Inhibition of F1-ATPase Rotational Catalysis by the Carboxyl-terminal Domain of the ϵ Subunit*

Background: The ϵ subunit inhibits F1-ATPase activity. Results: Truncation of helix 2 or a point mutation in loop 2 of the ϵ subunit decreased its inhibitory effects on subunit rotation. Conclusion: Helix 2 and loop 2 play pivotal roles in inhibitory regulation of F1 rotational catalysis. Significance: Structure-based studies on the ϵ subunit function are critical for understanding the mechanism underlying rotational catalysis of ATP synthase. Escherichia coli ATP synthase (F0F1) couples catalysis and proton transport through subunit rotation. The ϵ subunit, an endogenous inhibitor, lowers F1-ATPase activity by decreasing the rotation speed and extending the duration of the inhibited state (Sekiya, M., Hosokawa, H., Nakanishi-Matsui, M., Al-Shawi, M. K., Nakamoto, R. K., and Futai, M. (2010) Single molecule behavior of inhibited and active states of Escherichia coli ATP synthase F1 rotation. J. Biol. Chem. 285, 42058–42067). In this study, we constructed a series of ϵ subunits truncated successively from the carboxyl-terminal domain (helix 1/loop 2/helix 2) and examined their effects on rotational catalysis (ATPase activity, average rotation rate, and duration of inhibited state). As expected, the ϵ subunit lacking helix 2 caused about ½-fold reduced inhibition, and that without loop 2/helix 2 or helix 1/loop 2/helix 2 showed a further reduced effect. Substitution of ϵSer108 in loop 2 and ϵTyr114 in helix 2, which possibly interact with the β and γ subunits, respectively, decreased the inhibitory effect. These results suggest that the carboxyl-terminal domain of the ϵ subunit plays a pivotal role in the inhibition of F1 rotation through interaction with other subunits.

ATP synthase (F 0 F 1 ) plays a central role in biological energy transduction; this enzyme synthesizes most of the cellular ATP from ADP and phosphate (P i ), coupling with the electrochemical proton gradient through subunit rotation (for reviews, see Refs. [1][2][3][4][5][6][7]. It is composed of membrane extrinsic catalytic sector F 1 (␣ 3 ␤ 3 ␥␦⑀) and trans-membrane proton pathway sector F 0 (ab 2 stator and c-ring formed from multiple c subunits) (5)(6)(7). Protons are transported through a pathway formed by the Asp or Glu residue in the rotating c-ring and the Arg residue of the stator a subunit (4,5).
We have established an experimental system for observing the high-speed rotation of the Escherichia coli ␥ subunit located at the center of the ␣ 3 ␤ 3 hexamer (8 -10) or the ␥⑀/c-ring in the entire ATP synthase (11). To observe the ␥ subunit rotation, the ⑀ subunit-depleted F 1 sector was fixed on a glass surface, and a small gold bead was introduced to the ␥ subunit (8,9). In the presence of a high concentration of ATP, the bead showed continuous rotation defined as the "active state" and entered stochastically into a period of long pauses (Ն100 ms), defined as the "inhibited state" (9,10). Both states last for ϳ1 s on average (9). The rotation rate during the active state is ϳ400 revolutions/s (rps), 2 and the overall rate, including the inhibited state, is ϳ160 rps, which is comparable with that estimated from the bulk phase ATPase activity, assuming that three ATP molecules are hydrolyzed upon 360°revolution (9). A high data collection rate (4,000 -8,000 frames/s) allows us to detect the short pauses between the 120°rotation steps during the active state (9,10). We concluded that ATP is hydrolyzed during the short pauses (ϳ0.2 ms), which are thus defined as "catalytic dwells" (9). The 120°rotation steps are further divided into 40°and 80°rotation steps, and the short pause between the two substeps corresponds to the ATP-binding dwell (ATP-waiting dwell), during which ATP binds to one of the three ␤ subunits, and ADP is released from another ␤ subunit (6,7,9,12).
The addition of an excess amount of the ⑀ subunit extends the duration time of the inhibited state but has no effect on the time length of the active state (8,9). We found that the ⑀ subunit decreased the rotation rate in the active state by increasing the frequency and duration of the short pauses (8). We also found that the ⑀ subunit significantly decreased the activation energy required for subunit rotation (9).
The recently determined high-resolution crystal structure of the E. coli F 1 sector revealed that the ⑀CTD adopts a highly extended conformation, with helix 2 inserted deeply between the ␤ and ␥ subunits (Fig. 1a) (15). This structure should correspond to a transiently inhibited conformation because the ␥ rotor may be connected non-covalently to the stator (␣ 3 ␤ 3 ) by the ⑀ subunit, thereby preventing ␥ rotation. In contrast, the isolated ⑀ subunit adopts a compact conformation in which the two ␣-helices interact with each other and form a hairpin structure (Fig. 1b, right) (19 -21). An ⑀ subunit with a compact structure was also observed in the F 0 F 1 complex (22)(23)(24)(25) and may exhibit less inhibitory activity because the ⑀CTD including helix 2 cannot interact with stator subunits.
