Identification of the F1-binding Surface on the δ-Subunit of ATP Synthase

The stator function in ATP synthase was studied by a combined mutagenesis and fluorescence approach. Specifically, binding of δ-subunit to δ-depleted F1 was studied. A plausible binding surface on δ-subunit was identified from conservation of amino acid sequence and the high resolution NMR structure. Specific mutations aimed at modulating binding were introduced onto this surface. Affinity of binding of wild-type and mutant δ-subunits to δ-depleted F1 was determined quantitatively using the fluorescence signals of natural δ-Trp-28, inserted δ-Trp-11, or inserted δ-Trp-79. The results demonstrate that helices 1 and 5 in the N-terminal domain of the δ-subunit provide the F1-binding surface of δ. Unexpectedly, mutations that impaired binding between F1and δ were found to not necessarily impair ATP synthase activity. Further investigation revealed that inclusion of the soluble cytoplasmic domain of the b subunit substantially enhanced affinity of binding of δ-subunit to F1. The new data show that the stator is “overengineered” to resist rotor torque during catalysis.

ATP synthase is the membrane enzyme responsible for ATP synthesis in oxidative and photophosphorylation of prokaryotes and eukaryotes, and also for ATP-driven proton pumping to generate the transmembrane proton gradient in bacterial membranes. In Escherichia coli, ATP synthase consists of a complex of eight subunits, ␣ 3 ␤ 3 ␥␦⑀ab 2 c n . It was defined in earlier times in terms of a membrane-peripheral F 1 sector (␣ 3 ␤ 3 ␥␦⑀) containing three catalytic sites, and a membrane-embedded F 0 sector (ab 2 c n ), which carries out transmembrane proton transport. Recent work has demonstrated that the enzyme functions as a rotary motor. The centrally located "rotor" consists of subunits ␥⑀c n . At the top, it rotates inside the ␣ 3 ␤ 3 hexagon, and thus modifies the activities of the catalytic sites; at the base, it rotates against subunit a, thereby facilitating proton movement. In this way, the energy of the proton gradient is transduced into the energy of ATP synthesis/hydrolysis. Understanding the mechanism by which this occurs is currently of major interest. To ensure that subunits a and ␣ 3 ␤ 3 remain firmly fixed in relation to each other they are connected by a peripheral structure, the "stator" stalk, consisting of subunits b 2 and ␦. For recent reviews of the structure and function of ATP synthase, see Refs. 1-3. The stator must be able to resist strain resulting from rotor torque, thus its construction is of considerable importance. The dimer of b-subunits forms an elongated helical connection between subunit a and the C-terminal domain of the ␦-subunit, it lies at one side of the ␣ 3 ␤ 3 hexagon, and its functional domains have been well characterized (4 -6). There may exist functional interactions between b 2 and the ␣ 3 ␤ 3 hexagon (7). Currently only partial high-resolution structure has been reported for the b subunit (8,9). The ␦-subunit has been shown by electron microscopy to bind to the very top ("crown") of the ␣ 3 ␤ 3 hexagon (10), and the homologous mitochondrial OSCP 1 subunit also binds at the top of the molecule, with its C-terminal domain at the side (11). The N-terminal region of ␣-subunit (residues ␣1-15 or ␣1-19) was shown to be necessary for ␦ binding by proteolysis experiments (12), and cross-linking between an inserted Cys at residue ␣2 and natural Cys residue(s) in ␦ (13) provided further evidence that the N terminus of ␣, which is not seen in high-resolution x-ray structures, binds to ␦. In our laboratory, quantitative determination of the affinity of binding between ␦-subunit and the ␣ 3 ␤ 3 ␥⑀ complex (also called "␦-depleted F 1 "), using a novel fluorescence assay dependent upon the fluorescence of the single natural Trp in ␦, revealed a K d of 1.4 nM in the E. coli enzyme (14). This is equivalent to a binding energy of ϳ50 kJ/mol, approximately equivalent to rotor torque (15,16). A similar K d value was reported for the chloroplast enzyme (17) using a fluorescent probe attached to ␦. We were able to show that a fragment of ␦ containing only N-terminal residues 1-134 2 (called ␦Ј) bound with the same affinity as intact ␦, demonstrating that all the determinants of binding lie in the N-terminal region of ␦ (14). The structure of the N-terminal domain of ␦ (residues 2-106 in the fragment ␦Ј) has been determined to high-resolution by NMR (18). Examination of this structure in combination with homology searches of the sequences of other species of ␦ and OSCP suggested possible residues that might be involved in binding interaction at the F 1 /␦ interface. Using the quantitative binding assay that we had developed, in combination with mutagenesis, we were able here to test these ideas. From the results, this paper presents the identification of the surface on the ␦-subunit that provides interactions with F 1 . In addition we show that inclusion of the soluble cytoplasmic domain of the b subunit substantially enhanced the affinity of binding of ␦-subunit to F 1 . tional properties, and has a low Trp background signal in the ␦-binding assay. For the ATP-driven proton pumping assay, the ␦-depleted F 1 was from strain pSWM92/DK8 (␦W28L), which is easier to completely deplete of ␦-subunit than wild-type F 1 (14). Growth yields on limiting glucose were measured as in Ref. 19.
