ATP Hydrolysis-linked Structural Changes in the N-terminal Part of the y Subunit of Escherichia coli F,-ATPase Examined by Cross-linking Studies*

A mutant of Escherichia coli F1-ATPase (ECF1) in which the serine residue in position 8 of the y subunit has been replaced by a cysteine residue (yS8C) has been used to study nucleotide-dependent cross-linking of the y subunit to a P subunit. When examined in the presence of ADP + M P , either supplied directly or as produced during catalytic turnover of ATP + Me’, the main cross- linked product generated using the heterobifunctional, photoactivatable, cross-linker tetrafluorophenylazide maleimide-6 had a Mr(app) of 108,000. When ATP hydrolysis was inhibited, either by cold or by reaction with sodium azide, or when ATP hydrolysis was prevented by the use of adenyl-5’-yl P,y-imidodiphosphate, the main cross-linked products were species with Mr(app) of 102,000 and 84,000. The nucleotide-dependent switching from one cross-linking pattern to another could only be observed when the E subunit was bound to ECF1; it was not seen in ECF1*, an enzyme preparation missing S and E subunits, but was observed in preparations selectively depleted of the S subunit. We conclude that the changes detected in these cross-linking experiments are occur- ring during the hydrolysis of ATP when the P-y phosphate bond is cleaved and that they are related to the coupling of ATP hydrolysis to proton translocation. chlo-roplast catalyzes in re-sponse a transmembrane gradient. This can also generate a proton gradient by using the energy released by ATP hydrolysis (reviews in Refs. 1-3). best FIFO-type ATPase, the Escherichia coli enzyme, ECFlFo,’ is composed of eight different subunits, five of which, a, p, y, 6, and E, are present in the membrane extrinsic F1 part in the stoichiometry 3:3:1:1:1 (1, 2). biochemical

order to understand how catalytic sites of the F1 are coupled conformationally to the proton channel in the membrane intercalated Fo part of the complex. Our general approach is to create mutants by site-directed mutagenesis to include a reactive cysteine in a subunit of choice, a site that can then be selectively modified with reporter groups such as fluorescent reagents, spin labels, and bifunctional cross-linking reagents (4)(5)(6). Recently we described the generation of mutants of the y subunit including the mutant yS8C (6). Reaction of this mutant with the cross-linker TFPAMB led to covalent linkage of the y to a p subunit. Preliminary experiments indicated that this cross-linking depended on which nucleotides were present during the photolysis reaction (6). We have now explored the nucleotide dependence of the cross-linking from the cysteine residue at position 8 of the y subunit to a p subunit in more detail and here provide evidence that this cross-linking is tied to ATP hydrolysis in catalytic sites in a manner that is dependent on the tight binding of the E subunit to the core of the enzyme complex.
EXPERIMENTAL PROCEDURES Enzyme Preparation-E. coli ATPase from the mutant yS8C, containing the plasmid pRA112 in the unc-strain AN888 (61, was isolated and purified by a modified method of Senior et al. (7) and Wise et al. (8). ECF1*, an ECF, preparation depleted of 6 and e subunits, was isolated essentially as described by Tuttas-Dorschug and Hanstein (9), followed by two passages through a monoclonal anti-e antibody affinity column according to Dunn (10). The e antibody used (E-4 in Ref. 10) was a kind gift from Dr. Stanley Dunn (University of Western Ontario). ATPase, depleted only of the &subunit, was prepared by the procedure of Lotscher et al. (11).
Cross-linking of E. coli FI-ATPase-Cross-linking experiments were carried out with TFPAM-6 essentially as described in Aggeler et al. (4). ATPase was partially depleted of nucleotides by the use of centrifuge columns (121, and cysteine groups were modified at a protein concentration of between 2 and 4 mg/ml in labeling buffer (50 m M MOPS, pH 7.0, 0.5 m M EDTA, and 10% glycerol) with 200 PM tetrafluorophenylazide maleimide. Excess label was removed by two consecutive passages through centrifuge columns, and nucleotides were added at final concentrations of 5 mM, either in the absence or in the presence of 5 m M Mgz+. The photolysis was camed out either at room temperature for 2 h with a 6-watt 365-nm UV lamp ( U " , Inc., model UVL-56, Blak-Ray lamp) or in ice water for 15 min with a 100-watt 365-nm lamp (UVP, Inc., model B 100 A P , Blak-Ray lamp).
Other Methods-Protein concentrations were determined with the BCA protein assay from Pierce Chemical Co. Cross-link products were 0.15% SDS as described by Laemmli (13), followed by staining with analyzed by electrophoresis on 6-12% polyacrylamide gels containing Coomassie Brilliant Blue R according to Downer et al. (14).

