Characterization of a b 2δ Complex fromEscherichia coli ATP Synthase*

The δ subunit of Escherichia coliATP synthase has been expressed and purified, both as the intact polypeptide and as δ′, a proteolytic fragment composed of residues 1–134. The solution structure of δ′ as a five-helix bundle has been previously reported (Wilkens, S., Dunn, S. D., Chandler, J., Dahlquist, F. W., and Capaldi, R. A. (1997) Nat. Struct. Biol. 4, 198–201). The δ subunit, in conjunction with δ-depleted F1-ATPase, was fully capable of reconstituting energy-dependent fluorescence quenching in membrane vesicles that had been depleted of F1. A complex of δ with the cytoplasmic domain of the b subunit of F0 was demonstrated and characterized by analytical ultracentrifugation using b ST34–156, a form of the b domain lacking aromatic residues. Molecular weight determination by sedimentation equilibrium supported ab 2δ subunit stoichiometry. The sedimentation coefficient of the complex, 2.1 S, indicated a frictional ratio of approximately 2, suggesting that δ and the b dimer are arranged in an end-to-end rather than side-by-side manner. These results indicate the feasibility of the b 2δ complex reaching from the membrane to the membrane-distal portion of the F1 sector, as required if it is to serve as a second stalk.

The ␦ subunit of Escherichia coli ATP synthase has been expressed and purified, both as the intact polypeptide and as ␦, a proteolytic fragment composed of residues 1-134. The solution structure of ␦ as a five-helix bundle has been previously reported ( The proton-translocating ATP synthases couple the generation of ATP to the protonmotive force present across membranes involved in energy transduction (for reviews, see Refs. [1][2][3][4]. These complex enzymes consist of a peripheral F 1 sector, which catalyzes ATP synthesis and hydrolysis, and an integral F 0 sector, which catalyzes movements of protons across the membrane. In the relatively simple ATP synthase of Escherichia coli, F 1 contains five types of subunits in a stoichiometry of ␣ 3 ␤ 3 ␥␦⑀, while F 0 contains three types of subunits in a stoichiometry of ab 2 c 9 -12 . Subunit interactions at the interface of the two sectors are responsible for coupling their catalytic activities. Recent work has strongly indicated that hydrolysis of ATP by F 1 is accompanied by rotation of the ␥ and ⑀ subunits relative to the ␣ 3 ␤ 3 hexameric ring (5)(6)(7)(8)(9)(10)(11), consistent with proposals from Paul Boyer's laboratory (12). The high resolution structure of the mitochondrial F 1 (13) reveals that the N and C termini of ␥ form an antiparallel coiled-coil running up the center of the ␣ 3 ␤ 3 ring; this structure appears to function as an asymmetric spindle, which, by rotating, plays the major role in directing conformational changes at the catalytic sites. In the intact ATP synthase, the rotation of ␥ and ⑀ should be coupled to proton conduction through F 0 . The a and c subunits provide those residues that are essential for proton conduction.
In all systems, ␦ or the analogous mitochondrial protein called oligomycin sensitivity conferral protein (OSCP), 1 is essential for the coupling of the catalytic activities of the two sectors. The ␦ subunit (reviewed in Ref. 14) has no significant effect on steady-state ATP hydrolysis rates by isolated F 1 -ATPase but does alter unisite hydrolysis (15). ␦ binds to F 1 through interactions with the external surface of the N-terminal third of the ␣ subunit (16 -20). In some systems, ␦ alters the proton permeability of F 0 (21). This effect is not seen in E. coli, but here ␦ is essential for the interaction of F 1 and F 0 , implying a link between ␦ and F 0 (22). The physical and functional nature of the ␦-F 0 interaction is currently the subject of intense interest. In recent work, an interaction of ␦ or OSCP with the b subunit of F 0 has been demonstrated (23)(24)(25). Nearest neighbor analysis by chemical cross-linking had not revealed crosslinks between b and ␦ in the E. coli system (26), but they had been reported for the corresponding subunits in the chloroplast (27) and mitochondrial (28) enzymes. The importance of the cytoplasmic domain of b to the F 1 -F 0 interaction has also been demonstrated through proteolysis (29 -31) and direct binding (32) studies.
