Hybrid Rhodospirillum rubrumF0F1 ATP Synthases Containing Spinach Chloroplast F1 β or α and β Subunits Reveal the Essential Role of the α Subunit in ATP Synthesis and Tentoxin Sensitivity*

Trace amounts (∼5%) of the chloroplast α subunit were found to be absolutely required for effective restoration of catalytic function to LiCl-treated chromatophores ofRhodospirillum rubrum with the chloroplast β subunit (Avital, S., and Gromet-Elhanan, Z. (1991) J. Biol. Chem. 266, 7067–7072). To clarify the role of the α subunit in the rebinding of β, restoration of catalytic function, and conferral of sensitivity to the chloroplast-specific inhibitor tentoxin, LiCl-treated chromatophores were analyzed by immunoblotting before and after reconstitution with mixtures of R. rubrum and chloroplast α and β subunits. The treated chromatophores were found to have lost, in addition to most of their β subunits, approximately a third of the α subunits, and restoration of catalytic activity required rebinding of both subunits. The hybrid reconstituted with theR. rubrum α and chloroplast β subunits was active in ATP synthesis as well as hydrolysis, and both activities were completely resistant to tentoxin. In contrast, a hybrid reconstituted with both chloroplast α and β subunits restored only a MgATPase activity, which was fully inhibited by tentoxin. These results indicate that all three copies of the R. rubrum α subunit are required for proton-coupled ATP synthesis, whereas for conferral of tentoxin sensitivity at least one copy of the chloroplast α subunit is required together with the chloroplast β subunit. The hybrid system was further used to examine the effects of amino acid substitution at position 83 of the β subunit on sensitivity to tentoxin.

The photosynthetic F 0 F 1 ATP synthases found in the thylakoids of chloroplasts and in the cytoplasmic membranes of photosynthetic bacteria couple the movement of protons down an electrochemical proton gradient to the synthesis of ATP during photophosphorylation. The general structure of these ATP synthases is highly conserved, consisting of F 0 , the membrane-spanning proton channel, and F 1 , the peripheral membrane sector, which contains the catalytic sites for reversible ATP synthesis. The F 0 is composed of four different subunits labeled a, b, bЈ, and c in photosynthetic bacteria (1) and I-IV in chloroplasts (2)(3)(4). The chloroplast F 0 subunits IV, I, II, and III are analogous to the bacterial a, b, bЈ, and c, respectively, with a probable stoichiometry of a 1 b 1 bЈ 1 c 9 -12 . F 1 from all sources is composed of five different subunits designated ␣ to ⑀ in order of decreasing molecular weight with a stoichiometry of The x-ray crystal structure of bovine heart mitochondrial F 1 (MF 1 ) 1 at 2.8-Å resolution (5) defined the three-dimensional structures of alternating ␣ and ␤ subunits as forming a closed hexamer having a portion of the ␥ subunit embedded in its central cavity. Among the nucleotide binding sites, which are located one at each of the six ␣/␤ interfaces, the three catalytic sites, located predominantly on ␤ subunits, were found to exist in three different conformational states. This asymmetric feature is compatible with the binding change mechanism, which proposed that the catalytic sites interconvert between three different conformational states during ATP synthesis via energydependent affinity changes in substrate binding and product release (6,7). Several recent studies of isolated (8 -10) or membrane-bound (11) F 1 have suggested that this is achieved via rotation of the ␥ subunit relative to the ␣ 3 ␤ 3 subassembly. The MF 1 crystal structure identified a number of ␣-␤, ␣-␥, and ␤-␥ contacts, all or some of which may be responsible for dictating the asymmetric properties of each of the catalytic sites during catalysis. The importance of each of these contacts, and the steps involved in changing them, remain to be determined.
