Single-molecule Analysis of F0F1-ATP Synthase Inhibited by N,N-Dicyclohexylcarbodiimide*

Background: DCCD is a classical inhibitor of F0F1 that modifies the proton binding site of the c subunit. Results: Single-molecule analysis showed that a single modification in the multiple c subunit complex significantly inhibits F0F1. Conclusion: Bound DCCD induces steric hindrance of the c subunit of the a subunit and hence blocks F0F1 rotation. Significance: This is the first direct evidence to reveal the DCCD inhibition mechanism. N,N-Dicyclohexylcarbodiimide (DCCD) is a classical inhibitor of the F0F1-ATP synthase (F0F1), which covalently binds to the highly conserved carboxylic acid of the proteolipid subunit (c subunit) in F0. Although it is well known that DCCD modification of the c subunit blocks proton translocation in F0 and the coupled ATP hydrolysis activity of F1, how DCCD inhibits the rotary dynamics of F0F1 remains elusive. Here, we carried out single-molecule rotation assays to characterize the DCCD inhibition of Escherichia coli F0F1. Upon the injection of DCCD, rotations irreversibly terminated with first order reaction kinetics, suggesting that the incorporation of a single DCCD moiety is sufficient to block the rotary catalysis of the F0F1. Individual molecules terminated at different angles relative to the three catalytic angles of F1, suggesting that DCCD randomly reacts with one of the 10 c subunits. DCCD-inhibited F0F1 sometimes showed transient activation; molecules abruptly rotated and stopped after one revolution at the original termination angle, suggesting that hindrance by the DCCD moiety is released due to thermal fluctuation. To explore the mechanical activation of DCCD-inhibited molecules, we perturbed inhibited molecules using magnetic tweezers. The probability of transient activation increased upon a forward forcible rotation. Interestingly, during the termination F0F1, showed multiple positional shifts, which implies that F1 stochastically changes the angular position of its rotor upon a catalytic reaction. This effect could be caused by balancing the angular positions of the F1 and the F0 rotors, which are connected via elastic elements.

F 0 F 1 -ATP synthase (F 0 F 1 ) is a large multisubunit complex that catalyzes ATP synthesis from ADP and inorganic phosphate (P i ) by using the proton motive force that arises during oxidative phosphorylation and photosynthesis (1)(2)(3)(4). F 0 F 1 is composed of two structurally and functionally distinct parts: the F 1 domain, which is responsible for catalysis, and the F 0 domain, which conducts proton translocation across the membranes. Bacterial F 0 F 1 has the simplest subunit composition: ␣ 3 ␤ 3 ␥␦⑀ for F 1 and ab 2 c 10 -15 for F 0 . Although the number of c subunits can vary between species, it is 10 in Escherichia coli (5)(6)(7)(8). Both F 0 and F 1 act as rotary molecular motors. F 1 hydrolyzes ATP into ADP and inorganic phosphate (P i ) to rotate the ␥⑀ complex against the surrounding ␣ 3 ␤ 3 stator ring. Its catalytic sites reside on the three ␣-␤ interfaces. In contrast, F 0 rotates the oligomer ring of the c subunit (c-ring) against the a subunit, and the ␥⑀ complex binds to the c-ring to form the common rotary shaft. The stator components of F 1 and F 0 (the ␣ 3 ␤ 3 ring and a subunit, respectively) are connected via the peripheral stalk, which is composed of the ␦ subunit and the b 2 complex.
In physiological conditions, F 0 , when powered by the proton motive force, rotates the rotor complex (the ␥⑀-c 10 -15 complex) against the ␣ 3 ␤ 3 ␦-ab 2 stator complex, leading to a cycle of conformational changes in the ␣ 3 ␤ 3 assembly and ATP synthesis from ADP and P i . Conversely, when proton motive force diminishes, F 1 hydrolyzes ATP, leading to reverse rotation of the rotor complex and causing F 0 to pump protons in the reverse direction.
Single-molecule studies have extensively elucidated the chemomechanical coupling scheme and torque generation mechanism of F 1 (5). The rotation consists of an elementary 120°step, which is consistent with the three-fold structural symmetry of F 1 (9). Each 120°step is coupled with a single turnover of ATP hydrolysis and can be further resolved into discrete 80 and 40°substeps, which are triggered by ATP binding and ADP release or ATP-hydrolysis and P i release, respectively (10 -13). Although some uncertainties remain, a detailed coupling reaction scheme for F 1 has been proposed (5).
