The N-terminal region of the ϵ subunit from cyanobacterial ATP synthase alone can inhibit ATPase activity

ATP hydrolysis activity catalyzed by chloroplast and proteobacterial ATP synthase is inhibited by their ϵ subunits. To clarify the function of the ϵ subunit from phototrophs, here we analyzed the ϵ subunit–mediated inhibition (ϵ-inhibition) of cyanobacterial F1-ATPase, a subcomplex of ATP synthase obtained from the thermophilic cyanobacterium Thermosynechococcus elongatus BP-1. We generated three C-terminal α-helix null ϵ-mutants; one lacked the C-terminal α-helices, and in the other two, the C-terminal conformation could be locked by a disulfide bond formed between two α-helices or an α-helix and a β-sandwich structure. All of these ϵ-mutants maintained ATPase-inhibiting competency. We then used single-molecule observation techniques to analyze the rotary motion of F1-ATPase in the presence of these ϵ-mutants. The stop angular position of the γ subunit in the presence of the ϵ-mutant was identical to that in the presence of the WT ϵ. Using magnetic tweezers, we examined recovery from the inhibited rotation and observed restoration of rotation by 80° forcing of the γ subunit in the case of the ADP-inhibited form, but not when the rotation was inhibited by the ϵ-mutants or by the WT ϵ subunit. These results imply that the C-terminal α-helix domain of the ϵ subunit of cyanobacterial enzyme does not directly inhibit ATP hydrolysis and that its N-terminal domain alone can inhibit the hydrolysis activity. Notably, this property differed from that of the proteobacterial ϵ, which could not tightly inhibit rotation. We conclude that phototrophs and heterotrophs differ in the ϵ subunit–mediated regulation of ATP synthase.

ATP synthase synthesizes ATP from ADP and P i using membrane potential, which is generated as a proton gradient across the bacterial plasma membrane, mitochondrial inner membrane, or chloroplast thylakoid membrane (1)(2)(3). ATP synthase is composed of the hydrophilic portion F 1 and the membraneembedded hydrophobic portion F o . The subunit composition of F 1 is ␣ 3 ␤ 3 ␥␦⑀ (4), and that of F o is ab 2 (or bbЈ)c 10 -15 in bacteria and chloroplasts (5)(6)(7). F o of mitochondrial ATP synthase contains additional subunits, such as d, f, and h, which were not observed in those of bacterial and chloroplast-type ATP synthase (8). F 1 is often referred to as F 1 -ATPase because F 1 itself can catalyze the ATP hydrolysis reaction. This activity was thoroughly studied, and finally the rotary motion during the ATP hydrolysis reaction was directly visualized using singlemolecule observation (9,10). In addition, the intrinsic regulation of this rotary motion caused by the ⑀ subunit has also been reported in the case of thermophilic bacterial and cyanobacterial F 1 -ATPases (11)(12)(13).
Because ATP synthase can potentially catalyze ATP hydrolysis when the membrane potential is insufficient for ATP synthesis, regulation of the activity should be important for living cells to avoid futile ATP hydrolysis reactions. One of the most well-known regulatory mechanisms is MgADP inhibition (ADP inhibition), which is induced by occupation of the catalytic site with MgADP and prevents the ATP hydrolysis reaction (14). The ␥ subunits of chloroplast-type ATP synthases, including the cyanobacterial one, possess an insertion region composed of 30 -40 amino acids between the Rosmann-fold domain and the C-terminal domain (CTD) 2 of ␣-helices, which is not observed in the ␥ subunit of other F 1 ; this region has a function in regulating F 1 -ATPase (15, 16). In addition, the ␥ subunit of the chloroplast ATP synthase has an additional nine-amino acid insertion containing a pair of Cys residues at this insertion region, and this Cys pair is key for the redox control by thioredoxin (17,18). Under light conditions, this pair of Cys, which forms a disulfide bond under dark conditions, is reduced by thioredoxin, and consequently the catalytic activity is accelerated (12,19,20).
In addition, F 1 -ATPases from chloroplasts and proteobacteria adopt the ⑀ subunit as an intrinsic inhibitor for the ATP hydrolysis reaction (21,22). The ⑀ subunit is composed of two domains, the N-terminal domain (NTD) with a ␤-sandwich structure and the CTD containing two tandem ␣-helices (23)(24)(25). The inhibitory mechanism by the ⑀ subunit (⑀-inhibition) has been proposed as follows: the ⑀ subunit changes its conformation of two C-terminal ␣-helices from the retracted to the extended form in the enzyme complex in a manner dependent on a change in the microenvironment (26). In the case of the bacterial ⑀ subunit, this conformational change is caused by the binding and release of the ATP molecule at the CTD of ␣-helices (27) and/or the change in membrane potential (28). The extended CTD of ␣-helices of the ⑀ subunit is then inserted into the cavity between the ␣ and ␤ subunits. This conformational change of the ⑀ subunit enables the interaction between the positively charged CTD of the ⑀ subunit and the negatively charged DELSEED motif of the ␤ subunit (29). This electrostatic interaction appeared to be the cause of inhibition of the enzyme because the deletion of the positively charged amino acid residues or the truncation of the CTD ␣-helices of the ⑀ subunit diminished the inhibitory effect (29,30). In contrast, the cyanobacterial ⑀ subunit mutant exhibited different behavior. Truncation of the CTD containing ␣-helices of the ⑀ subunit of cyanobacterial ATP synthase resulted in the decrease of ATP synthesis activity (31), whereas truncation to the same extent of the CTD containing ␣-helices of the ⑀ subunit from proteobacteria resulted in an increase of ATP synthesis activity (32,33). This implies that the ⑀ subunit of cyanobacterial ATP synthase may have a different inhibitory mechanism from those in other organisms. In this study, to reveal the function of the ⑀ subunit from cyanobacteria, three C-terminal ␣-helix null ⑀ subunit mutants (CTD null ⑀-mutants) were constructed; the mutant lacking the C-terminal part containing ␣-helices (⑀ N ), the mutant whose C-terminal conformation can be locked by a disulfide bond formed between two ␣-helices (⑀ CC_SS ), and the other one whose C-terminal conformation can be locked by a disulfide bond formed between ␣-helix in the CTD and ␤-sandwich structure in the N-terminal domain (NTD) (⑀ NC_SS ). The inhibitory properties of these mutant ⑀ subunits were thoroughly examined.

