Introduction

FoF1-ATP synthase (FoF1) is a rotary motor protein that catalyzes ATP synthesis reaction from ADP and inorganic phosphate (Pi) using the proton motive force (pmf) across membranes. FoF1 is composed of two rotary motor proteins: Fo and F11,2,3,4,5. Fo is a membrane-embedded protein complex forming the proton translocation pathway. When FoF1 synthesizes ATP, Fo transports protons from the outside to the inside of the membrane, accompanied by the clockwise rotation of the rotor complex inside FoF1 when viewed from the outside of the membrane. F1 is a water-soluble portion of FoF1 that contains catalytic centers for ATP synthesis. Proton translocation by Fo and ATP synthesis/hydrolysis reactions in F1 are tightly coupled through the rotation of the rotor complex. When pmf is sufficiently high, Fo forcibly rotates F1, powered by proton translocation down pmf, resulting in ATP synthesis via F1. When pmf is insufficient, F1 reverses the rotation and hydrolyzes ATP, which induces Fo to pump protons to generate pmf 6.

F1 is an ATPase that hydrolyzes ATP to rotate its rotor part counterclockwise, when viewed from the membrane side, against the surrounding stator α3β3-ring7, where the central rotor γ subunit is accommodated in the central cavity8,9,10,11,12. F1 possesses three catalytic sites for ATP hydrolysis/synthesis, each at the interface between the α and β subunits. Amino acid residues critical for catalysis are mainly located in the β subunit. The “ground-state” crystal structure of bMF113 revealed that the three β subunits have different conformations and nucleotide-bound states; β with nucleotide analog (βTP), β with ADP (βDP), and β with none (βE). The βTP and βDP adopt a closed form with their C-terminal domain swinging towards the γ subunit, whereas βE adopts an open conformation4,12,14. The rotary dynamics of F1 have been investigated extensively using single-molecule studies1,15,16,17,18,19. The reaction scheme of rotary catalysis in bMF120 (Fig. 1a) was proposed by considering the structural features of bMF1 and kinetic analysis in single-molecule studies on F1 from thermophilic Bacillus PS3 (TF1)21. Several experiments have shown that most of the crystal structures of bMF1 represent catalytic dwell states22,23,24.

Fig. 1: Outline of this work.
figure 1

a Rotation scheme of bMF1. The circles represent the catalytic states of bound nucleotides at the β subunit. ATP* in the circles at 80°, 200°, and 320° represents the catalytically active state in which the bound ATP is to be hydrolyzed. The arrows represent the rotary angles of γ subunit. 0° is defined as the position of the γ subunit where the blue β subunit binds to ATP. Short pauses at 10°-20°, 130°-140°, and 250°-260°, observed in our previous paper20, are omitted from this figure for clarity. b Crystal structure of bMF1 with IF1 bound to the αβDP subunit (PDB: 2v7q). αDP, βDP, γ, and IF1 are colored by dark red, pink, blue, and green, respectively. The βDP subunit has been omitted for clarity in the enlarged figure. c An illustration of the single-molecule rotation assay system of bMF1. The stator α3β3-ring is immobilized on a glass surface. A magnetic bead (φ ~ 300 nm) is attached to the rotor γ subunit as a rotation probe via biotin-streptavidin interaction. Magnetic tweezers, consisting of two sets of electromagnets, were equipped onto the sample stage of the microscope.

Full size image

Most FoF1 play the primary role in ATP synthesis in vivo, although proton- or sodium-pumping activity driven by ATP hydrolysis is dominant in some species. In general, ATP hydrolysis activity of F1 can result in futile ATP consumption. Therefore, several types of regulatory systems can suppress or block the ATP hydrolysis activity of F125,26,27,28,29,30. The most common mechanism universally found in bacterial and mammalian F1 is ADP inhibition in which F1 spontaneously lapses into a resting state during rotation31,32,33. Mitochondrial F1s have a unique inhibitor protein, termed the ATPase inhibitory factor 1 (IF1)34. It is found in eukaryotic cells and is highly conserved particularly among mammalian cells25. Unlike the ε subunit9,35,36,37,38 that some of the bacterial F1s employ for inhibition, IF1 is not a built-in inhibitor, but a dissociative one that associates with F1 when pH decreases to acidic state or F1 is isolated before association with Fo to form the whole complex of FoF16. The most remarkable feature of IF1 is its condition-dependent manner of inhibition: IF1 inhibits ATP hydrolysis activity of F1 almost completely under ATP hydrolysis conditions, whereas, in ATP synthesis conditions, IF1 dissociates from F1 and does not interfere with ATP synthesis reactions after dissociation34,39. Notably, a few recent studies claim that IF1 also suppresses the rate of ATP synthesis when IF1 is overexpressed in cells40,41.

The atomic details of the interaction between IF1 and bMF1 have been investigated in structural studies42,43,44. Full-length bovine mitochondrial IF1 forms a homodimer complex via an antiparallel coiled coil45, which associates two molecules of the F1-ATPase with each N-terminal region, whereas the C-terminal region of IF1 forms a coiled-coil structure for dimerization42. Deletion of the C-terminal residues 61–84 produces a stable monomeric form of IF1, which achieves full inhibition capacity irrespective of pH change46. This simple platform of IF1 is often used in biochemical47,48 and structural studies43,44, including this study. The crystal structure of the bMF1-IF1 complex44 showed that the long α-helical structure in the middle section of IF1 was bound to the interface of the αβDP pair, whereas the short helix of the N-terminus was in contact with the γ subunit. The N-terminal short helix was linked to the middle-long helix via a distinct kink (Fig. 1b). The crystal structure of bMF1 with three IF1 units, referred to as bMF1-(IF1)3, has been resolved43, where each αβ interface was bound with IF1. Because the structures of the αβ interface are different, IF1’s show different conformational states. The structure of IF1 on αβDP is consistent with that of the 1:1 bMF1-IF1 complex, whereas the N-terminal short helix of IF1 is not resolved on αβTP. The structure of IF1 on αβE shows the second half of a long helix, and the remaining was unsolved. Based on these observations, the progressive folding of IF1 coupled with γ rotation has been proposed. Progressive folding process has also been suggested in biochemical studies48,49, although the progressive processes are simplified as a two-step process: the initial binding process and the subsequent isomerization process.