Partially dissociated membrane ATP synthase complexes, which can be reassembled by adding isolated ␣ and ␤ subunits, can provide suitable tools for identifying and characterizing the interacting protein domains responsible for the binding change process. One such system, where both ATP synthesis and hydrolysis can be followed, was obtained by LiCl treatment of chromatophores isolated from the photosynthetic bacterium Rhodospirillum rubrum (12). This treatment was found to release the bulk of their ␤ subunits (12,13) resulting in the loss of over 90% of their ATP synthesis and hydrolysis activities. Both activities could be restored upon reconstituting the treated chromatophores with the released subunits. The treated chromatophores could also be reconstituted with native spinach chloroplast CF 1 ␤, although the protein preparation was contaminated with trace amounts of CF 1 ␣ (14). More recently it was shown that the presence of small amounts of the ␣ subunit was a requirement for the reconstitution of a hybrid ATP synthase with the CF 1 ␤ subunits (15,16) or a native enzyme with RrF 1 ␤ (17, 18), suggesting the possibility that the LiCl treatment also removed some of the ␣ subunit from the chromatophores. The identification of a small amount of an ␣ 1 ␤ 1 dimer in addition to the large amount of ␤ in the LiCl extract (19) has confirmed this possibility. Further analysis (51) has shown that the amount of ␣ subunit, but not of ␤ subunit, released by the LiCl treatment is dependent upon the concentration of the chromatophores during the treatment.
The hybrid RrF 0 F 1 /CF 1 containing the CF 1 ␤ subunit with trace amounts of the CF 1 ␣ subunit was shown to have little, if any, proton-coupled ATP synthesis but 30 -40% of the normal MgATPase activity (14). The hybrid MgATPase activity was fully sensitive to tentoxin, a specific CF 1 inhibitor (20), whereas the control or restored native RrF 0 F 1 ATP synthesis or hydrolysis activities were completely resistant to tentoxin (14,21). This result suggested that the ␤ subunit might be responsible for conferring tentoxin sensitivity to the F 1 enzyme. An aspartic acid residue at position 83 of CF 1 ␤ was indeed implicated as being essential for tentoxin sensitivity of the chloroplast CF 0 F 1 ATP synthase (22,23). However, the observation that the MgATPase activity of membranes isolated from an uncD-deleted Escherichia coli strain complemented with the chloroplast atpB gene was insensitive to tentoxin (24) suggests that additional CF 1 subunits might also be involved in conferral of this sensitivity.
The fact that the LiCl treatment of R. rubrum chromatophores releases some of the ␣ subunits as well as the ␤ subunits has allowed us to examine the interplay between the ␣ and ␤ subunits, which results in coupled ATP synthesis/hydrolysis and in sensitivity to tentoxin. In this study we folded insoluble recombinant CF 1 ␤ (25) into a fully functional monomer using the method developed for folding the recombinant RrF 1 ␣ subunit (26) and prepared two different hybrid RrF 0 F 1 enzymes containing RrF 1 ␣ and CF 1 ␤ or CF 1 ␣ and CF 1 ␤ subunits. The results reveal that (a) the CF 1 ␤ subunit can restore a significant amount of proton-coupled ATP synthesis to treated chromatophores but only when all three copies of the RrF 1 ␣ subunit are present, and (b) the CF 1 ␣ subunit is required, along with the CF 1 ␤ subunit, to confer sensitivity to inhibition by tentoxin as well as high (Ͼ8-fold) stimulation by sulfite of the restored MgATPase activity.
Radioactive phosphate was obtained from DuPont. ATP (grade II) and tentoxin were purchased from Sigma. Tentoxin was dissolved in ethanol, diluted to a final concentration of 5 mM, and stored at Ϫ80°C. All other chemicals were the highest quality reagent grade available.
Preparation of CF 1 ␤ Mutants-All mutations were constructed by enzymatic amplification of the expression plasmid pET3a-␤NE3 described previously (25) and modified as described below, using inverse PCR. The primers employed (synthesized and 5Ј-phosphorylated by Macromolecular Resources, Colorado State University) had abutting 5Ј termini allowing for replication of the whole plasmid along with the mutation. The reverse primer was 24 nucleotides long with a base sequence corresponding to bases ϩ214 to ϩ237 of the wild-type atpB sequence and was used for generation of all three codon 83 mutations. The three forward primers had base sequences corresponding to bases ϩ238 to ϩ257. The mutations were generated by substituting the codon for aspartate, GAT, starting at position ϩ247, with GAG, CTT, or GCT to generate codons for glutamate, leucine, or alanine, respectively. Oligonucleotides were generated with the aid of the Primer Design program (SciEd Software). The pET3a-␤NE3 plasmid used only for the ␤ mutations contained a modified ␤ subunit gene in which the codon at position 378 (AGG) was modified to UGG, which replaced the arginine with tryptophan. This construct was found to increase the yield of folded monomeric ␤ subunits without affecting any of the catalytic properties of the protein. 2 This construct is designated as the wild-type protein in the results presented in Table II.
