alpha-Aspartate 261 is a key residue in noncatalytic sites of Escherichia coli F1-ATPase.

X-ray structure analysis of the noncatalytic sites of F1-ATPase revealed that residue α-Asp261 lies close to the Mg of bound Mg-5′-adenylyl-β,γ-imidodiphosphate. Here, the mutation αD261N was generated in Escherichia coli and combined with the αR365W mutation, allowing nucleotide binding at F1 noncatalytic sites to be specifically monitored by tryptophan fluorescence spectroscopy. Purified αD261N/αR365W F1-ATPase showed catalytic activity similar to wild-type. An important feature was that, without any resort to nucleotide-depletion procedures, the noncatalytic sites in purified native enzyme were already empty. Binding studies with MgATP, MgADP, and the corresponding free nucleotides led to the following conclusions. Residue α-Asp261 interacts with the Mg of Mg-nucleotide in noncatalytic sites and provides a large component of the binding energy (∼3 kcal/mol). It is the primary determinant of the preference of noncatalytic sites for Mg-nucleotide. The natural ligands at these sites in wild-type enzyme are the Mg-nucleotides and free nucleotides bind poorly. Under conditions where noncatalytic sites were empty, αD261N/αR365W F1 showed significant hydrolysis of MgATP. This establishes unequivocally that occupancy of noncatalytic sites by nucleotide is not required for catalysis.

F 1 F 0 -ATP synthase is the enzyme responsible for ATP synthesis by oxidative phosphorylation and ATP-driven proton extrusion in Escherichia coli. The catalytic F 1 -sector is an oligomer with subunit composition ␣ 3 ␤ 3 ␥␦⑀, which contains six nucleotide-binding sites (Wise et al., 1983). Three sites are catalytic sites, whose structure is provided primarily from the ␤-subunits as shown by the x-ray structure resolution of bovine F 1 (Abrahams et al., 1994). Simultaneous occupancy of all three catalytic sites is necessary for physiological rates of ATP hydrolysis (Weber et al., 1993a). The other three sites are noncatalytic sites, formed primarily from ␣-subunits (Abrahams et al., 1994). Although the ligand-binding properties of these sites have been established in some detail (e.g. see Perlin et al. (1984), Milgrom and Cross (1993), Jault and Allison (1993), Weber et al. (1994), and Weber and Senior (1995)), their physiological role has proved difficult to ascertain.
One obstacle in studying the noncatalytic sites has been of a technical nature. Although it was possible to prepare F 1 -ATPase which was completely depleted of nucleotides, and therefore had empty noncatalytic sites (Garrett and Penefsky, 1975;Wise et al., 1983;Senior et al., 1992), there was until recently no method of specifically monitoring noncatalytic site occupancy. The use of "chases" with GTP or PPi, or with adenine nucleotides for various times, in order to replace catalytic site nucleotides and thus allow estimation of noncatalytic site nucleotides (presumed to be nonexchangeable) is complicated by the findings that both GTP and PPi bind to noncatalytic sites (Weber et al., 1994;Hyndman et al., 1994;Weber and Senior, 1995) and by the fact that catalytic site-bound nucleotide does not rapidly exchange with medium nucleotide under all conditions. However, this problem was recently solved by use of fluorescence responses from a tryptophan residue specifically placed at residue ␣-365 by ␣R365W mutagenesis. 1 Residue ␣-Arg 365 interacts directly with the adenine ring of bound nucleotide in noncatalytic sites (Weber et al., 1993b;Abrahams et al., 1994) 2 and the fluorescence of residue ␣-Trp 365 is fully quenched on binding of nucleotide, providing a direct and specific probe of noncatalytic site occupancy (Weber et al., 1994).