Expression and Purification of the Cytoplasmic Domain of b Subunit in Soluble Form-The soluble form of the cytoplasmic domain of the b subunit named b ST34 -156 was expressed from plasmid pJB3 (20) after transformation into strain DK8. The procedure followed essentially that described in Ref. 20. Briefly, six 1-liter cultures of strain pJB3/DK8 were grown and harvested, cells were broken by a French press, and the supernatant fraction was collected after centrifugation. Ammonium sulfate was added to 60% saturation, the pellet was resuspended and dialyzed, the protein was then purified by sequential chromatography on Whatman DE-52 anion exchange resin and by gel filtration on Sephacryl S-100. Buffers were described as in Ref. 20. Yield of b ST34 -156 was 23 mg of protein/liter of culture.

RESULTS
A Potential F 1 -binding Surface on the ␦-Subunit- Fig. 1 shows the NMR structure of the N-terminal domain of the ␦-subunit of E. coli ATP synthase (18) as determined using the proteolytic fragment (␦Ј) consisting of residues 1-134 (intact ␦ has 177 residues). As noted in the Introduction, because pro-teolytic removal of the 43 C-terminal residues has no effect on the affinity for binding to ␦-depleted F 1 (14), it is highly likely that the domain shown in Fig. 1 contains most, if not all, of the amino acid residues involved in binding of ␦ to F 1 . Comparison of the sequences of ␦ or OSCP from 90 species by BLAST search (21) showed that the highly conserved residues are clustered on a relatively small portion of the ␦ surface, as demonstrated in Fig. 2. Because ␦ sequences are in general not highly conserved, for example, in comparison with those of ␣or ␤-subunits (22), it appeared probable that this conserved part of ␦, formed by helices 1 and 5 (colored dark blue and yellow, respectively, in Fig. 1), constituted the F 1 -binding site. We decided to test this hypothesis by mutagenesis of four conserved residues, ␦-Tyr-11, ␦-Ala-14, ␦-Asn-75, and ␦-Val-79. From the BLAST search (above) we found the following degrees of conservation of these residues. At position ␦-11, Tyr occurred in 95% of species and Phe in 5%; at ␦-14, Ala occurred in 90%; at ␦-75 Asn occurred in 78% and Asp in 6%; and at ␦-79 a hydrophobic residue (Val, Leu, or Ile) occurred in 92% of species. Fig. 2A shows the location of these residues on the hypothesized binding surface.
Mutagenesis and Purification of the ␦-Subunit-Two mutations were introduced in each of the four conserved positions, namely: ␦Y11A,␦Y11W; ␦A14D,␦A14L; ␦N75A,␦N75E; ␦V79A,␦V79W. The Ala mutations were designed to give an estimate of the contribution of the natural residue to subunit- subunit binding energy. The introduced Trp residues could give new probes for assaying the interaction of ␦ with F 1 and in both cases they were combined with the ␦W28L mutation to remove the only naturally occurring Trp in ␦. The remaining mutations were designed to perturb binding of ␦ to F 1 . The mutation ␦G150D was included as a control. This mutation is known to completely impair ATP synthase function (23), but because it occurs in the C-terminal domain of the ␦-subunit it is expected to interfere with binding of ␦ to the b subunit (24) and not with binding to F 1 .