Plasmid DNA was prepared by ethanol precipitation following phenol:chloroform extraction (30). PCR was carried out in 50 l of cloned Pfu DNA polymerase reaction buffer containing 4 mM total MgSO 4 , 20 pmol of each primer, 0.4 mM dNTPs, 60 ng of the pET3a-␤NE3 plasmid DNA, and 2.5 units of cloned Pfu DNA polymerase (Stratagene). All components were mixed on ice and the placed in a GenAmp PCR System 2400 (Perkin-Elmer) prewarmed to 94°C. The PCR reaction was continued for 25 cycles of: 94°C for 1 min, 55°C for 1 min, and 72°C for 12 min. PCR products were purified by gel electrophoresis followed by electroelution using an ISCO micro trap. The eluted DNA was precipitated with ethanol and circularized by incubating the eluted DNA with 3 units of T4 DNA ligase (Promega) in T4 ligase buffer overnight at room temperature. The plasmid was transformed into competent E. coli XL1-blue cells (24). Cloned plasmid was isolated from the XL1-blue cells using boiling lysis, followed by isopropanol and ethanol precipitation, and transformed into the expression host E. coli BL21(DE3)/pLysS (25). The mutations were confirmed by sequencing of the entire ␤ gene using the fluorescent dideoxy method (31).
Overexpression, Solubilization, and Refolding of WT and Mutant CF 1 ␤-Recombinant E. coli cells containing either the WT or mutant atpB gene were grown, harvested, and washed as described by Chen et al. (25), except that the washing buffer contained 50 mM Tris-HCl (pH 8.0), 2 mM EDTA, and the following protease inhibitors: 1 mM phenylmethanesulfonyl fluoride, 2 mM benzamidine, and 10 g/ml N ␣ --tosyl-L-lysine chlorolethyl ketone. The cells were lysed by two cycles of freezing and thawing and sonicated to shear the DNA. The inclusion bodies containing the ␤ polypeptide were washed twice, resuspended in 10 -12 ml of the above buffer, and stored at Ϫ20°C. The ␤ polypeptide was solubilized and folded using the method developed for the refolding of RrF 1 ␣ (26), except that CF 1 ␤ was solubilized using 4 M urea and folded at a protein concentration of 0.3 mg/ml.
Reconstitution of LiCl-treated R. rubrum Chromatophores and Assays for Restored Activities-Reconstitution was carried out by incubating, for 1 h at 35°C, LiCl-treated chromatophores equivalent to 5 g of BChl with the indicated amounts of recombinant or native RF 1 and CF 1 ␣ and ␤ monomers in a final volume of 0.2 ml of reconstitution buffer containing 50 mM Tricine-NaOH (pH 8.0), 25 mM MgCl 2 , 4 mM ATP, 1 mM dithiothreitol, and 10% glycerol. To assay restored ATP synthesis, 0.04 -0.12 ml of the reconstitution mixture was diluted into a 1-ml assay mixture containing 50 mM Tricine-NaOH (pH 8.0), 5 mM MgCl 2 , 4 mM sodium phosphate (containing 0.5-1.0 ϫ 10 6 cpm 32 P i ), 2 mM ADP, 15 mM glucose, 24 units of hexokinase, and 66 M N-methylphenazonium methosulfate. The mixture was incubated in the dark at 35°C for 3 min before starting the reaction by illumination. The reaction was terminated after 3 min by turning off the lights and adding 0.1 ml of 2 M trichloroacetic acid. The amount of [␥-32 P]ATP synthesized was determined as described by Avron (32).
