Energy Coupling, Turnover, and Stability of the F0F1 ATP Synthase Are Dependent on the Energy of Interaction between γ and β Subunits

Replacement of the F0F1 ATP synthase gamma subunit Met-23 with Lys (gammaM23K) perturbs coupling efficiency between transport and catalysis (Shin, K., Nakamoto, R. K., Maeda, M., and Futai, M. (1992) J. Biol. Chem. 267, 20835-20839). We demonstrate here that the gammaM23K mutation causes altered interactions between subunits. Binding of delta or epsilon subunits stabilizes the alpha3beta3gamma complex, which becomes destabilized by the mutation. Significantly, the inhibition of F1 ATP hydrolysis by the epsilon subunit is no longer relieved when the gammaM23K mutant F1 is bound to F0. Steady state Arrhenius analysis reveals that the gammaM23K enzyme has increased activation energies for the catalytic transition state. These results suggest that the mutation causes the formation of additional bonds within the enzyme that must be broken in order to achieve the transition state. Based on the x-ray crystallographic structure of Abrahams et al. (Abrahams, J. P., Leslie, A. G. W., Lutter, R., and Walker, J. E. (1994) Nature 370, 621-628), the additional bond is likely due to gammaM23K forming an ionized hydrogen bond with one of the betaGlu-381 residues. Two second site mutations, gammaQ269R and gammaR242C, suppress the effects of gammaM23K and decrease activation energies for the gammaM23K enzyme. We conclude that gammaM23K is an added function mutation that increases the energy of interaction between gamma and beta subunits. The additional interaction perturbs transmission of conformational information such that epsilon inhibition of ATPase activity is not relieved and coupling efficiency is lowered.

The F 0 F 1 ATP synthase links two disparate functions: transport of protons across a membrane and catalysis of ATP synthesis or hydrolysis (for reviews see Refs. [1][2][3][4][5]. The fully cooperative mechanism of ATP hydrolysis requires a minimum of three different subunits in a complex containing ␣ 3 ␤ 3 ␥ 1 . The transport mechanism is most likely assembled from F 0 sector subunits. In the Escherichia coli complex, transport requires three different membrane-spanning subunits, a 1 b 2 c ϳ10 (6). In addition, two more soluble subunits, ␦ and ⑀, are needed to reconstitute catalytic and transport sectors so that they are coupled to carry out ATP-driven proton pumping or ⌬ Hϩdriven ATP synthesis (7)(8)(9)(10).
Catalysis and transport mechanisms most likely communicate indirectly through a series of conformational and electrostatic interactions. Conformational changes relevant to the catalytic state of the enzyme or the presence of a ⌬ H ϩ have been detected by several methods including altered cross-linking patterns, protease susceptibility, environmentally sensitive fluorescent probes, accessibility of epitopes, x-ray diffraction, cryoelectron microscopy, and spectroscopic analyses (reviewed in Refs. 11 and 12). High resolution structural information based on crystals of the bovine mitochondrial F 1 has also provided a great deal of information about possible subunit interactions that may be involved in linking transport and catalysis (13).
Mutagenic analysis has also yielded important information about the coupling mechanism (reviewed in Ref. 5). For example, mutations in the single hydrophilic loop of subunit c, an F 0 subunit that is involved in proton transport, disrupt coupling between transport and catalytic mechanisms (14 -16). Furthermore, genetic and chemical cross-linking results strongly suggest that ⑀ subunit interacts with this portion of subunit c (17,18).
Likewise, mutations near the catalytic sites have also been found to affect coupling. The most clear example of these mutations is replacement of ␥Met-23 with Arg or Lys (19). These mutations caused greatly reduced ATP-dependent proton pumping and ATP synthesis rates without strongly affecting catalytic or transport functions. The F 0 sector was unaffected as were interactions between F 0 and F 1 . Restoration of efficient coupling in the ␥Met- 23 3 Lys (␥M23K) mutant enzyme was conferred by several second site mutations near the carboxyl terminus of the ␥ subunit including the replacement of amino acid ␥Arg-242 and seven different residues between ␥Gln-269 and ␥Ala-280 (20). Furthermore the temperature sensitivity caused by the ␥M23K mutation was suppressed by each of the second site mutations. In the simplest interpretation of these results, direct interactions within the ␥ subunit are perturbed by the ␥M23K mutation, and each of the suppressor mutations is close enough to the ␥M23K residue to directly counteract its influence on structure and function. However, many observations do not coincide with this interpretation (21). Analysis of the x-ray crystallographic structural map of the bovine F 1 (13) indicates that ␥Met-23 is a considerable distance from the ␥269 -280 region. Furthermore, the types of amino acid changes that suppress the effects of the ␥M23K mutant are not of any particular functional group, and in some cases multiple changes at a single site result in suppression (20,21). It is difficult to imagine how multiple and diverse amino acid replacements would be able to compensate for a specific perturbation.
