Analysis of Sequence Determinants of F1Fo-ATP Synthase in the N-terminal Region of α Subunit for Binding of δ Subunit

The stator in F1Fo-ATP synthase resists strain generated by rotor torque. In Escherichia coli, the b2δ subunit complex comprises the stator, bound to subunit a in Fo and to the α3β3 hexagon of F1. Previous work has shown that N-terminal residues of α subunit are involved in binding δ. A synthetic peptide consisting of the first 22 residues of α (αN1–22) binds specifically to isolated wild-type δ subunit with 1:1 stoichiometry and high affinity, accounting for a major portion of the binding energy between δ and F1. Residues α6–18 are predicted by secondary structure algorithms and helical wheels to be α-helical and amphipathic, and a potential helix capping box occurs at residues α3–8. We introduced truncations, deletions, and mutations into αN1–22 peptide and examined their effects on binding to the δ subunit. The deletions and mutations were introduced also into the N-terminal region of the uncA (α subunit) gene to determine effects on cell growth in vivo and membrane ATP synthase activity in vitro. Effects seen in the peptides were well correlated with those seen in the uncA gene. The results show that, with the possible exception of residues close to the initial Met, all of the αN1–22 sequence is required for binding of δ to α. Within this sequence, an amphipathic helix seems important. Hydrophobic residues on the predicted nonpolar surface are important for δ binding, namely αIle-8, αLeu-11, αIle-12, αIle-16, and αPhe-19. Several or all of these residues probably make direct interaction with helices 1 and 5 of δ. The potential capping box sequence per se appeared less important. Impairment of α/δ binding brings about functional impairment due to reduced level of assembly of ATP synthase in cells.

direction, ATP hydrolysis in the catalytic sites drives rotation of the rotor (3), which then generates uphill transport of protons across the bacterial plasma membrane to form the electrochemical gradient essential for nutrient uptake, locomotion, and other functions. Again, rotor strain must be resisted by the stator for efficient function. The mechanisms by which catalysis, proton gradient formation, and subunit rotation are functionally integrated are subjects of active investigation (4 -7).
The structure and function of the stator have been reviewed recently (8,9). The stator (b 2 ␦) interacts with the ␣ 3 ␤ 3 catalytic unit via ␦/F 1 interactions and with the proton-translocating machinery via b 2 /a interactions. ␦ and b 2 interact together via their C-terminal regions. There is also interaction between b 2 and ␣ or ␤ subunits (10 -13). In recent papers, we have studied the binding of the ␦ subunit to F 1 . Using novel tryptophan fluorescence assays, we established quantitative parameters for ␦ binding to F 1 (14) and demonstrated that helices 1 and 5 of the N-terminal domain of the ␦ subunit form the F 1 -binding surface (15). An earlier report had used NMR to establish the structure of the N-terminal domain of ␦ subunit, which is composed of a six-helix bundle, and had shown that helices 1 and 5 form a hydrophobic groove (16).
In Ref. 17, we also studied the ␦ binding surface on F 1 . ␦ subunit (and its mitochondrial homolog oligomycin-sensitivity conferral protein) was known from electron microscopy studies to bind at the "top" of F 1 (18,19). Proteolysis (20) and crosslinking (21) experiments had suggested that the extreme Nterminal residues of the ␣ subunit were involved in binding of ␦. For example, removal of the first 15 residues of ␣ by trypsin or of the first 19 residues by chymotrypsin was sufficient to greatly reduce ␦ binding to F 1 (20). X-ray crystallography studies have not yet been able to determine the structure of these ␣ subunit residues (7,22,23). In Ref. 17, we showed that a 22-residue synthetic peptide corresponding in sequence to the N-terminal residues of ␣ subunit with free N and C termini (␣N1-22) 1 was able to bind to wild-type ␦ subunit with high affinity and specificity (K d ϭ 130 nM) and with 1:1 stoichiometry, effectively mimicking the binding of intact F 1 to ␦. Mutations on the F 1 -binding surface of ␦, which impaired binding of F 1 , were seen to impair binding of ␣N1-22, providing further evidence for the specificity of binding of the peptide. The data provided clear evidence that the N terminus of ␣ subunit provides the major fraction of total binding energy between the stator component ␦ and F 1 . We noted in Ref. 17 that secondary structure algorithms consistently predict residues ␣6 -18 to be ␣-helical and that a potential helix capping box sequence (24) occurs at residues ␣3-8. 2 A helical wheel diagram reveals that the predicted helix would be amphipathic. These features are shown in Fig. 1, A and C. Circular dichroism studies of the ␣N1-22 peptide supported the view that it assumed a helical structure (17). NMR studies of ␣N1-22 bound to ␦ subunit in a 1:1 complex have now shown that the peptide adopts a helical conformation when bound to ␦. 3 These considerations lead to a general model of how ␣N1-22 fits into the hydrophobic groove between helices 1 and 5 of ␦, presented in Fig. 1B.
