Interaction of the δ and b Subunits Contributes to F1 and F0 Interaction in the Escherichia coli F1F0-ATPase*

Interactions of the F1F0-ATPase subunits between the cytoplasmic domain of the b subunit (residues 26–156, bcyt) and other membrane peripheral subunits including α, β, γ, δ, ε, and putative cytoplasmic domains of the a subunit were analyzed with the yeast two-hybrid system and in vitroreconstitution of ATPase from the purified subunits as well. Only the combination of bcyt fused to the activation domain of the yeast GAL-4, and δ subunit fused to the DNA binding domain resulted in the strong expression of the β-galactosidase reporter gene, suggesting a specific interaction of these subunits. Expression of bcyt fused to glutathione S-transferase (GST) together with the δ subunit in Escherichia coli resulted in the overproduction of these subunits in soluble form, whereas expression of the GST-bcyt fusion alone had no such effect, indicating that GST-bcyt was protected by the co-expressed δ subunit from proteolytic attack in the cell. These results indicated that the membrane peripheral domain of b subunit stably interacted with the δ subunit in the cell. The affinity purified GST-bcyt did not contain significant amounts of δ, suggesting that the interaction of these subunits was relatively weak. Binding of these subunits observed in a direct binding assay significantly supported the capability of binding of the subunits. The ATPase activity was reconstituted from the purified bcyttogether with α, β, γ, δ, and ε, or with the same combination except ε. Specific elution of the ATPase activity from glutathione affinity column with the addition of glutathione after reconstitution demonstrated that the reconstituted ATPase formed a complex. The result indicated that interaction of b and δ was stabilized by F1 subunits other than ε and also suggested that b-δ interaction was important for F1-F0interaction.

dient across the membrane formed by the respiratory chain in mitochondria and bacteria or photosynthetic electron transport in chloroplasts. In the reverse reaction, this enzyme pumps protons from the inside to the outside of bacterial cells and mitochondria coupled with hydrolysis of ATP.
Based on the three-dimensional structure of ␣␤␥ complex of the bovine F 1 -ATPase (8) and the results of kinetic analyses of nucleotide binding and catalysis (9), rotation of ␣ 3 ␤ 3 around ␥ was proposed to occur during catalysis (8). Data to support this hypothesis have been reported (10 -12), and recently direct visualization of this rotation of the ␥ subunit within the ␣␤␥ complex has been performed for thermophilic bacteria (13). Although evidence of the rotation within the F 1 complex has been thus presented, its structural basis, especially in terms of the connecting portion between the ␣␤␥ complex and F 0 is not well understood (14). The ␦ and ⑀ subunits were shown to be important for this interaction (1,2,5,6,14). The atomic structure of the ⑀ subunit and a portion of the ␦ subunit was resolved by NMR analysis for E. coli (14 -16). The ⑀ subunit interacts with the DELSEED region of the ␣ and ␤ subunits (14,19) and also binds to the cytoplasmic region of the c subunit (17). However, the topological arrangement of the ␦ subunit in the F 1 F 0 complex is less well known than that of the ⑀ subunit. In an early study of the biochemical properties of the purified ␦ subunit from E. coli (18), an elongated structure was estimated. This subunit is also known to open sealed proton channel activity of F 0 during enzyme biosynthesis (20). Based on the results of hydropathy analysis of the primary structure of the b subunit of F 0 (2,3), the majority of this subunit is estimated to be hydrophilic and possibly extruded into the cytoplasm, whereas the small amino-terminal portion is essential for its integration into membranes (22)(23)(24). The region between residues 25 and 146 of this subunit was overproduced as a soluble form and was shown to be capable of binding to the F 1 complex (24). Although these previous studies strongly suggested interactions between the b and ␦ or ⑀ subunits, direct evidence of b-␦ or b-⑀ interaction has not been reported. The a subunit of F 0 has been shown by genetic analysis to be required for F 1 -F 0 binding (21) and its interacting partners within F 1 subunits have been also determined by chemical cross-linking experiments (25). However, the precise regions involved in the binding of this subunit to a specific F 1 subunit have not been elucidated.
Here, we have found a specific interaction between the b and ␦ subunits with two approaches; a genetic approach with the yeast two-hybrid system (26,27) and a biochemical approach with overproduction, purification, and in vitro binding of the subunits (28). The results directly indicate that the interaction of b and ␦ has an important role for F 1 and F 0 interaction.