⑀-Mutants at the C-terminal domain
To study the molecular mechanism of the ⑀-inhibition in cyanobacterial ATP synthase, three ⑀ subunit mutants were prepared: ⑀ N , which consists of 83 amino acid residues at the NTD and lacks the CTD part containing ␣-helices; ⑀ CC_SS , whose Ala 99 and Phe 122 were substituted to Cys to allow disulfide bond formation between two ␣-helices (Fig. 1A); and ⑀ NC_SS, whose Thr 46 and Arg 124 were substituted with Cys to allow disulfide bond formation between the ␤-sandwich structure in the NTD and the ␣-helix in the CTD. We then determined the oxidant concentration that is sufficient for disulfide bond formation in ⑀ CC_SS and ⑀ NC_SS, and the reductant concentration to completely reduce these disulfide bonds (Fig. 1B).
In the presence of more than 300 M aldrithiol-2, disulfide bonds were formed between the two ␣-helices of ⑀ CC_SS , and between the ␤-sandwich structure in the NTD and the ␣-helix in the CTD of ⑀ NC_SS . The disulfide bond in ⑀ CC_SS was cleaved in the presence of 300 M DTT. In contrast, the disulfide bond in ⑀ NC_SS was cleaved when 1 mM DTT was added. We therefore used 300 M aldrithiol-2, and 300 M or 1 mM DTT to control the disulfide bond formation in the mutants for further experiments. Hereafter, we describe ⑀ CC_SS _Ox as the oxidized ⑀ CC_SS , ⑀ CC_SS_Red as the reduced ⑀ CC_SS , and ⑀ CC_SS_Non as the untreated ⑀ CC_SS . In addition, ⑀ NC_SS_Ox as the oxidized ⑀ NC_SS , ⑀ NC_SS_Red as the reduced ⑀ NC_SS , and ⑀ NC_SS_Non as the untreated ⑀ NC_SS were used.

