F0 of ATP Synthase Is a Rotary Proton Channel

Coupling of proton flow and rotation in the F0 motor of ATP synthase was investigated using the thermophilic Bacillus PS3 enzyme expressed functionally inEscherichia coli cells. Cysteine residues introduced into the N-terminal regions of subunits b and c of ATP synthase (bL2C/cS2C) were readily oxidized by treating the expressing cells with CuCl2 to form predominantly a b-c cross-link with b-b andc-c cross-links being minor products. The oxidized ATP synthases, either in the inverted membrane vesicles or in the reconstituted proteoliposomes, showed drastically decreased proton pumping and ATPase activities compared with the reduced ones. Also, the oxidized F0, either in the F1-stripped inverted vesicles or in the reconstituted F0-proteoliposomes, hardly mediated passive proton translocation through F0. Careful analysis using single mutants (bL2C or cS2C) as controls indicated that the b-c cross-link was responsible for these defects. Thus, rotation of the c-oligomer ring relative to subunit b is obligatory for proton translocation; if there is no rotation of the c-ring there is no proton flow through F0.

ATP synthases catalyze ATP synthesis/hydrolysis coupled with a transmembrane H ϩ (proton) translocation in bacteria, chloroplasts, and mitochondria (1)(2)(3). The enzyme is composed of two portions, a water-soluble F 1 , which has catalytic sites for ATP synthesis/hydrolysis (4), and a membrane-integrated F 0 , which mediates H ϩ translocation. The bacterial enzyme has the simplest subunit structure, ␣ 3 ␤ 3 ␥ 1 ␦ 1 ⑀ 1 for F 1 and a 1 b 2 c 10 -11(?) for F 0 . F 1 is easily and reversibly detached from F 0 by removal of Mg 2ϩ in a low ionic strength solution. F 1 is by itself a rotary motor driven by ATP hydrolysis (5,6) in which a central stalk made of ␥ and ⑀ subunits rotates relative to the surrounding ␣ 3 ␤ 3 hexamer ring (7,8). Remaining F 0 sector in the membrane acts as a proton channel that mediates passive proton translocation across the membrane. F 0 in ATP synthase is thought to work as a rotary motor driven by the energy of proton translocation down the electrochemical potential. Structural studies on F 0 with electron microscopy (9) and atomic force microscopy (10 -12) have suggested that subunits a and b 2 are peripherally located outside of a ring of subunit c oligomers (c-ring). A low resolution crystal structure of an F 1 ϩc 10 subcomplex from yeast mitochondria revealed a tight interaction between ␥⑀ subunits of F 1 and the c-ring of F 0 (13). The cross-links were readily made between introduced cysteines of subunit c and ␥⑀ subunits of F 1 without losing functional coupling between F 1 and F 0 (14). ATP facilitated the movement of subunit c relative to subunit a that was assessed by the a-c cross-link (15). A side stalk, made of b 2 and ␦, connects the stator of F 0 and that of F 1 and prevents the stators from being dragged by rotation of the central stalk (16). From these and other observations, it is generally accepted that the c-ring rotates relative to stator subunits ab 2 (17)(18)(19). Thus, proton influx into the cytoplasm through F 0 (in the case of mitochondria, into the matrix) would cause rotation of the c-ring and hence the central ␥⑀ stalk, which then enforces each catalytic site in F 1 to synthesize ATP. As a reverse reaction, ATP hydrolysis in F 1 drives reverse rotation of the ␥⑀ stalk and c-ring, which causes proton efflux through F 0 . To explore the mechanism of proton flow through F 0 and its coupling with rotation, measurement of the proton flow together with the c-ring rotation is absolutely required. The study has been impeded by the unstable nature of ATP synthases; the structural and functional integrity of the enzyme is easily damaged during experimental procedures (20,21). For example, the cross-linking between subunits b and c in Escherichia coli ATP synthase impairs coupling between proton pumping and ATPase activity (21). To overcome this difficulty, we have established an expression system for a stable ATP synthase of thermophilic Bacillus PS3 (TF 0 F 1 ) 1 in E. coli cells. Cysteine residues introduced into subunits b and c of TF 0 F 1 were readily cross-linked in the presence of CuCl 2 . The resultant enzyme lost both proton pumping and ATPase activity, an indication of retaining tight coupling after the cross-linking. The passive proton translocation through F 0 was also disabled by this cross-link. Thus, the c-ring must rotate for protons to pass through F 0 .