MATERIALS AND METHODS
Bacterial Strains and Culture Conditions-E. coli strains BL21 and JM109 were used (28) for overproduction of various peptides and genetic manipulations including preparation of various plasmids, respectively. E. coli cells were cultured in a minimal medium (Tanaka me-dium) (29) supplemented with glucose or glycerol at 37°C. For the selection of transformants with plasmids, appropriate antibiotics were added to these media.
Construction of Expression Plasmids-For the two-hybrid system, plasmids pGAD424 carrying the activation domain of GAL4, and pGBT9 carrying the binding domain of GAL4 (27) were used. A DNA fragment corresponding to the cytoplasmic domain (residues 26 -156) of the b subunit (b cyt ) was amplified by the polymerase chain reaction (30) with primer oligonucleotides (Table I), bT26-F (forward primer) and bT156-R (reverse primer), and genomic DNA from the wild-type DNA or plasmid pKM02 carrying the entire unc operon. The nucleotide sequence and deduced primary structure of the b subunit were based on the published results (31)(32)(33). The amplified DNA was digested with EcoRI and BamHI, and the fragment subsequently recovered from agarose gels after electrophoresis was ligated into EcoRI and BamHI digested pGAD424 or pGBT9. The chimeric plasmids thus constructed were named pGAD-b cyt and pGBT-b cyt , respectively. To prepare b cyt peptides, DNA corresponding to the same region (residues 26 -156) or this region plus the ␦ subunit gene was amplified by polymerase chain reaction with primers bG26-F and bT156-R or bG26-F and DER-C1. The amplified DNA was integrated into the BamHI site of the expression vector pGEX-2T (28) and resultant chimeric plasmids were named pGEX-b cyt and pGEX-b cyt and ␦. Following essentially the same steps, portions of the a subunit between residues 60 and 100, and 160 and 200 were amplified by polymerase chain reaction with the oligonucleotide primers aT60-F and aT100-R, and aT160-F and aT200-R, respectively, and the wild-type sequences of the subunit on plasmid pKM02 as a template. The amplified DNAs were digested with the appropriate restriction enzymes shown in Fig. 1 and ligated to pGAD424 or pGBT9. The resultant chimeric plasmids were named pGAD-a 60 -100 , pGBT-a 60 -100, pGAD-a 160 -200 , and pGBT-a 160 -200 . The nucleic acid and deduced amino acid sequence of the a subunit were cited from (34). Polymerase chain reaction was performed using the conditions as described previously (35).
Yeast Two-hybrid Assay of Subunit Interaction-The chimeric plasmids of pGAD424 and pGBT9 derivatives constructed as described above or as previously reported for F 1 subunits (26) were introduced into yeast SFY526 carrying the ␤-galactosidase reporter gene. Expression of ␤-galactosidase was measured as described previously (26). The optical density at 420 nm of o-nitrophenol released from the substrate o-nitrophenyl ␤-galactoside was normalized by the cell density of yeast measured photometrically at 600 nm and was expressed in Miller units (26).
Overexpression of GST 1 Fusion Subunit and Its Purification-The expression plasmid pGEX-b cyt was introduced into E. coli BL21 (28). Transformed E. coli was cultured in 500 ml of M9ZB medium supplemented with 0.2% glucose with vigorous shaking at 37°C. At 0.6 A 600 , isopropyl-1-thio-␤-D-galactopyranoside was added to the culture (0.4 mM) and then the incubation was continued for another 2 h. Cells were harvested, washed, suspended in 6 ml of phosphate-buffered saline with 1% Triton X-100, and disrupted by sonication. The disrupted materials were subjected to low (10,000 Ϫ g for 10 min) and high speed (100,000 Ϫ g for 60 min) centrifugation to fractionate proteins into soluble (supernatant) and membrane (precipitate) fractions as described previously (35). The amounts of protein recovered were 92.4 mg and 14.3 mg for the supernatant and membrane fractions, respectively, after high speed centrifugation. The supernatant fraction was subjected to glutathione-Sepharose (Pharmacia Biotech Inc.) (2.0 ml) column chromatography as described previously (28). The fraction eluted with 10 mM glutathione contained 26.6 mg of protein in which the major band stained by Coomassie Brilliant Blue was shown to be GST-b cyt .
