 domain to protrude from the bottom of the structure. The shorter N-terminal ␣-helix forms a coiled-coil with the C-terminal helix and also extends from the bottom of the central cavity. A third short ␣-helix of the ␥ subunit is resolved in the structure, positioned at about 45°to the other two helices, along the bottom of the ␣-␤ hexagon. The remainder of the ␥ subunit (about half) was unresolved, but biochemical studies have shown that this subunit extends the full length of the stalk to interact with the c subunits of the F 0 . (7). The ⑀ subunit is also known to be part of the stalk region of the enzyme. This subunit has a two-domain structure (8,9). The C terminus is a helix-loop-helix that interacts with the ␣ and ␤ subunits (10,11), while the N terminus, a 10-stranded ␤-sandwich, interacts with the ␥ subunit along one side and the c subunits at its bottom (12)(13)(14).
Also necessary for the interactions of the F 1 portion of the enzyme with the F 0 are two other subunits, the ␦, which is the subject of this study, and the b subunit (15,16).
Reconstitution of a coupled F 1 F 0 is impaired by the removal of as few as four C-terminal residues from ␦ (17) and 5 Cterminal residues of oligomycin sensitivity conferring protein (18) (the equivalent of ␦ in MF 1 F 0 ). These C-terminal truncations have no effect on binding of these subunits to the F 1 part. Protease digestion of the b subunits also prevents rebinding of F 1 to F 0 (19). Evidence from cross-linking studies in CF 1 F 0 and MF 1 F 0 shows that the ␦ and b subunits are in close proximity. For example, Beckers et al. (20) were able to form a zero length cross-link between ␦ and the chloroplast counterpart of b (subunit I). Similarly, Belogrodov et al. (21) found that the b subunit of MF 1 F 0 cross-links to OSCP. Recent studies in this laboratory show that the ␦ subunit is required for binding of a water-soluble C-terminal fragment of subunit b to the core ECF 1 complex (22). Removal of the C-terminal 43 residues from ␦ prevented binding of this b subunit fragment.
Most structural models of F 1 F 0 ATPase place the ␦ and b subunits along with ␥ and ⑀ in the stalk region of the enzyme (1, 2, 23), but this seems unlikely given the relatively narrow dimensions of the stalk. The alternative is that the ␦ and b subunits form a separate connection between the F 1 and F 0 parts. In this regard, earlier studies implicating the N terminus of the ␣ subunit in binding the ␦ subunit to F 1 (24 -29) are important, because the N-terminal domain of ␣ is located at the top of the ␣-␤ hexagon and away from the stalk region.
To further investigate the role of the N terminus of the ␣ * 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. subunit in binding of ␦, we have introduced a Cys residue at position 2 (mutant ␣Q2C). The introduced Cys is shown to form disulfide bonds with intrinsic Cys residues in the ␦ subunit. The high yield of these linkages has allowed us to examine the effect that fixing the ␦ to a single ␣ subunit has on both ATPase activity and ATP hydrolysis-driven proton translocation.

EXPERIMENTAL PROCEDURES
Construction of the Mutant ␣Q2C and Isolation of ECF 1 and ECF 1 F 0 -A 657-base pair ClaI fragment containing the region encoding the N terminus of the ␣ subunit was isolated from pRA100 (30). The fragment was inserted into the AccI site of M13mp18 (New England Biolabs). Site-directed mutagenesis (31) was then used to replace the Gln at position 2 of the ␣ subunit with a Cys residue. The oligonucleotide used for this procedure is as follows: TGGAGCATGTGTCTGAAT-TCCA. The incorporation of the mutation was checked by restriction enzyme analysis as the presence of the mutation resulted in the loss of an SphI site. The mutation was subsequently inserted back into the unc operon by ligating a 212-base pair BfrI-XhoI fragment from M13mp18 into the corresponding restriction site on pRA100 to create the plasmid pIO1 containing the mutation ␣Q2C.