Compared to studies under hydrolysis conditions48,50,51, studies on IF1 under synthetic conditions are less advanced, despite its importance for the understanding of the condition-dependent inhibition mechanism of IF1. IF1-inhibited state of F1-ATPase is known to be so stable that F1 is unable to eject IF1 by thermal agitation alone48. A typical condition for the unlocking from IF1 inhibition is to charge sufficiently high pmf to membrane vesicles containing IF1-inhibited FoF152,53,54,55,56,57. However, the principal mechanism for IF1 ejection from the catalytic site of F1, followed by the recovery of catalysis in F1 (hereafter, “activation from IF1 inhibition” in this paper), remains elusive. Many fundamental questions are unsolved such as ‘Does reversible rotation lead to activation from IF1 inhibition?’, ‘Are there factors to enhance the activation?’, and ‘Which interactions among IF1 and F1 are responsible for IF1 inhibition and for condition-dependent inhibition?’.

This study investigates the experimental conditions required for the dissociation of IF1 from the inhibition complex using magnetic tweezers, which enable control of the angular orientation of the rotor during single-molecule rotation assays for bMF1 (Fig. 1c). Similar to the mechanical activation of F1 from the ADP-inhibited form, we forcibly rotate the rotor of IF1-inhibited F1, and define activation as the resumption of F1 molecule rotation. The activation probability is determined as a function of the angle, similar to our previous stall-and-release experiments20,21,33,58. Further, we investigate the roles of the N-terminal short helix and the middle-long helix in inhibition and activation. These results highlight the molecular mechanism of IF1 dissociation, which is critical for the unidirectional inhibition mechanism of IF1.

Results

IF1-inhibited states of bMF1

The rotation of bMF1 was monitored using magnetic beads (beads diameter, φ ~300 nm) attached to the γ subunit as a rotation probe and recorded at 30 frames per second (fps). During rotation, bMF1 showed a transient pause in the absence of IF120. This was attributed to ADP inhibition. Supplementary Fig. 1 summarizes the kinetic analyses of ADP inhibition. Although it is dependent on ATP concentrations, the mean time for ADP inhibition range between 10–30 s. For IF1 inhibition, monomeric bovine IF1 (Δ61–84) was used, as described in previous studies47,48.

The experimental procedure for the analysis and manipulation of IF1-inhibited bMF1 molecules was as follows (Fig. 2a); after identification of rotating particles in the presence of 1 mM Mg-ATP, a solution containing IF1 and Mg-ATP was gently introduced into the flow cell. After rotations for several tens of seconds, all bMF1 molecules stopped rotation without exception (Fig. 2b), and none of the observed molecules resumed rotation during the observation time of 480 s (Supplementary Fig. 2a). The pause duration was evidently longer than the duration of ADP inhibition, and thus attributed to IF1 inhibition. Under 3 μM IF1 and 1 mM ATP, the mean time to lapse into IF1 inhibition was approximately 20 s (Supplementary Fig. 2b). The time constant for IF1 inhibition obtained in our biochemical experiments at saturated ATP and IF1 concentrations (Supplementary Fig. 9, gray) was 30 s, which is approximately 1.5 times longer than that in the single-molecule rotation assay. This is probably due to differences in experimental conditions; the single-molecule rotation assay selectively analyzes actively rotating molecules, whereas biochemical measurement is based on ensemble averaging of active and ADP-inhibited molecules. The latter would require a longer time to lapse into IF1 inhibition than actively rotating molecules when ADP-inhibited form is off the pathway to IF1 inhibition.

Fig. 2: Single-molecule analysis of IF1-inhibited bMF1.
figure 2

a Experimental method. After observing a rotating bMF1 molecule in IF1-free solution, we introduced a mixed solution with IF1 into the reaction chamber. The bMF1 continued to rotate for a while but finally stopped rotation. b Typical time courses of bMF1 in the presence of 3 μM IF1 and 1 mM ATP. See Supplementary Fig. 2 for more detailed analyses. Source data are provided in the Source Data file. c Stall positions of IF1 inhibition. (Left) An example of the IF1-inhibited pauses. After observing the ATP-binding waiting dwell at 100 nM ATP, 5 μM IF1 with 100 nM ATP was introduced into the reaction chamber. Blue data points represent stalls of IF1 inhibition. (Right) The angular distance (Δθ) of IF1-inhibited states from the left-side ATP-binding waiting dwell (pink) (N = 29 pauses). Values represent mean ± SD estimated from Gaussian fitting of the plots. Source data are provided in the Source Data file.

Full size image

Next, we investigated the dwell position of bMF1 when inhibited by IF1. Experiments were performed under 100 nM ATP condition, where bMF1 showed three distinct pauses at the ATP-binding dwell angles (Fig. 2c). Transient pauses, such as ADP inhibition, are rarely observed20. When a solution containing 100 nM ATP and 5 μM IF1 was introduced into the flow chamber, bMF1 stopped rotating completely. The time constant for IF1 inhibition at 100 nM ATP was 461 s (Supplementary Fig. 3), which was longer than that obtained with 1 mM ATP, i.e., 19.6 s (Supplementary Fig. 2b). The observed ATP-dependent effect of IF1 inhibition was consistent with previous biochemical analysis48,49,59,60. Fig 2c shows a representative dataset for the angular position analysis of IF1 inhibition. The angular position of IF1 inhibition (Fig. 2c, blue) was determined using the ATP-binding dwell angles as the reference; it was defined as the angular distance from the left-side ATP-binding pause (Fig. 2c, pink). The position of IF1 inhibition was determined as 89 ± 14° from the ATP-binding angle. This value was almost identical to the pause positions of bMF1 stalled by AMP-PNP (76°)20 and sodium azide (79° in Supplementary Fig. 4), which corresponds to the position of the catalytic dwell (80°). Thus, we confirmed that IF1-inhibited bMF1 pauses at the catalytic dwell angle, as seen in human mitochondrial F119.