NTD of the ⑀ subunit inhibits F 1 -ATPase activity
Inhibition of the ATP hydrolysis activity of F 1 -ATPase was examined in the presence of the WT ⑀ subunit (⑀ WT ) or its mutants (Fig. 2, A and B). The extent of the inhibition by ⑀ WT was very similar to that reported previously (22). In contrast, one of our interesting findings is that ⑀ N clearly inhibited F 1 -ATPase activity even at 300 nM, which was comparable with the findings for ⑀ WT (Fig. 2A). This result was unexpected because CTD of the ⑀ subunit was thought to be a key for the inhibition of F 1 -ATPase activity, as mentioned previously (28,29), and to date no reports of the inhibition of F 1 -ATPase activity by NTD of the ⑀ subunit have been published. In addition, we found that ⑀ CC_SS and ⑀ NC_SS can also inhibit F 1 -ATPase activity irrespective of their redox states (Fig. 2, B and C). These results also contradict the previous findings that the ⑀ subunit mutant, which is incapable of conformational change at CTD, cannot inhibit the activity of F 1 -ATPase obtained from proteobacteria (34,35). The apparent dissociation constant (K D(app) ) values between F 1 -ATPase and the ⑀ subunit or its mutants were determined based on the ⑀-dependent decrease of ATP hydrolysis activity ( Table 1). The K D(app) value for ⑀ WT was slightly lower than the previously reported value, 2.1 Ϯ 0.3 nM (36), but comparable with those in other reports (13,34). The K D(app) value for ⑀ N (21 Ϯ 6.3 nM) was about 10 times greater than those for ⑀ WT and ⑀ CC_SS_Ox (2.5 Ϯ 0.1 nM) and ⑀ NC_SS_Ox (2.3 Ϯ 2.2 nM).
Previously, we showed that the level of ⑀-inhibition of F 1 -ATPase containing the mutant ␥⌬198 -222 subunit, which lacks the insertion region from Leu 198 to Val 222 , apparently decreased to 20% (22). We therefore applied this mutant ATPase complex (F 1 -ATPase ⌬ins ) to investigate the inhibitory properties of ⑀ N . Neither ⑀ WT nor ⑀ N inhibited the ATP hydrolysis activity of the mutant ATPase complex (Fig. 2D), indicating that ⑀ N does not affect the activity of PK or LDH, which are used for our coupling assay system, and does not affect the ␣ or ␤ subunit of the complex as well.
The K D(app) values of ⑀ SS under various oxidation or reduction conditions are shown in Table 1. No significant differences between the values that we obtained for ⑀ CC_SS or ⑀ NC_SS and those from the previous study on ⑀ WT were observed (36).

The properties of the ⑀-mutants
To confirm that ⑀ N maintains the ␤-sandwich structure, we measured the CD spectrum of the protein. ⑀ N showed a negative peak at around 220 nm, which indicates the formation of the typical ␤-sheet ( Fig. 3A) (37), and is different from the possible unfolded structure, because the latter shows a positive peak at 220 nm and negative peak at 200 nm (38). The CD spectra of ⑀ CC_SS suggested that the folding of ⑀ CC_SS is identical to that of ⑀ WT irrespective of its redox state (Fig. 3B) . In addition, ⑀ CC_SS_Ox inhibited F 1 -ATPase activity, like ⑀ WT and ⑀ N (Fig. 2B). These results imply that the introduced Cys residues on the ⑀ CC_SS and the formation and dissociation of the disulfide bond in the CTD do not affect the affinity of ⑀ CC_SS to F 1 -ATPase. As shown in Table 1, K D(app) values of ⑀ N were weaker than those of ⑀ WT and the other mutants. To examine the binding of ⑀ N to F 1 -ATPase, we then tested the co-migra-The ⑀-inhibition of cyanobacterial F 1 -ATPase tion of F 1 -ATPase and ⑀ WT or ⑀ N by gel-filtration chromatography and analyzed the subunit composition in the peak by Western blotting (Fig. 4, A-D). When only F 1 -ATPase was subjected to the gel-filtration chromatography, a single peak was observed ( Fig. 4A, peak 1). The ␤ subunit was detected by Western blotting as indicated (Fig. 4D, lane 1). In contrast, two peaks were obtained when F 1 -ATPase was incubated with ⑀ WT or ⑀ N (Fig. 4B, peaks 2, 3, 5, and 6). After collecting these peak fractions, co-migrations of the ⑀ subunits with F 1 -ATPase were examined by anti-⑀ subunit antibody (Fig. 4D, lanes 2, 3, 5, and 6). These protein bands showed the protein mass at around 15 and 10 kDa, which correspond to the molecular mass of ⑀ WT and ⑀ N , respectively. As a control, ⑀ N was subjected to gel-filtration chromatography (Fig. 4C, peak 7) solely, and the collected peak was analyzed by Western blotting (Fig. 4D, peak 7). Based on these results, we concluded that ⑀ N can directly bind to the F 1 -ATPase complex.

