Conformational changes in the Escherichia coli ATP synthase (ECF1F0) monitored by nucleotide-dependent differences in the reactivity of Cys-87 of the gamma subunit in the mutant betaGlu-381 --> Ala.

Cys-87, one of two intrinsic cysteines of the gamma subunit of the Escherichia coli ATP synthase (ECF1F0), is in a short segment of this subunit that binds to the bottom domain of a beta subunit close to a glutamate (Glu-381). Cys-87 was unreactive to maleimides under all conditions in wild-type ECF1 and ECF1F0 but became reactive when Glu-381 of beta was replaced by a cysteine or alanine. The reactivity of Cys-87 with maleimides was nucleotide-dependent, occurring with ATP or ADP + EDTA in catalytic sites, in the presence of AMP.PNP + Mg2+ but not with ADP + Mg2+ bound, whether Pi was present or not, and not when nucleotide binding sites were empty. Binding of N-ethylmaleimide had no effect, whereas 7-diethyl-amino-3-(4'-maleimidylphenyl)-4-methylcoumarin increased the ATPase activity of ECF1 more than 2-fold by reaction with Cys-87. In ECF1F0, these reagents inhibited activity. The nucleotide dependence of the reaction of Cys-87 of the gamma subunit depended on the presence of the epsilon subunit. In epsilon subunit-free ECF1, maleimides reacted with Cys-87 under all nucleotide conditions, including when catalytic sites were empty. These results are discussed in terms of nucleotide-dependent movements of the gamma subunit during functioning of the F1F0-type ATPase.

F 1 F 0 -type ATPases catalyze oxidative or photo-phosphorylation by using a transmembrane proton motive force to drive ATP synthesis (reviewed in Senior, 1988;Futai et al., 1989;Hatefi, 1993). In the reverse direction, these enzymes use ATP hydrolysis to generate a proton gradient that can be used in ion transport processes. The simplest F 1 F 0 -type ATPases are found in bacteria. The hydrophilic F 1 part of the Escherichia coli enzyme (ECF 1 ) 1 contains five different subunits in the stoichiometry ␣3, ␤3, ␥, ␦, and ⑀, whereas the membrane-integrated F 0 part (ECF 0 ) contains three different subunits in the molar ratio a1, b2, c10 -12.
As first demonstrated by electron microscopy studies (Tiedge et al., 1985;Gogol et al., 1989aGogol et al., , 1989b, the three ␣ and three ␤ subunits alternate in a hexagonal arrangement surrounding a central cavity containing the ␥ subunit. The recent high reso-lution structure of the beef heart F 1 (MF 1 ) confirms this arrangement and shows that the part of the ␥ subunit within the ␣ 3 ␤ 3 domain is in the form of two large ␣ helices, one provided by residues 1-45 (ECF 1 numbering system) and the other by residues 223-286 (Abrahams et al. 1994). A third short ␣ helix of the ␥ subunit has been resolved in the x-ray analysis (Abrahams et al. 1994). This short segment of residues 82-99, including an intrinsic Cys residue (Cys-87), binds to the so-called DELSEED region (residues 380 -386) of one of the ␤ subunits.
Recent evidence indicates that the ␥ subunit runs from within the ␣ 3 ␤ 3 domain of the F 1 through the stalk region that connects the F 1 to F 0 (Gogol et al., 1987;Lü cken et al., 1990) and binds to the c subunits of the F 0 that are a part of the proton channel (Watts et al., 1995). It is now generally agreed that energy coupling within the F 1 F 0 complex is by conformational changes involving the stalk-forming subunits, including the ␥ subunit (reviewed in Boyer, 1993;Capaldi et al., 1994). Previously, we have provided evidence of nucleotide-dependent conformational changes in the ␥ subunit around Cys residues site-directed into positions 8 and 106 of this subunit Capaldi, 1992, 1993;Capaldi, 1994a, 1994b). Here, we describe studies in which one of the two intrinsic Cys residues of the ␥ subunit is reacted with various maleimides in both ECF 1 and ECF 1 F 0 . This residue, shown to be Cys-87, is shielded in wild-type enzyme but becomes available for reaction in ECF 1 (and ECF 1 F 0 ) when Glu-381 of the ␤ subunit is replaced by a smaller side chain, e.g. by a Cys or an Ala. The interaction of the short ␣ helix of ␥ with the DELSEED region is shown to be nucleotide-dependent and, as with ATP hydrolysis-driven structural changes already observed at residues 8 or 106, requires binding of the ⑀ subunit.