To assay Mg-and Ca-dependent ATP hydrolysis, the reconstituted chromatophores were washed three times with 50 mM Tricine-NaOH (pH 8.0), 50 mM NaCl, and 20% glycerol to remove residual MgCl 2 and unbound subunits. Washed chromatophores equivalent to 1-3 g of BChl were incubated for 10 min at 35°C in a 0.5-ml reaction mixture containing 50 mM Tricine-NaOH (pH 8.0), 50 mM NaCl, and either 2 mM MgCl 2 and 4 mM ATP, or 5 mM CaCl 2 and 5 mM ATP. The reactions were stopped by addition of 50 l of 2 M trichloroacetic acid, and P i release was determined as described by Taussky and Shorr (33).
When assaying for sensitivity to tentoxin, the reconstituted chromatophores were pre-incubated with 4 M inhibitor for 20 min at 35°C in the complete ATP synthesis assay mixture in the dark, or in the MgATPase assay mixture lacking ATP. ATP synthesis was initiated by illumination and ATP hydrolysis by addition of 20 l of 0.1 M ATP.
Other Procedures-Protein concentrations were determined by the method of Lowry et al. (34) or by the method of Bradford (35) using bovine serum albumin as a standard. The BChl content of chromatophores was determined at 880 nm using the in vitro extinction coefficient given by Clayton (36). SDS-polyacrylamide gel electrophoresis was carried out on 10% polyacrylamide gels according to Laemmli (37).

Folding and Isolation of Recombinant CF 1 ␤ Monomers-
Size exclusion chromatography was used to assess the difference between ␤ polypeptides solubilized and folded from inclusion bodies using an earlier procedure (25) as compared with the method described under "Experimental Procedures." The earlier method (Fig. 1A) yielded a protein preparation containing approximately equal amounts of monomers and dimers together with a small amount of aggregated protein. This could explain the earlier observation that only about 50% of the folded CF 1 ␤ was in a conformation capable of binding nucleotides (25). On the other hand, preparations of CF 1 ␤ solubilized and folded in the presence of 50 mM MgATP by the method developed for RrF 1 ␣ (26) was found to contain mainly monomeric ␤ subunits together with a small amount of aggregated protein, but no detectable dimers (Fig. 1B). Similar results were obtained for RrF 1 ␤ (Fig. 1C). At the protein concentrations used to fold the ␤ subunits (ϳ0.3 mg/ml), the ␣ subunit tended to aggregate and less than 20% of the RrF 1 ␣ was present in the monomeric form (data not shown). This increased to about 50% at protein concentrations of 50 g/ml (26). The several mutant CF 1 ␤ subunits used in this study all folded as well as the wild type protein, as judged by the percentage of monomeric protein present. The monomeric protein fractions of all folded proteins used in subsequent work were collected, pooled, concentrated, and could be stored indefinitely at Ϫ80°C without loss of reconstitutive activity.
Reconstitution of a Hybrid RrF 1 F 0 from Recombinant RrF 1 ␣ and CF 1 ␤ Subunits-When LiCl-treated chromatophores were incubated with a fixed excess of CF 1 ␤ and increasing amounts of RrF 1 ␣, a significant rate of sulfite-stimulated Mg-dependent ATP hydrolysis activity was restored. At a saturating ratio of added RrF 1 ␣/BChl of about 3, this activity reached nearly 60% of that restored by reconstitution with the RrF 1 ␣ and RrF 1 ␤ monomers. When reconstituted by themselves, neither RrF 1 ␤ nor CF 1 ␤ restored any activity. Similarly, the RrF 1 ␣ alone did not restore appreciable amounts of activity to the LiCl-treated chromatophores (Fig. 2). The RrF 0 F 1 /CF 1 hybrid containing RrF 1 ␣ and CF 1 ␤ also showed a significant rate of restored ATP synthesis, which closely paralleled the restoration of MgATPase activity in its dependence on increasing ␣ subunit concentrations (Fig. 3). The maximal ATP synthesis activity restored at saturating ␣ concentrations amounted to about 30% of the activity restored in the presence of identical ratios of RrF 1 ␣ and ␤ subunits ( Table I).
