F 0 F 1 -ATPase/Synthase Is Geared to the Synthesis Mode by Conformational Rearrangement of (cid:1) Subunit in Response to Proton Motive Force and ADP/ATP Balance*

The (cid:1) subunit in F 0 F 1 -ATPase/synthase undergoes drastic conformational rearrangement, which involves the transition of two C-terminal helices between a hairpin “down”-state and an extended “up”-state, and the enzyme with the up-fixed (cid:1) cannot catalyze ATP hydrolysis but can catalyze ATP synthesis (Tsunoda, S. P., Rodgers, A. J. W., Aggeler, R., Wilce, M. C. J., Yoshida, M., and Capaldi, R. A. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, 6560–6564). Here, using cross-linking between introduced cysteine residues as a probe, we have investi-gated the causes of the transition. Our findings are as follows. (i) In the up-state, the two helices of (cid:1) are fully extended to insert the C terminus into a deeper position in the central cavity of F 1 than was thought previously. (ii) Without a nucleotide,

F 0 F 1 -ATPase/synthase (F 0 F 1 ) 1 catalyzes ATP synthesis/hydrolysis coupled with a transmembrane H ϩ (proton)-translocation in bacteria, chloroplasts, and mitochondria (1)(2)(3)(4)(5). The enzyme is composed of two portions, i.e. a water-soluble F 1 , which has catalytic sites for ATP synthesis/hydrolysis, and a membrane-integrated F 0 , which mediates proton 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 reversibly detached from F 0 and is by itself a rotary motor driven by ATP hydrolysis (6 -8) in which a central stalk made of ␥ and ⑀ subunits rotates relative to the surrounding ␣ 3 ␤ 3 hexamer ring where hydrolysis occurs (9,10). The remaining F 0 portion in the membrane acts as a proton channel that mediates passive proton translocation across the membrane (11).
The ⑀ subunit is known as an endogenous inhibitor of ATPase activity of F 1 and F 0 F 1 (12,13). Structures of the isolated ⑀ from Escherichia coli, as determined by x-ray crystallography (14) and NMR spectroscopy (15,16), show that ⑀ consists of two distinct domains. A C-terminal helical hairpin domain of ϳ50 residues lies on an N-terminal 10-stranded ␤ sandwich domain of ϳ80 residues (Fig. 1A). The ␦ subunit (equivalent to the bacterial ⑀ subunit) in the crystal structure of bovine mitochondrial F 1 also has a two-domain conformation that is very similar to that of the isolated bacterial ⑀ and is associated with the "bottom" globular part of the ␥ subunit (we refer to this conformational state of ⑀ as the "down"-state hereafter) (Fig. 1B) (17). However, the down-state ⑀ does not exhibit an inhibitory effect on ATPase activity because, when downstate conformation is locked by cross-linking between the two domains, the inhibitory effect of ⑀ is lost, and apparent activation of ATP hydrolysis is observed (18,19). Actually, in the structure of mitochondrial F 1 , the ␦ subunit does not have any contact with the ␣ 3 ␤ 3 (Fig. 1B). Another conformation of ⑀ was suggested from observations that the residue (⑀Ser 108 , E. coli numbering) in the C-terminal domain of the E. coli ⑀ subunit has interactions with the residues (␤Glu 381 ) in the "DELSEED" region of the ␤ subunit (and the homologous region of the ␣ subunit) (9, 20 -22). Also, it was shown that positive residues in the C-terminal domain of the ⑀ subunit of thermophilic F 1 from thermophilic Bacillus PS3 would make electrostatic interaction with the DELSEED region of the ␤ subunit (23). The dynamic and flexible nature of the ⑀ subunit has been also reported for chloroplast F 0 F 1 (24). In accordance with these biochemical results, a new conformation of ⑀ was found in the crystal structure of the complex of truncated-␥ (␥Ј) and ⑀ of E. coli F 1 (Fig.  1C) (25). In this ␥Ј⑀ complex, a helical hairpin in the previous structures of ⑀ is opened, and the helices are lifted up. Such a location of the ⑀ subunit could be an obstacle for the rotation of the ␥ subunit and, indeed, F 0 F 1 , with the ⑀ locked to this lifted-up conformation by ␥-⑀ cross-linking, did not show ATP hydrolysis activity. Interestingly, however, the activity of ATP synthesis of this cross-linked enzyme was fully retained (26). Thus, it has been established that ⑀ can adopt at least two conformational states, i.e. the down-state in which C-terminal helices form a hairpin and the up-state in which the helices are extended. Only the ⑀ subunit in the up-state can exert an inhibitory effect on ATPase activity.
