Interactions among γR268, γQ269, and the β Subunit Catch Loop of Escherichia coli F1-ATPase Are Important for Catalytic Activity*

Removal of the ability to form a salt bridge or hydrogen bonds between the β subunit catch loop (βY297-D305) and the γ subunit of Escherichia coli F1Fo-ATP synthase significantly altered the ability of the enzyme to hydrolyze ATP and the bacteria to grow via oxidative phosphorylation. Residues βT304, βD305, βD302, γQ269, and γR268 were found to be very important for ATP hydrolysis catalyzed by soluble F1-ATPase, and the latter four residues were also very important for oxidative phosphorylation. The greatest effects on catalytic activity were observed by the substitution of side chains that contribute to the shortest and/or multiple H-bonds as well as the salt bridge. Residue βD305 would not tolerate substitution with Val or Ser and had extremely low activity as βD305E, suggesting that this residue is particularly important for synthesis and hydrolysis activity. These results provide evidence that tight winding of the γ subunit coiled-coil is important to the rate-limiting step in ATP hydrolysis and are consistent with an escapement mechanism for ATP synthesis in which αβγ intersubunit interactions provide a means to make substrate binding a prerequisite of proton gradient-driven γ subunit rotation.

The F 1 F o -ATP synthase 1 uses a non-equilibrium transmembrane proton gradient to catalyze the formation of ATP from ADP and inorganic phosphate. The enzyme consists of two protein complexes, the membrane embedded F o complex, which couples proton translocation to the synthesis of ATP, and the membrane extrinsic F 1 complex, which contains the catalytic sites. The F 1 portion consists of five subunits that occur with a stoichiometric ratio of (␣␤) 3 ␥␦⑀. The ␣ and ␤ subunits are arranged alternately similar to the sections of an orange around a central ␥ subunit stalk (see Fig. 1A). Catalytic activity occurs at the ␣␤ interfaces, primarily within each ␤ subunit. The F 1 portion can be isolated from F o and function as an ATPase (1). The F 1 functions as an ATP driven rotary motor (2), which moves through a complete 360°rotation in three discrete 120°steps. The binding of Mg 2ϩ -ATP to a catalytic site initiates a 90°rotation of the ␥ subunit to form an intermediate state. Following a 2-ms pause, a 30°rotation concurrent with product release completes the catalytic cycle (3).
Differences in the conformation of the ␥ subunit have been observed between ground state F 1 structures (4,5) and the (ADP⅐AlF 4 Ϫ ) 2 F 1 structure by Menz et al. (6), a putative posttransition state structure that contains Mg 2ϩ ADP and SO 4 2Ϫ at the low affinity catalytic site and the transition state analog Mg 2ϩ -ADP-fluoroaluminate bound at the other two catalytic sites. In this intermediate state structure, the position of the ␥ subunit used to attach the probe for rotation studies is rotated approximately 30°from its position in ground state structures. As a consequence, the coiled-coil of the ␥ subunit is more tightly wound in the (ADP⅐AlF 4 Ϫ ) 2 F 1 structure than in the ground state.
The major specific interaction between the helical coiled-coil of the ␥ subunit and the (␣␤) 3 subcomplex occurs with ␥R268 2 and ␥Q269 in all of the F 1 structures. These residues form a "catch" with a loop of the ␤ subunit in the empty catalytic site conformation that encompasses residues 297-305 (Fig. 1A). In the ground state structure, this catch results from a salt bridge between ␥R268 and ␤D302 and from H-bonds among ␥Q269, ␤D302, and ␤T304 (Fig. 1B). In the (ADP⅐AlF 4 Ϫ ) 2 F 1 structure, the catch loop of the catalytic site with bound Mg 2ϩ -ADP and SO 4 2Ϫ also interacts with ␥R268 and ␥Q269. However, the interactions of these residues with the catch loop in (ADP⅐AlF 4 Ϫ ) 2 F 1 differ from the ground state, such that ␥R268 and ␥Q269 are rotated approximately 15°as shown in Fig. 1C. Among other differences, ␥R268 forms a second salt bridge with ␤D305 of the catch loop in (ADP⅐AlF 4 Ϫ ) 2 F 1 . The region of the ␥ subunit that interacts with the ␤ subunit catch loop was one of three ␥ subunit locations where second site mutations suppressed the deleterious effects of ATP synthase activity caused by the F 1 ␥M23K mutant (7). Based on these observations, Al-Shawi et al. (7) concluded that ␥M23K decreases the coupling efficiency of F 1 F o because of an increase in the energy of interaction between ␤ and ␥ subunits. Catch loop residue ␤Y297 is approximately 5.5 Å from the terminal phosphate in the conformation of the catalytic site that binds Mg 2ϩ -AMPPNP (4). Site-directed mutants of this residue in Chlamydomonas chloroplast F 1 decreased Mg 2ϩ -ATPase activity and changed the electron paramagnetic resonance spectrum of vanadyl bound as VO 2ϩ -ATP to the low affinity catalytic site (8). These changes indicated that this functional surrogate for the Mg 2ϩ cofactor used ␤Y297 as a metal ligand during the initial binding of metal-nucleotide to the empty catalytic site. Based on these observations, it was proposed that intersubunit H-bonds between the ␥ subunit and the (␣␤) 3 ring prevent rotation driven by the proton gradient until the empty catalytic site binds substrate. Deformation of the catch loop-␥ subunit * This work was supported by National Institutes of Health Grant GM50202. 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.
