Replacement of Amino Acid Sequence Features of a- and c-Subunits of ATP Synthases of Alkaliphilic Bacillus with the Bacillus Consensus Sequence Results in Defective Oxidative Phosphorylation and Non-fermentative Growth at pH 10.5*

Mitchell's (Mitchell, P. (1961) Nature 191, 144–148) chemiosmotic model of energy coupling posits a bulk electrochemical proton gradient (Δp) as the sole driving force for proton-coupled ATP synthesis via oxidative phosphorylation (OXPHOS) and for other bioenergetic work. Two properties of proton-coupled OXPHOS by alkaliphilic Bacillus species pose a challenge to this tenet: robust ATP synthesis at pH 10.5 that does not correlate with the magnitude of the Δp and the failure of artificially imposed potentials to substitute for respiration-generated potentials in energizing ATP synthesis at high pH (Krulwich, T. (1995) Mol. Microbiol. 15, 403–410). Here we show that these properties, in alkaliphilic Bacillus pseudofirmus OF4, depend upon alkaliphile-specific features in the proton pathway through the a- and c-subunits of ATP synthase. Site-directed changes were made in six such features to the corresponding sequence in Bacillus megaterium, which reflects the consensus sequence for non-alkaliphilic Bacillus. Five of the six single mutants assembled an active ATPase/ATP synthase, and four of these mutants exhibited a specific defect in non-fermentative growth at high pH. Most of these mutants lost the ability to generate the high phosphorylation potentials at low bulk Δp that are characteristic of alkaliphiles. The aLys180 and aGly212 residues that are predicted to be in the proton uptake pathway of the a-subunit were specifically implicated in pH-dependent restriction of proton flux through the ATP synthase to and from the bulk phase. The evidence included greatly enhanced ATP synthesis in response to an artificially imposed potential at high pH. The findings demonstrate that the ATP synthase of extreme alkaliphiles has special features that are required for non-fermentative growth and OXPHOS at high pH.

during catabolism is used to produce a bulk ⌬p, acid and positive out, across the mitochondrial or bacterial cell membrane by respiration-dependent proton extrusion. Inward proton flux through the proton-coupled ATP synthase, energized by the ⌬p, then leads to ATP production (1)(2)(3)(4). Among the important unresolved issues is whether protons are always captured from the bulk medium as in Mitchell's (1) chemiosmotic model or whether they can be sequestered as they emerge from the respiratory chain (2,(5)(6)(7)(8)(9). Robust H ϩ -coupled OXPHOS by extremely alkaliphilic Bacillus strains growing on non-fermentable carbon sources at external pH values Ն10.5 poses one of the most striking challenges to the strictly bulk energization model (10 -13). At such pH values, maintenance of a cytoplasmic pH that is much lower than the external pH, i.e. a ⌬pH that is acid in, lowers the total chemiosmotic driving force, and yet OXPHOS proceeds optimally (10,13).
A variety of solutions to the energetic conundrum of alkaliphile OXPHOS have been proposed (for reviews, see Refs. 10 and 14). We have hypothesized that special properties of the alkaliphile ATP synthase are needed for OXPHOS at high pH that depend upon the presence of specific amino acid residues or stretches of amino acids in functionally important regions of the membrane-embedded a-and c-subunits of the enzyme (10,15). Other hypotheses have suggested global features of the alkaliphile membrane, membrane surface, or cell wall-associated polymers (see Refs. 10 and 14) that contribute to the resolution of the alkaliphile energetic problem. It has recently been proposed that a single global feature of this kind, i.e. a sufficiently low pH near the membrane surface, can completely account for alkaliphile OXPHOS (14), eliminating any need for special features of the enzymes that directly participate in OXPHOS. The goal of the current studies was to test whether in fact apparent alkaliphile-specific features of two membraneembedded ATP synthase subunits are specifically important for non-fermentative growth of alkaliphilic Bacillus pseudofirmus OF4 by changing them to the consensus sequence for non-alkaliphilic Bacillus species. If the alkaliphile F 0 sequence features are required for OXPHOS at high pH, mutants expressing normal levels of an altered ATP synthase would be expected to be specifically defective in non-fermentative growth and OXPHOS at pH 10.5. If, on the other hand, there are features of the surface layers that obviate the need for special adaptations of the alkaliphile OXPHOS machinery itself, the mutants would be without a specific OXPHOS phenotype at high pH.
Two signatures of alkaliphile OXPHOS are exhibited by whole cells and ADP ϩ P i -loaded membrane vesicles of alkaliphilic B. pseudofirmus OF4 that lack cell wall polymers (10 -13). Along with non-fermentative growth at high pH, these signature properties were used in this study to evaluate the OXPHOS phenotypes of the mutants and develop information about potential roles of particular ATP synthase features. First, the "quantitative signature" of alkaliphile OXPHOS is the generation of a higher ⌬Gp (reflecting the [ATP]/[ADP][P i ] ratio expressed in mV) at pH 10.5 than at pH 7.5 even though the ⌬p is about 3 times higher at the lower pH. In a strictly bulk chemiosmotic model, the ⌬Gp should exhibit a direct relationship to the bulk ⌬p. The lower ⌬p at high pH results from a large "chemiosmotically reversed" bulk pH gradient, acid in, that is required at pH 10.5. Although the other component of the ⌬p, the ⌬⌿, increases in the productive direction (positive out) at the higher pH, this increase does not come close to offsetting the chemiosmotically adverse pH gradient (10,13).