Here we have studied mutations at two positions within the F 1 noncatalytic sites which are expected to disrupt nucleotide binding, namely ␣-Lys 175 and ␣-Asp 261 . The former lies in the Walker Homology A (P-loop) sequence and the latter in the Walker Homology B sequence (Walker et al., 1982). Residue ␣-Lys 175 is seen in the x-ray structure of noncatalytic sites to lie close to the terminal phosphate of bound Mg-AMPPNP 3 (Abrahams et al., 1994). Previous work (Rao et al., 1988;Yohda et al., 1988) showed that the mutations ␣K175I and ␣K175E abolished or greatly reduced nucleotide binding to the ␣-subunit. Residue ␣-Asp 261 lies close to the Mg moiety of the Mgnucleotide (Abrahams et al., 1994), and since nucleotide binding in noncatalytic sites is Mg-dependent (Weber et al., 1994) it seemed likely that removal of the carboxyl group by mutagenesis might have large effects. Yohda et al. (1988) have shown that the ␣D261N mutation in Bacillus PS3 reduced MgADP binding affinity in isolated ␣-subunit.
In each case, the "disrupting" mutation was combined with the ␣R365W mutation in order to allow direct analysis of nucleotide binding by fluorescence measurement. In this paper, we describe the properties of the ␣K175E/␣R365W, ␣K175I/ ␣R365W, and ␣D261N/␣R365W mutants.

MATERIALS AND METHODS
Construction of ␣K175I/␣R365W, ␣K175E/␣R365W, and ␣D261N/ ␣R365W Mutant Strains-The ␣R365W mutation from plasmid pAW7 (Weber et al., 1994) was moved into plasmids pRR1 and pRR2 (Rao et al., 1988), which contain the ␣K175I and ␣K175E mutations, respectively, on a 1.7-kilobase SacII-SacII fragment, screening for the correct * This work was supported by National Institutes of Health Grant GM25349 (to A. E. S.). 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. 1 The E. coli residue numbers are used throughout. Residues in bovine enzyme corresponding to E. coli residues ␣-Lys 175 , ␣-Asp 261 , and ␣-Arg 365 , are ␣-Lys 175 , ␣-Asp 269 , and ␣-Arg 362 .
2 The x-ray structure of bovine mitochondrial F 1 (Abrahams et al., 1994) shows that the hydrophobic aliphatic side chain of the arginine arches up to approach the adenine ring, with the guanidinium group pointing away from the adenine (J. E. Walker, A. G. W. Leslie, and J. P. Abrahams, personal communication). 3 The abbreviation is: AMPPNP, 5Ј-adenylyl-␤,␥-imidodiphosphate.
orientation by PstI digestion, and confirming the presence of the mutations by DNA sequencing. The resultant plasmids pSWM17 (␣K175E/ ␣R365W) and pSWM18 (␣K175I/␣R365W) were transformed into strain JP2, which carries a deletion of the ␣-subunit gene (Rao et al., 1988) to give strains SWM17 and SWM18. Oligonucleotide-directed mutagenesis (Vandeyar et al., 1988) was used to generate the ␣D261N mutation in an M13 mp18 template which already contained the ␣R365W mutation (ClaI-EcoRI fragment from plasmid pAW7 (Weber et al., 1994) ligated into M13 mp18 between AccI and EcoRI sites). The mutagenic oligonucleotide used was TG ATC ATT TAC AAT GAC CTT TCG AAA CAG GC, where the underlined base A introduces the mutation ␣D261N (GAT 3 AAT) and the underlined bases T and G introduced a new BstBI site. Introduction of the desired mutation was screened by digesting RF phage with BstBI and MscI (which identifies the ␣R365W mutation). The mutations were moved into plasmid pDP34N (Weber et al., 1993a) on a 1.7-kilobase SacII-SacII fragment, screening for correct orientation with PstI. The final plasmid was named pCB6 and the whole DNA sequence between the SacII restriction sites in this plasmid was determined and shown to be wild-type except for the ␣D261N/␣R365W mutations and the silent mutations associated with the introduced BstBI and MscI restriction sites. Plasmid pCB6 was transformed into strain JP2 and the final strain was designated CB6.
Growth of E. coli Cells; Preparation of Membrane Vesicles; Assay of ATPase Activity; Purification of F 1 ; Preparation of Nucleotide-depleted F 1 ; Fluorescence Measurements-These were all as described in Weber et al. (1994).

