Oxidation of the α3(βD311C/R333C)3γ Subcomplex of the Thermophilic Bacillus PS3 F1-ATPase Indicates That Only Two β Subunits Can Exist in the Closed Conformation Simultaneously*

In the crystal structure of the bovine heart mitochondrial F1-ATPase (Abrahams, J. P., Leslie, A. G. W., Lutter, R., and Walker, J. E. (1994)Nature 370, 621–628), the two liganded β subunits, one with MgAMP-PNP bound to the catalytic site (βT) and the other with MgADP bound (βD) have closed conformations. The empty β subunit (βE) has an open conformation. In βT and βD, the distance between the carboxylate of β-Asp315 and the guanidinium of β-Arg337 is 3.0–4.0 Å. These side chains are at least 10 Å apart in βE. The α3(βD311C/R333C)3γ subcomplex of TF1 with the corresponding residues substituted with cysteine has very low ATPase activity unless it is reduced prior to assay or assayed in the presence of dithiothreitol. The reduced subcomplex hydrolyzes ATP at 50% the rate of wild-type and is rapidly inactivated by oxidation by CuCl2 with or without magnesium nucleotides bound to catalytic sites. Titration of the subcomplex with iodo[14C]acetamide after prolonged treatment with CuCl2 in the presence or absence of 1 mm MgADP revealed nearly two free sulfhydryl groups/mol of enzyme. Therefore, one pair of introduced cysteines is located on a β subunit that exists in the open or partially open conformation even when catalytic sites are saturated with MgADP. Since V max of ATP hydrolysis is attained when three catalytic sites of F1are saturated, the catalytic site that binds ATP must be closing as the catalytic site that releases products is opening.

In the crystal structure of the bovine heart mitochondrial F 1 3 ␥ subcomplex of TF 1 with the corresponding residues substituted with cysteine has very low ATPase activity unless it is reduced prior to assay or assayed in the presence of dithiothreitol. The reduced subcomplex hydrolyzes ATP at 50% the rate of wild-type and is rapidly inactivated by oxidation by CuCl 2 with or without magnesium nucleotides bound to catalytic sites. Titration of the subcomplex with iodo[ 14 C]acetamide after prolonged treatment with CuCl 2 in the presence or absence of 1 mM MgADP revealed nearly two free sulfhydryl groups/mol of enzyme. Therefore, one pair of introduced cysteines is located on a ␤ subunit that exists in the open or partially open conformation even when catalytic sites are saturated with MgADP. Since V max of ATP hydrolysis is attained when three catalytic sites of F 1 are saturated, the catalytic site that binds ATP must be closing as the catalytic site that releases products is opening.
The proton-translocating F 0 F 1 -ATP synthases couple proton electrochemical gradients to condensation of ADP with P i in energy-transducing membranes. The F 0 moiety is a membraneembedded protein complex that mediates proton translocation. F 1 is a peripheral membrane protein complex composed of five different subunits with ␣ 3 ␤ 3 ␥␦⑀ stoichiometry. When removed from the membrane, F 1 is an ATPase containing six nucleotide binding sites. Three are catalytic sites that are predominantly on ␤ subunits at ␣/␤ interfaces. The three other sites, called noncatalytic sites, are located predominantly on ␣ subunits at different ␣/␤ interfaces (1)(2)(3).
The crystal structures of the bovine heart and rat liver mi-tochondrial F 1 -ATPases as well as that of the ␣ 3 ␤ 3 subcomplex of the TF 1 -ATPase have been determined (4 -6). In the crystals of the bovine heart enzyme (BH-MF 1 ), 1 which form in media containing AMP-PNP, ADP, Mg 2ϩ , and N 3 Ϫ , noncatalytic sites are homogeneously liganded with MgAMP-PNP, whereas catalytic sites are heterogeneously liganded (4). One, designated ␤ T , contains MgAMP-PMP, another, designated ␤ D , contains MgADP and the third catalytic site, designated ␤ E , is empty. Except for small differences in the regions of the terminal phosphates of bound nucleotides, the arrangements of functional amino acid side chains in catalytic sites in ␤ T and ␤ D are essentially identical. In contrast, these residues are arranged differently in ␤ E . In the crystals of the ␣ 3 ␤ 3 subcomplex of TF 1 that form in media free of Mg 2ϩ and nucleotides, the ␣ subunits are in closed conformations corresponding to the liganded ␣ subunits of the mitochondrial enzymes. The ␤ subunits in the ␣ 3 ␤ 3 structure are in open conformations corresponding to ␤ E of BH-MF 1 (5).
