Mutagenesis of β-Glu-195 of the Rhodospirillum rubrum F1-ATPase and Its Role in Divalent Cation-dependent Catalysis*

We introduced mutations at the fully conserved residue Glu-195 in subunit β of Rhodospirillum rubrum F1-ATPase. The activities of the expressed wild type (WT) and mutant β subunits were assayed by following their capacity to assemble into the earlier prepared β-depleted, membrane-bound R. rubrum enzyme (Philosoph, S., Binder, A., and Gromet-Elhanan, Z. (1977)J. Biol. Chem. 252, 8742–8747) and to restore ATP synthesis and/or hydrolysis activity. All three mutations, β-E195K, β-E195Q, and β-E195G, were found to bind as the WTβ into the β-depleted enzyme. They restored between 30 and 60% of the WT restored photophosphorylation activity and 16, 45, and 105%, respectively of the CaATPase activity. The mutants required, however, much higher concentrations of divalent cations and could not restore any significant MgATPase or MnATPase activities. Only β-E195G could restore some of these activities when assayed in the presence of 100 mm sulfite and high MgCl2or MnCl2 concentrations. These results suggest that the observed difference in restoration of ATP synthesis and CaATPase, as compared with MgATPase and MnATPase, can be due to the tight regulation of the last two activities, resulting in their inhibition at cation/ATP ratios above 0.5. The R. rubrumF1β-E195 is equivalent to the mitochondrial F1β-E199, which points into the tunnel leading to the F1 catalytic nucleotide binding sites (Abrahams, J. P., Leslie, A. G. W., Lutter, R., and Walker, J. E. (1994) Nature 370, 621–628). Our findings indicate that this residue, although not an integral part of the F1catalytic sites, affects divalent cation binding and release of inhibitory MgADP, suggesting its participation in the interconversion of the F1 catalytic sites between different conformational states.

All respiratory and photosynthetic cells contain a membrane-embedded F 0 F 1 ATP synthase that generates ATP at the expense of the electrochemical proton gradient formed during electron transport. The catalytic F 1 component of the enzyme has been solubilized as a functional ATPase from many bacterial, mitochondrial, and chloroplast sources. It is a very conserved multimeric assembly with a stoichiometry of ␣ 3 ␤ 3 ␥␦⑀ and has up to six nucleotide binding sites. Three of them are catalytic sites, residing mainly on the ␤ subunits, and three are noncatalytic, located mainly on the ␣ subunits (1)(2)(3)(4)(5)(6).
Two recently published x-ray crystallographic structures of rat liver mitochondrial MF 1 at 3.6 Å (7) and bovine heart MF 1 at 2.8 Å (8) have confirmed the alternate arrangement of the six large ␣ and ␤ subunits in a closed hexamer. The structure at 2.8 Å resolution provides direct proof for the earlier suggested location of all six nucleotide binding sites at ␣/␤ interfaces (see Ref. 9). It has also resolved two long C-and Nterminal helical domains of the ␥ subunit, which are embedded within the internal cavity of the ␣ 3 ␤ 3 hexamer, and impose on it an asymmetric structure. This asymmetry is also reflected in different conformational states displayed by the three catalytic sites on the ␤ subunits (8). The binding change mechanism (10) suggested that ATP synthesis and hydrolysis involves transitions between such different conformational states via rotation of the ␥-subunit relative to an ␣ 3 ␤ 3 subassembly. So this structural information initiated a drive to demonstrate that such rotation is possible (11)(12)(13).
These important observations do not explain how rotation drives catalytic site transitions, which must involve intricate protein-protein interactions within the ␣/␤ catalytic interfaces as well as between them and various domains on the ␥ subunit. Full elucidation of the unique mechanism of action of the ATP synthases will therefore require clear definition of the role of amino acid residues that are involved in catalysis, as well as identification and detailed characterization of all interacting subdomains on the ␣, ␤, and ␥ subunits, and their relation to the catalytic sites.
