Modulation of Charge in the Phosphate Binding Site of Escherichia coli ATP Synthase*

This paper presents a study of the role of positive charge in the Pi binding site of Escherichia coli ATP synthase, the enzyme responsible for ATP-driven proton extrusion and ATP synthesis by oxidative phosphorylation. Arginine residues are known to occur with high propensity in Pi binding sites of proteins generally and in the Pi binding site of the βE catalytic site of ATP synthase specifically. Removal of natural βArg-246 (βR246A mutant) abrogates Pi binding; restoration of Pi binding was achieved by mutagenesis of either residue βAsn-243 or αPhe-291 to Arg. Both residues are located in the Pi binding site close to βArg-246 in x-ray structures. Insertion of one extra Arg at β-243 or α-291 in presence of βArg-246 retained Pi binding, but insertion of two extra Arg, at both positions simultaneously, abrogated it. Transition state stabilization was measured using phosphate analogs fluoroaluminate and fluoroscandium. Removal of βArg-246 in βR246A caused almost complete loss of transition state stabilization, but partial rescue was achieved in βN243R/βR246A and αF291R/βR246A. βArg-243 or αArg-291 in presence of βArg-246 was less effective; the combination of αF291R/βN243R with natural βArg-246 was just as detrimental as βR246A. The data demonstrate that electrostatic interaction is an important component of initial Pi binding in catalytic site βE and later at the transition state complex. However, since none of the mutants showed significant function in growth tests, ATP-driven proton pumping, or ATPase activity assays, it is apparent that specific stereochemical interactions of catalytic site Arg residues are paramount.

ATP synthase is the terminal enzyme of oxidative phosphorylation and photophosphorylation, which synthesizes ATP from ADP and phosphate (P i ). The energy for ATP synthesis comes from transmembrane movement of protons down an electrochemical gradient, generated by substrate oxidation or by light capture. Initially, as the protons move through the interface between a and c subunits in the membrane-bound F 0 -sector of the enzyme, the realized energy is transduced into mechanical rotation of a group of subunits (␥⑀c 10 -14 ), which comprise the "rotor". A helical coiled coil domain of ␥ projects into the central space of the ␣ 3 ␤ 3 hexagon, in the membraneextrinsic F 1 -sector. ␣ 3 ␤ 3 hexagon contains three catalytic sites at ␣/␤ interfaces. In a manner that is not yet understood, rotation of ␥ vis-à -vis the three ␣/␤ subunit pairs brings about ATP synthesis at the three catalytic sites using a sequential reaction scheme (1). "Stator" subunits b 2 and ␦ are present to prevent co-rotation of ␣ 3 ␤ 3 with the rotor. Detailed reviews of ATP synthase mechanism may be found in Refs. 2-5. Binding of P i is an important step of the ATP synthase mechanism that has been extensively studied by biochemical approaches and may be directly coupled to rotation of subunits (3, 6 -11). Recent studies of the rotational mechanism have begun to illuminate which steps in the enzymic pathway of ATP synthesis and hydrolysis are likely coupled to the two substeps (80 o and 40 o ) of subunit rotation and which steps occur in the intervening stationary dwells (12)(13)(14)(15). While it has not yet been possible to directly correlate the step of P i binding/ release with a specific mechanical event or an intervening dwell, it seems likely that this will soon be achieved. Thus we can foresee that it may be possible in the near future to correlate molecular features of P i binding, derived from mutational and biochemical studies, with mechanical function in this nanomotor system.