[ 14 C]NEM Incorporation into ECF 1 -For labeling experiments, ECF 1 was precipitated in 70% ammonium sulfate for 1 h at 4°C. The protein was collected by centrifugation at 10,000 ϫ g for 15 min at 4°C and resuspended in 50 mM MOPS, 1 pH 7.0, 0.5 mM EDTA, and 10% glycerol. The enzyme was then passed through two consecutive Sephadex G-50 centrifuge columns equilibrated in the same buffer, as described by Aggeler et al. (30). Mutant and wild-type ECF 1 (200 g) was labeled with [ 14 C]NEM (200 M) for 1 h at 22°C. The stoichiometry of labeling was determined relative to a control of a second aliquot of each sample that had been denatured and dissociated with 2% SDS prior to treatment with [ 14 C]NEM. Samples were electrophoresed on a 10 -18% SDSpolyacrylamide gel electrophoresis, and the radioactivity in individual bands was measured as before (37).
CuCl 2 Induced Cross-linking in ECF1 and ECF 1 F 0 -ECF 1 for crosslinking was diluted with 50 mM MOPS, pH 7.0, 10% glycerol to give a final concentration of 0.1 mM EDTA and then 2.5 mM MgCl 2 added, followed by 100 M CuCl 2 and the protein incubated for 1 h at 22°C. ECF 1 F 0 (1 mg) was reconstituted into egg-lecithin vesicles as described in Aggeler et al. (37). Enzyme complex collected from the Sephadex G-50 column was pelleted at 175,000 ϫ g for 30 min at 4°C in a Beckman TLA100.2 rotor. The pellets were resuspended in 50 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 , 2 mM DTT, and 10% glycerol and stored in liquid nitrogen. Prior to cross-linking experiments with CuCl 2 , the reducing agent was removed by pelleting the enzyme as above. The sample was then washed twice by successive resuspension and centrifugation steps in 50 mM MOPS, pH 7.0, 2 mM MgCl 2 , and 10% glycerol. Final resuspension was in the same buffer at a concentration of about 1 mg/ml. Cross-linking was carried out at 22°C for 1 h using concentrations of between 20 and 200 M CuCl 2 . All cross-linking reactions were stopped by the addition of 5 mM EDTA, and ATPase activities (38) were measured with and without incubation of the samples with 25 mM DTT for 2 h at 22°C.
Effect of NEM Labeling of ECF 1 F 0 on Formation of CuCl 2 -induced Cross-linking-Reconstituted ECF 1 F 0 (50 g), that contains 20 M DTT, was incubated with 300 M NEM for 1 h at 22°C. The excess NEM and DTT were then removed by centrifuging at 175,000 ϫ g, 4°C for 30 min in a Beckman TLA100.2 rotor. Samples were washed twice with 50 mM MOPS, pH 7.0, 2 mM MgCl 2 , 10% glycerol before cross-linking with 100 M CuCl 2 in the same buffer.
ACMA Quenching of Fluorescence in Reconstituted ECF 1 F 0 and Inner Membranes from Mutants and Controls-ECF 1 F 0 was reconstituted into egg lecithin vesicles as above, but using 0.75% deoxycholate and 2 mg/ml of lipid. Inner membranes (2 mg of protein) were prepared as described by Aggeler et al. (39) and resuspended in 50 mM Tris-HCl, pH 7.5, 5 mM MgSO 4 , and 10% glycerol. (1 mM dithioerythritol was added for storage in liquid nitrogen). Cross-linking was carried out with 20 M to 1 mM CuCl 2 for 1 h at 22°C. ACMA fluorescence quenching was measured with, and without, prior incubation of the cross-linked samples with 50 mM DTT for 1 h at 22°C. The samples (0.01 mg/ml) were resuspended in 10 mM Hepes, pH 7.5, 100 mM KCl, 5 mM MgCl 2 . To each assay was added 3.6 M valinomycin, 1 M ACMA, 0.5 mM NADH, 2 mM KCN, 2 mM ATP, and 3.6 M nigericin (39). ATPase activities and dicyclohexylcarbodiimide sensitivity from the above cross-linked inner membrane samples were measured and 100 g of protein from each sample was loaded on 10 -18% SDS-polyacrylamide gel electrophoresis. The protein was transferred to Immobilon membrane (Millipore) and probed using a monoclonal anti-␦ antibody (11). The blot was developed using enhanced chemiluminescence (NEN Life Science Products).