Here we have introduced truncations, point mutations, and deletions into ␣N1-22 peptide in order to examine the importance of specific regions of its sequence for binding to ␦. Deletions and mutations were also made in the uncA (␣ subunit) gene, to examine effects on cell growth in vivo and membrane ATP synthase function in vitro. The results were well correlated and allow conclusions regarding specific sequence determinants of ␦ binding and overall ATP synthase function in the N-terminal region of ␣.

EXPERIMENTAL PROCEDURES
Purification of ␦ Subunit, Purification of F 1 , Preparation of Membrane Vesicles, Assay of ATPase Activity, Assay of ATP-driven Proton Pumping in Membrane Vesicles, and Growth Yield Assays-These were as described previously (14,15). For assay of membrane ATPase activity, 50 g of membranes were assayed at 30°C in 1 ml of assay medium (50 mM Tris-SO 4 , pH 8.5, 10 mM NaATP, 4 mM MgCl 2 ) for appropriate times (1-2 min for wild type, 10 -20 min for mutants with low activity). To assess sensitivity of ATPase to dicyclohexylcarbodiimide (DCCD) inhibition, membranes (1 mg of protein/ml) were preincubated at 30°C in 10 mM HEPES-KOH, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 0.1 mM DCCD for 30 min. Growth yields on limiting glucose were measured as in Ref. 25.
E. coli Strains and uncA Gene Mutagenesis-For generation of deletions plasmid pBWU13.4 (26) expressing the unc operon (ATP syn-thase) structural genes digested with PstI; then 200-bp fragments containing the first part of the uncA (␣ subunit) gene and its upstream region were amplified by PCR. The reverse primer corresponded to bp ϩ216 -235 of uncA. Forward primers began at bp Ϫ11 and incorporated the codon deletions required. Amplified fragments were phosphorylated at 3Ј-ends and cloned into pUC118 pretreated with alkaline phosphatase and cut with SmaI. pUC118 derivatives with inserts containing uncA sequence were identified by digestion with XhoI and ApaLI. Deletions could be readily recognized by fragment size and/or loss of a natural EcoRI site and/or gain of a new restriction site introduced by the PCR primer and were confirmed routinely by DNA sequencing. It was also routinely confirmed that no PCR-generated errors were present between SphI and XhoI sites. The deletions were transferred to pBWU13.4 on SphI-XhoI fragments, and the resultant plasmids were transformed into E. coli strain DK8 (27) for functional studies. Point mutations in the uncA gene were generated by oligonucleotide-directed mutagenesis following previous methods (14) except that the template was M13mp18 containing the SphI-SalI fragment from pBWU13.4. After mutagenesis the point mutations were moved into plasmid pBWU13.4 as above and expressed in strain DK8.
Synthetic Peptides-Peptides were purchased from United States Biological (immunological grade). All peptides with the single exception of ␣N1-22Cam (see "Results") had free N and C termini. (In intact ATP synthase, the N terminus of ␣ subunit is free Met.) For fluorescence binding assays, the peptides were dissolved at 10 mg/ml in dry Me 2 SO and used for 1 day only. Me 2 SO was rendered anhydrous by use of Molecular Sieves (catalog no. M-6141; Sigma). Concentration of Me 2 SO in the fluorescence binding assays did not exceed 0.75% (v/v).
Fluorescence Binding Assays-Tryptophan fluorescence titrations were carried out as described in Ref. 17, with individual conditions given in the figure and table legends. Excitation was at 295 nm, and emission was measured at 325 nm. The fluorophore was the single natural Trp in the ␦ subunit at residue 28. The buffer was 50 mM HEPES/NaOH, 5 mM MgSO 4 , pH 7.0, at room temperature. A fixed concentration of ␦ subunit (0.6 -3.2 M) was titrated with peptide, and several concentrations of ␦ were used with each peptide. A control titration of buffer alone with peptide was always conducted and sub-3 S. Wilkens, J. Weber, and A. E. Senior, manuscript in preparation.