Although the importance of the conformational transition of ⑀ has been thus recognized, the following critical questions on this transition remain unanswered. (i) What is the actual upstate conformation of ⑀ in native F 0 F 1 ? The present knowledge on the up-state conformation of ⑀ is largely based on the crystal structure of the ␥Ј⑀ complex. However, it is obvious that truncated ␥Ј imposes an artificial constraint on the conformation of ⑀ (as well as ␥) in the ␥Ј⑀ structure. Indeed, if the extreme C-terminal helix of ⑀ were to have the same conformation as in the ␥Ј⑀, it would clash sterically with the closest ␤ subunit. In addition, in the model reconstituted from the ␣ 3 ␤ 3 ␥ part of the mitochondrial F 1 structure and the ␥Ј⑀ structure, ⑀Ser 108 in the ␥Ј⑀ is apparently too far from ␤Glu 381 to account for efficient cross-linking (Fig. 1C). In a 4.4-Å resolution electron density map of E. coli F 1 , the first ␣ helix of the ⑀ subunit in the extended conformation was barely seen as continuous density, but the second ␣ helix was unable to be traced (27). Therefore, the conformation and arrangement of the up-state ⑀ in intact F 0 F 1 is yet unclear. (ii) What is the effect of ATP and ADP on the conformational transition of ⑀ in F 0 F 1 ? In E. coli F 1 , depending on whether the added nucleotide is ATP or ADP, the same residue of the ⑀E108C changes the cross-linking partner subunit; ⑀-␣ in Mg 2ϩ ϩ ATP state (in the presence of MgCl 2 ϩ 5Ј-adenylyl-␤,␥-imidodiphosphate) and ⑀-␤ in the Mg 2ϩ ϩ ADP state (20,28). However, the individual roles of ATP and ADP were not obvious for F 1 from thermophilic Bacillus PS3 in our previous paper (19). The distinct role of ATP and ADP in the conformational transition of the ⑀ must be clarified. (iii) Do the enzyme with the up-state ⑀ and the enzyme with the downstate ⑀ have different affinities to ATP and ADP? If ATP and ADP have different effects on the conformational transition of ⑀, binding affinity to ATP and ADP, conversely, might be different between the enzyme with the up-state ⑀ and the enzyme with the down-state ⑀. (iv) Does the proton motive force affect the transition of ⑀? Because the enzyme with the up-state ⑀ can apparently catalyze ATP synthesis but not ATP hydrolysis, the enzyme with the up-state ⑀ can be regarded as the enzyme species geared to the ATP synthesis mode. If so, it is natural to expect that proton motive force would facilitate the down-to-up transition of the ⑀ subunit. To address these questions, we generated a new set of F 0 F 1 mutants from thermophilic Bacillus PS3 that enabled us to detect and fix the down-and upstates of ⑀ in the working enzyme.