Because the mechanism of coupling must involve interactions among subunits, we hypothesized that the ␥M23K mutation perturbed interactions between subunits. In this paper, we provide evidence that the ␥M23K mutation is an added function mutation that increases the energy of interaction between ␥ and ␤ subunits. The consequences of this increase are destabilization of the ␣ 3 ␤ 3 ␥ complex and inefficient coupling between transport and catalysis. In turn, the suppressor mutations described above counteract by decreasing the energy of interaction.
The ␥M23K mutation was introduced into the uncG gene on plasmid pBWU13 to give pBMU13-␥M23K. Oligonucleotide-directed mutagenesis with the Stratagene (La Jolla, CA) Chameleon Kit (24) was used to introduce the ␥M23K mutation on plasmid pBWG11 (23) using the synthetic oligonucleotide, 5Ј-CACTAAAGCGaaaGAGATGGTCGCC-3Ј (lowercase letters denote the ␥M23K mutation). After sequence verification, the mutation was isolated on the AsuII to RsrII restriction fragment and ligated into pBWU13. Introduction of the mutation into the expression plasmid and presence of the mutation after growth were verified by phenotype and DNA sequencing (25).
Molecular biological manipulations were performed as described (26) or according to the manufacturer's instructions. Restriction enzyme and DNA modifying enzymes were obtained from Amersham Corp., Boehringer Mannheim, Life Technologies, Inc., New England Biolabs, (Beverly, MA), or Promega (Madison, WI).
Isolation of Membranes and Purification of F 1 and F 1 Subunits-Membranes from KF10rA were prepared as described previously (8). F 1 complex was purified as detailed by Duncan and Senior (27) and Al-Shawi and Senior (28). Purified wild-type F 1 was used as a source for isolation of ␦ and ⑀ subunits (9).
Determination of F 1 Content in E. coli Membranes-Determination of F 1 in membrane preparations was performed by quantitative immunoblot analysis and comparing results with known amounts of purified F 1 . Proteins from various amounts of membranes were prepared as in Nakamoto et al. (21), separated on a 12.5% SDS-polyacrylamide gel (29), and transferred to a nitrocellulose filter (30). The filter was immunostained with a 1:1000 dilution of a polyclonal rabbit anti-␣ subunit antibody (obtained from Dr. Alan Senior of the University of Rochester) followed by a secondary anti-rabbit IgG antibody conjugated to fluorescein diluted 1:1000 (Boehringer Mannheim). The fluorescence of the band corresponding to the ␣ subunit was quantified by a Molecular Dynamics FluorImager, and the values were plotted against total protein. By increasing amounts of total membrane protein in adjacent lanes, a linear line was generated for each membrane preparation. The slope of each line was compared with that obtained from a titration of purified E. coli F 1 to determine the percentage of total protein that could be attributed to F 1 . Duplicate experiments were performed, and the values were averaged.
Enzymatic Assays-Protein concentrations were determined by the method of Lowry et al. (31). ATPase activities were measured in the buffers given below and in the figure legends by procedures described in Al-Shawi et al. (32). The experimental conditions detailed in the footnotes to Table II were chosen to optimize for coupling efficiency and enzyme stability. Free Mg 2ϩ and Mg⅐ATP concentration were determined by the algorithm of Fabiato and Fabiato (33). ATPase reactions were stopped by the addition of 5% sodium dodecylsulfate or 10 mM ice-cold H 2 SO 4 . Liberated P i was determined by the methods of Taussky and Shorr (34) or Van Veldhoven and Mannaerts (35) depending on the sensitivity required. The Van Veldhoven and Mannaerts assay was slightly modified by stopping the final color development reaction with the addition of 0.5 M H 2 SO 4 after 20 min of incubation at room temperature. ATP synthesis was measured as described previously (36). Pyruvate kinase was obtained from Boehringer Mannheim, and hexokinase was from Sigma.