RESULTS
Nucleotide Dependence of Cross-linking between $ys-8 a n d the p Subunit-For the experiments reported here, ECFl isolated from the mutant yS8C was freed of loosely bound nucleotide by two centrifugation column steps in EDTA-containing buffer, thereby generating an enzyme retaining only 2 mol of tightly bound nucleotides in (two) non-catalytic sites (15). This enzyme preparation was reacted with TFPAMB via the maleimide and then cross-linked by photolysis after first adding different combinations of Mg2+ a n d o r nucleotides. Fig. 1 shows the results of cross-linking from Cys-8 of the y subunit by TFPAM-6 under various nucleotide conditions. Three different major cross-linked products were seen in Coomassie Brilliant Blue stained gels, and each contained the p and y subunits as determined by Western blotting using subunit-specific monoclonal antibodies as before (6). lane, and electrophoresis was carried out on a SDS-containing 6-12% gradient polyacrylamide gel. The doublet of cross-linked products migrating below the 84-kDa band is a-S and p-8, respectively.
The key data are shown in lanes 6-9. With ADP + Mg2+ (lane 61, when ADP + Mg2+ + Pi were generated in catalytic sites by turnover of ATP + Mg" on the enzyme (lane 7) and in the presence of AMP-PN + Mg2+ (lune 8), the predominant crosslinked product was the 108-kDa species, with a small amount of a 102-kDa species also produced. In contrast, with the nonhydrolyzable ATP analogue AMP-PNP + Mg2+ present (lane 91, cross-linked products of 102 and 84 kDa were generated with little or no 108-kDa species produced. Fig. 1, lunes 1-5, shows cross-linking under several different control conditions. In EDTA alone (lane 2 ) or Mg2+ alone (lane 3 ) the 102-and 84-kDa species were formed, but in low yield.
With ADP + EDTA (lune 4 ) or ATP + EDTA (lane 5 ) the same two products were generated, but in higher amounts.
The difference in cross-linked products with AMP-PNP + Mg2+ versus AMP-PN + Mg2+, or ADP + Mg2+ suggests that a conformational change occurs around Cys-8 of the y subunit related to the cleavage of the P-y bond ofATP by the enzyme. To explore this possibility more fully, cross-linking was conducted in ATP + Mg2+ at 0 "C to slow down ATP hydrolysis and retain ATP in a catalytic site during the photolysis reaction. As shown in Fig. 2A, the presence of uncleaved ATP along with Mg2+ increased the amount of the 102-and 84-kDa species with a concomitant reduction of the 108-kDa species (cf. lune 3 with lane 2, where there had been complete hydrolysis of ATP to ADP + Pi). In the cold, therefore, the cross-linking patterns of the M$+ + ATP and Mg2+ + AMP-PNP (lane 4 ) were identical, while that was not the case at room temperature. This difference could not be attributed to denaturation of ATPase in the cold. Preincubation in the cold followed by photolysis at room temperature, as well as preincubation at room temperature followed by photolysis in the cold, showed the 108-kDa crosslinked species as the main product (data not shown). A fbrther indication of a native structure of ECFl in the cold is provided by the obsemation that the cross-linking pattern in the presence of Mg2+ + ADP was identical on ice (lane 5 ) and at room temperature (Fig. 1, lane 6).
Cross-linking was also conducted in the presence of azide, a potent non-competitive inhibitor of F1-ATPase activity, that abolishes cooperative interactions between catalytic sites (16, 17). Fig. 2B shows the concentration dependence of the effect of sodium azide. With an increasing concentration of the inhibitor, the amount of the 108-kDa species was reduced while the amount of the 102-and 84-kDa species increased.