Activation of IF1-inhibited bMF1 via forcible rotation

In the rotation assays, IF1-inhibited bMF1 did not resume the rotations once it lapsed into IF1 inhibition. By contrast, it was reported that FoF1 is activated from IF1 inhibition when the pmf is charged on the vesicle membrane, in which the FoF1 is embedded55. To explore the crucial conditions and factors for unlocking from IF1 inhibition, IF1-inhibited bMF1 molecules were forcibly rotated using magnetic tweezers. First, we tested the clockwise rotation for one turn (Fig. 3a), considering that forcible rotation in the ATP synthesis direction for one turn is sufficient for the unlock from IF1 inhibition. Before the forcible rotation, the buffer in a flow cell was exchanged with IF1-free ATP solution (1 mM) to prevent possible rebinding of IF1. However, the molecules did not show activation in most cases (Fig. 3b). Only a small fraction of molecules (10%) resumed continuous rotation. When forcibly rotated in the counterclockwise direction, the activation probability was even lower (<4%), suggesting the rotational-direction-dependence for the activation from the IF1 inhibition. The reactivation is a unique phenomenon that was never observed without manipulation with magnetic tweezers, but the probability is too low compared to the reported pmf-induced full activation of FoF152,53,54,55,56,57.

Fig. 3: Single-molecule manipulation of IF1-inhibited bMF1.
figure 3

a Schematic images of the manipulation procedure. When bMF1 was stalled by IF1, the magnetic tweezers were turned on to stall bMF1 to rotate one clockwise or counterclockwise revolution at the rate of 0.5 revolutions per second (rps). After manipulation, released bMF1 either resumed its rotation (ON) or did not (OFF). These behaviors indicate whether or not IF1 dissociates from bMF1 under the stalling time. b Reactivation probability of IF1-inhibited bMF1 after manipulation. “Hyd” and “Syn” represent the direction of hydrolysis (counterclockwise) and synthesis (clockwise), respectively. Values represent reactivation probability (\({P}_{{ON}}\)) ± SD. \({P}_{{ON}}\) was defined as the probability of an ON event against total molecules. The SD of \({P}_{{ON}}\) is given as \(\sqrt{{P}_{{ON}}\,(100-{P}_{{ON}})/N}\), where \(N\) is the number of total molecules (N = 19-26 molecules). Source data and the exact number of molecules in each data point are provided as the Source Data file.

Full size image

Next, we tested the effect of ADP and Pi in a forcible rotation to further mimic ATP synthesis conditions. After confirming IF1 inhibition, the buffer in the flow cell was gently exchanged with IF1-free ATP synthesis buffer (100 μM ATP, 100 μM ADP, 1 mM Pi), and then, IF1-inhibited bMF1 molecules were forcibly rotated in the clockwise or counterclockwise direction for 360°. As shown in Fig. 3b, a remarkable increase in activation was observed when IF1-inhibited bMF1 was rotated in the clockwise direction in the presence of ADP and Pi; the probability of activation was about 60%. This is in contrast to the reactivation probability for the clockwise manipulation in ADP/Pi-free solution (10%). An evident rotational-direction-dependent manner was observed; the counterclockwise rotation failed the activation. These observations indicate that both directional manipulation and the presence of substrates for ATP synthesis are requisite for the efficient activation of the IF1-inhibited state. To investigate the effect of individual substrates on activation, we performed the same manipulation experiment in the presence of ADP or Pi, respectively (Supplementary Fig. 5). Pi was more effective for reactivation (28%), whereas the activation achieved only 4% in the presence of ADP. Notably, once activated, bMF1 molecules did not show long pauses that are attributable to IF1 inhibition in IF1-free solution. This observation indicates that activation is accompanied by dissociation of IF1 from F1.

Angle-dependence of IF1 dissociation

We determined the activation probability as a function of the rotation angle by performing a “stall-and-release” experiment. The experimental procedure was as follows (Fig. 4a); after confirming IF1 inhibition, the inhibited bMF1 molecules were rotated to stall at the target angle for the programmed period of 0.5–5.0 s with the magnetic tweezers. After the set time had elapsed, the magnetic tweezers were turned off to release the F1 molecule. The released molecule showed two types of behaviors (Fig. 4b): starting active rotation or returning to the initial position of IF1 inhibition to resume the pause. The former behavior was identified as activation from IF1 inhibition, and the latter as failure of activation. In contrast to the abovementioned activation by a forcible 360° rotation, we conducted a stall-and-release experiment without washing IF1 in solution. The molecules again lapsed into IF1 inhibition after activation due to the rebinding of IF1 in a solution containing 3 μM IF1. The mean rotation time until the re-inactivation of activated F1 molecules was 18.3 s (Supplementary Fig. 6), and the result is consistent with the time constant of IF1 inhibition (15.2 s), indicating that re-inactivation was due to the rebinding of IF1 from solution. We repeated the manipulation of each molecule to confirm reproducibility. After manipulation, a few molecules occasionally exhibited unusual behaviors, such as random tethered Brownian motion, and nonspecific binding to or detachment from the coverslip. These data were omitted from the analyses. A detailed description of the analysis of molecules in this experiment is presented in Supplementary Note and Supplementary Fig. 7.

Fig. 4: The stall-and-release experiment of IF1-stalled bMF1.
figure 4

a Schematic images of the manipulation procedure in the stall-and-release experiment. When bMF1 was stalled by IF1, the magnetic tweezers were turned on to stall bMF1 at the target angle. After the set time had elapsed, the magnetic tweezers were turned off, and the molecule returned to the initial angle. Released bMF1 either resumes its rotation (ON) or stays at the initial position (OFF). These behaviors indicate whether or not IF1 dissociates from bMF1 under the stalling time. b A representative time course of the stall-and-release experiment under 100 μM ATP, 100 μM ADP, and 1 mM Pi in the presence of 3 μM IF1. In this figure, the stall time for both trials (blue) was 5 s and the stall angles were -26° and -168°, respectively. c Angle dependence of reactivation probability under 100 μM ATP, 100 μM ADP, and 1 mM Pi in the presence of 3 μM IF1. Each data point was obtained from 10 to 57 trials using 4 to 14 molecules. Counterclockwise rotation (blue) and clockwise rotation (orange) are defined as positive and negative direction, respectively. The colors on the plots represent the stall time of 0.5 s (red), 2 s (gray) and 5 s (blue), respectively. Values represent reactivation probability (\({P}_{{ON}}\)) ± SD. \({P}_{{ON}}\) was defined as the probability of an ON event against total trials. The SD of \({P}_{{ON}}\) is given as \(\sqrt{{P}_{{ON}}\,(100-{P}_{{ON}})/N}\), where \(N\) is the number of total trials in each data point. Source data and the exact number of trials in each data point are provided in the Source Data file.

