In FoF1 (FoF1-ATP synthase), proton translocation through Fo drives rotation of the oligomer ring of Fo-c subunits (c-ring) relative to Fo-a. Previous reports have indicated that a conserved arginine residue in Fo-a plays a critical role in the proton transfer at the Fo-a/c-ring interface. Indeed, we show in the present study that thermophilic FoF1s with substitution of this arginine (aR169) to other residues cannot catalyse proton-coupled reactions. However, mutants with substitution of this arginine residue by a small (glycine, alanine, valine) or acidic (glutamate) residue mediate the passive proton translocation. This translocation requires an essential carboxy group of Fo-c (cE56) since the second mutation (cE56Q) blocks the translocation. Rotation of the c-ring is not necessary because the same arginine mutants of the ‘rotation-impossible’ (c10-a)FoF1, in which the c-ring and Fo-a are fused to a single polypeptide, also exhibits the passive proton translocation. The mutant (aR169G/Q217R), in which the arginine residue is transferred to putatively the same topological position in the Fo-a structure, can block the passive proton translocation. Thus the conserved arginine residue in Fo-a ensures proton-coupled c-ring rotation by preventing a futile proton shortcut.
FoF1-ATP synthase, often simply called FoF1, is composed of two portions, a water-soluble F1, which has catalytic sites for ATP synthesis/hydrolysis, and a membrane-integrated Fo, which mediates H+ (proton) translocation [1,2]. The FoF1 from thermophilic Bacillus PS3, which was used in the present study, as well as that from Escherichia coli, have the simplest subunit structure; α3β3γ1δ1ε1 for F1 and a1b2c10 for Fo . Proton translocation down the gradient through Fo drives rotation of a central rotor shaft made of an Fo-c oligomer ring (c-ring) and γε subunits, causing conformational changes in F1 that result in ATP synthesis. Conversely, ATP hydrolysis in F1 induces a reverse rotation of the rotor that enforces Fo to pump protons in the reverse direction. The mechanism and structure of the Fo motor remain less clear than the F1 motor. The stator portion of Fo motor, ab2, is located on the periphery of the c-ring [4,5]. Fo-c folds in a hairpin-like structure [6–8] with two transmembrane helices, and an essential proton-binding carboxy group (cD61 in E. coli and cE56 in Bacillus PS3) is located in the middle of the second helix. Without knowledge of atomic structure, Fo-a has been assumed to have five transmembrane helices [9,10] and the fourth helix has an essential arginine residue (aR210 in E. coli and aR169 in Bacillus PS3) [11,12]. A current model assumes that the arginine residue is located at the middle of two half-channels for proton passage in Fo-a and is close to the carboxy group in Fo-c [13–15]. In the present study, we showed that mutants with replacement of this Arg169 of Bacillus PS3 FoF1 with a small or acidic residue, but not other residues, can mediate passive proton translocation. This translocation is dependent on an essential carboxy group in Fo-c, but not on c-ring rotation. It appears that Arg169 plays a role to block a futile proton shortcut via cE56 and ensures proton-coupled rotation of the c-ring.
MATERIALS AND METHODS
Proteins and vesicles
FoF1 from thermophilic Bacillus PS3 with a tag of ten histidine residues at the N-terminus of the β-subunit was used throughout the present study. Culture of E. coli cells expressing thermophilic Bacillus PS3 FoF1, and purification of wt (wild-type) and mutant FoF1 with Ni-NTA (Ni2+-nitrilotriacetate) affinity chromatography were carried out as described previously . Inverted membrane vesicles (membrane vesicles, hereafter) from E. coli cells were prepared as described in , except that MgCl2 was omitted from the buffers. F1-stripped membrane vesicles were prepared by treatment with 2 mM EDTA as described previously . Proteoliposomes were reconstituted using a freeze–thaw method  with some modifications. Liposomes (100 μl, 44 mg/ml) and FoF1 (20 μl, 10 mg/ml) were mixed vigorously, left at 25 °C for 15 min, and frozen in liquid nitrogen. The suspension was thawed, added to 1 ml of water, and centrifuged (180000 g, 30 min, 4 °C). The precipitate was suspended in 1 ml of 0.5 M KCl, incubated at 55 °C for 30 min, chilled in ice, added to 5 μl of MgSO4 and centrifuged (180000 g, 30 min, 4 °C). The precipitate was suspended in 50 μl of 10 mM Hepes/NaOH (pH 7.5), 0.25 M sucrose and 5 mM MgSO4, and used for experiments immediately.