Affinity Chromatography of the Reconstituted ATPase-After dialysis, the reconstituted materials that contained the subunits in 3.0 ml were applied to glutathione-Sepharose (0.3 ml). After washing the column with 7.0 ml of reconstitution buffer, the ATPase was eluted with 10 mM glutathione in reconstitution buffer. Aliquots of the eluted materials were used for the ATPase assay (37), protein measurement by the published procedure (29,30), and SDS-gel electrophoretic analysis as described previously (21).
Binding Assay of Subunits on Membrane Filters-Aliquots of 2 or 4 g of b cyt or GST were fixed on Millipore GVHP filters (39), which were activated with methanol and equilibrated with buffer (100 mM Tris-HCl, pH 7.4, 190 mM glycine, 5% MeOH) prior to fixation. After blocking nonspecific binding of proteins by soaking the filters in skimmed milk, the filter with b cyt and GST was washed with reconstitution buffer and then soaked again in 2 ml of reconstitution buffer containing 100 g of the ␦ subunit for 8 h at 26°C for binding. Then, the filter was washed with reconstitution buffer to remove unbound ␦ subunits and nonspecific protein binding was blocked with 10% skimmed milk solution. Bound ␦ was detected by anti-␦ mouse IgG raised against the purified E. coli ␦ and visualized with an ABC Vectastain kit as described previously (35). DNA Modification and Sequencing-Preparation of plasmids, digestion and ligation of the DNA fragments, and other techniques related to handling DNA were performed according to the published procedures (40). Nucleotide sequences of the amplified DNA in the expression plasmids were verified by the dideoxy method with appropriate primers and 35 S-␣-deoxy CTP (37 TBq/mmol, Amersham) (41) or with a DNA sequencer (Pharmacia, Alfexpress DNA sequencer).
Immunological Detection of the ATPase Subunits-Aliquots of eluted materials after affinity chromatography with glutathione-Sepharose were subjected to SDS-polyacrylamide gel electrophoresis (12.5% acrylamide) (29) and separated peptides were blotted onto GVHP filters (Millipore) (35). The membrane filters were soaked in a solution of polyor monoclonal antibodies raised against the purified E. coli ␣, ␤,␥, ␦, or GST, and the reacted bands were visualized with an ABC Vectastain kit as described previously (35).
Reagents and Enzymes-Restriction endonucleases, T4 DNA ligase, Tth and Pfu DNA polymerase, and T7 DNA polymerase were purchased from Bethesda Research Labs, Toyobo Co., New England Biolabs, and Takara Co. Oligonucleotides used as primers were synthesized by DNAgency (Malvern, PA). Other materials were of the highest grade commercially available.

RESULTS
Construction of F 1 and F 0 Subunit Expression Plasmids for the Yeast Two-hybrid System and Overproduction of Cytoplas- 1 The abbreviation used is: GST, glutathione S-transferase.

TABLE I
Oligonucleotide primers used for polymerase chain reaction Position 1 corresponds to the first base (A) of the initiation codon of the a subunit gene in the unc operon. Sequences of the top two lines and the other lines correspond to the sequences of the b and a subunits, respectively. The first 5Ј bases of the given reading frame (sequence) are shown next to a comma, and the residue number is shown in the left column (position). The first four nucleotides in all oligonucleotides were random and the subsequent GGATTC and GAATTC sequences are recognition sites for the restriction endonucleases BamHI and EcoRI, respectively.
mic Domain of the b Subunit-To identify subunits capable of binding the b subunit, we took two approaches; a genetic approach with the yeast two-hybrid system (27) and a biochemical approach with in vitro binding assay of subunits. For these approaches, we constructed two types of chimeric plasmids ( Fig. 1). For biochemical approaches, DNA fragment spanning the membrane peripheral portion of the b subunit (residues 26 -156, b cyt ) that was deduced from hydropathy analyses was fused to the GST gene to create a fusion peptide, and for genetic analysis with the two-hybrid system, the same portion of the b subunit was fused to the activation or DNA binding domain of GAL4 gene. The b cyt region of the b subunit gene was amplified by the polymerase chain reaction and fused to the GST or GAL4 gene. Based on the previous model of membrane topology in the a subunit of F 0 (3, 42-44), we amplified DNA of this gene corresponding to the putative peripheral membrane domain between residues 60 and 100, and 160 and 200, and also fused it to the binding or activation domain of the GAL4. To determine the binding partners of b cyt and portions of the a subunit, we used chimeric plasmids of F 1 subunits (␣, ␤, ␥, ␦, and ⑀) fused to the activation or binding domain of the GAL4 gene for the yeast two-hybrid system constructed in the previous study (26).