In contrast to F 1 , the reaction scheme of the H ϩ translocation and the torque generation mechanism of F 0 are poorly understood. One widely accepted working model is the two halfchannels model (14). Here, the H ϩ -conducting path of F 0 is formed by two hemi-channels in the a subunit (15)(16)(17) and the H ϩ -binding sites on the c-ring located on each of the individual c subunits. Under a proton motive force, H ϩ enters from the half-channel that faces the periplasm and reaches the H ϩ -binding site of the c subunit. After one revolution of the c-ring, the bound H ϩ is transferred to the other half-channel that faces the cytoplasmic side and is finally released into the cytoplasmic space. Although structural details of the a-c interface are lacking, it is thought that torque is generated upon the H ϩ -trans-ferring steps between the a and c subunits. Thus, the postulated elementary angular step of the c-ring rotation is 360°divided by the number of c subunits, which translates into 36°for the bacterial c-ring (c 10 ). Recent single-molecule studies have observed F 0 rotation with multiple steps (up to 10) under ATP hydrolysis conditions, probably reflecting the structural symmetry of the c-ring (7,8), whereas a three-step rotation was reported under ATP synthesis conditions (18).
The c subunit is composed of two transmembrane helices connected by a short loop on which the ␥⑀ complex binds (19 -22). The H ϩ -binding site that forms the H ϩ translocation path is a highly conserved carboxyl residue (Asp-61 in E. coli) at the middle of the secondary helix located on the outer circumference of the c-ring. One NMR study showed that the pK a of the H ϩ -binding carboxyl residue is higher than that of the carboxylate in aqueous condition, suggesting that the H ϩ -binding carboxyl residue predominantly stays protonated in the membrane (23). Pogoryelov et al. (21,24) proposed that when facing the exit half-channels, the H ϩ -binding carboxyl residue orients out of the binding pocket and releases H ϩ .
N,N-Dicyclohexylcarbodiimide (DCCD) 2 is the gold standard inhibitor of F 0 F 1 and has been widely used to study the corresponding coupling reaction. DCCD specifically reacts with the H ϩ -binding carboxyl residue of the c subunit by forming a stable N-acyl urea (25) and blocks the coupling reaction of F 0 F 1 (26,27). Because DCCD reacts with carboxyl groups in the protonated state, carboxyl groups exposed in water solvent are much less reactive. Although F 1 has a DCCD-reactive glutamate at the catalytic site, the glutamate residue is in a more aqueous condition than the H ϩ -binding carboxyl residue of the c subunit. In addition, the glutamic acid of F 1 requires a Mg 2ϩfree solution to react with DCCD (28). Thus, reactivity of the glutamate residue of F 1 is negligible. It was reported that the incorporation of a single DCCD molecule per F 0 is sufficient to inhibit coupled ATPase activity (29). The crystal structure of the DCCD-modified c-ring (Protein Data Bank codes 2XQU, 2XQS, and 2XQT) showed that the cyclohexyl moieties are too large to enter the proton binding pocket of the c-subunit and protrude outwardly into the hydrophobic environment (24), implying that steric hindrance of the N-acyl urea with the a subunit blocks the rotation of the c-ring.
In single-molecule rotation assays, DCCD has been used to verify whether F 0 F 1 retains the activity of the coupling reaction. Although pretreatment of the sample with DCCD decreases the probability of rotating particles (30), DCCD inhibition during rotation observation has not been attained. Therefore, it remains unclear how DCCD blocks the rotation. In the present study, we analyzed DCCD inhibition using the single-molecule rotation assay by introducing DCCD while observing rotating particles to characterize the rotary dynamics of DCCD-inhibited F 0 F 1 .

EXPERIMENTAL PROCEDURES
Preparation of F 0 F 1 -ATPase-Briefly, the 1.3 S subunit of Propionibacterium shermanii transcarboxylase, which contains a biotinylated domain, was fused to the N terminus of the ␤ subunits for attaching the magnetic beads, and three histidine residues were introduced at the C terminus of the c subunits. Construction and purification of mutant E. coli F 0 F 1 -ATPase (EF 0 F 1 ) for the rotational analysis were previously described by Iino et al. (31).