E. coli ␦ purified following pyridine treatment of F 1 -ATPase was shown to be an elongated monomer (33), but the low yield of the preparation limited the scope of work that could be carried out. The current studies were undertaken to produce recombinant ␦ in quantities appropriate for high resolution structural analysis and to permit the characterization of interactions of ␦ with other subunits in ATP synthase. Here we describe the preparation of recombinant ␦ and a proteolytic fragment called ␦Ј as well as the hydrodynamic analysis of a complex of ␦ with the cytoplasmic domain of the b subunit of F 0 . The solution structure of ␦Ј has been previously reported (34).

EXPERIMENTAL PROCEDURES
Construction of Plasmids-Recombinant DNA procedures were carried out as described by Sambrook et al. (35) using E. coli strain MM294 (36) as the host. Plasmid pHN2 (37), which carries the tac promoter, lacI q , and the unc transcription terminator, was used as the vector. The E. coli uncH DNA sequence encoding the ␦ subunit was amplified, and the translation initiation region was altered using the expression cassette polymerase chain reaction procedure of MacFerrin et al. (38). The upstream primer, CGCGGAATTCTGGAGGATTTTAAAATGTCTGAA-* This work was supported by Medical Research Council of Canada Grant MT-10237. The XL-A analytical ultracentrifuge was obtained with the support of the Academic Development Fund of the University of Western Ontario. 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.
‡ To whom correspondence should be addressed. Tel.: 519-661-3055; Fax: 519-661-3175; E-mail: sdunn@julian.uwo.ca. 1 The abbreviations used are: OSCP, oligomycin sensitivity conferral proteins; b sol and b syn , forms of the cytoplasmic domain of the b subunit containing residues Val 25 -Leu 156 and Tyr 24 -Leu 156 , respectively; b 34 -156 and b 53-156 , forms of the cytoplasmic domain of the b subunit containing residues Glu 34 -Leu 156 and Asp 53 -Leu 156 , respectively, preceded by the leader sequence Ser-Tyr-Trp, assuming removal of the initiating methionine; b ST34 -156 , a form of the cytoplasmic domain of the b subunit containing residues Glu 34 -Leu 156 preceded by a leader sequence of Ser-Thr, assuming removal of the initiating methionine; s obs , the sedimentation coefficient observed under the experimental conditions. TTTATTACGG, contained an EcoRI cloning site, a Shine-Dalgarno sequence, an A/T-rich spacer region, and the first 19 bases of the uncH coding sequence. The downstream primer, GCATCCCGGGTTAAGAC-TGCAAGACGTCTG, contained a SmaI cloning site and the last 20 bases of the uncH coding sequence. The PCR product was cut with EcoRI and SmaI and then ligated into pHN2 that had been digested with the same enzymes, to produce plasmid pJC1. The uncH gene in pJC1 was sequenced and compared with the published sequence (39) to confirm that no mutations had been introduced within the coding region.
A form of the b subunit lacking aromatic residues, b ST34 -156 , was expressed from plasmid pJB3. This plasmid was constructed using polymerase chain reaction to mutate the N-terminal sequence of b syn , encoded by plasmid pDM3 (40). The mutagenic primer, CGCGCATAT-GAGTACTGAAAAGCGCCAAAAAG, included successive NdeI and ScaI sites, which encoded a Met-Ser-Thr leader followed by 16 bases encoding the b sequence beginning with residue Glu 34 ; the M13 universal primer was used as the downstream primer. The product was cut with NdeI and BglII and inserted into pJB2, a derivative of pSD80 (41) containing a modified b syn coding sequence (40), in which an NdeI site had been placed at the start codon. The insert was sequenced to ensure that no undesired mutations had been introduced.