Studies of molecular aspects of P i binding in ATP synthase have been held back by lack of a suitable system to which both mutagenesis and a P i binding assay were applicable. Penefsky (16,17) reported that P i binding to mitochondrial ATP synthase F 1 could be assayed using the centrifuge column procedure with an estimated K d (P i ) of 30 M. However Al-Shawi and Senior (8) found that in Escherichia coli F 1 , no P i binding was detectable by this procedure. Further work by Weber and colleagues (18 -20) was carried out to determine whether P i binding could be assayed in E. coli F 1 using competition assays with MgAMPPNP or ATP, but these attempts also proved negative. 7-Chloro-4-nitrobenzo-2-oxa-1,3-diazole (NBD-Cl) 1 is a potent inhibitor of ATPase activity that covalently reacts at stoichiometry of 1 mol/mol ATP synthase, specifically with residue ␤Tyr-297, 2 situated at the end of the P i binding pocket (21)(22)(23). Following the terminology of Walker, Leslie, and colleagues (23), the three catalytic sites are conventionally referred to as ␤E, ␤DP, and ␤TP. NBD-Cl was found to react in the ␤E (empty) site. Perez et al. (24) reported that P i protects against NBD-Cl inhibition of ATPase activity of ATP synthase in mitochondrial membrane preparations, potentially providing a tool to assay P i binding in ␤E catalytic site. From their work, a K d (P i ) of 0.2 mM was calculated. In recent work we confirmed that this assay was applicable, both with membrane-bound enzyme and with purified F 1 from E. coli (11). Concentration dependence of P i protection against NBD-Cl inactivation in E. coli enzyme was similar to that found by Perez et al. (24) in mitochondrial enzyme. Studies of NBD-Cl inactivation kinetics and of MgADP protection characteristics confirmed that reaction occurred in the ␤E site in E. coli enzyme (11). Subsequently using mutagenesis we found this assay to be successful in assessing the functional roles of various catalytic site residues in P i binding (11,25,26). X-ray crystal structures of catalytic sites containing the P i analogs AlF 3 (27) and SO 4 2Ϫ (28) were valuable in suggesting residues within the P i binding pocket that were suitable targets for mutagenesis. Finally it may be noted that Penefsky (29) has recently confirmed, using purified [ 32 P]P i , that P i binding to E. coli F 1 is not detectable by the centrifuge column procedure but that a pressure ultrafiltration method did detect P i binding, with a K d (P i ) in the range of 0.1 mM, consistent with data obtained from the NBD-Cl inactivation assay. It is apparent that P i dissociates more rapidly from E. coli F 1 than it does from mitochondrial F 1 , unfortunately rendering the convenient centrifuge assay inapplicable with the E. coli enzyme.
In proteins, arginine residues show the highest propensity for occurrence and functional interaction at P i binding sites (30). Our earlier work established that natural Arg residues at positions ␣-376, ␤-182, and ␤-246 were important for P i binding in the ␤E catalytic site of ATP synthase, with the latter playing a key role (11,26). Mutagenesis of ␤Arg-246 to Ala, Gln, or Lys abolished P i binding (11). Residue ␤Asn-243, although totally conserved and located very close to bound P i , was found to be not directly involved in interacting with P i . Rather it was found to be necessary for correct organization of the transition state complex (25). However, if Asp was introduced at this position it prevented P i binding, presumably because it nullified the positive charge of the neighboring ␤Arg-246 (25). Therefore balance of charge in the P i binding pocket also appeared important.
After binding, P i must be condensed with MgADP via a chemical transition state, for which a molecular mechanism has been proposed in (3). The transition state analog MgADP-AlF 4 Ϫ trapped in catalytic sites has been visualized by x-ray crystallography (28), and it is clear that the fluoroaluminate group occupies the position of phosphate in the transition state complex. Contribution of different residues to stabilization of the transition state complex can be compared by assay of inhibition of ATPase activity by MgADP-fluoroaluminate (or MgADP-fluoroscandium) in mutant and wild-type enzymes (11,25). By comparing effects on P i binding and transition state stabilization one can further infer roles of each potential P i residue at early and later steps of the catalytic pathway.
In this paper we modulated charge within the P i binding site by introduction of extra Arg at residues ␤-243 and ␣-291, both in presence of the natural ␤Arg-246 and in its absence (␤R246A mutant). We also combined ␤Arg-243 and ␣Arg-291 with the natural ␤Arg-246 to test effects of excess positive charge. P i binding and transition state stabilization were assessed in each of the new mutants.