FIG. 1. The N-terminal residues of ␣ subunit. A, a, the sequence of the first 22 residues of ␣ corresponding to the peptide ␣N1-22 is shown, with the predicted helix at residues ␣6 -18. H, indicates helix; C, random coil. b, the potential helix capping box at residues ␣3-8. Terminology of capping box residues is from Ref. 24. In the capping box signature motif ⌽ is a hydrophobic residue, and X is variable. B, model for binding of N-terminal region of ␣ to ␦ subunit. The NMR structure of the N-terminal domain of ␦ is from Ref. 16. It consists of a six-helix bundle, with helices 1-6 shown in dark blue, light blue, teal, green, yellow, and red, respectively. The N-terminal region of ␣ comprised by the ␣N1-22 peptide is shown in silver, with residues 6 -18 forming an ␣-helix, and fitted into the hydrophobic groove formed by helices 1 (dark blue) and 5 (yellow) of ␦. The actual orientation of ␣N1-22 (i.e. N 3 C) is not specified here. C, helical wheel diagram of residues ␣6 -19. The helix predicted for residues ␣6 -18 is displayed as a helical wheel (with the addition of ␣19). The amphipathic nature of the predicted helix is obvious. Residues on the nonpolar surface are shown in italic type; residues on the polar surface are in normal Roman type.
tracted, although usually negligible. Binding-induced changes in Trp fluorescence were plotted versus peptide concentration, and from the resulting curves, K d values were calculated by nonlinear regression following the methods of Eftink (28), using the equation, where F is the measured fluorescence change at fixed ␦ concentration E o and added peptide concentration L total , n is the stoichiometry of binding, and m is a factor that converts fluorescence data to ligand bound (equals F at saturation divided by [E o ]). Values of n (stoichiometry of binding of peptide to ␦ subunit) were consistent with one peptide molecule binding per ␦ subunit molecule in all cases.

RESULTS
Use of Me 2 SO to Dissolve Peptides-The assay that we use in this work to determine quantitatively the binding of peptides to ␦ subunit is a fluorescence assay that measures the strong (up to 50%) enhancement of fluorescence of the single natural Trp-28 residue in ␦ that occurs upon binding of wild-type ␦ to F 1 or to synthetic peptides (14,15,17). We had previously encountered difficulty due to insolubility of the synthetic peptides in aqueous media at millimolar concentration, as needed to perform fluorescence titrations. For instance, peptide ␣N1-22Cam (the analog of ␣N1-22 but with an amidated C terminus) could not be studied (17). Here the same problem arose with several of the deletion or mutation peptides. We found, however, that the majority of peptides dissolved at up to 8 mM in dry Me 2 SO, and we confirmed that the addition of associated small amounts of Me 2 SO (Յ0.75% v/v) had no effect on the fluorescence titration of ␣N1-22 with ␦. Thus, all peptides were dissolved in Me 2 SO for fluorescence studies.
In fact, it turned out that ␣N1-22Cam behaved identically to ␣N1-22 in fluorescence assays and bound to ␦ with a similar K d value of 150 nM; thus, the presence of a free or blocked carboxyl at the C terminus of the peptide had no effect.
Finding the Optimal Length of Peptide for ␦ Binding by Incrementally Truncating the C Terminus-Originally (17), we identified ␣N1-22 as a suitable peptide that might bind to purified ␦, because secondary structure predictions suggested a continuous helix for residues ␣6 -18, proteolysis experiments showed that removal of residues ␣1-19 prevented ␦ binding to F 1 (20), and in x-ray structures the ␤-barrel domain at the crown of F 1 is seen to commence in ␣ subunits after residue ␣23 (22,23). In the first part of this work, we tested binding of a series of peptides ␣N1-24, ␣N1-22, ␣N1-20, and so on in increments of two-residue truncations, down to ␣N1-12. Fig. 2 shows the results of typical fluorescence titrations measuring binding of the peptides to purified wild-type ␦ subunit. Fig. 3 shows similar titrations, but using ␦V79A mutant purified ␦ subunit instead of wild type. The mutation ␦V79A, occurring at the highly conserved hydrophobic residue ␦79, is known to impair binding of purified ␦ subunit to F 1 (15) and to ␣N1-22 (17). It is located on the surface of helix 5 of the F 1 -binding face of ␦. Therefore, it provides additional evidence as to the specificity of binding. K d values for truncated peptides were calculated from several titration curves carried out as in Figs. 2 and 3, as described under "Experimental Procedures," and are presented in Table I.