Genetic/biochemical assay systems that are crucial for such studies have been developed in respiratory bacteria and yeast mitochondria (see Refs. 1, 2, and 6), but the application of this approach to photosynthetic systems lagged behind the respiratory ones. Although the conserved structure of all isolated F 1 -ATPases suggests a common catalytic mechanism, there are clear differences in various properties between the respiratory and photosynthetic F 0 F 1 and F 1 -complexes, for instance in regulation of activity and sensitivity to some inhibitors (3)(4)(5). It is, therefore, important to develop recombinant molecular techniques and direct in vivo and/or in vitro assays of activity for photosynthetic complexes containing mutated F 1 -subunits. Two such systems have recently been reported for Chlamydomonas reinhardtii and spinach chloroplast F 1 -␤ subunits (14 -17).
The F 1 -␤ subunit of the photosynthetic bacterium Rhodospirillum rubrum (RrF 1 ␤) 1 provides an especially suitable system for mutational analysis, based on the available MF 1 crystal structure. Thus, the published sequence of the RrF 1 operon (18) revealed that RrF 1 ␤ is most closely related to MF 1 ␤. Their amino acid sequences show Ͼ76% identity, as compared with only 72% for EcF 1 ␤, 69% for tobacco CF 1 ␤ (18), and 68% for TF 1 ␤ (19). The very high similarity between RrF 1 ␤ and MF 1 ␤ extends also to the R. rubrum and mitochondrial F 0 F 1 and F 1 which, unlike the chloroplast and Escherichia coli enzymes, exhibit an identical sensitivity to inhibition by oligomycin and efrapeptin (20 -22). Furthermore, a large number of in vitro assays of activity have been developed for the RrF 1 ␤ subunit that was isolated in functional form from the chromatophore membrane-bound RrF 0 F 1 . They include the ability of the isolated RrF 1 ␤ to bind ATP, ADP, and P i (23,24) as well as to rebind to the ␤-depleted enzyme and restore its lost ATP synthesis and hydrolysis activities (25,26).
Baltscheffsky et al. (27) have cloned the RrF 1 ␤ gene and expressed it in E. coli as a fusion protein with glutathione Stransferase but did not carry out any assays of activity. We have recently cloned this gene and have expressed the RrF 1 ␤ subunit in E. coli lacking the whole unc operon as a soluble protein that could restore ATP synthesis to ␤-depleted chromatophores (28). Here we describe the expression and full purification of RrF 1 ␤ mutated in . A detailed comparison of the activities of these mutants with those of the expressed WT ␤ subunit revealed that RrF 1 ␤-E195 plays an important role in divalent cation-dependent ATP synthesis and hydrolysis.
Cloning and Site-directed Mutagenesis-The RrF 1 ␤ gene was amplified from R. rubrum genomic DNA by PCR and cloned as described in (28). Since the PCR is known to cause mutations, two PCR products, one obtained with Vent and the second with Taq DNA polymerase, were cloned and fully sequenced by the Taq dye deoxy chain termination method, using a 373A DNA Sequenase. Both clones were found to contain one mutated nucleotide. In the first clone, nucleotide 222 was changed, altering the published ACC codon of threonine 74 (18) into an ACA one that also codes for threonine. So this clone encodes the WT RrF 1 ␤ polypeptide. But the second clone had a mutation in nucleotide 584, which changed the GAG codon of glutamic acid 195 into the GGG codon of glycine. This clone thus encodes an RrF 1 ␤-E195G mutant polypeptide.
