Mutations in the β-Subunit Thr159 and Glu184 of the Rhodospirillum rubrumF0F1 ATP Synthase Reveal Differences in Ligands for the Coupled Mg2+- and Decoupled Ca2+-dependent F0F1 Activities*

In the crystal structure of the mitochondrial F1-ATPase, the β-Thr163 residue was identified as a ligand to Mg2+ and the β-Glu188 as directly involved in catalysis. We replaced the equivalent β-Thr159 of the chromatophore F0F1 ATP synthase of Rhodospirillum rubrum with Ser, Ala, or Val and the Glu184 with Gln or Lys. The mutant β subunits were isolated and tested for their capacity to assemble into a β-less chromatophore F0F1 and restore its lost activities. All of them were found to bind into the β-less enzyme with the same efficiency as the wild type β subunit, but only the β-Thr159 → Ser mutant restored the activity of the assembled enzyme. These results indicate that both Thr159and Glu184 are not required for assembly and that Glu184 is indeed essential for all the membrane-bound chromatophore F0F1 activities. A detailed comparison between the wild type and the β-Thr159 → Ser mutant revealed a rather surprising difference. Although this mutant restored the wild type levels and all specific properties of this F0F1 proton-coupled ATP synthesis as well as Mg- and Mn-dependent ATP hydrolysis, it did not restore at all the proton-decoupled CaATPase activity. This clear difference between the ligands for Mg2+ and Mn2+, where threonine can be replaced by serine, and Ca2+, where only threonine is active, suggests that the β-subunit catalytic site has different conformational states when occupied by Ca2+ as compared with Mg2+. These different states might result in different interactions between the β and γ subunits, which are involved in linking F1 catalysis with F0proton-translocation and can thus explain the complete absence of Ca-dependent proton-coupled F0F1catalytic activity.

The F 0 F 1 ATP synthase-ATPase complexes, found in the inner membranes of mitochondria and chloroplasts and in bacterial plasma membranes, couple ATP synthesis and hydrolysis to electrochemical proton gradients. These complexes are composed of a membrane-intrinsic F 0 sector, which mediates proton translocation, and an extrinsic F 1 sector, which carries the catalytic sites. All isolated F 1 complexes are composed of five subunits with a stoichiometry of ␣ 3 ␤ 3 ␥␦⑀ (1-6).
The crystal structure of bovine mitochondrial MF 1 (7) pre-sents the large ␣ and ␤ subunits arranged alternately in a closed hexamer around the N-and C-terminal helices of the ␥ subunit. The three catalytic ␤ subunits show a clear difference in bound nucleotides resulting in different conformational states. The asymmetric structure imposed on this hexamer by the interaction of the ␥-subunit with these different ␤-subunits, supports the binding change mechanism (8), which proposed that ATP synthesis and hydrolysis involve transitions between different but interacting catalytic sites via rotation of the ␥ subunit relative to an ␣ 3 ␤ 3 subassembly. Several models suggested that proton-translocation through F 0 results in a coupled rotation of the F 0 -c and F 1 -␥ subunits (8 -11). However full elucidation of the detailed mechanism of action of the F 0 F 1 ATP synthase will depend on identification of the specific residues and/or whole domains that participate in proton-coupled ATP synthesis and hydrolysis as well as in the regulation of these reversible activities.
Tight regulation of ATP hydrolysis is especially important in photosynthetic organisms where it prevents the depletion of essential cellular ATP pools in the dark (3)(4)(5). One stringent regulatory pathway operating in plant chloroplasts (12) as well as in bacterial chromatophores (13,14) is their high sensitivity to inhibition by excess free Mg 2ϩ ions, which results in optimal MgATPase activity at Mg 2ϩ /ATP ratios around 0.5 and its drastic decrease at higher ratios. F 1 -␣, ␤, and ␥ subunits from photosynthetic sources are therefore very interesting targets for mutational analysis, based on the available MF 1 crystal structure, of amino acid residues participating in coupled catalysis and its regulation.