Mutants of FoF1
Plasmids for FoF1 mutants were made from pTR19-ASDS  using the megaprimer method and were used for transformation of a FoF1-deficient E. coli strain DK8 . FoF1 with a fused c10-a subunit [(c10-a)FoF1], in which a c10-polypeptide and Fo-a subunit was fused by a linker LAGLVPRGSP (underlined residues are the thrombin-recognition sequence), was obtained as follows. A region containing an Fo-a gene in pTR19-ASDS was amplified by PCR to introduce a new NheI site at an upstream region of the Fo-a gene and new SpeI and PstI sites at a downstream region of the Fo-a gene. The DNA fragment obtained was digested with EcoRI and PstI, and ligated into the pTR19-ASDS previously digested with both enzymes (pTR19-NASP). A region containing a gene for c-dimer (Fo-c2) of the expression plasmid pTR-AL2 for FoF1  was amplified by PCR to introduce a new EcoRI site at an upstream region of the Fo-c2 gene and new NheI and PstI sites at a downstream region of the Fo-c2 gene. The DNA fragment obtained was digested with EcoRI and PstI and ligated into the pTR19-ASDS previously digested with both enzymes to obtain a recombinant DNA for Fo-c2 gene (pTR19-C2). pTR19-C2 was digested with EcoRI and NheI and ligated into the pTR19-AL8C  previously digested with EcoRI and AvrII to obtain a recombinant DNA for c10-fused FoF1 lacking the Fo-a gene (pTR19-C10N). pTR19-C10N was digested with EcoRI and NheI, and the 2.2-kbp EcoRI-NheI fragment was ligated into an EcoRI-NheI site of pTR19-NASP (pTR19-C10NASP). pTR19-C10NASP was digested with EcoRI and SpeI and ligated into the pTR19-ASDS previously digested with both enzymes to obtain the recombinant DNA (pTR19-C10TA) for (c10-a)FoF1. A plasmid to express (c10-a)FoF1 with a replacement of Arg169 of Fo-a with a lysine residue [(c10-aR169K)FoF1], alanine residue [(c10-aR169A)FoF1] or glutamate residue [(c10-aR169E)FoF1] was constructed by the same method described above. Plasmids obtained as above were individually expressed in an Fo-deficient E. coli strain JJ001  (a gift from Dr J. Hermolin, University of Wisconsin Medical School, Madison, WI, U.S.A.). When indicated, (c10-a)FoF1 (500 μg in 60 μl) was treated with thrombin (50 units) in 50 mM Tris/H2SO4 (pH 8.0), overnight at room temperature (25 °C).
ATPase activity was measured with an ATP-regenerating system at 45 °C . When indicated, the reaction mixture contained 100 μM DCCD (dicyclohexylcarbodi-imide). ATP-driven proton-pumping activity was measured at 45 °C by quenching of the fluorescence (excitation, 410 nm; emission, 480 nm) of ACMA (9-amino-6-chloro-2-methoxyacridine) in 10 mM Hepes/KOH (pH 7.5), 100 mM KCl, 5 mM MgCl2, 0.3 μg/ml ACMA and 50 μg of protein/ml of proteoliposomes or membrane vesicles . The reaction was initiated by adding 1 mM ATP and terminated by 1 μg/ml FCCP (carbonyl cyanide p-trifluoromethoxyphenylhydrazone). For the tests of proton permeability of membrane vesicles prepared from E. coli expressing FoF1 and its variants, protons were pumped by respiratory NADH oxidation and the proton gradient was assessed by ACMA quenching. The reaction mixture containing 10 mM Hepes/KOH (pH 7.5), 100 mM KCl, 5 mM MgCl2, 0.3 μg/ml ACMA, 100 μg/ml (Figures 2A, 2B, 5B and 5C) or 15 μg/ml (Figures 4A and 4B) membrane vesicles, was pre-incubated at 45 °C and the reaction was started by the addition of NADH [final concentration, 0.3 mM (Figures 2A, 2B, 5B and 5C) or 0.15 mM (Figures 4A and 4B)]. For the tests of proton permeability of proteoliposomes, proteoliposomes (15 μl) pre-incubated in 100 mM KCl were added to the reaction mixture (1.2 ml) containing 10 mM Hepes/NaOH (pH 7.5), 0.25 M sucrose, 5 mM MgSO4 and 0.3 μg/ml ACMA. After 1 min at 40 °C, valinomycin (final concentration, 1 ng/ml) was added to induce inside-negative membrane potential. After the reaction, FCCP (final concentration, 1 μg/ml) was added to see the level without a proton gradient. For each set of experiments, membrane vesicles and proteoliposomes were prepared on the same day using the same procedures. Proteins were analysed by SDS/PAGE (12–20% polyacrylamide in 0.1% SDS). Proteins were stained with Coomassie Brilliant Blue. Protein concentrations were determined with the BCA (bicinchoninic acid) Protein Assay Kit from Pierce, with BSA as a standard.