The E. coli strain RA1 (unc Ϫ /cyo Ϫ ) was transformed with a F 0 F 1 mutant plasmid and cultured in 1.2 liters of medium containing 30 g/ml chloramphenicol for 16 h at 37°C. Inverted membrane vesicles were prepared as follows. After the cell wall was digested with lysozyme treatment, the spheroplast was collected by centrifugation, resuspended, and broken down by sonication. The suspended mixture was centrifuged at 6,000 ϫ g for 10 min at 4°C to remove any cell debris. The supernatant containing the inverted membrane was transferred to a new tube and centrifuged at 300,000 ϫ g for 20 min at 4°C. The supernatant was discarded, and the pellet of membranes was resuspended in buffer A (100 mM HEPES-KOH (pH 7.5) and 50 mM KCl).
ATPase Activity Measurements-The ATP hydrolysis activity of solubilized F 0 F 1 was measured with an ATP regeneration system using a UV-visible spectrophotometer (VP-550, JASCO). 0.7 pmol of EF 0 F 1 was suspended in 200 l of buffer and incubated for a given time with various amounts of DCCD solubilized in ethanol at 25°C. ATPase activity was started by introducing the F 0 F 1 solution into buffer E (20 mM HEPES-NaOH (pH 7.5), 2 mM MgSO 4 , 100 mM KCl, 0.3% (w/v) C 12 E 8 , 0.1% (w/v) E. coli total lipid) containing 1.8 mM ATP, 2.5 mM phosphoenolpyruvate, 0.1 mg/ml pyruvate kinase, 0.1 mg/ml lactate dehydrogenase, and 0.2 mM NADH. The NADH absorbance at 340 nm was monitored for 900 s, and the ATP hydrolysis rate was calculated from the time course of the change in [NADH] using a molecular extinction coefficient value of 6,220 at 340 nm. After the ATP hydrolysis activity reached steady state, 0.2% (w/v) lauryldimethylamine oxide (LDAO) was added to confirm the reactivation of the activity. It has been reported that LDAO activates DCCD-inhibited F 0 F 1 by inducing the uncoupling of F 0 and F 1 (32).
Rotation Assay-Coverslips coated with nickel-nitrilotriacetic acid were prepared as described previously (33). A flow cell of 10 -30 l in volume was made using two coverslips (bottom, 24 ϫ 36 mm 2 ; top, 24 ϫ 24 mm 2 ) separated by two spacers of 50-m thickness. F 0 F 1 was infused into the flow cell and incubated for 5 min. Unbound F 0 F 1 was washed out with 100 l of buffer E, and buffer E containing 1% BSA was infused into the flow cell to reduce nonspecific binding of the beads. A solution of streptavidin-coated magnetic beads of 200-nm diameter (Seradyn, Thermo Scientific) was infused into the flow cell. After a 10-min incubation, unbound beads were washed out by the infusion of 500 l of buffer E containing 1.8 mM ATP. Rotation of the bead was observed under phase-contrast microscopy (IX70, Olympus) using a 60ϫ objective lens. Images were captured with a charge-coupled device camera (FC300M; Takenaka) and recorded with a DV-CAM device (DSR-11; Sony) at 30 frames/s. Image analysis was performed using custom software (Digimo). Time-averaged rotation speed was calculated from over five consecutive revolutions. All experiments were carried out at 23°C.
Mechanical Manipulation by Magnetic Tweezers-The experimental setup of the magnetic tweezers was previously described in detail by Hirono-Hara et al. (34). A schematic image of the magnetic tweezers is shown in Fig. 1A. The tweezers comprise four electromagnets, each constructed from a soft iron core (10 ϫ 10 ϫ 40 mm) and a copper wire with 100 turns around the core. Each pair of tweezers was electrically connected in series and separated by an interval of 15 mm. The two electromagnet pairs were crossed at each center space and positioned 10 mm above the microscope stage. The magnetic tweezers generated a magnetic field parallel to the stage, and the angle of the composite magnetic field was controlled by applying the sine components of the electric current to one pair (y axis) and the cosine components to the other (x axis). The field intensity was controlled by changing the current amplitude. The magnetic field was measured with a gauss meter (421 gaussmeter, Lake Shore Cryotronics, Westerville, OH) to confirm that the magnetic tweezers generated ϳ200 G at the center of the focal plane with a precision of 4% for the intensity and 5°for the angle.