RESULTS
Effects of the Mutations ␣K175E, ␣K175I, and ␣D261N in Whole Cells and in Membrane Vesicles-The mutations ␣K175E, ␣K175I, and ␣D261N were each combined with the mutation ␣R365W on plasmids as described under "Materials and Methods." Then the mutations were expressed by transforming the plasmids into haploid strain JP2, which contains a deletion of the chromosomal ␣-subunit gene (Rao et al., 1988). The growth characteristics of each of the three double mutants were studied alongside those of strain AW7 (␣R365W single mutation in JP2), an isogenic wild-type strain (pDP34N/JP2), and an Unc Ϫ strain (pUC118/JP2). Table I shows the growth characteristics on succinate plates and the growth yields in limiting (3 mM) glucose liquid medium. Membrane vesicles were also prepared from each strain and assayed for ATPase activity (Table I).
Consistent with previous results, the ␣R365W single mutant grew well on succinate plates and in limiting glucose medium, and had significant membrane ATPase activity. As we described previously, F 1 can be readily purified from this strain and it is similar to wild-type F 1 in properties (Weber et al., 1994).
The ␣K175E/␣R365W mutant did not grow on succinate plates and had a growth yield in limiting glucose the same as the Unc Ϫ control. Its membrane ATPase activity was not significant. We were unable to prepare F 1 from this mutant because in the final step of purification (S-300 gel filtration) there was no protein peak at the position corresponding to F 1 . With the ␣K175I/␣R365W mutant, growth on succinate plates and in growth yield tests was strong, however, the membrane-ATPase activity was low (Table I). We were not able to purify F 1 from this mutant either, because again in the S-300 gel filtration step, no protein peak corresponding to F 1 was present. Our earlier studies of the ␣K175E and ␣K175I single mutants (Rao et al., 1988) showed that the former mutation partially impaired assembly of F 1 F 0 in the cells, whereas the latter mutation allowed assembly but impaired subunit-subunit interaction, such that both membrane-bound and purified F 1 tended to dissociate. We were previously able to purify F 1 from the ␣K175E and ␣K175I single mutants, but only in low yield. The present results show that combination of the ␣K175E and ␣K175I mutations with ␣R365W exacerbated their effects. Jounouchi et al. (1993) reported that five other mutations at residue ␣-175 had similar effects, in that they all impaired assembly and/or subunit stability of F 1 . Yohda et al. (1988) found that the ␣K175I mutation in Bacillus PS3 F 1 impaired subunit interactions. Therefore, it is clear that mutation of residue ␣-Lys 175 interrupts correct assembly of the enzyme and oligomeric stability of F 1 .
The ␣D261N/␣R365W mutant grew well on succinate plates, had almost normal growth yield, and had 25% of the membrane ATPase activity of the isogenic wild-type strain run alongside ( Table I). The membrane ATPase activity was 76% inhibited by dicyclohexylcarbodiimide (150 M at pH 8.0, 30°C, for 60 min), which is similar to wild-type. The membrane vesicles showed normal ATP-dependent pH gradient formation when assayed using acridine orange fluorescence quenching as described by Perlin et al. (1983) (data not shown), consistent with the fact that the specific ATPase activity of the membranes from the ␣D261N/␣R365W mutant (ϳ0.8 mol of ATP hydrolyzed per min/mg of protein) was similar to that of a haploid wild-type strain (Cox et al., 1978).