Detection of Subunits Capable of Binding to the b Subunit-First, we examined the interaction between b cyt and one of the F 1 subunits, and then we analyzed b cyt -b cyt and also b cyt -a interactions with the putative peripheral membrane domains of the a subunit. The pair of ␣ and ␤ subunits gave strong expression of the reporter gene as we described previously (26), whereas the vectors alone did not (Table II). Among the various combinations tested, the combination of b cyt fused to GAL4-ad (activation domain of GAL4) and ␦ fused to GAL4-bd (binding domain of GAL4) alone resulted in strong expression of the reporter gene, whereas all other combinations showed no significant expression (Table II). These results demonstrated that the interaction between b cyt and ␦ occurred under in vivo conditions and other combinations might not cause interaction in those pairs. Although the purified soluble fraction of b (residues 25-146 or 26 -156) was reported to form a dimer in vitro (24,45), we did not observe this interaction in the two-hybrid system. The putative membrane peripheral region of the a subunit also did not show reporter gene expression, suggesting that these portions may not interact stably with the ␦ subunit.
Overproduction of the b cyt and ␦ Subunits Fused to GST-When we induced expression of the GST-b cyt in E. coli carrying the GST-b cyt expression plasmid, GST alone without GST-b cyt was observed as shown in Fig. 2A. This observation suggested that the b cyt portion of the overproduced GST-b cyt was degraded in the cells, whereas the GST portion in the fusion protein was intact. Next, we tried to overproduce the GST-b cyt together with the ␦ subunit, because the interaction of these subunits indicated in the yeast two-hybrid system might protect the b cyt portion from degradation by proteases. In fact, the overproduction of GST-b cyt and ␦ were observed on SDS-polyacrylamide gel electrophoresis of cell extract by staining (Fig.  2B) and immunological detection with specific antiserum for GST and ␦, respectively (data not shown), consistent with the results of the two-hybrid system and our expectations. The stoichiometry of GST-b cyt and ␦ was 1.5 based on densitometric analysis of the stained bands. This result supported the notion that GST-b cyt interacted with ␦ in E. coli. It should be noted that the overproduced ␦ subunit has not been degraded, as shown immunologically (data not shown). According to our previous results regarding to GST-␦ expression (28), this fusion subunit is susceptible to proteolytic attack in E. coli cells.
Affinity Purification of GST-b Fusion Subunit-The soluble fraction of E. coli cells overproducing GST-b cyt and ␦ was subjected to glutathione-Sepharose affinity chromatography. The GST-b cyt fusion protein was eluted specifically with the addition of glutathione (Fig. 2B, final lane), whereas the ␦ subunit FIG. 1. Cloned portions of the a, b, or ␦ subunit genes in various expression vectors. Cloned portions of the b and a subunits in the expression plasmids for the two-hybrid system and GST fusion proteins are shown schematically in (1) and (2), and (3), respectively. For GAL4-ad (activation domain of GAL4) and GAL4-bd (binding domain of GAL4) plasmids, pGAD424 and pGBT9 were used, respectively. pGEX-2T was used for the expression of fusion proteins. Details of plasmid construction are described under "Materials and Methods."

TABLE II
Interaction of the b cyt and the F 1 F 0 ATPase subunits in the yeast twohybrid system The expression plasmids carrying the genes for the indicated fusion proteins were cotransfected into yeast SFY526 by the lithium acetate method as described under "Materials and Methods." Transformants were selected by plating onto SD medium lacking Trp and Leu. Transformants that contained both plasmids were grown to mid-log phase of cell growth in SD medium lacking Trp and Leu and then assayed for ␤-galactosidase activity, shown here in Miller units (c). The pairs in the top six lines indicate combinations b cyt fused to the activation domain of GAL4 (GAL4-ad) and various F 1 F 0 subunits fused to the binding domain of GAL4 (GAL4-bd). Similarly, the next five lines indicate combinations of b cyt fused to the binding domain of GAL4 and various subunits fused to the activation domain of GAL4. In the bottom two lines, the pair of ␣ and ␤ subunits fused to the activation and binding domains, respectively, and the combination of vectors alone (Ϫ) are shown as positive and negative controls, respectively.