Additional mutations at RrF 1 ␤-E195 were introduced by site-directed mutagenesis with the U.S.E. mutagenesis kit (Pharmacia Biotech Inc.) according to the instructions of the supplier, using the EcoRI-BamHI PCR-prepared fragment containing the complete WT RrF 1 ␤ gene (28) cloned into pUC18. The mutagenic primers, which must anneal to the same strand of the heat-denatured plasmid DNA as the U.S.E. selection primer, were therefore designed with the 5Ј-end to right. The oligonucleotides used for the E195K and E195Q substitutions were, respectively: 3Ј-A-GAA-ATA-GTG-TTC-TAC-TAG-CTA-CGG-C-CC-TAA-T-5Ј and 3Ј-A-GAA-ATA-GTG-GTC-TAC-TAG-CTA-CGG-CC-C-TAA-T-5Ј. They contained a new, underlined site for ClaI, introduced by a single base change, indicated by a bold letter as the bases changed to give the Lys or Gln codons.
The U.S.E. selection primer eliminates the ScaI site in the pUC18 amp gene, but the repair defective E. coli NM 522 mutS strain, transformed with the ScaI-resistant mutated plasmid DNA according to the instructions of the supplier, failed to grow in liquid medium. The transformed cells did, however, form ampicillin-resistant colonies when plated on solid agar. Introduction of the mutations into the RrF 1 ␤ gene was confirmed by hybridization of these colonies with a 32 P-labeled mutagenic primer. The mutations were further confirmed by ClaI restriction analysis, using the ClaI site introduced into the mutagenic primers, and by full DNA sequencing.
The ␤-E195Q gene had only the stated mutation. But ␤-E195K had two additional changes: 1) in nucleotide 159, altering the codon from GTG to GTT, both coding for valine 53 (18); and 2) in nucleotide 1069, altering the codon from CTG to TTG, both coding for leucine 357 (18). So this gene also encodes only the stated mutation.
Identification of Expressed RrF 1 ␤-Cultures of the various transformed unc-depleted E. coli strains were centrifuged and resuspended in buffer containing 50 mM Tricine-NaOH, pH 8.0; 4 mM MgATP, and 10% glycerol, which is optimal for isolating active native RrF 1 ␤ (25). The protease inhibitors benzamidine, phenylmethanesulfonyl fluoride, and N ␣ -p-tosyl-L-lysine chloromethyl ketone were added at 1.5 mM, 1 mM, and 30 M, respectively, and the cells were disrupted under argon in a Yeda press (26). The soluble cytoplasmic and insoluble membrane fractions were analyzed for the presence of expressed RrF 1 ␤ by SDS-PAGE (31,32) and Western immunoblotting (33).
Preparation of ␤-less R. rubrum Chromatophores and Their Reconstitution with RrF 1 ␤-␤-less chromatophores were obtained as described in (26). This technique releases all their RrF 1 ␤ (Refs. 25 and 34; see Fig. 4, lane 6) together with trace amounts of RrF 1 ␣ (28, 35), leading to loss of their ATP synthesis and hydrolysis activity. Reconstitution was carried out by incubating ␤-less chromatophores, at 5 g of Bchl, for 1 h at 35°C in 0.2 ml of a reaction mixture containing 50 mM Tricine-NaOH (pH 8.0), 25 mM MgCl 2 , 4 mM ATP, 1 mM dithiothreitol, and unless otherwise stated, 1 g of RrF 1 ␣ and saturating amounts of RrF 1 ␤ at various stages of purification.
Assays of Restored ATP Synthesis and Hydrolysis-ATP synthesis was usually assayed by a 5-fold dilution of the 0.2-ml mixture of reconstituted chromatophores into an assay mixture which contained in 1 ml a final concentration of 50 mM Tricine-NaOH (pH 8.0), 5 mM MgCl 2 , 4 mM sodium phosphate (containing 0.4 -0.8 ϫ 10 6 cpm of 32 P i ), 2 mM ADP, 15 mM glucose, 24 units of hexokinase, and 66 M N-methylphenazonium methosulfate. When assaying the Mg 2ϩ requirement of the restored ATP synthesis (and hydrolysis), about 3 ml of the reconstituted chromatophores were subjected to two rounds of centrifugation and resuspension in 50 mM Tricine-NaOH (pH 8.0) and 10% glycerol to remove the 25 mM MgCl 2 , which are essential for the reconstitution (25,26). These washed chromatophores, at 3-5 g of Bchl, were added to the synthesis assay mixture, together with the indicated concentrations of MgCl 2 , and preequilibrated for 5 min at 35°C in the dark. ATP synthesis was started by illumination and stopped after 3 min by turning off the lights and adding 0.1 ml of 2 M trichloroacetic acid. The synthesized [␥-32 P]ATP was determined as described by Avron (36).