In this study, we prepared a series of truncated ⑀ subunits lacking part of the ⑀CTD and ones with amino acid replacements and examined their effects on ATPase activity, rotation speed, and duration of the inhibitory state. These studies indicated the importance of the ⑀CTD interaction with the ␤ and ␥ subunits. The concerted effect of loop 2 and helix 2 was also suggested.

EXPERIMENTAL PROCEDURES
Preparation and Materials-A recombinant plasmid carrying the F 1 F 0 gene was used as the wild type throughout this study. E. coli F 1 with six histidine residues and cysteine substitutions (␥S193C and ␥K108C) introduced into the ␣ subunit amino terminus and ␥ subunit, respectively, was constructed from pBUR17 by replacing the Csp45I-RsrII segment with that containing the ␥K108C substitution (26).
The F 1 sector was prepared from E. coli strain DK8 harboring the plasmid as described previously (26) with minor modifications. Briefly, membrane vesicles were prepared from the cells, suspended in 50 mM Tris-HCl buffer (pH 8.0) containing 0.5 mM dithiothreitol, 140 mM KCl, 1 mM EDTA, and 10% (w/v) glycerol, and then centrifuged at 160,000 ϫ g for 1 h at 15°C to remove the ␦ subunit. The following procedures were performed at room temperature. The precipitate was incubated in 2 mM Tris-HCl buffer (pH 8.0) for 10 min. After centrifugation, 0.5 M MOPS-NaOH buffer (pH 7.0) and 1 M Na 2 SO 4 were added to the supernatant (final concentrations of 20 and 50 mM, respectively), and the mixture was incubated with 100 M biotin-PEAC 5 -maleimide (Dojindo, Kumamoto, Japan) for 1.5 h. The ⑀ subunit was removed from F 1 by binding the biotinylated F 1 (ϳ5 mg of protein) to a nickel-nitrilotriacetic acid resin column (Qiagen, Hilden, Germany) (0.5-ml bed volume), followed by extensive washing with 30 ml of 10 mM Tris-H 2 SO 4 buffer (pH 8.0) containing 10% glycerol and 1 mM ATP (pH adjusted to 8.0 with HCl). Bound F 1 was eluted with 3 ml of 50 mM Tris-H 2 SO 4 buffer (pH 8.0) containing 200 mM imidazole, 50 mM Na 2 SO 4 , 10% glycerol, and 1 mM ATP. The elution speed was ϳ5 ml/h. Imidazole was removed by dialysis against 50 mM Tris-H 2 SO 4 buffer (pH 8.0) containing 50 mM Na 2 SO 4 , 25% glycerol, and 1 mM ATP. Purified enzyme (ϳ1 mg/ml protein) was quickly frozen in liquid nitrogen and stored at Ϫ80°C until use. Protein concentrations were determined by the method of Bradford using bovine serum albumin (Sigma, Fraction V) as a standard (27). The concentration of F 1 (␣ 3 ␤ 3 ␥⑀) was determined assuming its molecular weight to be 365,000. Gel elec-trophoresis analysis indicated that Ͼ80% of the ⑀ subunit was removed from F 1 , as estimated from the densities of the ⑀ band before and after the nickel column chromatography using recombinant ⑀ as a standard. The small amount of remaining ⑀ subunit was dissociated during the course of the ATPase or rotation assay (8).
Gold beads (60-nm diameter) were obtained from British BioCell International and coated with biotinylated bovine serum albumin (8). Other materials used were of the highest grade commercially available.
Plasmid Construction and Mutagenesis-To generate plasmids encoding truncated ⑀CTD mutants (Fig. 1b), DNA fragments were amplified by means of polymerase chain reaction (PCR) using a plasmid encoding the ⑀ subunit with a histidine tag and a recombinant tobacco etch virus protease site at the amino terminus (28) as a template. The sense primer used was the T7 promoter primer, and the antisense primers contained sequences encoding a truncated ⑀CTD and an EcoRI site inserted downstream of a stop codon ( Table 1). Insertion of the PCR products into the same plasmid (28) using NdeI and EcoRI yielded recombinant plasmids encoding a series of ⑀CTD-truncated ⑀ subunits. A plasmid encoding ⌬H2/L2/H1 ⑀ (Fig. 1b) containing the ⑀A39C substitution was constructed by the same method using the plasmid containing the ⑀A39C substitution (28) as a template. To construct a plasmid encoding ⑀CTD (⑀Glu 91 -⑀Met 138 ), a DNA fragment was amplified using a sense primer containing an NcoI site and an antisense primer containing the ⑀ carboxyl terminus and an EcoRI site (carboxyl terminus/EcoRI primer) ( Table 1) and then subcloned into the plasmid (28) using NcoI and EcoRI.
For alanine-scanning mutagenesis, 10 sets of sense and antisense complementary primers encoding the ⑀ subunit with a single amino acid residue to alanine substitution were synthesized (Table 1). These antisense primers and the T7 promoter primer and the sense primers and carboxyl terminus/EcoRI primer sets were used to amplify the amino-terminal ϳ500 base pairs and carboxyl-terminal ϳ100 base pairs, respectively. The resulting PCR products were annealed and then used as templates with the T7 promoter and carboxyl terminus/EcoRI primers. The obtained DNA fragments encoding ⑀ subunits with an alanine substitution were subcloned into the plasmid (28) using NdeI and EcoRI. A ⑀Ser 108 to Lys, Asn, or Asp replacement was also introduced by the same method using primers encoding the ⑀ subunit with a single amino acid substitution (Table 1). All mutant sequences were confirmed by DNA sequencing.