For each mutant, we purified ␦-subunit following previously published procedures (14,20), except in the cases of ␦A14D and ␦G150D, where the concentration of saturated ammonium sulfate necessary to precipitate ␦ was 50% rather than 32%. Yields ranged from 0.5 to 3 mg/liter culture. Purity of the mutant ␦ was checked by SDS gels, and in each case was the same as with wild-type as previously shown in Ref. 14. Each purified mutant ␦ was assayed to determine the affinity of binding to F 1 using the fluorescence assay developed previously (14). This assay depends on the fact that binding of ␦ to F 1 that has been depleted of ␦ produces a large enhancement of fluorescence of the natural ␦-Trp-28 residue. As noted above, for the ␦Y11W and ␦V79W mutants, we removed the natural ␦-Trp-28 (by combining with ␦W28L) and used the new Trp as the probe. For each mutation we also assayed the effect on function in vivo, by measuring growth of mutant cells on succinate plates or in limiting glucose medium, and we measured effects on ATPdriven proton pumping in vitro by reconstitution with membrane vesicles and ␦-depleted F 1 using purified mutant ␦.
Fluorescence Spectra of Purified Mutant ␦- Fig. 3, A-E, shows the Trp fluorescence spectra of the purified mutant ␦-subunits in the absence of F 1 . Except for the ␦Y11W/W28L and ␦V79W/W28L mutants (dashed curves in Fig. 3, A and D, respectively), the fluorophor is the natural Trp in position ␦-28. Several of the spectra of mutant ␦ (␦A14L, ␦V79A, and ␦G150D) resemble closely that of wild-type ␦ (dotted lines in Fig. 3, A-E) with a maximum at 326 nm, indicating a relatively unpolar environment for the tryptophan side chain. The wavelength position of the spectra for ␦N75A and ␦N75E (Fig. 3C) is also very similar; however, the fluorescence intensity is 20 -30% higher, suggesting that the mutations caused minor changes in the environment of ␦-Trp-28. In the ␦A14D mutant, the spectrum of ␦-Trp-28 was red-shifted by 2 nm, in the ␦Y11A mutant by 5 nm, indicative of an increased polarity experienced by the fluorophor. In general, however, the mutations appeared not to perturb the tertiary structure to any large extent, by this criterion.
A tryptophan inserted in position ␦-11, in the ␦Y11W/W28L mutant, has an emission maximum of 337 nm (Fig. 3A), suggesting clearly a more polar environment than for ␦-Trp-28 in wild-type. A tryptophan in position ␦-79 (Fig. 3D) has nearly aqueous surroundings, as indicated by fluorescence maximum at 350 nm of the ␦V79W/W28L mutant. These data are consistent with the predicted location from the structure (Figs. 1 and 2).
Effect of Addition of ␦-depleted F 1 on the Fluorescence Spectra of Trp in ␦-Subunits-Wild-type and most of the mutant ␦ preparations contained ␦-Trp-28 as the sole Trp. The response of the ␦-Trp-28 fluorescence upon binding of ␦-depleted F 1 to wild-type ␦ is shown in Fig. 4A. As described previously (14), there was an increase in fluorescence intensity of about 50%, combined with a slight blue-shift (ϳ4 nm), indicative of a change to a more unpolar environment upon binding. The same change was seen in the ␦V79A and ␦G150D mutants (data not shown). In the ␦N75A and ␦N75E mutants the fluorescence signal of unbound ␦ was higher to begin with (Fig. 3C), but the final signal after binding to ␦-depleted F 1 was very similar to wild-type ␦, as in Fig. 4A (data not shown). In the ␦Y11A mutant, which had a lower intensity and a more red-shifted spectrum before addition of F 1 , the fluorescence increase upon F 1 binding was about 40%, and the blue-shift was by 6 nm. Mutants ␦A14D and ␦A14L did not show any change in fluorescence under the experimental conditions used (discussed below).
The engineered ␦-Trp-11, in the ␦Y11W/W28L mutant, experiences a very pronounced fluorescence increase upon F 1 binding, by 80 -90% at 330 nm, combined with a blue-shift of 4 nm (Fig. 4B). In contrast, the fluorescence of the engineered ␦-Trp-79, in the ␦V79W/W28L mutant, is quenched upon F 1 binding, by about 50% at 350 nm, accompanied by a blue-shift of 14 nm (Fig. 4C). The blue-shifts of all three Trp residues (␦-Trp-11, ␦-Trp-28, and ␦-Trp-79) reflect a more unpolar environment of the tryptophan side chains upon F 1 binding, suggesting that the fluorophors are better shielded from the medium in this state. This, in turn, provides good evidence that all three fluorophors are actually located at, or close to the F 1 -binding surface on the ␦-subunit. These data indicated that for all except the ␦A14D and ␦A14L mutants a direct fluorescence assay was available for binding of ␦-subunit to ␦-depleted F 1 , and that in the cases of ␦A14D and ␦A14L a competition assay would be required.