Purification of ␦-Strain MM294/pJC1 was grown at 30°C in LB broth with vigorous shaking. When the cells reached a density that gave an A 600 of 0.8, isopropylthiogalactoside was added to a concentration of 1 mM, and growth was continued for 3-4 h. After harvesting and washing, the cell pellet was suspended in 10 volumes of cold 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 . All subsequent steps were carried out at 4°C. Phenylmethylsulfonyl fluoride was added to a concentration of 1 mM, and the cells were broken by one passage through a French pressure cell at 20,000 p.s.i. Cell debris, ribosomes, and membranes were removed by centrifugation for 1.5 h at 40,000 rpm in a Beckman Ti-60 rotor. Soluble proteins were fractionated by ammonium sulfate precipitation and column chromatography. Material precipitating between 20 and 32% of saturation was collected, redissolved in one-half the original volume, and again brought to 32% saturation with ammonium sulfate. The precipitate was collected, redissolved in 5 ml of buffer containing 20 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 1 mM dithiothreitol (TED buffer), and dialyzed overnight against 1 liter of the same buffer. The sample was applied to a column of DEAE-Sepharose (1.5 ϫ 30 cm) equilibrated with TED buffer, and the proteins were eluted with a linear gradient of 0 -400 mM NaCl in TED buffer. The ␦ subunit eluted as the major peak midway through the gradient. The protein was precipitated by the addition of ammonium sulfate to 40% of saturation, redissolved in a small volume of TED buffer, and applied to a column (1.5 ϫ 90 cm) of Sephadex G-75 Superfine. The column was developed at a flow rate of 3 ml/h. Pure ␦ eluted as the major peak. The yield was between 10 and 20 mg/liter of LB culture.
Analytical Ultracentrifugation-Analytical ultracentrifugation was performed at 20°C using a Beckman model XL-A analytical ultracentrifuge. The buffer contained 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, and 0.5 mM dithiothreitol. In sedimentation velocity experiments, the rotor speed was 60,000 rpm, and scans were taken at 10-min intervals. Data were analyzed using the software provided by Beckman. Observed sedimentation coefficients were calculated using the time derivative (dc/dt) analysis of Stafford (42), fitting the peak of the g(s*) versus s* plot to a Gaussian distribution. Molecular weight determinations using sedimentation equilibrium experiments were carried out using six-sector cells at various rotor speeds. Data sets were individually fitted to determine molecular weights. Partial specific volumes were calculated by standard methods (43).
Other Materials and Methods-b ST34 -156 was purified using techniques similar to those employed for other forms of the b cytoplasmic domain (32,40), namely ammonium sulfate precipitation, DEAE-Sepharose ion exchange chromatography, and size exclusion chromatography on Sephacryl S-200. Fractions used in analytical ultracentrifugation experiments had absorbance at 280 nm of less than 0.005/mg/ ml. The expression and purification of two other forms of the b cytoplasmic domain, b 34 -156 or b 53-156 , have been described previously (40). ␦-Depleted F 1 -ATPase was the generous gift of Drs. S. Wilkens and R. Capaldi of the University of Oregon (Eugene, OR).
Protein was determined using the method of Bradford (44). SDSpolyacrylamide gel electrophoresis was carried out on 15% polyacrylamide gels using the buffer system of Laemmli (45). Reconstitution of energy-dependent quenching of quinacrine fluorescence was measured as described previously (32, 46) using F 1 -ATPase-depleted membranes (0.9 mg of protein) in a volume of 2.7 ml. In one set of experiments, various amounts of ␦ were mixed with 44 pmol of ␦-depleted F 1 -ATPase and 20 pmol of ⑀ and then reconstituted with the depleted membrane vesicles. In a second set of experiments, various amounts of ␦-depleted F 1 -ATPase were mixed with 200 pmol of ␦ and 20 pmol of ⑀ and then reconstituted and assayed as above. After dilution to 2.7 ml, the quenching of quinacrine fluorescence by 1 mM ATP was determined.