Full size image

Experiments were conducted with 100 μM ATP, 100 μM ADP, and 1 mM Pi. Fig 4c shows the experimental results, where the counterclockwise direction of the stall angle was defined as positive relative to the initial IF1-inhibited state. In principle, activation was rarely observed in the counterclockwise manipulation, which is consistent with the experimental results for forcible 360° rotation manipulation (Fig. 3b). For the clockwise manipulation, rotation up to 120° did not activate IF1-inhibited bMF1. When rotated over 200°, the activation probability, i.e., PON was remarkably increased. PON was 60% when stalled at −200° for 2 s, and reached over 90% when stalled at –320° for 5 s. At each stall angle, PON increased with stall times. These features are similar to those observed in our previous studies on the angle dependence of ATP binding and hydrolysis in TF158 and bMF120 as well as the angle dependence of activation from ADP inhibition32. These results on angle-dependent activation are discussed in detail in the Discussion section.

N-terminal-truncated mutants of IF1

With an aim to investigate which structural elements of IF1 are responsible for the observed rotation-direction-dependent activation, we generated IF1 mutants with N-terminal truncations and tested their inhibitory functions in biochemical and single-molecule manipulation experiments (Fig. 5). IF1 consists of the N-terminal region that is missing in the crystal structure (1–7), unstructured region (8–13), the short helix region (14–18), glycine loop (19–20), and the long helix region (21–60) (Fig. 5a, b). The long helix of IF1 mainly interacts with the βDP subunit, whereas the proximity contacts with the γ subunit are found at S11 and F22 (Fig. 5a)43. We prepared four truncation mutants of IF1: IF1(Δ1–7), IF1(Δ1–12), IF1(Δ1–19), and IF1(Δ1–22) (Fig. 5c). The inhibitory effects of the mutants were first analyzed using a solution ATPase assay (Supplementary Figs. 8 and 9). IF1(Δ1–7) and IF1(Δ1–12) showed similar inhibitory effects to IF1(WT); upon association with bMF1, the mutants of IF1 completely inhibited ATPase activity. In contrast, IF1(Δ1–19) and IF1(Δ1–22) showed a reduced inhibitory effect. At concentrations lower than 0.5 μM, IF1(Δ1–19) did not completely inhibit ATPase activity, suggesting a reversible association/dissociation. IF1(Δ1–22) showed a lower inhibitory effect with a peculiar kinetic behavior. Upon addition of IF1(Δ1–22) into the assay mixture, ATPase activity decreased, followed by slow recovery. Although the mechanisms underlying such complex behavior are unclear, these biochemical measurements showed that the inhibitory effect of IF1(Δ1–19) and IF1(Δ1–22) was less than that of IF1(Δ1–7) and IF1(Δ1–12).

Fig. 5: Analysis of mutant IF1’s with the N-terminal truncation.
figure 5

a Enlarged illustration of the interactions between IF1 and γ subunit in the bMF1-IF1 complex (PDB: 2v7q). S11 and F22 in IF1 (green) can interact with N15 and I16 in γ subunit (blue), respectively. b Details of the IF1 structure (PDB: 2v7q), where residues 8–50 are resolved. Residues 1–7 are not resolved. Residues 8-13 are resolved and form an extended structure. Residues 14–18 and 21–50 form short and long helix, respectively, linked by glycine loop in residues 19–20. c Sequence of bovine IF11–60 and definition of N-terminal truncated mutants. d, e Reactivation probability of inhibited bMF1 by mutant IF1s after d counterclockwise and e clockwise manipulation. Experiments were performed under 100 μM ATP, 100 μM ADP, and 1 mM Pi. The data for IF1(WT) are also plotted for comparison (Fig. 3b). Values represent reactivation probability (\({P}_{{ON}}\)) ± SD. \({P}_{{ON}}\) was defined as the probability of an ON event against total molecules. The SD of \({P}_{{ON}}\) is given as \(\sqrt{{P}_{{ON}}\,(100-{P}_{{ON}})/N}\), where \(N\) is the number of total molecules (N = 11–25 molecules). Source data and the exact number of molecules in each data point are provided in the Source Data file.

Full size image

We analyzed the rotation of bMF1 in the presence of truncated mutants of IF1. Similar to the biochemical experiments, we observed that IF1(Δ1–7) and IF1(Δ1–12) showed nearly the same properties as those of IF1(WT) (Supplementary Fig. 10). As shown in the time course and pause time analysis, inhibition by these mutants was essentially irreversible; IF1-inhibited bMF1 did not spontaneously resume active rotations within the observation time. In contrast, further N-terminal truncation had a significant effect on IF1 inhibition. With IF1(Δ1–19) and IF1(Δ1–22), bMF1 molecules did not completely stop the rotation, but the molecules showed frequent transitions between rotation and intervening pauses (Supplementary Fig. 10j and m). The mean duration of pauses with IF1(Δ1–19) or IF1(Δ1–22) was 250 s or 60 s, respectively. These observations show that the short helix and N-terminus of the long helix of IF1 have critical roles in the stabilization of IF1 inhibition.

