F1-ATPase is a rotary motor protein driven by ATP hydrolysis. The rotary motion of F1-ATPase is tightly coupled to catalysis, in which the catalytic sites strictly obey the reaction sequences at the resolution of elementary reaction steps. This fine coordination of the reaction scheme is thought to be important to achieve extremely high chemomechanical coupling efficiency and reversibility, which is the prominent feature of F1-ATPase among molecular motor proteins. In this study, we intentionally change the reaction scheme by using single-molecule manipulation, and we examine the resulting effect on the rotary motion of F1-ATPase. When the sequence of the products released, that is, ADP and inorganic phosphate, is switched, we find that F1 frequently stops rotating for a long time, which corresponds to inactivation of catalysis. This inactive state presents MgADP inhibition, and thus, we find that an improper reaction sequence of F1-ATPase catalysis induces MgADP inhibition.
F1-ATPase (α3β3γδε), a catalytic sub-complex of F0F1-ATP synthase, is a rotary motor protein fuelled by ATP hydrolysis1,2,3. Three subunits (α3β3γ) function as the minimum component of the rotating system, where the α3β3 unit forms a cylindrical stator and the γ rotor subunit penetrates the centre of the cylinder4. The catalytic sites for ATP hydrolysis are located on each interface between the α and β subunits, mainly on the β subunits, and the three β subunits are always in different catalytic states4. The interconversion of the catalytic states of β subunits, which induces their conformational transition, drives the rotation of the γ subunit5,6,7.
The rotary motion of F1 can be directly visualized by optical microscopy8,9,10. The unitary step size of the rotation of F1 is 120°, and each step is coupled to a single turnover of ATP hydrolysis11. The 120° step is further divided into 80° and 40° sub-steps12,13. The 80° sub-step is triggered by ATP binding and ADP release, each of which occurs on different β subunits14. The 40° sub-step is triggered by ATP hydrolysis and release of inorganic phosphate (Pi), which also occurs on different β subunits14,15. The catalytic state of the β subunits is tightly coupled to the rotation of the γ subunit; thus, F1 can synthesize ATP from ADP and Pi when the γ subunit is forcibly rotated in the reverse direction16,17. Mechanically induced ATP synthesis is the physiological role of F1 of F0F1-ATP synthase, in which F0, a proton-driven motor protein, compels F1 to rotate in the reverse direction for ATP synthesis. This reversibility of chemomechanical coupling is the prominent feature of F1 among other motor proteins.
To achieve reversibility of chemomechanical coupling, fine coordination of the reaction sequence, which properly induces the conformational transition of β subunits, is required. To understand the fine coordination mechanism, extensive studies have been dedicated to unveiling the basic reaction scheme for the rotation and catalysis of F1, focusing on the catalytic turnover of individual β subunits (Fig. 1a). In this scheme, an individual β subunit binds to ATP when the γ subunit is oriented at a specific angle, and the binding angles for the individual β subunits differ by ±120°. Each β subunit hydrolyses the bound ATP into ADP and Pi after γ rotates another 200° from the ATP-binding angle18. In addition, the produced ADP and Pi are released from the catalytic site after additional 40° and 120° rotations, respectively14,15,19,20. When the γ subunit returns to the original angular position, the β subunit initiates the next round of catalysis by binding to a new ATP. Thus, the reaction sequences among the three β subunits are well coordinated at the resolution of elementary reaction steps, which contributes to the reversibility of the chemomechanical coupling on F1.
Then, the question arises as to what happens when the reaction sequence of F1 is changed. In the case of the hydrolysis step, although the timing of hydrolysis shifts by 80° relative to the authentic reaction angle, F1 drives the rotation without interruption, demonstrating the robustness of the chemomechanical coupling of F1 (ref. 21). In contrast, in the case of other elementary reaction steps, such as Pi release, the effect of timing on rotary motion has not been examined so far, although we established a method to change the timing of Pi release by using single-molecule manipulation15. In particular, the free energy change resulting from Pi release was relatively high among reaction steps; that is, the rotary torque was mainly generated by Pi release, and therefore, the effect of the timing of Pi release on rotary motion is expected to be larger than that of other reaction steps14,15,22. In addition, the timing of Pi release in F1 is different from that in other motor proteins fuelled by ATP hydrolysis, such as kinesin and myosin, in which Pi release occurs before ADP release23,24, and this difference is thought to be important for the reversibility of the chemomechanical coupling mechanism of F1 (ref. 15).