Preparation of E. coli Membranes; Measurement of Growth Yield in
Limiting Glucose Medium; Assay of ATPase Activity of Membranes; Measurement of Proton Pumping in Membrane Vesicles; SDS-gel Electrophoresis; Immnunoblotting-E. coli membranes were prepared as described previously (31). It should be noted that this procedure involves three washes of the initial membrane pellets, once in buffer containing 50 mM TES, pH 7.0, 15% glycerol, 40 mM 6-aminohexanoic acid, 5 mM p-aminobenzamidine, then twice in buffer containing 5 mM TES, pH 7.0, 15% glycerol, 40 mM 6-aminohexanoic acid, 5 mM paminobenzamidine, 0.5 mM DTT, 0.5 mM EDTA. Prior to the experiments, membranes were washed twice more by resuspension and ultracentrifugation in 50 mM Tris/SO 4 , pH 8.0, 2.5 mM MgSO 4 . Growth yield in limiting glucose was measured as described previously (32). ATPase activity was measured in 1 ml of assay buffer containing 10 mM NaATP, 4 mM MgCl 2 , 50 mM Tris/SO 4 , pH 8.5 at 37°C. Reactions were started by addition of membranes and stopped by addition of SDS to 3.3% final concentration. P i released was assayed as described previously (33). For wild-type membranes (5-10 g of protein), reaction times were 2-10 min. For mutant membranes (20 -100 g of protein), reaction times were 30 -120 min. All reactions were shown to be linear with time and protein concentration. ATP-driven proton pumping was measured by following the quench of acridine orange fluorescence as described previously (34). SDS-gel electrophoresis on 10% acrylamide gels was as described previously (35). Immunoblotting with rabbit polyclonal anti-F 1 -␣ and anti-F 1 -␤ antibodies was as described previously (36). Densitometry of immunoblots was performed using software from Scion Corp. (Scion Image Release Beta 4.02, www.scioncorp.com/).
Construction of Mutant Strains of E. coli-Mutagenesis was by the method of Vandeyar et al. (38). For ␤N243R/␤R246A and ␤N243R mutants, the template for oligonucleotide-directed mutagenesis was M13mp18 containing the HindIII-XbaI fragment from pSN6. pSN6 is a plasmid containing the ␤Y331W mutation from plasmid pSWM4 (18) introduced on a SacI-EagI fragment into pBWU13.4 (37), which expresses all the ATP synthase genes. Mutagenic oligonucleotides were as follows: ␤N243R/␤R246A, GCTGTTCGTTGACCGCATCTATGCATA-CACCCTGGCCG (where the underlined bases introduce the mutation and a new Nsi1 restriction site); ␤N243R, GTGTTCGTCGACCGCATC-TATCGTTAC (where the underlined bases introduce the mutation and a new SalI restriction site). DNA sequencing was performed to confirm the presence of mutations and absence of undesired changes in sequence, and the mutations were transferred to pSN6 on SacI-EagI fragments, generating the new plasmids pZA8 (␤N243R/␤R246A/ ␤Y331W) and pZA15 (␤N243R/␤Y331W). Each plasmid was transformed into strain DK8 (39) containing a deletion of ATP synthase genes for expression of the mutant enzymes. For ␣F291R, ␣F291R/ ␤R246A, and ␣F291R/␤N243R mutants, the template for oligonucleotide-directed mutagenesis was M13mp18 containing the SphI-SalI fragment from pSN6. The mutagenic oligonucleotide for ␣F291R was: CGGGCGACGTCCGCTACCTCCACTCTCG (where the underlined bases introduce the mutation and a new Aat2 restriction site). DNA sequencing was performed to confirm the presence of mutations and absence of undesired changes in sequence. The mutation was transferred to pSN6 on a XhoI-PmlI fragemt generating the new plasmid pZA10 (␣F291R/␤Y331W). For new plasmid pZA9 (␣F291R/␤R246A/ ␤Y331W) a XhoI-PmlI fragment was transferred to pZA7 (11). For new plasmid pZA16 (␣F291R/␤N243R/␤Y331W) a SacI-EagI fragment was transferred from plasmid pZA15 to plasmid pZA10. Each plasmid was transformed into strain DK8 (39) containing a deletion of ATP synthase genes for expression of the mutant enzymes. It may be noted that all of the new mutant strains contained the ␤Y331W mutation, which is valuable for measurement of nucleotide binding parameters (18) and does not affect function significantly. While it was not utilized in this work, the Trp mutation was included for possible future use.