The F 1 -␣ and ␤ subunits of the photosynthetic bacterium Rhodospirillum rubrum are most suitable for such studies. RrF 1 ␤ 1 was isolated in large amounts from the chromatophore membrane-bound RrF 0 F 1 by a specific LiCl treatment (15,16) and recently also cloned and expressed in soluble form (17) in Escherichia coli cells lacking the whole unc operon. The recombinant WT RrF 1 ␤ was found to be as active as the native ␤-subunit (18) in a large number of earlier developed in vitro assays, including the binding of nucleotides (19) and rebinding to the ␤-less RrF 0 F 1 resulting in restoration of both ATP synthesis and hydrolysis activities (15,16). The rebinding of the highly purified native and WT RrF 1 ␤ subunits was, however, found to require the presence of small amounts of RrF 1 ␣ (18), which were released with the bulk of the ␤ subunit from the LiCl-treated chromatophores (17,20,21). This finding provided a direct assay also for the isolated ␣-subunit. It has been used to follow the refolding of the recombinant RrF 1 ␣, expressed only in insoluble inclusion bodies (22), into a functional monomer that assembles with the WT RrF 1 ␤ monomer into active ␣ 1 ␤ 1 dimers (23). If the released ␣/␤ ratio can be controlled, the ␣-depleted ␤-less chromatophores together with the recombinant RrF 1 ␣ and ␤ monomers could provide suitable systems for studying the effect of mutagenized ␣ as well as ␤ on the in vitro assembly and activity of both the membrane-bound F 0 F 1 and soluble F 1 complexes.
In this investigation we have defined the conditions for LiCl treatment of R .rubrum chromatophores that release the bulk of their RrF 1 ␤ together with specific amounts of RrF 1 ␣. ␤-less chromatophores, which lost at least a third of this ␣ subunit were used for preparing hybrid RrF 0 F 1 /CF 1 complexes containing either only CF 1 ␤ or CF 1 ␤ and at least one copy of CF 1 ␣ (50). ␤-less chromatophores containing 90% of their ␣ subunit were used here for testing the effect of mutations in RrF 1 ␤-Thr 159 and Glu 184 , which are equivalent to MF 1 ␤-Thr 163 and Glu 188 in the catalytic nucleotide binding site (7). These fully conserved F 1 ␤ residues have been mutated only in respiratory F 1 -ATPases (6), except for the equivalent CF 1 ␤-Thr 168 of Chlamydomonas reinhardtii, which has recently been mutated to serine and shown to increase dramatically the MgATPase activities of the soluble CF 1 and ␣ 3 ␤ 3 ␥ complexes (24). Our studies revealed that in the membrane-bound F 0 F 1 the RrF 1 ␤-T159S, but not the T159A or T159V, could restore the proton-coupled Mg-and Mn-dependent ATP synthesis and hydrolysis activities to the extent restored by WT RrF 1 ␤. Moreover, even the active ␤-T159S did not restore the proton-decoupled CaATPase activity. These results indicate that the conserved WT ␤-Thr 159 is an absolutely essential ligand for Ca 2ϩ , which could not be replaced even by serine. The RrF 1 ␤-E184Q and E184K mutants did bind to the ␤-less chromatophores, but the assembled mutant RrF 0 F 1 s were completely inactive.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-R. rubrum cells were grown as described previously (15). The E. coli LM3115 strain (25) lacking the unc operon was used as host for the recombinant plasmids (17) carrying the WT and mutated RrF 1 ␤ genes. This strain was found to express large amounts of RrF 1 ␤ as a soluble protein when grown in LB medium, supplemented as described by Nathanson and Gromet-Elhanan (18), at 22°C to about A 0.65 .