Fluorescence Titration of Binding of ␦-Subunit to ␦-depleted F 1 and Determination of Binding Affinities of Mutant ␦-Subunits-Binding of mutant ␦ to ␦-depleted F 1 was measured by monitoring the tryptophan fluorescence of ␦, either because of the natural ␦-Trp-28 or because of the engineered ␦-Trp-11 or ␦-Trp-79. For each mutant, initially an F 1 concentration of 0.05 M was chosen. If this concentration was too high to obtain a reliable K d value (because the resulting titration curve was "stoichiometric," for example, the curve for wild-type ␦ in Fig.  5A), the titrations were repeated using 0.01 M F 1 . If the initial F 1 concentration was too low (i.e. the titration curves did not reach saturation), the experiments were repeated using 0.5 M F 1 . Typical titration curves are shown in Figs. 5 and 6 and the resultant calculated K d values are given in Table I. In all cases where a binding stoichiometry could be determined reliably (i.e. with stoichiometric or close-to-stoichiometric binding curves), it was very close to 1 (0.9 to 1.2) mol of ␦ per mol of F 1 .
Both mutations in position ␦-11 resulted in a significantly decreased binding affinity. Based on the results obtained with the ␦Y11A mutant, the ␦-Tyr-11 side chain in wild-type ␦ contributes about 12 kJ/mol of binding energy (Table I). In this case, tryptophan is not a good substitute for tyrosine, because the ␦Y11W mutation also causes a loss of close to 12 kJ/mol in binding energy (calculated by comparing ␦Y11W/W28L to ␦W28L mutant).
Titrations with the ␦A14D and ␦A14L mutants, even using 0.5 M F 1 and up to 2.5 M ␦, did not result in significant changes in the ␦-Trp-28 fluorescence (Fig. 5B). One possible explanation is that the binding affinity is so low that even at the highest concentrations used, binding is negligible. In this case, K d would be Ͼ5 M (and the loss in binding energy Ͼ20 kJ/mol). Alternatively, it is possible that binding occurs, but that the mutations affect the interaction between ␦ and F 1 in such a way that the fluorescence of ␦-Trp-28 does not respond to binding. To address this question, we performed competition titration experiments, which showed that a 6 -10-fold excess of ␦A14D or ␦A14L mutant ␦-subunit did not significantly reduce binding of wild-type ␦. On the basis of these results, we can calculate that the K d for ␦A14D or ␦A14L mutant is at least 0.1 M. Thus, from either calculation, both mutations at residue ␦-14 perturb interaction between ␦ and F 1 considerably.
In position ␦-75, neither mutation ␦N75A nor ␦N75E has a significant effect on the F 1 binding affinity (Fig. 6A and Table  I). Specifically the results obtained with the ␦N75A mutant suggest that the natural asparagine side chain does not contribute binding energy. In contrast, the valine side chain of residue ␦-Val-79 makes a moderate contribution of about 6 kJ/mol (Table I). A tryptophan in this position strongly interferes with binding to F 1 , with a loss of binding energy of more than 11 kJ/mol (comparing ␦V79W/W28L to ␦W28L). As expected, the mutation ␦G150D in the C-terminal region of ␦, used as a control here, had no effect on F 1 binding affinity.
Effect of the ␦ Mutations on ATP Synthesis in Vivo-Plasmids containing mutant uncH (␦-subunit) genes were transformed into strain AN2015, which contains the chromosomal mutation ␦-Trp-28 3 stop and is therefore unable to grow by oxidative phosphorylation. All plasmids carrying mutations in positions 11, 14, 75, and 79 of ␦ were able to restore oxidative phosphorylation, as demonstrated by their growth yields on limiting glucose (Table II) and their ability to grow on plates containing succinate as the sole carbon source (not shown). In contrast, the mutation ␦G150D prevented ATP synthesis by the ATP synthase in vivo (Table II).