‡ To whom correspondence should be addressed. interactions induced by substrate binding would provide an escapement mechanism that would maintain tight coupling between the proton-motive force and ATP synthesis.
In this study, we have mutated ␥R268 and ␥Q269 as well as residues in the catch loop of the ␤ subunit to assess the importance of the intersubunit interactions at this location to catalytic activity. The results presented here suggest that these subunit interactions are very important to ATP hydrolysis and synthesis.

EXPERIMENTAL PROCEDURES
Plasmid p3U ϩ containing the entire Escherichia coli unc operon and the AN887 strain of E. coli, which contains a Mu phage suppression of the endogenous unc operon, were a generous gift from T. Duncan and R. Cross. Plasmid pLysS (Novagen) that encodes T7 lysozyme was inserted to assist with breaking of the E. coli cells and to provide chloramphenicol resistance. Transformation of the AN887 strain by plasmid pLysS resulted in the strain ANLS. Construction of pXL1 was performed by insertion of a six-histidine residue tag immediately after the start codon of unc A, which encodes the F 1 ␣ subunit, using the Stratagene Chameleon double-stranded site-directed mutagenesis kit. The mutagenic primer was 5Ј-TAAGGGGACTGGAGCATGCATCACCATCACCATCA-CCAGCTGAATTCCACCGAA-3Ј. The underlined base insertions add a His 6 tag, whereas the boldface base change introduces a PvuII site (CAA to CAG). The selection primer utilized was 5Ј-CTGTGACTGGT-GACGCGTCAACCAAGTC-3Ј. Here the underlined base changes convert the unique restriction site ScaI to MluI (AGTACT to ACGCGT). The desired mutations were first identified by screening with PvuII and then confirmed by DNA sequencing using ABI prism automatic sequencing as were the sequences of all of the other mutations. Strain AN887 was transformed via insertion of pXL1 by electroporation (9). A further mutation, ␥S193C, was made to facilitate attachment of rotation probes for future studies. The mutagenic primer utilized was 5Ј-C-TGCCGTTACCGGCATGCGATGATGATGATCTG-3Ј, and an XmnI restriction site was deleted through the following mutagenic primer 5Ј-CATCATTGGAAAACGCTCTTCGGGGCG-3Ј. The resulting cell line that contains the His 6 tag, ␥S193C mutation, and XmnI deletion is referred to as XL10.
The F 1 -ATPase was purified from E. coli using a modified form of the procedure described by Wise (10). Cells were grown on LB agar plates enriched with 10 mM MgSO 4 and 20 mM glucose containing 50 g/ml ampicillin and 34 g/ml chloramphenicol. Single colonies were picked and transferred to 250-ml flasks containing LB medium with the same supplements and grown overnight at 37°C and shaken at 250 rpm. Overnight cultures were then transferred to 2-liter baffled flasks containing 1 liter of minimal medium (60 mM K 2 HPO 4 , 40 mM NaH 2 PO 4 , 15 mM (NH 4 ) 2 SO 4 ) with 50 g/ml ampicillin, 34 g/ml chloramphenicol, and glucose to a final concentration of 30 mM. Cells were grown until late log phase at 37°C and shaken at 250 rpm. The cultures were then harvested by centrifugation at 6000 ϫ g.
