The b and δ Subunits of the Escherichia coli ATP Synthase Interact via Residues in their C-terminal Regions*

An affinity resin for the F1sector of the Escherichia coli ATP synthase was prepared by coupling the b subunit to a solid support through a unique cysteine residue in the N-terminal leader.b 24–156, a form of b lacking the N-terminal transmembrane domain, was able to compete with the affinity resin for binding of F1. Truncated forms ofb 24–156, in which one or four residues from the C terminus were removed, competed poorly for F1binding, suggesting that these residues play an important role inb-F1 interactions. Sedimentation velocity analytical ultracentrifugation revealed that removal of these C-terminal residues from b 24–156 resulted in a disruption of its association with the purified δ subunit of the enzyme. To determine whether these residues interact directly with δ, cysteine residues were introduced at various C-terminal positions ofb and modified with the heterobifunctional cross-linker benzophenone-4-maleimide. Cross-links between b and δ were obtained when the reagent was incorporated at positions 155 and 158 (two residues beyond the normal C terminus) in both the reconstituted b 24–156-F1 complex and the membrane-bound F1F0 complex. CNBr digestion followed by peptide sequencing showed the site of cross-linking within the 177-residue δ subunit to be C-terminal to residue 148, possibly at Met-158. These results indicate that theb and δ subunits interact via their C-terminal regions and that this interaction is instrumental in the binding of the F1 sector to the b subunit of F0.

In the process of oxidative phosphorylation or photophosphorylation, the electron transport chain generates a transmembrane proton gradient. The ATP synthase, or F 1 F 0 -ATPase, allows protons to flow down this electrochemical gradient and uses the energy obtained to synthesize ATP (for reviews, see Refs. [1][2][3][4]. Under appropriate conditions ATP synthase can hydrolyze ATP to pump protons. The enzyme is composed of two sectors; the F 0 sector is membrane-integral and is responsible for proton translocation, and the F 1 sector is attached to the membrane via F 0 and houses the catalytic sites for ATP synthesis. F 1 is easily detached from the membrane and can be purified as a soluble protein with ATPase activity. ATP synthases contain at least eight types of subunits. In the relatively simple enzyme from Escherichia coli, the F 1 sector has the stoichiometry ␣ 3 ␤ 3 ␥ 1 ␦ 1 ⑀ 1 , whereas F 0 is com-posed of three subunits of stoichiometry a 1 b 2 c 9 -12 . The a and c subunits, but not b, contain residues essential for the translocation of protons across the membrane (3). The 156-residue b subunit is believed to span the membrane once at its hydrophobic N terminus, whereas the remainder of the protein is very hydrophilic. b is thought to exist as a dimer in the complex (5)(6)(7), and proteolysis studies have shown that the hydrophilic region of b is required for the association of F 1 with the membrane (7)(8)(9). Removal of two residues from the C terminus of b disrupts normal assembly of the complex (10), as does mutation of Gly-131 to aspartate (11). Thus the b subunit is essential for linking the F 1 and F 0 sectors and likely plays a key role in the coupling of energy from proton translocation to ATP synthesis.
The crystal structure of the mitochondrial F 1 has shown that the ␣ and ␤ subunits alternate in a hexagonal ring structure, with two long ␣-helices from ␥ extending into a hole in the center of the ring (12). Recent evidence strongly suggests that ␥ and probably ⑀ rotate relative to ␣ and ␤ during catalysis (13)(14)(15)(16). Several studies have implied that the ␦ subunit is located near the top of the ␣␤ cluster (17)(18)(19)(20). Like b, ␦ is required for the binding of F 1 to F 0 . Membranes of mutant E. coli strains expressing truncated forms of ␦ showed little ATPase activity (21), suggesting that F 1 cannot bind to the membrane in the absence of ␦. In truncation and mutagenesis studies using the mitochondrial (22,23) and yeast (24) homologues of ␦, called OSCP 1 for oligomycin sensitivity conferring protein, the C-terminal region of OSCP was implicated in F 0 binding.