The second property of alkaliphile OXPHOS that is at odds with strict coupling to a bulk ⌬p is a "qualitative signature" that is also assessed in one group of mutants studied here. This "signature" is the inability of a valinomycin-mediated potassium diffusion potential to energize ATP synthesis comparably to a respiration-generated ⌬⌿ of the same magnitude. At elevated pH values at which the ⌬⌿ is the sole ⌬p component energizing ATP synthesis, an artificially imposed potential of the same magnitude would be predicted by the chemiosmotic model to work just as well as the respiration-generated potential. Rather, overall energization of ATP synthesis by the imposed potential is much lower than energization by respiration in alkaliphile cells and ADP ϩ P i -loaded membrane vesicles (11,12). Most notably, although the imposed potentials can still energize transport up to external pH values of 10.5, they lose their ability to energize ATP synthesis totally as the external pH rises above 9.2 where alkaliphile growth on OXPHOSrequiring substrates is still optimal (10 -13). We hypothesized that failure of imposed potentials to energize ATP synthesis above pH 9.2 reflects a pH-dependent gating of proton entry into the alkaliphile ATP synthase. Presumably such gating would also block proton loss to the bulk through the synthase when the alkalinity is particularly high and the bulk ⌬p is particularly low (Ref. 10 and Fig. 1A). An element of the ATP synthase responsible for such gating would be expected to reside in the proton uptake pathway of the a-subunit through which protons entering from the bulk pass en route to the critical carboxylate of the rotary c-subunit assembly (4, 16 -20). At pH values above the hypothesized gating pH, protons presumably could gain entry without first equilibrating with the bulk phase as proposed by others, e.g. by a protongathering element (21,22), from anionic surface lipids (23), from protons trapped in the surface water layer (14), or from an interacting proton-extruding respiratory chain complex (24).
In this study, we first built on the initial observations of alkaliphile-specific sequence features in a-and c-subunits of the F 0 -ATP synthase (15) using the expanded data base and greatly increased information about the Escherichia coli enzyme to identify the a-and c-subunit features to be mutagenized. Two single amino acids and one hydrophilic loop segment in the alkaliphile a-subunit and two single amino acids and one membrane-embedded segment in the alkaliphile c-subunit differed significantly from the homologous neutro-phile sequences. These sequences of B. pseudofirmus OF4 were changed to that of non-alkaliphilic Bacillus megaterium, which represents the consensus sequence for Bacillus. Five of the six mutants were found to assemble an active ATP synthase, but they still exhibited a defect in non-fermentative growth at high pH. Assays of the two signatures of alkaliphile OXPHOS at pH 10.5 further indicated that the sequence features in these mutants are likely to have different mechanistic roles.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Genetic Manipulations, and Growth Conditions-Alkaliphilic B. pseudofirmus OF4 strain 811M, a methionine auxotroph of B. pseudofirmus OF4 (25), was the parent strain for the F 0 mutant studies. Homologous recombination using the temperature-sensitive plasmid pGϩhost4 (Appligene, Pleasanton, CA) was used to construct a ⌬atpB-F (⌬F 0 ) strain by methods described previously (26). Mutations were made in a cassette that was reintroduced into the deletion strain (26) so that a new EcoRI restriction site was created in codons 162-163 of atpB. The primers used for this and other constructs made in this study are shown in Table I. The wild type used for all the experiments was prepared by reintroducing the wild type sequence, with the added EcoRI site, into the deletion strain. This strain exhibited no difference from its parent wild type strain in its capacity for OXPHOS and nonfermentative growth. All the other mutations changed B. pseudofirmus OF4 sequences for single amino acids or small stretches of amino acids to the sequence(s) found in the equivalent position(s) of the B. megaterium ATP synthase. Single amino acid mutations were made by the method of Kunkel et al. (27) with a bacteriophage M13m19 F 0 template that was identical to the cassette cloned into pGϩhost4. The multiple amino acid replacements (a-loop and cTMH1) were made by the PCR overlap extension method described by Ho et al. (28). The mutations that were constructed included the following in the a-subunit: a-loop, residues 97-106 (equivalent to 128 -137 in E. coli), from FELYNPTTHE to FAIVIDHN; a180 (218 in E. coli), from Lys to Gly, a212 (245 in E. coli), from Gly to Ser; double mutant a180 and a212, Lys to Gly and Gly to Ser, respectively. The mutants in the c-subunit were the following: cTMH1, residues 15-23 (22-30 in E. coli), from VAGAIAVAI to LGAGIGNGL; c33 (40 in E. coli), from Thr to Ala; and c51 (58 in E. coli), from Pro to Ala. The F 0 segment of each mutant was entirely sequenced and verified to have only the desired mutation(s). Initial characterizations of growth properties were nonetheless conducted on four to six independent mutants of each type to be sure that each strain used for subsequent detailed studies represented a phenotype that was reproducibly found as a result of the particular mutagenic change. Cloning was performed in E. coli XL-1 Blue (Promega, Madison, WI) grown in Luria broth using 250 g/ml erythromycin as selection for pGϩhost4. B. pseudofirmus OF4 strain 811M and mutant derivatives were routinely grown at 30°C in a semidefined medium containing 0.1% yeast extract with mineral salts and buffered with 0.1 M MOPS-NaOH at pH 7.5 or 0.1 M Na 2 CO 3 /NaHCO 3 at pH 10.5 (26). Growth experiments were conducted in which the sole carbon source was the yeast extract or with added 50 mM glucose or 50 mM sodium DL-malate. Growth on glucose is reported as the A 600 after subtracting the amount of growth on yeast extract alone. The ⌬F 0 strain exhibited very little growth on malate in excess of the growth on yeast extract alone, but this small fermentative component was subtracted from the raw A 600 data for malate growth so that the values presented represent the growth yield from non-fermentative growth.