was eluted nonspecifically by washing the column with phosphate-buffered saline (data not shown). A small amount of ␦ subunit was eluted with addition of glutathione (Fig. 2B, final lane), suggesting that ␦ was dissociated from GST-b cyt during chromatography. Although a significant amount of the ␦ subunit was not observed in the purified GST-b cyt fraction, trace amounts of ␣, ␤, ␥, ⑀, and ␦ might bind to the GST-b cyt , leading to the formation of the active ATPase. The ATPase activity in the purified GST-b cyt and GST fractions (100 g each) was measured, and significant activity was found for GST-b cyt (0.044 units/mg of protein) compared with GST alone (0.023 units/mg of protein). The presence of the ␣, ␤, and ␥ subunits in the fraction was detected by immunological assay (data not shown). These results suggested that a small portion of the GST-b cyt contained the ATPase complex with GST-b cyt formed in vivo, and also that the GST-b cyt could behave as the b subunit itself in vivo.
We tested in vitro binding of the ␦ and b cyt by assay of the binding product with glutathione-Sepharose affinity chromatography, but no significant binding was detected. Then we tested binding of these subunits on membrane filters, a procedure which is more sensitive than affinity chromatography. The interaction was detected by this binding assay: b cyt bound to ␦ but not to control GST (Fig. 3).
Reconstitution of the ATPase from the b cyt and ␣, ␤, ␥, ␦ and ⑀-Since interaction of b cyt and ␦ was demonstrated by the two different approaches, namely in vivo and in vitro binding assays, and a trace but a significant amount of ATPase formed in vivo was further detected as described above, we tested the interaction of b cyt and F 1 complex as follows. The ATPase was reconstituted from the purified ␣, ␤, ␥, ␦, ⑀, and GST-b cyt . The reconstituted ATPase was then subjected to glutathione affinity chromatography. As shown in Fig. 4, the ATPase activity was eluted after washing the column and also with addition of glutathione (26% of the applied activity) with specific activity of 19.0 units per mg of protein at the peak fraction. The eluted materials were analyzed by SDS-polyacrylamide gel electrophoresis and subsequent staining of proteins (Fig. 5A) or immunological detection with antisera against the subunits (Fig.   5B). The ␣, ␤, ␥, and GST-b cyt and these subunits together with ␦ and ⑀ were detected by staining (Fig. 5A) and immunological detection (Fig. 5B), respectively. The stoichiometry of ␣, ␤, ␥, ␦, ⑀, and GST-b cyt contained in the peak fraction based on densitometric analysis of the stained bands was 2.5:2.6:1.4:1.0:0.7: 1.9, which is similar to that in the native F 1 F 0 . When GST instead of GST-b cyt was used for reconstitution, the specific elution of the ATPase was not observed with the addition of glutathione (data not shown). These results indicated that at least one-third of the subunits were reconstituted with b cyt as a native F 1 F 0 -like complex.
Reconstitution of the ATPase from Various Combinations of the Subunits-To determine the interacting partner of the b cyt , we tested reconstitution of the ATPase from the combination of F 1 subunits except ␦ or ⑀ and b cyt . As shown in Fig. 6, A and B, the combination with ␦ but not ⑀ reconstituted the ATPase, the activity of which was specifically eluted with glutathione. The reconstituted activity was 38.1 units per mg of protein at the peak fraction. The subunits contained in this peak were those expected as the ␣, ␤, ␥, ␦, and b cyt (Fig. 7A). The combination with the ⑀ subunit instead of ␦ did not yield the ATPase activity after addition of glutathione, and no subunits other than GST-  3. Binding of b cyt to ␦ subunit on membranes. Two or 4 mg of the purified b cyt without GST and GST as a negative control was blotted on GVHP filters and soaked in a reconstitution buffer containing purified ␦ subunit (20 g in 2 ml) to bind both peptides. After removing unbound ␦ by washing the filter with a reconstitution buffer, bound ␦ was visualized with serum against the purified ␦ and a Vectastain kit.  5 g) were mixed together in a reconstitution buffer (2.4 ml) and dialyzed against the reconstitution buffer for 8 h at 26°C. After reconstitution, whole materials were subjected to glutathione-Sepharose column (0.3 ml) chromatography. After washing the column with 7 ml of reconstitution buffer, 10 mM glutathione was added at fraction 15 (n). Aliquots of the eluate were used for the ATPase assay and protein measurement. b cyt alone were observed (Fig. 7B). These results indicated that GST-b cyt did not bind ␣␤␥⑀ complex but to ␣␤␥␦ complex and that interaction of b cyt and ␦ is more important than that of b cyt and ⑀ in F 1 and F 0 interaction.