In the crystals of the rat liver F 1 -ATPase (RL-MF 1 ), which form in media containing ATP, P i , and EDTA, noncatalytic sites are homogeneously occupied with MgATP. In contrast to the crystals of BH-MF 1 , all three catalytic sites of RL-MF 1 contain bound nucleotides in the absence of Mg 2ϩ (6). One catalytic site contains ADP alone, whereas the other two contain ADP and P i . Since Mg 2ϩ is not associated with ADP and P i bound to catalytic sites of RL-MF 1 , side chains in catalytic sites that interact with the Mg 2ϩ ion in the crystal structure of BH-MF 1 are arranged differently in the crystal structure of RL-MF 1 . However, these differences are minor compared with the difference between the closed conformations of ␤ D and ␤ T and the open conformation of ␤ E in BH-MF 1 . In the crystal structure of RL-MF 1 , the conformations of all three ␤ subunits resemble the closed conformation of ␤ D and ␤ T in the crystal structure of BH-MF 1 . Bianchet et al. (6) suggest that the open conformation of ␤ E in the crystal structure of BH-MF 1 is the consequence of low concentrations of nucleotides in the crystallization medium. Extending this argument, they propose that the RL-MF 1 structure with three closed ␤ subunits plays a central role in catalysis, whereas the BH-MF 1 structure with two closed ␤ subunits and one open ␤ subunit exists transiently during catalysis when products dissociate and substrates rebind.
To determine which of these structures is more consistent with properties of F 1 -ATPases in solution, we took advantage of pairs of amino acid residues in ␤ subunits that do not contrib-ute to catalytic sites in BH-MF 1 that have side chains within interaction distance in ␤ D and ␤ T but are considerably distant from each other in ␤ E . For example, the carboxylate of ␤-Asp 315 is within 3.0 and 3.7 Å of the guanidinium of ␤-Arg 337 , in ␤ T and ␤ D , respectively, whereas these side chains are 10.3 Å apart in ␤ E . Other pairs of amino acid residues that meet these criteria are ␤-Ala 158 /␤-Tyr 311 ; ␤-Met 167 /␤-Phe 183 , and ␤-Lys 175 / ␤-Ser 203 . Double mutants of the ␣ 3 ␤ 3 ␥ subcomplex of TF 1 (7) were prepared in which residues corresponding to ␤-Asp 315 /␤-Arg 337 , ␤-Ala 158 /␤-Tyr 311 , ␤-Met 167 /␤-Phe 183 , and ␤-Lys 175 /␤-Ser 203 of BH-MF 1 were substituted with cysteine.
The aim of preliminary studies was to find a double mutant that was stable and also could be converted reversibly between an oxidized, inactive form and a reduced, active form. Of the four double mutants generated, only the ␣ 3 (␤D311C/R333C) 3  Toyopearl Butyl-650S resin was from TosoHaas. The oligonucleotides used for mutagenesis were purchased from Life Technology, Inc. The Escherichia coli strains and the plasmids used to prepare the wild-type and mutant ␣ 3 ␤ 3 ␥ subcomplexes were described by Matsui and Yoshida (7). The purified enzyme subcomplexes were stored as suspensions in 75% ammonium sulfate at 4°C.