Effect of ␦ Mutations on ATP-driven H ϩ Pumping in Vitro-KSCN-stripped membranes were reconstituted with wild-type or mutant ␦ together with ␦-depleted F 1 and proton pumping was initiated by addition of ATP. It was found that none of the mutants at positions 11, 14, 75, or 79 of ␦ caused significant impairment of ATP-driven H ϩ pumping (Table II). These data indicate that, in the presence of intact F 0 , functional binding of the mutant ␦-subunits to F 1 did occur. From the considerations presented in the Introduction, one likely mechanism for such FIG. 4. Tryptophan fluorescence spectra of purified wild-type and mutant ␦-subunits after addition of ␦-depleted F 1 . Spectra are shown before (dotted lines) and after (solid lines) addition of 2-3-fold excess of ␦-depleted F 1 . All spectra are corrected for the contribution by F 1 and unbound ␦. A, wild-type ␦. B, ␦Y11W/W28L. C, ␦V79W/W28L.

FIG. 5. Fluorescence titrations to assess binding of ␦-Tyr-11
and ␦-Ala-14 mutant ␦ to ␦-depleted F 1 . Pure ␦-subunit was mixed with ␦-depleted F 1 , and the resulting fluorescence enhancement at 325 nm (after subtraction of the contribution of ␦ alone and of F 1 alone) was plotted against the concentration of ␦. an effect would be through involvement of the b subunit dimer, and specifically the cytoplasmic domain of b that interacts with ␦. In contrast, the ␦G150D mutant prevented formation of a proton gradient upon ATP hydrolysis (Table II) as expected from previous work (23). As already noted, this mutation occurs in the C-terminal region of ␦ and is expected to interrupt interaction of ␦ with b.
Effect of the Soluble Cytoplasmic Domain of the b Subunit on Binding of ␦ to ␦-depleted F 1 -For these experiments the soluble cytoplasmic domain of the b subunit that we used was purified b ST34 -156 (20). It consists of residues 34 through 156 (C terminus) of the b subunit with an additional Ser-Thr-sequence at the N terminus. The purified protein showed a single major band on SDS gels with the expected mobility (20). Functional integrity of b ST34 -156 was evaluated using the method of Dunn (25). In this assay, inhibition of reconstitution of ATP-

TABLE II Effect of mutations in ␦-subunit on growth of cells in vivo
and ATP-driven proton pumping in vitro Growth yields in limiting glucose medium and ATP-driven proton pumping in membrane vesicles were measured as described under "Experimental Procedures." For the former, plasmid containing wildtype or mutant ␦ (uncH) gene was transformed into strain AN2015, which contains a nonsense mutation ␦W28stop. For the latter, 500 g of KSCN-stripped membranes were reconstituted with 100 g of ␦-depleted F 1 plus 10 g of wild-type or mutant ␦-subunit, and ATP-induced quench of acridine orange fluorescence was measured. dependent proton pumping by F 1 in stripped membranes is measured. Inhibition comes about as a result of competition between F 0 and added b subunit cytoplasmic domain for a limited number of F 1 molecules. We found that reconstitution was inhibited by b ST34 -156 in a dose-dependent manner. The ratio of b ST34 -156 /F 1 giving 50% inhibition of reconstitution was 2 g/g, the same as in Ref. 25. Theoretically b ST34 -156 should contain no Trp residue. Calculation of the Trp content of our preparation, from the fluorescence spectrum in 6 M guanidine hydrochloride, revealed a Trp content of 0.1 mol/mol. The fluorescence because of Trp contamination in b ST34 -156 was corrected routinely.
We demonstrated that addition of b ST34 -156 to wild-type or any of the mutant ␦-subunits in the absence of F 1 had no effect on the Trp fluorescence of the ␦-subunit. In initial experiments with F 1 present, we chose to use a mutant ␦ with a high Trp signal and a relatively low affinity for binding to F 1 , namely ␦Y11W/W28L. Fig. 7A demonstrates an experiment in which b ST34 -156 and ␦-depleted F 1 were added at a constant concentration and the amount of added ␦Y11W/W28L subunit was varied. This fluorescence titration showed that addition of b ST34 -156 to the binding assay had a very large effect, reducing K d for binding of ␦Y11W/W28L to F 1 from 0.5 M to Ͻ5 nM. In Fig. 7B we determined the apparent K d for b  . Here, with constant concentrations of ␦-depleted F 1 and ␦Y11W/ W28L subunit present, the concentration of b ST34 -156 was varied. The apparent K d for the b ST34 -156 dimer (the physiological form, Refs. 4 and 5) was 150 nM, with two independent experiments showing excellent agreement. Fig. 8 shows the effect of inclusion of b ST34 -156 on the binding of wild-type and mutant ␦-subunits to ␦-depleted F 1 . We used a concentration of 4 M b ST34 -156 in these experiments to be sure of saturation while reducing the impact of the Trp fluorescence because of contaminants in b ST34 -156 to a minimum. Presence of this contamination restricted the concentration range accessible for titration experiments to Ն100 nM F 1 for all ␦ mutants except ␦Y11W/W28L, with its larger signal, where 20 nM F 1 could be used. Nevertheless, it is clear from Fig. 8, A-C, that in mutants ␦Y11W/W28L, ␦Y11A, ␦V79A, and ␦V79W/W28L, inclusion of b ST34 -156 in the binding assay brought about a substantial increase in binding affinity between the ␦-subunit and F 1 . The titration curves (Fig. 8) are all fully or close to ''stoichiometric,'' thus we can only stipulate that in these cases that the K d was reduced to Ͻ5 nM (except for ␦Y11W/W28L where we can say K d was reduced to Ͻ1 nM). Still, by comparison with the K d values obtained in the absence of b ST34 -156 (shown in Table I), it is clear that the changes are large, indeed Ն500-fold in the case of ␦Y11W/W28L mutant.