Isolation of Everted Membrane Vesicles and ATPase Assays-Everted membrane vesicles were prepared from overnight cultures grown at pH 7.5 with malate as the energy source as described earlier (29) except that a battery of protease inhibitors was included in the French press buffer (1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 0.06 M bestatin, 1 g/ml pepstatin A, 1 M E-64 (N-(trans-epoxysuccinyl)-L-leucine 4-guanidinobutylamide), 1 M phosphoramidon, and 1 mM phenylmethylsulfonyl fluoride, final concentrations) (29). Octylglucoside-stimulated ATPase assays, which were carried out for 3 min at 37°C, contained, in a 0.5-ml volume, 20 mM Tricine-NaOH, pH 8.0, 5 mM ATP (sodium salt, Sigma), 2.5 mM MgCl 2 , 30 mM octylglucoside, 50 mM Na 2 SO 3 , and 50 g of membrane protein (29). Unstimulated ATPase activity using 0.4 mg of membrane protein was measured in a 1-ml volume containing 50 mM Tricine-NaOH, pH 8.0, 5 mM MgCl 2 , and 5 mM ATP. The reactions were incubated for 15 min at 37°C and terminated by trichloroacetic acid addition (5% final concentration). The precipitated protein was removed by centrifugation, and the supernatants were analyzed for liberated P i by the method of LeBel et al. (30). Blank and P i standard tubes received an equivalent volume of membranes after trichloroacetic acid addition. For DCCD inhibition, everted membrane vesicles were preincubated with 4.8 l of methanol (no DCCD) or 25 mM DCCD (to 100 M final concentration) in 1.2 ml at a concentration of 2.6 mg of protein/ml for 30 min at room temperature in 20 mM MOPS-NaOH, pH 7.0, 5 mM MgCl 2 . They were then diluted 4-fold into assay buffer (67 mM Tricine-NaOH, pH 8.0, 5 mM MgCl 2 ) and equilibrated at 37°C, and the reaction was carried out for 30 min after addition of 5 mM ATP. Reactions were terminated by addition of trichloroacetic acid and analyzed as described above. Control experiments were carried out with purified B. pseudofirmus OF4 F 1 -ATPase prepared as described earlier (31). Instead of membranes, 50 g of purified F 1 was included in the 1.2-ml preincubation mixture, and P i was determined by the malachite green assay (32). The experiments showed no DCCD inhibition of F 1 under these conditions.
Western Analyses-Everted membrane vesicles equivalent to 5 g of protein were loaded in each lane (except for the cTMH1 mutant for which 25 or 50 g of protein was used), resolved on 12% SDS-polyacrylamide minigels (33), and transferred electrophoretically to nitrocellulose overnight in a Tris/glycine/methanol buffer. Western blots were carried out by the chemiluminescence method according to the manufacturer's instructions (Amersham Biosciences). The ␤ subunit of the F 1 F 0 was detected using a monoclonal antibody against the E. coli ␤ subunit (Molecular Probes, Eugene, OR). Scanned blots were quantitated using Quantity One analysis software from Bio-Rad.
Measurements of ⌬Gp and ⌬p-Cells were grown overnight in semidefined malate medium at pH 8.5. A fresh culture at pH 8.5, with the yeast extract lowered from 0.1 to 0.02%, was inoculated from the overnight culture and allowed to grow for 3 h. Cells were harvested and washed free of nutrients with pH 8.5 buffer (100 mM Na 2 CO 3 /NaHCO 3 , 1 mM MgSO 4 , 1 mM KP i ), then diluted 1:15 into pH 7.5 or 10.5 buffer (100 mM MOPS-NaOH and 100 mM Na 2 CO 3 /NaHCO 3 , respectively) containing 10 mM malate, and shaken at 30°C. The dilution marked the zero time for the experiment and was ϳ30 min after the end of the 3-h growth period. During that period, cytoplasmic ATP levels dropped about 75%. This depletion is less than used in earlier studies that included long starvation periods before re-energization (11). Preliminary experiments indicated that the gentler protocol was required for the wild type and mutant panel to retain uniform viability and a normal capacity for ⌬⌿ generation. Samples were taken at zero time, 1 h, and 3 h for assays of [ATP], [ADP], [P i ], ⌬pH, and ⌬⌿. The intracellular pH was calculated using measurements of the external pH and ⌬pH. ⌬pH was determined from the distribution of radiolabeled probes, methylamine for ⌬pH acid in and dimethyloxazolidine-2,4-dione for ⌬pH acid out, which were used at 2 M and whose accumulation was assayed via a filtration assay (26). The ⌬⌿ was calculated from the distribution of the ⌬⌿ probe tetraphenylphosphonium bromide using the Nernst equation. Probe distribution was assayed using 2 M tritiated probe and a filtration assay that had been validated using imposed diffusion potentials of known magnitude (34). ATP was measured by the luciferinluciferase assay (35,36), and ADP was similarly assayed after its enzymatic conversion to ATP using pyruvate kinase (37). Measurements of the starting ATP and ADP concentrations of cells and of background ATP and ADP of unenergized vesicles were conducted. Inorganic phosphate was measured by the method of LeBel et al. (30). For calculation of the ⌬Gp, the value of ⌬G 0 for the measured cytoplasmic pH was used (38).