Construction of Plasmids Containing Mutant Genes-Plasmid pKK223-3, which carries genes encoding the ␣, ␤, and ␥ subunits of TF 1 , was used for both mutagenesis and gene expression. Expression plasmids were constructed using polymerase chain reaction with the Quick Change TM site-directed mutagenesis kit from Stratagene. The plasmids were purified using the Wizard TM Plus miniprep kit from Promega. The first mutation was introduced into wild-type pKK223-3 by polymerase chain reaction and confirmed by sequence analysis. The resulting mutant plasmid was used as template for the second polymerase chain reaction substitution, which was subsequently confirmed. The resulting pKK223-3 mutant plasmids were expressed in E. coli strain JM103 (unc Ϫ ). The primers used in the polymerase chain reactions are as follows with mismatched bases underlined: Y307C, 5Ј-CGATTCAAGC-GATTTGCGTCCCGGCCG; Q177C, 5Ј-CACAACATCGCCTGTGAGCA-CGGCGG; S205C, 5Ј-GAGATGAAAGATTGCGGCGTCATCAGC; Q169C, 5Ј-GGTCTTGATCTGTGAGCTGATTCACAACAT; F185C, 5Ј-G-GGATTTCCGTCTGTGCTGGCGTCGGC; D311C, 5Ј-CGTCCCGGCCT-GCGACTATACGGACC; R333C, 5Ј-CGACGAACCTGGAGTGTAAGCT-CGCGG and the corresponding complementary primers.
Enzyme Preparation and Assays-The wild-type and mutant subcomplexes were expressed and purified as described previously (7,8). Unless stated otherwise, stock solutions of the enzyme subcomplexes were prepared by the following procedure. After centrifugation of the ammonium sulfate suspensions, the pellets were dissolved in 50 mM Tris-Cl, pH 8.0, containing 1 mM CDTA. The solutions were incubated for 1 h at 23°C, at which time they were passed through 1-ml centrifuge columns of Sephadex G-50 equilibrated with 50 mM Tris-Cl, pH 8.0, containing 0.1 mM EDTA (9). Preparations treated in this manner are designated CDTA-treated subcomplexes.
The ␣ 3 (␤D311C/R333C) 3 ␥ mutant subcomplex was reduced as follows. Ammonium sulfate precipitates of the subcomplex were collected by centrifugation and dissolved in 50 mM Tris-HCl, pH 8.0, containing 5 mM CDTA and 10 mM DTT. After incubating for 2 h, the resulting solution was passed through a centrifuge column of Sephadex G-50 equilibrated with 50 mM Tris-HCl, pH 8.0, containing 0.1 mM EDTA. After reduction, ATPase activity did not decline for at least 2 h. When the reduced enzyme was subsequently treated with iodoacetate or iodoacetamide, the Sephadex G-50 centrifuge columns (9) were equilibrated with 50 mM Tris-HCl, pH 8.0, containing 0.1 mM EDTA plus 0.2 mM DTT.
ATPase activity was determined spectrophotometrically with 2 mM ATP plus 3 mM Mg 2ϩ using ATP regeneration with phosphoenolpyruvate and pyruvate kinase coupled to NADH oxidation by lactate dehydrogenase under conditions described earlier (10).
Other Analytical Methods-Protein concentrations were determined by the method of Bradford (11). Endogenous nucleotides bound to the enzyme subcomplexes were determined by high pressure liquid chromatography using ion pairing with tetrabutyl ammonium hydrogen sulfate as described by Bullough et al. (12). Table I shows the distances in the crystal structure of the ␣ 3 ␤ 3 subcomplex of TF 1 (5) between side chains of amino acid residues that were substituted with cysteine in the four double mutants examined in this study. The distances between the corresponding side chains in ␤ T , ␤ D , and ␤ E in the crystal structure of MF 1 (4) are also tabulated.

Comparison of the ATPase Activities of the
The effects of DTT on the ATPase activities of the purified double mutant subcomplexes were determined. Each isolated enzyme subcomplex was treated with CDTA before assays were performed. This procedure removes endogenous MgADP from a catalytic site of the purified, wild-type ␣ 3 ␤ 3 ␥ subcomplex (8). The ␤A160C/Y307C double mutant was isolated in an inactive form that could not be activated by treatment with DTT before assay or by including DTT in the assay medium. Lauryl dimethylamine oxide, which stimulates ATPase activity of the wild-type and certain other mutant subcomplexes did not activate the ␣ 3 (␤A160C/Y307C) 3 ␥ subcomplex in the presence or absence of DTT. Defective assembly was not responsible for the lack of ATPase activity. A normal pattern of ␣, ␤, and ␥ subunits was observed in a 3:3:1 ratio after the mutant subcomplex was submitted to SDS-polyacrylamide gel electrophoresis.