Assays of Membrane Vesicles-Inverted membrane vesicles from E. coli cells expressing thermophilic F 0 F 1 were prepared by the procedures described previously (29) except for a modification in which 5 mM DTT was supplemented to the cell extract just after disruption of the cells. The thermophilic F 0 F 1 used in this work has a histidine tag of 10 residues at the N terminus of the ␤ subunit. Prior to use, the membrane vesicles were washed twice with PA3 buffer to remove DTT. ATPase activities of the membrane vesicles containing the F 0 F 1 mutants were inactivated by N,NЈ-dicyclohexylcarbodiimide down to Ͻ 20% of the initial activities, which were almost the same as in the case of the wild-type (15-20%). ATP-driven proton pump activity of membrane vesicles was assayed with fluorescence quenching of 9-amino-6-chloro-2-methoxyacridine at 40°C in PA4 buffer (10 mM HEPES/KOH, pH 7.5, 100 mM KCl, and 5 mM MgCl 2 ) as described previously (29). The reaction was started by the addition of 1 mM ATP and terminated by the addition of 1 M carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). ATP synthesis activity was measured at 50°C in PA4-buffer containing 1 mM ADP, 25 mM KPi, pH 7.5, and membrane vesicles (4.5 g of protein/ml). Oxidized and reduced membrane vesicles were prepared by treating with 20 M CuCl 2 for 30 min and 10 mM DTT for 30 min, respectively. EDTA (final concentration, 1 mM) was added to the oxidized vesicle solution prior to the assay of ATP synthesis to chelate free Cu 2ϩ . EDTA was also added to the reduced vesicle solution to adjust the conditions. After a 5-min preincubation, the reaction was initiated by adding 5 mM NADH and terminated at 2, 4, 6, 8, 10, and 12 min by adding 2.5% trichloroacetic acid. The solution was neutralized to pH 7.7 with 0.25 M Tris acetate (pH 9.5), and the amount of synthesized ATP was determined with ATP bioluminescence assay kit CLSII (Roche Applied Science).
Other Assays-ATPase activity was monitored in triplicate in 50 mM HEPES/KOH, pH 7.5, containing 100 mM KCl, 5 mM MgCl 2 , and 3 mM ATP with an ATP-regenerating system (31), and average hydrolysis rates in a time period from 3 to 6 min after initiation of the reactions at 40°C were measured. The activity that hydrolyzed 1 mol of ATP per minute was defined as one unit. 2Ј,3Ј-O-(2,4,6-trinitrophenyl)-ATP (TNP-ATP) and 2Ј,3Ј-O-(2,4,6-trinitrophenyl)-ADP (TNP-ADP) were purchased from Molecular Probes (Eugene, OR). Fluorescence change induced by the binding of TNP-nucleotide to the enzyme was monitored in a Spectrofluorometer model FP-6500 (Jasco, Tokyo, Japan) as performed previously (32). Protein concentrations were determined by using the BCA protein assay kit from Pierce, with bovine serum albumin as a standard.

RESULTS
Mutants-We generated three mutants. To obtain the enzyme with the up-fixed ⑀ by cross-linking between ␥ and ⑀, one cysteine residue was introduced into the N-terminal region of the ␥ subunit (␥S3C) and another was added to the C-terminal end of ⑀ subunit (⑀Cys 134 ) (Fig. 1D). To obtain the enzyme with the down-fixed ⑀ by cross-linking two domains within ⑀, a mutant that had ⑀A85C and ⑀Cys 134 was used (Fig. 1E). To assess the relative population of enzymes with up-and downstate ⑀ under various conditions, three mutations, ␥S3C, ⑀A85C and ⑀Cys 134 were introduced. The ⑀Ala 85 is located in a region between two domains of ⑀ and abuts on ⑀Cys 134 (C␣ distance is 9.0 Å) when the C-terminal domain adopts a hairpin structure (14). Therefore, ⑀Cys 134 is expected to make the cross-link with ␥S3C when the ⑀ is in the up-state or with ⑀A85C when the ⑀ is in the down-state (Fig. 1F). These mutations were termed ␥ c ⑀ c , ⑀ cc , and ␥ c ⑀ cc , respectively. The enzymes have one endogenous cysteine residue in the F 0 a subunit. This cysteine is buried inside of the transmembrane region and does not respond to the CuCl 2 and 5,5Ј-dithiobis-(2-nitro)benzoic acid (DTNB) treatment employed in this report. The enzymes containing ␥ c ⑀ c , ⑀ cc , and ␥ c ⑀ cc were as active as the wild-type enzyme in their reduced forms (Table I). Also ATPase activities of the inverted membrane vesicles prepared from the cells expressing the wild-type and the F 0 F 1 mutants were similar to each other under the reducing conditions. SDS-PAGE analysis of the membrane vesicles showed almost identical band patterns for the three mutants and the wild-type (not shown). Therefore, the amounts of expressed mutant F 0 F 1 in the inverted membranes are similar to that of the wild-type F 0 F 1 .