EXPERIMENTAL PROCEDURES
Construction of an Expression Vector for TF 0 F 1 -DNA manipulations were carried out by following the methods of the literature (22). A plasmid pTR-k⑀, which is an expression vector for ␣ 3 ␤ 3 ␥⑀ subcomplex of Bacillus PS3 F 1 -ATPase (23), was used as a start material. A 4.4kilobase pair DNA fragment containing Bacillus PS3 uncBEFHAЈ genes (coding for a, c, b, ␦, and ␣ subunits of TF 0 F 1 ) was amplified by PCR from the plasmid pUC119/TF 0 F 1 (24) using primers 5Ј-CCGCGG-GAATTCTAAGAAGGAGATATACATATGGAGCATAAAGCGCCGCT-TGTCG-3Ј and 5Ј-GGCCGATCGGTACCAGCGCGTCGATCGCTTTAA-* This work was supported in part by Human Frontiers Science Program Organization Grant RG15/1998-M. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Expression and Purification of TF 0 F 1 -E. coli DK8 (23) harboring pTR19-ASDS was aerobically cultivated in 2ϫ YT medium (22) containing ampicillin (100 g/ml) at 37°C for 20 h. Cells (40 g, wet weight) harvested from the culture were dissolved in 280 ml of buffer PA3 (10 mM HEPES/KOH, pH 7.5, containing 5 mM MgCl 2 and 10% glycerol) and disrupted by a French press (1200 millibars, twice). After removing the cell debris, the membrane fraction was collected by centrifugation (200,000 ϫ g, 1 h, 4°C). The membrane fraction obtained was washed, precipitated, and then suspended in 40 ml of buffer PA3. This suspension was used as the inverted membrane vesicles in this study. After washing the inverted vesicles with a 2-fold volume of buffer PA3 by centrifugation, the vesicles were dissolved in 120 ml of buffer PA3 containing 2% Triton X-100 and 1% sodium cholate and subjected to sonification for 2 min at room temperature. The solution containing solubilized TF 0 F 1 was obtained by a centrifugation (140,000 ϫ g, 15 min, 25°C). The solution was diluted 3-fold with buffer M (20 mM potassium phosphate buffer, pH 7.5, containing 100 mM KCl) containing 20 mM imidazole and applied to a Ni-nitrilotriacetic acid column (Qiagen, Germany). After washing with 10 volumes of buffer MT (buffer M supplemented with 0.5% Triton X-100), TF 0 F 1 was eluted with buffer MT containing 200 mM imidazole. Then the eluted fraction was precipitated by ammonium sulfate in the presence of 1% sodium cholate (27). The precipitate was dissolved in a small volume of 50 mM Tris/HCl, pH 8.0. About 10 mg of the purified TF 0 F 1 were routinely obtained from a 1-liter culture. The homogeneity of the purified TF 0 F 1 was judged by a SDS-PAGE analysis (see Fig. 1A).
Oxidation of cysteine residues in F 0 was performed as follows. Recombinant cells expressing TF 0 F 1 (10 g) were washed and suspended in 40 ml of buffer PA3 containing 200 g/ml CuCl 2 and incubated at 25°C for 30 min. The recombinant TF 0 F 1 has one cysteine residue at position 27 of subunit a, but this residue has no reactivity under these oxidizing conditions. After adding 10 mM EDTA to the solution, the cells were washed with 40 ml of buffer PA3 twice, and inverted membrane vesicles were prepared by the procedures described above.
Reconstitution of TF 0 F 1 into Liposomes-Reconstitution of proteoliposomes was performed by the method of Kaim and Dimroth (28) with some modifications. Phospholipid (soybean L-␣-phosphatidylcholine, type II-S, Sigma) was suspended in buffer PA3 at the concentration of 44 mg/ml and sonicated for 3 min on ice water to prepare liposomes. Either TF 0 F 1 (0.5 mg) or TF 0 (0.2 mg) was added to 200 l of the liposome solutions. The solution was frozen with liquid nitrogen and thawed at room temperature. After adding an equal volume of distilled water, the liposomes were briefly sonicated for 15 s, collected by centrifugation (200,000 ϫ g, 20 min, 4°C), and resuspended in 200 l of buffer PA3.