RESULTS
Design of pJC1 and Expression of ␦-Previous plasmids carrying uncH, which encodes ␦, and its natural translation initiation region gave poor expression of ␦, even when transcription was from a high level promoter. 2 Two factors that may have contributed to this low expression were corrected in the construction of pJC1. First, evidence that an mRNA secondary structure encompassing the region of uncH from the Shine-Dalgarno to residue 36 of the coding sequence strongly reduces expression has been presented by Pati and co-workers (47). In the construction of plasmid pJC1, the expression cassette polymerase chain reaction strategy of MacFerrin et al. (38) was used to replace C and G residues in the spacer region between the Shine-Dalgarno and the initiation codon with A or T to weaken this secondary structure, making the translation initiation region more accessible to ribosomes. Second, in pJC1 the unc transcriptional terminator was placed downstream of the uncH sequence. We have previously shown that the terminator placed immediately after a coding sequence can increase expression by stabilizing the transcript against exonucleolytic attack (48).
Induction of pJC1 with isopropyl-1-thio-␤-D-galactopyranoside at 37°C led to the synthesis of large amounts of ␦, but essentially all of the expressed subunit was in the form of insoluble inclusion bodies (data not shown). Growth and induction of cells at 30°C led to the expression of good levels of soluble ␦ (see Fig. 1A), although if the induction period was extended beyond 3 h, some of the subunit was insoluble. The soluble portion of ␦ was purified to homogeneity using the standard procedures of ammonium sulfate precipitation and using ion exchange and size exclusion chromatography, as described under "Experimental Procedures." Samples of the preparation at various stages were examined by SDS-polyacrylamide gel electrophoresis (Fig. 1A).
The ␦ subunit was unstable in the crude extract. Coincident with the loss of ␦ was the appearance of a new 14-kDa polypeptide, designated ␦Ј because it could be identified immunochemically as a fragment of ␦ (data not shown). For the purification of intact ␦, proteolysis was minimized by designing the ammonium sulfate fractionation to provide higher purification in exchange for a lower yield. Nevertheless, a portion of ␦ was cleaved during the subsequent dialysis in preparation for DEAE-Sepharose chromatography, and ␦Ј eluted from this column just after ␦ with some overlap, as can be seen in the samples of sequential column fractions in Fig. 1B (also note the material at position of the arrow in the DEAE pool of Fig. 1A). The residual ␦Ј and the higher molecular weight impurities were removed from ␦ by size exclusion chromatography on Sephadex G-75. The ␦Ј fragment has also been purified using Sephadex G-75. 3 The recombinant ␦ was capable of fulfilling the role of the subunit in energy transduction. In the experiment shown in Fig. 2, the indicated amounts of either ␦ or ␦-depleted F 1 -ATPase were mixed with an excess of the complementary component and then reconstituted with membrane vesicles depleted of their endogenous F 1 , and ATP-dependent quinacrine quenching was measured. Essentially the same molar amounts of ␦ or ␦-depleted F 1 -ATPase were required to restore maximal quenching, indicating that the recombinant ␦ is folded properly and fully functional.
Interaction of b and ␦-Preliminary experiments conducted by techniques such as size exclusion chromatography indicated that any interaction between ␦ and b sol (32), the cytoplasmic domain of b, would be relatively weak. 2 To see such an interaction, we carried out sedimentation velocity experiments in the analytical ultracentrifuge. This technique has the advantages that relatively high concentrations may be analyzed, the sample is not significantly diluted during the experiment, and information about the size and shape of the complex can be obtained.
Isolated recombinant ␦ sedimented at 20°C with s obs ϭ 1.67 S (Table I), a value similar to that of the expressed cytoplasmic domain of b, which forms dimers (32,40). These proteins sediment slowly considering their molecular weights, indicating that both, especially b, are extended. A complex of the two proteins should sediment more rapidly because of its greater molecular weight, and the extent of the increase will depend on the overall shape of the complex. Experiments summarized in Table I show that mixtures of the soluble b domains b 34 -156 or b 53-156 with ␦ sedimented slightly more rapidly than either component alone, indicating complex formation to some extent. The complex did not sediment rapidly enough to form a boundary that could be resolved from that of the excess b domains that were present in these experiments, so the observed sedi-mentation coefficients reflect contributions from both complexed and unassociated species. Accordingly, it was difficult to ascertain whether the observed s values reflected substantial formation of a complex that sedimented only marginally faster than either individual component or weaker formation of a complex that would sediment substantially faster.