Cross-linking of ⑀ in the Up-state-In the previous experiments using E. coli F 0 F 1 (26), cysteine residues were introduced at positions ␥99 and ⑀118 (E. coli numbering) to fix the conformation of ⑀ in the up-state by a ␥-⑀ cross-link. These positions were chosen based on the crystal structure of the ␥Ј⑀ complex in which ␥99 is located in the globular domain of ␥Ј and ⑀118 is in the extreme C-terminal helix of ⑀. In this structure, two helices do not fully extend but rather entwine the globular domain of ␥Ј (Fig. 1C). Expecting that helices of the up-state ⑀ in the native enzyme could extend straighter, we introduced cysteine residues to a near N-terminal position of ␥ (␥Ser 3 ) atop the central helical coiled-coil of the ␥ subunit and to the C terminus of ⑀ (⑀134) (Fig. 1D). The membrane vesicles of E. coli expressing ␥ c ⑀ c -F 0 F 1 were oxidized in 20 M CuCl 2 for 20 min at 25°C and analyzed with non-reducing SDS-PAGE (SDS-PAGE without prior reducing treatment) ( Fig. 2A). Compared with a control ␥ c ⑀ c -F 0 F 1 , which was treated with 50 mM DTT prior to electrophoresis ( Fig. 2A, lane 1), a new band appeared just below the band of the ␤ subunit ( Fig. 2A, lane 4, indicated by an arrow). Peptide sequencing of this band gave two kinds of amino acid sequences corresponding to the N terminus sequences of ␥ and ⑀, indicating that this band is a cross-link product of these two subunits. Consistently, band intensities of ␥ and ⑀ decreased. The same ␥-⑀ cross-link product was also readily generated in the purified ␥ c ⑀ c -F 1 under the same oxidizing conditions (Fig. 2A, lane 5). Cross-linking yields in the F 1 and F 0 F 1 were estimated from the band intensities to be 80 -85%. It is worth noting that ␥-⑀ cross-link was generated spontaneously in ϳ40% of F 1 during the purification (2 days) that was carried out without DTT and EDTA. Also, ϳ60% of F 0 F 1 in membrane vesicles were spontaneously oxidized during preparation. The efficient cross-linking between ␥S3C and ⑀Cys 134 suggests that their proximal location is in the up-state conformation of ⑀ in F 1 and F 0 F 1 and that the C-terminal helix of ⑀ inserts itself deep into the central cavity of the ␣ 3 ␤ 3 . Because the isolated ␥ c ⑀ c -F 1 used in the above experiments was mostly free from endogenous nucleotide, the ⑀ mostly adopts the up-state in the absence of bound nucleotides.