Site-directed Mutagenesis-The RrF 1 ␤-Thr 159 mutants were obtained by a modification of the PCR-based mutagenesis method of Ho et al. (26). For each mutation the start and end fragments of the cloned WT RrF 1 ␤ gene (17) were amplified separately in two independent PCR reactions, using the start-forward primer with the earlier introduced EcoRI site (17), and the reverse primer with a newly introduced BamHI site. Each PCR reaction contained a complementary mutagenic primer (26). The following mutagenic primers for T159V, T159A, and T159Sforward were, respectively: 5Ј-C-GTC-GGC-AAG-GTA-GTA-CTG-ATC-CAG-G-3Ј; 5Ј-C-GTC-GGC-AAG-GCA-GTA-CTG-ATC-CAG-G-3Ј and 5Ј-C-GTC-GGC-AAG-TCA-GTA-CTG-ATC-CAG-G-3Ј; and for T159V, T159A and T159S-reverse were, respectively: 5Ј-C-CTG-GAT-C AG-TAC-TAC-CTT-GCC-GAC-G-3; 5Ј-C-CTG-GAT-CAG-TAC-TGC-CTT-GCC-GAC-G-3Ј and 5Ј-C-CTG-GAT-CAG-TAC-TGA-CTT-GCC-GAC-G-3Ј. Each primer contained a new, underlined site for ScaI, introduced by two base changes, indicated by bold letters as the bases changed to give the Val, Ala, or Ser codons. Amplification was carried out as described earlier (17), except that 50 pmol of each set of primers, 4 ng of pBSKS ϩ -WT␤, and 1.25 units of Pwo DNA polymerase were used, in the buffer supplied with the enzyme. The two amplified fragments of each mutated gene were cut by EcoRI and ScaI or ScaI and BamHI, ligated with each other and with the pBSKS ϩ plasmid between its EcoRI and BamHI sites. The resulting recombinant plasmids were transformed into E. coli HB101 cells (17), and plasmids from ampresistant colonies were screened for the mutations by ScaI restriction analysis and by full DNA sequencing. All three mutants had only the above stated nucleotide changes.
For mutagenesis of RrF 1 ␤-Glu 184 , the pBSKS ϩ -WT␤ plasmid (17) was transformed into E. coli CJ236 and single-stranded uracil-contain-ing DNA was prepared after infection with helper phage VCS M13 (27). Site-directed mutagenesis was performed as described by Kunkel et al. (28), using the following oligonucleotides to introduce the ␤-E184Q and E184K mutations, respectively: 5Ј-GCC-CTC-ACG-CGT-GCG-CTG-GCC-GAC-GCC-G and 5Ј-GCC-CTC-ACG-CGT-GCG-CTT-GCC-GAC-GCC-G. They also contained a new, underlined site for MluI, introduced by a single base change, indicated by a bold letter as the bases changed to give the Glu or Lys codons. The mutagenized DNA was transformed into E. coli HB101 cells, and the mutations were confirmed by MluI restriction analysis and full DNA sequencing. The ␤-E184K gene had only the stated mutation. However, ␤-E184Q had one additional change in nucleotide 249, altering the codon from GGC to GGT, both coding for glycine.
All the expressed WT and mutant RrF 1 ␤ subunits were isolated from the cytoplasmic fraction of the E. coli LM3115 cells and purified as described previously (18).
Preparation of LiCl-treated R. rubrum Chromatophores and Determination of the Amounts of Released RrF 1 ␣ and ␤ Subunits-LiCl treatment of the chromatophores was carried out as outlined by Gromet-Elhanan and Khananshvili (16) with the following modifications. 1 mM protease inhibitor PMSF was present in the final 1.9 M LiCl buffer, and the concentration of the treated chromatophores was varied as stated in the text. The LiCl supernatant was separated from the treated chromatophores, and the dissolved 60% ammonium sulfate precipitate (16) was subjected to SDS-PAGE (29,30), transferred to nitrocellulose (31), and probed with antibodies raised against RrF 1 ␣ and ␤ subunits.
Other Procedures-Chromatophores treated with 1.9 M LiCl at 1.2 mg of BChl/ml were washed to remove all traces of LiCl (16), reconstituted with WT or mutated RrF 1 ␤ in presence of RrF 1 ␣ at a ratio of ␣/␤ of 0.2, and assayed for restored ATP synthesis and hydrolysis as described previously (18). Published methods were used for measurements of protein concentration (32,33) and the BChl content of chromatophores (34).
Materials-E. coli LM3115 was a gift of Dr. P. R. Jensen (The Netherlands Cancer Institute, Amsterdam). Oligonucleotides were synthesized by Dr. Ora Goldberg (Biological Services, Weizmann Institute of Science, Rehovot, Israel). Restriction enzymes, T4 DNA polymerase, and ligase were from New England Biolabs. Pwo DNA polymerase, dNTP, and pBTacI were purchased from Roche Molecular Biochemicals. Plasmid pBSKS ϩ and helper phage VCS M13 were from Stratagene. [ 32 P]P i was obtained from the Nuclear Research Center, Negev, Israel. All other reagents were of the highest purity available.