Inhibition of ATPase Activity by NBD-Cl and Protection by MgADP or P i -NBD-Cl was prepared as a stock solution in dimethyl sulfoxide and protected from light. Membranes (0.2-2.0 mg/ml) were reacted with NBD-Cl for 60 min in the dark, at room temperature, in 50 mM Tris/ SO 4 , pH 8.0, 2.5 mM MgSO 4 , then 50-l aliquots were transferred to 1 Ϫ ϩ 56 a Wild type, pBWU13.4/DK8; null, pUC118/DK8. All mutants were expressed with the ␤Y331W mutation also present, which does not significantly affect growth. Data are means of four to six experiments each.
b Growth on succinate plates after 3 days estimated by eye. ϩϩϩϩ, heavy growth; Ϫ, no growth; ϩ, light growth. ml of ATPase assay buffer to determine ATPase activity. Where protection from NBD-Cl inhibition by ADP or P i was determined, membranes were preincubated 60 min with protecting agent at room temperature before addition of NBD-Cl. MgSO 4 was present, equimolar with ADP or P i . Control samples containing the ligand without added NBD-Cl were included. Neither P i (up to 50 mM) nor MgADP (up to 10 mM) had any inhibitory effect alone. Where reversal of NBD-Cl inhibition by DTT was measured, membranes were first reacted with NBD-Cl (150 M) for 1 h at room temperature, then DTT (final ϭ 4 mM) was added and incubation continued for 1 h at room temperature before ATPase assay. Control samples without NBD-Cl and/or DTT were incubated for the same times.
Inhibition of ATPase Activity by Fluoroaluminate or Fluoroscandium-Membranes were incubated for 60 min at room temperature in 50 mM Tris/SO 4 , 2.5 mM MgSO 4 , 1 mM NaADP, and 10 mM NaF at a protein concentration of 0.2-1.0 mg/ml in the presence of AlCl 3 or ScCl 3 added at varied concentration (see "Results"). 50-l aliquots were then added to 1 ml of ATPase assay buffer and activity measured as above. It was confirmed in control experiments that no inhibition was seen if MgSO 4 , NaADP, or NaF was omitted.

RESULTS
Growth Properties of New Mutants of E. coli ATP Synthase-A series of mutants was generated to modulate charge in the proximity of residue ␤Arg-246, which was shown earlier to be a key residue for binding of P i into the catalytic sites on the F 1 -sector of ATP synthase (11). Mutation of ␤Arg-246 to Ala abrogates P i binding (11). We introduced Arg at two residues located close to ␤Arg-246, namely ␤Asn-243 and ␣Phe-291, to generate the new mutants ␤N243R, ␤N243R/␤R246A, ␣F291R, and ␣F291R/␤R246A. These mutants are designed to test the  d The very low ATPase activity in the null mutants is attributable to the fact that all membrane preparations were washed several times before assays (see "Materials and Methods").
"Materials and Methods," were run on 10% SDS-polyacrylamide gels together with purified wild-type F 1 (0.1-0.4 g) as reference. Protein bands were transferred to nitrocellulose and immunoblotted using anti-F 1 -␣ antibody (36). Densitometry was performed as described under "Materials and Methods." A, immunoblot. B, densitometric scans. The Growth yields on limiting glucose medium and growth on succinate plates are shown in Table I. It was evident that introduction of a new Arg residue at ␣-291 or ␤-243 was debilitating either in combination with or in absence of the ␤246A mutation, although it may be noted that the ␤R246A mutation alone consistently displayed even lower growth. Similar results were seen in ␣F291R/␤N243R. Therefore oxidative phosphorylation is defective in each of the mutants containing Arg at ␣-291 or ␤-243 or both.
SDS-gel Electrophoresis and Immunoblotting of Membrane Preparations-Previous work (11) had established that P i binding by mutant and wild-type ATP synthase can be assayed using either membrane preparations or purified F 1 . For a series of mutants, as studied here, it was more efficient to use membrane preparations. However, the possibility existed that the mutations may have compromised assembly and/or oligomeric stability, leading to membrane preparations with low ATP synthase content. This could account for the low growth yields in Table I. We therefore performed SDS-gel electrophoresis and immunoblotting experiments.
Coomassie Blue-stained SDS-gels of mutant and wild-type membranes (with purified wild-type F 1 as reference) established that all the mutant membrane preparations had bands running at the position of F 1 -␣ and F 1 -␤ subunits, with similar intensities to the ␣ and ␤ bands seen in wild-type membranes (data not shown). Immunoblotting and densitometry was performed with anti-␣ subunit and anti-␤ subunit antibodies (36). Preliminary experiments using purified wild-type F 1 revealed that the response was linear in the range 0.1-0.4 g of protein, and further tests showed that 4 g of wild-type or mutant membrane preparations gave a response that fell within this range. An immunoblot using anti-F 1 -␣ subunit is shown in

. Inhibition of ATPase activity in membrane-bound wild-type and mutant ATP synthase by NBD-Cl.