yCys-8 to the p Subunit Requires the Binding of the e Subunit
-ECFl* is a preparation of ECFl missing both the 6 and e subunits. This preparation has a high ATPase activity due to release of the inhibitory e subunit (i.e. 90 pmol of ATP hydrolyzed per min per mg uersus 7-12 pmol per mg for the fivesubunit enzyme) (18). Fig. 3 shows cross-linking studies similar to those presented in Fig. 1

DISCUSSION
There are several interesting aspects to the cross-linking data presented here. There is a clear difference in cross-linking from the Cys at position 8 of the y subunit to a P subunit when the photolysis is performed in ATP + Mg2+ as opposed to ADP + Mg2+. Thus, when uncleaved ATP was present along with Mg2+ during photolysis, whether the rate of hydrolysis of the substrate has been slowed by cold (0 "C), or azide inhibition (171, or prevented by using the non-cleavable analogue AMP-PNP (19), the main products of cross-linking had Mdapp) of 102 and 84 kDa. This cross-linking pattern with ATP + Mg2+ bound is the same one seen when nucleotides ADP or ATP were bound without Mg2+ present. In the presence of ADP + Mg2+, the main cross-linked product was a 108-kDa species with the 102-kDa product much reduced in amount and the 84-kDa species absent. The temperature effect and the azide sensitivity of the observed changes in cross-linking pattern make it unlikely that these result from nucleotide binding changes in the non-catalytic sites. Instead, we conclude that the structural changes around Cys-8 of y are related to ATP hydrolysis in catalytic sites.
Supporting evidence for the conclusion comes from fluorescence studies. We have labeled Cys-8 of the y subunit in the mutant yS8C with the fluorescent reagent coumarin maleimide and followed changes in the fluorescence spectrum from this site as a function ofATP hydrolysis. There is a quenching of the fluorescence spectrum on addition of ATP + Mg2+ that follows the cleavage of the P-y Pi bond of ATP. This quenching is not seen when ATP + EDTA, ADP + Mg2+, or AMP-PNP + Mg2+ are added, and it is blocked by prior treatment of the enzyme with the inhibitor azide.2 The difference in cross-linking of the y to p subunit in ATP + Mg2+ compared with ADP + Mg2+ was found to depend on the presence of the E subunit. In ECFI*, which is missing the 6 and E subunits, the 102-and 84-kDa cross-linked species were obtained in both nucleotide conditions. This must be a consequence of the loss of the E subunit because enzyme selectively depleted of the 6  bound at Cys-8 of the y subunit are also lost on removal of the E subunit.2) ECF1* is a much more active ATPase than intact ECFl because of the removal of the inhibitory E subunit (18). The interaction of the E subunit with the core ECFl complex has been shown to involve both the y and P subunits (4,(20)(21)(22)(23). In the intact ATP synthase (where the E subunit is tightly bound), we propose that changes in the y subunit couple catalytic site events with proton pumping in the Fo part of the complex.
The generation of the 108-kDa species obtained when ADP is bound to intact ECFl is Mg2+-dependent. In the absence of Mg2+ (i.e. when ADP + EDTA are present), the 108-kDa species does not form, and instead, the 102-and 84-kDa species are obtained. There is good evidence that the catalytic sites on the P subunits of F1 are similar to the GTP binding site of Ras protein (24, 25). In particular, the glycine-rich or phosphate binding loop is conserved in the two structures and probably plays the identical function in F1 (26,27) as in Ras, where it participates in binding the Pand y-phosphates of ATP and in liganding the Mg2+ ion (28, 29). Movements of the phosphate binding loop along with rearrangements of other close by segments of the Ras protein occur on GTP hydrolysis, and these changes alter the binding of Ras with other proteins (28, 29). Similar changes in and around the catalytic site of F1 could trigger changes in the structure of the y subunit under coupled conditions, i.e. in the presence of the E subunit. We are in the process of mapping the sites of cross-linking of Cys-8 of y with the P subunit. In the 108-kDa cross-linked product, this site has been localized to the peptide fragment (residues 145-155), which includes the phosphate loop r e g i~n .~