Analyses of Proton Flow through F 0 -F 1 was removed from the inverted vesicles by washing with 0.2 mM EDTA (29), and the resultant F 1 -stripped inverted vesicles were used for analysis of proton efflux through F 0 . Isolation of pure F 0 complexes and the analysis of proton influx through F 0 were performed as described previously (30).
Other Methods-ATPase activity was measured at 37°C using an ATP-regenerating system (23). The activity that hydrolyzes 1 mol of ATP per min is defined as 1 unit. DCCD sensitivity was measured as follows. Either the inverted vesicles (10 mg of protein/ml) or proteoliposomes (prepared as above) were added to an equal volume of 100 mM Tris-HCl (pH 8.8) containing 100 M DCCD. These conditions were chosen to avoid undesired DCCD labeling of the catalytic residue (Glu-190) of ␤ subunits of F 1 . After an incubation at 25°C for 30 min, the vesicles were diluted 100-fold and subjected to the measurement of ATPase activity. ATP-driven H ϩ -pumping activity of the vesicles was measured by the method of Aggeler et al. (31) using a Jasco model 720 spectrofluorometer at 42°C at 480 nm using an excitation wavelength of 410 nm. Authentic Bacillus PS3 F 0 F 1 was purified from Bacillus PS3 cells as described previously (27). Two-dimensional SDS-PAGE was carried out as performed previously (23). Protein concentrations were determined using the BCA protein assay kit from Pierce, with bovine serum albumin as a standard.

RESULTS
Expression of TF 0 F 1 in E. coli Cells-An operon structure containing unc genes coding for TF 0 F 1 was introduced into a downstream of a strong promoter, trc, and then expressed in a F 0 F 1 -deficient E. coli strain, DK8. The resultant recombinant strain (DK8/pTR19-ASDS) acquired an ability to form colonies on succinate minimum medium plates within 2 days (the diameter is about 2 mm), which is almost the same growth rate of a wild-type E. coli strain (JM109), indicating that the recombinant TF 0 F 1 is functional as ATP synthase in the E. coli cell. TF 0 F 1 consisting of eight subunits was constitutively expressed in the plasma membranes, which amounted to ϳ20% of the whole membrane proteins (Fig. 1A, lane 1). The purified TF 0 F 1 was comprised of eight kinds of subunits, the same as authentic TF 0 F 1 purified from the original Bacillus PS3 cells 2 (Fig. 1A,  lanes 2 and 3). The membranes of expressing E. coli exhibited 1.0 unit/mg of protein (at 37°C) of ATPase activity. More than 80% of the activity was inactivated by a 50-min incubation with 50 M DCCD or by a 20-min incubation with 100 M DCCD (Fig. 1B). This inhibition is comparable or slightly more efficient than that (75%) observed for the authentic TF 0 F 1 purified from Bacillus PS3 cells (27). This inactivation is due to labeling subunit c but not labeling catalytic glutamic acid (Glu-190) in the ␤ subunit of F 1 because ATPase activity was unaffected by DCCD treatment when measured in the presence of lauryldimethylamine oxide (LDAO), which unleashes ATPase activity of F 1 from F 0 . The membrane ATPase was inhibited almost completely by 5 mM azide, an inhibitor of ATPase activity of F 1 and ATP synthase.
Cross-linking b and c Subunits at the N-terminal Regions-Two cysteine residues were introduced into TF 0 F 1 by substituting Leu-2 of subunit b and Ser-2 of subunit c to obtain a mutant TF 0 F 1 , bL2C/cS2C. To use as controls, we also made two single mutants, bL2C or cS2C, that had one substituted cysteine in subunit b or c. These mutants were expressed in E. coli DK8 cells. N-terminal regions of subunits b and c are 2 On the gel of Fig. 1A, a band of recombinant b subunit (lanes 1 and  2) moved slightly slower than the counterpart of the TF 0 F 1 purified from Bacillus PS3 cells (lane 3). N-terminal sequencing of the recombinant subunit revealed that an N-terminal segment comprising 11 residues is not processed in the E. coli cells, which is different from the authentic enzyme (23). The low mobility of recombinant ␤ subunit is ascribed to 10 residues of a histidine tag introduced at the N terminus. located in the periplasmic surface of plasma membrane (21) and as expected, the introduced cysteine residues in bL2C/ cS2C formed a disulfide cross-link by treating expressing E. coli cells with an oxidant, 200 M CuCl 2 . After careful removal of the oxidant, inverted membrane vesicles were prepared from the cells. In the presence of LDAO, the vesicles from the mutants showed ATPase activities similar to that of the wild type, indicating that the wild type and mutants were expressed at comparable levels (wild type, 6.6 Ϯ 0.1 units/mg; bL2C, 5.8 Ϯ 0.1 units/mg; cS2C, 5.3 Ϯ 0.1 units/mg; bL2C/cS2C, 6.8 Ϯ 0.2 units/mg of membrane protein).