Assays of ATP Synthesis in ADP ϩ P i -loaded Membrane Vesicles-Measurements of ATP synthesis in right-side-out membrane vesicles were performed as in earlier studies (12) except that the cells from which vesicles were prepared were grown at pH 8.5 instead of pH 10.5 on the semidefined malate medium. For experiments in which ATP synthesis was energized by an electron donor, 10 mM ascorbate plus 0.1 mM phenazine methosulfate, the vesicles were loaded with 20 mM potassium phosphate, 5 mM sodium phosphate, 0.25 M sucrose, 5 mM MgCl 2 , and 5 mM potassium ADP at a final pH of 8.3. The same mixture was used to load the vesicles for generation of a diffusion potential at pH 8.3; for the vesicles to be energized by a diffusion potential at pH 9.3, the final pH of the loaded solution was 9.3. For assays of respirationdependent ATP synthesis, the ADP ϩ P i -loaded vesicles were diluted 1:20, to 500 g of protein/ml, into either pH 7.5 buffer containing 25 mM sodium phosphate, 0.25 M sucrose, 5 mM MgCl 2 , and 200 mM K 2 SO 4 or into pH 10.5 buffer containing 25 mM Na 2 CO 3 , 0.25 M sucrose, 5 mM MgCl 2 , and 200 mM K 2 SO 4 , and immediately thereafter 10 mM ascorbate plus 0.1 mM phenazine methosulfate was added to start the reaction. For energization of ATP synthesis by a valinomycin-mediated potassium diffusion potential, the right-side-out vesicles were loaded at pH 8.3 with the same concentrations of ADP ϩ P i as in the experiments with ascorbate-phenazine methosulfate energization. The vesicles were concentrated to ϳ20 mg of protein/ml after which valinomycin was added to a final concentration of 1 M. For DCCD treatment, 200 M DCCD was added to the concentrated cell mixture 5 min before the valinomycin addition. It was also included in the dilution buffer for those samples. To initiate synthesis, the vesicles were diluted 500-fold into a potassium-free buffer of the same pH as the internal pH, either 8.3 or 9.3, consisting of 25 mM sodium phosphate, 0.25 M sucrose, 5 mM MgCl 2 , 200 mM Na 2 SO 4 , and 1 M valinomycin. This was calculated, by the Nernst equation, to generate a potential of Ϫ160 mV. A set of controls in which the external buffer contained the equivalent potassium ion concentration as the internal concentration was conducted routinely.  Fig. 1B, the alkaliphile-specific sequence features included in the study after analyses of new sequence alignments are indicated on topological models of the a-and c-subunits by numbered boxes. The alignments used to identify these features included the two alkaliphiles that were in the earlier alignments, B. pseudofirmus OF4 and Bacillus alcalophilus, both of which are categorized as extreme alkaliphiles because they can grow at pH values of 11 and above (13,39); only partial sequence of the a-subunit is available for B. alcalophilus. Also included in the alignments are the data from the completely sequenced genomes of alkaliphilic Bacillus halodurans C-125 and Oceanobacillus iheyensis (40,41). B. halodurans C-125 is another extreme alkaliphile (42), while O. iheyensis is a marine organism with an upper pH limit for growth of about 9.5 and, importantly, little apparent ability to grow on non-fermentable carbon sources that would necessitate a capacity for OXPHOS (43). The other new alkaliphile sequence data in the alignments is for the H ϩ -coupled ATP synthase of the alkaliphilic, thermophilic Bacillus strain TA2.A1 that grows only up to pH 10 but still exhibits the quantitative signature found for B. pseudofirmus OF4 (44 -46). Sequence data from three different neutrophilic Bacillus species and the extensively studied E. coli are also included in the alignments.