Expression and Purification of Mutant ⑀ Subunits-The wild type and mutant ⑀ subunits having a histidine tag and a recombinant tobacco etch virus protease site at the amino terminus were expressed in E. coli cells, BL21(DE3)pLysS, harboring the corresponding recombinant plasmid, as described previously (28); cells were cultured in minimal medium supplemented with 1.1% glucose, and the ⑀ subunits were expressed by incubation with isopropyl ␤-D-thiogalactopyranoside to the midlog phase for 18 h at 18°C. Cells were harvested and disrupted in buffer A (20 mM Tris-HCl (pH 8.0) and 150 mM NaCl) containing protease inhibitors, 1 g/ml DNase I, and 1 mM dithiothreitol, passed through a French press twice, and then centri-fuged at 12,000 ϫ g for 12 min twice. The supernatant (ϳ150 mg of protein) was applied to a nickel-nitrilotriacetic acid resin column (1-ml bed volume). The column was washed with 6 ml of buffer A containing 45 mM imidazole, and the ⑀ subunit was eluted with 3 ml of the same buffer containing 300 mM imidazole. The elution speed was ϳ10 ml/h. After the removal of imidazole by dialysis against buffer A, the histidine tag was digested with recombinant tobacco etch virus protease (Invitrogen); dithiothreitol and the protease were added (final concentration of 0.8 mM and 250 units/ml, respectively), and the mixture was incubated for 2.5 h. To remove the digested histidine tag, the resulting solution was mixed with one-tenth volume of nickelnitrilotriacetic acid resin and incubated for 30 min with gentle shaking. After removal of the resin by quick centrifugation (10,000 ϫ g for 1 min), the supernatant (50 -100 g/ml protein) was used as the purified ⑀ subunit. Purified protein was quickly frozen in liquid nitrogen and stored at Ϫ80°C until use. Cell harvest and disruption were performed at 4°C, and other procedures were at room temperature.
Assay Procedures-Bulk phase ATPase activity was estimated essentially as described previously (8); rotation buffer (10 mg/ml bovine serum albumin, 10 mM MOPS-KOH (pH 7.0), 50 mM KCl, and 2 mM MgCl 2 ) was used to compare the results with the rotation rates. The reaction was started by the addition of 2 mM ATP and terminated with trichloroacetic acid immediately after a 10-min incubation at the indicated temperature. Each mixture was centrifuged, and then the phosphate released was determined. To examine the effect of the ⑀ subunit, 100 nM ⑀ subunit in rotation buffer was incubated with 4 nM F 1 for 10 min before the addition of ATP. These assays were repeated at least three times.
The rotation assay was carried out as described previously (8): two sizes of coverglasses (18 ϫ 18 mm and 24 ϫ 32 mm) were washed extensively with 0.1 N KOH and used to construct a flow cell with spacers (ϳ30 m deep). To observe the rotation of a gold bead attached to the ␥ subunit, the flow cell was filled with rotation buffer containing 200 nM biotinylated F 1 and then incubated for 10 min. After washing off unbound F 1 with rotation buffer, streptavidin (4 M) in the same buffer was introduced into the cell, followed by incubation for 5 min and then extensive washing with the buffer. The biotinylated gold beads (0.2%) were introduced and washed similarly. To examine the effect of the ⑀ subunit, 100 nM ⑀ subunit in rotation buffer was introduced into the flow cell, followed by incubation for 10 min. Rotation buffer containing 2 mM ATP, 1 mM phosphoenolpyruvate, 50 g/ml pyruvate kinase, and 100 nM ⑀ subunit was introduced to start the rotation.
To observe the rotation of a gold bead attached to the ⑀ subunit, 25 nM unbiotinylated F 1 was incubated with 100 nM biotinylated ⑀ for 10 min and then introduced to the flow cell. After streptavidin and the biotinylated gold beads were sequentially applied to the flow cell as described above, the rotation buffer containing 2 mM ATP and its regenerating system was introduced to start rotation. The rotation assay was performed at 24°C.
Images of beads were observed after the addition of 2 mM ATP at 24°C by dark field microscopy (BX51WI-CDEVA-F, Olympus, Tokyo), and recorded with an intensified charge- coupled device camera at speeds of 4,000 -8,000 frames/s. The results were analyzed with modified MATLAB (The Mathworks, Inc.) and ImageJ (National Institutes of Health). Rotations of more than 10 beads were followed, and the rotation rates were estimated as follows (10). We determined the single revolution time (time required for 360°revolution), which is suitable for evaluating stochastic fluctuation of rotation. The geometric mean of the single revolution time was calculated, and the rotation rate was estimated as a reciprocal of the geometric mean. The deviation of the rotation rates of individual beads was ϳ10%.