Activities of F 1 and F 0 F 1 with the Up-fixed ⑀ -The ATPase activity of ␥ c ⑀ c -F 1 was severely inhibited by oxidation with its residual activity being only 21% of that of the reduced ␥ c ⑀ c -F 1 , whereas the ATPase activity of the wild-type F 1 was hardly affected whether it was oxidized or reduced (Fig. 2B, left panel). The degree of inhibition by oxidation (79%) agreed well with the yield of cross-link by oxidation (81%). Similarly, the ATPase activity of the ␥ c ⑀ c -F 0 F 1 contained in the vesicles was inhibited (77%) in proportion to the yield of cross-linking (80%) (Fig. 2B, right panel). Because ATP hydrolysis was blocked, oxidized ␥ c ⑀ c -F 0 F 1 was unable to mediate ATP-driven proton translocation, whereas reduced ␥ c ⑀ c -F 0 F 1 was fully capable of it (Fig. 2C). ATP synthesis activities of membrane vesicles containing reduced or oxidized ␥ c ⑀ c -F 0 F 1 were also measured. Membrane vesicles containing the wild-type and ␥ c ⑀ c -F 0 F 1 treated with DTT catalyzed ATP synthesis at 44.5 Ϯ 2.8 and 34.7 Ϯ 2.4 nmol of ATP/min/mg of membrane protein, respectively. Oxidized vesicles showed 71% (wild-type) and 75% (␥ c ⑀ c -F 0 F 1 ) of the ATP synthesis activity of the vesicles treated with DTT (Fig. 2D). Thus, ATP synthesis activity was retained after the formation of the ␥-⑀ cross-link to lock the ⑀ in the up-state. These results are consistent with the previous reports, demonstrating remarkable asymmetric inhibition by the up-state ⑀ toward ATP hydrolysis (18,26,33).
Effect of ATP and ADP on the Conformational State of ⑀ -To assess the distribution of the ⑀ either in the up-state or down- FIG. 1. Conformations of the ⑀ subunit. A, crystal structure of the isolated E. coli ⑀ subunit (16). N-terminal and C-terminal domains were shown with green and red/yellow colors, respectively. B, crystal structure of the down-state conformation of the ␦ subunit (equivalent to the ⑀ subunit in bacterial F 1 ) observed in bovine mitochondrial F 1 (17). Only subunits of ␤ TP , ␥, ␦, and ⑀ (no equivalent subunit in bacterial F 1 ) are depicted in the figure. A loop that contains a DELSEED sequence was colored purple. C, crystal structure of the ␥Ј⑀ complex of E. coli F 1 (25) superimposed with the ␤ TP of bovine F 1 . Blue spheres indicate ␤E395 residue of the DELSEED region (second Glu residue). C␣ distance between ␤Glu 381 and ⑀Ser 108 is 24 -27 Å, too far to be cross-linked. D-F, schematic diagrams of cross-link formation in ␥ c ⑀ c -F 1 (D), ⑀ cc -F 1 (E), and ␥ c ⑀ cc -F 1 (F). In the up-state, ⑀Cys 134 -␥S3C is to be cross-linked. In the down-state, ⑀Cys 134 -⑀A85C is to be cross-linked. These figures were prepared by using a program package, MOLMOL (45). ⑀S110, ⑀Ser 110 ; ␤E395, ␤Glu 395 ; ⑀C134, ⑀Cys 134 .

Two States of ⑀ Subunit in ATP Synthase
state, ␥ c ⑀ cc -F 1 and ␥ c ⑀ cc -F 0 F 1 were used. With oxidation procedures, the down-state ⑀ can be detected as a band corresponding to an internally cross-linked ⑀ (⑀A85C and ⑀C 134 ) and the up-state ⑀ as a ␥-⑀ band (␥S3C and ⑀Cys 134 ). The ⑀ subunit behaved very similarly in ␥ c ⑀ cc -F 1 and ␥ c ⑀ cc -F 0 F 1 (Fig. 3, A and  B). In the absence of nucleotides, the ⑀ in F 1 and in F 0 F 1 was mostly in the up-state (Fig. 3, A and B, lanes 2), and the up-state conformation was stabilized when 3 mM ADP was present (Fig. 3, A and B, lanes 3). 2 The further addition of 5 mM P i caused no significant change (not shown). However, the ␥-⑀ band disappeared, and the internally cross-linked ⑀ band (Fig.  3, A and B, arrowheads) appeared when ADP was converted into ATP by pyruvate kinase (Fig. 3, A and B, lanes 4). Also, the internally cross-linked ⑀ band appeared when ATP was added from the beginning (Fig. 3, A and B, lanes 5). The addition of hexokinase and glucose to the sample of the lanes 5 (Fig. 3, A  and B) resulted in the appearance of the ␥-⑀ band (Fig. 3, A and  B, lanes 6). Thus, it is clear that the ⑀ subunit in F 1 and F 0 F 1 adopts reversibly the up-state conformation in the presence of ADP and the down-state conformation in the presence of ATP. As shown previously (19), hydrolysis of ATP is not necessary to stabilize the down-state ⑀, because 3 mM AMP-PNP also stabilized the down-state conformation of ⑀ (not shown).