Identification of Alkaliphile-specific Sequence Features in the a-and c-Subunits of the ATP Synthase-In
The first feature of the alkaliphile a-subunit, the a-loop feature (numbered 1 in Fig. 1B), is part of the periplasmic loop between aTMH2 and aTMH3. As illustrated by the boxed area of hydropathy plots in Fig. 1C, the region nearest aTMH2 displays a more polar, hydroxyamino acid-rich character in alkaliphiles than in Bacillus neutrophiles and E. coli. Modeled   FIG. 1. Alignments of a-subunit regions and c-subunits illustrating alkaliphile-specific sequence features and models illustrating their location and a possible function. A, the ATP synthase is shown with the usual proton entry route from the bulk phase blocked (hatched arrow) by a hypothesized pH-dependent gating of the proton uptake pathway in subunit a. This closure is proposed to account for the failure of an artificially imposed potassium diffusion potential to energize the wild type synthase above this putative gating pH. The property is expected to minimize loss of protons to the alkaline bulk phase. Numbers correspond to features 1-3 (see numbered boxes in B) that are candidates for a role in the putative gating property. B, topological models illustrate positions of both the a-and c-subunit features using boxes that are numbered. Several additional residues of importance in these subunits are also annotated. The numbered features are: 1, a-loop; 2, aLys 180 ; 3, aGly 212 ; 4, cTMH1; 5, cThr 33 ; and 6, cPro 51 . C, Kyte-Doolittle hydropathy plots (Gene Runner) of different a-subunits. The boxed region is the hypothesized periplasmic loop between aTMH2 and aTMH3 corresponding to E. coli residues 128 -137 (47)  according to Valiyaveetil and Fillingame (47), the alkaliphiles have 6 -8 hydroxy or charged amino acids of 9 -10 total residues (FELYNPTTHE in B. pseudofirmus OF4) compared with the 2-4 found in the 8 -10-amino acid stretch of various neutrophiles (FAIVIDHN in B. megaterium). If modeled as a longer loop proposed at this position (48), the difference is preserved. The a-loop extends from a region of aTMH2 proposed to be part of the proton uptake pathway through the ATP synthase and to initiate the ⌬p-dependent conformational changes during proton uptake (17,19). In the current study, the a-loop mutant has the FAIVIDHN sequence of B. megaterium in place of its slightly longer and more polar native sequence. The two other a-subunit features included in the study were the single amino acid deviations noted earlier in aTMH4 and aTMH5 (15), numbered 2 and 3, respectively, in Fig. 1B. The alkaliphiles have a lysine at aTMH4 position 180 (Gly in neutrophiles) and glycine in position 212 (Ser in Bacillus neutrophiles and His in E. coli). The E. coli residues corresponding to aLys 180 and aGly 212 are among the "second site suppressor pairs" found to functionally interact and considered likely to be in physical proximity within the proton uptake pathway (17) as modeled for the alkaliphile subunit by methods used in studies of E. coli (49) (Fig. 1E). Hartzog and Cain (50) replaced the E. coli equivalents of aLys 180 and aGly 212 with the alkaliphile residues in those positions. Single mutants lost the capacity for growth on succinate, whereas this capacity was restored in the double mutant. Therefore, both the single mutations of the alkaliphile residues and a double mutant with changes to the Bacillus consensus sequence (aK180G, aG212S, and aK180G/aG212S) were constructed in B. pseudofirmus OF4 for the current study.
The three c-subunit features chosen for study so far have been found only in extreme alkaliphiles (Fig. 1, B and F), a correlation re-enforced by analyses of newly isolated alkaliphile strains (51). One of these features is a proline residue that is 3 residues away, on the N-terminal side, from the essential c-subunit carboxylate (cGlu 54 in B. pseudofirmus OF4 and cAsp 61 in E. coli, see residue numbered 6 in Fig. 1B). It is 6 residues away from a conserved proline (cPro 57 in B. pseudofirmus OF4) that flanks the carboxylate on the C-terminal side. Arechaga and Jones (52) called PXXEXXP the "alkaliphile motif " and suggested that it might influence the ion binding site. On the opposite helix of the c-subunit hairpin, cTMH1, a consensus sequence of XGXGXGXGX in the middle of the helix has been reported from alignments of 38 c-subunits, and the glycines were suggested to allow for the tight packing of adjacent helices in the ring (53). The presence of small, uncharged residues in cTMH1 was also suggested to accommodate movements of cTMH2 that would be essential during the protonation-deprotonation cycle of the carboxylate (49, 54). Thus the substitution of at least 3 of the glycine residues of this motif with alanines in extreme alkaliphiles represents another feature of interest (numbered 4 in Fig. 1B). The mutant in which the B. pseudofirmus OF4 sequence VAGAIAVAI was changed to the B. megaterium sequence LGAGIGNGL is designated as the cTMH1 mutant. The third feature of the c-subunit included in the study is threonine, cThr 33 of the alkaliphiles, substituting for a conserved alanine in the loop region involved in binding of F 1 (55). This feature is numbered 5 in Fig. 1B.
Growth Phenotypes of the ATP Synthase Mutant Panel-None of the atp mutant strains exhibited a reduced growth yield on glucose at either pH 7.5 or 10.5 (Fig. 2) even though one of the mutants, the cTMH1 mutant, exhibited a large growth defect on malate at both pH 7.5 and 10.5. Although not shown, the ⌬F 0 strain of alkaliphilic B. pseudofirmus OF4 similarly did not exhibit a defect in growth on glucose, although such a deficit has been observed in a ⌬atp strain of Bacillus subtilis (56). This lack of a significant contribution of the hydrolytic activity of the alkaliphile F 1 F 0 -ATP synthase to fermentative growth is consistent with observations that the alkaliphile enzymes are substantially "locked" in a synthetic conformation (29,31,46). This in turn has been hypothesized by Keis et al. (46) to result from the especially basic character of the C-terminal domain of alkaliphile ⑀-subunits, resulting in interactions strongly favoring the synthetic over the hydrolytic conformation of the enzyme (57,58). In contrast to their normal growth yield on glucose, the mutants exhibited diverse deficits in non-fermentative growth. The aGly 212 mutant grew to a normal growth yield on malate at pH 10.5 in semidefined medium (Fig. 2), but this mutant as well as the aK180G/aG212 double mutant and cP51A mutant failed to grow in a defined medium that had a high amine-nitrogen concentration and more strongly stresses the pH homeostatic mechanism at high pH (Refs. 59 and 60 and data not shown). The remaining mutants exhibited a substantial defect in growth yield on the semidefined malate-containing medium that was specific for pH 10.5. The defect was particularly severe in the a-loop, aK180G (single or double), and cP51A mutants (Fig. 2), although the ATPase activity of their membranes was at 80% or more of wild type levels (see below). Growth rates on malate at pH 10.5 examined side-by-side with the wild type and ⌬F 0 strains were zero or negligible for the cTMH1, cP51A, and both the single and double aK180G mutants. The a-loop mutant exhibited modest growth after a prolonged lag. The aG212S mutant grew at the same rate as wild type, and the cT33A mutant grew at a rate that was about 15% greater than that of the wild type (data not shown).