The ␣ 3 (␤Q177C/S205C) 3 ␥ double mutant had negligible ATPase activity in the absence of activation with dithiothreitol. When 10 mM DTT was included in the assay medium, its specific activity for hydrolyzing 2 mM ATP increased to 0.8 mol of ATP mg Ϫ1 min Ϫ1 . ATPase activity was stimulated about 10-fold when 0.03% lauryl dimethylamine oxide was present in the assay medium. Lauryl dimethylamine oxide stimulates the ATPase activity of the wild-type ␣ 3 ␤ 3 ␥ subcomplex by a factor of 3-4 (8).
The specific activity of the ␣ 3 (␤Q169C/F185C) 3 ␥ double mutant was 7 mol of ATP hydrolyzed mg Ϫ1 min Ϫ1 in the absence of activation with DTT. Prior treatment with DTT or the inclusion of 10 mM DTT in the assay medium did not stimulate ATPase activity of this mutant. Surprisingly, including lauryl dimethylamine oxide in the assay medium in the concentration range of 0.01-0.06% lowered the ATPase activity to less than 0.07 mol of ATP hydrolyzed mg Ϫ1 min Ϫ1 . The ATPase activity of this double mutant was inactivated by 2 mM o-iodosobenzoate with a pseudo-first order rate constant of 1.4 ϫ 10 Ϫ2 min -1 .  (5) and between corresponding residues in the crystal structure of BH-MF 1 (4) Fig. 1A shows that the isolated ␣ 3 (␤D311C/ R333C) 3 ␥ subcomplex has negligible ATPase activity unless the assay medium is supplemented with DTT or another thiol. Whereas traces a and c of Fig. 1A were obtained in the absence of thiols, trace b shows the time-dependent activation of ATPase activity observed when the assay medium contained 10 mM DTT. Fig. 1B shows that after reduction as described under "Experimental Procedures" the ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex was inactivated during assay unless DTT or 0.1 mM EDTA was included in the assay medium. Trace a in Fig. 1B illustrates assay of the enzyme in the absence of DTT or EDTA. Traces b and c of Fig. 1B represent assays conducted in the presence of 10 mM DTT or 100 M EDTA, respectively. The dependence of protection against oxidation on EDTA concentration revealed that maximal protection was achieved with 50 M EDTA. Other experiments indicated that a contaminant in the MgCl 2 was responsible for oxidation in the absence of EDTA. Therefore, 100 M EDTA was included in the assay medium in experiments with the reduced ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex. Since the presence of EDTA in the assay medium does not affect ATPase activity of the isolated enzyme, illustrated by trace c of Fig. 1A, the isolated ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex exists in the oxidized form.
Iodoacetate Inactivates the Reduced Form of the ␣ 3 (␤D311C/ R333C) 3 ␥ Subcomplex-Following reduction of the ␣ 3 (␤D311C/ R333C) 3 ␥ subcomplex with 1 mM DTT for 30 min, the addition of iodoacetate to 4 mM led to 80% inactivation of ATPase activity within 10 min followed by slower inactivation, illustrated in Fig. 2. The addition of increasing concentrations of ADP plus Mg 2ϩ to the reduced subcomplex prior to adding iodoacetate slowed the rate of inactivation in both phases. Fig. 3 correlates the mol of [ 3 H]acetate incorporated per mol of the ␣ 3 (␤D311C/ R333C) 3 ␥ subcomplex with the extent of inactivation observed during inactivation of the reduced, mutant enzyme with iodo[ 3 H]acetate. It is clear from Fig. 3 that multiple hits are required to cause full inactivation. To reconcile this unusual stoichiometry, it is possible that modification of a single cysteine in a given ␤ subunit wounds the enzyme, and modifica-tion of both Cys 311 and Cys 333 in the same ␤ subunit wounds the enzyme to a greater extent. The introduced cysteines appear to be the only residues that were appreciably carboxymethylated. The wild-type subcomplex was not inactivated when treated with iodo[ 3 H]acetate under the same conditions and incorporated no more than 0.3 mol of reagent/mol of enzyme in the absence of nucleotides and much less than that in the presence of MgADP. The wild-type subcomplex, like the mutant, contains ␣-Cys 193 that is resistant to modification by iodoacetate.