TNP-AT(D)P Binding to F 1 with Up-or Down-state ⑀ -It has been known that a nucleotide analogue, TNP-AT(D)P, increases its fluorescence upon binding to F 1 (32). Taking advantage of this, we compared initial kinetics of nucleotide binding to the enzymes that contained the up-or down-state ⑀. To measure the binding to the nucleotide binding site with the highest affinity, a sub-stoichiometric amount of TNP-AT(D)P was mixed with ␥ c ⑀ c -F 1 or ⑀ cc -F 1 , and fluorescence changes were monitored. Time courses of TNP-ADP binding were almost the . ATP hydrolysis by the reduced (white bars) and oxidized (black bars) ␥ c ⑀ c -F 1 and membrane vesicles containing ␥ c ⑀ c -F 0 F 1 were assayed at 40°C. The same procedures were applied to the wild-type F 1 and F 0 F 1 . C, effect of the ␥-⑀ cross-linking on proton pump activity. Proton pump activities of the reduced or oxidized membrane vesicles were analyzed by monitoring the fluorescence of 9-amino-6-chloro-2methoxyacridine at 40°C. Prior to the analysis, 1 mM EDTA was added to the solutions to remove free Cu 2ϩ . At the indicated times, pumping was initiated by adding 1 mM ATP and terminated by 1 g/ml FCCP. D, effect of the ␥-⑀ cross-linking on ATP synthesis activity. The reactions were started by addition of 5 mM NADH to the membrane vesicle solutions containing reduced or oxidized wild-type F 0 F 1 and ␥ c ⑀ c -F 0 F 1 . The reactions were carried out at 50°C, and the amount of generated ATP was measured with luciferase. same for ␥ c ⑀ c -F 1 and ⑀ cc -F 1 , irrespective of whether they were reduced or oxidized (Fig. 4, A and B). The wild-type F 1 , with or without oxidizing treatment, also showed the same kinetics of TNP-ADP binding (not shown). These results indicated that TNP-ADP binding to F 1 was not affected by the conformational states of the ⑀ subunit. The time course of TNP-ATP binding to reduced ␥ c ⑀ c -F 1 was also the same as that of ⑀ cc -F 1 (Fig. 4C) and wild-type F 1 (not shown), ensuring no significant effect of the introduced cysteines on the TNP-ATP binding kinetics of F 1 . The time course of TNP-ATP binding to the oxidized ␥ c ⑀ c -F 1 (Fig. 4D, bottom curve) was similar to that of TNP-ADP binding to the oxidized ␥ c ⑀ c -F 1 , indicating that TNP-ATP and TNP-ADP bind to the same site of F 1 with the up-fixed ⑀. The oxidized ⑀ cc -F 1 , on the contrary, bound TNP-ATP much faster, and the fluorescence reaches a higher magnitude than with the oxidized ␥ c ⑀ c -F 1 . (Fig. 4D, upper curve). Thus, F 1 with the downstate ⑀ binds TNP-ATP quickly, whereas F 1 with the up-state ⑀ binds it slowly. Accordingly, results of TNP-ATP binding to the reduced ␥ c ⑀ c -F 1 and ⑀ cc -F 1 in Fig. 4C are well interpreted as a mixture of F 1 s with the up-and down-state ⑀.