Arrhenius Analysis and Derivation of Transition State Thermodynamic Parameters-Apparent enzyme activation energies of ATP hydrolysis were calculated from measurements of maximal rates of ATPase activities as a function of temperature. Activation energies and entropic and enthalpic components of the transition state for ATP hydrolysis were calculated from plots of log velocity at saturating ATP (in the presence of an ATP regenerating system) versus the reciprocal temperature as detailed in Al-Shawi and Senior (37). For membrane preparations (0.13-0.3 mg/ml), turnover numbers were calculated using the determined F 1 content of the membrane (measured as described above), and assays were performed in the presence of 5 M carbonylcyanide-m-chlorophenylhydrazone (CCCP). 1 For purified F 1 , enzyme preparations were preincubated in buffer at 23°C for at least 30 min, and the assay concentration was 65 nM F 1 . The ATPase buffer, comprised of 0.5 ml of 50 mM HEPES-KOH, 10 mM ATP, 5 mM MgSO 4 , 5 mM phosphoenolpyruvate, and 32 g/ml pyruvate kinase, was incubated at the required temperature, and the pH was adjusted to 7.5 using a pH electrode (Sigma) that had been calibrated at that temperature. The linear rate of P i liberation was determined as described above.

RESULTS
The ␥M23K Mutation Does Not Alter ⑀ or ␦ Subunit Interactions with F 1 -We first assessed if the ␥M23K mutation altered interactions with the single copy F 1 subunits, ␦ and ⑀. In the wild-type E. coli complex, ⑀ subunit binds tightly to the F 1 complex with a K D of approximately 10 Ϫ8 M (38), and the association constant of isolated ␥ and ⑀ subunits is similar to this value (39). These results suggested that ⑀ interactions with the remainder of the F 1 complex are mostly through the ␥ subunit. In turn, the affinity for ⑀ subunit to ␥ subunit can be assessed by the inhibitory properties of the ⑀ subunit on ATPase activity. ATPase activity of the ␣ 3 ␤ 3 ␥␦ complex is inhibited approximately 90% by ⑀ subunit; the K I is very close to the binding constant between isolated ␥ and ⑀ subunits (39).
Sternweis and Smith (38) showed that dilution of the purified F 1 complex to a concentration below the K D for ⑀-F 1 results in activation of ATPase specific activity. Fig. 1 reproduces this result where the enzyme is activated as F 1 is diluted below 100 nM. The same activation is observed for wild-type and ␥M23K enzymes. Titration of the wild-type F 1 indicates a half-maximal activation at approximately 10 nM, which is the same value as previously reported (9,38). ␦ subunit is generally believed to bind F 1 with lower affinity than ⑀ subunit and is most likely dissociated at the concentration of 10 nM F 1 (40).
In a marked difference from wild type, further dilution of ␥M23K F 1 caused inactivation of activity with half-maximal inactivation occurring at 0.2 nM. The inactivation was not detected with the wild-type F 1 even at the lowest F 1 concentration measured, 0.01 nM. This indicates that the ␥M23K mutation causes destabilization of ␣ 3 ␤ 3 ␥, the minimum complex capable of ATPase activity (32,41).
Interestingly, the inactivation of the ␥M23K enzyme does not occur until ⑀ and ␦ subunits have dissociated and suggests that binding of ⑀ and ␦ subunits may help to stabilize the complex. To test this notion, superstoichiometric amounts of purified ⑀ or ␦ subunits were added, while ␥M23K F 1 was diluted to various concentrations (Fig. 2). In the presence of 72 nM ⑀ subunit alone, the ATPase specific activity of 13 nM ␥M23K F 1 (the concentration that gave maximal activation) was decreased as expected due to ⑀ inhibition. At 1.3 nM ␥M23K F 1 , the ATPase specific activity remained about the same, and at 0.13 nM ␥M23K F 1 , the activity increased more than 2-fold compared with the absence of added ⑀ subunit. This behavior reflects the balance between the dissociation/association of ⑀ subunit and inactivation of ␣ 3 ␤ 3 ␥.