To obtain insight into the topological arrangement of the amino terminus of the ␦ or ⑀, and the b cyt within the F 1 F 0 -ATPase, we tested reconstitution of the ATPase from combinations of subunits, ␣, ␤, ␥, ⑀, b cyt and GST-␦, and ␣, ␤, ␥, ␦, b cyt and GST-⑀. The ATPase activities reconstituted with 21.4 and 25.2 units per mg of protein for the complex with GST-␦ and GST-⑀, respectively, were eluted in both cases after addition of glutathione (Fig. 8, A and B) on glutathione affinity chromatography. This result indicated that the amino termini of both subunits did not exist within the ATPase complex but were exposed on the surface of the complex. DISCUSSION It was shown by gel filtration that the purified membrane peripheral portion of the b subunit (residues 25-146) was integrated in a complex together with F 1 (24). This portion also inhibited the binding of F 1 to the F 0 portion. These previous results suggest that the cytoplasmic region of the b subunit interacts with the F 1 subunits. However, the subunits capable of binding to this subunit have not been identified. Here, we provide the evidences of b-␦ interaction with the yeast twohybrid system and in vitro binding of the purified subunits. However, the levels of the ␦ subunit in the affinity-purified GST-b cyt fraction of E. coli extract and in a mixture of purified GST-b cyt and ␦ in in vitro binding assay were low, indicating that the interaction of the subunits might be weaker than that observed for ␥-⑀ interaction reported previously (26,28).
Here, we clearly showed the reconstitution of the active ␣␤␥␦⑀GST-b cyt complex from the purified subunits, which has a similar structure to that of a native complex. Therefore, the relatively weak b-␦ interaction might be stabilized by complex formation with the other F 1 subunits. This suggests that the ␦ subunit interacts not only with b cyt but also with other subunits of F 1 . We have reported previously that this subunit interacts with the ␣, ␤, and ␥ subunits weakly based on the results of the yeast two-hybrid system (26). Dunn and coworkers (54) reported that deletion of the amino-terminal 15 residues of the ␣ subunit caused loss of F 1 binding to ␦, sug-gesting ␣-␦ interaction. Aris and Simoni (25) described chemical cross-linkage between ␣-␦, ␤-␦, and ␥-␦. Recently, extensive analyses in terms of subunit interaction of the ␦ with other subunits have been reported for chloroplast F 1 by cross-linking between Cys introduced into the ␦, and residues in the ␣ or ␤ subunits (46). This previous study showed the interaction of Ser-10 substituted by Cys in the ␦ subunit with the aminoterminal 62 residues in the ␤ subunit. Ser-57, -82, and -166 substituted by Cys were cross-linked within the amino-terminal 192 residues of the ␣ subunit. Therefore, stabilization of b-␦ interaction requires complex interactions of most other subunits of F 1 .
The stoichiometry of b and ␦ was shown to be 2 to 1 based on the purified F 1 F 0 -ATPase (18,45,47). According to semi-quantitative analyses of GST-b cyt and ␦ expression in total extracts of E. coli cells in this experiment, the stoichiometry b and ␦ was 1.5, indicating that both GST-b cyt and ␦ subunits were protected from proteolytic attack in E. coli cells to a certain extent. The specific interaction of these subunits based on the stoichiometry (GST-b cyt :␦ ϭ 1.9:1) was also observed for the reconstituted ␣␤␥␦⑀GST-b cyt complex. These results suggest that GSTb cyt and ␦ interact with each other in a manner similar to the native complex.