Rotation-direction-dependent activation was investigated with the mutant IF1’s. Whereas the activation experiments for IF1(Δ1–7) and IF1(Δ1–12) were conducted in IF1-free solutions, those were done in the presence of IF1 for IF1(Δ1–19) and IF1(Δ1–22). This is because these mutants can dissociate from F1 during buffer exchange, hampering the activation experiment. Fig 5d and e show the activation probabilities after a forcible 360° rotation in the counterclockwise and clockwise directions, respectively. The probabilities of activation from inhibition with IF1(Δ1–7) and IF1(Δ1–12) showed similar trends to that of IF1(WT) in both directions; i.e., no activation after counterclockwise rotation, whereas clockwise rotation induced activation with a significant probability of 80%, which is higher than that for IF1(WT) at 58%. These results indicate that the unstructured N-terminal region from positions 1 to 13 of IF1 is not responsible for the principal mechanism of IF1 inhibition, although the region contributes to the structural stabilization of the bMF1-IF1 complex. Unlike the truncation of the unstructured N-terminal region, the truncation of the short helix and the subsequent N-terminal region of the long helix had the distinctive impact on rotation-direction-dependent activation. For the mutants IF1(Δ1–19) and IF1(Δ1–22), the activation probabilities after clockwise or counterclockwise rotation were both higher than those for IF1(WT), IF1(Δ1–7), and IF1(Δ1–12). In particular, a significant fraction of events showed activation after counterclockwise rotation: 13% for IF1(Δ1–19) and 80% for IF1(Δ1–22). The activation probability after clockwise rotation was also higher than that for IF1(WT), IF1(Δ1–7), and IF1(Δ1–12), and was close to 100%. These results suggest that IF1 readily dissociates from bMF1 without the short helix and the N-terminal tip of the long helix, as suggested by biochemical data. Importantly, IF1(Δ1–22) almost loses the rotation-direction-dependent feature, and the activation probabilities after clockwise or counterclockwise rotation are more than 80%. Considering that IF1 shows direct contact with the γ subunit in the truncated regions, these results suggest that the rotation-direction-dependent dissociation mechanism is based on contact with the γ subunit. One may consider that IF1(Δ1–22) dissociates from bMF1 not by forcible rotation, but by its nature of spontaneous dissociation. However, the manipulation time is 2 s which is too short for spontaneous dissociation of IF1(Δ1-22); the probability of spontaneous dissociation within 2 s is only 3% when estimated from the mean time constant of spontaneous activation, 60 s (see Supplementary Note). Thus, the observed activation with IF1(Δ1–22) results due to forcible rotation.

Discussion

In this study, we observed the activation from IF1 inhibition by forcible rotation with magnetic tweezers. Activation is accompanied by the ejection of bound IF1 from bMF1. This is supported by the following observations: re-inactivation by IF1 was observed only when the solution contained IF1. In the absence of IF1, the activated molecules did not show long pauses attributable to IF1 inhibition. Second, the mean rotation time before re-inactivation was in good agreement with that for IF1 inhibition (Supplementary Fig. 6). The analysis of activation from IF1 inhibition provides important implications for the mechanism of IF1 inhibition. bMF1 molecules during IF1 inhibition were activated when the γ subunit was forcibly rotated with magnetic tweezers for over 200° in the clockwise (synthesis) direction. The activation probability was significantly enhanced in the presence of Pi in the solution by a factor of 15, although ADP also had an impact on the activation by a factor of 2 or more (Supplementary Fig. 5). Thus, it is evident that reactivation from IF1 inhibition occurs under ATP synthesis conditions. With IF1(WT), the probability of activation reached 100% when stalled at –320° for 5 s. This is sufficiently high to explain the pmf-induced activation of the IF1-inhibited FoF152,53,54,55,56,57. Thus, the principal mechanism for the pmf-induced activation of IF1-inhibited FoF1 is the ejection of IF1 from F1 by forcible rotation powered by pmf-driven Fo motor in the FoF1 complex. The requirement of ADP and Pi for efficient activation indicates that forcible clockwise rotation should be coupled with the ATP synthesis reaction for IF1 ejection. As the presence of substrates enhances the cooperative nature of F1, IF1 may likely be ejected through a concerted conformational transition of the α and β subunits coupled with γ subunit rotation.

We observed a significant increase in the probability of activation when the IF1-inhibited bMF1 was rotated over –200° (Fig. 4c). The crystal structure of bMF1-IF1 complex show that IF1 binds to βDP which represents the +200° state from the ATP-binding state (0°) in the scheme. These results suggest that IF1 is ejected through the state transitions of the β subunit from the +200° state to the 0° state or –40° (Fig. 6a). This state transition should be coupled with the conformational transition of the β subunit from a closed to an open conformation. The closed-to-open transition accompanies the swing-out motion of the C-terminal domain of the β subunit to which IF1 binds via the long α helix. Thus, IF1 dissociates from the β subunit in an open conformation that facilitates IF1 dissociation. The closed-to-open conformational transition would destabilize the bMF1-IF1 complex by pulling the N-terminal regions of IF1 out of the γ subunit, because the C-terminal domain of the β subunit in the open conformation is apart from the axis of the γ subunit, compared with the closed conformation. This dissociation model is almost the reverse process of the IF1 association proposed for the bMF1-(IF1)3 structure, which suggests the progressive folding of IF1 coupled with the conformational transition of the β subunit from βE to βDP via βTP.

Fig. 6: The proposed mechanisms presented in this study.
figure 6

a Coupling scheme of IF1 dissociation with the rotary mechanism in bMF1. A part of the reaction scheme of rotary catalysis in bMF1 (Fig. 1a) is described with only the essential elements. The green oval represents IF1. The green line in the scheme represents the IF1 binding angle in the range of –40°-0°. The blue line represents the angular difference (0°–200°) required for the conformational transition of the blue β subunit from βE to βDP via βTP. The black line represents the angle distance for IF1 dissociation, estimated in the stall-and-release experiment. b Schematic representation of the role of IF1 (PDB: 2v7q). The N-terminal region of IF1 wraps around the γ subunit, and the long helix of IF1 interacts with the βDP subunit. In addition to the N-terminal amino acid residues, including the short helix (residues 14-18), residues 20–22 (blue) play a critical role in the rotation-direction-dependent activation. Residues starting at residue 23 (green) in the central long helix work as a prototypical inhibitor.