Other Methods-Protein concentrations were determined using the BCA protein assay from Pierce. 10 -18% SDS-polyacrylamide gels were used for the analysis of cross-linking products (40). Protein bands were visualized by staining with Coomassie Brilliant Blue R according to Downer et al. (41). 1 and ECF 1 F 0 from the Mutant ␣Q2C-As a preliminary to cross-linking, the reactivity of the Cys replacing Gln 2 of the ␣ subunit was examined in both ECF 1 and ECF 1 F 0 by labeling with the water-soluble maleimide [ 14 C]NEM. The five subunits of wild-type ECF 1 , ␣, ␤, ␥, ␦, and ⑀, contain 4, 1, 2, 2, and 0 Cys residues, respectively (42,43). Of these, only one of the Cys residues of ␦, previously identified as Cys 140 , and a Cys in one of the three copies of the ␣ subunit (probably Cys 90 ), are reactive to [ 14 C]NEM in the native complex (29).

Accessibility of Cys Residues in ECF
When ECF 1 from the mutant ␣Q2C was reacted in native form with [ 14 C]NEM, the labeling pattern obtained was similar to that of wild-type enzyme, i.e. only around 1 mol of reagent/ mol of F 1 was incorporated into ␣ subunits (Fig. 1). In SDS, all 5 Cys residues of each ␣ subunit (total of 15 per F 1 ) were reacted. These results show that Cys 2 is not available for reaction with NEM in ECF 1 . Cys 2 was also inaccessible in ECF 1 F 0 (result not shown). Unfortunately, the N-terminal approximately 20 residues of the ␣ subunits were not resolved in the structure determination of MF 1 (6). The lack of reactivity of the Cys at position 2 suggests interaction of this very N-terminal part of each ␣ subunit with the core complex, at least in the E. coli enzyme.
Cross-linking between Cys Residues in the Mutant ECF 1 -CuCl 2 treatment of ECF 1 from the mutant ␣Q2C led to high yield disulfide bond formation between ␣ and ␦ subunits ( Fig.  2A). On less heavily loaded SDS-polyacrylamide gels, two closely spaced bands could be seen. One of these migrates at the position of the ␣-␦ band obtained in wild-type enzyme (indicated as ␣*-␦), which has been established to involve Cys 140 of ␦ and an intrinsic Cys of ␣ (probably Cys 47 or Cys 90 ). Running just above this is the product involving Cys 2 and ␦. Several low yield cross-linked products were formed including ␣-␣, ␣-␣-␦, and ␦ internal. The identification of these products is based on size and on monoclonal antibody blotting (not shown). Disulfide bond formation within the ␦ subunit between Cys 64 and Cys 140 has been reported in wild-type ECF 1 and ECF 1 F 0 (44). This product stains poorly with Coomassie Blue but is clearly detected by monoclonal antibody binding (see later). Although not seen clearly in Fig. 2, CuCl 2 treatment also induces disulfide bond formation between two intrinsic Cys of the ␣ subunit (between Cys 47 and Cys 90 from the structure determination of MF 1 ), which shifts the migration of the ␣ subunit slightly. This product is also seen in wild-type ECF 1 after CuCl 2 treatment (e.g. see Watts et al. (14)).
Cu 2ϩ treatment to yield essentially full cross-linking between ␣ and ␦, as well as the other products mentioned above, reduced the ATPase activity of ECF 1  wild-type ECF 1 . The effect of fixing the ␦ subunit to one ␣ subunit on ATP hydrolysis is, therefore, minimal.
Cross-linking of ␦ to ␣ in ECF 1 F 0 -An ␣-␦ cross-link was also obtained in high yield by incubating ECF 1 F 0 from the mutant ␣Q2C with Cu 2ϩ . This product must be between Cys 2 of ␣ and an intrinsic Cys of ␦, because cross-linking via Cys 47 /Cys 90 does not occur in the intact ATP synthase (27)(28)(29). As with ECF 1 , small amounts of ␣-␣ and ␦ internal were also formed. The yield of ␣-␣ was always low relative to, and inversely related to the amount of, the ␣-␦ product. Also, the amount of ␣ subunits in all cross-linked products never appeared to exceed 1 mol/mol of enzyme. Finally, the yield of the internally cross-linked ␦ was greatest when the amount of ␣-␣ was highest. Taken together, these findings suggest that the same ␣ subunit is involved either in the ␣-␣ product or in the ␣-␦ product. Consistent with this interpretation, when high concentrations of Cu 2ϩ were used, a small amount of the product ␣-␣-␦ was generated. Such a product would then involve a disulfide bond from ␦ to Cys 2 of one ␣, and from Cys 90 or Cys 47 of this same ␣ to Cys 2 of a second ␣ subunit.