Effect of Proton Motive Force on the State of ⑀ -The inverted membrane vesicles containing ␥ c ⑀ cc -F 0 F 1 were incubated for 3 min in the varying amounts of ATP and ADP, and conformational states of the ⑀ subunit were analyzed with non-reducing SDS-PAGE after fixing the conformation by cross-linking (Fig.  5A, lanes 1-6). As ATP increased and ADP decreased, intensity of the ␥-⑀ band decreased, as is expected from the results mentioned above. However, when the incubation was continued for another 5 min after the addition of NADH to impose proton motive force, the intensity of the ␥-⑀ band did not significantly decrease even at high ATP concentrations (Fig. 5A,  lanes 7-12). When FCCP, an uncoupler that dissipates proton motive force, was added in addition to NADH, the intensity of the ␥-⑀ band was decreased as ATP increased, similar to Fig.  5A, lanes 1-6 (Fig. 5A, lanes 13-18). These results suggest that when proton motive force is provided, the ⑀ subunit in F 0 F 1 strongly favors the up-state conformation irrespective of ADP/ ATP balance. In other words, proton motive force counteracts the effect of ATP in the conformational transition of the ⑀ subunit.

DISCUSSION
C Terminus of the ⑀ Subunit Reaches the Center of F 1 -The questions listed in the Introduction were mostly answered by the present study. Concerning the first question (i), it becomes evident that C terminus of ⑀ in the up-state is located near the N terminus of the ␥ subunit. 3 To reach this position, referring to the structure of mitochondrial F 1 , the C-terminal helices of ⑀ have to extend ϳ70 Å from the exit (⑀A85) of the N-terminal ␤ sandwich domain. Considering the length of ␣-helix per residue (1.5 Å/residue) (34), a peptide stretch of 48 residues from ⑀Ala 85 to ⑀Lys 133 can extend by 72 Å as an ␣-helix or longer as two helices with a connecting segment. Previous cross-linking results of E. coli F 1 between ␤Glu 381 and ⑀Ser 108 (9, 20 -22) are explained by this new arrangement rather than by the ␥Ј⑀  6, 12, and 18), respectively. The experiments were carried out at 50°C in 50 mM HEPES/ NaOH, pH 7.5, containing 100 mM NaCl and 5 mM MgCl 2 . It was confirmed that NADH oxidation by respiratory chains of vesicles under these conditions was as active as that at 37°C and could generate proton motive force. In this figure, ␥-⑀ bands formed by cross-linking are shown. B, the relative staining intensities of the ␥-⑀ bands in panel A were plotted against the ATP concentrations. Closed triangles, the control samples (lanes 1-6); closed circles, NADH (lanes [7][8][9][10][11][12]; open circles, NADH ϩ FCCP (lanes 13-18). structure (Fig. 1, compare C and D) (25). Probably, the ␥Ј⑀ structure represents an intermediate conformation that appears during the transition of ⑀ from the down-state to the up-state. Our study suggests that three helices, the coiled-coil of the ␥ subunit and C-terminal helix of the ⑀ subunit, rather than two as previously thought, rotate as a body within the ␣ 3 ␤ 3 ring when F 0 F 1 with the up-state ⑀ is synthesizing ATP.
ATP and ADP Have Opposite Effects on the Conformational States of the ⑀ Subunit-As to the second question (ii), it is now clear that the ⑀ subunit, either in F 1 or F 0 F 1 , is in the up-state conformation in the absence of nucleotide or the presence of ADP and is in the down-state conformation in the presence of ATP. Thus, ATP and ADP counteract each other (Fig. 6). Reciprocally, an ATP analogue, TNP-ATP, binds quickly to F 1 with down-state ⑀ but slowly to F 1 with up-state ⑀. An ADP analogue, TNP-ADP, does not show binding preference between F 1 s with up-and down-state ⑀. Therefore, if TNP-AT(D)P mimics the AT(D)P correctly in binding to F 1 , the answer to the third question (iii) will be that F 1 s with down-state ⑀ indeed prefer ATP to ADP, whereas F 1 s with up-state ⑀ bind both ATP and ADP in the same slow kinetics. The results are consistent with the previous observation that the ␣ 3 ␤ 3 ␥ complex binds TNP-ATP quickly, but the reconstituted ␣ 3 ␤ 3 ␥⑀ complex does this slowly (32), because without previous exposure to nucleotide, the ⑀ subunit in the reconstituted ␣ 3 ␤ 3 ␥⑀ must be in the up-state.