The ␣ and ␤ subunit compositions of chromatophores before and after LiCl treatment and after reconstitution with different mixtures of ␣ and ␤ subunits were compared using immunoblots probed with a mixture of antisera produced against the RrF 1 ␣ and ␤ subunits (Fig. 4). To minimize nonspecific subunit binding, the reconstitution of all tested native and hybrid complexes was carried out with the minimal amounts of the ␣ and ␤ pairs required to restore maximal activity. The antibody mixture was found to cross-react with CF 1 ␤ but not with CF 1 ␣ (Fig. 4, lane 1). The LiCl-treated chromatophores appeared to have lost almost all of the ␤ subunits but also a significant amount of the ␣ subunits (Fig. 4, lanes 2 and 3). Reconstitution of the treated chromatophores with a mixture of the refolded RrF 1 ␣ and ␤ monomers resulted in their binding to an approximately equimolar ratio (Fig. 4, lane 4). Reconstitution with a mixture of RrF 1 ␣ and CF 1 ␤ led to the formation of a hybrid complex containing both RrF 1 ␣ and CF 1 ␤ together with the residual traces of RrF 1 ␤ (compare Fig. 4, lanes 4 and 5). Interestingly, when the reconstitution was carried out with either ␤ or ␣ alone, the amount of subunit rebound to the LiCl-treated chromatophores was much lower than when chromatophores were exposed to both subunits together (Fig. 4,  compare lanes 4, 6, and 7 and lanes 5 and 8). This suggests that binding of the ␣ and ␤ subunits might be coordinated.
Reconstitution of a Hybrid RrF 0 F 1 Containing CF 1 ␣ and ␤ Subunits-An attempt to fold insoluble recombinant CF 1 ␣ subunit via the methods described for CF 1 ␤ (25) or RrF 1 ␣ (26) was unsuccessful, probably due to the rapid aggregation of the ␣ polypeptide upon removal of the solubilizing reagent. It was, therefore, not possible to test whether purified CF 1 ␣ can, as RrF 1 ␣, reconstitute hybrids with monomeric RrF 1 ␤ or CF 1 ␤. However, since CF 1 ␣␤ isolated from LiCl-treated chloroplasts was reported to supply the low amounts of CF 1 ␣ required for reconstitution of a hybrid RrF 0 F 1 with isolated pure native CF 1 ␤ (16), we checked whether it can also form a hybrid with RrF 1 ␤. The CF 1 ␣␤ used for this assay was isolated from soluble CF 1 , which lacked the ␦ and ⑀ subunits according to Gao et al. (27), and size exclusion HPLC revealed that it was completely dissociated into a mixture of ␣ and ␤ monomers (Fig. 1D). The results of titrating the LiCl-treated chromatophores with increasing amounts of the ␣␤ preparation are shown in Fig. 5 and demonstrate that this mixture could, in the absence of additional added ␤ subunits, restore sulfite stimulated MgATPase activity to nearly the same level obtained with the RrF 1 ␣ and ␤ subunits (compare Figs. 2 and 5). Moreover, the reconstitution showed the typical lag in attaining significant rates of ATP hydrolysis, which is associated with the normal reconstitution process (Fig. 2).