The only subunit of F 0 able to span the distance from the membrane to the top of F 1 is b. The hydrophilic portion of b is dimeric, highly ␣-helical, has an elongated shape, and binds to F 1 (25). Although chemical cross-linking of E. coli ATP synthase has failed to reveal b-␦ cross-links, such products have been obtained with the chloroplast (26) and mitochondrial (27) enzymes. In the chloroplast work, the cross-link produced by 1-ethyl-3,3-(dimethylaminopropyl)-carbodiimide was mapped to the C-terminal part of b. The site in ␦ was determined to be within the cyanogen bromide fragment encompassing residues Val-1 to Met-165 of the 187-residue polypeptide. Recent studies (28 -31) have demonstrated the interaction of E. coli b and ␦ in the absence of other subunits.
In the present work we examine the interaction between E. coli b and F 1 . An F 1 affinity resin has been generated by linking the hydrophilic portion of b to a solid support. By using binding assays based on this resin, analytical ultracentrifugation, and chemical cross-linking, we have gathered evidence to demonstrate an interaction between residues at the C terminus of b and the C-terminal region of ␦.

EXPERIMENTAL PROCEDURES
Construction of Plasmids-Molecular biological procedures were carried out as described by Sambrook et al. (32). Plasmid pMR2, which was used as an intermediate during the construction of pJB2, was generated from pSD80 (33) by elimination of an NdeI site outside of the multiple cloning region, followed by insertion of the PCR-amplified uncF gene from pSD51 (25) into the EcoRI and HindIII sites. The PCR primer was designed such that the initiating ATG codon is part of an NdeI site.
Plasmid pJB2, encoding the b MERC protein, was constructed as follows. pDM3 (34) was used as the template for PCR, using the mutagenic primer 5Ј-GCGCATATGGAACGTTGCTCGAATTCCCACTACG-3Ј, the 5Ј end contains an NdeI site and the 3Ј end is complementary to the beginning of the gene encoding b 24 -156 . The initial ATG codon is underlined. The second primer for PCR was the M13 forward sequencing primer. The resulting PCR product was inserted into the NdeI and HindIII sites of pMR2 to encode a protein beginning with the amino acid sequence MERCSNSH followed by residues Tyr-24 through Leu-156 of the b sequence. The entire open reading frame was sequenced to ensure that no mutation had arisen during the PCR.
Plasmids expressing other variants of b 24 -156 (b 24 -155 , b 24 -152 , and the mutations D150C, K151C, E155C, and 158C) were based on pDM3 (34). In general, mutagenic PCR primers were designed to encode the desired mutation and were cloned into pDM3 using unique restriction endonuclease sites. Correct incorporation of the desired mutations into all of the above plasmids was determined by DNA sequencing. To introduce mutations into the full-length b protein, appropriate restriction endonucleases were used to cut the desired fragment from the pDM3-based plasmid and transfer it to pDM8 (34). The construction of plasmid pSD114 encoding b 34 -156 has been described previously (34).
Expression and Purification of Proteins-b 24 -156 and b 34 -156 were expressed and purified as described (34). Cysteine-containing and truncated forms of b 24 -156 as well as b MERC were expressed and purified in the same manner as b 24 -156 , except that proteins containing cysteine residues were purified in the presence of 1 mM dithiothreitol (DTT). Plasmids encoding wild type or mutated full-length b were transformed into the uncF E. coli strain KM2 (35). Expression of the proteins and preparation of membranes were performed as described (34). The ␦ subunit was expressed and purified as described previously (31). F 1 was purified by standard methods (36).
Production of the F 1 Affinity Resin-DTT was removed from b MERC by passing it through a Sephadex G-25 size exclusion column equilibrated with 50 mM triethanolamine HCl (TEA-HCl), pH 7.5, 1 mM EDTA. Sulfo-link Coupling gel, obtained from Pierce, was washed with 8 volumes of 50 mM TEA-HCl, pH 7.5, 5 mM EDTA before addition of 2.5 mg of b MERC per ml of resin. The mixture was incubated at room temperature with agitation for 1 h, and then excess buffer was removed and assayed for protein. Based on the amount of protein remaining in the solution, the amount of b MERC bound to the resin was inferred to be 2.2 mg per ml of resin. The resin was then incubated at room temperature in 50 mM Tris-HCl, pH 8.5, 1 mM EDTA containing 50 mM cysteine to block any unreacted thiol-binding sites. After 1 h, excess buffer was removed; the resin was washed with 1 M NaCl, and it was resuspended to its original volume in buffer at pH 7.4. The resin was stored at 4°C. A control resin was prepared by incubating the Sulfo-link Coupling gel with 50 mM Tris-HCl, pH 8.5, 1 mM EDTA containing 50 mM cysteine instead of the b MERC protein.