Mutant TF 0 F 1 s were purified from the vesicles and analyzed by SDS-PAGE after incubation with or without 50 mM DTT at 25°C for 1 h (Fig. 2A). The mutant TF 0 F 1 s incubated with DTT showed eight bands, the same as that of wild type (lanes 1-4). However, those not exposed to DTT had an additional one (bL2C and cS2C, lanes 6 and 7) or three band(s) (bL2C/cS2C, lane 8). Based on the N-terminal peptide sequences and estimated molecular sizes of the bands, these new bands were identified as disulfide cross-linked products of b-b, b-c, and c-c as indicated by arrows in Fig. 2A. Cross-link yields in bL2C/ cS2C were analyzed by two-dimensional SDS-PAGE, first in non-reducing and second in reducing conditions (Fig. 2B). The three bands were separated in the second electrophoresis into spot(s) corresponding to monomeric subunit b and/or c. In bacterial ATP synthase, subunit b has been known to exist as a homodimer (32), and only one of two copies of subunit b is assumed to lie adjacent to the c-ring that is able to form a cross-link (21). Taking this into account, the yields of b-b and b-c cross-links in bL2C/cS2C were estimated from the densities of spots to be 16 Ϯ 5 and 68 Ϯ 7%, respectively. A further increase of the cross-links of b-b or b-c was not observed in bL2C/cS2C even though either 500 M CuCl 2 or 1 mM copper/ phenanthroline was used as an oxidant. It is worth mentioning that a significant amount of b-c cross-link was formed (yield ϳ40%) even in the membranes that were not treated with CuCl 2 (data not shown). 3 It appears evident that there is close proximity between the N-terminal ends of subunit b(s) and one (or more) of subunit c(s).
Effects of b-c Cross-link on the Activities of TF 0 F 1 -The inverted vesicles prepared from the cells oxidized by CuCl 2 treatment were incubated with or without 50 mM DTT for 1 h, and H ϩ -pumping activity was analyzed. DTT-treated vesicles of all mutants showed substantial ATP-driven H ϩ -pumping activities, comparable with that of the wild type (Fig. 3A, left panel). Therefore, the introduced cysteine residue(s) at position 2 of subunit b and that of subunit c, alone or together, do not significantly affect the F 0 function. This was also the case for the vesicles of mutants bL2C and cS2C that were not reduced by DTT. However, the vesicles of a mutant bL2C/cS2C (without DTT treatment) had drastically decreased H ϩ -pumping activity (Fig. 3A, right panel). Membranes of the oxidized vesicles from bL2C/cS2C are capable of holding the electrochemical potential of protons generated by NADH oxidation, as described later, and proton leak cannot be a reason for the apparent loss of H ϩ pumping. The inactivation of bL2C/cS2C by oxidation was also observed for ATPase activity. Oxidized vesicles from bL2C/cS2C without DTT treatment retained only 37% of the ATPase activity of that of the DTT-treated ones whereas activities of the vesicles from the single mutants (and wild type) were hardly affected by oxidation-reduction treatment (Fig. 3B). The inhibition of ATPase was completely recov- supplemented with 0.3 g/ml ACMA. The inverted vesicles prepared from the oxidized cells were added to the solution to a final concentration of 0.5 mg of membrane protein/ml. The reaction was initiated by adding 1 mM ATP and terminated by 1 g/ml carbonyl cyanide ptrifluoromethoxyphenylhydrazone (FCCP). B, ATPase activities of the inverted vesicles. ATPase activities were measured by the ATP-regenerating system in buffer PA5 (50 mM HEPES/KOH, pH 7.5, containing 100 mM KCl, 5 mM MgCl 2 ) supplemented with 1 mM ATP, 5 mM