Prior to dissociation of F 1 from F o , membranes were washed once with 50 mM TES (pH 8.0), 40 mM ⑀-aminocaproic acid, 5% (v/v) glycerol, and 4.8 mM para-aminobenzamidine. Removal of F 1 was accomplished via washing of membranes with 5 mM TES (pH 8.0), 40 mM ⑀-aminocaproic acid, 5%(v/v) glycerol, and 1.0 mM EDTA. Dithiothreitol was excluded from all of the buffers to avoid reducing the nickel column material. Membranes were then centrifuged at 100,000 ϫ g, and the supernatant containing F 1 was concentrated down to ϳ5 ml via pressure dialysis using an Amicon YM100 membrane. Nickel affinity chromatography was utilized to purify the His 6 -tagged F 1 under native conditions as described in the Qiagen nickel-nitrilotriacetic acid Superflow product manual. Proteins removed from membranes in earlier steps were exchanged into nickel column Binding Buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 10% glycerol, 1 mM imidazole, 1 mM ATP). Buffer exchange was accomplished by concentrating the crude F 1 extract to Ͻ5% of the original volume by pressure dialysis using an Amicon YM100 membrane followed by dilution in Binding Buffer. Diluted F 1 was stirred in the presence of nickel-nitrilotriacetic acid column material for 1 h and packed into a column. Unwanted proteins were removed by flushing the column in Wash Buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 10% glycerol, 10 mM imidazole, 1 mM ATP). The purified His 6 -tagged F 1 was then removed with Elution Buffer (25 mM FIG. 1. Interactions between the ␥ subunit and the ␤ E -catch loop of bovine F 1 -ATPase. A, location of the interactions between the catch loop of ␤ E (green) and the ␥ subunit (red) in the structure reproduced from Protein Data Bank file 1E79 using Web Lab Viewer from Molecular Simulations, Inc. Subunits ␣ E and ␤ TP are not shown for clarity. Details of the ␤ E -catch loop-␥ subunit interactions in the ground state F 1 structure (B) and the (ADP⅐AlF 4 Ϫ ) 2 F 1 structure (C) are reproduced from Protein Data Bank files 1E1Q and 1H8E, respectively, using Web Lab Viewer. Residues are labeled using E. coli numbering.
Succinate-dependent growth measurements were determined by growing single colonies picked from LB agar plates overnight at 37°C and shaken at 250 rpm in 50-ml cultures of minimal medium described above with 50 g/ml ampicillin, 34 g/ml chloramphenicol, 30 mM succinate, and 600 mg/liter casein hydrolysate. Overnight cultures were then inoculated into 1-liter cultures of the same medium, grown at 37°C, and shaken at 250 rpm. Absorbance at 600 nm was used as a measure of cell density, and the doubling time of each culture was determined in the log phase of growth.
The rate of ATP hydrolysis was measured with an ATP-regenerating coupled assay that consisted of 50 mM Tris-HCl (pH 8.0), 10 mM KCl, 2.5 mM phosphoenolpyruvate, 0.15 mM-0.3 mM NADH, 50 g/ml pyruvate kinase, 50 g/ml lactate dehydrogenase, and 3 nM F 1 with 2 mM Mg 2ϩ -ATP. The rate was determined as the change in absorbance at 340 nm using a Varian Cary 50 Bio UV-visible spectrophotometer equipped with a stirred cell Peltier temperature control. The reaction was initiated by the addition of F 1 to the assay mixture. Reaction rates were calculated from data collected 8 -10 min after initiation of the reaction to allow for dissociation of the ⑀ subunit and to minimize inhibition by entrapped Mg 2ϩ -ADP (11). Arrhenius analyses of data to determine the entropic and enthalpic components of the free energy of activation were performed using standard Equations 1-3 (12), where k cat is the turnover number, E A is the Arrhenius activation energy, N is Avogadro's number, and ⌬H ‡ and ⌬S ‡ are the enthalpic and entropic components of the changes in Gibbs free energy of activation (⌬G ‡ ). All of the Mg 2ϩ -ATPase assays were accomplished within 5 days of the date at which the F 1 -ATPase preparation was completed. After this period, the preparations were found to lose activity as the result of an increase in the entropy of activation (data not shown).