Assays for F 1 Binding by the Affinity Resin-Two assays for F 1 binding were used. In the "soluble ATPase activity" assay, 31.3 g of F 1 supplemented with 0.8 g of the ␦ subunit was incubated with 20 l of the affinity resin in a final volume of 50 l of buffer containing 50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 2 mM ATP, 10% glycerol, and 0.15 mg/ml BSA (binding buffer). The ␦ subunit was added to the assay to make up for any deficiency of that subunit in the F 1 -ATPase, since a fraction of ␦ is easily lost during purification of the complex (37). After gentle agitation for 1 h at room temperature, the mixture was centrifuged to pellet the resin. The ATPase activity of the supernatant was assayed as described (37).
In the "SDS-PAGE" assay, F 1 , ␦, and the affinity resin in the amounts described above were mixed in 250 l of the binding buffer. After the 1-h incubation and centrifugation, the supernatant solution was discarded, and the pellet was resuspended in 50 l of SDS-PAGE sample buffer containing DTT. After heating at 100°C for 10 min, the mixture was centrifuged to sediment the resin, and 10 l of the supernatant solution were analyzed by SDS-PAGE.
Analytical Ultracentrifugation-Analytical ultracentrifugation was carried out using a Beckman XL-A ultracentrifuge at 20°C. Buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 1 mM EDTA (centrifugation buffer) was used. In sedimentation velocity experiments the rotor speed was 60,000 rpm, and scans were taken at 10-min intervals. During analysis of ␦, 1 mM DTT was added to the centrifugation buffer; the buffer used during experiments in which ␦ and forms of b were mixed contained 0.5 mM DTT. The data were analyzed with the Beckman software using the time derivative method of Stafford (38). The values of Cohn and Edsall (39) were used to calculate partial specific volumes. Sedimentation equilibrium experiments were performed at 20°C and 20,000 rpm using b 24 -152 at concentrations of 0.5, 1.0, or 2.0 mg of protein per ml in centrifugation buffer. Three molecular weight determinations were done at each concentration.
Cross-linking of Membranes-Membranes containing F 1 F 0 complexes including either wild type or mutated b were diluted to 0.25 mg of total protein per ml with buffer containing 50 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , and 10% glycerol. Benzophenone-4-maleimide (BPM) (Molecular Probes, Eugene, OR) dissolved in dimethylformamide (DMF) was added to a final concentration of 1 mM, and the mixture was allowed to stand at room temperature for 30 min. Controls were performed in which only DMF was added to the membranes. The treated membranes were exposed to long wave ultraviolet light from an Ultra-Violet Products model TM-36 transilluminator for 5 min. As a control some BPM-modified samples were placed on the transilluminator but were removed before it was turned on. After illumination, SDS-PAGE sample buffer was added to the samples, which were then heated at 100°C for 5 min and analyzed by SDS-PAGE followed by Western blotting.
Cross-linking of Reconstituted F 1 b-Purified F 1 and wild type or cysteine-containing b 24 -156 were passed separately through 1-ml centrifuge columns (40) containing Bio-Gel P-10 resin (Bio-Rad) equilibrated with 50 mM sodium phosphate, pH 7.5, and 1 mM EDTA. Tris-(2-carboxyethyl)phosphine (Molecular Probes, Eugene, OR) was added to all b 24 -156 samples in a 1.1-fold molar excess to ensure the reduction of any disulfide bonds. BPM in DMF was added in a 5-fold molar excess over b 24 -156 ; in control samples DMF alone was added. After incubation for 15 min at room temperature, an excess of DTT was added to consume unreacted BPM. The b 24 -156 was then mixed with the columncentrifuged F 1 at a molar ratio of 1.2 F 1 per b 24 -156 dimer, in the presence of 5 mM MgCl 2 . Cross-linking and analysis was carried out as described above for the membrane samples.