Effects of the mutation at position aR169
Mutants of thermophilic FoF1 with replacement of aR169 with glutamate, alanine, valine, isoleucine, lysine, phenylalanine, aspartate or tryptophan were expressed in the membranes of the host E. coli cells at approximately half-yield of that of wt-FoF1. All of the mutants, except for the aR169D mutant, contained a set of eight subunits of FoF1 (Figure 1A). FoF1 containing aR169D lacked the Fo-a subunit and was not analysed further. The purified FoF1s were reconstituted into liposomes and ATPase activities with or without pre-treatment by DCCD were measured (Figure 1B). For wt-FoF1, 85% of the ATPase activity was sensitive to DCCD inhibition, which represents proton-coupled activity (Figure 1B). ATPase activities of all aR169 mutants were low, approx. 30% of wt-FoF1, and DCCD-insensitive. As expected, these mutant FoF1s did not show ATP-driven proton-pumping activity (Figure 1C). Membrane vesicles prepared from E. coli cells expressing wt-FoF1 synthesized ATP by respiratory NADH oxidation at a rate of 45 nmol of ATP/min per mg of protein, but vesicles containing the mutant FoF1s did not (<1% of wt-FoF1). Thus the mutant FoF1s with substituted aR169 lost the proton-coupled ATP hydrolysis/synthesis activity. This implies that rotation of the rotor shaft (c-ring/γε) of FoF1 is prevented in these mutants.
Purified FoF1s that have mutation at aR169 of Fo-a
Proton permeability of the aR169 mutants
When ADP and Pi, substrates for ATP synthesis, are omitted from the assay mixtures for ATP synthesis described above, the maximum proton gradient should be built by NADH oxidation. The vesicles containing wt-FoF1 and the mutant FoF1s with the substitutions aR169I, aR169K, aR169F or aR169W maintained the maximum proton gradient (Figure 2A). The mutants of FoF1 with substitutions aR169E, aR169A or aR169V, however, exhibited only weak ACMA fluorescence quenching, indicating that these three mutants mediated passive proton translocation along the gradient. In contrast with wt-FoF1, the F1 sector cannot block the proton permeation of the Fo sector in these mutant FoF1s. Next, to test proton permeability of Fo, F1-stripped membrane vesicles were prepared with EDTA treatment. Without the F1 sector, wt-Fo acted as a proton channel and the vesicles containing wt-Fo showed only poor quenching of ACMA fluorescence (Figure 2B). Also, vesicles containing Fos with substitutions aR169E, aR169A or aR169V mediated passive proton translocation and showed small ACMA quenching. Since membrane vesicles containing FoF1 (Figure 2A) and Fo (Figure 2B) with the double substitution aR169E/cE56Q did not permeate proton, the carboxy group of cE56 is involved in this proton permeation pathway. In contrast, vesicles containing mutant Fos with substitutions aR169I, aR169K, aR169F or aR169W showed maximum ACMA quenching, indicating that transfer of protons from Fo-a to a carboxy group of Fo-c would be blocked. Proton permeability of proteoliposomes containing purified FoF1 was also tested. In this case, KCl (0.5 M) was loaded into the inside of proteoliposomes and the inside-negative membrane potential was induced by valinomycin. As shown in Figure 2(C), protons flowed into proteoliposomes containing the mutant FoF1s with the substitutions aR169E, aR169A or aR169V. Among the three mutants, aR169E was most proton-permeable, aR169A the next and aR169V the least proton-permeable, as observed for the membrane vesicles. The proteoliposomes containing other mutant FoF1s including aR169E/cE56Q, as well as wt-FoF1, blocked proton influx. From the results in Figure 2, we conclude that the three mutant Fos with substitutions aR169E, aR169A or aR169V can mediate the passive proton translocation and that cE56 is involved in this translocation.