Inhibition of F 1 -ATPase by ⑀ CC_SS at the single-molecule level
To understand the molecular mechanism behind the inhibition by NTD of the ⑀ subunit, the inhibition of rotation of the ␥ subunit by one of the CTD null ⑀-mutants ⑀ CC_SS_Ox was analyzed at the single-molecule level (Fig. 5). In this assay system, 3 M ⑀ WT or ⑀ CC_SS_Ox was used. Although we examined 3 M ⑀ N in this system, the marked change of rotation of the ␥ subunit was not observed. This must be due to the low affinity of ⑀ N to the ␥ subunit, which was lower than ⑀ WT or ⑀ CC_SS_Ox . We therefore tried to prepare 10 times higher concentrations of ⑀ N for this experiment. However, we failed to handle the higher concentration of ⑀ N in this study due to the low solubility of this  (Table 1), we used ⑀ CC_SS_Ox for this rotation analysis. The rotary motion of the ␥ subunit was observed for 5 min under an optical microscope using 340-nm duplex polystyrene beads as a probe, followed by exchange of the assay buffer in the absence or presence of ⑀ WT or ⑀ CC_SS_Ox . As expected from the activity measurement (Fig.  2B), ⑀ WT and ⑀ CC_SS_Ox entirely stopped the rotation of F 1 -ATPase (Fig. 5, E and I). When a low concentration of ATP at 250 nM was used, the step of the ␥ subunit rotation at around 120°was observed (Fig. 5, B, F, and J). This step shows the ATP-waiting position when F 1 -ATPase is waiting for the next ATP binding (10,22). When ⑀ WT was infused into the observation chamber, rotation of the ␥ subunit stopped at an angular position that differed from the ATP-waiting position (Fig. 5, E-H), namely around 80°forward from that of the ATP-waiting position ( Table 2). This observation was consistent with a previous report (22). When ⑀ CC_SS_Ox was used, a similar cessation of rotation was observed (Fig. 5I), and the angular position was also around 80° (Fig. 5, I-L).

Magnetic tweezer manipulation cannot recover the cessation of rotation by ⑀ CC_SS_Ox
To distinguish the inhibition of F 1 -ATPase by ADP and that by ⑀, restoration of rotation of the ␥ subunit was thoroughly studied using the magnetic tweezer technique (36,39). For this purpose, a magnetic bead was attached to the ␥ subunit instead of the polystyrene beads, and the ␥ subunit stopped by the inhibition was forced 80°in the counterclockwise direction using the magnetic tweezers (Fig. 6A). ADP inhibition is a common way to inhibit F 1 -ATPase caused by the tightly bound ADP at the catalytic site(s) of the enzyme, and is conserved among the ATPases from mitochondria, proteobacteria, and chloroplasts. In the case of cyanobacterial F 1 -ATPase (36) and thermophilic bacterial F 1 -ATPase (39), restoration by 80°forcing was observed at the ADP inhibition state, whereas the ⑀-inhibition was not recovered by this procedure. We therefore applied this technique to the ⑀ CC_SS_Ox -inhibited F 1 -ATPase. This time, restoration of rotation was observed in ADP-inhibited F 1 -ATPase by 80°forcing (Fig. 6, B and C), and the frequency of restoration was about 70% (Table 3). This value is fairly similar to that reported previously (86% in Ref. 36). In contrast, ⑀ WTinhibited F 1 -ATPase did not restore the rotation after 80°forcing (Fig. 6D). Restoration was also not observed in the case of ⑀ CC_SS_Ox -inhibited F 1 -ATPase (Fig. 6E). These results are summarized in Table 3 and suggest that the mechanism of inhibition of ⑀ CC_SS_Ox is identical to that of ⑀ WT .