It is apparent that the original choice of ␣N1-22 was optimal. Here we found a K d for ␣N1-22 binding to wild-type ␦ of 113 nM, slightly lower than the value of 130 nM reported in our previous work (17), and for binding of ␣N1-22 to V79A ␦ we found a K d of 1.95 M compared with 2.4 M in the previous report. Increasing the peptide length by two residues in ␣N1-24 reduced affinity of binding to wild-type ␦ by 43-fold. Reduction in length by two residues (␣N1-20) also decreased affinity in wild-type ␦, by 11-fold. Further reduction in length of the peptide yielded greater decreases in binding affinity, such that neither ␣N1-14 nor ␣N1-12 peptides showed measurable binding to wild-type ␦ subunit. The same general pattern of reduced affinity was seen for the series of peptides binding to V79A mutant ␦. It may be noted that binding of the parent ␣N1-22 peptide to V79A ␦ is already 17-fold weaker than to wild-type ␦. Overall, the data show that for optimal binding to ␦, a peptide with a length of 22 residues is required, suggesting that residues located along the entire sequence make interactions with ␦. Possibly, the two additional residues in ␣N1-24 cause formation of novel structure in the peptide or simply interfere with binding to ␦ by folding back on the rest of the peptide.
Testing the Idea That the N-terminal Region of the ␣ Subunit Forms an Amphipathic Helix and That Hydrophobic Residues on One Surface Make Interactions with ␦ Subunit- Fig. 1C shows a helical wheel diagram of residues ␣6 -␣19. A series of hydrophobic residues shown in italic type (Leu-11, Ile-8, Phe-19, Ile-12, and Ile-16) clusters on one surface of the helical  wheel. (Arg-15 on this surface may also be viewed as potentially hydrophobic, depending on its orientation). Conversely, the other surface shows mainly polar residues (Ser-9, Lys-13, Thr-6, Glu-10, Gln-14, Glu-7, Gln-18). Given that we had identified helices 1 and 5 of ␦ as forming the F 1 -binding surface (15) and that the NMR structure of ␦ shows these helices forming a hydrophobic groove (16), it seemed likely that some or all of the hydrophobic residues listed above might make direct interaction with ␦. Furthermore, an analysis of sequences of ␣ subunit from multiple species revealed that hydrophobic residues are strongly favored at positions ␣8, ␣11, ␣12, ␣16, and ␣19 (data not shown).
Initially, we tested binding of peptides containing point mutations at position ␣Ile-8 in ␣N1-22, namely ␣N1-22(I8E) and ␣N1-22(I8K), both containing a charged residue in place of Ile. Fig. 4 shows typical fluorescence titrations with wild-type and V79A mutant ␦, and Table I shows the calculated K d values. The substitutions Glu and Lys for Ile at residue 8 in ␣N1-22 peptide caused approximately equal losses of binding affinity with either wild-type or V79A mutant ␦ in the range of 5-8fold. Therefore, it was clear that residue ␣Ile-8 is important for binding of ␣ to ␦. In further work (Table II), we found that when the mutations ␣I8E and ␣I8K were incorporated into the uncA (␣ subunit gene) and expressed in E. coli cells, they caused reduction of growth yield in vivo, indicating loss of ATP synthesis capacity and reduction of ATPase activity and ATPdriven proton pumping in membrane preparations in vitro. The ATPase activity showed somewhat reduced DCCD inhibition. Experiments using NADH to induce proton pumping showed that none of these membrane preparations was leaky to protons (data not shown). In contrast, the mutations ␣I8A and ␣I8W (where hydrophobic residues are substituted for ␣Ile-8) had much lesser effects on growth yield, membrane ATPase, and ATP-driven proton pumping (Table II). The results show that a hydrophobic residue is preferred at position ␣8.