For assaying restored ATP hydrolysis, the washed chromatophores, at 3-5 g of Bchl, were preincubated for 5 min at 35°C in 0.66 ml of 50 mM Tricine-NaOH (pH 8.0) with the divalent cation concentrations indicated in the text. Hydrolysis was started by addition of 40 l of ATP to a final concentration of 4 mM and stopped by 0.1 ml of 2 M trichloroacetic acid. The released P i was measured according to (37).
Other Procedures-For measurement of ATP binding, purified RrF 1 ␤ preparations were depleted of bound MgATP by elution-centrifugation through Sephadex G-50 columns (38) preequilibrated with TGN buffer containing 50 mM Tricine-NaOH (pH 8.0), 20% glycerol, and 50 mM NaCl. Binding was assayed by incubating 10 M depleted RrF 1 ␤ for 90 min at 23°C in TG buffer with the indicated concentrations of MgCl 2 and ATP, containing 1-2 ϫ 10 6 cpm of [2, H]ATP. Incubation was started by addition of RrF 1 ␤ and stopped by subjecting 50-l samples to elution-centrifugation, and the effluent was assayed for protein and radioactivity as described by Gromet-Elhanan and Kananshvili (23).
Protein concentrations were measured either by the BCA method (39) or according to Lowry et al. (40). The Bchl content of chromatophores was determined at 880 nm using the absorption coefficient in vivo given by Clayton (41).

Construction of RrF 1 ␤-E195
Mutants-The ␤-E195G mutation was obtained as a mistake, which changed the GAG codon of glutamic acid to the GGG codon of glycine, during amplification of the RrF 1 ␤ gene from genomic R. rubrum DNA by PCR.
Since the equivalent MF 1 ␤-E199 is described in the crystal structure as pointing into the conical tunnel leading to the catalytic nucleotide binding site (8), it was interesting to check whether the E195G mutation affects activity. Using the system described previously (28), the expressed, partially purified mutated ␤ subunit was found to restore a much lower rate of ATP synthesis in ␤-less chromatophores than the native or expressed WT RrF 1 ␤. We have therefore prepared two additional RrF 1 ␤-E195 mutants by oligonucleotide-directed mutagenesis: from glutamic acid to lysine (␤-E195K), carrying a positive charge, and to glutamine (␤-E195Q), carrying no charge.
Expression and Purification of WT and Mutant RrF 1 ␤ Subunits-Cloned WT and mutant ␤ genes were ligated into the expression vector pBTacI, and the recombinant plasmid was transformed into E. coli LM3115 (28). This strain was found to express larger amounts of RrF 1 ␤ than two other unc-deleted strains, LM2800 (29) and DK8 (30), under all tested growth conditions (not shown). Optimal conditions for expression of large amounts of RrF 1 ␤ as a soluble protein include growth of E. coli LM3115 at 22°C in the presence of 5% glycerol to about A 0.65 . Growth at higher temperatures or to a higher optical density increased the fraction of expressed ␤ subunit appearing in inclusion bodies (42).