Cyanogen Bromide Cleavage and Analysis of Cross-linked Products-Cross-linked b 24 -156 E155C-F 1 and b 24 -158 158C-F 1 were analyzed by SDS-PAGE and stained briefly with Coomassie Blue. After destaining for 10 min with 10% acetic acid, gel slices containing bands to be cleaved were excised from the gel and were dried by lyophilization. The dried gel slices were then treated with 2% CNBr in 70% formic acid. After 30 min the slices had swollen back to their original sizes. At this time the excess CNBr solution was removed, and the tubes containing the gel slices were sealed and incubated at 37°C overnight. The pH of the slices was then adjusted by three successive 15-min incubations in 150 l of 1.0 M Tris-HCl, pH 8.0, followed by 15-min incubations in 150 l of 1.0 M Tris-HCl, pH 6.8, 150 l of 67 mM Tris-HCl, pH 6.8, and 150 l of SDS-PAGE sample buffer containing DTT. The slices were placed in the wells of a second SDS-polyacrylamide gel that had been pre-electrophoresed for 1 h at 100 V in the presence of 50 mM Tris-HCl, pH 8.0, 0.1% SDS, containing 0.1 mM sodium thioglycolate to scavenge free radicals in the gel. Electrophoresis was carried out in the presence of 0.1 mM thioglycolate. The gel was either stained with Coomassie Blue or Western blotted onto a polyvinylidene difluoride (PVDF) membrane. After blotting, the membrane was stained briefly with Coomassie Blue and destained with 30% methanol before the bands of interest were excised and analyzed by peptide sequencing.
Other Methods-SDS-PAGE was performed by the method of Laemmli (41) using 15% separating gels. The proteins were stained with Coomassie Brilliant Blue R-250. Protein blotting onto PVDF membranes was carried out using carbonate blot buffer (42). The anti-b monoclonal antibodies 10-1A4 and 10-6D1 were generous gifts of Drs. Karlheinz Altendorf and Gabriele Deckers-Hebestreit of Universität Osnabrü ck, Germany. Anti-␦ polyclonal antiserum was raised against purified ␦ subunit (43), and the anti-␦ antibodies were affinity purified on a column containing immobilized recombinant ␦ (31). Antibodies were labeled with 125 I by the IODO-GEN method (44). Protein concentrations were determined by the method of Bradford (45) or Lowry et al. (46). Peptide sequencing was performed at the Laboratory for Macro-molecular Structure at Purdue University (West Lafayette, IN) using an Applied Biosystems 470A sequencer.

F 1 Affinity
Resin-To characterize better the binding between the F 1 sector and the hydrophilic portion of b, we coupled this region of b to a solid matrix to form an affinity resin for F 1 . The Sulfo-link coupling gel from Pierce, to which proteins can be coupled specifically via thiol groups, was chosen as a matrix. Since the hydrophilic region of b has no cysteine residues, the site of coupling to the resin can be specified by site-directed mutagenesis of b. Because the C-terminal region was thought most likely to be involved in b-F 1 contacts, we introduced a cysteine residue near the N terminus of the hydrophilic region. A construct encoding the polypeptide sequence MERCSN-SHY 24 -L 156 (b MERC ) was produced as described under "Experimental Procedures." The first four amino acids of this sequence were taken from the E. coli enzyme 3-methyladenine-DNA glycosylase I, because of the polar nature of the sequence and the high expression of this enzyme in a recombinant system (47).
The purified b MERC was coupled to the Sulfo-link resin as described under "Experimental Procedures," at a final concentration of 2.2 mg per ml of resin. Upon incubation of the modified resin with the F 1 complex followed by centrifugation, a significant amount of F 1 sedimented with the resin, as determined by SDS-PAGE (Fig. 1A, first lane). Only a small amount of F 1 co-sedimented with the resin that had been modified with cysteine (Fig. 1A, last lane). The F 1 present in this pellet was probably not bound specifically to the resin but rather was trapped between and within the resin particles. As a control for the volume of trapped liquid, BSA was included in all incubations; similar amounts of BSA were observed to co-sediment with both resins under all conditions used (Fig. 1). Thus the b MERC -modified resin is able to bind F 1 specifically. The faint band migrating at the position of b 24 -156 in the first lane represents a trace of b MERC eluted from the resin during the incubation in SDS sample buffer. Most likely this arose from instances in which only one subunit of the dimer became covalently coupled to the resin.