Membranes were preincubated for 60 min at 23°C with the indicated concentration of NBD-Cl, then aliquots were added to 1 ml of assay buffer and ATPase activity determined. For details see "Materials and Methods." E, wild type; •, mutant, as indicated in the separate panels. Each data point represents average of at least four experiments, using two independent membrane preparations of each mutant. In each case the "100%" point is the uninhibited rate of ATPase as shown in Table II. presented relative to wild type. Three different experiments gave similar results. It is evident that the mutant membranes were similar in ATP synthase content to wild type. Immunoblotting using anti-F 1 -␤ antibody (data not shown) confirmed this conclusion. Table II shows the ATPase and proton pumping activities of the mutant ATP synthase enzymes in membranes compared with wild type and with two different null controls. It may be noted that the membrane preparations were washed extensively before assay. Data from the null controls showed that this removed virtually all contaminating ATPase activity. The following conclusions are evident. First, insertion of one or two new Arg residues close to the P i binding site (␣F291R, ␤N243R, ␣F291R/␤N243R) in otherwise wild-type background (i.e. with ␤Arg-246) reduced membrane ATPase activity to a very low level. ATPase activities were far too low to support ATP-driven proton pumping. Second, insertion of ␤Arg-243 in presence of ␤Ala-246 (␤N243R/␤R246A) did not restore ATPase activity. Third, insertion of ␣Arg-291 in presence of ␤Ala-246 (␣F291R/␤R246A) did significantly restore ATPase activity (by 10-fold over ␤R246A alone), and in this case there was detectable, although low, ATP-driven proton pumping. It is apparent that the effects seen on ATPase and proton pumping are consistent with growth characteristics described in Table I; in the case of ␣F291R/␤R246A the partial "rescue" of ␤R246A was apparently not substantial enough to translate into significant growth.

ATPase Activity and Proton Pumping Activities of Mutant ATP Synthase Enzymes in Membranes-
Inhibition of ATPase Activity of ATP Synthase in Membranes by NBD-Cl- Fig. 2 shows NBD-Cl inhibition of each of the new mutant ATP synthase enzymes generated in this work, together with wild type and ␤R246A mutant for comparison. In each panel the mutant enzyme is represented by filled circles and wild type by open circles. Please note that the 100% value in each case is the uninhibited ATPase rate as shown in Table  II; this rate varied widely in wild type versus the different mutants. However percent inhibition is plotted to allow easy comparison of the degree of inhibition by NBD-Cl. Wild type was almost completely inhibited by NBD-Cl at higher concentrations. The data show that each mutant enzyme was inhibited by NBD-Cl but to a lesser final extent than in wild type and with different concentration dependence. In previous work (11,26) we have noted several instances where mutant ATP synthases were incompletely-inhibited by NBD-Cl. To test whether the residual activity was a real activity of NBD-Clinhibited enzyme, for each mutant in Fig. 1 we first incubated for 1 h with 150 M NBD-Cl, then added a further pulse of NBD-Cl, equivalent to additional 200 M NBD-Cl, and incubated for a further hour before assaying ATPase activity. In each case additional inhibition of ATPase was small or zero, consistent with Fig. 1 data. Two lines of evidence further supported the idea that ATPase activity in mutant membranes is due to ATP synthase. First, in each case inhibition by NBD-Cl was completely reversed, up to starting activity, by incubation of inhibited enzyme with 4 mM DTT for 1 h at room temperature. This also occurred in wild type and is known to be due to release of the NBD-adduct from ␤Tyr-297, the reactive residue (21,22). Second, reaction with NBD-Cl was prevented by presence of MgADP in the reaction incubation, and in each case dependence on MgADP concentration was the same as in wild type (EC 50 ϭ 4.5 mM). This protection is referable to loose

FIG. 3. Protection by P i of ATPase activity in wild-type and mutant membranes from inactivation by NBD-Cl.
Membranes were preincubated with P i at zero, 2.5, or 10 mM concentration as shown, for 60 min at 23°C. Then NBD-Cl (125 M) was added and aliquots withdrawn for assay at time intervals as shown. ATPase activity remaining is plotted against time of incubation with NBD-Cl. E, no P i added; Ⅺ, 2.5 mM P i ; ‚, 10 mM P i . Each data point represents the average of four different experiments using two independent membrane preparations of each mutant.