Characterization of a (b ST34 -156 ) 2 ␦ Complex by Sedimentation Velocity-To eliminate the contribution of the excess free b subunit to the observed s value, we produced a form of b lacking aromatic residues (see "Experimental Procedures" for details). The only tryptophan and tyrosine residues present in b 34 -156 and b 53-156 were added at their N termini during plasmid construction to make them detectable at 280 nm, so removing them would not be expected to affect the interaction with ␦. In the new polypeptide, a leader sequence Met-Ser-Thr was fused to residues Glu 34 to Leu 156 of b; since the N-terminal methionine is ordinarily removed, this polypeptide was called b ST34 -156, to indicate the Ser-Thr leader and to distinguish it from the polypeptide previously denoted b 34 -156 . Lacking aromatic residues, b ST34 -156 is transparent at 280 nm but can still be observed at wavelengths below 250 nm due to absorbance by the peptide bond and certain side chains. Sedimentation equilibrium experiments showed b ST34 -156 to be essentially dimeric Panel B, samples of successive fractions from the DEAE-Sepharose column were analyzed, revealing that ␦Ј eluted at a slightly higher salt concentration (peak at fraction 59) than intact ␦ (peak at fraction 54).

FIG. 2. Reconstitution of ATP-dependent quinacrine quenching by ␦ and ␦-depleted F 1 -ATPase.
Assays were carried out as described under "Experimental Procedures." In one set of experiments, the indicated amounts of ␦ (q) were mixed with excess ␦-depleted F 1 -ATPase and then reconstituted with depleted membrane vesicles. After dilution to 2.7 ml, the quenching of quinacrine fluorescence by 1 mM ATP was determined. In a second set of experiments, the indicated amounts of ␦-depleted F 1 -ATPase (f) were mixed with excess ␦ and then reconstituted and assayed as above. (see observed and inferred molecular weights in Table II) but with a slight tendency to become monomeric at low protein concentrations. In sedimentation velocity analysis, a low concentration of b ST34 -156 (0.5 mg/ml) sedimented at 1.47 S, and the sedimentation coefficient increased to 1.71 at higher concentrations (Fig. 3, OE). Again, this pattern suggests some degree of dissociation into monomers at low protein concentration, so the value observed at the higher concentrations is more likely an accurate reflection of the sedimentation coefficient of the dimer. A set of sedimentation velocity experiments was conducted in which ␦ at a concentration of 0.5 mg/ml was sedimented in the presence of various concentrations of b ST34 -156 (Fig. 3, q) Three or four determinations were made at each concentration, with the S.D. indicated by the error bars. The addition of the b ST34 -156 caused the observed s value of ␦ to rise rapidly from 1.67 to 2.08 S and then to rise more slowly with further increases in b concentration. Calculation of s 20,w by correction of the data for solvent density provides a value of 2.07-2.16, depending on whether one uses the s obs value obtained at 4 mg/ml b ST34 -156 or extrapolates to zero b concentration as shown in Fig. 3. Regardless of the choice, these results demonstrate that the complex sediments only moderately faster than either component alone.
Sedimentation coefficients were also measured for (b ST34 -156) 2 -␦ mixtures in the presence of 5 mM Mg 2ϩ , since this ion is particularly effective in maintaining the F 1 -F 0 interaction. At low concentrations of the b domain, Mg 2ϩ caused a slight reduction in the rate of sedimentation, but there was little effect at b ST34 -156 concentrations of 3 or 4 mg/ml (data not shown).