Preparation of ␤-less
Chromatophores Containing about 90% of the RrF 0 F 1 ␣ Subunit-Earlier preparations of ␤-less chromatophores were obtained by treatment with 2 M LiCl and no added protease inhibitors (16). Similar treatments of spinach (35) as well as lettuce and tobacco (36) chloroplasts were also found to release some CF 1 ␣ together with all the CF 1 ␤. Furthermore Western immunoblots probed with antibodies raised against spinach CF 1 ␣ and ␤ subunits revealed that some of the released CF 1 ␣ was nicked by proteases and ran in SDS-PAGE together with the ␤ subunit. This proteolysis was fully blocked only in the presence of a mixture of three protease inhibitors (37).
Since antibodies against CF 1 ␣ show no cross-reaction with RrF 1 ␣ (50), we have raised antibodies against RrF 1 ␣ and found proteolyzed RrF 1 ␣ in LiCl extracts of earlier treated chromatophores. Addition of 1 mM PMSF blocked completely this proteolysis, thus enabling a clear determination of the relative amounts of the RrF 1 ␣ and ␤ subunits released from chromatophores treated by 1.9 M LiCl at 0.4, 0.8, or 1.2 mg of BChl/ml (Fig. 1, lanes 2-4). About 0.65, 0.32, and 0.13 of their RrF 1 ␣ were, respectively, released as compared with practically all their ␤ subunit ( Fig. 1 and Fig. 2, lane 5). Chromatophores treated with LiCl at 1.2 mg of BChl/ml were used for evaluating the effect of the ␤-T159 and ␤-E184 mutants when reconstituted with the various ␤ subunits in presence of RrF 1 ␣ at a ratio of ␣/␤ of 0.2.
Activities of RrF 1 ␤ Mutants at Position 159 -RrF 1 ␤-Thr 159 is the last residue in the glycine rich p-loop sequence found in ␣ and ␤ subunits of all F 1 -ATPases and many other nucleotidebinding proteins (38). The parallel MF 1 ␤ O ␥ -T163 was identified in the crystal structure as a ligand to Mg 2ϩ (7). In the GTP-binding Ras protein, a serine, which replaces the threonine in this sequence, was also found to be a ligand to Mg 2ϩ (39). We have mutated the RrF 1 ␤-Thr 159 into serine (␤-T159S) as well as alanine (␤-T159A) or valine (␤-T159V). All of them were found to assemble into the ␤-less R. rubrum chromatophores as efficiently as WT RrF 1 ␤ (Fig. 2, lanes 1-4). However, the RrF 1 ␤-T159V and T159A mutants were unable to restore any ATP synthesis or hydrolysis activities, whereas the RrF 1 ␤-T159S mutant restored ATP synthesis as well as Mg-and Mn-dependent ATP hydrolysis (Figs. 3 and 4).
The ATP synthesis restored by the ␤-T159S was similar or slightly higher than the WT rate and showed a similar MgCl 2 requirement (Fig. 3). Values of K m of 173 and 238 M MgCl 2 and V max of 352 and 429 mol of ATP formed/h per mg of BChl were calculated for chromatophores reconstituted with the WT and the T159S mutant, respectively. The maximal rate of this restored ATP synthesis, as that observed in control R. rubrum chromatophores (14), was obtained at MgCl 2 concentrations below 10 mM and Mg 2ϩ /ADP ratios below 5. At 40 mM MgCl 2 , the rates restored by the WT and T159S mutant were inhibited by 50% and 30%, respectively (Fig. 3). The lower sensitivity of the mutant to inhibition by increasing MgCl 2 concentrations could explain its capacity to restore somewhat higher maximal rates of ATP synthesis.
A very similar pattern was observed also for restoration of MgATPase activities by the WT and T159S mutant (Fig. 4A). Mg-and Mn-dependent ATP hydrolysis in control R. rubrum chromatophores (14), as in ␤-less chromatophores reconstituted with WT RrF 1 ␤ (18), were reported to be much more sensitive to inhibition by excess free Mg 2ϩ or Mn 2ϩ ions than their respective ATP synthesis activities. This difference in sensitivity provides the basis for the tight regulation of ATP hydrolysis in photosynthetic organisms, which enables them to limit the depletion of essential cellular ATP pools in the dark (2)(3)(4)(5). Chromatophores reconstituted with the RrF 1 ␤-T159S mutant retained this tight regulation more efficiently in presence of MgCl 2 than in presence of MnCl 2 (Fig. 4, compare A and B).