Manipulation of the magnetic tweezers was performed as follows. The tweezers were activated to trap the magnetic bead at the targeted angle. After a 3-s stall, the external magnetic field was turned off to observe whether the F 0 F 1 rotated forward or backward. All procedures were controlled with custommade software (Disimo). The image of the beads was recorded using the same method as that described under "Rotation Assay" above.

RESULTS
Rotation Behavior of F 0 F 1 in the Presence of DCCD-To visualize the rotation of F 0 F 1 using optical microscopy, a magnetic bead was attached to a biotin-binding domain introduced at the N terminus of the ␤ subunit. Although the rotor complex of ␥⑀c 10 was kept immobilized, the stator ␣ 3 ␤ 3 ␦ab 2 complex rotated counterclockwise in the presence of ATP.
We selected smoothly rotating F 0 F 1 molecules that had no obvious interference by steric interactions with the glass for the analysis. Typical time courses of the rotations are shown in Fig.  1B. Although F 0 F 1 showed smooth consecutive rotation under saturating ATP (1.8 mM), they also showed short pauses during the rotation. The average duration time of these pauses was 4.9 s, which is consistent with a previous study (35) that showed EF 0 F 1 transiently enters an inhibition state during the ATPdriven rotation, and the average duration time of this state is 2.57 Ϯ 0.77 s. When 1 mM DCCD was introduced into the flow chamber, 27 of 29 F 0 F 1 molecules eventually terminated their continuous rotations. F 0 F 1 never resumed consecutive rotations within the observation time (more than 10 min). The other two molecules (Fig. 1B, red lines) showed rotations for more than 13 min after the DCCD injection. In contrast, when DCCD-free buffer was introduced, most molecules continued rotating ( Fig. 2A, diamonds). Thus, we concluded that the terminations that appeared after the DCCD injection represent the DCCD inhibition state of F 0 F 1 . In the following analysis, we define the termination state as the DCCD inhibition state that appeared after DCCD injection and lasted more than 2 min.
Kinetic Analysis of DCCD Inhibition-To analyze the kinetics of the DCCD inhibition, we measured the rotating time, which was defined as the period from the DCCD injection until the time when the molecules lapsed into the DCCD-inhibited state, at various concentrations of DCCD (200 M, 600 M, and 1 mM). The time courses of the rotation probability were calculated from the rotating time ( Fig. 2A, circles, triangles, and squares). The probability exponentially decayed to approach a constant value that corresponded to the DCCD-insensitive fraction. The time courses were fitted with a single exponential equation, y ϭ 100 Ϫ a)eϪ kt ( Fig. 2A, solid lines), where k and a represent the rate constant of the DCCD inhibition and the fraction of DCCD-insensitive molecules, respectively. We also conducted kinetic analysis of the DCCD inhibition on the ATPase activity in solution. ATPase activity decreased with the incubation time of DCCD, similar to the rotation assay results (Fig. 2B). It should be noted that it is reasonable to assume that preincubation of F 0 F 1 with DCCD should show the same DCCD inhibition rate as real-time monitoring of DCCD inhibition because the concentration of the remaining DCCD in the ATPase assay mixture is low enough to prevent further F 0 F 1 inhibition. In addition, we confirmed that the presence of ATP does not affect any DCCD inhibition rate of F 0 F 1 . Fig. 2C shows k increased in proportion to the DCCD concentration, suggesting that DCCD inhibition is a first ordered reaction and that a single DCCD modification of the c-ring is sufficient to block the rotation. Additionally, the values of k are in good agreement with those values from the rotation assay. Finally, the rate constant of the DCCD inhibition standardized with the DCCD concentration was 2.4 ϫ 10 Ϫ5 and 1.8 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 according to the ATPase and rotation assays, respectively.