Competition for F 1 Binding by the SDS-PAGE Assay-To test the ability of the hydrophilic portion of b to compete with the resin for binding to F 1 , the resin was incubated with F 1 in the presence of increasing amounts of b 24 -156 (formerly known as b syn ; Ref. 33). After centrifugation and analysis of the pellet by SDS-PAGE, the amount of F 1 bound to the resin was observed to decrease as the concentration of b 24 -156 increased, until essentially no F 1 was bound by the resin at 6.5 M of b 24 -156 dimer (Fig. 1A). Note that the amounts of soluble b 24 -156 trapped within the pelleted resin provide an internal representation of the amount of competitor added. These experiments demonstrate that b 24 -156 is able to compete with the b MERCmodified resin for binding to F 1 , establishing a simple competition assay for determining the relative affinity of any mutant form of b for F 1 .
Such experiments were carried out using b 24 -155 and b 24 -152 , which lack one and four residues from the C terminus, respectively. It was found that these forms of b competed very poorly, relative to b 24 -156 , with the b MERC -modified resin (Fig. 1B). At a dimer concentration of 10 M, b 24 -155 showed a small amount of competition, whereas at the same concentration b 24 -152 showed no detectable competition by the SDS-PAGE assay. It is evident, however, that very weak competition is difficult to detect in this manner, as it requires seeing a small difference in band intensity.
Competition for F 1 Binding by the Soluble ATPase Activity Assay-To determine weak competition more reliably and in a quantifiable way, the assay was modified such that soluble ATPase activity, rather than bound ATPase protein, was measured. Under the conditions used in these assays, more than 90% of the added enzyme was bound by the resin. Competition for F 1 binding by b 24 -156 , b 24 -155 , and b 24 -152 was determined by the increase in ATPase activity remaining in the supernatant solution when these forms of b were added to the incubations. As expected, the amount of F 1 in the supernatant solution increased sharply with increasing concentration of b 24 -156 , whereas the ability of b 24 -155 and b 24 -152 to compete with the b MERC -modified resin was far weaker, although still detectable (Fig. 2). These results show that the C-terminal residues of b are essential for its proper interaction with F 1 -ATPase. The preparations of soluble b were tested directly for ATP hydrolysis activity to make certain that trace contamination with an enzyme such as alkaline phosphatase could not account for the increase in soluble ATPase activity. In no case could such a contaminant account for more than 2% of the observed soluble activity.
Sedimentation Velocity Analysis-Recent ultracentrifugation results from this laboratory have provided evidence for the formation of an elongated complex by two molecules of the hydrophilic region of b and one molecule of the ␦ subunit (31). We performed further centrifugation experiments to determine whether the reduced binding of the C-terminal truncations of b to F 1 could be due to loss of contacts with ␦. The isolated b 24 -156 and ␦ subunits showed sedimentation coefficients (Table I)  ing of b to ␦ is significantly disrupted by removal of one or four residues from the C terminus. Thus the essential nature of these residues for binding F 1 derives from their role in binding ␦. The molecular mass of b 24 -152 was determined by sedimentation equilibrium, as described under "Experimental Procedures," to be 28,100 Ϯ 1000 Da confirming that the dimeric nature of b was unaffected by the C-terminal truncation. Unlike b 24 -156 (34), no aggregation of b 24 -152 was observed at high concentrations.