Results with 72 nM ␦ subunit were even more dramatic because the activity was stabilized even at the most dilute F 1 concentrations. The activities at the two lower concentrations were maximal, indicating that the concentration of ⑀ subunit was well below its K I . Clearly, ␦ subunit alone binds to the complex independent of the ⑀ subunit and in a manner that stabilizes ␣ 3 ␤ 3 ␥. With both ␦ and ⑀ subunits added, activity was consistent with ⑀-inhibited levels.
The stabilization of activity by ␦ subunit alone provided a way to directly assess the K I for ⑀ subunit. ␥M23K F 1 was diluted to 0.13 nM in the presence of 72 nM ␦ subunit to maintain the stabilized complex. Titration of ⑀ subunit resulted in an apparent K I of 13 nM (data not shown), which is in good agreement with the dilution experiments in Fig. 1.
␥M23K F 1 Does Not Release ⑀ Inhibition upon Binding to F 0 -In order to more fully analyze the kinetic and thermodynamic properties of the ␥M23K enzyme, we accurately determined the concentration of F 0 F 1 complex in the membrane of E. coli strain KF10rA. This was done by a quantitative immunoblot analysis described under "Materials and Methods." In brief, the amount of immunostaining obtained for each of the membrane preparations was compared with the amount of staining of known amounts of purified F 1 loaded on the same gel (data not shown).
Knowing the amount of F 1 on the membranes, we were able to derive turnover numbers for native F 0 F 1 in membranes. Fig.  3 compares the turnover numbers for wild-type and ␥M23K enzymes as soluble F 1 or membranous F 0 F 1 . The wild-type F 0 F 1 turnover was 425 s Ϫ1 (at 30°C) compared with 92 s Ϫ1 for soluble F 1 , which is about 85% replete with the inhibitory ⑀ subunit (Figs. 1 and 2). This result confirms the early work of Sternweis and Smith (38) that ⑀ inhibition is relieved upon binding to F 0 . Interestingly, the turnover numbers of the ␥M23K F 1 and F 0 F 1 are quite similar (132 s Ϫ1 and 92 s Ϫ1 , FIG. 1. Effect of dilution on the ATPase activity of wild-type and ␥M23K F 1 . Various concentrations of F 1 indicated were preincubated for 1 h at 37°C in 50 mM Tris-OH, 50 mM HEPES, 20 mM Na 2 SO 4 , and 0.1 mM MgSO 4 , adjusted to pH 7.5 with H 2 SO 4 . These buffer conditions were predetermined to be optimal for coupling efficiency and stability of the F 0 F 1 complex. ATP hydrolysis was started by the addition of a stock solution of Mg⅐ATP in the above buffer such that the final concentration was 10 mM ATP and 5 mM MgSO 4 , and the final pH remained constant at 7.5. 50 -200-l samples of reaction mix were quenched at various times with 1 ml of 10 mM H 2 SO 4 , and P i liberation was quantitated as described under "Materials and Methods." The linear time course of the reactions were followed in steady state conditions for 10 s to 60 min as required. The results shown are average values of at least three independent determinations (Ϯ standard deviation bars). q, Normal wild-type F 1 ; å, ␥M23K F 1 .
FIG. 2. Effect of adding ␦ and ⑀ subunits on the ATPase activities ␥M23K F 1 at various concentrations. Various concentrations ␥M23K F 1 were preincubated as in Fig. 1 with superstoichiometric concentrations of purified ␦ and ⑀ subunits as indicated. ATPase activities were determined as in Fig. 1. Each result is the average of at least three independent experiments with the standard deviations indicated .   FIG. 3. Turnover of wild-type and ␥M23K F 1 or F 0 F 1 . ATPase assays were performed at 30°C as detailed under "Materials and Methods" in a buffer containing 50 mM HEPES-KOH, 10 mM ATP, 5 mM MgSO 4 , pH 7.5, with 5 mM phosphoenolpyruvate and 32 g/ml pyruvate kinase as an ATP regenerating system. 65 nM F 1 or 0.13-0.30 mg of membrane protein/ml was used in each assay. 5 M carbonylcyanidem-chlorophenylhydrazone was included with the membrane assays to ensure that there was no back inhibition from ⌬ Hϩ . F 1 concentration in the membranes was quantitated as described under "Materials and Methods." Turnover numbers were calculated using a molecular mass of 3.82 ϫ 10 5 Daltons for the F 1 complex.