It is known that, along with the c subunit, DCCD can react with a highly conserved glutamate residue in the catalytic sites of F 1 and inhibits the ATPase activity, although with a much lower reactivity (28,36). To confirm that DCCD predominantly reacted with the c subunit in our experiments, we tested whether the bulk ATPase activities of F 0 F 1 recovered from the DCCD inhibition by adding the detergent LDAO. LDAO is known to decouple the ATPase activity of F 1 from the H ϩ -conducting activity of F 0 and thus restore the F 0 F 1 ATPase activity in the presence of LDAO (32). As expected, when 0.2% (w/v) LDAO was added after a 10-min incubation of 1 mM DCCD, more than 80% of the ATPase activity was recovered (Fig. 2B,  diamonds), demonstrating that DCCD predominantly reacted with F 0 . LDAO was not tested in the rotation assay, however, because it caused nonspecific attachment of the magnetic beads onto the glass coverslip, compromising the results.
Transient Activation from the DCCD Inhibition-We analyzed the rotary dynamics of the DCCD-inhibited state. To avoid additional DCCD binding to F 0 F 1 , DCCD in the flow chamber was washed out by injecting DCCD-free buffer after the DCCD-induced termination was observed. Fig. 3A shows typical time courses of the revolutions of the molecules in the DCCD-inhibited state. Interestingly, some molecules showed transient rotations; that is, they rotated only once and abruptly stopped. In rare cases, inhibited molecules showed a maximum of two or three turns (the bottom time course in Fig. 3A). DCCD-inhibited F 0 F 1 always stopped at the same position of the initial termination (Fig. 3A, green arrows), resulting in a single peak in the histogram of the angular positions (Fig. 3B). This result implies that the rotation of F 0 F 1 was inhibited by a single DCCD modification. The angle distribution of the DCCD-inhibited state was fitted with a Gaussian function. The standard deviation of the angle distribution was relatively large (58.1°, Fig. 3B) when compared with that of the ATP-waiting state of F 1 (37). Finally, the average frequency of the transient activation was determined as 0.018 s Ϫ1 , which corresponds to an ATPase activity of 0.054 s Ϫ1 , a value unlikely detectable in bulk ATPase measurements.
The Acceleration of Transient Activation by Mechanical Manipulation-We investigated whether the spontaneous transient activation could be accelerated by external force. After observing the DCCD-induced termination, the buffer was exchanged to the DCCD-free buffer. A magnetic field was applied to forcibly rotate and stall DCCD-inhibited F 0 F 1 molecules at a target angle for 3 s. After these 3 s, the molecule was released from the tweezers, resulting in two behaviors (Fig. 4A). One behavior was deemed "activation," in which F 0 F 1 rotated forward one turn and terminated (blue arrow). The other was deemed "nonactivation," in which F 0 F 1 returned to the original termination angle (green arrow). In both cases, the angular posi-FIGURE 1. Single-molecule observation of DCCD inhibition of F 0 F 1 . A, experimental system for the observation of the F 0 F 1 rotation. After the immobilization of purified F 0 F 1 on a nickel-nitrilotriacetic acid (Ni-NTA)-coated glass, magnetic beads were attached by fusing the biotin-binding domain to the ␤ subunit of F 0 F 1 . The rotation of the bead was observed using optical microscopy. The rotor (red) rotates counterclockwise against the stator (blue). In stalling experiments, the rotational angle of the bead was controlled by magnetic tweezers. B, time courses of the revolutions of rotating molecules before and after adding 1 mM DCCD. After 5 min of recording the rotation, a buffer containing 1 mM DCCD was infused into the chamber. When the buffer change was finished (ϳ1 min), the recording was restarted. Of the eight rotating molecules, six terminated their rotations, and the remaining two (red lines) kept rotating. Of the former six molecules, four had already terminated their rotation before restarting the recording, and the other two stopped after restarting the video recording. tion approximated the same angle as that before stalling. The inset in Fig. 4A shows histograms of the angle displacements for the two behaviors, with Gaussian distribution peaks at 3 Ϯ 38 and 341 Ϯ 57°for the activation and nonactivation states, respectively. It should be noted that F 0 F 1 terminated at different positions from the original termination angle after being released from the magnetic tweezers when F 0 F 1 was incubated with DCCD for a long time to allow multiple DCCD modifications of the c-ring (data not shown).