The ⑀ Subunit Transits between Two States Depending on Proton Motive Force and ADP/ATP-This study has revealed that proton motive force counteracts the effect of ATP by stabilizing the up-state ⑀ (answer to the fourth question (iv)). Therefore, the two conformational states of ⑀ in F 0 F 1 are alternated by two factors, i.e. proton motive force and ADP/ATP balance (Fig. 6). At high proton motive force and low ATP, ⑀ is predominantly in the up-state, and F 0 F 1 is geared to the ATP synthesis mode. At low proton motive force and high ATP, ⑀ adopts the down-state and F 0 F 1 hydrolyzes ATP to pump out protons, generating proton motive force with enough magnitude to drive uptake of nutrients and flagella motion.
Role of C-terminal Helices of the ⑀ Subunit-In some bacteria, such as Chlorobium limicola (35) and Thermotoga neapolitana (36), the native ⑀ subunit lacks the C-terminal helical domain. Without the C-terminal helical domain, the ⑀ subunit cannot adopt the up-state arrangement and should be always in the state that is functionally similar to the down-state. These bacteria grow in anaerobic environments, and F 0 F 1 should work as an ATP hydrolysis-driven proton pump. Because the F 0 F 1 with up-state ⑀ is unable to mediate ATP hydrolysis-driven proton pumping, these bacteria do not need, or even had better delete, the C-terminal domain of the ⑀ subunit. F 0 F 1 with down-state ⑀ can catalyze both ATP synthesis and ATP hydrolysis (26). Therefore, it is not surprising that a mutant E. coli F 0 F 1 containing the ⑀ subunit with deleted C-terminal helical domain or an artificially fused protein at the C terminus can support aerobic growth by oxidative phosphorylation (37,38). A similar observation was reported recently for chloroplast F 0 F 1 (33). Then, a critical question should be asked, i.e. what is the essential function of the F 0 F 1 whose ⑀ subunit is in the up-state? Probably, the F 0 F 1 with the up-state ⑀ plays an important role under starving conditions rather than in rich nutritional environments. In E. coli cells, total concentration of cellular adenine nucleotides is maintained to be ϳ3 mM (39), but the fraction of ATP in total adenine nucleotide pool varies from 3 to 0.3 mM in parallel with growth rate (40 -42) through ribosome synthesis (43) and transcription (44). As ATP concentration decreases from 3 to 0.3 mM in the absence of proton motive force, the population of the F 0 F 1 with up-state ⑀ increases about 3-fold (Fig. 5B) so that hydrolysis of the precious ATP by F 0 F 1 is suppressed. For any organisms, regulation of ATP synthesis/hydrolysis to meet physiological demand in quickly changing nutritional conditions is a critical matter, and conformational transition of the ⑀ subunit in F 0 F 1 might constitute a part of an elaborately integrated regulatory system that awaits further study. FIG. 6. Schematic diagram of two forms of the F 0 F 1 with upstate ⑀ (left) and down-state ⑀ (right). The F 0 F 1 with up-state ⑀ can catalyze ATP synthesis but not ATP hydrolysis (ATP synthesis mode). This form is stabilized by ADP and proton motive force. The F 0 F 1 with down-state ⑀ can catalyze ATP synthesis as well as ATP hydrolysis (proton pump/ATP synthesis mode). This form is favored when ATP is present. Transition between two forms is determined by proton motive force and ADP/ATP balance.