RESULTS
Effects of the ⑀CTD on F 1 -ATPase Activity-The ⑀ subunit inhibits F 1 -ATPase catalysis by decreasing the rotation rate and extending the inhibited state (8,9). Considering the two conformations of the ⑀ subunit (Fig. 1b), we can assume that the ⑀CTD of the extended ⑀ subunit interacts with other subunits, as shown in the recent E. coli F 1 structure (Fig. 1a). This structure should lead to inhibition of rotational catalysis.
The bulk phase ATPase activity of ⑀-depleted F 1 was 20.0 Ϯ 1.2 mol/mg/min, i.e. slightly higher than the previous result (8) because of extensive washing of the ⑀ subunit. The inhibitory effects of the wild type and mutant ⑀ subunits were expressed as relative values, taking the control without the ⑀ subunit as 100%. As shown in Fig. 2b, the bulk phase activity decreased with increasing ⑀ concentration and reached a plateau level with more than 30 nM, indicating that the ⑀ subunit concentration used (100 nM) is enough to examine the inhibitory effects.
About 80 Ϯ 0.2% of the bulk phase activity was inhibited by 100 nM wild type ⑀ subunit (Fig. 2c), consistent with the previous results (18,29). The inhibition decreased to 36 Ϯ 1.3% with the ⌬H2 mutant, indicating that helix 2 plays a pivotal role in the inhibitory effect. On the other hand, the inhibition decreased to 20 Ϯ 5.0% with the further truncated ⌬H2/L2/H1 mutant. Thus, helix 1 contributes slightly to the inhibition. The ⌬H2/L2 mutant caused almost the same inhibition as ⌬H2 (Fig.  2c), suggesting that the loop 2 region does not have a significant effect. It is noteworthy that the shortest mutants, ⌬H2/L2/H1 and ⌬H2/L2/H1(A39C), retained weak inhibitory effects. The ⑀CTD protein (between ⑀Glu 91 and ⑀Met 138 ) without the amino-terminal ␤ sandwich domain caused no inhibition (Fig. 2c), suggesting that the ␤ sandwich domain is required for the binding of the ⑀ subunit to F 1 , which is consistent with the previous result (30).
To determine the relative affinities of F 1 for the wild type and mutant ⑀ subunits, competition assays were carried out: we added various amounts of the mutant ⑀ subunit to the ATPase reaction mixture together with the wild type ⑀ subunit. As shown in Fig. 2d (1), 100 nM wild type and ⌬H2 mutant ⑀ subunit inhibited ϳ80 and 16% of the ATPase activity, respectively ( Fig. 2d (1), closed and open bars, respectively). When increasing amounts of the ⌬H2 mutant were added together with 100 nM wild type ⑀ subunit, the inhibitory effect of the wild type decreased (Fig. 2d (1), gray bars). The addition of 100 nM ⌬H2 mutant decreased the inhibition to 44%, which corresponded to the value when half of the F 1 molecules were inhibited by the wild type and the rest by the ⌬H2 mutant ((80 ϩ 16%)/2 ϭ 48%) (Fig. 2d (1), horizontal dotted line). These results suggest that the amounts of F 1 molecules with ⌬H2 and ones with the wild type ⑀ subunit were essentially the same. Similar experiments suggested that the numbers of F 1 molecules with the ⌬H2/L2 mutant and ones with the wild type were the same when 30 -100 nM ⌬H2/L2 and 100 nM wild type ⑀ were included in the assay mixture ( Fig. 2d (2)). These results suggest that the two mutant ⑀ subunits bound to F 1 similarly to the wild type.
On the other hand, 0.3-1 M ⌬H2/L2/H1 mutant was required to obtain similar results when 0.1 M wild type was present in the assay mixture (Fig. 2d (3)), indicating that the mutant's affinity to F 1 is 3-10-fold lower than that of the wild type. Because the dissociation constant (K d ) of the wild type ⑀ is ϳ1 nM (31,32), K d for this mutant should be 3-10 nM. We performed ATPase and rotation assays in the presence of 100 nM ⑀ subunit, which was ϳ10-fold higher than the K d for the mutant. These results suggest that at least 90% of the F 1 bound to the mutant ⑀ subunits under experimental conditions and that the reduced inhibition by the mutants is mainly due to the truncation of the ⑀CTD.
Next we examined the effect of temperature on ATPase activity (Fig. 2e) and calculated the activation energy of F 1 with and without the ⑀ subunit. The wild type ⑀ subunit lowered the activation energy, confirming the previous results (9, 33); the activation energy values calculated for ATP hydrolysis by F 1 with ⑀ and without ⑀ were 4.6 and 35.8 kJ/mol, respectively (Fig.  2e, triangles and circles, respectively). Truncation of helix 2 affected the temperature dependence (Fig. 2e, squares); the activation energy value was 48.8 kJ/mol, i.e. significantly higher than that with the wild type. Other truncated mutants showed high activation energy values similar to ⌬H2, i.e. 49.1 and 39.1 kJ/mol for ⌬H2/L2 and ⌬H2/L2/H1, respectively. These results clearly indicated that helix 2 is essential for lowering of the activation energy for ATP hydrolysis.