Proton permeation mediated by the aR169 mutants
Fusion of Fo-a and decamer Fo-c
To investigate whether the rotation of the c-ring would be necessary for the mutant FoF1s to undergo proton translocation, a ‘rotation-impossible’ mutant was produced. We previously produced (c10)FoF1, in which ten copies of the Fo-c subunit in the c-ring were fused into a single polypeptide, and demonstrated that (c10)FoF1 was active in proton-coupled ATP hydrolysis/synthesis . Starting from this (c10)FoF1, we generated (c10-a)FoF1 in which the c-ring is made up of c10 and Fo-a connected by a linker containing a thrombin recognition sequence (Figure 3A). (c10-a)FoF1 was expressed in membranes of E. coli cells with the expression level being approximately half of that of wt-FoF1. It was remarkable that membrane insertion of a nascent polypeptide that had 25 transmembrane helices (20 for c10 and five for Fo-a) occurred normally. (c10-a)FoF1 was purified as a stable complex and SDS/PAGE showed the appearance of a new protein band above the α subunit band and disappearance of bands of c10 (overlapped with β subunit band) and Fo-a (Figure 3B). The new band was confirmed to be the fused c10-a protein by Western blotting with anti-Fo-c and anti-Fo-a antibodies (results not shown). When (c10-a)FoF1 was treated with thrombin, the c10-a band disappeared and the bands of Fo-a and c10 appeared instead. Relative staining intensities of each subunit band were similar to those of wt-FoF1, ensuring their normal subunit stoichiometry. Because c-ring rotation relative to Fo-a was impossible, (c10-a)FoF1 was unable to catalyse ATP-driven proton translocation (Figure 3B). However, as expected, the thrombin-treated (c10-a)FoF1 catalysed ATP-driven proton-pumping. Therefore subunits of (c10-a)FoF1 were assembled into a native-like structure. These results also provided evidence for the proposed membrane topology of Fo-a, i.e. periplasmic location of its N-terminus.
Rotation-impossible FoF1 with a fused c10-a subunit
Proton permeation through (c10-a)FoF1 and its variants
The (c10-a)FoF1 variants with mutation of aR169E [(c10-aR169E)FoF1] or aR169A [(c10-aR169A)FoF1] were also expressed in E. coli at the same expression level as that of (c10-a)FoF1. They were purified as stable complexes (Figure 3C). Membrane vesicles containing (c10-a)FoF1 maintained the proton gradient generated by NADH oxidation (Figure 4A). Since rotation of the c-ring is prevented, (c10-a)Fo can no longer work as a proton translocator. In contrast, the vesicles containing (c10-aR169A)FoF1 or (c10-aR169E)FoF1 partially lost the gradient. Parallel results were obtained for the F1-stripped membrane vesicles (Figure 4B). The proton permeability of proteoliposomes containing purified (c10-a)FoF1 and its variants was also tested (Figure 4C). Even though the rate was low, (c10-aR169A)FoF1 and (c10-aR169E)FoF1 in the proteoliposomes carried out passive proton translocation, but (c10-a)FoF1 did not. In any of the above tests, (c10-aR169E)FoF1 was always more efficient in proton translocation than (c10-aR169A)FoF1. Because c-ring rotation was impossible in (c10-a)FoF1 and its variants, these results provide evidence that protons can pass through Fo without accompanying c-ring rotation in the case when Arg169 of Fo-a is replaced with a glutamate or alanine residue.