Sedimentation Equilibrium Analysis of the b ST34 -156 -␦ Interaction-Sedimentation equilibrium experiments were carried out to confirm the subunit stoichiometry of the complex as (b ST34 -156 ) 2 ␦. In this experiment, the starting concentrations of ␦ and b ST34 -156 were 1 and 4 mg/ml, respectively; this represents a 2.8-fold molar excess of b ST34 -156 dimer over ␦. With these high concentrations and the low rotor speeds used (12,000 -18,000 rpm), substantial concentrations of all species were maintained even near the meniscus so that complex formation would be favored. As in the sedimentation velocity analyses, the distribution of ␦ and complexes containing ␦ could be determined at 280 nm without any direct contribution from b ST34 -156 ; representative data are shown in Fig. 4. The apparent concentration gradient of ␦ in the presence of the b domain (OE) was significantly steeper than when the subunit was present alone (q), indicative of formation of a complex with higher molecular weight. The results were analyzed in two ways (Table II). Determination of the molecular weight that gave the best fit for a single component provided a value of just over 40,000, which can be compared with the inferred weight of 46,708 for the (b ST34 -156 ) 2 ␦ complex. Alternatively, the data could be fitted to two noninteracting components, ␦ and a second component of unknown molecular weight. The best fit for the second component was about 45,000, which closely approaches the expected value for the (b ST34 -156 ) 2 ␦ complex.
Summary of Properties of (b ST34 -156 ) 2 ␦-Information about both the shape of the complex and the affinity of the subunits can be obtained from the ultracentrifugation data shown in Fig.  3 and Table II. Given a stoichiometry of two copies of b ST34 -156 to one of ␦, which was indicated by the sedimentation equilibrium results, the sedimentation coefficient allows calculation of a frictional ratio of 2.03-2.13 (Table III). In comparison, the frictional ratio of ␦, a moderately extended protein, was 1.42, while that of the highly extended b ST34 -156 was 1.76. These results indicate that the complex is more extended than either the b dimer or the ␦ subunit alone.
Three of the sets of velocity data shown in Fig. 3 were collected at b ST34 -156 concentrations of 0.5, 0.72, and 1.0 mg/ml, which correspond to molar ratios of b ST34 -156 dimer to ␦ of 0.7, 1.0, and 1.39. Calculation of the K d of the complex based on the fraction of ␦ migrating as a complex gave values of 5-10 M.  a The partial specific volumes of b ST34 -156 and ␦ were calculated to be 0.740 and 0.742 cm 3 per g, respectively. b More details of this experiment can be found in the legend to Fig. 4. c The same data were fitted for either one or two components. For the two-component fit, the first component was taken to be free ␦, and the best fit for the molecular weight of a second, noninteracting component was determined.

DISCUSSION
The ␦ subunit of E. coli could be expressed either as an inclusion body or as a soluble protein, depending on the temperature of growth. Previously, bovine OSCP was expressed as an inclusion body that could be solubilized with guanidine hydrochloride (23,49), while yeast OSCP was expressed as a soluble protein (50). The expressed ␦ subunit of spinach chloroplast F 1 -ATPase was partially soluble, while the solubility of ␦ of Synechocystis depended, like the E. coli subunit, on the temperature of induction (51). Generally, it would appear that ␦ does not fold as efficiently as most proteins of its small size. We have concentrated on purifying the soluble form, obtaining preparations with very high activity for reconstitution of energy coupling. These results imply that essentially all of the ␦ is folded correctly, which was essential for the ultracentrifugal analysis.
The proteolytic sensitivity of the expressed, soluble ␦ subunit after breaking the cells is not surprising, given the ease with which the subunit is cleaved while it is incorporated into F 1 -ATPase (18). In this case, the cleavage turned out to be useful, since ␦Ј was more amenable to structural analysis by NMR than was the entire subunit (34). Intact ␦ was obtained by removing as much contaminating protein as feasible during the early stages of the purification. The cleavage of ␦ and the defined structure of ␦Ј imply the existence of the subunit as a two-domain protein: a well folded N-terminal domain and a less folded, or less stable, C-terminal domain, which is particularly susceptible to proteolytic attack.
Evidence for the interaction of the cytoplasmic domain of b with ␦ or OSCP has appeared recently. Collinson and co-workers (23) found that an insoluble complex was formed from mixtures of bЈ, the cytoplasmic domain of mitochondrial b, and OSCP. Sawada and co-workers (24) have demonstrated the interaction of E. coli b and ␦, primarily through in vivo studies (24). Rodgers et al. (25) found that the addition of cytoplasmic b domain to 15 N-labeled ␦ subunit caused a broadening of NMR signals from the well folded N-terminal domain, which implied slower tumbling of the entire b-␦ complex compared with ␦ alone. The site of the interaction could not be determined, although the effect was much reduced when the ␦ was replaced by ␦Ј. Because of the qualitative nature of these demonstrations, it has been possible to extract little information regarding the structure of the b-␦ complex from these experiments.