Properties of Purified F 1 from ␣D261N/␣R365W Mutant-Purified ␣D261N/␣R365W mutant F 1 was obtained in low yield (0.025 mg/g wet weight cells), but showed normal chromatographic profile on S-300 gel filtration, and normal subunit composition in SDS gels. The V max of ATPase activity was 20 units/mg, with K m (MgATP) ϭ 90 M, and k cat /K m ϭ 1.41 ϫ 10 6 M Ϫ1 s Ϫ1 , i.e. similar to wild-type and ␣R365W F 1 . The pH dependence of ATPase activity between pH 6.0 and 8.5 was assayed and was similar to that of wild-type and ␣R365W F 1 , indicating that the oligomeric stability of the enzyme was not impaired by the ␣D261N mutation in this pH range. Above pH 8.5 the mutant enzyme did appear to show loss of activity, possibly due to subunit dissociation. All the assays reported below were done at pH Յ 8.5. Yohda et al. (1988) found previously that the ␣D261N mutation (in Bacillus PS3) did not significantly affect subunit assembly when the mutant ␣-subunit was reconstituted into an ␣␤␥ subcomplex. Compared to the wild-type subcomplex, the mutation caused nearly 5-fold reduction in V max of ATPase activity of the mutant ␣␤␥ subcomplex, and abolished apparent negative cooperativity that was evident in the wild-type in Lineweaver-Burk plots. The effect on V max was larger than was seen here in intact F 1 , but comparisons are difficult because both the species and enzyme forms are different. In our ATPase assays, with the intact F 1 enzymes from wild-type, ␣R365W, or ␣D261N/␣R365W, plots of V versus [S] showed no deviation from simple monophasic Michaelis-Menten kinetics between 5 and 100% of V max . We did not scrutinize the kinetic behavior at very low substrate concentrations.
The fluorescence properties of the purified mutant F 1 were interesting. It may be recalled that in the ␣R365W single mutant F 1 as purified ("native F 1 ") the noncatalytic sites are essentially filled with endogenous adenine nucleotide and the tryptophan fluorescence spectrum ( exc ϭ 295 nm) is the same as for wild-type F 1 , because fluorescence of the ␣-Trp 365 residues is fully quenched (Weber et al., 1994). When ␣R365W F 1 is depleted of nucleotide by gel filtration in 50% (v/v) glycerolcontaining buffer, the marked fluorescence signal of the ␣-Trp 365 residues is revealed. In contrast, in the case of ␣D261N/␣R365W purified F 1 , the fluorescence of the ␣-Trp 365 residues was already fully apparent even in the native F 1 , without requiring nucleotide depletion (Fig. 1). This showed that the ␣D261N mutation had a strong effect, and that in native F 1 from the ␣D261N/␣R365W mutant, the noncatalytic sites were empty. Addition of high concentrations of nucleotide quenched the fluorescence of the ␣D261N/␣R365W mutant F 1 almost down to the level of the wild-type enzyme, and it is reasonable to ascribe the residual difference between the signal of quenched ␣D261N/␣R365W mutant F 1 and wild-type in Fig.  1 to the presence of a small amount of contaminating protein in the mutant enzyme. Binding of Free ATP or MgATP to Noncatalytic Sites of Purified ␣D261N/␣R365W F 1 -The fluorescence signal of the ␣-Trp 365 residues was used to characterize binding of nucleotides to noncatalytic sites in the mutant enzyme. Fig. 2, A and B, show titration of native ␣D261N/␣R365W F 1 with MgATP and free ATP, together with titration of nucleotide-depleted ␣R365W F 1 for comparison. The titration with free ATP was done as described earlier (Weber et al., 1994) in buffer containing EDTA. For the MgATP titration we had to modify the conditions used earlier (Mg/ATP ratio of 4/10 and a maximal MgSO 4 concentration of 2.5 mM) because the binding affinity for MgATP was very low. Here we used 25 mM MgSO 4 in the buffer and titrated with ATP up to 25 mM. Binding parameters were calculated by fitting theoretical curves to the data and calculated K d values are tabulated in Table II.
MgATP was seen to bind with the same affinity to all three noncatalytic sites of ␣R365W F 1 ( Fig. 2A; Table II, first line), similar to previous results (Weber et al., 1994). In ␣D261N/ ␣R365W F 1 , MgATP binding affinity was drastically reduced ( Fig. 2A, Table II, first line). In all likelihood the affinity was reduced by the same factor at all three sites, although as seen in Fig. 2A it was only possible to fill ϳ2 sites under the experimental conditions. Free ATP was seen to bind with the same affinity to all three sites in ␣R365W F 1 (Fig. 2B, open circles); however, for ␣D261N/␣R365W F 1 a better fit was obtained assuming one site of higher and two sites of lower affinity (Fig.  2B, closed circles). Nevertheless, as is clear from inspection of Fig. 2B the two enzymes were not actually very different in overall behavior toward free ATP. The calculated K d (free ATP) values are given in Table II, second line. It is obvious from Fig.  2 that free ATP bound with much weaker affinity than MgATP in ␣R365W F 1 , but in contrast, free ATP bound with similar or slightly higher affinity than MgATP in ␣D261N/␣R365W F 1 .