With regard to wild-type ␦, the above mentioned technical constraints prevented us from determining whether the binding affinity for F 1 was increased by inclusion of b ST34 -156 (the K d in the absence of b ST34 -156 was already 1.4 nM, Table I). If we assume that a 500-fold increase in affinity occurs for wildtype, as it did for ␦Y11W/W28L, this would give a K d for wild-type ␦ in the presence of b ST34 -156 of Յ3 pM.
In contrast, with mutants ␦A14D and ␦A14L the titration curves (not shown) looked the same as those in Fig. 5B (squares and diamonds) even when b ST34 -156 was present. From this we can conclude that either these mutant ␦-subunits do not bind significantly to F 1 even in the presence of b ST34 -156 , or that if they do bind, the binding does not engender a change in the fluorescence signal of ␦-Trp-28.
Titration of Mutant ␦-Subunits with KSCN-stripped Membranes and ␦-depleted F 1 in the Reconstituted Proton-pumping Assay- Table I demonstrated that the mutant ␦A14D and ␦A14L subunits did reconstitute ATP-driven proton pumping in membrane vesicles when added back under saturating conditions with ␦-depleted F 1 to KSCN-stripped membranes. To investigate the binding properties of these mutant ␦-subunits in the presence of intact F 0 further we titrated them in this assay with constant amounts of stripped membranes and ␦-depleted F 1 . Whereas we had shown before (14) that such a titration cannot be used to determine absolute values of K d for binding of ␦ to F 1 , nevertheless, we expected that by comparison of the mutants with wild-type we could get qualitatitve information regarding the binding of the ␦A14D and ␦A14L proteins. Fig. 9 indicates that the titration curves obtained with these (and other) mutants are not very dissimilar from wild-type. We therefore conclude that the failure to see enhancement of fluorescence when these mutant subunits are added to ␦-depleted F 1 in the presence of b ST34 -156 (above) is most likely because of the fact that the mutations affect the environment of residue ␦-Trp-28 and prevent enhancement of its fluorescence signal upon binding. DISCUSSION The goal of this study was to investigate in detail the interaction between the ␦-subunit and F 1 in ATP synthase, and specifically to identify the F 1 -binding surface on the ␦-subunit. Based on the location of conserved or conservatively replaced residues in one particular region of the high resolution NMR structure of ␦, we hypothesized that the F 1 -binding surface might be formed by helices 1 and 5 (see Figs. 1 and 2). Our results confirmed the hypothesis. However, several mutations in ␦ that clearly disrupted binding of ␦ to F 1 did not lead to impaired function, an unexpected finding. Additional studies showed that inclusion of the soluble cytoplasmic domain of the b subunit substantially enhanced the binding affinity between ␦ and F 1 and compensated for loss of binding affinity caused by ␦ mutations.