EXPERIMENTAL PROCEDURES
Materials-CM and BM were obtained from Molecular Probes; Sephadex G-50 was purchased from Pharmacia Biotech Inc.; all other chemicals were of analytical grade and obtained from Sigma.
Plasmids and Bacteria Strains-Routine cloning was carried out in XL1-Blue and site-directed mutagenesis in CJ236 according to Kunkel et al. (1987). Mutant ATPase and ATP synthase was isolated from AN888 (unc -), transformed with unc operon containing plasmids.
The Cys residue at position 87 of the ␥ subunit was replaced with Ser by using M13mp18 that contained the 1.4-kb EcoRI/SmaI fragment  and the oligonucleotide GACCGTGGTTT-GAGCGGTGGTTTG. Successful mutagenesis was shown by testing for the newly created BsrBI restriction site. The mutation was incorporated in an unc operon-containing plasmid in two steps. (i) The 1.1-kb SfuI/EcoRI fragment from M13mp18 was inserted in the pBluescript derivative pRA13 (Aggeler et al., 1995). (ii) The 2.8-kb SstI/XhoI fragment of this plasmid was then introduced in pRA134 ( Aggeler et al., 1995), creating pRA149 with the mutation ␤E381C/␥C87S/⑀S108C.
The Glu in position 381 of the ␤ subunit was replaced with an Ala by using M13mp18 that contained the 1.01-kb NcoI insert, described in , and the oligonucleotide TTCTTCAGACAAT-GCATCCATACC (nucleotide A was introduced to obtain a new NsiI restriction site for analysis). The NcoI fragment was introduced in * This work was supported by National Institutes of Health Grant HL24526 (to R. A. C.). 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.
Preparation of ECF 1 and ECF 1 F 0 -ECF 1 was isolated by a modification of the method of Wise et al. (1981) described in Gogol et al. (1989b). The enzyme was precipitated for 1 h at 4°C in 70% (NH 4 ) 2 SO 4 , pelleted by centrifugation at 10,000 ϫ g for 15 min, and the protein then dissolved in 50 mM MOPS, pH 7.0, 0.5 mM EDTA, and 10% glycerol (v/v). Loosely bound nucleotides were removed by passing samples of enzyme through two consecutive centrifuge columns (Sephadex G-50, fine, 0.5 ϫ 5.5 cm) (Penefsky, 1977) equilibrated in the same buffer. The resulting ECF 1 preparations retain 1.6 -1.8 mol of ADP ϩ ATP bound in noncatalytic sites (see also Haughton and Capaldi, 1995). ECF 1 F 0 was prepared according to Foster and Fillingame (1979) with modifications described in Aggeler et al. (1987). This ATP synthase was reconstituted into egg-lecithin vesicles by the method described in Aggeler et al. (1995).
Maleimide Reaction of ECF 1 and ECF 1 F 0 -For modification by maleimides, nucleotide-depleted ECF 1 (2-3 M) was equilibrated at room temperature in 50 mM MOPS, pH 7.0, 0.5 mM EDTA, and 10% glycerol (v/v) buffer for 0.5-1 h. After addition of nucleotide, as stated, the enzyme was incubated for 5 min before the various maleimides were added. Samples were incubated in the dark at room temperature and at specific time intervals, aliquots withdrawn, and the reaction quenched by the addition of 20 mM DTT. The reaction of ECF 1 F 0 with maleimides was done similarly, except in 50 mM MOPS, pH 7.0, 5 mM MgCl 2 , and 10% glycerol. Labeling of ECF 1 from the mutant ␤E381C with [ 14 C]NEM (DuPont NEN) was conducted in 50 mM MOPS, pH 7.0, 0.5 mM EDTA, 10% glycerol, in the presence of different nucleotides as indicated, using 20 M of the maleimide. Data were analyzed as described in Haughton and Capaldi (1995).