The soluble cytoplasmic fraction of E. coli LM3115 cells containing the expressed WT and mutated RrF 1 ␤ was subjected to ammonium sulfate fractionation as described for native RrF 1 ␤ (34). The 30 -60% ammonium sulfate precipitate was dissolved in TGN buffer to a concentration of about 10 mg of protein/ml and thoroughly dialyzed against the same buffer. The WT and ␤-E195G mutant were further purified by dyeligand chromatography on a Red A column. As is illustrated in Fig. 1, the dialyzed WT ␤ subunit was loaded on a column preequilibrated in TGN buffer at 4°C. After 1 h, the column was washed with TGN buffer, to elute all the unbound protein (Fig. 1, A, peak I, and B, lane 3). The RrF 1 ␤ was then eluted with TGN buffer containing 1 mM MgATP (Fig. 1, A, peak II,  and B, lane 4). Further washing with TGN buffer containing 1 mM MgATP and 1.5 M NaCl did not elute any additional RrF 1 ␤ subunit (Fig. 1, A, peak III, and B, lane 5).
This column yielded a highly purified expressed WT RrF 1 ␤ (Fig. 1B, lane 4), which contained, of course, no trace of RrF 1 ␣. On the other hand, the earlier prepared native RrF 1 ␤, which was removed by LiCl extraction of R. rubrum chromatophores (25) and purified by chromatography through two ion-exchange columns (26), was recently found to contain 5-10% of RrF 1 ␣ (28, 35). Most of this ␣ subunit was removed by dye-ligand chromatography, leaving a pure native RrF 1 ␤ containing less than 1% of RrF 1 ␣.
Activities of the Highly Purified Native, Expressed WT and E195G RrF 1 ␤-All three types of RrF 1 ␤ subunits were found to bind, in the presence of Mg 2ϩ , up to 2 mol of ATP/mol of ␤. But, whereas the binding of ATP to the expressed WT as to pure native RrF 1 ␤ (Fig. 2), saturated at 200 M ATP and showed a pronounced cooperativity, with a Hill coefficient (n H ) of 2.2, the RrF 1 ␤-E195G mutant required a higher concentration of ATP for saturation and showed a much lower cooperativity with a Hill coefficient of only 1.5. The native RrF 1 ␤, which contained at least 5% of RrF 1 ␣ (28, 35), was earlier shown to bind up to 2 mol of ATP/mol (23) but with somewhat different properties than those reported here for the highly purified native ␤ subunit.
The difference, between the native RrF 1 ␤ containing 5% of RrF 1 ␣ and the highly purified one containing ϳ1% RrF 1 ␣, was much more pronounced in assays of assembly into ␤-less chromatophores and restoration of their lost activities. RrF 1 ␤ with at least 5% of RrF 1 ␣ could restore these activities (35), whereas pure native and expressed WT RrF 1 ␤, containing either 1% or no ␣, respectively, could not restore any ATP synthesis by ␤-less chromatophores (Fig. 3). They could, however, restore high rates of ATP synthesis when a fixed trace amount of RrF 1 ␣, which by itself was completely inactive, was added to the reconstitution mixture together with increasing amounts of the ␤ subunits. Under these conditions, the degree of restoration of ATP synthesis (Fig. 3) and hydrolysis (not shown) was dependent on the amount of RrF 1 ␤ present during the reconstitution of a fixed amount of ␤-less chromatophores. An identical saturation curve, which leveled off around a ratio of 1 g of added RrF 1 ␤/1 g of Bchl, was obtained for both pure native and expressed WT RrF 1 ␤ (Fig. 3).
The RrF 1 ␤-E195G mutant exhibited the same pattern of concentration-dependent saturation curve, in the presence of a fixed trace amount of RrF 1 ␣, but the restored rate was much lower, reaching only 30% of the rate restored by pure native or expressed WT RrF 1 ␤ (Fig. 3). This lower rate was not due to a lower capacity of the E195G mutant to bind into ␤-less chromatophores. Fig. 4 illustrates that reconstitution of these chromatophores with the same, saturating amounts of WT or the E195G mutant, resulted in binding of similar amounts of RrF 1 ␤ (Fig. 4, lanes 2 and 3). Moreover, the other two RrF 1 ␤-E195Q and -E195K mutants were also found to bind in similar amounts into the ␤-less chromatophores (Fig. 4, lanes 4 and 5). The presence of ␣ in these ␤-less chromatophores (Fig. 4, lane  6) indicates that the additional trace of soluble RrF 1 ␣, required for an effective reconstitution of all types of tested RrF 1 ␤ (Fig.  3), might indeed exert a chaperonin-like activity, keeping the pure RrF 1 ␤ correctly folded during reconstitution, as has earlier been suggested for CF 1 ␣ (43).