KCN, and 1 g/ml carbonyl cyanide p-trifluoromethoxyphenylhydrazone at 37°C. Bars indicate ATPase activities of samples treated with (black) or without DTT (white). C, H ϩ -pumping activity of reconstituted TF 0 F 1 proteoliposomes. The proteoliposomes were added to a final concentration of 21 g of TF 0 F 1 /ml. D, ATPase activity of reconstituted TF 0 F 1 proteoliposomes. Analytical conditions of C and D are the same as those of A and B. In these analyses, the inverted vesicles and reconstituted TF 0 F 1 proteoliposomes were incubated with or without 50 mM DTT at 25°C for 1 h before the measurement. ered by adding 0.1% LDAO in the ATPase assay mixture (data not shown), confirming that the failure was not in F 1 but in F 0 . The same experiments were repeated for the proteoliposomes reconstituted from purified TF 0 F 1 and soybean phospholipids (Fig. 3, C and D). As observed for the inverted vesicles, inactivation of H ϩ -pumping and ATPase activities was evident only for the oxidized bL2C/cS2C. To summarize the results, only oxidized TF 0 F 1 containing double mutations bL2C/cS2C has a defect in ATP hydrolysis and in H ϩ pumping. This defect is caused by a b-c cross-link and cannot be ascribed to the b-b and c-c cross-links. This is because (as shown in Fig. 2A) the amount of b-b and c-c cross-links produced in the oxidized bL2C/cS2C was too little to account for the observed inactivation. Furthermore, the oxidized single mutants, bL2C and cS2C, contained more b-b and c-c cross-links, respectively, than those in the oxidized bL2C/cS2C as shown in Fig. 2A but still their activities were not inactivated significantly. On the contrary, the amount of b-c cross-link in oxidized bL2C/cS2C (ϳ68%) agrees fairly well with the degree of inactivation (ϳ60% for ATPase activity). Thus, prevention of movement of the c-ring relative to subunit b is fatal for the catalytic function of ATP synthase with proper coupling.
Effects of b-c Cross-link on Proton Flow through F 0 -Inverted vesicles that were prepared from CuCl 2 -oxidized cells were washed with 0.2 mM EDTA to obtain F 1 -stripped inverted vesicles. The electrochemical potential of protons was generated across the membrane of inverted F 0 vesicles by the respiratory chain on the vesicles using NADH as a substrate, and the downhill proton efflux through F 0 was assessed by monitoring fluorescence quenching of ACMA. Without F 1 , protons taken up in vesicles by respiration easily diffused out through F 0 , and only a small fluorescence quenching was maintained at steady state as a balance between activities of respiration and proton flow through F 0 (Fig. 4A, wild type). Prior reducing treatment of the vesicles by DTT did not change the result significantly. Inverted F 0 vesicles prepared from single mutants, bL2C and cS2C, behaved similarly; the quenching was small and the effect of DTT treatment was minor (Fig. 4A,  bL2C and cS2C). Also the extent of fluorescence quenching was small for the DTT-treated inverted F 0 vesicles of bL2C/cS2C (Fig. 4A, bL2C/cS2C). However, when the same inverted F 0 vesicles of bL2C/cS2C were subjected to the test without prior DTT treatment, remarkable fluorescence quenching was induced in response to the addition of NADH. The magnitude of quenching by NADH matched well the one observed for the inverted vesicles, without F 1 -stripping treatment, prepared from cells expressing wild-type F 0 F 1 (data not shown). This result clearly indicates that b-c cross-link blocks proton efflux through F 0 . It also implies that b-c cross-link does not make F 0 leak protons.