Full size image

Estimating the energy required for activation from IF1 inhibition is important to discuss under what conditions ATP synthase is activated in the cell. Although this study does not measure the torque of magnetic tweezers applied to F1 molecules, rough estimation is possible based on several assumptions. When bMF1 generates torque of 40 pN·nm like other F1s, the minimum value for proton motive force as voltage to reverse F1 is 196 mV, considering the proton stoichiometry of 8 protons/turn. In addition, the minimum energy for IF1 ejection is assumed to be 140 pN·nm (84 kJ/mol).

The above discussion indicates the reversible association/dissociation processes of IF1. This can be viewed as an inevitable property for IF1 to avoid being Maxwell’s demon that violates the second law of thermodynamics by allowing only one direction of motion in a microscopic system. If IF1 can dissociate from F1 without coupling with clockwise rotation, while IF1 association is tightly coupled with counterclockwise rotation, the second law of thermodynamics will be violated as Maxwell’s demon. This is evident when one considers the rotary motion of FoF1 under conditions where the torques of F1 and Fo are balanced. In IF1-free conditions, the rotor complex should show non-biased Brownian motion in both directions: clockwise and counterclockwise, in the FoF1 complex. Even though the rotor can cause occasional rotary steps in both directions, the mean rotational displacement does not increase. In contrast, in the presence of IF1, F1 should make a +200° rotation coupled upon the association of IF1. This is because ‘the tight coupling of IF1 association and the rotation’ means that IF1 association induces +200° rotation. Although the pause by IF1 inhibition should be long, IF1 should eventually dissociates from F1. Along the above assumption, IF1 is dissociated without biasing rotation. Thus, each cycle of association and dissociation of IF1 should cause the rotation by +200° in hydrolysis direction. After multiple events of IF1 association and dissociation, F1 undergoes unidirectionally biased rotation in hydrolysis direction. Thus, the assumption of the asymmetric coupling model should violate the second law of thermodynamics, and the reversibility of IF1 association/dissociation seems inevitable. For the same reason, the requirement for ADP and Pi for IF1 dissociation is also inevitable, considering that IF1 association is coupled with ATP hydrolysis.

This consideration would be applicable to other regulatory systems of FoF1. In addition to IF1, several types of inhibitory mechanisms can block the ATP hydrolysis. The ζ subunit of α-proteobacteria inhibits hydrolysis by binding to the interface of the αβDP pair in a manner similar to IF161,62,63. Mycobacteria have a specific insertion sequence at the C-terminal of the α subunit called the α-extension loop64,65. Recent cryo-EM analysis showed that the α-extension loop binds to a specific loop on the γ subunit, which is considered to block rotation in a counterclockwise direction26,66. When these inhibitory processes are coupled with the rotation of the γ subunit, the dissociation of the ζ subunit or the α-extension loop has to be coupled with rotation in the opposite direction, according to the above contention.

All the experiments described in the main text were performed at room temperature; 23 ± 2 °C for single-molecule experiments and 25 °C for solution experiments. In order to confirm that IF1 kinetics at the physiological temperature of bovine mitochondria (i.e., 37 °C) is essentially the same as that at room temperature, the biochemical assay of IF1(WT), IF1(Δ1-19) and IF1(Δ1-22) at 37 °C was performed (Supplementary Figs. 13 and 14). As a result, the trends for the residual ATPase activity at 37 °C were principally identical to that at 25 °C as described in Supplementary Fig. 8: IF1(WT) showed complete inhibition with almost no residual activity, whereas IF1(Δ1-19) and IF1(Δ1-22) did not completely inhibit ATPase activity. These results suggest that the principle of IF1 inhibition and dissociation does not differ between room temperature to physiological temperature.

Structural analyses of the bMF1-IF1 complex showed that IF1 has interactions with bMF1 via two parts: an N-terminal region with an unstructured loop and a short helix (1–20), and a central long helix (21–60) (Fig. 5a, b). The major interaction between IF1 and F1 is formed by the central long helix, which is accommodated on the C-terminal domain of the β subunit. The N-terminal region of IF1 wraps around the γ subunit, with evident contacts at S11 and F22 of IF1. Accordingly, two scenarios for the principal mechanism of IF1 inhibition are possible42. First is the mechanical hindrance of the γ subunit rotation by the N-terminal region of IF1. The second model assumes the prevention of β conformational transitions by the central long helix of IF1. The investigation of IF1 mutants with N-terminal truncation provides important clues to this question. We observed that IF1(Δ1–19) and IF1(Δ1–22) maintained inhibitory potency to halt the catalysis and rotation of bMF1, although these mutants were remarkably less effective than IF1(WT) or other mutants. IF1(Δ1–22) was truncated before F22, losing all residues that were in contact or close proximity to the γ subunit. Thus, our study suggests that the interaction with the γ subunit is dispensable, supporting the latter model. Our result was also supported by the research using F1 and IF1 from yeast mitochondria: deletion of all the residues preceding F17 in yeast IF1, corresponding to F22 in bovine IF1, still maintained the inhibitory capacity67. This result implies a universal IF1 inhibition mechanism beyond the species level.

In the atomic structures of bMF1-IF1 and bMF1-(IF1)3, hydrophobic residues in the central long helix, including Y33, form strong interactions with the C-terminal domain of the β subunit, providing most of the binding energy. The binding of IF1 to F1 is further augmented by salt-bridge formation between E30 in IF1 and R408 in the βDP subunit. The contribution of these residues was experimentally confirmed in mutagenetic approaches47,48. A possible molecular mechanism for IF1 inhibition is that the tight binding of IF1 with the β subunit prevents conformational transition required for rotary catalysis. There are several highly conserved charged residues in the C-terminus of IF1, and the roles of these residues have not been clarified. As a preliminary trial, we conducted molecular dynamics simulation68 where the γ subunit is forcibly rotated in synthesis direction. The interaction between the charged residues of the γ subunit and negative charge of IF1 was suggested, which is reminiscent of the ionic track69. Based on this observation, we tested the role of these conserved charged residues by selectively substituting five residues with alanine (Supplementary Figs. 11 and 12). Except for a slight decrease in binding affinity to bMF1, no clear differences were found between IF1(WT) and the alanine-substitution mutant in the IF1 inhibition assay as well as in manipulation assay. Our experiments suggest that C-terminal charged residues have little impact on the inhibitory mechanism and the rotation-direction-dependent dissociation.