FIG. 4. Arrhenius analysis of wild-type and ␥M23K F 1 or F 0 F 1 . Log V (maximal velocity at saturating ATP in mol/mg/min for clarity) is plotted against the reciprocal of absolute temperature. ATPase activities were assayed from 5 to 45°C as detailed under "Materials and Methods." The lines plotted were generated by linear least squares regression of the data. q, wild-type F 1 ; f, wild-type F 0 F 1 ; å, ␥M23K F 1 ; ç, ␥M23K F 0 F 1 -containing membranes. respectively). Two important observations are to be made from these values. First, the ␥M23K mutation does not affect the catalytic mechanism because wild-type and the ␥M23K enzymes have relatively similar turnover numbers, and second, the ⑀ inhibition of the ␥M23K enzyme is not relieved when bound to F 0 . The latter results may indicate that the ␥M23K mutation perturbs the functional interaction of F 1 with F 0 as mediated by the ⑀ subunit.
Altered Transition State Thermodynamic Parameters in the ␥M23K Mutant Enzyme-In order to understand the effects of the ␥M23K mutation, we investigated the results of the mutation on the catalytic mechanism under steady state conditions. We already saw that the turnover of the ␥M23K F 0 F 1 was similar to that of wild type. Furthermore, both enzymes had a pH optimum of around 8.5-9.0 as well as the K m values for Mg⅐ATP hydrolysis being similar (0.16 and 0.32 mM for wildtype and ␥M23K F 0 F 1 , respectively; data not shown). These results reinforce the conclusion that the general reaction schemes and cooperative mechanisms of the mutant enzyme were similar to those of the wild-type enzyme. Additional support for this conclusion can be seen later in the "isokinetic" plots (see Fig. 8) in that the mutant enzyme preparations were close to the regression lines.
However, we have previously demonstrated that the catalytic transition state of the F 0 F 1 ATP synthase was very sensitive to changes in catalytic site conformation and the utilization of binding energy to drive catalysis (42). Thus, in order to probe the effects of the ␥M23K mutation on catalysis and coupling, we measured the thermodynamic parameters of the transition state of ATP hydrolysis by Arrhenius analysis of steady state turnover. In contrast to the results above, the activation energies for the catalytic transition state were strongly affected. Fig. 4 shows an Arrhenius plot of steady state ATPase activity (for clarity, log of the actual ATPase specific activities are plotted instead of turnover). Temperature dependence of maximal velocities were measured with 5 mM Mg⅐ATP. In the case of membrane-bound enzymes, the protonophore, carbonylcyanide-m-chlorophenylhydrazone, was added to prevent back inhibition from the electrochemical gradient of protons. Purified F 1 from wild type and ␥M23K had linear plots from 5 to 45°C, whereas the membranous F 0 F 1 had a break in the plot around 19°C. The break in the Arrhenius plot is clearly due to an effect of F 0 on F 1 . This effect is likely a manifestation of the influence of the lipid phase on the function of the F 0 , which is communicated to the catalytic mechanism through coupling. Significantly, the change in the slope for ␥M23K F 0 F 1 is much less pronounced than that for the wildtype enzyme, suggesting that the influence of F 0 on catalysis is decreased in the mutant complex.
From these plots the enthalpic, entropic, and free energy terms can be calculated for the transition state of the reaction pathway. These values for the wild-type and ␥M23K F 1 at 30°C are listed in Table I and plotted in Fig. 5A. The differences for each parameter between wild type and ␥M23K are plotted in Fig. 5B. Note that the membrane-bound enzyme had the same trends as soluble F 1 , but the ␥M23K F 0 F 1 had considerably larger differences in both enthalpic and entropic terms. For both soluble and membrane-bound complexes, the ␥M23K enzyme has a more positive ⌬H ‡ and a less negative ⌬S ‡ which together add up to a small difference in ⌬G ‡ . According to transition state theory, these results suggest that the mutation causes the formation of additional bonds between substrate and enzyme or, more likely in this case, bonds within the enzyme that must be broken in order to achieve the transition state.
Suppressor Mutations of ␥M23K Reverse the Effects on Activation Energies-If the changes in transition state thermodynamic parameters are due to the same perturbations that causes the uncoupling phenotype of the ␥M23K mutation, then second site suppressor mutations of ␥M23K (20) would be expected to reverse the altered thermodynamic parameters. This was the case with the suppressor mutation, ␥Q269R.