Binding of Free ADP or MgADP to Noncatalytic Sites of Purified ␣D261N/␣R365W F 1 -MgADP and free ADP binding were also studied in native ␣D261N/␣R365W and nucleotidedepleted ␣R365W mutant F 1 preparations following the same procedures as for free ATP and MgATP above. The binding curves are shown in Fig. 3, A and B. In each case all three noncatalytic sites behaved identically, and the calculated binding parameters are in given in Table II, third and fourth lines. It is seen that MgADP bound much more weakly to ␣D261N/ ␣R365W F 1 than to ␣R365W F 1 (Fig. 3A), and that in ␣R365W F 1 , free ADP bound more weakly that MgADP (Table II, third and fourth lines). However, free ADP binding to the ␣D261N/ ␣R365W F 1 showed, somewhat unexpectedly, a relatively high affinity (Table II,  Fluorescence spectra of mutant F 1 preparations. F 1 was passed once through a 1-ml centrifuge column containing 50 mM Tris SO 4 , pH 8.0, at 23°C before use. The fluorescence spectra are uncorrected and were taken at protein concentration of 100 nM in 50 mM Tris-SO 4 , 0.5 mM EDTA, pH 8.0, at 23°C. Excitation was at 295 nm. Curve 1, ␣D261N/␣R365W F 1 in absence of nucleotide; curve 2, ␣D261N/␣R365W F 1 plus 10 mM ADP; curve 3, wild-type F 1 Ϯ ADP. are noted under "Discussion." MgATPase Activity of ␣D261N/␣R365W F 1 -As was noted above, under standard ATPase assay conditions, the K m (MgATP) for ␣D261N/␣R365W F 1 was 90 M. Fig. 2A showed that at 90 M concentration of MgATP, the noncatalytic sites of ␣D261N/␣R365W F 1 were empty, as judged by the fluorescence signal. However, the fluorescence experiments were carried out in buffer containing 25 mM MgSO 4 , at room temperature, whereas the standard ATPase assays are carried out at 30°C, and with an ATP/Mg 2ϩ concentration ratio of 2.5/1, which was shown previously to optimize ATPase activity and to minimize inhibition by free Mg 2ϩ (Wise et al., 1983;Al-Shawi et al., 1988). In order to test whether the ␣D261N/ ␣R365W enzyme could hydrolyze MgATP when the noncatalytic sites were empty, we therefore assayed ATPase activity under the exact buffer conditions of the fluorescence titration of Fig. 2A, using a concentration of 100 M MgATP. Wild-type and ␣R365W F 1 enzymes, which contained filled noncatalytic sites, were assayed alongside. The ATPase activities seen were: wildtype, 0.45 unit/mg; ␣R365W F 1 , 0.31 unit/mg; ␣D261N/ ␣R365W F 1 , 1.06 unit/mg. These values represent 1.8, 2.1, and 5.3%, respectively, of the corresponding specific ATPase activities under standard assay conditions. Inhibition by the 25 mM MgSO 4 present in the buffer, and the fact that the assays were conducted at room temperature, are responsible for reducing the activities, nevertheless, the results unequivocally establish that MgATP hydrolysis occurs in the ␣D261N/␣R365W enzyme even when the noncatalytic sites are empty. Interestingly, the ␣D261N/␣R365W F 1 seemed less susceptible to inhibition by Mg 2ϩ than the other two enzymes. DISCUSSION We studied nucleotide binding to the noncatalytic sites of E. coli F 1 -ATPase using the fluorescence signal of the genetically engineered residue ␣-Trp 365 as a direct probe of noncatalytic site nucleotide occupancy (Weber et al., 1994). Residues ␣-Lys 175 and ␣-Asp 261 are both known to lie close to bound nucleotide in noncatalytic sites (Abrahams et al., 1994). The double mutations ␣K175E/␣R365W, ␣K175I/␣R365W, and ␣D261N/␣R365W were constructed. Unfortunately, purified F 1 could not be obtained from the first two mutants because of impaired assembly and oligomeric instability. This appears to be a general feature of mutations at the ␣-Lys 175 locus. However, purified F 1 was obtained from the ␣D261N/␣R365W mutant, and it provided valuable information about the noncatalytic sites.