Divalent Cation-dependent Restoration of ATP Hydrolysis and Synthesis by WT and Several RrF 1 ␤-E195 Mutants-
The similar binding of WT and the E195G mutant RrF 1 ␤ to ␤-less chromatophores, but the much lower ability of the mutant to restore ATP synthesis, suggested that ␤-E195 plays an important role in the catalytic activity of the ATP synthase. To test this possibility, we compared the capacity of WT and all three ␤-E195 mutants to restore ATP hydrolysis (Fig. 5). ATP hydrolysis, unlike its synthesis, was shown to occur in control R. rubrum chromatophores in the presence of Ca 2ϩ as well as Mg 2ϩ and Mn 2ϩ but was coupled to proton translocation only in the presence of the last two divalent cations (44). Reconstitution of ␤-less chromatophores with WT RrF 1 ␤ restored their capacity to hydrolyze ATP with all three divalent cations (Fig.  5, A-C). As in control chromatophores (44), hydrolysis in the presence of Mg 2ϩ and Mn 2ϩ showed a narrow range of dependence on the cation concentration, resulting in maximal activity at cation/ATP ratios around 0.5 followed by a drastic inhibition at a cation/ATP ratio of 2 (Fig. 5, A and B). The CaATPase activity in control (44), as well as WT RrF 1 ␤ reconstituted, chromatophores (Fig. 5C) was much less sensitive to inhibition by an excess of free Ca 2ϩ -ions, remaining optimal at cation/ ATP ratios between 0.5 and 2.0. Even with 32 mM CaCl 2 , at a cation/ATP ratio of 8, this restored ATPase activity was inhibited by only 35%.
Chromatophores reconstituted with the ␤-E195 mutants showed very little MgATPase or MnATPase activities at the whole range of tested cation concentrations (Fig. 5, A and B). They did, however, exhibit a CaATPase activity that required higher CaCl 2 concentrations than the activity restored with WT RrF 1 ␤. At 32 mM CaCl 2 , the activities restored by RrF 1 ␤-E195G, -E195Q, or -E195K mutants reached 105, 45, and 16%, respectively, of the activity restored with WT RrF 1 ␤ already at 4 mM CaCl 2 (Fig. 5C). These results suggest that the Glu-195 mutants also might require higher concentrations of MgCl 2 and MnCl 2 for restoration of ATP hydrolysis. But, since high concentrations of these cations are already inhibitory (Fig. 5, A  and B), the mutants could not restore MgATPase or MnATPase activity.