In further work, we studied peptides with Glu substitutions at positions 11, 12, 16, and 19 (␣N1-22(L11E), ␣N1-22(I12E), ␣N1-22(I16E), and ␣N1-22(F19E)). Fig. 5 shows fluorescence titrations with these peptides using wild-type ␦. The calculated K d values are given in Table I. Substitution of Glu at any one of these positions impaired binding impressively. Binding affinity was decreased at least 10-fold in all cases. Interestingly, the ␣N1-22(I16E) peptide brought about a quench of the fluorescence with wild-type ␦ rather than the usual fluorescence enhancement. Of 23 peptides tested in this and other work, this was the only peptide with this property, suggesting that residue ␣Ile-16 may come very close to residue ␦Trp-28 when ␣ is bound to ␦.
We also tested binding of the same series of point mutant peptides to V79A mutant ␦ (titrations not shown), and resultant calculated K d values are listed in Table I. A similar pattern of reduction of binding affinity was apparent. It may be noted that a fluorescence quench was again seen when ␣N1-22(I16E) was titrated with V79A mutant ␦; however, the fluorescence changes were too small to permit confident calculation of K d values.
Overall, this section of work supports the concept that the nonpolar surface of an amphipathic helix at the N terminus of ␣ subunit is important for ␣/␦ binding. It also suggests that residues ␣Ile-8, ␣Leu-11, ␣Ile-12, ␣Ile-16, and ␣Phe-19 on this surface make direct contact with ␦.
Testing the Idea That a Postulated Helix Capping Box at Residues ␣3-8 Is Important for Binding of ␣ Subunit to ␦ Subunit- Fig. 1A showed that within the amino acid sequence at the N terminus of ␣ subunit, residues ␣3-8 conform to a consensus helix capping box motif (24). To test whether the postulated capping box plays a critical role in binding of ␣ to ␦, we introduced point mutations and deletions in this region, both in synthetic peptides and in the uncA gene in cells. Residue ␣Ile-8 corresponds to "N 4 " of the capping box motif, where hydrophobic residues are preferred (24). Evidence described above showed that hydrophobic residues are favored at this position, with the charged Glu and Lys substitutions impairing both peptide/␦ binding and ATP synthase function.  Residue ␣Glu-7 is the most strongly conserved of any of the first 22 residues in ␣ subunits, and one possible reason is that it forms part of the capping box sequence at position N 3 (Fig.  1A), where Glu is strongly preferred (24). In the helical wheel diagram, ␣Glu-7 lies on the polar face (Fig. 1C). As shown in Table II, point mutations Gln and Lys at this position in the uncA gene did yield impairment of ATP synthase function, whereas Ala and Asp substitutions were not very detrimental. However, since Gln should be preferred at the N 3 position of a capping box, whereas Asp is disfavored (24), the results appear ambiguous in relation to the role of residue ␣Glu-7 in a capping box motif. The synthetic peptide ␣N1-22(E7Q) was found to be totally insoluble in aqueous media, Me 2 SO, or other innocuous organic solvents; thus, we could not study its binding properties, and we decided not to pursue other single substitutions in peptides at this position.
Effects of these deletions on growth of cells and function of ATP synthase in membranes were also studied (Table III). It is seen that the ␣⌬3-6 deletion reduced growth yield, showing that ATP synthesis by oxidative phosphorylation was partly impaired in vivo. In contrast, the other deletions in this series appeared to completely prevent ATP synthesis, because they gave growth yields the same as the null control. Growth yield data in Table III were corroborated by analysis of growth on succinate plates, with ␣⌬3-6 showing partial growth and the rest showing no growth. Also shown in Table III are ATPase activities of membrane vesicles prepared from the deletion strains. The ␣⌬3-6 strain showed partial activity (13.5% of wild-type), and the ␣⌬3-8 showed a small residual activity (1% of wild-type, see footnote c of Table III for absolute activity), whereas ␣⌬3-7, ␣⌬3-9, and ␣⌬2-10 showed less. The ATPase activity in ␣⌬3-6 had somewhat reduced sensitivity to inhibition by DCCD (Table III); for the other strains, the activity was low, so DCCD sensitivity was not tested. ATP-driven proton pumping in membrane vesicles was measured by the acridine orange fluorescence quenching technique. The deletion strain ␣⌬3-6 showed only partial quench with ATP, consistent with its reduced ATPase activity, whereas all the other strains showed zero quench with ATP. These data corroborate the peptide binding data in Table I. It may be noted that NADHinduced proton gradient formation in membrane vesicles from all of the deletion strains was the same as for wild type (data not shown).