Discussion
In this study, we aimed to clarify the mechanism of ⑀-inhibition of cyanobacterial ATP synthase in detail. First, we found that three CTD null mutants, ⑀ N , ⑀ CC_SS , and ⑀ NC_SS , inhibited F 1 -ATPase activity based on the enzymatic analysis (Fig. 2,  A-C). This was unexpected because the significance of CTD of the ⑀ subunit for the inhibition of ATP hydrolysis has already been reported (29,30). Single-molecule observation revealed that ⑀ CC_SS_Ox stopped rotation of the ␥ subunit at around 80° (  Fig. 5, J-L). The restoration of rotation of the ␥ subunit by forcing the ␥ subunit was not observed when it was inhibited by ⑀ CC_SS_Ox (Fig. 6E), although ADP-inhibited F 1 -ATPase was easily restored by 80°forcing. We therefore concluded that the ⑀ CC_SS -inhibitory mechanism was identical to that of ⑀ WT inhi- The ⑀-inhibition of cyanobacterial F 1 -ATPase bition. This result indicates that ⑀-inhibition of cyanobacterial ATP synthase can occur irrespective of the conformation of CTD of the ⑀ subunit, although CTD of the ⑀ subunit is required for the tight binding to the ␥ subunit (Table 1). To our knowledge, all previous studies on the proteobacterial ⑀ subunit indicated that CTD is indispensable for ⑀-inhibition (40 -42), and no reports have described that only NTD of the ⑀ subunit inhibited the ATP hydrolysis activity of F 1 -ATPase or F o F 1 -ATPase.
Cyanobacteria are believed to be the origin of chloroplasts, which were incorporated into ancestor cells during symbiosis. Consequently, many metabolic processes in chloroplasts are very similar to those in cyanobacteria. Therefore, the ⑀ subunit of cyanobacterial ATP synthase may also have inhibitory mechanisms similar to those of chloroplast ATP synthase. However, Nowak et al. (30) reported that the mutant ⑀ subunit of chloroplast ATP synthase from spinach, which lacks CTD, cannot inhibit F 1 -ATPase activity. In contrast, our results clearly indicate that ⑀ N inhibits the activity (Fig. 2), although the affinity to the complex was lower than that for ⑀ WT (Fig. 2 and Table 1). There is therefore a possibility that the ⑀ subunit that lacks CTD did not sufficiently associate with the ␥ subunit of chloroplast ATP synthase in their experimental conditions (30) and did not inhibit F 1 -ATPase activity very well. Recently, the whole molecular structure of chloroplast ATP synthase was determined by cryo-EM (43), and our group also determined the X-ray crystal structure of the cyanobacterial ␥-⑀ subcomplex (16). In both structures, CTD of the ⑀ subunit showed a retracted form, although CTDs of the ⑀ subunits from Escherichia coli and Geobacillus stearothermophilus (formerly known as thermophilic Bacillus PS3) were extended in the crystal structures (44,45). Recently, the structure of F 1 -ATPase from Caldalkalibacillus thermarum was reported, and the ⑀ subunit was found as a retracted form in this complex, whereas CTD of the ⑀ subunit exerted the inhibitory effect on the ATPase activity (46,47). This might be another example of inhibition by the retracted form CTD of the ⑀ subunit.
The ␥ subunits of chloroplast-type (cyanobacteria and chloroplast-type) ATP synthase equip the insertion region (Fig. 7), which also functions to inhibit ATP hydrolysis activity (16). We therefore investigated the involvement of this region with the ⑀-inhibition of cyanobacterial ATP synthase in Fig. 2C. We then prepared a ␥ subunit mutant of F 1 -ATPase lacking the insertion region of the ␥ subunit (F 1 -ATPase ⌬ins ). The inhibition of the ATP hydrolysis activity of F 1 -ATPase ⌬ins by the ⑀ subunit and its mutants (Fig. 2D) was not remarkable compared with that of the WT F 1 -ATPase. In the crystal structure of the cyanobacterial ␥-⑀ subcomplex, the ␤5 strand at the proximal end of ⑀-NTD appeared to form a mixed parallel ␤-sheet with the ␤-strand of the ␥ subunit to provide tight coupling between the ␥ and ⑀ subunits (see Fig. 1C of Ref. 16). Therefore, both the insertion region of the ␥ subunit of cyanobacterial ATP syn-   The ⑀-inhibition of cyanobacterial F 1 -ATPase thase and ⑀-NTD must be involved in the regulation of ATP hydrolysis by the tight interaction between the ␥ and ⑀ subunits. In addition, the ⑀ subunit may affect the conformation of the insertion region of the ␥ subunit, which was found to be a ␤-hairpin structure (16,43), to control the ATP hydrolysis activity in cyanobacteria and have an impact on the redox state of the ␥ subunit from chloroplast ATP synthase as well (48). Only from these in vitro analyses, we could not draw a definitive conclusion on whether CTD of the ⑀ subunit of cyanobacterial ATP synthase can change the conformation in the F o F 1 complex and exert the inhibitory effect on ATP hydrolysis. However, it should be noted that there are some organisms whose ⑀ subunit of ATP synthase lacks CTD (42,49). In the case of mitochondria, the homologous protein of the ⑀ subunit is the ␦ subunit, although the ␦ subunit is not supposed to inhibit ATP hydrolysis activity (50). In addition, the mitochondrial ␦ subunit cannot change conformation due to the covered structure  Table 2 Stop angular position of rotation of the ␥ subunit inhibited by the ⑀ WT or ⑀ CC_SS_Ox The most stopped angular positions in Fig. 5 were averaged on the indicated number of the particles. The ⑀-inhibition of cyanobacterial F 1 -ATPase of the ⑀ subunit in the complex. Instead, IF 1 acts as an ATP hydrolysis inhibitor protein in mitochondrial F 1 . IF 1 binds to the interface between the ␣ and ␤ subunits of F 1 -ATPase, whereas the orientation for insertion into the interface differs from that of ⑀-CTD of F 1 from E. coli or G. stearothermophilus (44,45,51). In these bacterial F 1 s, the decrease of membrane potential or change in ATP concentration is suggested to induce the conformational change of ⑀-CTD and regulate the ⑀-inhibition (28,52). Interestingly, both the bacterial ⑀ subunit and the mitochondrial IF 1 use the ␣-helix part to regulate F 1 -ATPase activity, whereas the binding positions in the complex are very different. In contrast, our study clearly shows that the cyanobacterial ⑀ subunit does not require conformational change in the part of CTD containing ␣-helices for its inhibitory function, although we cannot rule out the possibility that the conformational change of CTD may occur in the F o F 1 complex. Cyanobacterial ATP synthase works for both the photophosphorylation and the oxidative phosphorylation reaction, and this unique feature might be the origin of the unique regulatory system.