ATP synthesis in control chromatophores was found to be much less sensitive than the MgATPase activity to inhibition by high MgCl 2 concentrations, remaining optimal at MgCl 2 / ADP ratios between 0.4 and 4.0 (44). Furthermore, ␤-less chromatophores reconstituted with the ␤-E195G mutant could Reconstitution was carried out as described under "Experimental Procedures." The reconstituted chromatophores were centrifuged to remove all unbound RrF 1 ␤ subunits and resuspended in buffer containing 50 mM Tricine-NaOH, pH 8.0, and 10% glycerol. Chromatophores corresponding to 3 g of Bchl were incubated with 1% SDS at 100°C for 5 min and applied to SDS-PAGE. After electrophoresis, the gel was transferred to nitrocellulose (33) and probed with antisera produced against native RrF 1 ␣ and ␤ subunits. Lane 1, control chromatophores; lanes 2-5, chromatophores reconstituted with WT, E195G, E195Q, and E195K RrF 1 ␤, respectively; lane 6, unreconstituted ␤-less chromatophores. clearly restore some ATP synthesis (Fig. 3). We have therefore investigated the Mg 2ϩ requirement for restoration of ATP synthesis in chromatophores reconstituted with WT and all three RrF 1 ␤-E195 mutants. Fig. 6 does indeed demonstrate that all tested RrF 1 ␤-E195 mutants can restore at least some Mg-dependent ATP synthesis but that they require higher concentrations of MgCl 2 than the activity restored with WT RrF 1 ␤. In this reaction, unlike in CaATPase (Fig. 5C), both ␤-E195G and ␤-E195Q could restore at 5 mM MgCl 2 up to 60%, whereas ␤-E195K restored less than 25%, of the WT restored ATP synthesis (Fig. 6). However, higher concentrations of 10 mM MgCl 2 already started to inhibit the WT restored rate, and 40 mM MgCl 2 inhibited by about 50% the ATP synthesis restored by the WT and all three RrF 1 ␤-E195 mutants (Fig. 6).
The results summarized in Figs. 5 and 6 have indicated that the inability of the RrF 1 ␤-E195 mutants to restore MgATPase and MnATPase activities while restoring ATP synthesis might be due to their requirement of high concentrations of Mg 2ϩ and Mn 2ϩ ions, which inhibit ATP hydrolysis much more than its synthesis. Since anions, such as sulfite were found to stimulate the MgATPase activity of RrF 1 by relieving the inhibition caused by excess free Mg 2ϩ ions (45), we have compared the effect of sulfite on the MgATPase and MnATPase activities in chromatophores reconstituted with WT and the E195 mutant RrF 1 ␤ (Fig.  7, A and B). Addition of 20 mM sulfite stimulated by 3-4-fold both restored ATPase activities in chromatophores reconstituted with WT RrF 1 ␤, reducing the inhibition by high concentrations of MgCl 2 and MnCl 2 and shifting the cation/ATP ratio for maximal activity to 1 (compare Fig. 5, A and B, and Fig. 7, A and B). This effect of sulfite was already saturated at 20 mM since 100 mM caused no further stimulation of the WT RrF 1 ␤ restored activities. Addition of sulfite to chromatophores reconstituted by ␤-E195G did indeed uncover both MgATPase and MnATPase activities; but in this case, 100 mM sulfite was much more effective than 20 mM, and maximal activity was obtained at a cation/ ATP ratio of 3. With ␤-E195Q and ␤-E195K, even 100 mM sulfite did not uncover any MgATPase or MnATPase activities (Fig. 7, A and B), and higher sulfite concentrations did already inhibit the activities restored by WT RrF 1 ␤.

DISCUSSION
Glutamic acid 195 of RrF 1 ␤ is equivalent to MF 1 ␤-Glu-199, which in the bovine heart crystal structure (8) is described as pointing into the conical tunnel leading to the nucleotide binding sites. This fully conserved glutamic acid residue was shown to bind the inhibitor DCCD in MF 1 ␤ (46) and EcF 1 ␤ (47). But a mutation in the equivalent EcF 1 ␤-E192 to Gln was found to grow on succinate and showed normal assembly of F 0 F 1 , with only 6 -8-fold inhibition of its membrane ATPase activity and small perturbations in rate constants and equilibrium constants of unisite catalysis (48). It was, therefore, suggested that the full inhibition of F 0 F 1 by DCCD could be due to the presence of the two bulky cyclohexyl moieties in the conical tunnel (8).