MgADP binding in catalytic site ␤E where NBD-Cl reacts (11,23). Fig. 3 demonstrates that P i protected well against NBD-Cl inhibition of ATPase activity in wild type but not in ␤R246A mutant, confirming previous work (11). It is seen that mutants ␤N243R/ ␤R246A and ␣F291R/␤R246A both showed clear protection by P i . It is apparent that insertion of an Arg at position ␤-243 or ␣-291 compensates for the loss of the natural Arg at ␤-246 in P i binding. P i binding was retained in ␣F291R and ␤N243R. Therefore introduction of one extra Arg did not interrupt P i binding. However, introduction of two extra Arg (␣F291R/ ␤N243R mutant) prevented P i binding. Fig. 4 shows inhibition of ATPase activity by fluoroaluminate in each of the mutants (closed circles) as compared with wild type (open circles). The top left panel shows results obtained for ␤R246A and wild-type membranes, and it may be noted that the data are very similar to the parallel data reported previously (11) where purified F 1 was used. Wild type was very strongly inhibited (Ͼ95%), and ␤R246A was inhibited by 15% at the highest AlCl 3 concentration. Inclusion of "re-placement" Arg in ␤N243R/␤R246A or ␣F291R/␤R246A mutants increased inhibition markedly, to maximally 42 and 62%, respectively. Inclusion of one additional Arg (␤N243R, ␣F291R) gave maximal inhibition of 34 and 45%, respectively, i.e. less than wild type by far, and less than when in combination with ␤Ala-246 but higher than ␤Ala-246 alone. Inclusion of two additional Arg (␣F291R/␤N243R) gave 17% inhibition, similar to ␤R246A. An exactly similar pattern was seen when fluoroscandium was the inhibitor (Fig. 5). The maximal inhibition reached with fluoroscandium was: Ͼ98% for wild type, 4% for ␤R246A, 28% for ␤N243R/␤R246A, 62% for ␣F291R/␤R246A, 24% for ␤N243R alone, 49% for ␣F291R alone, and zero for ␣F291R/␤N243R. DISCUSSION P i binding is a primary step in ATP synthesis by ATP synthase, so that understanding the molecular basis of P i binding is an important goal. Earlier work using the NBD-Cl inactivation assay described in the Introduction has shown that positively charged residues are functionally important for P i binding in the ␤E catalytic site of E. coli ATP synthase (11,26). X-ray crystallography structures of ATP synthase catalytic sites containing ADP with bound AlF 3 (27) or SO 4 2Ϫ (28) as phosphate analogs are consistent with these conclusions. To- gether these studies are supportive of the molecular mechanism for ATP synthesis proposed in (3). Other work (25) indicated that introduction of negative charge in the P i binding pocket, close to ␤Arg-246, prevented P i binding. This suggested that modulation of charge in the P i binding site could be used to illuminate the molecular mechanism of P i binding. It is established that Arg residues occur particularly commonly in P i binding sites in proteins (30), therefore varying the number Ϫ -inhibited enzyme (28). E. coli residue numbering is shown, with corresponding bovine mitochondrial residue numbers in parentheses.

Inhibition of ATPase Activity by Fluoroaluminate and Fluoroscandium in Membranes Containing Mutant ATP Synthase-
of Arg residues in the P i binding site of ATP synthase seemed a useful approach.
Residue ␤Asn-243 lies 3.2 Å from ␤Arg-246 in both AlF 3 and SO 4 2Ϫ -containing catalytic sites (nearest atom distances quoted) and close to either AlF 3 or SO 4 2Ϫ (see Fig. 6). Thus one experimental approach was to introduce the mutation ␤N243R in wild-type background (with ␤Arg-246) and in presence of the ␤R246A mutation. Residue ␣Phe-291, located at the end of the P i binding pocket across the catalytic ␣/␤ interface with its side chain pointing toward the bound P i analogs, also appeared to be a suitable location at which to introduce a new Arg. It lies at a distance from ␤Arg-246 of 3.2 Å in the AlF 3 -containing catalytic site and 7.5 Å in the SO 4 2Ϫ -containing catalytic site (27,28). 3 We introduced the ␣F291R mutation in wild-type background and in the presence of the ␤R246A mutation. Table III shows the actual distances of residues ␤Arg-246, ␤Asn-243, and ␣Phe-291 from bound AlF 3 and SO 4 2Ϫ as determined by x-ray crystallography (27,28), together with speculative distances (in parentheses) calculated for mutant residues ␤Ala-246, ␤Arg-243 and ␣Arg-291 using the "Deep View Swiss-Pdb Viewer" (described in Table III). It is apparent that the mutations would place extra positive charge relatively close to P i and that the ␤Ala-246 mutation leaves a relatively large "hole" into which a new Arg might fit. No other suitable location at which to introduce new Arg close to bound P i was apparent.