Two further deletions were studied, namely ␣⌬6 -9, which removed the later part of the helix capping box sequence, and ␣⌬10 -13, which served as a control involving a four-residue deletion located away from the capping box. Fluorescence titrations with peptides containing these deletions were carried out using wild-type and mutant V79A ␦ subunit, and the K d values are listed in Table I. The deletion in ␣N1-22(⌬6 -9) was detrimental, reducing binding affinity by 70-fold in wild-type ␦ and abolishing it in V79A mutant ␦. The deletion in ␣N1-22(⌬10 -13) abolished binding to either ␦ preparation. These deletions were also introduced into the uncA gene, and their  effects on growth yield in cells and ATP synthase function in membrane preparations are described in Table III. The ␣⌬6 -9 deletion had a significant impairing effect on growth yield, membrane ATPase, and ATP-driven proton pumping. DCCD sensitivity of ATPase was much reduced. The ␣⌬10 -13 deletion abolished ATP synthase function. NADH-induced proton gradient formation was normal in both strains (data not shown).
Overall, whereas the data show significant impairment caused by point mutations or deletions within the postulated helix capping box sequence, it is also notable that point mutations at residue ␣Glu-7 were only partly debilitating, and the results were not consistent with those expected from statistical analyses of preferred residues at this position of the capping box sequence (24). Also, deletions ␣⌬3-6 and ␣⌬6 -9 were only partly debilitating, whereas ␣⌬10 -13, away from the capping box sequence, was extremely debilitating.
Deletion of Residue ␣Gln-2-Ogilvie et al. (21) found that when the mutation ␣Q2C was introduced into ATP synthase, the new Cys was able to cross-link to a Cys in ␦ subunit, implying that residue ␣Gln-2 might play a functional role in ␦ binding to ␣. However, as the last line of Table III shows, the deletion ␣⌬2 had no effect on growth yield in vivo. Membrane ATPase activity was 54% of wild type, this reduction in activity apparently not sufficient to affect growth in vivo. (Note that the ATPase activities in membranes of pBWU13.4/DK8 wild type are severalfold higher than in a haploid strain.) The ␣⌬2 strain also showed unimpaired ATP-driven proton pumping in membrane vesicles. Thus, we concluded that residue ␣Gln-2 is not directly involved in binding ␦ to ␣. Taking into account the location of Cys residues in ␦, it may be speculated that the cross-linking seen in Ref. 21 occurs between ␦ and one of the two ATP synthase ␣ subunit N termini to which ␦ is not directly bound but may be spatially close. DISCUSSION In recent studies of the stator subunits (b 2 ␦) of ATP synthase, we established quantitative parameters for binding of ␦ subunit to ␣ subunit (14) and identified (15) the F 1 binding surface on the NMR structure (16) of the N-terminal domain of ␦. We further demonstrated, using a synthetic peptide called ␣N1-22, that the N-terminal 22 residues of ␣ form a major component of the ␦-binding surface on F 1 and that just one of the three ␣ subunits in ATP synthase binds ␦ (17). As noted in the Introduction, residues ␣6 -18 are predicted to be ␣-helical, and residues ␣3-8 appear to form a potential helix capping box (Fig. 1A). This, together with circular dichroism and NMR information that ␣N1-22 does form helical structure, lead to a model of how the N-terminal region of ␣ subunit might bind to ␦, shown in Fig. 1B. In the model, a continuous helix comprising most of the residues ␣1-22 binds to the F 1 -binding face presented by helices 1 and 5 of the N-terminal domain of the ␦ subunit.
In this paper, we have examined sequence characteristics within the N-terminal 22 residues of ␣, looking for determinants of ␦ binding. Our approach was to introduce truncations, point mutations, and internal deletions into the synthetic ␣N1-22 peptide and into the ␣ subunit in ATP synthase, determining K d values for binding of the peptide to ␦ and measuring effects on growth yield, membrane ATPase, and ATPdriven proton pumping activities in cells.