Effects of the ⑀CTD on F 1 ␥ Subunit Rotation-Because the wild type ⑀ subunit causes a decreased rotation rate of the ␥ subunit in the active state and an extended duration time of the inhibited state, we determined the effects of the ⑀CTD on the rate and the duration of the inhibited state. We observed the rotation of the F 1 ␥ subunit for 2 s with or without a superstoichiometric concentration (100 nM) of the ⑀ subunit or one of its truncated derivatives (Fig. 3a). Rotation rates were calculated as reciprocal values of the geometric means of the single revolution time (time required for 360°revolution). The rotation rate (without ⑀, ϳ530 rps) decreased to about one-half (ϳ310 rps) upon the addition of the wild type ⑀ subunit (Fig.   FIGURE 2. Effects of ⑀CTD-truncated mutants on F 1 -ATPase activity. a, purification of the wild type (WT) and truncated mutant ⑀ subunits. The purified proteins (0.4 g) were subjected to polyacrylamide gel electrophoresis in the presence of SDS and then stained with Coomassie Blue. The arrows indicate the positions of 10-and 15-kDa molecular markers. b, effects of the wild type and CTD-truncated ⑀ subunits on F 1 -ATPase activity. F 1 was incubated with the indicated amounts of the wild type (circles), ⌬H2 (squares), ⌬H2/L2 (triangles), or ⌬H2/L2/H1 (diamonds) mutant ⑀ for 10 min, and then ATPase activity was assayed at 24°C. Each relative activity is shown with the S.D. value (error bar), taking the control without ⑀ (21.8 Ϯ 1.0 mol/mg/min) as 100%. c, inhibitory effect by the wild type and mutant ⑀ subunits on F 1 -ATPase activity. After incubation with 100 nM wild type or a mutant ⑀ subunit for 10 min, the ATPase activity was assayed at 24°C, and inhibition is shown as the relative value, taking the control without ⑀ (21.8 Ϯ 1.0 mol/mg/min) as 100%. ⌬H2/L2/H1(A39C) and ⑀CTD are the ⌬H2/L2/H1 mutant with the ⑀Ala 39 to Cys substitution and the ⑀CTD protein (between ⑀Glu 91 and ⑀Met 138 ) without the ⑀NTD, respectively. d, competition of ⑀CTD-truncated mutants with the wild type ⑀. Varying concentrations of a mutant ⑀ subunit were incubated with F 1 in the presence of 100 nM wild type ⑀ for 10 min, and then ATPase activity was assayed. The relative inhibition is shown with the S.D. value, taking the control without ⑀ (22.4 Ϯ 2.1 mol/mg/min) as 100%. Closed bars, with 100 nM wild type ⑀; open bars, with 100 nM each mutant ⑀; gray bars, with increasing amounts of each mutant ⑀ together with 100 nM wild type. Horizontal dotted lines indicate the estimated values when half of the F 1 molecules were inhibited by the wild type and the rest by the mutant ⑀. e, effect of temperature on the ATPase activity with or without the wild type or ⌬H2 ⑀ subunit. Activities are shown as relative values, taking that without ⑀ at 24°C (19.2 Ϯ 0.8 mol/mg/min) as 100%. Circles, without ⑀; squares, ⌬H2 ⑀; triangles, wild type ⑀. 3b), consistent with the previous results (8,9). As expected, the ⌬H2 and ⌬H2/L2 mutants were less inhibitory, giving rotation rates of 370 and 420 rps, respectively (Fig. 3b). The ⌬H2/L2/H1 mutant gave a similar speed to ⌬H2/L2. These results suggest that helix 2 and loop 2 are involved in the decrease in the rotation rate, although helix 1 plays only a minor role.
For further confirmation, we connected gold particles to the mutant ⑀ subunit and examined whether or not they rotated at a rate similar to that of those attached to the ␥ subunit; the ⑀Ala 39 to Cys substitution was introduced into the ⌬H2/L2/H1 mutant, which was then biotinylated and attached to a gold bead. The ⌬H2/L2/H1(A39C) mutant showed the same effect on ATPase activity as the ⌬H2/L2/H1 mutant (Fig. 2c). The rotation rate of beads attached to the ⌬H2/L2/H1(A39C) mutant was 430 rps, i.e. essentially the same rate as for the ones attached to the ␥ subunit (Fig. 3b). These results indicate that even the shortest mutant can bind to the F 1 sector and rotate together with the ␥ subunit.
Effects of ⑀CTD on the Durations of the Inhibited and Active States-As shown in Fig. 3a, F 1 stochastically enters into an inhibited state (Ͼ100 ms) during continuous rotation (active state) (9, 10). We followed the rotation for 32 or 16 s with or without a mutant ⑀ subunit, respectively, and compared the durations of the inhibited and active states. The observation time was long enough, because ϳ80% of the inhibited states (from the start to the end of a pause) could be defined. Without the ⑀ subunit, the durations of the active and inhibited states were about 1 s (Fig. 3c, None): pink and blue bars indicate the inhibited and active states, respectively. Upon the addition of the wild type ⑀ subunit, the duration time of the inhibited states was extended ϳ3-fold, whereas that of the active state did not change (Fig. 3c, WT), confirming the previous observation (9). However, the duration of the inhibited state became shorter with the ⌬H2/L2 or ⌬H2/L2/H1 mutant than that with the wild type (Fig. 3c, ϩ ⑀). On the other hand, the truncated and wild type ⑀ subunits showed no effect on the duration of the active state (Fig. 3c, blue bars). In summary, the ⑀ subunit extends the duration time of the inhibited state mainly through its ⑀CTD.