This mutation was most effective at restoring efficient coupling. The ratio of NADH-driven ATP synthesis versus ATP hydrolysis can be used as a parameter of coupling efficiency (21,43). Table II shows that the synthesis:hydrolysis ratio of the ␥M23K mutant is restored to wild-type levels in the presence of ␥Q269R. As predicted, the Arrhenius analysis shows that the ␥Q269R mutation counteracted the effects of the FIG. 5. Transition state thermodynamic parameters for steady state ATP hydrolysis by wild-type and ␥M23K F 1 or F 0 F 1 . The activation energy parameters of k cat were calculated at 30°C from the data of Fig. 4 as described under "Materials and Methods." A illustrates activation energy parameters for wild-type membranes F 0 F 1 (open bars) or wild-type F 1 (hatched bars). B illustrates the differences between activation energy parameters for ␥M23K preparations and the corresponding wild-type preparations. Open bars represent F 0 F 1 preparations, and hatched bars represent F 1 preparations. ␥M23K mutation on the thermodynamic parameters of the ATPase transition states (Fig. 6).
In contrast, another suppressor mutation, ␥R242C, did not increase the coupling efficiency even though it genetically suppressed the ␥M23K mutation (20) and resulted in increased ATP synthesis rates (Table II). Instead, the mutation suppressed the effects of the ␥M23K mutation by increasing the turnover rate of the enzyme by 1.44-fold. This result is reminiscent of our observations that overexpression of the ␥M23K F 0 F 1 resulted in increased ATP synthesis rates and increased growth yields on the nonfermentable carbon source, succinate (20). 2 With more enzyme present, albeit an inefficient one, there was sufficient ATP synthesis to allow growth of the cell.
Consistent with the lack of recovery of efficient coupling, differences in the transition state thermodynamic parameters of the ␥M23K mutant were only slightly reduced by introduction of ␥R242C (Fig. 6). In this case, it appears that the reduced energy of interaction between ␥ and ␤ subunits resulted in a faster turnover rate. DISCUSSION We have sought to understand how the ␥M23K mutation perturbs linkage between transport and catalysis to provide information about the mechanism of coupling. We have seen that the mutation does not affect F 1 interactions with ⑀ and ␦ subunits (this report) nor the F 0 sector (19); however, dissociation of ␦ and ⑀ subunits leaves a destabilized ␣ 3 ␤ 3 ␥ complex (Fig. 1). Based on chemical cross-linking experiments, ⑀ subunit is believed to interact with ␥, ␤, and ␣ subunits (44 -47) and apparently does so with a stabilizing effect on the ␣ 3 ␤ 3 ␥ complex. An important property of the ⑀ subunit is its inhibitory activity on the hydrolysis activity of F 1 . Most significantly, in contrast to the situation in wild type, inhibition by ⑀ subunit in the ␥M23K mutant is not relieved upon binding to F 0 (Fig.  3), suggesting a perturbation in communication between transport and catalytic mechanisms. This notion was supported by the decreased change in slope of the Arrhenius plot for the membranous ␥M23K enzyme (Fig. 4). Both of these effects illustrate the impaired coupling between F 1 and F 0 functions in the mutant enzyme ␥M23K.
From the above data, it seems that the ␥M23K mutation affects the regions of interactions between ␥ and ␤ subunits. Based on suppressor mutagenesis results, we earlier concluded that three highly conserved regions of the ␥ subunit, ␥18 -35, ␥238 -246, and ␥269 -280, functionally interact as a domain that mediates energy coupling (21). Upon close inspection of the x-ray crystallographic structural model of Abrahams et al. (13), we observed that each of these three regions is in contact with the surrounding ␤ subunits and that at least two of the residues form a hydrogen bond with specific ␤ subunit residues, namely ␥Gln-269 with ␤Thr-304 and ␥Arg-242 and ␤Glu-381 (see below and Fig. 7A). It is apparent that effects of ␥ subunit mutations in these regions on catalysis are due to perturbation of the interactions with the catalytic ␤ subunits. Thermodynamic analysis of transition state activation energies for ATP hydrolysis reactions strengthen this notion. Significantly, ␥M23K mutant enzyme had dramatically more positive ⌬H ‡ and less negative ⌬S ‡ parameters, suggesting that the amino acid replacement caused an extra bond to form that must be broken in order to achieve the catalytic transition state. Fig. 7B shows details of the bovine F 1 structural map. ␥M23 is a member of a conserved triad consisting of ␥R242 and one of the three ␤E381. The illustrated ␤E381 is in the ␤ subunit conformer known as ␤ DP , which has ADP bound in the F 1 crystal (13). We note that all three residues are in highly conserved regions of the ␥ and ␤ subunits and are identical in all known ␥ subunit sequences; therefore, in the analysis of mutant E. coli complexes, using the positions of these residues as determined from crystals of the bovine enzyme is valid. In turn, we note that the results presented here are entirely consistent with the structural model of Abrahams et al. (13). We propose that when ␥M23 is changed to lysine, the ⑀-amino group forms an additional ionized hydrogen bond with ␤Glu-381 during a step of the catalytic cycle, hence the extra bond detected by Arrhenius analysis. Clearly, ␥M23K is an added function mutation. This conclusion is consistent with the similar uncoupling effect of replacing ␥Met-23 with arginine and the lack of effect when substituted by neutral or negatively charged amino acids (19).