Materials
Biotin-PEAC 5 -maleimide was purchased from Dojindo (Kumamoto, Japan). ATP, phosphoenolpyruvate, and BSA were obtained from Sigma. Pyruvate kinase, lactate dehydrogenase, and NADH were purchased from Roche Diagnostics (Basel, Switzerland). Other chemicals were of the highest grade commercially available.

Protein preparation
In this study, the ␣ 3 ␤ 3 ␥ subcomplex from a thermophilic cyanobacterium (T. elongatus BP-1) (36) was used as a WT F 1 -ATPase. The expression and purification of the F 1 -ATPase complex were described previously (36). ⑀ N was generated by the PrimeSTAR Mutagenesis Basal kit (Takara, Shiga, Japan) using the mutation primers shown in Table 4, and ⑀ SS was generated by an infusion method (Takara) using the mutation primers shown in Table 4.
The ⑀ subunit and its mutants were expressed in E. coli and purified as described previously (22) with some modification. For purification of the ⑀ subunits, the combination of anionexchange chromatography using DEAE Sephacel (GE Healthcare) and hydrophobic interaction chromatography using a Phenyl-Toyopearl column (Tosoh, Tokyo, Japan) was used. For further purification, size-exclusion chromatography using a Superdex 75 column (GE Healthcare) equilibrated with 50 mM HEPES-KOH (pH 8.0) and 100 mM KCl (Buffer A) was performed. All proteins were stored at Ϫ80°C until use.

Oxidation or reduction of ⑀ SS
⑀ CC_SS or ⑀ NC_SS mutant was incubated with various concentrations of Aldrithiol-2 or DTT in Buffer A for 60 min at room temperature to obtain the oxidized form ⑀ CC_SS or ⑀ NC_SS (⑀ CC_SS_Ox or ⑀ NC_SS_Ox) or the reduced form ⑀ CC_SS or ⑀ NC_SS (⑀ CC_SS_Red or ⑀ NC_SS_Red ). To remove oxidants or reductants, the protein solution was loaded onto a Microcon column (10-kDa cut-off; Millipore) and centrifuged repeatedly. To confirm the oxidation and reduction state of ⑀ SS , the mutant proteins were precipitated by adding 5% (w/v, final concentration) TCA. After centrifugation, the supernatant was removed, and the remaining oxidant was washed away with 500 l of acetone. After the removal of acetone by centrifugation, the pellet was dissolved in 50 mM Tris-HCl (pH 8.0), 1% (w/v) SDS, and thiol-modifying reagent 4-acetamido-4Јmaleimidylstilbene-2,2Ј-disulfonate (AMS). After labeling for 1 h at room temperature, protein samples were subjected to nonreducing SDS-PAGE, and the redox state of ⑀ CC_SS or ⑀ NC_SS was confirmed by determining the mobility in the gel.