The total revolutions of each bead that alters the active and inhibited states stochastically during the observation time, are not the same. Thus, the bulk phase ATPase activity should not correspond to the total revolutions of each bead attached to F 1 . However, the rates on single molecule observation should become similar to the bulk phase activity if the total revolutions of significant numbers of beads are averaged. As expected, the bulk phase ATPase activities (Fig. 3d, closed bars) essentially corresponded to the average total revolutions of 10 rotating beads (Fig. 3d, open bars), and both were affected similarly by the mutant and wild type ⑀ subunits (Fig. 3d, ϩ ⑀).
Effects of Substitutions of ⑀Ser 108 -As shown previously (34), a proper rotor-stator interaction through ␤Glu 381 of ␤ 1 corresponding to bovine ␤ DP (ADP-bound ␤ subunit) (35) and ␥Met 23 of the ␥ subunit is essential for efficient rotation; the rotation rate of the ␥Met 23 to Lys mutant F 1 became ϳ2-fold slower and recovered to the wild type level upon introduction of the ␤Glu 381 to Asp substitution, indicating that a hydrogen FIGURE 3. Effects of the ⑀CTD-truncated mutants on F 1 subunit rotation. a, time courses of ␥ subunit rotation in the F 1 sector with the wild type or a ⑀CTD-truncated mutant. A gold bead (60-nm diameter) was attached to the ␥ subunit, and its rotation was followed for 2 s in the presence of 2 mM ATP, with or without the ⑀ subunit. ⌬H2L2H1(A39C) indicates the ⌬H2L2H1 mutant with substitution of ⑀Ala 39 to Cys, to which residue a gold bead was introduced. b, rotation rates of the ␥ subunit in F 1 with or without the wild type or an ⑀CTD-truncated ⑀. We observed the rotation of more than 10 beads for 2 s, and the average rotation rates were estimated as reciprocal values of the single revolution time. c, durations of the inhibited and active states with or without the wild type or a ⑀CTD-truncated ⑀ subunit. A gold bead was introduced to the ␥ subunit, and rotation was followed for 32 or 16 s. The duration times of the inhibited (pink) and active (blue) states are shown with S.E. (error bars) values. None, without ⑀; WT, wild type ⑀. Because the time courses of F 1 with the ⌬H2 mutant did not show clear inhibited states for an unknown reason, the duration of the states could not be estimated. d, comparison of the ATPase activity in the bulk phase and the average total revolutions. We estimated the average total revolutions of 10 rotating beads and expressed them as relative values, taking the control without the ⑀ subunit (296 rps) as 100% (open bars). The relative ATPase activity in the bulk phase (closed bars) was cited from Fig. 2b, 100 nM mutant ⑀. None, without ⑀; WT, wild type ⑀. OCTOBER 31, 2014 • VOLUME 289 • NUMBER 44 bond formed between ␤Glu 381 and ␥Lys 23 of the ␥M23K mutant possibly affected the proper rotor-stator interaction. The recently determined E. coli F 1 structure revealed that ⑀CTD in an extended conformation was inserted between ␤Glu 381 of ␤ 1 (bovine ␤ DP ) and ␥Met 23 (Fig. 4a) (15).

Inhibition of F 1 -ATPase by ⑀ Subunit Carboxyl Terminus
The determined structure suggests that ␤Glu 381 interacts with ⑀Ser 108 in loop 2. The distance between the oxygen atoms of the ⑀Ser 108 hydroxyl group and the ␤Glu 381 carboxyl group is 2.56 Å, which is close enough for the formation of a hydrogen bond. The predicted hydrogen bond could interfere with highspeed rotation of the rotor, leading to ATPase inhibition, because the ⑀ subunit rotates with the ␥ subunit against the ␣ 3 ␤ 3 hexamer.
To address the role of ⑀Ser 108 in F 1 -ATPase inhibition, we replaced it with Ala, Lys, Asp, or Asn and purified the mutant subunits (Fig. 4b). The inhibitory effect of the ⑀ subunit with the Lys or Asn substitution on the F 1 -ATPase activity was similar to that of the wild type (Fig. 4c), suggesting that residues having an amino group in their side chains can replace ⑀Ser 108 functionally. Similar to the case of the wild type ⑀ subunit, the Lys or Asn residue may form a hydrogen bond with ␤Glu 381 , possibly through the amino or amide group in its side chain, and thereby maintain the inhibitory effect.
On the other hand, the Ala and Asp substitutions decreased the inhibitory effect to 61 Ϯ 1.2 and 47 Ϯ 0.3%, respectively (Fig.  4c). These results suggest that the replaced Ala or Asp residue interacts with ␤Glu 381 only weakly, and thus the effect decreases. The carboxyl group of substituted ⑀S108D may induce repulsion against ␤Glu 381 and decrease the interference with smooth rotation. Therefore, we substituted ␤Glu 381 with Asp to reduce the repulsion between ⑀S108D and ␤Glu 381 . As expected, the inhibitory effect of ⑀S108D recovered from 47 Ϯ 0.3 to 56 Ϯ 1.3% with this mutation (Fig. 4c). The ␤E381D substitution showed no effect on the F 1 -ATPase activity without the ⑀ subunit (34).