In this study, to further investigate the role of the finely coordinated reaction sequence of F1, we evaluate the effect of changing the timing of Pi release on the chemomechanical coupling mechanism. When Pi is released before ADP, we find that F1 frequently stops rotating for a long time, which corresponds to inactivation of catalysis. This inactive state presents MgADP inhibition, and thus, we find that an improper timing of Pi release induces MgADP inhibition.
Rotation of hybrid F1
We observed the rotation of hybrid F1 carrying one β(E190D), that is, α3β2β(E190D)γ, in the presence of 1 mM ATP. Glu190 of the β subunit, the so-called ‘general base’, is known to be one of the most important residues for hydrolysis25,26,27, and its substitution into aspartic acid causes distinctively slow hydrolysis of ATP28 and strong temperature sensitivity29. Therefore, the kinetic steps of the mutated catalytic site are easily distinguished from those of the other two sites. In the presence of 1 mM ATP, the hybrid F1 showed three distinctive dwells caused by the incorporated β(E190D); the dwell for the temperature-sensitive reaction at 0° (τ=32 ms)15,29, which is the same as the dwell at the ATP-binding angle, ATP hydrolysis at 200° (τ=318 ms), and Pi release at 320° (τ=9 ms) (cyan in Fig. 1a,b). The dwell time for Pi release at 320° was prolonged because of the viscous drag of the rotary probe on γ as previously reported30,31. Hereafter, we refer to the angle for ATP hydrolysis by β(E190D) as the catalytic angle.
Single-molecule manipulation to induce Pi release
During free rotation, one β hydrolyses ATP and another β releases the produced Pi at the catalytic angle, resulting in a rotation of the γ subunit in an anticlockwise direction (Fig. 1a). In contrast to the free rotation, we recently succeeded in inducing the release of another produced Pi from the same catalytic site used for ATP hydrolysis by arresting the γ subunit for a long time (>10 s)15,22. In this study, we used the same method to induce the release of produced Pi immediately after ATP hydrolysis. For manipulation of γ rotation, a magnetic bead was attached to the γ subunit of F1, and the α3β3 ring was immobilized on the glass surface. When the hybrid F1 showed a pause for ATP hydrolysis by β(E190D) at 200°, we turned on the magnetic tweezers within ~450 ms (Fig. 2a,b) and arrested F1 at ±10° from 200°. After a certain time period had elapsed, we turned off the magnetic tweezers and released F1 from arrest. Then, F1 roughly showed one of two behaviours without exception, as previously reported: rotating forward or restarting the hydrolysis-waiting pause (Fig. 2a). When F1 shows the former behaviour, β(E190D) has already hydrolysed ATP and exerted a torque on the magnetic beads, whereas when β(E190D) shows the latter behaviour, it has not hydrolysed ATP because F1 cannot generate torque unless β(E190D) catalyses hydrolysis. These behaviours are hereafter referred to as ‘on’ and ‘off’, respectively. Then, we measured the probability of the post-hydrolysis state as the probability of the ‘on’ case against the total number of trials, PON. In the absence of Pi in solution, PON increased toward 100% as the arrest time increased; in contrast, in the presence of 100 mM Pi, PON did not change depending on the arrest time, which suggests that PON depends on the time course of Pi release from the catalytic site (Fig. 2c).