Other Methods-ATPase activity was measured with a regenerating system described by Lötscher et al. (1984). ⑀-Depleted ECF 1 was prepared according to Dunn (1986) but using Sephacryl S300 (Pharmacia) followed by two passages through an ⑀-4 monoclonal antibody affinity column. Protein concentrations were determined with the BCA protein assay (Pierce Chemical Co.). SDS-polyacrylamide gel electrophoresis was performed with a 10 -18% SDS-containing gradient gel (Laemmli, 1970). Protein bands on gels were stained with Coomassie Brilliant Blue R (Downer et al., 1976).

RESULTS
Cys-87 of the ␥ Subunit shows a Nucleotide-dependent Reactivity with Maleimides When Glu-381 of ␤ Is Mutated to a Smaller Residue-In earlier studies, using the mutant ␤E381C, we had noted that an intrinsic Cys of the ␥ subunit was reactive to various maleimides. This preliminary observation was followed up as an approach to examining the conformation of the ␥ subunit under different nucleotide conditions. As shown in Fig. 1, there was incorporation of [ 14 C]NEM into the ␥ subunit of ECF 1 isolated from the mutant ␤E381C when the reaction was carried out in ATP ϩ EDTA, but no significant labeling if the reaction was carried out in EDTA alone (no nucleotide in catalytic sites) or in ADP ϩ Mg 2ϩ ϩ P i . In these experiments, NEM was incorporated rapidly into the Cys at position 381 of ␤, as well as into the ␦ subunit (not shown). The ␦ subunit is reactive to maleimides in wild-type ECF 1 , but its modification (at Cys-140) has no effect on activity (Mendel-Hartvig and Capaldi, 1991;Ziegler et al., 1994). Activity measurements showed that NEM incorporation, into either the ␥ subunit, or into Cys-381 of ␤, or both, activated the enzyme more than 2-fold. In contrast, CM modification of one or both sites caused essentially full inhibition (Fig. 2B).
Recently, Duncan et al. (1995aDuncan et al. ( , 1995b) used a mutant ␤D380C/␥C87S to distinguish which of the two Cys in the ␥ subunit (Cys-87 or Cys-112) was involved in disulfide bond cross-linking between ␥ and the ␤ DELSEED region. Following the same approach, we constructed the mutant ␤E381C/␥C87S/ ⑀S108C to identify which of the intrinsic Cys in ␥ was being reacted by maleimides. There was reactivity of the ␤ subunit in the Cys at 381 and modification of both ␦ and ⑀ (via Cys-108) but no labeling of the ␥ subunit by CM in this mutant (Fig. 3,  lanes 10 and 11). Therefore, Cys-87 must be the site of male-imide incorporation into the ␥ subunit. CM modification of the mutant ␤E381C/␥C87S/⑀S108C reduced the ATPase activity by 90% (Fig. 2C). This inhibition is not due to modification of Cys-␦140 as discussed above. Moreover, CM modification of the ECF 1 isolated from mutant ⑀S108C had no effect on activity (result not shown). Therefore, it must be modification of Cys-381 in the DELSEED region of the ␤ subunit that caused the observed inhibition of activity in this mutant.
To explore the reactivity of Cys-87 more fully, the mutant ␤E381A was prepared. This change preserves the short side chain but avoids a maleimide-reactive cysteine in the ␤ subunit. The reactivity of Cys-87 with CM under different nucleotide conditions is shown in Fig. 3 (lanes 1-8). There was rapid and strong incorporation of CM in EDTA ϩ ATP, EDTA ϩ ADP, or AMP⅐PNP ϩ Mg 2ϩ , a low incorporation of reagent in EDTA or Mg 2ϩ alone, but essentially no modification of ␥ with Mg 2ϩ ϩ ADP-bound, either when added directly, or as generated on the protein by addition of ATP ϩ Mg 2ϩ followed by
Nucleotide Dependence of the Reactivity of ␥ Cys-87 Is Lost in ⑀-Free ECF 1 -The reactivity of CM was monitored in ECF 1 from the mutant ␤E381A that had been freed of ⑀ subunit by affinity chromatography with a monoclonal antibody against the ⑀ subunit (Dunn, 1986). Fig. 4 shows that Cys-87 is labeled by CM under all nucleotide conditions including ADP ϩ Mg 2ϩ or AMP⅐PNP ϩ Mg 2ϩ . This site was also labeled by CM in EDTA or Mg 2ϩ alone (results not shown). The ATPase activity of the ⑀-free ECF 1 from mutant ␤E381A was high, i.e. 70 mol of ATP hydrolyzed per min per mg. There was no significant increase in the activity on reaction of CM (Fig. 2E), in contrast to the activation observed with enzyme that had not been freed of ⑀ subunit.