This explanation is also consistent with the effects of second site mutations that suppress the effects of the ␥M23K mutation. The ␥Q269R suppressor mutation restored efficient coupling and negated the increase in transition state activation energy. As mentioned before, the amino acid replacement of ␥Gln-269 affected the interactions between ␥ and ␤ subunits, causing reduced coupling efficiency (Table II) and stability (20). We propose that the ␥Q269R mutation suppresses the effect of ␥M23K in part by reducing the energy of interaction between ␥ and ␤ subunits. It is likely that similar effects were observed by Jeanteur-De Beukelaer et al. (48) with ␤ subunit mutations that suppressed the effects of an altered ␥ subunit carboxyl terminus. Related to the effect on coupling is an effect on complex stability. Loss of ␦ and ⑀ subunits leaves a destabilized 2 R. K. Nakamoto, unpublished observation.   Fig. 1), and all of the known suppressor mutations of ␥M23K confer temperature stability (20,21). Clearly, protein-protein interactions between ␥ and ␤ subunits are critical for both complex stability and coupling.
In contrast, changing ␥R242C resulted in overall higher proton pumping rates and ATP synthesis because the enzyme turned over faster and not because coupling efficiency was restored. Interestingly, the structure suggests that ␥R242C should reduce the energy of interaction between ␥ and ␤ subunits because this amino acid change should remove the ionized hydrogen bond between ␥Arg-242 and ␤ DP Asp-381 (Fig.  7B). It is possible that the cysteine may reduce ␥-␤ interactions even more if the environment of ␥R242C induces a lower pK and causes the residue to ionize at neutral pH. The thiolate ion, a strong nucleophile, would form an ion pair with ␥M23K and in addition create a repulsive pair with ␤Glu-381. We are analyzing the properties of complexes with additional mutations of both residues to clarify their roles. Not surprisingly, certain amino acid replacements of ␤Glu-381 have a similar uncoupling phenotype to the ␥M23K mutation. 3 Without question, the ␥-␤ interactions involving ␥R242 and ␤E381 play an important role in turnover and coupling. The ␤ subunit residue is a part of the conserved sequence 380 DELSEED 386 (E. coli numbering). Residues in this sequence have also been implicated in interactions with ⑀ subunit residue ⑀S108 (45,49). It seems quite plausible that the perturbations on the conserved ␤DELSEED sequence induced by the ␥M23K mutation are communicated to ⑀S108 or nearby residues and perturbs the functional interaction of F 1 with F 0 as demonstrated by the fact that ⑀ inhibition of F 1 ATPase activity in not relieved by F 0 binding in the ␥M23K mutant (Fig. 3).