RESULTS
With the exception of ␤D305V and ␤D305S, the yield of F 1 purified from the site-directed mutants to remove ␥ subunitcatch loop interactions was approximately the same as that isolated from the XL10 strain. Intact F 1 -ATPase that contained the ␤D305V or the ␤D305S mutants was never successfully isolated, and these mutations were presumed to interfere with assembly of the F 1 F o -ATP synthase. The F 1 -ATPase isolated from E. coli with the other mutations (␥R268L, ␥Q269L, ␤D302T, ␤D302V, ␤T304A, or ␤D305E) contained all five subunits as determined by SDS-polyacrylamide gel electrophoresis (data not shown). These results suggest that these latter mutations did not significantly affect the synthesis and assembly of the enzyme.
The relative ability of the mutant and wild type strains to grow via oxidative phosphorylation on minimal medium in the presence of succinate was assessed by determining growth curves. The inability of ␤D305V and ␤D305S mutants to assemble intact F 1 -ATPase coupled with their inability to grow on minimal medium with succinate was utilized as a negative control in which the growth rate was dependent on the rate of ATP synthesis catalyzed by the F 1 F o -ATP synthase. The doubling times calculated from the growth curves are summarized in Table I.
The F 1 F o demonstrated very little tolerance for the substitution of ␤D305. In addition to the fact that substitution of this carboxyl for a hydroxyl prevented assembly of F 1 , the most conservative mutation, ␤D305E, completely impaired the succinate-dependent growth rate. The strain that contained the ␥Q269L mutation was also unable to grow on succinate, indicating that these residues are very important for ATP synthase activity. The ability of the strains that contained ␥R268L, ␤D302V, or ␤D302T mutants to grow on succinate was also significantly decreased. The doubling time of the strain containing ␥R268L was nearly twice that of XL10, whereas that of the strains that contained mutants of ␤D302 increased by approximately 1.5-fold. However, the rate of growth of the ␤T304A mutant on succinate was not affected. None of the mutations made to either the ␤ or ␥ subunits affected the ability of the bacteria to grow on minimal medium in the presence of glucose. These results suggest that the poor growth on minimal medium with succinate was the result of impairment of the ability of the F 1 F o complex to synthesize ATP.
The mutations that interfered with ␥ subunit-catch loop interactions also affected the ATPase activity of isolated F 1 significantly. Fig. 2 shows an Arrhenius plot the Mg 2ϩ -ATPase activity. With the exception of the ␥Q269L mutant, the mutations decreased the temperature stability of the enzyme. Although XL10-F 1 was stable to 50°C, ␤T304A-F 1 and ␤D305E-F 1 were stable only to 32.5°C, whereas ␤D302T-F 1 and ␥R268L-F 1 reached their stability limits at 40 and 27.5°C, respectively. Consequently, a direct comparison of the effects of the mutations on k cat was made at 25°C where all of the enzymes were stable as shown in Table I.
The F 1 -ATPase that contained the ␤D302V mutation had no detectable ATP hydrolysis activity over the temperature range of 5-50°C despite the fact that this mutant retained partial ability to grow on minimal medium with succinate ( Table I). The more conservative ␤D302T mutation retained approximately 3% of the XL10-F 1 activity. The effect of this mutation on ATPase activity was substantially greater than its effect on succinate-dependent growth. The k cat values of ␤D305E-F 1 and ␥Q269L-F 1 were only 4 and 1% XL10-F 1 , respectively, which were comparable to their effects on the succinate-dependent growth rate. It is noteworthy that the temperature stability of the ␥Q269L-F 1 mutant was comparable to XL10-F 1 despite the fact that both ATP hydrolysis and the growth rate on succinate were significantly lower than XL10-F 1 . The difference in ATPase activity and succinate-dependent growth of the ␤T304A and ␥R268L mutations were 2.5-3-fold with k cat values for ATP hydrolysis that were 45 and 16% XL10-F 1 , respectively.