Complete carboxymethylation lowered the capacity of the ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex to bind ADP in the presence of Mg 2ϩ . This was established as follows. After carboxymethylation of the nucleotide-depleted mutant subcomplex by treatment with 4 mM iodoacetate in the presence of 1 mM DTT for 16 h, the modified enzyme was passed through a 1-ml centrifuge column of Sephadex G-50 equilibrated with 50 mM Tris-Cl, pH 8.0, to remove excess iodoacetate and DTT. The gel-filtered Carboxamidomethylation Converts the Reduced ␣ 3 (␤D311C/ R333C) 3 ␥ Subcomplex to a Slightly More Active Form-Modification of the ATPase activity of the ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex by sulfhydryl reagents other than iodoacetate was also explored. Treatment of the reduced subcomplex with Nethylmaleimide or [(1-trimethylammonium)methyl]methanethiosulfonate bromide, a reagent that derivatizes cysteine residues with a quarternary ammonium cation, using the same protocol described for iodoacetate, caused slower inactivations that were partially protected by MgATP or MgADP. In contrast, carboxamidomethylation of the subcomplex with excess iodoacetamide in the presence of DTT converted the enzyme to a form that no longer required DTT or EDTA in the assay medium to retain constant activity. This is illustrated in Fig.  1C. Trace a of Fig. 1C represents ATP hydrolysis by the reduced, mutant subcomplex in the absence of DTT or EDTA after treating it exhaustively with iodoacetamide. In the absence of carboxamidomethylation, the reduced enzyme is inactivated during assay in the absence of DTT or EDTA as illustrated by trace a of Fig. 1B.
The rate of conversion of the reduced mutant subcomplex to a more active form by iodoacetamide was not affected by MgADP. When the subcomplex was treated with 4 mM iodoacetamide in the presence of 1 mM DTT, 50% conversion was observed within 2 min. Maximal conversion occurred within 60 min. The addition of ADP with or without MgCl 2 during treatment with iodoacetamide did not affect the rate of conversion.

Prolonged Oxidation of the Reduced ␣ 3 (␤D311C/R333C) 3 ␥ Subcomplex Fails to Promote Formation of Disulfide Bonds in
All Three ␤ Subunits-The reduced ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex is rapidly inactivated in the presence of CuCl 2 , tetrathionate, o-iodosobenzoate, 5,5Ј-dithiobis-(2-nitrobenzoic acid), or H 2 O 2 . To determine the number of ␤ subunits containing disulfide bonds on conversion of the reduced to the oxidized form, the reduced ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex was inactivated with CuCl 2 in the presence or absence of MgADP. In each case, complete inactivation was observed within 5 min. The reaction mixtures were then incubated an additional 2 h to allow further oxidation. After removing CuCl 2 , free ADP, and Mg 2ϩ , inactivated enzyme was treated with iodo[ 3 H]acetate or iodo[ 14 C]acetamide for 16 h to derivatize free sulfhydryl groups. These experiments are summarized in Table II. The results obtained when iodo[ 14 C]acetamide was used to monitor residual free sulfhydryl groups, strongly indicate that disulfide bonds are formed in only two ␤ subunits during inactivation. In this case, nearly two sulfhydryl groups were derivatized after inactivating the enzyme with CuCl 2 . The slightly lower values of free sulfhydryl groups titrated with iodo[ 3 H]acetate may reflect charge repulsion encountered during carboxymethylation of Cys 311 and Cys 333 in the same ␤ subunit that does not occur during carboxamidomethylation with iodoacetamide.