Furthermore, we observed ␥ subunit rotation in the presence of the ⑀S108D mutant. Compared with the rotation rate with the wild type ⑀ subunit (ϳ310 rps), that with the ⑀S108D mutant increased to ϳ430 rps, i.e. it was almost the same as that with the ⌬H2L2H1 mutant. In addition, the duration of the inhibited states with the ⑀S108D subunit was 1.3 s, i.e. more than ϳ2-fold shorter than that with the wild type (Fig. 4d). In summary, the interaction between ⑀Ser 108 and ␤Glu 381 is important to the inhibitory effect of the ⑀ subunit on F 1 -ATPase activity through a decreased rotation rate and an extended duration of the inhibited state.
Identification of a Key Amino Acid Residue between ⑀Ile 105 and ⑀Tyr 114 -The region between ⑀Ile 105 and ⑀Tyr 114 contains loop 2 and the first three amino acids of helix 2 (Fig. 1b). As shown in the E. coli x-ray structure (15), this region interacting with multiple subunits (␣, ␤, and ␥) (Fig. 4a) may contain amino acid residues involved in F 1 inhibition. We substituted amino acid residues in the region one by one with alanine, purified the mutants (Fig. 5a), and then examined their effects on F 1 -ATPase activity. Among the 10 substitution mutants, the ⑀Y114A mutant decreased the inhibition to 41 Ϯ 3.1% (Fig. 5b), suggesting that this residue in helix 2 is involved in the ⑀ subunit function. Because ⑀Tyr 114 is close to ␥Gly 85 (2.91 Å) (Fig. 4a) (15), a hydrogen bond could be formed between the hydroxyl group of the ⑀Tyr 114 side chain and the amide group of the ␥Gly 85 main chain. Interaction between ⑀Tyr 114 and ␥Gly 85 is possibly important for holding the ⑀ subunit in a certain conformation for F 1 inhibition.
To determine the relative affinities of F 1 for the wild type and mutant ⑀ subunit, competition assays were performed. The wild type and ⑀S108D mutant ⑀ (both 100 nM) inhibited ATPase activity ϳ80 and ϳ30%, respectively (Fig. 5c (1), closed and open bars, respectively). With an increase in the mutant ⑀ subunit relative to 100 nM wild type, the inhibition decreased ( Fig. 5c  (1), gray bars), and the addition of 300 nM ⑀S108D mutant decreased the inhibition to 52%, which corresponded to the value when equal populations of F 1 were inhibited by the wild type and ⑀S108D mutant subunits. These results indicate that the affinity of the mutant is ϳ3-fold lower than that of the wild type. 0.3-1 M ⑀Y114A mutant was required to attain equal inhibition (Fig. 5c (2)), indicating that the mutant has 3-10-fold lower affinity. We estimated that the K d values of these mutants were less than 3 and 10 nM, respectively. Therefore, at least 97 and 90% of the F 1 molecules should bind to ⑀S108D and ⑀Y114A under our assay conditions, respectively. The wild type F 1 (␤381E(WT)) or F 1 with amino acid substitution of ␤Glu 381 to Asp (␤381D) was used. Inhibition is shown as the relative value, taking the control without ⑀ (19.6 Ϯ 0.9 mol/mg/min) as 100%. The open bar indicates inhibition by the wild type ⑀ subunit. d, durations of the inhibited and active states with or without the wild type (WT) or ⑀S108D mutant. F 1 rotation was followed for 32 or 16 s, and the duration times of the inhibited (pink) and active (blue) states are shown with S.E. values (error bars). None, without ⑀; WT, wild type; S108D, the mutant ⑀ subunit with substitution of ⑀S108D. The durations for without ⑀ and the wild type are cited from Fig. 3c.
As described above, the ⑀ subunit decreases the activation energy of F 1 ATP hydrolysis and participates in the effective energy coupling (9). Interactions between the ⑀ and other subunits through ⑀Ser 108 and ⑀Tyr 114 should be responsible for lowering of the activation energy. We measured ATPase activities with the ⑀S108D or ⑀Y114A mutant at various temperatures (Fig. 5d) and calculated the activation energies. The estimated values were 45.9, 18.7, 38.4, and 29.5 kJ/mol for without ⑀, the wild type, ⑀Y114A, and ⑀S108D, respectively, indicating that both the ⑀Ser 108 and ⑀Tyr 114 residues contribute to lowering of the activation energy, at least partly. Interactions between the ⑀ and other subunits through these residues may be required for efficient energy coupling.