Assays of ATPase Content and Activity of Membrane Vesicles-The cTMH1 mutant, with its alanine-rich region replaced by a consensus glycine-rich region, had greatly reduced levels of ATPase protein in the membrane and was not included in further experiments. All the other mutants assembled an ATPase at protein levels comparable to the wild type (Table II). The ATPase (hydrolytic) activity of the enzymes from alkaliphilic Bacillus strains is very low, but it can be activated by inclusion of certain detergents in the reaction mixture (29,31,46). In the presence of octylglucoside, the ATPase activity of the wild type and mutant strains was similar except for the low activity of the cTMH1 mutant. In the absence of octylglucoside, there was a reproducible elevation in the hydrolytic activity of the aK180G mutant relative to the wild type, both alone and together with the aG212S mutation. There was also a smaller increase in the hydrolytic activity of the cP51A mutant. DCCD sensitivity of the individual aK180G and aG212S mutants was also elevated, but consistent with aK180G and aG212S being a second site suppressor pair (17), DCCD sensitivity was not elevated in the double mutant. The ATPase assays of the remaining two mutants, the a-loop and cT33A mutants, revealed no features of note.
ATP Synthesis and ⌬p Generation upon Re-energization of ATP-depleted Cells-ATP resynthesis was assayed in wild type and mutant cells that were grown at pH 8.5 and then washed with nutrient-free pH 8.5 buffer under conditions that resulted in depletion of cytoplasmic ATP (see "Experimental Procedures"). The cells were then re-energized in buffered malate at pH 7.5 or 10.5. Growth of the wild type at pH 10.5 enhances the ability of re-energized cells to synthesize ATP at pH 10.5 (11), but pH 8.5 had to be used for growth here since most of the mutants were incapable of substantial growth at the higher pH. The lower growth pH led to generation of significantly lower ATP/ADP ratios than observed in re-energization experiments using pH 10.5-grown cells and than those found in growing cells (11,13). This has been observed before with pH 7.5-grown cells (11) and probably results from the presence of a suboptimal complement of membrane lipids and respiratory chain components (10). However, upon energization in the current protocol, the wild type and all the mutants generated normal ⌬⌿ values in a range of Ϫ137 to Ϫ163 mV at pH 7.5 and Ϫ168 to Ϫ193 mV at pH 10.5, showing that each population of cells was energetically functional. The wild type ATP/ADP ratio increased over 3-fold during the 3-h incubation period at pH 10.5 under these conditions. Bioenergetic parameters (i.e. [ATP], [ADP], [P i ], ⌬pH, and ⌬⌿) were determined at zero time, 1 h, and 3 h, and the ⌬Gp/⌬p ratio was calculated. The ⌬Gp/⌬p ratio would reflect the protons traversing the ATP synthase per ATP synthesized if the synthase functions in equilibrium with the bulk force but would reflect the discrepancy between the actual driving force and measured bulk ⌬p if protons are captured without equilibration. The typical ⌬Gp/⌬p ratio during wild type OXPHOS at pH 10.5 is 13 (10, 13), a value that was reached within the 3-h resynthesis experiment (Fig. 3, top  panel). By contrast, the typical ⌬Gp/⌬p ratio during wild type OXPHOS at pH 7.5 is 2.5-3, and this ratio was observed throughout the resynthesis experiment at pH 7.5 for all the mutants as well as the wild type strain (Fig. 3, bottom panel). At pH 10.5, the mutants showed diverse deviations from the wild type capacity to resynthesize ATP. ATP resynthesis was compromised severely in the a-loop, aLys 180 , and cPro 51 mutants, was much less affected in the aGly 212 mutant, and was significantly enhanced in the cThr 33 mutant (Fig. 3, top panel). A correlation was observed between the capacity to synthesize ATP at a high ⌬Gp/⌬p ratio at pH 10.5 and a capacity to acidify the cytoplasm. This interesting relationship suggests that proton recapture may be a physiologically important adjunct to other mechanisms of alkaliphile pH homeostasis (10). Notably, the double aK180G and aG212S mutant exhibited a significant increase in cytoplasmic pH relative to all the other strains right after the upward pH shift to 10.5. This supported a role for this pair of residues in preventing proton loss through the ATP synthase. The individual aK180G, aG212S, and cP51A mutants exhibited somewhat less cytoplasmic alkalinization. The a-loop mutant exhibited no difference in initial cytoplasmic pH change from the wild type, and the cT33A mutant reproducibly exhibited a slightly lower cytoplasmic pH than wild type, paralleling its larger ⌬Gp generation.