We extensively analysed the ‘on’ case and found that it can be classified into three behaviours; (I) rotating forward to 320°, (II) rotating forward to 360°; that is, skipping the pause at 320°, and (III) rotating forward to 320° and spontaneously stopping the rotation for a long time (Fig. 2d,e). When β(E190D) shows (I), it has hydrolysed ATP during its arrest but has not released the produced Pi because the rate-limiting step at 320° is not hydrolysis by β(WT) but rather Pi release by β(E190D) (Fig. 1a). F1 cannot generate torque at 320° unless β(E190D) releases Pi. When β(E190D) shows (II), it has hydrolysed ATP and simultaneously released the produced Pi, and therefore, F1 has already exerted torque for the step from 320° to 360° driven by Pi release. Regarding the behaviour of (III), we have not identified the chemical state of β(E190D) thus far; however, the pausing angle is the same as that of the MgADP-inhibited state32. Hereafter, we characterize (III). First, we analysed the dwell time at 320° and calculated the probability of occurrence for each behaviour. As shown by the blue bars in Fig. 3a, distinctively long dwell times (longer than 100 ms), which correspond to the behaviour of (III), can be recognized. Then, we calculated the probability of occurrence of (III), that is, PIII, as the number of events corresponding to the blue bars divided by total number of events in the histograms, and found that PIII increased as the arrest time increased. As shown in Fig. 3b, we changed the time scale of the histogram to shorter bins (4 ms), and found that the histogram deviated from a single exponential decay model as the arrest time increased. For example, at 60 s of arrest, the first bin (<4 ms) contained many more events in which β(E190D) skipped the pause at 320° than expected from a single exponential decay that fit the data well after the first bin with a time constant of 9 ms (black line in Fig. 3b), which was consistent with that of Pi release in free rotation (rightmost panel in Fig. 3b). Therefore, the first bin deviated from single exponential decay because of the behaviour of (II); accordingly, the first bin potentially includes behaviours of both (I) and (II). To calculate the probability of occurrence of (II), that is, PII, we fit the data points of the histograms (except for the first bin) to a single exponential curve with 9 ms as the time constant of Pi release. PII was determined as the number of events that deviated from the single exponential curve in the first bin (green bars in Fig. 3b) divided by total number of events of the histogram in Fig. 3a. In addition, we also calculated the probability of occurrence of (I), that is, PI, as the number of remaining events (light yellow bars in Fig. 3b) divided by total number of events.
Kinetic analysis of Pi release and the inactive state
From the analysis of PON based on the irreversible reaction scheme (Fig. 2c) F1·ATP↔ F1·(ADP+Pi)→F1·ADP+Pi, we can calculate the probability of Pi release during arrest as PoffPi, as shown by the red line in Fig. 3c. For example, PoffPi is 37% and 60% for 30-s and 60-s arrest, respectively. Here, we compared PII and PoffPi, and found that PII was much smaller than PoffPi (Fig. 3c, green). As mentioned above, we could not identify the chemical state of (III). Thus, we suspected that (III) was caused by Pi release immediately after ATP hydrolysis and summed PII and PIII. The sum of PII and PIII was almost the same as PoffPi (Fig. 3c, orange), suggesting that (III) was caused by Pi release immediately after ATP hydrolysis. To confirm the relationship between (III) and Pi release, we conducted the arrest measurement in the presence of high [Pi] (~100 mM), where Pi did not affect the kinetics of ATP binding and produced ADP release14,33. When we added more than 10 mM Pi in solution, PIII was completely suppressed toward 0% (Fig. 4a,b), suggesting that (III) was induced after the Pi release. Then, the kinetic scheme of (III) is given below (equation (1)), where, khydATP, ksynATP, koffPi, konPi, and ka→i are the rate constants of ATP hydrolysis, synthesis, Pi release, binding and inactivation of β(E190D), respectively.
As previously reported, khydATP, ksynATP, koffPi, and konPi were given as 2.3 s−1, 0.8 s−1, 0.021 s−1 and 2 × 103 M−1 s−1, respectively15. Then, from the fitting of PIII in Fig. 4a based on the kinetic scheme as shown above, the only fitting parameter; ka→i, was determined as 0.019 s−1. To verify this kinetic scheme, we fixed the arrest time as 60 s, and we re-plotted PIII against Pi concentration. Based on this kinetic scheme with the determined ka→i, we mathematically reproduced the experimental result of PIII as shown in Fig. 4b. From these results, we confirmed that this inactive state was induced after Pi release at 200° with a rate constant of 0.019 s−1.
We analysed the dwell time for activation from this inactive state (Fig. 4c). As shown in Fig. 4c, the histogram of the dwell time fit well to a single exponential curve, and we determined the time constant of activation, τi→a, as 39 s, which was almost the same as that for activation from an MgADP-inhibited state32. In addition, the aforementioned phenomena in which the addition of a large amount of Pi prohibited F1 from lapsing into the inactive state (Fig. 4a) is a well-known feature of the MgADP-inhibited state34, suggesting that the inactive state revealed in this study corresponds to MgADP inhibition.