ATP hydrolysis activity assay
ATP hydrolysis activity was measured in the presence of an ATP-regenerating system in 50 mM HEPES-KOH (pH 8.0), 100 The ⑀-inhibition of cyanobacterial F 1 -ATPase mM KCl, 2 mM MgCl 2 , 2 mM ATP, 50 g/ml PK, 50 g/ml LDH, 2 mM phosphoenolpyruvate, and 0.2 mM NADH (53). The assay was carried out at 25°C. The ATP hydrolysis rate after the addition of F 1 -ATPase was determined by monitoring the decrease in NADH absorption at 340 nm using a spectrophotometer, V-550 (Jasco, Tokyo, Japan). The results of three independent experiments were averaged.

Estimation of ⑀ subunit binding based on ATP hydrolysis activity assay
The dissociation constant was estimated from the extent of ATP hydrolysis activity in the presence of the ⑀ subunit or its mutants. The ATP hydrolysis activity was determined from the steady-state slope of ATP hydrolysis (22). The change in the extent of inhibition depending on the ⑀-concentration was then fitted using the following equation, where y is the residual activity of ATP hydrolysis in the presence of each concentration of the ⑀ subunit, A min is the minimum residual ATP hydrolysis activity, K D(app) is the dissociation constant between F 1 -ATPase and the ⑀ subunit or its mutant, [⑀] is the final concentration of the ⑀ subunit, and [F 1 ] is the final concentration of F 1 -ATPase.

CD spectrum
The ⑀ subunit or its mutants were diluted in 20 mM Tris-HCl (pH 8.0), and their CD spectra were obtained using a spectrophotometer, J-820 (Jasco, Tokyo, Japan), at room temperature. The concentration of the ⑀ subunit or its mutants was 0.1 mg/ml.

Estimation of ⑀ subunit binding based on gel filtration chromatography
The amount of ⑀ N bound to F 1 -ATPase was estimated from the fraction of gel-filtration chromatography. 1 M F 1 -ATPase and 5 M ⑀ WT or ⑀ N were incubated at room temperature for 10 min in Buffer B (50 mM HEPES-KOH (pH 8.0), 100 mM KCl, 0.1 mM MgCl 2 , 0.1 mM ATP). Then the mixture was subjected to gel-filtration chromatography using a Superdex 200 increase column equilibrated with Buffer B, and the peaks were collected as indicated (peaks 1-7). The proteins in the peak fractions were then precipitated by 5% (w/v, final concentration) TCA. After centrifugation, the supernatant was removed, and the remaining protein pellet was washed away with 500 l of acetone. After the removal of acetone by centrifugation, the pellet was dissolved in 50 mM Tris-HCl (pH 8.0), 1% (w/v) SDS and subjected to SDS-PAGE. Proteins were transferred to a polyvinylidene difluoride membrane and detected by anti-␤ or -⑀ subunit antibodies.

Single-molecule observation
Rotation assays were carried out as reported previously (22) with some minor modifications. Streptavidin-coated beads with a diameter of 340 nm were used. Observation of rotation of the ␥ subunit was performed at room temperature. In general, solution exchange in the flow chamber took 1-2 min. The rota- Table 3 Frequency of mechanical restoration of the ADP-, ⑀ WT-, or ⑀ CC-SS_Ox -inhibited ␥ subunit Activated, the number of the particles that showed rotation at each of the stall angles. Not activated, the number of the particles that did not show rotation at each of the stall angles.
Author contributions-K. I. and T. H. conceived the study, and K. I., K. K., and K. Y. performed the experiments. K. W. and T. H. supervised the research. K. I., K. K., K. Y., K. W., and T. H. discussed the data. K. I. and T. H. wrote the paper, and K. K., K. Y., and K. W. commented on the manuscript.