We can first make the general point that there was strong correlation between effects of deletions and point mutations on quantitative impairment of binding of synthetic peptides to purified ␦ subunit in fluorescence titrations and effects on impairment of ATP synthase function in cells and membrane preparations. This not only validates the approach used; it also has ramifications for future work, in that it will be less expensive by far to examine multiple mutations in this region by mutating uncA than by synthesizing many novel peptides. Interesting mutations identified in uncA can later be assayed for quantitative effects on ␦ binding using peptides.
Initially, we determined the optimal length of the peptide for binding to ␦ using a series of peptides of differing length. From these data, it was apparent that ␣N1-22 was optimal, with extension to ␣N1-24 or truncation to ␣N1-20 leading to loss of binding affinity. Further truncation to ␣N1-18 and ␣N1-16 led to further loss of binding affinity, and ␣N1-14 and ␣N1-12 showed no significant binding. One might conclude from this series that residues between positions ␣14 and ␣22 provide most of the binding interactions; however, this conclusion is negated by the facts that deletions ␣⌬2-10 and ␣⌬10 -13, as well as e.g. ␣⌬3-7 and others, also abrogated binding. The summary picture that emerges is that (with the possible exception of residues at the extreme N terminus) the whole sequence ␣N1-22 is required. This region of ␣ subunit is disordered in x-ray structures (7,22,23). A speculative possibility is that this region of the ␣ subunit normally exists in an equilibrium between disordered and helical structure, and only when it collides productively with the binding surface on ␦ does it assume the fully folded helical structure.
A helix formed by the ␣N1-22 peptide will be strongly amphipathic as shown in Fig. 1C. In the N-terminal domain of ␦ subunit, helices 1 and 5 form a hydrophobic groove (16). These a Activity of wild type varied from 4.2 to 6.8 mol/min/mg membrane protein in different experiments. A wild type was always run alongside to allow comparison with mutants.
b Assayed in membrane preparations by measuring quench of acridine orange fluorescence. c Absolute values (mol/min/mg) were as follows: pUC118/DK8, 0.014 Ϯ 0.0058 (S.D., n ϭ 10); ␣⌬3-8, 0.058 Ϯ 0.0034 (S.D., n ϭ 10). d ND, not determined. same helices form the F 1 -binding surface on ␦ (15). It seemed likely, therefore, that the nonpolar side of the ␣N1-22 helix would bind to ␦, and this was made more likely by the fact that the specific residues on this surface (␣Ile-8, ␣Leu-11, ␣Ile-12, ␣Ile-16, and ␣Phe-19) are almost invariably hydrophobic in multiple species. Results obtained using point mutations of these residues to Glu or Lys showed that both binding of peptides to ␦ and ATP synthase function in cells and membranes were impaired. These data indicate that it is the nonpolar surface of the ␣ subunit N-terminal helix that binds to ␦. The importance of the postulated helix capping box (Fig. 1A) was studied using point mutations and deletions. As noted under "Results," point mutations at residue ␣Glu-7 gave ambiguous data in relation to the involvement of this residue in a capping box motif. Effects of mutation of residue ␣Ile-8 were consistent with its playing a role as part of the helix capping motif, but this residue is also part of the nonpolar surface of the helix, probably involved directly in interacting with ␦. Results obtained with deletions placed in and around the capping box motif were similarly equivocal. Therefore, in regard to the potential importance of the helix capping box sequence, we lean toward the view that this sequence per se is not critical for function.
In all cases where membrane ATPase activity was reduced, the membranes were not proton-leaky as judged by NADHinduced proton pumping assay. The lowered activity was due therefore to a reduced level of assembly of the enzyme into the membranes rather than to dislocation of F 1 from F o during membrane isolation. A further conclusion from our work is therefore that correct ␣/␦ interaction is necessary for membrane assembly of ATP synthase. Interestingly, with the ␣⌬6 -9 deletion, the resultant partial membrane ATPase activity was less sensitive to inhibition by DCCD than wild type, suggesting transient dissociation/association between ␣ and ␦ during turnover. In other strains, there was less indication that the mutant complexes, once formed, were "uncoupled." This might be explained as follows. As we showed in Ref. 15, the b subunit has a strong effect to increase the affinity of binding between ␣ and ␦, possibly reducing the K d to Յ3 pM in wild type. Therefore, once the intact complex is formed, the stator could well be strong enough to resist uncoupling and to hold F 1 firmly to F o even in strains where interaction between the N terminus of ␣ and ␦ is impaired.