In another set of experiments, the effect of removing the ⑀ subunit on the inhibition of ATPase activity by CM was investigated in the mutant ␤E381C/␥C87S/⑀S108C. As shown in Fig.  2F, CM inhibited ⑀-free ECF 1 from this mutant. Therefore, it is the interaction between ␤ and ␥, rather than between ␤ and ⑀, which is perturbed when the Cys at residue 381 of ␤ is reacted with CM.
Cys-87 Is Reactive to Various Maleimides in ECF 1 F 0 from the Mutant ␤E381A-ECF 1 F 0 purified from the mutant ␤E381A had normal ATPase activity, i.e. around 20 mol of ATP hydrolyzed per min per mg protein, which was inhibited to 90% by 50 M dicyclohexylcarbodiimide, results similar to those obtained with wild-type enzyme. Reaction of this preparation with CM gave a similar pattern of labeling to that with ECF 1 from this mutant, i.e. strong labeling of the ␥ subunit in AMP⅐PNP ϩ Mg 2ϩ , but little or none in Mg 2ϩ alone, or in ADP ϩ Mg 2ϩ (result not shown). There was also some reaction of the reagent with the ␦ subunit, but no significant reaction of the intrinsic Cys in the b subunit under the labeling conditions employed. Fig. 2G summarizes the effects on the ATPase activity of modification by three different maleimides. Modification by NEM, CM, and BM all led to an inhibition of activity that was not seen with wild-type ECF 1 F 0 , indicating that the effect is due to reaction of Cys-87 and not Cys-140 of the ␦ subunit. The highest amount of inhibition was with the BM (more than 90%).

DISCUSSION
Cys-87 is at the end of a short ␣ helix of the ␥ subunit that interacts with the so-called DELSEED region of the ␤ subunit (Abrahams et al. 1994). We have found that Cys-87 can be cross-linked in essentially 100% yield by disulfide bond formation to a Cys replacing Glu at 381 (Aggeler et al., 1995), indicating the close proximity of the two residues, consistent with the ϳ 4-Å spacing from side chain S to S, estimated from the x-ray structural data (Abrahams et al., 1994). Cys-87 is buried in wild-type ECF 1 and ECF 1 F 0 . However, when Glu-381 of the ␤ subunit is exchanged for a smaller and uncharged side chain, such as Cys or Ala, this residue of ␥ becomes exposed for reaction with maleimides at least as large as CM. This exposure is nucleotide-dependent.
In enzyme from which catalytic site nucleotide has been removed, Cys-87 is essentially buried. Addition of nucleotide, either ADP or ATP in the presence of EDTA, exposes Cys-87 for reaction with various maleimides. In the absence of Mg 2ϩ , the binding constants for nucleotide in each of the three catalytic sites, including that in the ␤ which is linked to the short ␣ helix of ␥, is around 100 M (Weber et al., 1994;Grü ber and Capaldi, 1996), similar to that of isolated ␤ subunit, suggesting an open arrangement of the sites (as in ␤ E in the structure of MF 1 ). In the presence of Mg 2ϩ , Cys-87 is exposed when ATP is bound, as demonstrated by the data for AMP⅐PNP, but the residue is buried in ADP or ADP ϩ P i . It appears, therefore, that the short ␣ helix undergoes a release or reorganization that exposes Cys-87 when the catalytic sites are all open, or when ATP is bound, and that this is reversed on ATP hydrolysis.