When the data are plotted as in Fig. 2, the activation energy (E A ) is determined directly from the slope. The values for enthalpy of activation, ⌬H ‡ , of ATP hydrolysis are directly proportional to E A as per Equation 1 and were calculated from the data indicated by the solid lines in the Arrhenius plot as summarized at 25°C in Fig. 3. For every mutant that retained Mg 2ϩ -ATPase activity, ⌬H ‡ was significantly lower than ob-  Fig. 3 also shows the entropy of activation, T⌬S ‡ , and free energy of activation, ⌬G ‡ , derived from the data in Fig. 2. The free energy of activation is inversely proportional to k cat as per Equation 3. Because ⌬G ‡ is the result of the difference between ⌬H ‡ and T⌬S ‡ (Equation 2), the free energy barrier of the reaction will increase as the value of T⌬S ‡ becomes more negative. Although all of the mutants lowered ⌬H ‡ , lower rates of ATP hydrolysis were observed in every case because the entropic component of the energy barrier for the reaction increased (became more negative) to a greater extent than the decrease in ⌬H ‡ . The value of T⌬S ‡ for ATP hydrolysis of ␥R268L-F 1 and ␤D305E-F 1 was Ϫ39.4 kJ/mol, nearly a 6-fold difference from the XL10-F 1 value of Ϫ7.5 kJ/mol. A 3.5-and 4.5-fold difference in T⌬S ‡ from that of XL10-F 1 was observed with ␤D302T-F 1 (Ϫ32.7 kJ/mol) and ␥Q269L-F 1 (Ϫ35.7 kJ/ mol), respectively. Mutant ␤T304A, which retained its ability to grow on succinate and had the highest k cat for ATP hydrolysis, had the smallest changes in both ⌬H ‡ and T⌬S ‡ . The more negative values of T⌬S ‡ observed in all of these mutants are consistent with an increase in disorder due to the weakening of or loss of either a hydrogen bond or salt bridge. DISCUSSION The results presented here imply that hydrogen bonds and salt bridges between the ␤ subunit catch loop and the ␥ subunit are very important to the catalytic function of the enzyme. Residues ␤T304, ␤D305, ␤D302, ␥Q269, and ␥R268 were found to be very important for ATP hydrolysis catalyzed by soluble F 1 -ATPase, and the latter four residues were also very important for oxidative phosphorylation. At a resolution of 2.4 Å, typical of the available F 1 crystal structures, it is difficult to determine whether a hydrogen bond is present or assesses its relative strength based on structural information alone. It is remarkable that single mutations to remove the possibility to make a single hydrogen bond or salt bridge outside the catalytic site can have such large effects on catalytic activity.
Several of the residues in the catch loop make a more important contribution to ATP hydrolysis than to ATP synthesis. The greatest differential effects were observed with mutants to ␤D302, although the ␥R268 and ␤T304 mutants also affect hydrolysis to a greater extent than synthesis. Even though mutations to these latter residues caused 2.5-3-fold differences in the decrease of hydrolysis and synthesis, the results indicate that ␥R268 was much more important to synthesis than ␤T304. These differential effects probably result from the fact that ATP hydrolysis was measured with isolated F 1 , whereas the competency of ATP synthesis was assessed by the growth rate of the E. coli strains on succinate. In the latter case, the fully assembled F 1 F o -ATP synthase may have a different rate-limiting step than isolated F 1 .
Relationship between Catch Loop-␥ Subunit Interactions and ATP Hydrolysis-The rate-limiting step of the ATPase reaction of isolated F 1 occurs after a 90°rotation of the ␥ subunit induced by Mg 2ϩ -ATP binding (3). The catalytic cycle is completed by a 30°rotation of the ␥ subunit concurrent with product release. In the (ADP⅐AlF 4 Ϫ ) 2 F 1 crystal structure, the portion of the ␥ subunit used to attach the rotation probe is rotated approximately 30°from the ground state structure (6). This crystal structure contains Mg 2ϩ -ADP and SO 4 2Ϫ at the low affinity catalytic site and the transition state analog Mg 2ϩ -ADP-fluoroaluminate at the other two catalytic sites.
Although the foot region of the ␥ subunit, the point of attachment for the rotation probe, was found to be rotated about 30°i n (ADP⅐AlF 4 Ϫ ) 2 F 1 , the C terminus of this subunit is in nearly the same position as the other F 1 structures. Consequently, the helical coiled-coil of the ␥ subunit is wound more tightly in (ADP⅐AlF 4 Ϫ ) 2 F 1 than the ground state structure. This implies that some of the torque on the ␥ subunit generated upon substrate binding may be used to wind the coiled-coil more tightly by inducing a 120°rotation of the C terminus while the foot of the ␥ subunit rotates 90°. Relaxation of the coiled-coil during the final 30°rotation could contribute to the energy needed to complete the rate-limiting step of the reaction.