When the CF 1 ␣␤ titration was carried out in the presence of a fixed amount of the folded CF 1 ␤ monomers, there was a marked stimulation of the activity restored at low CF 1 ␣␤ concentrations. The stimulation may result from the fact that more than twice as much ␤ than ␣ is required for the reconstitution and so increasing the ␤ to ␣ ratio provides a more optimal condition for reconstitution. A much more surprising result, however, was obtained upon addition of RrF 1 ␤, which completely blocked restoration of the MgATPase activity by the CF 1 ␣␤ (Fig. 5).  (14). The still greater sulfite stimulation observed for the hybrid enzyme containing similar amounts of CF 1 ␣ and ␤ subunits (see Fig. 6, inset, lane 3) might be a specific property of the CF 1 ␣/␤ interfaces. Tentoxin Sensitivities of Chromatophores Containing Hybrid F 0 F 1 Enzymes-Control R. rubrum chromatophores (14) or LiCl-treated chromatophores reconstituted with isolated native RrF 1 ␤ (14, 21), which was shown to contain ϳ5% of RrF 1 ␣ (17,40), are completely resistant to inhibition by tentoxin. In contrast, LiCl-treated chromatophores reconstituted either with CF 1 ␤ containing ϳ5% of CF 1 ␣ (14) or with CF 1 ␣␤ (21) were found to be fully sensitive to tentoxin. However, the hybrid F 0 F 1 enzyme reconstituted with CF 1 ␤ plus RrF 1 ␣ was unaffected by tentoxin at concentrations sufficient to fully inhibit the enzyme reconstituted with CF 1 ␣␤ (Fig. 6). Higher concentrations of tentoxin had no additional effect (data not shown). This result inferred that the ␣ subunit also plays a role in conferring tentoxin sensitivity. To confirm that the CF 1 ␣ does indeed bind to LiCl-treated chromatophores along with the CF 1 ␤, the membranes were solubilized and probed with antibodies raised against purified CF 1 ␣␤ (Fig. 6, inset). Chromatophores reconstituted with RrF 1 ␣ and ␤ subunits (Fig. 6, inset, lane 1) did not bind detectable amounts of the CF 1 antibodies, indicating a lack of cross-reactivity for the antibodies.  However, in chromatophores reconstituted with RrF 1 ␣ and CF 1 ␤, or CF 1 ␣␤ (Fig. 6, inset, lanes 2 and 3), similar amounts of the CF 1 ␤ were bound in each case. Furthermore, the chromatophores incubated with CF 1 ␣␤ bound a significant amount of the CF 1 ␣ subunit in addition to the ␤ subunit (Fig. 6, inset,  lane 3), thus confirming that sensitivity to tentoxin does indeed correlate with the binding of both CF 1 ␣ and ␤ subunits to the treated R. rubrum chromatophores. The absence of any crossreaction between the CF 1 ␣ antibodies and RrF 1 ␣ prevented an evaluation of the ratio of bound CF 1 ␣ to the remaining RrF 1 ␣. Mutation of CF 1 ␤-Asp 83 Block Tentoxin Inhibition-It was proposed earlier (22) that Asp 83 on the ␤ subunit plays a critical role in conferring tentoxin sensitivity to CF 1 , perhaps via a direct interaction with the toxin. We have used the chromatophore reconstitution system to test this hypothesis. Three mutant CF 1 ␤ subunits were prepared in which Asp 83 was substituted for the larger acidic side chain of Glu (␤-D83E), for Ala (␤-D83A) with a less bulky side chain, and for Leu (␤-D83L) with a bulky but uncharged side chain. The capacity of these mutants to stimulate the restoration of MgATPase activity was examined by reconstituting LiCl-treated chromatophores with a small amount of CF 1 ␣␤, which, by itself, restored only a low MgATPase activity, in the presence of a 100-fold excess of monomeric mutant or WT CF 1 ␤, which restored a 3-fold higher MgATPase activity (Table II). The hybrid chromatophores reconstituted with CF 1 ␣␤ minus or plus additional WT CF 1 ␤ were fully inhibited by tentoxin. In contrast, all three of the hybrid enzymes reconstituted with CF 1 ␣␤ and the three mutant CF 1 ␤ subunits were insensitive to tentoxin. Their complete insensitivity to tentoxin can be explained by the fact that the tentoxin-sensitive native CF 1 ␤ monomer present in the small amount of CF 1 ␣␤ used as a source of the ␣ subunit (Table  II) was diluted 100-fold with the added mutant CF 1 ␤ monomers.