Proton permeation mediated by rotation-impossible (c10-a)FoF1, (c10-aR169A)FoF1 and (c10-aR169E)FoF1
The arginine-switched mutant blocks proton permeation
In E. coli, the essential arginine residue in Fo-a can be transferred to the position of aQ252 and this arginine-switched (aR210Q/Q252R)FoF1 retains small proton-coupled activity [11,18,19]. E. coli aQ252 corresponds to aQ217 in Bacillus PS3, and we made a Bacillus version of the arginine-switched (aR169G/Q217R)FoF1, and the activity of the membrane vesicles containing the arginine-switched mutant was compared with that of (aR169G)FoF1. As far as we tested, both mutants could not catalyse ATP-driven proton translocation (Figure 5A). The arginine-switched FoF1, but not (aR169G)FoF1, can maintain the proton gradient established by NADH oxidation (Figure 5B). When the F1 sector was removed from membrane vesicles, wt-Fo and (aR169G)Fo were proton permeable, but the arginine-switched Fo was not (Figure 5C). These results showed that (aR169G)Fo allows passive proton translocation and, similar to E. coli (aR210Q/Q252R)Fo [11,18], the transferred arginine residue at 217 can prevent this.
ATP-driven proton-pumping activity and proton permeation of the aR169G and aR169G/Q217R mutants
Our observations on the mutant FoF1s are well explained by the ‘arginine proton barrier’ model (Figure 6A). It has been thought that, in the native FoF1, a proton coming through a half-channel in Fo-a is transferred to a carboxy group of Fo-c in the c-ring, enabling this Fo-c to move into the lipid-surrounding environment. The c-ring makes almost one revolution carrying the protonated carboxy group near the other half-channel of Fo-a, into which the proton is transferred. Shortcut of protons, that is, from one half-channel to the other via a carboxy group of Fo-c without c-ring rotation, is prevented by the arginine residue of Fo-a that would be located between two half-channels. Molecular simulation indicates that a positive charge of arginine might contribute to this function [20,21]. If this arginine residue is replaced by a small residue (glycine, alanine or valine) or an acidic residue (glutamate), protons could have a chance to take a shortcut pathway (Figure 6B). If this arginine residue is replaced by phenylalanine, isoleucine, lysine or tryptophan, a proton is not transferred to a carboxy group of Fo-c (Figure 6C). The model explains why the proton-permeable aR169 mutants do not catalyse uncoupled (DCCD-insensitive) ATPase (Figure 1B) and why F1 cannot block proton permeation of F1-stripped membranes of these mutants (Figures 2A, 4A and 6B). The usual cause of uncoupled ATPase is a loose association of F1 with Fo (the interface between γε and the c-ring is loose or the second stalk is not functional); the c-ring in Fo freely rotates and permeates protons without ATP synthesis/hydrolysis and F1 hydrolyses ATP without proton translocation. However, in these mutants, since protons permeate not by free c-ring rotation but through a proton shortcut pathway, the tight association between Fo and F1 is not damaged and uncoupled ATPase is not stimulated. In the intact FoF1, proton permeation by free c-ring rotation is blocked by tightly associated F1. In the proton-permeable mutant FoF1s, proton permeation is mediated through a shortcut without c-ring rotation and is therefore not prevented by F1. Thus the conserved arginine residue of Fo-a plays dual roles: to block a futile proton shortcut as a proton barrier and to facilitate proton transfer between Fo-a half-channels and the Fo-c carboxy group at the Fo-a/c-ring interface.
Arginine proton barrier model for the role of aR169 in FoF1
Noriyo Mitome and Sakurako Ono designed and performed experiments, analysed data and wrote the paper; Hiroki Sato performed experiments and analysed data; Toshiharu Suzuki and Nobuhito Sone designed experiments and analysed data; Masasuke Yoshida designed experiments, analysed data and wrote the paper.
We thank B. Feniouk for critical reading of the manuscript prior to submission.
This work has been supported in part by the Grants-in-aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan [grant number 18107004 (to M.Y.), and grant numbers 19042006, 19770103 (to N.M.)].