Analyses of N-terminal truncation mutants revealed the molecular mechanism of rotary-direction-dependent dissociation of IF1 from F1 (Fig. 6b). IF1(Δ1–22), which loses residues in contact with the γ subunit, showed a significantly high activation probability after the forcible rotation of 360°. The important finding in this mutant is that both of the clockwise and counterclockwise rotation resulted in the efficient activation from IF1 inhibition with almost equal probability (Fig. 5d, e). This is in contrast to the asymmetric features of IF1(WT), IF1(Δ1–7), and IF1(Δ1–12), which showed activation only via clockwise rotation. In the case of IF1(Δ1–19), the asymmetric effect of forcible rotation was retained. However, a higher activation probability is observed. These observations indicate that the N-terminal region of the central long helix plays a crucial role in rotary-direction-dependent activation, and the N-terminal short helix also contributes to this result. The truncated residues in IF1(Δ1–22), in addition to the 1–19 truncation, are G20, A21, and F22. Among them, F22 is bulky compared with the other residues. Further, in the crystal structures of the bMF1-IF1 complex, F22 is in direct contact with I16 of the γ subunit. Considering these points, F22 is likely one of the most critical residues for rotary-direction-dependent activation. Molecular dynamics simulations of the bMF1-IF1 complex would provide more detailed information on the molecular mechanism of rotation-direction-dependent activation.

Methods

Purification of bMF1 and IF1

F1-ATPase from bovine mitochondria with two mutated cysteine residues on A99 and S191 at the γ subunit and nine histidine residues (His-tag) at the N-terminus of the β subunit (hereafter referred to as bMF1) was purified as described previously20. IF11–60 with the linker and mScarlet fused to the C-terminus (referred to as IF1(WT)) was purified as described previously48. The N-terminal truncation mutants and the C-terminal alanine substitution mutant of IF1 were prepared as follows. Fragments of the N-terminal truncated mutants, deleted in the region encoding the corresponding N-terminal region, were generated by PCR using sets of mutation primers. The sequence encoding the C-terminal alanine mutant was amplified by PCR using primers containing mutations. The sequences of the PCR primers are provided in the Source Data file. The resulting PCR products were subjected to gel electrophoresis and purification. The insert plasmid and the vector IF1(WT) plasmid were digested with the same restriction enzymes. The products were ligated and introduced into Escherichia coli JM109. The sequence of the recombinant IF1 plasmid was confirmed using the Fasmac sequencing service (Fasmac, Japan). The purified mutant proteins were confirmed by SDS-PAGE and MALDI-TOF/TOF mass spectrometry (Genomine. Inc., Korea or IDEA Consultants, Inc., Japan) (Supplementary Fig. 15 and Supplementary Table 1).

Solution experiment

The ATPase activity of bMF1 was monitored as the rate of NADH oxidation48. The basal buffer contained 50 mM HEPES-KOH (pH 7.5), 50 mM KCl, 2 mM MgCl2, and an ATP-regenerating system (0.2 mg/mL pyruvate kinase and 2.5 mM phosphoenolpyruvate) supplemented with 0.2 mM NADH and 50 μg/mL lactate dehydrogenase, as described previously48. Experiments were performed at 25°C, unless otherwise indicated, using a V-660 (JASCO, Tokyo, Japan) UV/VIS spectrophotometer equipped with a peltier-type temperature controller (ETCS-761). ATP hydrolysis by bMF1 was initiated by adding purified bMF1 to the basal buffer containing 1 mM ATP. For the reaction to reach a steady state, we waited for 180 s before injecting IF1 into the reaction mixture. After the IF1 injection, the rate of ATPase activity changed. Inhibition by IF1 was quantified by estimating the apparent rate constants for IF1 (\({k}_{{inhibition}}^{{app}}\)), which were determined by fitting the decay using the following equation:

$$y\left(t\right)-{y}_{0}={V}_{{{\infty }}}t+\frac{{V}_{0}-{V}_{{{\!\infty }}}}{{k}_{{inhibition}}^{{app}}}\left\{1-{{\exp }}\left(-{k}_{{inhibition}}^{{app}}t\right)\right\}$$
(1)

where \(y\left(t\right)\) and \({y}_{0}\) are the absorbance values at the time t and 0 after IF1 injection into the solution cuvette, respectively, and \({V}_{0}\) and \({V}_{\infty }\) are the initial and final reaction rates, respectively. Fitting was performed using the time course of NADH absorbance at 2 s after IF1 addition. Because no exponential decay was observed in the time course of IF1(Δ1–19) and IF1(Δ1–22) inhibition (Supplementary Figs. 8g, 13c, e), curve fitting using Eq. (1) was not performed. Residual ATPase activity was measured by comparing the activity just before IF1 addition and 350 s after IF1 addition. In Supplementary Figs. 8 and 11, values were plotted over intermediate or higher IF1 concentrations (\({K}_{M}^{I{F}_{1}}\)) to see equilibrium at saturating conditions.

The sequential IF1 inhibition was analyzed using the following reaction scheme:

$${F}_{1}+I{F}_{1}\,\begin{array}{c}\mathop{\to }\limits^{\,{k}_{{on}}}\\ \mathop{\leftarrow }\limits_{{k}_{{off}}}\end{array}\,{F}_{1}\cdot I{F}_{1}\,\mathop{\to }\limits^{{k}_{{lock}}}\,{F}_{1}\cdot I{F}_{1}^{{lock}}$$

where \({F}_{1}\cdot I{F}_{1}^{{lock}}\) represents the inactive state, and \({F}_{1}\cdot I{F}_{1}\) is the intermediate state that is still catalytically active. \({k}_{{on}}\) and \({k}_{{off}}\) represent the rate constants of association and dissociation, respectively. \({k}_{{lock}}\) is the rate constant of isomerization to the locked state. According to this scheme, \({k}_{{inhibition}}^{{app}}\) can be expressed as follows:

$${k}_{{inhibition}}^{{app}}=\frac{{k}_{{lock}}\left[I{F}_{1}\right]}{{K}_{M}^{I{F}_{1}}+\left[I{F}_{1}\right]}$$
(2)
$${K}_{M}^{I{F}_{1}}\equiv \frac{{k}_{{off}}+{k}_{{lock}}}{{k}_{{on}}}$$
(3)

which shows a typical hyperbolic relationship.