As target residues for mutational analysis we selected ␦-Tyr-11 and ␦-Ala-14 in helix 1 and ␦-Asn-75 and ␦-Val-79 in helix 5. The first evidence that these residues are at or close to the F 1 -binding site came from tryptophan substitutions in positions 11 and 79, whose fluorescence signals responded strongly to binding of F 1 (Fig. 4, B and C). Further evidence came from the titrations in Figs. 5 and 6, and from competition binding experiments. Calculated binding affinity measurements (K d values, see Table I) indicated that three of the four residues are directly involved in binding. The tyrosine side chain of residue ␦-Tyr-11 contributes about 12 kJ/mol binding energy, possibly because ofor -cation interactions. Increasing the size of the side chain in position ␦-Ala-14, either by itself (␦A14L mutant) or in combination with introduction of a negative charge (␦A14D mutant) reduces the affinity for F 1 significantly, probably by affecting interprotein surface complementarity. The valine side chain in position ␦-Val-79 contributes about 6 kJ/mol binding energy, very likely because of hydrophobic interactions. The lack of binding energy contribution of the side chain of residue ␦-Asn-75 suggests this residue is not directly involved in binding. However, it might not necessarily be taken as an argument against its location at the F 1 -binding surface; in an analysis of the interaction between the human growth hormone and its receptor (26) it was found that only one-quarter of the residues at the protein-protein interface had a significant impact on the binding energy. Whereas this might be an extreme case, at many proteinprotein interface amino acid side chains can be found that should have the potential to participate in binding, but in fact play no or only a very minor role (see examples summarized in Ref. 27).
Surprisingly, despite losses in F 1 -binding energy of up to 15 kJ/mol, all mutations of residues in helices 1 and 5 were still fully or nearly fully functional in vivo and in vitro. In a previous study (14) we determined that the ␣G29D mutation reduces the binding affinity between ␦ and F 1 moderately, corresponding to a loss in binding energy of about 7 kJ/mol. We ascribed the strong functional impairment of this mutant ATP synthase to the interruption of binding of ␦ to F 1 . This led us to conclude that the stator resistance function was finely balanced. However, the results of the present study showed that much larger losses of binding energy between ␦ and F 1 were well tolerated.
Thus, we must now conclude that not all of the binding energy necessary to affix the stator stalk to F 1 , to resist the elastic strain generated by rotational catalysis, must necessarily be derived from ␦/F 1 interactions. In all likelihood, interactions between the b subunits and ␦ and/or F 1 also contribute.
To explore this possibility we included the soluble cytoplasmic domain (b Ref. 20) in ␦ binding assays and found that it substantially decreased the K d of binding of ␦ to F 1 . Due to technical limitations of the fluorescence assays, absolute values for this K d in presence of b ST34 -156 could not be obtained, but an enhancement of Ն500-fold in affinity was evident from the results, equivalent to an additional binding energy of Ն15 kJ/mol. The additional binding energy could come from interactions between b subunit and ␣ or ␤, and/or from b-induced conformational changes in ␦. Interestingly, the K d for interaction between b ST34 -156 and isolated ␦ (in the absence of F 1 ) was 5-10 M (20) but the K d measured here for binding of b  in the presence of isolated ␦ and ␦-depleted F 1 was 150 nM, indicating a considerable cooperativity between the stator subunits. Overall, the new data indicate that the wild-type stator stalk is "overengineered," i.e. it is equipped with excess binding energy. This might explain why only very few impairing point mutations in ␦ have been found (23,28).
The experiments presented here supplement and extend earlier studies on OSCP. A study using deletion mutants of bovine OSCP indicated that removal of the N-terminal 28 residues, corresponding approximately to all of helix 1 in E. coli ␦, impaired binding of OSCP to F 1 (29). Also a study of rat OSCP showed that the strongly conserved residue ␦-Arg-85 (Arg or Lys in 95% of sequences, residue Arg-94 in OSCP) contributed significantly to F 1 -binding energy (30). This residue occurs in the loop following helix 5 (see Fig. 2). Interestingly, despite the loss of binding energy, the ␦R85A and ␦R85Q mutants were still functional in vitro (30), and the authors concluded, as we do here, that other interactions provide binding energy. In E. coli, the mutation ␦R85Q reduced the membrane-bound ATPase activity, by 50%, but had little effect on growth characteristics (28).
The quantitative binding assays described here for wild-type and mutant ␦ binding to ␦-depleted F 1 in the presence or absence of the soluble cytoplasmic domain of the b subunit will allow us in future work to assess further aspects of stator subunit interactions. The influence of specific residues of the b subunits on ␦ binding affinity can readily be studied by mutagenesis, for example. In addition, we can now investigate, by mutagenesis and other techniques, the binding site for ␦ on F 1 , which appears to consist to a significant extent of one or more of the extreme N termini of the three ␣ subunits, for which a high resolution structure is not yet available.