From the nucleotide-binding data described under "Results" we can draw three conclusions. First, removal of the Asp carboxyl at position ␣-261 by the Asp 3 Asn mutation greatly reduces noncatalytic site affinity for MgATP and MgADP (Figs. 2A and 3A), indicating that the ␣-Asp 261 carboxyl interacts directly with the Mg moiety of bound Mg-nucleotide. This is consistent with the x-ray structure (Abrahams et al., 1994). Second, it is apparent that this interaction is the primary determinant of the preference of noncatalytic sites for Mgnucleotides, and provides a large component of the overall binding energy. From the K d values for MgATP and MgADP in Table II it is apparent that in combination residue ␣-Asp 261 and the Mg of the Mg-nucleotide provide about 3.2 kcal/mol of the MgATP binding energy, and that the corresponding value for MgADP is about 2.5 kcal/mol. Third, since Asp is the residue at position ␣-261 in wild-type, the natural ligands will be MgATP and/or MgADP.
It was interesting that free ADP was bound with higher affinity by the ␣D261N/␣R365W enzyme than by the ␣R365W enzyme (Fig. 3B). A possible explanation is that free ADP and the ␣-Asp 261 carboxyl in the noncatalytic sites of the ␣R365W enzyme undergo mutual electrostatic repulsion, and thus the Asp 3 Asn mutation can assist binding in this situation. This was not the case, however, with free ATP, which bound overall with relatively low affinity to both enzymes (compare Fig. 2B with Fig. 3B). The presence of two mutations in the binding site in the ␣D261N/␣R365W enzyme places limitations on the interpretations of these findings.
For some time we have propounded the view that F 1 catalysis is not dependent upon occupancy of the noncatalytic sites (Perlin et al., 1984;Wise and Senior, 1985), although this idea has been challenged (Bullough et al., 1988;Milgrom et al., 1990Milgrom et al., , 1991Allison et al., 1992;Harris, 1993). In a recent paper we provided evidence showing that rapid rates of catalysis were achieved in ␣R365W F 1 under conditions where less than one FIG. 3. Binding of free ADP and MgADP to noncatalytic sites of mutant F 1 preparations. The procedures were as described in the legend to Fig. 2 Weber et al. (1994). As noted in the text the concentration of MgSO 4 was much higher in this work (25 mM) to allow higher concentrations of MgATP to be achieved. b As Figs. 2A and 3A show, full saturation could not be achieved, hence n was set at 3 sites. c A value of 24 M was reported previously in Weber et al, (1994). As noted in the text the concentration of MgSO 4 was much higher in this work (25 mM) to allow higher concentrations of MgADP to be achieved. noncatalytic site was filled by nucleotide, on average, per enzyme molecule (Weber et al., 1994). The data in this paper reaffirm our conclusion, by demonstrating unequivocally that the ␣D261N/␣R365W enzyme, with empty noncatalytic sites, still hydrolyzes MgATP.
The ␣D261N/␣R365W mutant strain grew normally on succinate plates and in limiting glucose medium, showing that rates of oxidative phosphorylation in vivo were normal. Thus, it might appear that oxidative phosphorylation occurs in this strain without any requirement for the noncatalytic sites to be occupied by nucleotide. It could be argued that the proton gradient (⌬p) alters the behavior of the ␣D261N/␣R365W noncatalytic sites in membrane-bound enzyme, such that they are induced to bind nucleotide, although this seems unlikely given the clear structural rationale for the mutational effect. No doubt, however, the ␣D261N/␣R365W mutant will be helpful in seeking answers to this question, and also to the wider question of the possible regulatory or modulating roles of the noncatalytic sites under the variety of growth conditions encountered by E. coli in vivo.