Control chromatophores show no ATP synthesis in the presence of Ca 2ϩ , and their CaATPase, unlike the Mg-and Mn-dependent ATPase activities, is not coupled to proton-translocation. It is instead decoupled, since the presence of CaATP does not inhibit the light-induced, electron transport-triggered proton translocation (14). It was therefore rather surprising that the ␤-T159S mutant, which restored the proton-coupled ATP synthesis as well as the Mg-and MnATPase activities as efficiently as the WT RrF 1 ␤ (Figs. 3 and 4, A and B), did not restore at all the proton-decoupled CaATPase activity (Fig. 4C). ␤-T159S was in this specific assay similar to the completely inactive ␤-T159V and T159A mutants rather than to the WT RrF 1 ␤. These results revealed a clear difference between the ligands for Mg 2ϩ and Mn 2ϩ , where threonine could be replaced by serine, and for Ca 2ϩ , where only threonine is active.
Anions, such as sulfite, were found to stimulate the Mg-ATPase activities of control (40) as well as ␤-less chromatophores reconstituted with the WT RrF 1 ␤ (18). The MgATPase activity restored by the RrF 1 ␤-T159S mutant was also stimulated by sulfite although somewhat less than that restored by the WT ␤ subunit (Fig. 5). A similar effect was observed also with the MnATPase activity (Table I). The two other ␤-T159A and ␤-T159V mutants did not restore any activity even in the presence of 100 mM sulfite (Fig. 5).
Activities of RrF 1 ␤ Mutants at Position 184 -The x-ray structure of MF 1 showed a density for a water molecule, hydrogen-bonded to the equivalent MF 1 ␤-Glu 188 , at a distance of 4.4 Å from the ␥-phosphate of the ␤ TP (7). This residue could therefore be appropriately positioned to activate the water molecule for an inline nucleophilic attack on the ␥-phosphate during ATP hydrolysis. We have mutated the RrF 1 ␤-Glu 184 into glutamine (␤-E184Q) and lysine (␤-E184K), and both mutants were found to bind to the ␤-less chromatophores as efficiently as the WT RrF 1 ␤ (data not shown). However, no restoration of any tested activity was obtained with either mutant when measured under conditions found optimal for WT RrF 1 ␤ (Table I). There was also no restoration of any ATP synthesis or hydrolysis activity at higher concentrations of divalent cations, which enabled some restoration of activity in the RrF 1 ␤-E195Q Reconstitution of the ␤-less chromatophores obtained by LiCl treatment at 1.2 mg of BChl/ml was carried out as described under "Experimental Procedures." Washed reconstituted chromatophores corresponding to 3 g of BChl were incubated with 1% SDS at 100°C for 5 min, applied on SDS-PAGE, and probed with antibodies raised against RrF 1 ␣ and ␤ subunits as described in Fig. 1. and E195K mutants (18). These results demonstrate that the RrF 1 ␤-Glu 184 , although not required for assembly, is absolutely essential for all the above tested membrane-bound RrF 0 F 1 -activities. DISCUSSION RrF 1 ␤-Thr 159 and Glu 184 are fully conserved in all sequenced F 1 complexes. The equivalent MF 1 ␤-Thr 163 was identified in the catalytic nucleotide binding site of the bovine MF 1 crystal structure (7) as a ligand to Mg 2ϩ , and the MF 1 ␤-Glu 188 was suggested to be directly involved in catalysis. Our analysis of the equivalent RrF 1 ␤-E184Q and E184K mutants yielded assembled but fully inactive enzymes (Table I). Similar results were reported for the equivalent EcF 1 ␤-E181 mutations (41,42), where even the ␤-E181D mutant had only 1.4% of the control MgATPase activity. Only the equivalent TF 1 ␤-E190D mutant (43) had 7% of the WT MgATPase activity. These results confirm that this residue is essential for catalysis in respiratory as well as photosynthetic F 0 F 1 -ATP synthases, but do not really clarify its function (see Ref. 6). Residues equivalent to RrF 1 ␤-Thr 159 were mutated to serine in several respiratory systems (44 -46) and recently also in C. reinhardtii (24), and checked extensively on the MgATPase activity of isolated F 1 or ␣ 3 ␤ 3 ␥ complexes. In all tested cases this mutation resulted in a 3-10-fold increase in the MgATPase activity. It also eliminated the stimulation of the MgATPase by oxyanions or alcohols, and reduced the sensitivity of the MgATPase activity to inhibition by azide.