The protein b 34 -156 , which lacks 10 residues of the b sequence relative to b 24 -156 , was found to have a sedimentation coefficient comparable with b 24 -156 (Table I). Interestingly, the complex formed when b 34 -156 was mixed with ␦ had a significantly lower sedimentation coefficient than that of b 24 -156 and ␦. This lower sedimentation coefficient was much more in line with the value of 2.07-2.16 recently reported for the b 2 ␦ complex found with the b ST34 -156 construct that also contained residues 34 -156 of b (31). It seems unlikely that removal of 10 residues near the N terminus of b would have a significant effect on its binding to ␦, and the molecular masses of b 24 -156 and b 34 -156 differ by only about 1.5 kDa. It therefore appears that the presence of residues 24 -33, which are mostly hydrophobic, in the soluble b alters the conformation of the b 2 ␦ complex to make it less asymmetric, thereby increasing its sedimentation coefficient.
It is also of note that the sedimentation coefficients of b 24 -153 and b 24 -155 alone are substantially lower than that of b 24 -156 , which is only a few residues greater in length. This finding implies that the truncated forms have greater frictional coefficients than b 24 -156 , probably due to a conformational change in the C-terminal region upon deletion of the C-terminal residues. Such a conformational change could be responsible for the reduced binding of these forms of b to ␦. To distinguish between direct and conformational roles of the C terminus of b in binding ␦, we attempted to generate chemical cross-links to ␦ from sites within this region.
Cross-linking of b and ␦-Individual cysteine residues were introduced into full-length b at positions 150, 151, and 155. A fourth construct was made that encoded a protein, referred to as b158C, having two residues, glycine and cysteine, attached to the C terminus of b. It was anticipated that the glycine would provide conformational flexibility to the C-terminal cysteine, increasing the likelihood of obtaining a cross-link. Plasmids bearing these mutated forms of b were all able to complement the uncF strain KM2 for growth on minimal media with succinate as the sole carbon/energy source, indicating that the b-F 1 interaction was not disrupted in the mutants.
Membrane preparations bearing F 1 F 0 complexes containing the mutated b subunits were incubated with the photoreactive cross-linker benzophenone-4-maleimide (BPM) and then exposed to ultraviolet light. The samples were analyzed by Western blotting, using 125 I-radiolabeled monoclonal antibodies raised against b as probes. No cross-linking was observed with either bD150C or bK151C (data not shown). However, crosslinked products of about the same size were observed with both bE155C and b158C (Fig. 3). The new bands were approximately the size expected for a b-␦ cross-link and showed reac- tivity with anti-␦ polyclonal antibodies (Fig. 3), indicating that cross-links had been formed between b and ␦. Membranes containing cross-linked F 1 F 0 showed no apparent loss of activity compared with control membranes treated with either BPM or UV light (data not shown).
To characterize further the cross-linked products, the E155C and 158C mutations were incorporated into b 24 -156 . After modification of each of these proteins with BPM, reconstitution with F 1 , and exposure to ultraviolet light, new bands of an appropriate size were observed on SDS-PAGE (Figs. 4A and 5A). The new bands were recognized by antibodies directed against b and ␦ (Figs. 4B and 5B), confirming the identity of a b 24 -156 -␦ cross-link. Exposure of the BPM-modified E155C and 158C proteins to UV light in the absence of F 1 caused an apparent reduction in the total amount of protein on the stained gels (Figs. 4A and 5A). Western blotting of similar samples revealed a series of dimeric and higher order aggregates (not shown), which were probably not visible on the stained gels because of their heterogeneous nature.
Cross-linking of the soluble forms of b gave rise to a second cross-link in each case that had a slightly greater mobility than b 24 -156 on SDS-PAGE (Figs. 4A and 5A). These cross-links were recognized by anti-b antibodies (Figs. 4B and 5B), suggesting that in each case an internal cross-link had been formed in b 24 -156 . This internal cross-link was not formed to the same extent in the absence of F 1 .
Some cross-linked products of higher apparent molecular weight were observed with both the E155C and the 158C mutations (Figs. 4A and 5A). In each case, one of these cross-links was recognized by the anti-␦ antibodies, whereas the other was not (Figs. 4B and 5B). In the absence of other data we have tentatively identified one of these bands as a (b 24 -156 ) 2 -␦ crosslink. The other high molecular weight cross-linked product was recognized by an anti-␣ antibody (data not shown) and therefore could correspond to a b 24 -156 -␣ cross-link.