ATP Synthesis by ADP ϩ P i -loaded Right-side-out Membrane Vesicles-Greatly reduced respiration-dependent ATP synthesis was observed, relative to wild type preparations, in ADP ϩ P i -loaded vesicles from single or double aK180G mutants, the a-loop mutant, and the cP51A mutant at pH 10.5, while the aG212S mutant was only modestly affected (Fig. 4). Interestingly, vesicles of the cT33A mutant synthesized ATP at pH 10.5 after a lag and then quickly showed a decline in intravesicular ATP. At pH 7.5, ATP was synthesized by all the preparations, although there were small discernible deficits, relative to wild type, especially in preparations from the a-loop and two csubunit mutants.
Mutation of aLys 180 and aG212S together led to elevated ATPase activity in the absence of octylglucoside activation as well as greater cytoplasmic alkalinization than wild type upon an alkaline shift in external pH, but these properties were not observed in the a-loop mutant (Table II and Fig. 3). This suggested that aLys 180 and aGly 212 , but not the a-loop, have a role in pH-dependent modulation of proton fluxes through the asubunit, i.e. a "gating" function. To further probe this indication, we studied the qualitative signature of alkaliphile OXPHOS in the a-subunit mutants. Failure of an imposed valinomycin-mediated potassium diffusion potential to energize substantial ATP synthesis above pH 9.2 has been hypothesized to reflect a gating function (10,12). If this is correct and aLys 180 and aGly 212 are involved in gating, the mutants in these residues would be expected to exhibit enhanced ATP synthesis in response to the imposed potential, especially at pH Ͼ 9.2. We assayed diffusion potential-dependent ATP synthesis by ADP ϩ P i -loaded vesicles of the three a-subunit mutants and of the wild type at pH 8.3 and 9.3. These are, respectively, below and above the hypothesized gating pH of the proton uptake pathway (Ref. 12 and Fig. 1A). As observed previously, an artificially imposed valinomycin-mediated diffusion potential (Ϫ160 mV) supported DCCD-sensitive ATP synthesis by vesicles from the wild type at pH 8.3, but almost no synthesis was observed at pH 9.3. The a-loop mutant exhibited the same pattern, whereas mutants altered in aLys 180 , aGly 212 , or both exhibited almost twice the synthesis of ATP as wild type at pH 8.3 and over 10 times the synthesis at pH 9.3 (Table III). DISCUSSION This study shows that specific features of the membraneembedded subunits of proton-coupled alkaliphile ATP synthase are required for non-fermentative growth and for the robust OXPHOS that wild type B. pseudofirmus OF4 carries out at very high pH. Four of the five single a-and c-subunit mutants as well as the double mutant, with consensus Bacillus sequences replacing alkaliphile sequence features, assembled an active ATP synthase at wild type levels (Table II). All but the aG212S mutant exhibited deficits in non-fermentative growth at high pH in standard semidefined medium that were profound except in the cT33A mutant (Fig. 2). The details of their underlying deficits in OXPHOS were quite distinct from one another, indicating that a number of properties are involved in function of wild type ATP synthase at pH 10.5. The results are consistent with Williams' (2, 6) early expectation that specific properties of proton-consuming as well as proton-pumping OXPHOS complexes would be important for proton transfers that are faster than proton equilibration with the bulk. Various global factors that can increase proton retention near the membrane surface are also likely to contribute to alkaliphile OXPHOS as suggested by others (for reviews, see Refs. 10 and 14). The current results, however, show that the aggregate global adaptations are not sufficient to support the OXPHOS observed in B. pseudofirmus OF4.
The effects of single aK180G and aG212S mutations versus the effects of their combined mutation are consistent with a functional interaction between the residues at these positions.  3. Resynthesis of ATP by washed wild type and mutant cells upon energization by malate at either pH 7.5 or 10.5. Cells were grown at pH 8.5 for 3 h with malate as energy source and with the yeast extract lowered to 0.02% prior to washing the cells free of nutrients. Cytoplasmic ATP was depleted during the cell preparation (see "Experimental Procedures"). Re-energization was achieved by diluting the washed cells into buffer at either pH 7.5 or 10.5 that contained 50 mM sodium DL 4. Respiration-dependent synthesis of ATP by ADP ؉ P iloaded right-side-out membrane vesicles of wild type and mutants. Vesicles were prepared from cells that were grown at pH 8.5. Assays were initiated by the addition of electron donor, 10 mM ascorbate, plus 0.1 mM phenazine methosulfate and were conducted as described under "Experimental Procedures." Growth at pH 8.5 rather than 10.5 accounts for reduced ATP synthesis by wild type relative to synthesis at pH 7.5 in these experiments as compared with the earlier experiments where OXPHOS at pH 10.5 was higher (12). The S.D. of the values shown was within 10%, and the means were derived from duplicate assays on at least three independent vesicle preparations.