The three mutants RrF 1 ␤-T159S, T159A, and T159V have, however, been tested for their in vitro assembly into the ␤-less membrane-bound RrF 0 F 1 as well as for their capacity to restore all the divalent cation-dependent ATP synthesis and hydrolysis activities of the assembled RrF 0 F 1 complex. All three mutants did bind into the ␤-less chromatophores (Fig. 2), but only the assembled RrF 0 F 1 containing the ␤-T159S was active. It did restore the WT rates and specific properties of ATP synthesis as well as Mg-and Mn-dependent ATP hydrolysis, including the tight regulation of these ATPase activities, and their effective stimulation by sulfite. However, even this active mutant could not restore any CaATPase activity.
Two unexpected results emerged from the present study with the RrF 1 ␤-T159S. One is the large difference between the membrane-bound activities and properties obtained here with the mutant, which were very similar to those obtained with WT RrF 1 ␤, and the much higher activities and different properties than the WT ones that were earlier observed in soluble F 1 or ␣ 3 ␤ 3 ␥ complexes containing the equivalent threonine to serine mutants of CF 1 (24) MF 1 (44), and TF 1 (46).
The second unexpected result is the inability of the RrF 1 ␤-T159S mutant to restore the proton-decoupled CaATPase, while effectively restoring the proton-coupled Mg-and Mn-ATPases. The capacity to restore CaATPase has not been tested with any other equivalent mutant, but an opposite effect of restoration of Ca-but not MgATPase activity was earlier observed with the RrF 1 ␤-E195G mutant (18). These results indicate that the ligands for Ca 2ϩ and Mg 2ϩ must be different, since only in the case of Ca 2ϩ the threonine cannot be replaced  in ␤-less chromatophores reconstituted with WT and RrF 1 ␤-T159 and ␤-E184 mutants Conditions for reconstitution of the ␤-less chromatophores were as described under "Experimental Procedures." Restored ATP synthesis was assayed as described in Fig. 3, with 5 mM MgCl 2 , and ATP hydrolysis as described in Fig. 4 by serine (Fig. 4C). This difference in ligands might lead to a different conformational state of the catalytic site occupied by Ca 2ϩ as compared with Mg 2ϩ or Mn 2ϩ . It can explain the inability of the Ca 2ϩ -occupied catalytic sites to carry out ATP synthesis and proton-coupled ATP hydrolysis, by Ca-induced changes in the interactions between the ␤ and ␥ subunits that may be involved in linking catalysis to proton translocation. Omote et al. (47) have recently reported that they observed similar rotation and torque generation in engineered WT EcF 1 -␣ 3 ␤ 3 ␥ and an uncoupled mutant EcF 1 -␣ 3 ␤ 3 ␥M23K. This mutant is unable to translocate protons through F 0 but has about 60% of the WT EcF 1 -MgATPase activity. These properties are very similar to those recorded with the proton-decoupled membrane-bound and the soluble RrF 1 -CaATPase activity. It would therefore be most interesting to test whether rotation can occur also during CaATPase as well as MgATPase activity. Although CaATPase activity has been tested very thoroughly in various photosynthetic F 1 -ATPases (2-5), however, it has not been studied in detail in respiratory bacterial F 1 -ATPases, whereas a direct full rotation of ␥ has not been demonstrated as yet with any engineered photosynthetic F 1 -␣ 3 ␤ 3 ␥. Incubation of the recombinant WT RrF 1 ␣ and ␤ monomers (23) with a recombinant spinach CF 1 ␥ subunit (48,49) has recently been found to result in assembly of a very active hybrid RrF 1 ␣ 3 ␤ 3 /␥c, which has both Ca-and MgATPase activities. 2 This hybrid can be engineered to be used in direct rotation assays, and could thus enable the examination of rotation in presence of CaATP as well as MgATP.