The system developed in R. rubrum chromatophores enables reconstitution of ␤-less chromatophores with RrF 1 ␤ and detailed characterization of ATP hydrolysis as well as synthesis restored to the same preparation of reconstituted chromatophores (25,26). This system was used here for a direct comparison of the activities restored to ␤-less chromatophores reconstituted with WT and three Glu-195 mutants of RrF 1 ␤ and provided rather unexpected results. ATP synthesis was restored to about 60% by ␤-E195G and ␤-E195Q and to 30% by ␤-E195K (Fig. 6), whereas the Mg-and Mn-dependent ATP hydrolysis was not restored by E195Q and E195K mutants, even when assayed in the presence of saturating concentrations of sulfite (Fig. 7, A and B). The E195G mutant could restore some MgATPase and MnATPase activities in the presence of 100 mM sulfite, but they showed different divalent cation dependences and reached a much lower maximal rate than the activities restored by WT RrF 1 ␤.
Similar properties were recently reported for the MgATPase activity of S-carboxymethyl-␤185 EcF 1 (49). This residue is equivalent to MF 1 ␤-E192, which is located at the catalytic nucleotide binding sites of F 1 , whereas the RrF 1 ␤-E195 residue is pointing into the tunnel leading to these sites (8). Our results indicate that residues within the tunnel might also affect divalent cation and nucleotide binding, suggesting their participation in the interconversion of the catalytic sites between different conformational states. Interestingly, two earlier reports describing mutations in EcF 1 ␤-E192 to V (50) and TF 1 ␤-E201 to Q (51), which are equivalent to RrF 1 ␤-E195, have observed some type of interaction between this glutamic acid and EcF 1 -G149 or TF 1 ␤-E190, which are located in the catalytic site (8). It would be worthwhile to check whether other amino acid residues that are located in the tunnel show similar effects.
The strange results obtained with the RrF 1 ␤-E195 mutants, which enable restoration of ATP synthesis but not hydrolysis, could be explained by differences in affinity of the catalytic sites to MgADP as compared with MgATP. These differences are especially important during ATP hydrolysis in photosynthetic organisms, which is highly regulated to limit wasteful hydrolysis of low concentrations of ATP in the dark (3)(4)(5). One aspect of this regulation involves the onset of inhibition of ATPase activity, and of exchange of ADP tightly bound at a catalytic site with medium nucleotides, by added free Mg 2ϩ (52). Sulfite has been shown to overcome the Mg 2ϩ -induced inhibition of both ATPase activity and release of the tightly bound MgADP (53,54). Addition of a saturating sulfite concentration did indeed enable low MgATPase and MnATPase activities in the ␤-E195G mutant, but not in ␤-E195Q or ␤-E195K (Fig. 7, A and B). These mutations might impede more severely the release of the inhibitory MgADP as compared with the release of the more loosely bound MgATP.
The overall higher activities restored by ␤-E195G (Figs. 5C, 6, and 7) can be explained by the capacity of glycine to adopt a much wider range of conformations than the other residues and thus allow unusual main chain conformations in proteins (55). In RrF 1 ␤-E195G, it could enable conformational changes that compensate the change of charge and size, and lead to the higher activity of this mutant.
An additional interesting effect observed with the RrF 1 ␤-E195G and -E195Q, is their capacity to restore CaATPase, but not MgATPase and MnATPase activities (compare Fig. 5, A, B, and C). CaATPase, unlike the MgATPase and MnATPase activity, is not coupled to proton translocation and not subject to the tight regulation observed with MgATP and MnATP hydrolysis (44). Moreover CaCl 2 does not enable ATP synthesis. So the binding of Ca 2ϩ to the catalytic sites on RrF 1 ␤ might convert it to a different conformational state that is less impeded by the Glu-195 mutations. Work with RrF 1 ␤-T159 mutants has provided direct experimental support for this conclusion. 2 It would be most interesting to crystallize F 1 in the presence of CaCl 2 since any differences between the CaCl 2 -and MgCl 2 -containing crystals might illuminate the domains, and/or specific amino acid residues, that are involved in the coupling of proton translocation to ATP synthesis and hydrolysis. This information is essential for the understanding of the mechanism of ATP synthesis.