DISCUSSION
Previous work has shown that the MgATPase activity of LiCl-treated chromatophores could be restored by their reconstitution either with CF 1 ␤ containing trace amounts (ϳ5%) of CF 1 ␣ (14), or with a preparation of CF 1 ␣␤ containing an equimolar ratio of both subunits but not with a highly purified preparation of CF 1 ␤ (15). High rates of MgATPase activity could, however, be restored to the treated chromatophores when the purified CF 1 ␤ was supplemented with small amounts of the CF 1 ␣␤ subunit preparation (16). We have shown here that LiCl treatment of the R. rubrum chromatophores releases a significant amount of the ␣ subunit along with the ␤ subunits, thus explaining the requirement for small amounts of ␣ in addition to ␤ subunits for restoration of ATP synthesis and hydrolysis in the treated chromatophores. In an accompanying paper (51), it is further shown that the amount of ␣ released increases with decreasing chromatophore concentration during the LiCl treatment, whereas the bulk of the ␤ subunit is released at all tested concentrations. It is, therefore, also possible in this system to vary the ratio of the released ␣ and ␤ subunits.
The preferential release of RrF 1 ␤, which has been demonstrated by immunoblotting of the LiCl supernatant (51) as well as of the treated chromatophores (Fig. 4, lanes 2 and 3) is not readily explained by examining the crystal structure of MF 1 in which the ␣ 3 ␤ 3 hexamer appears to be stabilized by multiple contacts with the ␥ subunit (5). It can be better explained by the recently proposed structural models of the F 0 F 1 ATP synthase, which suggest the presence of two connections between the F 0 and F 1 parts of the enzyme (7,(41)(42)(43). One connection involves the ␥, ⑀, and c subunits and is proposed to function as the rotating portion of the enzyme. The second connection contains the ␦ subunit and the two b subunits and is suggested to act as a stator, which binds the F 1 ␣ 3 ␤ 3 hexamer to the F 0 sector allowing the smaller subunits to rotate with respect to the hexamer. Earlier biochemical evidence has shown that a readily formed disulfide link between the EcF 1 ␦ and one of the ␣ subunits did not inhibit ATPase activity (44,45). Furthermore, with bifunctional cross-linking reagents, ␣-␦ dimers and ␣-␣-␦ trimers, but not ␤-␦ dimers, were identified in EcF 1 (46) and TF 1 (47). The formation of such dimers and trimers as well as b-␦ and b-b dimers have recently been observed by crosslinking of introduced cysteine residues in EcF 0 F 1 ␣, ␦, and b subunits (48,49). The specific interactions between these subunits could stabilize a portion of the RrF 1 ␣ in the R. rubrum membranes during the LiCl treatment, thus resulting in the retention of at least one and possibly two ␣ subunits per RrF 0 F 1 complex while nearly all of the RrF 1 ␤ subunits are removed.
The treated R. rubrum chromatophores enabled us to form two types of RrF 0 F 1 /CF 1 hybrids. One contained mostly CF 1 ␤ and exclusively RrF 1 ␣, while the other also contained CF 1 ␣, which replaced the released portion of RrF 1 ␣. A comparison of the activities of these two hybrids with control RrF 0 F 1 demonstrated that F 1 ␣ plays an essential role in a number of catalytic properties. Two specific CF 1 ␣ functions were documented; it was absolutely required, together with CF 1 ␤, for obtaining full inhibition of the restored MgATPase activity by tentoxin, as well as for the high stimulation of this restored activity by sulfite ( Fig. 6 and Table I).
Earlier reports demonstrated that the binding of 1 mol of tentoxin/mol of heat-activated CF 1 was sufficient for obtaining full inhibition of its ATPase activity (20). So, although the absence of any cross-reaction between RrF 1 ␣ antibodies and CF 1 ␣ (Fig. 4, lane 1) or vice versa (Fig. 6, inset) did not enable the determination of the ratio of the bound CF 1 ␣ to the remaining RrF 1 ␣, the full inhibition of the hybrid containing both CF 1 ␣ and ␤ by tentoxin sets a lower limit of 1 mol CF 1 ␣ bound/mol of reconstituted RrF 0 F 1 /CF 1 . CF 1 ␤-Asp 83 was shown to be required for the inhibitory action of tentoxin and residues of different charge, such as lysine, or different spacer length such as glutamate, could not replace it (22,23). Table II illustrates that this aspartate is indeed essential for obtaining full inhibition by tentoxin in the presence of CF 1 ␣. It could not be replaced even by the noncharged leucine or by the smaller alanine, thus suggesting the importance of the CF 1 ␤-Asp 83 charge and spacer length for tentoxin binding and/or inhibition.