In marked contrast with ␣␤␥␦GST-b cyt complex formation, the ⑀ subunit did not reconstitute ␣␤␥⑀GST-b cyt complex, indicating the ⑀-b cyt interaction may be very weak even in the presence of other F 1 subunits. Since we have shown that purified ␣, ␤, ␥, and GST-⑀ reconstitute an active complex (28) Fig. 4 were subjected to SDS-polyacrylamide gel electrophoresis. The eluted materials were stained with Coomassie Brilliant Blue (A). After electrophoresis, the proteins in the peak fraction (fraction 16) were blotted onto GVHP filters and reacted with antibodies or serum for ␣, ␤, ␥, ␦, ⑀, or GST (B). The reacted materials were visualized with an ABC Vectastain kit. complex) does not bind to F 0 without the ⑀ subunit (18,48,49), which is inconsistent with the present results. However, the previous conclusion was based on the experiments in vivo or in vitro with membranes as the F 0 source, which was clearly different from the present experimental conditions. This difference suggests that the a and c subunits in F 0 (membrane) may destabilize the b-␦ interaction without the ⑀ subunit. It has been shown that the ⑀ subunit interacts with ␥ and c subunits (17). The ␥ subunit was also found to interact with the c subunit (50). Therefore, the F 1 F 0 interaction may be completed with two independent structures, interactions of ␥-c and ⑀-c, and b-␦.
It has been shown that 20 residues of the carboxyl-terminal region of the ␦ subunit are more susceptible to trypsin compared with the amino-terminal region (53). Recently based on NMR analyses, the atomic structure of ␦ subunit between residues 1-134 has been reported (16). In this study ␦ was shown to be a protein with two characteristic domains, with the compact six ␣-helix bundle at NH 2 terminus and a less well characterized COOH-terminal domain. These two domains may be cross-linked by a disulfide bridge between residues 64 and 140 as reported previously (51). Based on the present study together with these previous observations including CF 1 ␦ reported by Lill et al. (46), the ␦ subunit within NH 2 -terminal 104 residues is thought to bind the amino-terminal regions of the ␣ and ␤ subunits, and the rest of ␦ may bind to the dimer of b. Since random and site-specific mutagenesis studies suggested that no essential residues were present in the ␦ subunit (52), this subunit might be structurally, rather than functionally, required for catalysis. Formation of b-␦-␣ or b-␦-␤ interaction may thus be structurally important. It would be of interest to analyze whether the ␣, ␤, ␦, and b interaction is involved in the rotation as has been reported very recently (13) or whether this structure forms an immobile stator.
The previous study showed that the cytoplasmic region (residues 25-146) of b subunit tagged with a few amino acids was overproduced without degradation in the absence of ␦ subunit coexpression (24). The same region of ␦ in the present study fused to GST was also reported to be overproduced (45). Although the reason for the discrepancy between our result and those of the previous studies is not clear, differences in the experimental conditions for the tagged protein and different host cells used for overproduction as well in the these studies might be a cause. For the latter case (45), GroESL (chaperonin proteins) coexpressed with ␦ might lead to no degradation of ␦. It should be noted that point mutation in the ␦ subunit (Asp-161 to Pro) and total deletion the ␦ subunit causes degradation of the b subunit (52, 55), consistent with our results. Although FIG. 7. SDS-polyacrylamide gel electrophoresis and immunological detection of subunits in the reconstituted materials. The subunits in the eluates of affinity chromatography shown in Fig. 6 at the peak fraction were analyzed by SDS-polyacrylamide gel electrophoresis as described in the legend of Fig. 5 (a and b). Immunological detection (b) was also performed as described in the b subunit was reported to form a dimer (24, 45), we could not detect b cyt -b cyt interaction with the yeast two-hybrid system. Fusion of the DNA binding and activation domain of GAL4 gene product might interfere with this interaction. We also did not observe an interaction between the ␦ subunit and putative peripheral membrane portions of the a subunit (residues 60 -100 and 160 -200), regions that were selected based on the previous report (15). This negative result suggests that the model of membrane topology of the a subunit (15) may not be correct or that other subunits are involved for the b-a interaction. Yamada et al. (44) recently reported a new topology model of the a subunit in which residues 60 to 100 are in the exterior surface or within the membrane. Our results in terms of the region from residues 60 to 100 is consistent with this hypothetical structure. However, residues 160 to 200 are at the cytoplasmic surface of the a subunit (15,44). Based on this predicted structure and the present results in the two-hybrid system, residues between 160 and 200 do not appear to be involved in F 1 F 0 interaction.
In the present study, we have detected ␦-b interaction with the yeast two-hybrid system. This system may be useful to obtain mutants defective in ␦-b interaction. This approach was also successfully demonstrated for ␣-␤ interaction in our previous study (26).