Formation of MgADP-Fluoroaluminate Complexes in the Reduced and Oxidized ␣ 3 (␤D311C/R333C) 3 ␥ Subcomplex-Previous studies have shown that the MgADP-fluoroaluminate complex forms slowly when Al 3ϩ and F Ϫ were added to wildtype and mutant ␣ 3 ␤ 3 ␥ subcomplexes of TF 1 after loading a single catalytic site with MgADP (13,14). In contrast, when MgADP was loaded onto two catalytic sites or when the subcomplex was incubated with excess ADP plus Mg 2ϩ , MgADPfluoroaluminate complexes formed rapidly in two catalytic sites after the addition of Al 3ϩ and F Ϫ . This suggests that MgADPfluoroaluminate complexes are formed cooperatively in two catalytic sites. Table III 3 ␥ subcomplex was prepared in 50 mM Tris-Cl, pH 8.0, containing 100 M EDTA as described under "Experimental Procedures." The oxidized subcomplex was prepared by treating the reduced subcomplex at 1 mg/ml with 200 M CuCl 2 in the presence or absence of 1 mM ADP plus 2 mM MgCl 2 for 2 h, at which time the reaction mixture was passed through a centrifuge column of Sephadex G-50 column (9)   subcomplex formed fluoroaluminate complexes somewhat faster than wild-type enzyme when Al 3ϩ and F Ϫ were added to the subcomplexes containing stoichiometric MgADP or in the presence of excess ADP and Mg 2ϩ . The rates of formation of inhibitory fluoroaluminate complexes in the reduced mutant subcomplex and the wild-type subcomplex were about 10-fold faster when Al 3ϩ and F Ϫ were added after prior exposure of the enzymes to 200 M ADP plus Mg 2ϩ rather than stoichiometric ADP plus Mg 2ϩ . This suggests that the wild-type and reduced mutant subcomplexes display similar positive cooperativity between catalytic sites. In contrast, the rate of formation of the inhibitory MgADP-fluoroaluminate complex is the same when Al 3ϩ and F Ϫ are added to the oxidized mutant enzyme, indicating no cooperativity between catalytic sites. Whereas the presence of 2 mM P i inhibited formation of the inhibitory fluoroaluminate complex when MgADP was bound to a single catalytic site, 2 mM P i stimulated the rate of formation of MgADP-fluoroaluminate complexes when the wild-type and reduced mutant subcomplexes were treated with 200 M ADP plus 2 mM Mg 2ϩ before adding Al 3ϩ and F Ϫ . In contrast, the presence of 2 mM P i inhibited formation of the MgADP-fluoroaluminate complex when Al 3ϩ and F Ϫ were added to the oxidized subcomplex after exposure to 200 M ADP plus 2 mM Mg 2ϩ .

DISCUSSION
The proximity of the side chains of ␤-Asp 315 and ␤-Arg 337 in ␤ T and ␤ D , but not in ␤ E , in the crystal structure of BH-MF 1 implies that these residues may participate in functionally important electrostatic interactions during catalysis. However, substitution of the corresponding residues in the ␣ 3 ␤ 3 ␥ subcomplex of TF 1 with cysteine does not severely impair ATP hydrolysis. The reduced ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex hydrolyzes 2 mM ATP at about 50% the rate exhibited by the wild-type subcomplex. Furthermore, carboxamidomethylation of the introduced cysteines slightly enhances ATP hydrolysis. Nevertheless, it is clear that the side chains of these residues must be free to change position during catalysis. The mutant subcomplex is rapidly inactivated by several reagents that promote oxidation of adjacent cysteine side chains in proteins to disulfide bonds. This presumably locks affected ␤ subunits in the closed conformation. Perception of the consequences of locking a ␤ subunit in the disulfide form can be realized from inspection of Fig. 4. The different positions of the side chains of ␤-Asp 315 and ␤-Arg 337 in the nucleotide binding domain of ␤ T and ␤ E in the crystal structure of BH-MF 1 are illustrated. It is clear that the nucleotide binding domains, shown in yellow, are folded very differently in ␤ T and ␤ E . To highlight the different conformations of ␤ T and ␤ E , helices B and C, which contain components of catalytic sites are illustrated in cyan and magenta, respectively. Transition from the closed conformation of ␤ T to the open conformation of ␤ E moves the carboxylate of ␤-Asp 315 from the guanidinium of ␤-Arg 337 by more than 7 Å.