Peptide Analysis of b 24 -156 -␦ Cross-links-To identify the region of ␦ involved in the cross-links to b 24 -156 , a slice containing the cross-linked product from each cysteine mutation was cut from an SDS-polyacrylamide gel and treated with cyanogen bromide as outlined under "Experimental Procedures." CNBr cleavage of b 24 -156 should give rise to fragments of 2, 12, and 126 residues, with the large C-terminal fragment (residues Ala-31 to Leu-156) containing the site of the cross-link. The ␦ subunit should give rise to fragments between 10 and 49 residues in length. The difference between b 24 -156 and its largest CNBr fragment is readily apparent on SDS-PAGE (Fig. 6). CNBr cleavage of each cross-linked product gave rise to a predominant fragment that is markedly larger than the 126residue fragment derived from b 24 -156 alone (marked by an arrow in Fig. 6).  24 -156 containing the E155C mutation, as well as wild type b 24 -156 , was modified with BPM and reconstituted with purified F 1 . After exposure to UV light, the samples were analyzed by SDS-PAGE (A) and Western blotting (B), probing with 125 I-labeled antibodies raised against either b or ␦. Controls were performed in which the reconstituted b 24 -156 -F 1 complex was exposed only to BPM or to UV light, and in which the b 24 -156 , with or without the E155C mutation, was treated in the absence of F 1 . An experiment identical to that described in the legend to Fig. 4 was performed, except that the 158C mutation was used instead of the E155C mutation.
After blotting a gel containing the cross-linked b 24 -156 -␦ fragments to a PVDF membrane, the major CNBr product from each cross-link was cut from the membrane and analyzed by Edman degradation. In most cycles, both cross-links gave rise to two residues, corresponding to sequences from b and ␦ (Table  II). As expected, one set of amino acids was the sequence of b beginning after Met-30. The other set represented the sequence of ␦ starting after Met-148. Interestingly, the ␦ fragment sequence in both the E155C and the 158C cross-links continued past residue Met-158, indicating that ␦ had not been cleaved at this position by the CNBr. Met-158 was not observed in the appropriate cycle of the Edman degradation of either of the cross-links. DISCUSSION The results provided demonstrate that the residues at the C terminus of b are essential for proper interaction with F 1 , that the interaction of b with ␦ is weakened when these residues are lacking, and that the C-terminal region of b is proximal to that of ␦. Together these subunits are believed to form a "second stalk" reaching from the membrane to near the top of the F 1 sector, which may function to hold the ␣ 3 ␤ 3 subunits stationary during rotation of ␥ and ⑀ driven by proton flow through F 0 (48,49). Earlier truncation and mutagenesis studies showed that the C terminus of b was essential for proper assembly of ATP synthase (10) and implied that the C terminus of ␦ played a role in interaction of F 1 with the F 0 sector (21)(22)(23)(24).
Our present studies confirm and extend this earlier work by providing additional information about the b␦ interaction. The major site of cross-linking of the E155C and 158C proteins to ␦ was shown to be C-terminal to Met-148. It is interesting that the cross-linked ␦ subunit was not cleaved at Met-158 by cyanogen bromide treatment and that no signal corresponding to this residue was observed during the appropriate cycle of peptide sequencing. The most straightforward interpretation of these results is that the major site of benzophenone linkage was through the side chain of Met-158. The benzophenone moiety may have formed a cross-link at C ␥ of this methionine to create a tertiary carbon center that would be relatively unreactive to cleavage by CNBr. It is notable that methylene groups adjacent to nitrogen or sulfur are particularly reactive to benzophenone (50).
Modeling of the C-terminal region of b as an ␣-helix reveals an amphipathic structure (Fig.7). The cross-linking results imply that the hydrophobic face of this helix on one of the b subunits interacts with ␦. We suspect that the hydrophobic face of the second b subunit is in contact with another region of b. Removal of one or more hydrophobic residues might disrupt this interaction, causing a significant conformational change in the soluble b protein. Such a conformational change to a more asymmetric shape would explain why the sedimentation coefficients of b 24 -155 and b 24 -152 were markedly lower than that observed for b 24 -156 (Table I). In the absence of ␦, the hydrophobic face of the first b subunit would not have its normal partner and might interact nonspecifically with the similarly exposed hydrophobic face of another b 24 -156 dimer. This would explain why the slight aggregation of b 24 -156 observed during sedimentation equilibrium centrifugation at high concentrations (34) is not seen with b 24 -152 .