The probability of mechanically induced transient activation is shown in Fig. 4B. The activation probability depended on the stall angle; stalling at over ϩ0°significantly increased the activation probability, whereas stalling from 0°to Ϫ360°did not. The probability of activation reached 100% when the system stalled at over ϩ300°. Based on these results, we estimated the activation rate constant assuming a simple first order reaction model.
The probability of activation, p, is given as a function of t as follows  where k is the rate constant of activation. In our experiments, the stall time was 3 s. k was determined from p values at 30, 90, 150, 210, and 270°stalls (Fig. 4C) and found to increase exponentially with the stall angle. Considering the Arrhenius equation (k ϭ Ae (Ϫ⌬G a /k B T), where k B is the Boltzmann constant and T is the temperature, the increase in k means a decrease in the activation energy ⌬G a . The slope of the fitting in Fig. 4C, which represents the activation energy change against the stall angle, was Ϫ0.0146 k B T per 1°. As described above, activation spontaneously occurred at a rate of 0.018 s Ϫ1 without any mechanical manipulation. This spontaneous activation rate corresponds to the activation rate at ϩ44.1°in Fig. 4C, which suggests that spontaneous transient activation occurs when the ␣ 3 ␤ 3 ␦-ab 2 complex is thermally rotated in the forward direction. A similar role for thermal agitation in rate enhancement has been reported for the spontaneous activation of Mg-ADP inhibition and the ATP binding process of F 1 -ATPase (34,38).
The Angular Position of the DCCD-inhibited State-In DCCD-inhibited molecules of the wild-type F 0 F 1 , the positional shifts of the angles were observed (Fig. 5A, red arrows). To determine the angular position of the DCCD-inhibited state in relation with the catalytic angle of F 1 , we conducted a rotation assay using the F 1 (␤E181D) mutant in the presence of 200 M DCCD. The corresponding mutant of F 1 from thermophilic bacteria, TF 1 (␤E190D), is known to have very low ATP hydrolysis activity (11). The catalytic dwell of these mutants is 100fold longer than that of wild type, and hence the mutation causes clear stepwise rotations that pause at all catalytic angles, each 120°apart from the other (11) . Fig. 5, B and C, show two typical time courses of the F 1 (␤E181D) mutant before and after DCCD inhibition. The mutant showed a discrete 120°stepping rotation that paused at the catalytic angles and eventually lapsed into the DCCD-inhibited termination. Similar to the above experiments, after observing the termination, DCCD was washed out to avoid multiple DCCD modifications. During DCCD inhibition, the mutant showed the positional shifts in the termination angles as seen for the wild-type F 0 F 1 (Fig. 5, A-C, red arrows). The positional shifts were, however, more obvious in the mutant, ranging from Ϫ94°to 139°with an average value of 58°(44 transitions for 11 molecules). The histograms of the angular distributions during the terminations showed multiple peaks (Fig. 5, B and C, bottom), with 2-4 peaks seen depending on the molecule.
Relative differences between the angle of DCCD-induced pause and the catalytic angle were determined for all molecules. When differences in the peak position between the catalytic angle histogram and DCCD inhibition angles histogram were calculated (Fig. 5B, red vertical dotted lines), no distinct peak was seen (Fig. 5D). Similarly, when the initial pause position was selected for this analysis, the histogram again had no obvious peak (data not shown). These results suggest that DCCD modifications are random among the 10 c subunits.

DISCUSSION
The DCCD-sensitive Rotation of F 0 F 1 -Several studies on the ATP-driven rotation of F 0 F 1 have been reported since Sam- . Mechanical activation of DCCD-inhibited wild-type F 0 F 1 . A, a typical time course of revolutions during mechanical activation. DCCD-inhibited F 0 F 1 was forcibly rotated and stalled at a target angle with magnetic tweezers. After a 3-s window (red dots), the molecule was released from the magnetic tweezers. The released molecules showed two behaviors, activation and nonactivation. The inset shows a histogram of the angle displacements before and after stalling for activation (blue) and nonactivation (green) events. Gaussian peaks of these histograms were 341 Ϯ 57 and 3 Ϯ 38°, respectively. B, stall angle dependence on the probability of activation. Each column was obtained from 8 -66 data sets. C, stall angle dependence on the activation rate, k, which was calculated using the equation p ϭ 100(1 Ϫ e Ϫkt ) and the datasets at 30, 90, 150, 210, and  bongi et al. (39) first observed the rotation of purified E. coli F 0 F 1 (8,30,40,41). Although DCCD sensitivity has been the gold standard for ensuring that F 0 F 1 molecules retain their competency when coupling ATP hydrolysis with H ϩ translocation, DCCD has not been used in the rotation assay for studying F 0 F 1 . In the early stage of the rotation assay, Triton X-100 has been often used to solubilize F 0 F 1 (39). However, such a detergent risks impairing the DCCD sensitivity of F 0 F 1 , probably by loosening the subunit interaction, meaning that rotations observed in the early stage are not well coupled with the H ϩ translocation in F 0 (36). Ueno et al. (30) attempted to solve this issue by using a lipid-like reagent, lysophosphatidylcholine, finding that the probability of rotating molecules was significantly decreased by DCCD-pretreatment prior to the rotation assay (41). However, DCCD inhibition during rotation observation remained elusive.