In perturbing the interactions between ␥ and the ␣ 3 ␤ 3 complex, the mutation alters thermodynamic parameters of the ATP hydrolysis reaction transition state in both F 0 F 1 and purified F 1 complexes (Table I and Fig. 5). Isokinetic plots of ⌬H ‡ versus T⌬S ‡ can be used to investigate perturbations of coupling through the effects of mutations on the structure of the catalytic transition state. Fig. 8 shows isokinetic plots for F 1 and F 0 F 1 preparations. It is seen that for various F 1 ␤ and F 1 FIG. 6. Effect of suppressor mutations on transition state thermodynamic parameters of ␥M23K F 0 F 1 . Difference activation energy parameters for membranous F 0 F 1 preparations (mutant enzyme minus wild-type enzyme values) were obtained at 30°C by Arrhenius analysis as detailed in the legend to Fig. 4. Thermodynamic values were calculated as described under "Materials and Methods." Thermodynamic values for F 0 F 1 preparations containing individual ␥ subunit mutations were compared with those for F 0 F 1 preparations containing double ␥ subunit mutations (the original ␥M23K mutation in conjunction with a "suppressor" mutation). Values for membranous F 0 F 1 preparations containing the ␥M23K mutation are represented by the hatched bars.

FIG. 7. Regions of interactions between ␥ and ␤ subunits.
A illustrates the regions of interaction between the ␥ subunit (in blue) with the ␤ subunit conformers known as ␤ DP (in yellow) and ␤ E (in green). Coordinates for this figure were obtained from x-ray crystallographic data for bovine F 1 (13). The braces illustrate the three ␥ subunit helical regions involved in energy coupling (21). The upper asterisk shows the contact between conserved residues ␥Gln-269 and ␤Thr-304 (E. coli numbering) near the "hydrophobic sleeve" (13). The lower asterisk shows the contact region between the conserved residue ␥Arg-242 and ␤Glu-381 of the conserved ␤ 380 DELSEED 386 sequence in ␤ DP . This region is shown in detail in B along with ␥Met-23 and Van der Waals' contacts.
␥ mutants, the ␥M23K enzyme preparations lie very close to bovine mitochondrial F 1 and F 0 F 1 . Al-Shawi et al. (42) suggested that the mitochondrial enzyme is a "better" catalyst than the E. coli enzyme because it binds the substrate in a manner that is closer to the true transition state of pentacoordinate ␥-phosphate and thereby reduces the transition state energy. In order to achieve a lower transition state, however, the enzyme must utilize more binding energy. In the case of the E. coli ␥M23K mutant, the enzyme must also utilize more binding energy to break an extra bond that is created by replacement of the conserved ␥M23 with a positively charged residue. Another interesting feature revealed by the isokinetic plots ( Fig. 8) is that the ␥M23K F 1 and ␥M23K F 0 F 1 points have very similar values, whereas the wild-type F 1 point has a more positive ⌬H ‡ and T⌬S ‡ than the wild-type F 0 F 1 membrane preparations. The origin of this phenomenon was seen in Fig. 3 in that as the wild-type F 1 binds to F 0 the ⑀ inhibition is relieved on binding F 0 . As pointed out above, the effect of the ␥M23K mutation perturbs transmission of conformational information, which modulates ⑀ interactions with F 0 such that inhibition is not relieved and coupling efficiency is lowered. This is the primary effect of ␥M23K on coupling.
The effect of ␥M23K as well as other mutations in the ␥-␤ interface appear to perturb a balance of interactions between the subunits necessary for transmission of coupling information and energy. In addition, the ␥M23K mutation appears to modulate the functional interaction of ⑀ subunit with F 0 . The effect of the structural perturbation on catalytic mechanism appears to be creation of a branched pathway that either bypasses or skips the coupling step. These results demonstrate the mechanistic linkage between catalysis and coupling that minimally involves ␤, ␥, and ⑀ subunits. Furthermore, we suggest that residues ␥Met-23, ␥Arg-242, ␤ 380 DELSEED 386, and ⑀ residues (near ⑀Ser-108) form a common energy coupling domain that transmits conformational energy from F 0 to F 1 and vice versa at discrete point (times) within the turnover of the enzyme. These suggestions are currently being investigated by further experiments. FIG. 8. Isokinetic plot of membranous F 0 F 1 and soluble F 1 preparations containing various ␥ subunit mutations. The enthalpic term, ⌬H ‡ of activation for k cat , is plotted against the entropic term, T⌬S ‡ of activation for k cat at 30°C. Filled symbols and a solid line give the results for various wild-type, ␤-mutant F 1 preparations, bovine mitochondrial F 1 (42) as well as ␥ subunit mutant F 1 preparations determined in this study. Open symbols and a dashed line illustrate the results from wild-type membranous F 0 F 1 , bovine mitochondrial F 0 F 1 , and various ␥ subunit mutant F 0 F 1 preparations. Lines were fitted by linear least squares regression analysis.