MgADP inhibition, a common feature of F1-ATPase and ATP synthases from various sources32,34,35,36,37, is the catalytically inactive state. Although many biochemical studies have addressed MgADP inhibition, the mechanism of MgADP inhibition has remained elusive in the context of catalysis-rotation scheme. Here, we propose a simple model for how F1 lapses into the MgADP-inhibited state during catalysis (Fig. 5). We assume that the fundamental principle of MgADP inhibition is the loss of the driving energy of rotation derived from Pi release. Our experimental data indicate that F1 can continue rotating even if it releases Pi before ADP at 200°, suggesting that F1 principally stores the driving energy derived from Pi release in conformation for 320°–360° rotation (grey solid line in Fig. 5). However, with a time constant of 0.019 s−1, F1 dissipates the stored energy during rotation before reaching 320° and lapses into MgADP inhibition at 320° (grey dash line in Fig. 5). The driving force of 320–360° rotation after inhibition is a thermal agitation, as our previous work suggests that thermally agitated rotation of γ over 30° induced activation from the MgADP-inhibited state37. Thus, the MgADP-inhibited state is so stable that it takes a long time to resume rotation, τi→a=39 s (Fig. 4c)32.
The crystal structures co-crystallized with azide, including the original structure4,38, revealed that β at 320° (βE) does not bind nucleotides or phosphate. Considering that azide is well known to stabilize the MgADP-inhibited state of F1, these structures demonstrate the structural features of the MgADP-inhibited state. Thus, it is highly probable that β in the 320° state of MgADP-inhibited F1 has no bound nucleotide or phosphate, which is consistent with our model (Fig. 5).
It should be mentioned that the Walker group has challenged our reaction scheme of catalysis and rotation for active F1, on which the present work is based, that is, where Pi is released from β in 320° state. Based on the recent crystal structure of MF1 in which β in the 320° state is bound to Mg-free ADP39, they proposed that Mg and Pi are released before ADP. Although the reason for this apparent discrepancy is not clear, it should be noted that most of the crystal structures solved so far had no bound nucleotide on β in the 320° state; the only exceptions are the above-mentioned structure39 and a structure in a different study40. In addition, there are several experimental differences between the present study and the crystal structure study such as the species from which F1 is derived and the buffer contents. A precise correlation between the crystal structure and the reaction model based on single-molecule experiments remains to be established.
In vivo, F1 forms F0F1-ATP synthase and synthesizes ATP by coupling with a rotary motion driven by F0. Previous studies revealed that MgADP inhibition suppresses ATP hydrolysis but not synthesis41. Recently, the rotary motion of the F0F1 complex coupled to the synthesis of ATP was visualized at the single-molecule level42,43,44. For further understanding of the effect of MgADP inhibition on ATP synthesis, verification by single-molecule measurement of the F0F1 complex is desirable in the future.
To visualize the rotation of hybrid F1, α3β(E190D) β2γ, the stator region (α3β(E190D)β2) was fixed on the glass surface, and a magnetic bead (φ=0.3 μm; Seradyn, USA) was attached to the rotor part (γ) as a probe for rotation and manipulation. The experimental procedure was as follows. The flow chamber was constructed from an uncoated top coverslip and a bottom coverslip whose surface was modified with Ni2+NTA. F1 solution, which was diluted in buffer A (50 mM MOPS-KOH, 50 mM KCl and 5 mM MgCl2, pH 7.0) to a final concentration of 200 pM, was infused into the flow chamber. After 5 min, unbound F1 molecules were washed out with buffer A containing 10% bovine serum albumin, and then streptavidin-coated magnetic beads in buffer A were infused. After 10 min, unbound beads were washed out with buffer A. Finally, buffer A was infused with the indicated amount of ATP or Pi. The rotating beads were observed under a phase-contrast microscope (IX-70 or IX-71; Olympus, Japan) with a × 100 objective lens. The rotation assay was performed at 23 °C.
Manipulation with magnetic tweezers
The stage of the microscope was equipped with magnetic tweezers that were controlled using custom-made software (Celery, Library, Japan). The rotary motion of the bead was recorded at 30 and 3,000 frames per second, simultaneously (FC300M, Takex, Japan; FASTCAM 1024PCI-SE, Photron, Japan). Images were stored in the hard disk drive of a computer as an AVI file and analysed using custom-made software. From the recorded images at 3,000 frames per second, we judged whether the operations precisely arrested the γ subunit at the targeted angle or not.
How to cite this article: Watanabe, R. et al. Timing of inorganic phosphate release modulates the catalytic activity of ATP-driven rotary motor protein. Nat. Commun. 5:3486 doi: 10.1038/ncomms4486 (2014).
We thank all members of Noji Laboratory for technical support. This work was supported by Grants-in-Aid for Scientific Research (No. 18074005) to H.N. and (No. 30540108) to R.W. from the Ministry of Education, Culture, Sports, Science, and Technology, Japan.