A conformational change of the ␥ subunit related to ATP binding and hydrolysis has been seen previously by changes in cross-linking from a Cys introduced at position 8 of the ␥ subunit (in the long N-terminal ␣ helix) with the ␤ subunit(s) (Aggeler and Capaldi, 1993). This process has also been followed by fluorescence changes of CM bound to either the Cys introduced at position 8 or another Cys introduced at residue 106 of the ␥ subunit Capaldi, 1994a, 1994b). Fluorescence measurements under unisite catalysis conditions showed that the conformational change in the ␥ subunit occurs with bond cleavage of ATP to product ADP⅐P i , rather than with P i release Capaldi, 1994a, 1994b).
Importantly, the conformational rearrangements observed here by changes in the reaction of Cys-87 were lost on removal of the ⑀ subunit. Without the ⑀ subunit bound, Cys-87 is ex- posed for reaction under all nucleotide conditions, including when catalytic sites are empty. The conformational changes observed for Cys-8 by cross-linking and from both Cys-8 and Cys-106 by fluorescence measurements were also lost when the ⑀ subunit was removed (Aggeler and Capaldi, 1993;Turina and Capaldi, 1994a). The enzyme continues to show highly cooperative ATPase activity in the absence of the ⑀ subunit. The implication, therefore, is that the ⑀ subunit in some way controls or regulates structural changes in the ␥ subunit, and these changes are likely a part of the energy transduction mechanism.
The second interesting aspect of the reactivity of Cys-87 is the effect on activity of the enzyme. When Glu-381 is replaced by an Ala, reaction of Cys-87 with NEM has very little effect in isolated ECF 1 , whereas incorporation of CM activates the enzyme around 2.5-fold. This activation is related to ⑀ subunit binding, as it is lost when the ⑀ subunit is removed. By contrast, the reaction of Cys-87 in ECF 1 F 0 with either NEM or CM leads to inhibition of ATPase activity by 50% or more, while modification by BM induces almost full inhibition. These results for ECF 1 and ECF 1 F 0 can be compared with data for CF 1 and CF 1 F 0 . In the chloroplast enzyme, the equivalent residue of Cys-87, numbered Cys-89, is reactive to NEM even with a Glu in the DELSEED region (McCarty and Fagan, 1973;Moroney et al., 1984;Soteropoulos et al., 1994). Reaction of NEM with Cys-89 occurs in thylakoid membranes when these are energized by light (i.e. when ATP is being made) and does not occur in the dark (with ADP bound) (McCarty and Fagan, 1973). However, CF 1 isolated from thylakoids modified with NEM in the light is inhibited by the reagent, as is the ATPase activity of the membrane-bound enzyme (Soteropoulos et al., 1994).
Recent results by Soteropoulos et al. (1994) have shown an altered affinity of catalytic sites for ADP in CF 1 that has been modified with NEM, leading these authors to propose that inhibition is due to altered binding of nucleotides in one, or more, catalytic sites. This cannot explain the activity effects with ECF 1 , as the modification of Cys-87 by CM activates the enzyme when the ⑀ subunit is present and gives the same activity as unmodified enzyme in ⑀-free ECF 1 . Rather, the effect of modification of Cys-87 seems to be a steric effect, based on the results with ECF 1 F 0 from the mutant ␤E381A, where inhibition occurs with any of the maleimides used. Studies with the mutant ␤E381C/␥C87S/⑀S108C also point to the importance of steric constraints for conformational changes involving the short ␣ helix of ␥ and DELSEED region of ␤. Modification of a Cys-at 381 in the ␤ subunit with NEM activated, whereas reaction of this site with CM causes a dramatic inhibition of activity.
Taken together, the nucleotide dependence and activity effects suggest that there is a loosening and possibly a release of the ␥ subunit at its catch region with a ␤ subunit on ATP binding, which is reversed on ADP formation. Such a release, followed by rebinding, may be a part of coupling catalytic sites with the proton channel and would be a necessary step if the ␥ subunit moves relative to the ␣ 3 ␤ 3 domain, as suggested by the structural features of the enzyme (Abrahams et al., 1994) and as visualized by electron microscopy  and, more recently, by biochemical methods (Duncan et al., 1995a(Duncan et al., , 1995b.