In one set of experiments, the cross-linking site(s) on the ␦ subunit for Cys 2 of ␣ was examined, taking advantage of the selective labeling of Cys 140 with NEM. As shown in Fig. 2C, the ␣-␦ cross-linked product was still formed (indicated by ␣-␦*), albeit at lower yields than in the unmodified enzyme after essentially 100% modification of Cys 140 with NEM. Therefore, it appears that the cross-linking can occur from Cys 2 of ␣ to both Cys 64 and Cys 140 of ␦. This is not surprising as the two intrinsic Cys of ␦ are very close, as discussed already.
Effect of Cross-linking of an ␣ to the ␦ Subunits on the Functioning of ECF 1 F 0 -Cross-linking between ␣ and ␦ in isolated ECF 1 F 0 in yields above 90% (based on the disappearance of ␦) caused a 15.2 Ϯ 9% (six different experiments) loss of ATPase activity, again the same level of inhibition that was obtained by Cu 2ϩ treatment of wild-type enzyme in which there was no cross-linking of ␣ to ␦.
Preliminary experiments using enzyme reconstituted into liposomes also showed that high levels of cross-linking of ␣ to ␦ in the mutant ␣Q2C gave only a partial inhibition of ATPdriven proton translocation as measured by the ACMA quenching assay (result not shown). This was confirmed and established more definitively by experiments using inner membranes prepared from the mutant strain. As shown in Fig.  3, CuCl 2 treatment gave high yield cross-linking of ␣ to ␦ based on monoclonal antibody blotting with a monoclonal antibody against the ␦ subunit. No ␣-␦ product was obtained with the equivalent treatment of membranes from wild-type cells. The ␣-␣ product is also formed (not shown), but in low yield at CuCl 2 concentrations below 100 M. A small amount of the ␦ internal product is also generated (Fig. 3). At higher concentrations of Cu 2ϩ , the yield of ␣-␦ product was more complete, but the band on gels became more smeared, possibly because of generation of the internal cross-link within ␣. There was also some ␣-␣-␦ product formed at high Cu 2ϩ concentrations. As expected, DTT treatment reversed the cross-linking of ␣-␦ (Fig.  3, lanes 4 -6), but the ␦ internal cross-link, once generated, is much more difficult to reverse, presumably because of limited access of the reductant to the disulfide bond once formed. Cross-linking of ␣ to ␦ in inner membranes reduced ATPase activity by around 20% (e.g. from 7.5 mol of ATP hydrolyzed per min/mg to 6.1 mol of ATP hydrolyzed per min/mg in the  Fig. 3), a loss of activity similar to that of inner membranes containing wild-type enzyme.
The effect of Cu 2ϩ treatment on ATP hydrolysis-driven proton translocation is shown in Fig. 4. At 20 M CuCl 2 , where most of the ␦ is cross-linked to ␣ (see Fig. 3), ATP hydrolysisdriven proton translocation was not greatly affected. Even at 200 M CuCl 2 , where small amounts of ␣-␣ and ␣-␣-␦ products were formed, there was substantial coupled proton translocation remaining. The loss of proton pumping function at higher Cu 2ϩ concentrations appeared proportional to the generation of the ␣-␣-␦ cross-link. This modification disrupts the F 1 F 0 interaction and makes the membranes leaky to protons, as shown by a concomitant loss of NADH-coupled proton translocation (result not shown). This reduction in proton pumping efficiency was reversed on addition of DTT. DISCUSSION The studies described here use a mutant ␣Q2C in which a Cys has been introduced into the ␣ subunit in a region previ-ously identified by protease digestion (24,25) and by genetic studies (26) to be important in ␦ subunit binding. In the presence of CuCl 2 , a high yield cross-link was obtained from Cys 2 of one of the ␣ subunits to intrinsic Cys of the ␦ subunit in both ECF 1 and ECF 1 F 0 . Evidence is presented that disulfide bonds are formed between the Cys at position 2 in ␣ and both Cys 140 and Cys 64 of the ␦ subunit.