Rotation assay of bMF1

For the rotation assay, bMF1 was immobilized on the glass surface, and magnetic beads were attached to the two cysteine residues of the γ subunit (γA99C and γS191C) through a biotin-avidin interaction. The protocol was as follows20. A flow chamber (~5 μL) was prepared with two coverslips (18 × 18 mm2 at the top and 24 × 32 mm2 at the bottom; Matsunami Glass) with double-sided tape (7602 #25, Teraoka) as a spacer. Purified bMF1 (~500 pM) in observation buffer (50 mM HEPES-KOH (pH 7.5), 50 mM KCl, and 2 mM MgCl2) was gently introduced into the flow chamber and incubated for 5–10 min. After washing unbound bMF1 with more than 50 μL of observation buffer containing ~10 mg/mL BSA, streptavidin-coated magnetic beads (GE Healthcare), which were pre-centrifuged to allow the use of relatively small beads (~300 nm), were introduced into the flow chamber. Unbound beads were washed with more than 50 μL of observation buffer containing the prescribed concentrations of ATP and/or ADP/Pi. Rotations of the magnetic beads were observed using a phase-contrast microscope (IX-70, Olympus) with a 100× objective lens at a recording rate of 30 fps. The rotation assay was performed at 23 ± 2 °C. The ATP-regenerating system (0.2 mg/mL pyruvate kinase and 2.5 mM phosphoenolpyruvate) was added to the solution mixture, except when ADP was used.

To observe the IF1 inhibitory state, we identified rotating molecules among the particles attached to the coverslip in the IF1-free solution. Next, a solution containing IF1 in ATP solution or ATP synthesis buffer (100 μM ATP, 100 μM ADP, and 1 mM Pi) was infused into the flow chamber. After rotations for several tens of seconds, all bMF1 molecules stopped rotation (Fig. 2a).

Manipulation with magnetic tweezers

Magnetic tweezers, consisting of two sets of electromagnets, were equipped onto the microscope stage and controlled by custom-made software20,58. Rotations were imaged at 30 fps using a progressive-scan camera (FC300M, Takex, Kyoto, Japan), which allowed real-time manipulation with magnetic tweezers. Movies were stored on a computer as AVI files and analyzed using custom-made software.

Kinetic analysis of ADP inhibition

Rotations in the IF1-free solution were recorded for more than 10 min per molecule (Supplementary Fig. 1). Under these conditions, ATP binding occurs for less than 1 ms and cannot be detected as a pause of rotations with magnetic beads at a recording rate of 30 fps. However, bMF1 showed frequent transient pauses; we collected the data from all pauses longer than 1 s31, and rotary traces between pauses longer than 1 s were defined as rotations. Distributions of pausing times before the onset of rotations were fitted by a double exponential function, \(y={N}_{{sp}}{{\exp }}(-t/{\tau }_{{sp}})+{N}_{{lp}}{{\exp }}(-t/{\tau }_{{lp}})\), where \({N}_{{sp}}\) and \({N}_{{lp}}\) are constants, and \({\tau }_{{sp}}\) and \({\tau }_{{lp}}\) are time constants for the short and long pause, respectively. A relatively short pause (sp) corresponds to decelerated Pi release, as previously reported31,70. The long pause (lp) corresponds to the ADP-inhibited form. These results showed that the time scale for ADP inhibition was 10-30 s under the various conditions tested. Distributions of rotating times before lapsing into the pauses were fitted by a single exponential function, \(y={N}_{{rot}}{{\exp }}\left(-t/{\tau }_{{rot}}\right)\), where \({N}_{{rot}}\) is a constant and \({\tau }_{{rot}}\) is a time constant for rotations. In Supplementary Fig. 1, the histograms include most of the total data; the remaining data are not shown in the figure for clarity. All data are provided in the Source Data.

Analysis of IF1 inhibition

The experimental procedure was described in the main text (see also Fig. 2a). Rotating time (Supplementary Figs. 2, 3, 6, 10, and 12) was defined as the period of time from the completion of solution exchange until when the molecules fell into IF1 inhibition. Probability plots versus rotating time were fitted with a single-exponential decay function to estimate the mean rotation time, \({\tau }_{{rot}}^{I{F}_{1}}\). The figures include most of the total data; the remaining data are not shown in the figure for clarity. All data are provided in the Source Data. A pause was defined as a pause state for more than 1 s. Except for IF1(Δ1–19) and IF1(Δ1–22), most of the pauses showed extremely long pausing states of more than 480 s, corresponding to IF1 inhibition. A few traces showed relatively short pausing states of ~20 s. They were attributable to ADP inhibition because the timescales of these pauses were mostly identical to the values estimated from the IF1-free solution experiment (see Supplementary Fig. 1). All of them spontaneously recovered active rotation and eventually fell into IF1 inhibition without exception. For IF1(Δ1–19) and IF1(Δ1–22), rotations were frequently recovered without manipulation, suggesting a reversible inhibition/dissociation mechanism. The histogram of pausing time was fitted by single-exponential decay function for IF1(Δ1–19) to estimate the mean duration time, \({\tau }_{{pause}}^{I{F}_{1}}\), 254 s. For IF1(Δ1–22) analysis, the double-exponential decay function was applied to fit the histogram, estimating \({\tau }_{{lp}}^{I{F}_{1}}\) and \({\tau }_{{sp}}^{I{F}_{1}}\). Although the underlying molecular mechanism corresponding to \({\tau }_{{sp}}^{I{F}_{1}}\) is unclear, we selected \({\tau }_{{lp}}^{I{F}_{1}}\) for IF1(Δ1–22) inhibitory states.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.