The results presented here are consistent with a role for the intermolecular interactions of the catch loop serving to promote the formation of the tightly wound form of the coiled-coil in a The concentration of F 1 used was 3 nM for XL10-F 1 , ␤D305E-F 1 , and ␤T304-F 1 mutants, 9 nM for ␥R268L-F 1 , 40 nM for ␥Q270L-F 1 , and 25 nM for ␤D302T-F 1 . ATPase activities were assayed at 2 mM Mg 2ϩ -ATP every 2.5°C from 5-50°C as described under "Experimental Procedures." The lines plotted were generated by linear least squares regression of the data. manner that provides energy for the final 30°rotation during the rate-limiting step. Because product release is rate-limiting to F 1 -ATPase activity (3), the thermodynamic parameters measured here provide information concerning this step. Despite the fact that all of the mutants decreased ⌬H ‡ , the changes in entropy of activation more than compensated for the more favorable values of ⌬H ‡ , thereby significantly lowering k cat . These changes in T⌬S ‡ can be explained if the elimination of the salt bridges and/or H-bonds between the ␤ subunit-catch loop and ␥ subunit increased the number of allowable conformations of the F 1 subunits during the rate-limiting step. Although some of these additional conformations dramatically lower the activation energy barrier, and thus ⌬H ‡ , from that of XL10-F 1 , many more of them are nonproductive. The additional time needed for the enzyme to adopt a productive conformation in the absence of the salt bridge or H-bond leads to the decrease in k cat . It is noteworthy that ␤D305, the residue that tolerated mutations least, forms a salt bridge to the ␥ subunit in (ADP⅐AlF 4 Ϫ ) 2 F 1 , which may represent the rate-limiting conformation, but not in the ground state structures.
Relationship between Catch Loop-␥ Subunit Interactions and ATP Synthesis-The results presented here also indicate that catch loop-␥ subunit interactions are important to ATP synthesis. The succinate-dependent growth rate of the ␥Q269L and ␤D305E strains was 1% of wild type, suggesting that these residues were essential for ATP synthase activity. The ␥R268L and both ␤D302 mutants also decrease the growth rate by 4and 2-fold, respectively. These results are consistent with the hypothesis that the residues at the catch loop serve as an escapement mechanism during ATP synthesis (12). In this mechanism, the trans-membrane proton gradient provides constant torque to the ␥ subunit (via the c-subunit ring). The interactions between the ␥ and ␣␤ subunit rings prevent this rotation until the empty catalytic site binds substrate. When the H-bonds and salt bridges at the catch loop are broken as the result of substrate binding, the torque on the ␥ subunit is greater than the energy in the remaining H-bonds, such that rotation of the ␥ subunit induces the conformational changes in the catalytic sites necessary for ATP synthesis. The results presented here suggest that residues ␥Q269, ␥R268, ␤D305, and ␤D302 contribute to the restraint of the rotation of the ␥ subunit during ATP synthesis prior to the binding of substrate.
Zhou et al. (13) demonstrated that substrate binding was a prerequisite of proton gradient-driven rotation in E. coli F 1 F o by direct observation of rotation via the FLAG epitope. Mutations to residues in the catch loop have been found in naturally occurring second site revertants of the inhibitory ␥M23K mu-tant (7). The ␥M23K mutation was postulated to form an additional H-bond to the DELSEED region of the ␤ subunit. We note that in every case the reported revertants cause the loss or weakening of a salt bridge or an H-bond either in the catch loop or near the ␥ subunit C terminus. Second site mutations to restore ATP synthase activity to ␥Q269E or ␥T273V had similar effects and identified a third important location in the ␥ subunit N terminus (14). We note that several intersubunit H-bonds are present at the N terminus of the ␥ subunit. These results suggest that the sum of the energy in the intersubunit H-bonds and salt bridges, regardless of their location, must not be above or below a certain value for the enzyme to function.
If the steps in ATP synthesis are the reverse of those during hydrolysis, the first synthesis step involves a 30°rotation of the ␥ subunit that forms the tightly wound coiled-coil. The available data suggest that the intersubunit H-bonds and salt bridges appear to be stronger in this conformation. Consequently, the proton-motive force may be capable of driving the 30°rotation, even though substrate has not bound to the empty catalytic site. With the mutants described here, the energy of the proton gradient may be sufficient to induce rotations greater than 30°that induce conformations of the catalytic sites not conducive to product formation.