The data include the enhanced DCCD inhibition of ATPase activity (in the absence of octylglucoside) of the single mutants but not the double mutant (Table II), the greater stimulation of ATP synthesis by a diffusion potential in the double mutant than in the single mutants (Table III), and the greater cytoplasmic alkalinization of the double mutant than of the single mutants immediately after an alkaline shift (Fig. 3). However, the aGly 212 and aLys 180 residues do not exhibit complete functional equivalence. Whereas the aK180G-containing single and double mutants exhibited severe deficits in growth on malate in semidefined medium, the aG212S exhibited no such deficit (Fig. 2), showing a non-fermentative growth deficit only on the more stringent high amine nitrogen-malate medium. The absence of a growth phenotype for the aG212S mutant on malate in semidefined medium at pH 10.5 is consistent with earlier observations that small defects in pH homeostasis, as might be caused by compromised proton loss through the ATP synthase, are more evident in the more challenging pH shift protocol (or growth phenotypes at pH values of 11 and above) than in growth experiments on malate at pH 10.5 (61). The growth deficits of the mutants in the current panel on semidefined malate-containing medium at pH 10.5 appear to correlate best with loss of the quantitative signature of alkaliphile OXPHOS, i.e. a robust capacity of ATP synthesis per se and a high ⌬Gp/⌬p ratio at high pH. We hypothesize that the growth deficit of the aLys 180 -containing mutants on this medium at pH 10.5 reflects an important role for the positively charged lysine residue in ATP synthesis itself at high pH, e.g. proton flow to the csubunit carboxylate, in addition to its role in proton retention. The aGly 212 residue does not appear to have an equivalent role in the ATP synthesis function. Two pieces of evidence support this lack of functional equivalence between aLys 180 and aGly 212 in ATP synthesis. Both the ⌬Gp/⌬p profile of cells of the aG212S mutant at pH 10.5 (Fig. 3) and the capacity of vesicles from the aG212S mutant to carry out respiration-dependent ATP synthesis at pH 10.5 (Fig. 4) were much closer to wild type patterns than to those of the aK180G mutant. It will be of particular interest to probe the role of aLys 180 in ATP synthesis further by examining different substitutions at the aLys 180 position alone and in combination with changes in aGly 212 or the conserved aArg 172 .
Mutation of the a-loop did not alter any of the indices of a gating element, but this mutation was nonetheless accompanied by a severe deficit in non-fermentative growth at pH 10.5 and loss of the quantitative signature of alkaliphile OXPHOS, i.e. the ability to carry out OXPHOS at pH 10.5 with a high ⌬Gp/⌬p ratio. The a-loop is in a position near a region of aTMH2 that is proposed to have an important role in the proton uptake pathway (17,19) and might monitor the status of the aLys 180 -aGly 212 pair. It is a candidate for the hypothesized element that facilitates sequestered proton capture by the gated ATP synthase at high pH (Fig. 1A).
While removed from the proton uptake pathway of the a-subunit, the two c-subunit features studied in detail are also of great interest. The cP51A mutation severely compromised nonfermentative growth and OXPHOS at high pH. It might modify the structure of the region surrounding the critical cGlu 54 carboxylate so that proton loss from this part of the ATP synthase complex is specifically minimized. This is consistent with the presence of a proline at the cPro 51 position only in extreme alkaliphiles (Fig. 1F). Alternatively this feature might have a role in assembly of a high number of c-subunit monomers in the rotor at elevated pH values. The different c-subunit stoichiometry found in different ATPases is a topic of intense recent interest especially because this stoichiometry is expected to have bioenergetic implications (4, 62-65).
Finally the phenotype of the cT33A mutant was distinct from those of the other mutants in the panel. The growth rate and bioenergetic profile of the cT33A mutant in whole cells suggested that it should be a "superalkaliphile". Of particular note was its rapid development of an extraordinarily high ⌬Gp/⌬p relative to the wild type strain in the ATP resynthesis experiments (Fig. 3, top panel). Yet the cT33A mutant exhibited a deficit in growth yield on non-fermentative substrates at high pH. This deficit indicates that there is functional value at high pH of the unusual threonine in the interface region between the c-subunit and the stalk that connects to the catalytic subunits.ThisroleforcThr 33 isreflectedinthekineticsofrespirationdependent ATP synthesis by right-side-out vesicles from the cT33A mutant at pH 10.5, i.e. the pattern of initial synthesis followed by a decline in ATP (Fig. 4). We hypothesize that the cThr 33 residue prevents rapid torque generation at high pH that, in the cT33A mutant, is followed by slippage. This might result in an ATP synthesis-hydrolysis oscillation that would account for the reduced growth yield on malate at pH 10.5 even though the ATP synthase is highly capable of synthesizing ATP over the full pH range of the alkaliphile.
The goal of the current study was to test the hypothesis that the alkaliphile-specific features of the a-and c-subunits of ATP synthase are critically involved in OXPHOS and non-fermentative growth. The experimental design therefore involved making a specific mutagenic change to the consensus sequence for non-alkaliphiles that was likely to allow most of the mutants to assemble an active ATP synthase. The current observation of diverse pH 10.5-specific deficits in fermentative growth and OXPHOS features found in these mutants demonstrates that the alkaliphile-specific features of the synthase are important for OXPHOS at pH 10.5. It will now be of interest to focus on these different features and their particular mechanistic roles by replacing them with multiple substitutions that will further probe their function. Mechanistic insights into the adaptations of ATP synthesis by the alkaliphile may be expected to be of general interest with respect to proton coupling mechanisms in OXPHOS.