To confirm further the above contention, F 0 was isolated from purified TF 0 F 1 and reconstituted into proteoliposomes. F 0 proteoliposomes were incubated with or without DTT and then loaded with 0.5 M KCl. Membrane potential (inside negative) was generated by the addition of valinomycin, and downhill proton influx through F 0 was assessed by monitoring fluorescence quenching of ACMA (Fig. 4B). In the case of F 0 proteoliposomes of the wild type, irrespective of whether they were treated with DTT or not, valinomycin induced significant fluorescence quenching that reflected proton flow through F 0 . F 0 proteoliposomes of bL2C/cS2C with prior DTT treatment displayed a similar extent of fluorescence quenching. On the contrary, the quenching was greatly suppressed when F 0 proteoliposomes of bL2C/cS2C without prior DTT treatment were examined. In any case, preincubation of F 0 proteoliposomes with DCCD resulted in complete abolishment of valinomycin-induced quenching. These results, together with those of the inverted F 0 vesicles, led to the conclusion that proton efflux and influx through F 0 are blocked by b-c cross-link. DISCUSSION The major message of this report concerns the relation between the c-ring rotation and the proton flow through F 0 . As illustrated in Fig. 5, the F 0 with a disulfide cross-link between subunits b and c was unable to mediate proton translocation (Fig. 4). With prevented proton translocation, ATP hydrolysis was also prevented, suggesting the retention of tight coupling between F 0 and F 1 in the cross-link containing TF 0 F 1 (Fig. 3). Regardless of the directions of proton translocation, either the periplasmic side to the cytoplasmic side or the cytoplasmic side to the periplasmic side, translocations were equally blocked by the cross-linking. The inactivation is reversible; reduction of the disulfide restored the proton translocation by F 0 and ATPdriven proton pumping by TF 0 F 1 . These results strongly indicate that protons cannot pass through F 0 without rotation of the c-ring, or conversely, rotation of the c-ring must accompany proton translocation. Cross-linking of b and c subunits caused neither a proton leak nor the unleashing of activation of ATPase of F 1 . Thus the possibility that the cross-link itself disrupts the function of F 0 is minimal, if not null. In our experimental setups, proton translocations across membranes down the ⌬ Hϩ were measured. Previous data from other laboratories indicate that the coupling of proton translocation and c-ring rotation is maintained even in the absence of ⌬ Hϩ . Dimroth's group has detected 22 Na ϩ /Na ϩ exchange across proteoliposome membranes in the absence of ⌬ Hϩ through F 0 isolated from Na ϩ -transporting ATP synthase of Propionigenium modestum (33). They did not examine the movement of the c-ring but assumed that the back-and-forth thermal rotary motion of the c-ring in F 0 was responsible for the exchange. Vesicles and proteoliposomes were incubated with or without 50 mM DTT prior to the analysis. A, proton efflux through F 0 of the F 1 -stripped inverted vesicles (final concentration, 40 g of membrane protein/ml) was monitored by fluorescence quenching of ACMA. The electrochemical potential of protons (inside positive) was generated by adding 0.2 mM NADH to the assay mixture. B, proton influx through F 0 of the reconstituted F 0 proteoliposomes (final concentration, 5 g of TF 0 /ml) was analyzed in 10 mM HEPES/NaOH (pH 7.5) containing 5 mM KCl and 0.5 M sucrose as performed previously (30). Membrane potential (inside negative) was generated by supplementing 2.5 ng/ml valinomycin to proteoliposomes loaded with 0.5 M KCl inside. The effect of DCCD on the proton translocation was measured for the proteoliposomes that were treated with 50 mM DTT and then treated with 100 M DCCD before the analysis. FCCP, carbonyl cyanide p-trifluoromethoxyphenylhydrazone.
Fillingame and his colleagues (21) demonstrated by using the b-c cross-linking of uncoupled E. coli ATP synthase that the subunit c adjacent to subunit b is mobile and exchanges with subunits c that initially occupied other positions. This exchange occurs independently of ATP, suggesting thermal motion of the c-ring. Although not tested by their experiments, it is natural to assume that this thermal motion accompanies proton translocation. Taken together, it is safe to conclude that isolated F 0 is a rotary proton channel in which the "friction" between rotor and stator is small enough to allow free fluctuating rotary motion even by environmental thermal energy. Nevertheless, this friction may differentiate the F 0 channel from other open ion channels; the ion conductance of the former is several orders lower than that of the latter (F 0 , ϳ7 s Ϫ1 (33); potassium channel, Ͼ10 6 s Ϫ1 (34); aquapolin-1, Ͼ10 9 s Ϫ1 (35)). Once ⌬ Hϩ is applied, the F 0 rotary channel starts to rotate unidirectionally and protons are transported also unidirectionally across membranes.  5. Proton flow through F 0 coupled with c-ring rotation  and its prevention by the b-c cross-linking. A, F 0 in the reduced form. The c-ring can rotate in either direction corresponding to either efflux or influx of protons. B, F 0 in the oxidized form. The b-c cross-link physically prevents rotation of the c-ring, and protons can no longer pass through F 0 . F 0 is derived from TF 0 F 1 (bL2C/cS2C) that has cysteines at N-terminal regions of subunits b and c.