Hydrophobic residues are conserved at positions corresponding to E. coli residues 153 and 156 in the b subunit of many organisms (51). It is tempting to suggest that the importance of these hydrophobic residues in the b-␦ interaction may provide the explanation for the observation that low ionic strength disrupts the binding of F 1 to F 0 . However, other regions of b must also be involved in b-F 1 interactions, since b 24 -152 was able to compete to a minor extent with the affinity resin for binding to F 1 (Fig. 2). In this regard it is also noteworthy that some bacteria have b subunits that lack the hydrophobic residues at the C terminus. For example, the b subunit of the thermophile PS3 ends at the residue corresponding to position 148 in E. coli b (52).
The difference in sedimentation coefficients observed between the (b 24 -156 ) 2 ␦ and the (b 34 -156 ) 2 ␦ complexes (Table I) was surprising. One possible explanation is that flexibility in the N-terminal region of the soluble b construct allows the hydrophobic residues Y 24 VWPPLMAAI 33 , present in b 24 -156 but absent in b 34 -156 , to loop back and interact with ␦. Such an arrangement would be more compact, and the complex would therefore sediment faster than if the N-terminal helices were in a more extended conformation. Because the N termini of the b subunits are anchored in the membrane in ATP synthase, it seems unlikely that residues 24 -33 of b normally interact with ␦ in the intact complex.
Takeyama and co-workers (10) showed that one aspect of the defective ATP synthase assembly caused by deletion of residues from the C terminus of b was the failure of F 0 to form a functional proton pore in vivo. In subsequent work, Brusilow and co-workers (53,54) demonstrated that the ␦ subunit was required for the formation of the proton pore from the cloned F 0 subunits. Here we have shown that the same region of b implicated in the F 0 assembly process is essential for binding to ␦, strengthening the argument that the b-␦ interaction is critical for the assembly of functional F 0 . At present, however, we have

AGVIIRAGD-VIDGSVRGRLERL
a Inferred amino acid sequences taken from Ref. 56. The position of the first residue of the inferred sequence is indicated. A dash in the sequencing data indicates a residue that could not be clearly assigned.
b When reading the sequence of the E155C cross-link, assignment of residues Ile-152, Ile-153, and Ile-160 of ␦ was difficult due to the small amount of material present and the possibility of carry-over from previous cycles. FIG. 7. Depiction of residues 146 to 156 of b in a helical wheel. no evidence of how a signal arising from the interaction may be transmitted from the C-terminal end of b to the membrane, where interaction with the other F 0 subunits would occur.
The solution structure of residues 1-105 of ␦Ј, a proteolytic fragment consisting of residues 1-134 of ␦, has been solved by NMR spectroscopy, but the structure of the entire 177-residue subunit could not be determined (49). Although the C-terminal regions of both proteins are predicted to be largely ␣-helical, in the absence of concrete structural information from either region it is difficult to propose specific interactions between amino acid residues in b and ␦. The proposed site of crosslinking, Met-158, lies before a predicted C-terminal helix encompassing residues ␦167-175 (55).
Our current results demonstrate that b and ␦ interact via their C-terminal regions, and a recent study from our laboratory has shown that the b 2 ␦ complex is extended enough to span the distance from the membrane to the N-terminal domain of ␣ (31). Thus our work is consistent with the hypothesis that the b and ␦ subunits form a stator to prevent ␣ 3 ␤ 3 from moving relative to ␥. Since the stator must resist an appreciable torque (15), and since the K d of the b-␦ interaction is relatively high (5-10 M; Ref. 31), it is likely that the interaction of these subunits is stabilized by the presence of the other parts of the ATP synthase complex. A further possibility is that b makes direct contact with other subunits, such as ␣ or ␤, to stabilize the b-F 1 interaction. The newly developed competition assay for b-F 1 binding provides a simpler, more quantitative, and more sensitive way of detecting minor changes in b-F 1 affinity compared with earlier methods. We are currently using this assay to define other residues of b that affect b-F 1 interactions.