SDS-gel electrophoresis and immunoblotting ( Fig. 1 and "Results") showed that the mutant ATP synthase enzymes were present in membrane preparations in amounts that did not deviate strongly from wild type. Growth, ATPase, and ATPdriven proton pumping activities were impaired in all the mutants as compared with wild type (Tables I and II). Introduction of one or two extra positively charged Arg residues in the wild-type background at either ␤-243 or ␣-291, or both, was therefore detrimental. Introduction of new Arg at ␤-243 or ␣-291 in the ␤R246A background did not restore function to normal, although a significant compensatory effect on ATPase and ATP-driven proton pumping was seen in the latter case (␣F291R/␤R246A, Table II).
The ␤R246A mutant did not show P i binding but both ␤N243R and ␣F291R mutations "rescued" P i binding in combination with ␤Ala-246 (Fig. 3). Since neither ␤Arg-243 nor ␣Phe-291 could be expected to assume the same exact stereochemical interactions that ␤Arg-246 achieves, electrostatic interaction per se is therefore important, and we conclude that the presence of at least one positive charge at this general location is a requisite determinant of initial P i binding in catalytic site ␤E. ␤N243R or ␣F291R in wild-type background (representing one extra positive charge) did not prevent P i binding (Fig. 3), but the combination of ␣F291R/␤N243R (two extra charges) abrogated P i binding. Presumably the local concentration of charge in the latter becomes too disruptive and distorts the P i binding site.
A similar pattern of effects was seen when transition state stabilization was assessed by assaying inhibition of ATPase activity by the transition state analogs MgADP-fluoroaluminate and MgADP-fluoroscandium. It was shown previously in Ref. 11 that both inhibitors are potent against wild-type ATP synthase but inhibit ␤R246A mutant only to small extent, indicating that ␤Arg-246 is intimately involved in transition state stabilization. It was found here (Figs. 4 and 5) that mutant residues ␤Arg-243 or ␣Arg-291 partly rescued transition state stabilization when present with ␤Ala-246. Raising the number of positively charged residues to two (␤N243R and ␣F291R mutants in wild-type background) had an adverse effect as reflected by lesser inhibition of ATPase; and raising the number of local positive charges to three reduced transition state stabilization right back to where it was in ␤R246A. Even in the best cases among the mutants (␤N243R/␤R246A and ␣F291R/␤R246A) transition state stabilization was incomplete as compared with wild type, providing one explanation for the functional impairment seen in all the mutants.
In summary our results show that in the catalytic site ␤E of ATP synthase, P i binding is notably affected by local positive charge. Positive charge in the vicinity of the natural ␤Arg-246 is important; its removal in ␤R246A mutant can be compensated for partially by introduction of one Arg at either ␤-243 or ␣-291. Thus, electrostatic interaction is an important determinant of P i binding. The presence of two Arg by introduction of either ␤Arg-243 or ␣Arg-291 in presence of ␤Arg-246 does not prevent P i binding, but the presence of all three Arg abrogates P i binding. Effects on transition state stabilization followed a parallel pattern. However, restoration of P i binding in ␤E catalytic sites by charge compensation is not sufficient by itself to restore full function.

TABLE III
Distances within the P i binding subdomain of ATP synthase catalytic sites Distances shown are between the nearest atom of the residue and the P i analog, in Ångstroms (Å), as determined by x-ray crystallography (27,28). Values in parentheses are speculative distances calculated for mutant residues (also in parentheses) calculated using the Deep View Swiss-Pdb Viewer, Version 3.7 (N. Guex, A. Diemand, M.C. Peitsch, and T. Schwede (2004)) at www.us.expasy.org/spdbv/mainpage.html).