In the present study, we achieved more than 90% DCCD sensitivity of ATPase activity in solution by solubilizing F 0 F 1 in buffer containing 0.1% (w/v) E. coli lipid total extract and 0.3% (w/v) of the detergent C 12 E 8 . In this buffer condition, most F 0 F 1 molecules stopped the ATP-driven rotation after DCCD injection into the flow chamber. In addition, the time constant of the DCCD inhibition, as determined from the duration time of the rotation after DCCD injection, was consistent with that of the DCCD inhibition determined from ATP hydrolysis activity in solution. Thus, a lipid-like condition is critical for sustaining the integrity of the subunit interaction and the coupling nature of F 0 F 1 . The Stoichiometry of DCCD for Inhibition-The rate of the DCCD inhibition was linearly proportional with the DCCD concentration, suggesting that the inhibition is a first order reaction and that a single DCCD modification is sufficient for the inhibition of the F 0 F 1 coupling reaction, as reported previously (26,27,29). Interestingly, even after inhibition with DCCD, F 0 F 1 showed transient activation by making at most three turns. The angular position after the transient activation was always the same as the initial termination position, although a slight positional shift during the termination was observed in some molecules. This result also suggests that a single DCCD modification is sufficient to block rotation.
No Preference of DCCD Incorporation-The angular position of the DCCD-induced terminations was analyzed in relation with the catalytic pause of F 1 by using the mutant F 0 F 1 (␤E181D). The distribution of the position of the DCCDinhibited state showed no obvious peak between the catalytic angles of F 1 , indicating that DCCD randomly reacts with the c subunit regardless of the position of the c subunit relative to F 1 and that there is obvious preference for the DCCD reactivity. This result is consistent with a crystal structure of the F 1 -c-ring complex (Protein Data Bank code 1QO1), where the c subunit showed identical conformations (20).
Transient Activation from DCCD Inhibition-We found that DCCD-inhibited F 0 F 1 shows transient activation that can be largely enhanced when DCCD-inhibited molecules are forcibly rotated in the forward direction, but not in the backward direction. This asymmetric effect is explainable when considering a structural feature of F 0 . In the coupled rotation, H ϩ is translated between the a and c subunits of F 0 . Therefore, the c subunit involved in the H ϩ translocation should tightly interact with the a subunit, whereas other c subunits are exposed to the lipid bilayer. A crystal structure of a DCCD-modified c-ring from Spirulina platensis (Protein Data Bank codes 2XQU, 2XQS, and 2XQT) showed that covalently bound DCCD (N-acyl urea) protrudes outward from the H ϩ -binding carboxylate (24). This observation implies that once the c-ring is modified by DCCD, its rotation is locked due to the steric hindrance of the protruding DCCD and a subunit (Fig. 6A). The transient activation then might be caused by a thermally induced looseness at the a-c interface. Forcible rotation in the forward direction enhances the looseness, whereas backward rotation causes F 0 to be rotated until the DCCD clashes again with the a subunit.