In contrast to the mitochondrial system, the complex of E. coli ␦ with the cytoplasmic domain of b was soluble, permitting an analysis of size and shape. Sedimentation velocity and equilibrium experiments using the b ST34 -156 construct showed that they produce an extended complex with a subunit stoichiometry of b 2 ␦. While this stoichiometry might be expected, given the fact that the cytoplasmic domain of b forms dimers in solution (32), it was possible that ␦ might displace one of the subunits of the b dimer to produce a b␦ heterodimer. The molecular weight of such a complex containing the b ST34 -156 construct would be 32,956. In contrast, we determined that the average molecular weight of species containing ␦ in mixtures of that subunit and b ST34 -156 was in excess of 40,000. Since this average reflects contributions from both the complex and ␦ alone, it is more appropriate to fit the data obtained to a two-component system, where one component is assigned the mass of ␦ and the best molecular weight for the second component is obtained through the fitting process. The results of such an analysis gave a molecular weight of 45,000, in good agreement with the value of 46,708 expected for the b 2 ␦ complex.
The interaction of ␦ with the cytoplasmic domain of E. coli b appears to be rapidly reversible and relatively weak, since no interaction could be detected by size exclusion chromatography. The K d calculated from the sedimentation rate of ␦ in the presence of subsaturating levels of b ST34 -156 was in the range of 5-10 M, if b concentrations are expressed as the dimer. However, our results suggest that a small fraction exists as monomer at the low protein concentrations used, so this must be taken as a rough estimate of affinity. Although somewhat weak, the b 2 -␦ interaction nevertheless appears to be essential for the proper binding of E. coli F 1 -ATPase to F 0 , since ␦ is essential for this assembly and it seems unlikely to interact with any other subunit of F 0 .
The shape of the (b ST34 -156 ) 2 ␦ complex was revealed by sedimentation velocity ultracentrifugation to be highly extended, with a frictional ratio of 2.1. The high value of this frictional ratio reveals the complex to be at least as asymmetric as b ST34 -156 , the more asymmetric of its parts. If the b dimer and ␦ were to interact side-by-side, one would expect the complex to be less asymmetric with a lower frictional ratio. Instead, the b dimer and ␦ appear to interact in more of an end-to-end manner, making it a very elongated structure. This result enhances the feasibility that ␦, a subunit classically known to be involved in binding F 1 to the F 0 sector (22), may be located near the top of F 1 -ATPase, where it interacts with N-terminal sequences of the ␣ subunits (16,20). The results thus support a model of F 1 -F 0 structure in which b reaches well up the side of the F 1 sector to make contact with ␦, forming a second stalk (5, 34), which may function as a stator to hold the ␣ 3 ␤ 3 hexamer while Sedimentation equilibrium was carried out as described under "Experimental Procedures." Six-sector cells were loaded with ␦ at a concentration of 1 mg/ml (q) or ␦ at a concentration of 1 mg/ml and b ST34 -156 at a concentration of 4 mg/ml (OE). Centrifugation was carried out at 20°C at rotor speeds of 12,000, 15,000, and 18,000 rpm. Data were collected at intervals to assure that equilibrium had been reached. The rotor speed was then increased to 44,000 rpm for 6 h to determine the base line. Representative data collected after equilibration at 15,000 rpm are shown. the ␥ subunit rotates inside. The weakness of the interaction, however, suggests to us that the complex must be stabilized in the entire enzyme, possibly through interactions of b with other subunits in the F 1 sector.
From data presented here, the first 52 residues from the N terminus of b are not essential for the interaction of b and ␦. The lack of effect of b domains on the NMR signal of ␦Ј (25) and the effect of proteolytic digestion of ␦ on the binding of F 1 to F 0 (18) suggest that b will interact with the C-terminal part of the ␦ subunit. Further analysis of the regions of b and ␦ involved in the interaction is currently under way in this laboratory.