The finding that nearly two of the introduced cysteines can be derivatized with iodoacetamide after prolonged exposure of the ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex to oxidizing conditions in the presence or absence of saturating MgADP suggests that disulfide bonds are only formed in two of the three ␤ subunits. This is consistent with the results of a study recently reported by Tsunoda et al. (15) who used computer modeling of the crystal structure of BH-MF 1 to show that it is not possible for three ␤ subunits to exist in the closed conformation simultaneously.
It is interesting that carboxymethylation of the introduced cysteines of the ␣ 3 (␤D311C/R333C) 3 ␥ double mutant inactivates, rather than stimulates, ATPase activity as observed when they are carboxamidomethylated. Maximal inactivation is observed only after carboxymethylation of all six of the introduced sulfhydryl groups. This suggests that the diminution of ATPase activity that increases with increasing carboxymethylation is caused by charge repulsion of carboxymethylated cysteines on the same ␤ subunit. The observation that bound MgADP protects against carboxymethylation with   3 ␥ subcomplexes under various conditions The rates of formation (k form ) are the pseudo-first order rate constants for irreversible inactivation observed after adding Al 3ϩ and F Ϫ to the TF 1 subcomplexes in the presence of 200 M ADP plus 2 mM Mg 2ϩ or with stoichiometric MgADP bound to a catalytic site as described previously (13,14). To load a single catalytic site with MgADP, the isolated mutant subcomplex at 1 mg/ml in 50 mM Tris-Cl, pH 8.0, was incubated with stoichiometric ADP in the presence of 1 mM Mg 2ϩ for 30 min, at which time the solution was passed through a centrifuge column of Sephadex G-50 (9) that was equilibrated with the same buffer. The reduced ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex used in these experiments was prepared by incubating the isolated enzyme with or without MgADP bound to a single catalytic site with 5 mM DTT for 30 min before making subsequent additions. The oxidized ␣ 3 (␤D311C/R333C) 3 ␥ subcomplex was prepared by treating the isolated enzyme with 1 mM CDTA for 30 min and passing it through a Sephadex G-50 centrifuge column (9) that was equilibrated with 50 mM Tris-Cl, pH 8.0. Assays were conducted in the presence of 10 mM DTT. iodoacetate, whereas it has no effect on carboxamidomethylation with iodoacetamide, is consistent with this premise. Liganding of catalytic sites forces the introduced cysteine side chains together in a given ␤ subunit, thus making it more difficult to modify both of them with the negatively charged carboxymethyl group.
In the crystal of the (␣␤) 3 subcomplex of TF 1 deduced by Shirakihara et al. (5), the three ␤ subunits have identical, open conformations. As shown here, treatment of the ␣ 3 (␤D311C/ R333C) 3 ␥ double mutant with CuCl 2 in the absence of bound nucleotides leads to disulfides in two ␤ subunits. Therefore, two ␤ subunits exist in closed or partly closed conformations in the absence of liganding catalytic sites with nucleotides in the presence of Mg 2ϩ . This asymmetry must be induced by the coiled-coil composed of the amino and carboxyl termini of the ␥ subunit within the central cavity of the (␣␤) 3 hexamer. That two ␤ subunits can exist in the closed conformation in the absence of bound nucleotides was also demonstrated by Tsunoda et al. (15). In the crystal structure of BH-MF 1 , the side chain of Ile 390 in ␤ T is in contact with the side chain of Ile 390 in ␤ D . Computer modeling reported by Tsunoda et al. (15) showed that this contact only occurs when two ␤ subunits are in the closed conformation. The residue in TF 1 corresponding to ␤-Ile 390 of BH-MF 1 is ␤-Ile 386 . Treatment of the ␣ 3 (␤I386C) 3 ␥ subcomplex of TF 1 with 0.25 M CuCl 2 in the presence of MgATP, MgADP, or MgADP and NH 3 Ϫ inactivated the enzyme and was accompanied by cross-linking of two ␤ subunits. In the absence of nucleotides plus Mg 2ϩ , inactivation and cross-linking occurred more slowly. This indicates that two ␤ subunits are at least partly closed in the absence of ligation of catalytic sites with MgATP or MgADP.