DISCUSSION
In this study, we have shown that the carboxyl-terminal domain of the ⑀ subunit, ⑀CTD, especially loop 2/helix 2, plays pivotal roles in inhibition of F 1 -ATPase activity. The inhibition occurs through a decreasing rotation rate and an extending duration of the inhibited state, as shown in this and previous studies (8). Moreover, we have identified important amino acid residues, ⑀Ser 108 in loop 2 and ⑀Tyr 114 in helix 2, for the ⑀ subunit inhibitory function. Because the ⑀ subunit rotates with the ␥ subunit in the ␣ 3 ␤ 3 hexamer, interactions of the ⑀ subunit with other subunits, possibly with the ␣ and ␤ ones, are essen-tial to inhibit ATPase activity. Through these interactions, the ⑀ subunit maintains an extended conformation, and thereby interferes with smooth rotation of the ␥ subunit.
Our recent findings also support the importance of interactions between ⑀CTD and other subunits: the ⑀ subunit fused with a globular protein, cytochrome b 562 , at its carboxyl terminus does not inhibit the subunit rotation of F 1 (8). This finding suggests that the globular protein prevents ⑀CTD from interacting with the ␣ 3 ␤ 3 hexamer, which results in a decreased inhibition.
The ⑀Ser 108 and ⑀Tyr 114 residues involved in the ⑀ subunit inhibitory function are in the region between ⑀Ser 106 and Tyr 114 (Fig. 1b). However, truncation of this region did not have a significant effect on F 1 -ATPase activity; the ⌬H2L2 mutant showed almost the same reduced inhibitory effect as ⌬H2 (Fig.  2c). These results indicate that the region is involved in the inhibitory function in a cooperative manner with helix 2. Helix 2 may be essential for holding ⑀Ser 108 at a position close to ␤Glu 381 to form a hydrogen bond that can inhibit high-speed subunit rotation.
We have studied the detailed mechanism of F 1 rotational catalysis by analyzing mutant enzymes, such as ␥M23K and ␤S174F, and inhibitors (ATP␥S and piceatannol) (9,10,13,14,34,36), which extended the duration of the catalytic dwell but FIGURE 5. Effects of alanine substitutions in the region between ⑀Ile 105 and Tyr 114 on F 1 -ATPase activity. a, purification of the wild type (WT) and alanine-substituted mutant ⑀ subunits. The amino acid residues in the region between ⑀Ile 105 and Tyr 114 were substituted with alanine one by one. The purified proteins (2 g) were subjected to SDS-PAGE and then stained with Coomassie Blue. The numbers and arrows indicate the positions of alanine substitutions and molecular markers, respectively. b, effects of the ⑀ subunit with alanine substitutions on ATPase activity. Inhibitory effects of the mutants on F 1 -ATPase activity are shown with S.E. values (error bars), taking the control without ⑀ (20.2 Ϯ 0.9 mol/mg/min) as 100%. The amino acid sequence of the wild type ⑀ subunit is shown as one-letter symbols with the corresponding motifs. The open bar indicates inhibition by the wild type ⑀ subunit. c, competition of mutant ⑀ with the wild type. Varying concentrations of a mutant ⑀ subunit were incubated with F 1 in the presence of 100 nM wild type ⑀ for 10 min, and then ATPase activity was assayed. The relative inhibition is shown with S.D. values (error bars), taking the control without ⑀ (20.3 Ϯ 0.1 mol/mg/min) as 100%. Closed bars, with 100 nM wild type ⑀; open bars, with 100 nM each mutant ⑀; gray bars, with increasing amounts of mutant ⑀ together with 100 nM wild type. Horizontal dotted lines indicate the value when half of the F 1 molecules were inhibited by the wild type and the rest by the mutant ⑀. d, effect of temperature on ATPase activity with or without the wild type or a mutant ⑀ subunit. Activities are shown as relative values, taking that without the ⑀ subunit at 24°C (20.9 Ϯ 0.1 mol/mg/min) as 100%. Circles, without ⑀; diamonds, ⑀Y114A; squares, ⑀S108D; triangles, wild type ⑀. OCTOBER 31, 2014 • VOLUME 289 • NUMBER 44 not the duration of the 120°rotation. The catalytic dwell was observed more clearly with these mutations or in the presence of these inhibitors. As mentioned above, we have previously shown that the ⑀ subunit increases the frequency and duration of short pauses (ϳ ms) during continuous rotation, which results in a lowered rotation rate (8). These short pauses should be the catalytic dwell and/or ATP-binding dwell. Unlike other inhibitors and mutations, the catalytic dwell did not become clearer in the presence of the ⑀ subunit (8). Moreover, the recently determined E. coli F 1 structure suggests that the ⑀ subunit in its extended conformation traps the ␥ subunit in a rotary position close to the dwell before the ATP binding event (15). Therefore, at least a certain population of the increased short pauses may be ATP-binding dwells.

Inhibition of F 1 -ATPase by ⑀ Subunit Carboxyl Terminus
The temperature effect on ATPase activity with the wild type or mutant ⑀ subunit revealed that helix 2 is essential for lowering of the activation energy. As we discussed previously (9), lowering of the activation energy should be important for avoiding loss of energy due to ATP hydrolysis and for maintaining the efficient coupling between catalysis and rotation.
In conclusion, our results suggest that the ⑀CTD plays a pivotal role in inhibitory regulation of subunit rotation in F 1 and, at the same time, contributes to efficient coupling by lowering the activation energy. Our structure-based studies on the ⑀ subunit function will contribute to understanding of the mechanism underlying rotational catalysis.