We have recently obtained structural data on the ␦ subunit from NMR studies (45). This subunit has two domains. The N-terminal domain of residues 1-95 is organized as a six-␣helix bundle. The C-terminal domain of around 50 residues contains at least one ␣-helix but is partly disordered in the isolated ␦ subunit. A disulfide bond can be formed between Cys 64 and Cys 140 in the isolated ␦ subunit, as well as in ECF 1 and ECF 1 F 0 (44), placing these residues and, therefore, the two domains close together. The interaction of Cys 2 of ␣ with both intrinsic Cys of ␦ places the very N terminus of ␣ at the interface between the two domains of ␦. As the N terminus of ␣ is at, or very close to, the top of the F 1 and therefore a long way from the F 0 region (6), these data indicate a location of ␦ in ECF 1 F 0 on the outside of the F 1 and not in the stalk part seen in cryoelectron microscopy (4,5). A similar conclusion has been reached by Lill et al. (46). These authors have introduced Cys residues at several sites in the ␦ subunit of CF 1 and show that bifunctional reagents cross-link from these sites to the N-terminal parts of both the ␣ and ␤ subunits.
Lill et al. (46) were unable to examine the consequences of fixing an ␣ to the ␦ subunit as yields of their cross-links were relatively low. This is not the case here, as we have been able to obtain disulfide bond formation between the two subunits in yields greater than 90%. Our data show that fixing the ␦ to an ␣ subunit has minimal effect on either ATP hydrolysis or the coupling of proton translocation to this ATPase activity. This lack of effect is in contrast to the effects of cross-linking of either the ␥ or the ⑀ to ␣ or ␤ subunits, which in all cases causes full inhibition of activity, as summarized in Fig. 5. These crosslinking data and recent studies in other laboratories (47,48) point to a rotation of the domain formed by the ␥ and ⑀ subunits with respect to the ␣ 3 ␤ 3 domain. Our earlier electron microscopy experiments (49), which suggested rotation of the ␥ subunit or ⑀ subunit, but not the two together, are explained by the antibody used disrupting the ␥-⑀ interaction so that ⑀ remained fixed at one ␤ subunit. Our more recent studies (14,30) are all consistent with the ␥ and ⑀ subunits moving as one domain and further suggest that the mobile domain also includes the c subunit oligomer, which must then rotate relative to the a and b subunits of the F 0 part.
For efficient rotation to occur, the ␣ 3 ␤ 3 domain must be fixed relative to the a and b subunits by a stator. The fact that ATPase activity (and coupled proton pumping) occurs under conditions where the ␦ subunit is fixed to one of the ␣ subunits indicates that movements of ␣ 3 ␤ 3 relative to ␦ are not a part of the functioning of the enzyme. Therefore, the ␦ subunit is a good candidate for a part of the stator. We envision that the b subunit dimer extends from the F 0 to interact with the C domain of the ␦ subunit to complete this stator.
It is interesting that the Cys introduced at position 2 is unavailable for reaction with NEM in any of the three ␣ subunits. This can be explained for one ␣ subunit by its binding to the ␦ subunit. In one of the other two ␣ subunits, it appears that the very N terminus extends to a second ␣ and close to the region of Cys 47 /Cys 90 . This is based on the creation of the ␣-␣ and particularly the ␣-␣-␦ cross-linked product in ECF 1 F 0 , both of which appear to involve the same ␣ subunit that interacts with ␦.
In summary, the work presented here shows that one of the three ␣ subunits binds to the ␦ subunit in an interaction face that includes Cys 2 of ␣ and Cys 64 and Cys 140 of ␦. The interaction of ␦ differentiates one ␣ subunit from the other two. Crosslinking of these two subunits has little effect on enzyme functioning, consistent with the ␦ being fixed in the F 1 F 0 complex and a part of the stator that allows rotation of a ␥-⑀-c subunit domain during cooperative ATP hydrolysis and ATP synthesis.