The mitochondrial equivalent of CF 1 ␤-Asp 83 , MF 1 ␤-Glu 67 , is located at an ␣/␤ interface in the bovine heart MF 1 structure (5). If the CF 1 ␤-Asp 83 forms part of the tentoxin binding site, then CF 1 ␣ residues located near the CF 1 ␤-Asp 83 may also contribute to the binding of tentoxin. Comparison of known amino acid sequences of ␣ subunits from both tentoxin-sensitive and -insensitive species revealed that, while all ␣ subunits share considerable homology, a stretch of amino acid residues between 121 and 133 (chloroplast numbering) shows conserva-tion only among the sensitive ␣ subunits of spinach, pea, and C. reinhardtii. This stretch of amino acids is very divergent in the resistant ␣ subunits of RrF 1 and EcF 1 with several amino acids differing in charge and size from the consensus sequence of sensitive CF 1 ␣ subunits. In addition, several of the residues in the equivalent stretch of amino acids in MF 1 are located within 10 Å of Glu 67 on the ␤ subunit. Two such residues in CF 1 , Ser 131 and Pro 132 , are good candidates for having involvement in tentoxin binding. We are currently mutating these residues into RrF 1 ␣ to test this possibility.
The MgATPase activity of treated chromatophores reconstituted with CF 1 ␤ containing ϳ5% CF 1 ␣ were stimulated 5-7fold by sulfite as compared with the 2-3-fold stimulation obtained with chromatophores reconstituted with RrF 1 ␤ (14). These results could not specify whether the CF 1 ␣ or ␤ or both were responsible for the extra high sulfite stimulation. The results presented in Table I indicate that, in the hybrid containing only CF 1 ␤, the stimulation is very similar to the one obtained with the control R. rubrum chromatophores, but when CF 1 ␣ is also present, an 8-fold stimulation is obtained. This high stimulation was found to raise the level of the MgATPase activity restored in the hybrid enzyme formed with CF 1 ␣␤ to the activity restored with RrF 1 ␣ and ␤. The lower initial MgATPase activity of the hybrid containing CF 1 ␣ and ␤ subunits may in part reflect the overall latency of the ATPase activity of CF 1 , which is considered necessary to limit wasteful ATP hydrolysis by CF o F 1 in the dark (1-3). CF 1 ␣ exerts a negative effect on the restoration of protoncoupled ATP synthesis, which was obtained only when the treated chromatophores were reconstituted with RrF 1 ␣ in the presence of either RrF 1 ␤ or CF 1 ␤ (Table I). These results suggest that the chromatophores containing bound CF 1 ␣ are not properly coupled. In contrast, the hybrid chromatophores containing CF 1 ␤ in the presence of RrF 1 ␣ showed significant rates of ATP synthesis, indicating that CF 1 ␤ can replace at least some of the important energy coupling interactions of RrF 1 ␤. The formation of hybrid RrF 0 F 1 /CF 1 with chimeric CF 1 ␣/RrF 1 ␣ and CF 1 ␤/RrF 1 ␤, whose preparation is now under way, should help to identify the F 1 ␣ and/or ␤ domains that are essential for the tight protein-protein interactions required for efficient proton-coupled ATP synthesis or hydrolysis.
Interestingly, although CaATPase activity is not coupled to proton translocation (50) it was not restored by any hybrid RrF 0 F 1 /CF 1 (Ref. 14 and see Table I). Recent results obtained with an RrF 1 ␤-T159S mutant (51) have demonstrated that substitution of serine for threonine in the active site blocks the restoration of Ca-dependent ATPase activity while enabling full restoration of both Mg-dependent ATP synthesis and hydrolysis. These two sets of results likely demonstrate differences in the geometry of the active sites on CF 1 ␤ and RrF 1 ␤, as well as on RrF 1 ␤-T159S when occupied by CaATP as compared with MgATP.