The demonstration that disulfide bonds are formed in two ␤ subunits upon oxidation of the ␣ 3 (␤D311C/R333C) 3 ␥ double mutant in the absence of bound nucleotides and that two ␤ subunits are cross-linked on oxidation of the ␣ 3 (␤I386C) 3 ␥ subcomplex in the absence of bound nucleotides has important mechanistic implications. These findings are consistent with earlier observations, suggesting that the heterogeneous affinities of catalytic sites of F 1 -ATPases for MgATP reflect intrinsic asymmetry of catalytic sites dictated by the position of the coiled-coil of the ␥ subunit within the central cavity of the (␣␤) 3 hexamer rather than negative cooperativity of binding (16,17). Three K d values of about 2 nM, 0.2 M, and 34 M were detected when quenching of the tryptophan fluorescence of the ␣ 3 (␤Y341W) 3 ␥ subcomplex was titrated with MgADP (17). According to the results presented here, the two low K d values represent binding of MgADP to ␤ subunits in closed conformations, whereas the K d of 34 M represents binding of MgADP to the open catalytic site.
The findings reported here and those reported by Tsunoda et al. (15) demonstrating that only two ␤ subunits exist in the closed conformation simultaneously in the presence of MgADP or MgATP are consistent with the crystal structure of BH-MF 1 reported by Abrahams et al. (4). However, they are inconsistent with the crystal structure of RL-MF 1 reported by Bianchet et al. (6) that indicates that all three ␤ subunits are in the closed conformation in the absence of Mg 2ϩ with two of them liganded with ADP and P i and the third liganded with ADP only.
Evidence indicating that only two ␤ subunits can exist in closed conformations simultaneously is also inconsistent with models proposed for ATP hydrolysis and synthesis that include conformational states of F 1 in which three catalytic sites are closed (1,6). This does not mean that maximal rates of ATP hydrolysis are achieved when only two catalytic sites are saturated. Weber et al. have convincingly demonstrated that ATP hydrolysis catalyzed by E. coli F 1 achieves maximal velocity when three catalytic sites are saturated (18). The model illustrated in Fig. 5 is proposed to accommodate this finding with the observations presented here and by Tsunoda et al. (15), indicating that three ␤ subunits of F 1 cannot exist in the closed conformation simultaneously. The model, which is modified from a scheme in Ref. 3, depicts a round of trisite ATP hydrolysis in which one ␤ subunit (␤ E ) is transiently in the open conformation and the two others (␤ D and ␤ T ) are in the closed conformation and liganded with MgATP. Binding of MgATP to ␤ E causes 1) simultaneous movement of the ␥ subunit in a counterclockwise direction; 2) propagation of a conformational signal from the catalytic site of ␤ E to the catalytic site of ␤ D (indicated by the curved arrow between ␤ E and ␤ D ), and 3) opening of the catalytic site of ␤ D , which is accompanied by ATP hydrolysis. The transition state for ATP hydrolysis is represented by [ATP]. During this process, the ␥ subunit rotates 120°. This is necessary to overcome steric restrictions imposed by the ␥ subunit pointed out by Tsunoda et al. (15) that prevent closing of three ␤ subunits simultaneously. From computer modeling, they showed that the coiled-coil of the ␥ subunit within the central cavity of the (␣␤) 3 hexamer allows simultaneous closing of only two ␤ subunits. Therefore, in order to proceed from the initial state illustrated in Fig. 5, where ␤ D and ␤ T are in closed conformations, to the final state, where ␤ T and ␤ E are in closed conformations, the position of the coiledcoil of the ␥ subunit within the central cavity must change as ␤ E closes and ␤ D opens. To be consistent with the parameters of rotational catalysis deduced by Noji et al. (19) and Yasuda et al. (20), the coiled-coil must rotate 120°in the counterclockwise direction in this process.