The Positional Shift of the DCCD-inhibited State-Surprisingly, we found that DCCD-inhibited F 0 F 1 showed positional shifts in the termination angle (Fig. 5, A-C). Such positional shifts were not observed in other inhibitory states of F 1 such as ADP inhibition and ⑀-subunit inhibition (34,42). This difference is attributable to the inhibition target. The ADP and ⑀ subunit inhibitions directly block the rotation of F 1 by tightly binding to the catalytic site or extending the C-terminal helices of the ⑀ subunit into the stator-rotor interface of F 1 to cause steric hindrance (31,43). On the other hand, in DCCD inhibition, DCCD blocks F 0 rotation, whereas F 1 is still catalytically active. This situation is similar to a stalling experiment using magnetic tweezers where actively rotating F 1 molecules are stalled by an external torque when F 1 reaches a dwelling pause while waiting for a particular catalytic reaction step such as ATP binding or ATP cleavage (38). The probability of the catalytic reaction step was measured as a function of the rotary angle of the ␥ subunit (38). It was revealed that over a wide range of stall angles, F 1 can undergo both forward and reverse catalytic reactions. Forward reactions trigger torque generation to continue the rotation, whereas reverse reactions induce rotation back to the original dwelling position. Therefore, stalled F 1 molecules show a positional shift in the termination angle that can be determined by a balance between the torque of F 1 and an external torque generated by the magnetic tweezers. 3 The positional shift of the DCCD-inhibited state can be explained as follows. The F 0 -locked state describes mechanical stalling for F 1 . Therefore, catalysis of F 1 should be at equilibrium between the forward and reverse reactions, each accompanying the torque generation accordingly. Because F 0 and F 1 are connected via elastic stalk structures, the central rotor stalk and the peripheral stalk (3, 37, 44 -47), F 1 can change the angular position of the DCCD-induced termination; the position shifts forward when F 1 conducts a forward reaction and shifts backward when the reaction is backward. In the schematic model shown in Fig. 6B, we attribute the elasticity to the peripheral stalk that is bent and stretched upon the torsional stress by F 1 based on the studies of Cain and co-workers (48,49) that showed that the peripheral stalk subunit, b, is exceptionally robust against the deletion or insertion mutation, suggesting structural flexibility of the peripheral stalk. It should be noted that Junge and co-workers (40) reported that the peripheral stalk is more rigid than the central rotor stalk against the torsional stress. Because the peripheral stalk bears mainly bending and stretching force in circumferential direction and thus, the torsional stress is relatively minor, we prefer the elastic periph-3 M. Toei and H. Noji, unpublished data. FIGURE 6. The models of DCCD inhibition of F 0 F 1 . A, the model of the transient activation of the DCCD-inhibited F 0 when viewed from the F 1 side. The rotation of the peripheral stalk of F 0 F 1 is transiently activated and blocked again at the same angle after one rotation. B, the model of the DCCD-inhibited F 0 F 1 showing the twisted peripheral stalk. The F 1 part undergoes the ATP-driven rotation, whereas the F 0 part is locked by DCCD, which provokes the twist of the peripheral stalk. eral stalk model. However, this issue remains to be tested experimentally.
The reason why F 0 F 1 (␤E181D) showed obvious and large positional shifts is attributable to the slow kinetics and equilibrium of the catalysis. The catalytic reaction of the wild type is much faster (on a millisecond scale) than that of the mutant (ϳ200 ms) or the response time of the magnetic bead (ϳ100 ms). Therefore, it is highly likely that the relatively slow response of the magnetic tweezers obscured discrete positional shifts in wild-type F 0 F 1 .
Overall, the multiple positional shifts of the termination angles observed in DCCD-inhibited F 0 F 1 (␤E181D) implies that several catalytic reactions and rotational steps are involved in this phenomenon and that F 0 F 1 has very flexible components that can twist upon torque generation. For quantitative analysis of this phenomenon, F 0 F 1 from thermophilic bacillus PS3 (TF 0 F 1 ) would be more suitable because its step rotations and kinetics parameters have been well characterized.
Conclusions-We have demonstrated the DCCD inhibition of F 0 F 1 using single-molecule techniques. The first order reaction kinetics of the inhibition and the essentially single position of the DCCD-induced termination indicate that a single modification of the c subunit in the c-ring is sufficient for blocking the rotation. We also found spontaneous transient activation upon DCCD inhibition that could be stimulated by forcibly rotating the complex to the forward direction. Finally, we showed a positional shift during DCCD inhibition, indicating a stochastic torque by F 1 and large torsional flexibility in F 0 .