Rotation scheme of V₁-motor is different from that of F₁-motor

Abstract

V₁, a water-soluble portion of vacuole-type ATPase (V-ATPase), is an ATP-driven rotary motor, similar to F₁-ATPase. Hydrolysis of ATP is coupled to unidirectional rotation of the central rotor D and F subunits relative to the A₃B₃ cylinder. In this study, we analyzed the rotation kinetics of V₁ in detail. At low ATP concentrations, the D subunit rotated stepwise, pausing every 120°. The dwell time between steps revealed that V₁ consumes one ATP per 120° step. V₁ generated torque of ≈35 pN nm, slightly lower than the ≈46 pN nm measured for F₁. Noticeably, the angles for both ATP cleavage and binding were apparently the same in V₁, in sharp contrast to F₁, which cleaves ATP at 80° posterior to the binding of ATP. Thus, the mechanochemical cycle of V₁ has marked differences to that of F₁.

The F₀F₁-ATP synthase (F)- and vacuole-type (V-type) ATPase/ATP synthase-superfamily members utilize a rotary mechanism to perform their specific function (1–3). The cytoplasmic portion of F- and V-type ATPases (called F₁ and V₁, respectively), responsible for ATP hydrolysis/synthesis, is connected via the central rotor stalk and the peripheral stator stalk to the transmembrane portion (F₀ and V₀), housing the ion-transporting pathway. The rotation of the central rotor subunits is a key feature of these enzymes in coupling ATP hydrolysis/synthesis with ion transport across the membrane. Isolated F₁ is, itself, a well characterized ATP-driven rotary motor that converts a chemical reaction (ATP hydrolysis) into mechanical rotation of the central-shaft subunits (1). Bacterial F₁ is composed of five different subunits with a composition of α₃β₃γδε. The catalytic site of F₁ is formed at the interface between the α and β subunits, with the majority of the catalytic residues residing in the β subunit. The minimum rotary unit of F₁ is the α₃β₃γ subcomplex, in which the central γ subunit rotates within the α₃β₃ core by repeating a sequence of (at least) four processes: (i) binding of ATP to a catalytic β subunit, (ii) 80° rotation of the γ subunit, (iii) cleavage of ATP and/or release of hydrolysis product(s), and (iv) 40° rotation of γ (4–7). It is notable that F₁ is a reversible motor. When the γ subunit of F₁ is rotated in the reverse direction by external force, ATP is synthesized from ADP and phosphate (8, 9). It is believed that proton flow through F₀ drives the rotation of the proteolipid ring together with the γ subunit in the reverse direction to synthesize ATP in the cell. Recently, proton-powered reverse rotation of the γ subunit was observed for F₀F₁ incorporated into liposomes (10).

V-ATPases exist in the endomembranes of all eukaryotic cells and in the plasma membrane of some specific eukaryotic cells and are responsible for a variety of cellular functions (11). The enzyme has also been found in the plasma membrane of some bacteria (12–14). It is thought that V-ATPase is evolutionarily related to F-ATPase, because several subunits share some similarity with F-ATPase subunits. For example, the A and B subunits of V-ATPase have 20–25% identity to β- and α subunits of F-ATPase, respectively. However, there are also apparent differences between V- and F-ATPases. Electron microscopic studies have shown that the peripheral stalk of V-ATPase has a much more complicated structure (15, 16) compared with that of F-ATPase (17). In addition, it has been suggested that a proteolipid ring of V₀ is capped by a funnel-shaped d subunit (C subunit in bacteria) that does not exist in F-ATPase (18). The central stalk in the V₁ portion is constituted of the D and F subunits, which rotate relative to A₃B₃ (2). Although the D and F subunits have been considered functional homologues of the γ and ε subunits of F₁, respectively, recent work has shown that the F subunit is very different from the ε subunit. First, binding of the F subunit to the A₃B₃D subcomplex increases ATPase activity, whereas the ε subunit functions as an intrinsic inhibitor of F₁ (19). Second, the crystal structure of the F subunit, which has a Rossman fold, is completely different from that of the ε subunit, which has a β sandwich domain and a helical domain (20). Also the D and F subunits do not show apparent sequence similarity to the γ and ε subunits, respectively. These differences between V- and F-ATPases raise the possibility that the rotation mechanism of V₁ is also different from that of F₁. In contrast to F₁, the rotation mechanism of V₁ is poorly understood. Here, we report the detection of steps in the rotation of V₁. Analysis of the rotational steps of V₁ revealed the previously uncharacterized features in rotation of V₁.

Materials and Methods

Protein Preparation. The V₁ complex of Thermus thermophilus was expressed in Escherichia coli and purified as described in ref. 19. We used a mutant V₁ (A_{(His-10/C28S/S232A/T235S/C508S)3}B_(C264S)3D_(E48C/Q55C)F) throughout this study, unless otherwise noted. Purified V₁ was incubated with 10 mM DTT at room temperature for 15 min, followed by application to a gel-filtration column (Superdex HR200, Amersham Pharmacia Bioscience) equilibrated with 20 mM Mops-NaOH, pH 7.0, and 100 mM NaCl. Cysteine residues located in the D subunit of the eluted V₁ were labeled with a two-molar excess of (+)-biotinyl-3-maleimidopropionamidyl-3,6-dioxaoctanediamine (Maleimide-PEO₂-biotin, Pierce) at room temperature for 1 h. Nonreacted biotin reagent was removed with a NAP10 desalting column (Amersham Pharmacia Bioscience). Labeled V₁ was flash-frozen in liquid N₂ and stored at –80°C until use. The α₃β₃γ subcomplex of F₁ was purified and labeled with biotin, as described in ref. 6.

Biochemical Assays. The protein concentration of the V₁ complex was determined from UV absorbance calibrated by quantitative amino acid analysis; 1 μM gives 0.36 OD at 280 nm. We measured ATPase activity of V₁ complex by adding the enzyme solution to 2 ml of assay mixture consisting of 50 mM Tris·HCl, pH 8.0, 100 mM KCl, 2 mM MgCl₂, 2 mM phosphoenol pyruvate, 0.2 mg/ml NADH, 0.1 mg/ml pyruvate kinase, 0.1 mg/ml lactate dehydrogenase, and a range of concentrations of MgATP. The rate of ATP hydrolysis was monitored continuously as the rate of oxidation of the NADH, determined by the absorbance decrease at 340 nm.

Rotation Assay. A flow cell (5–10 μl) was made of two coverslips (bottom, 24 × 36 mm² and top, 18 × 18 mm²) separated by two spacers of 50-μm thickness. The glass surface of the bottom coverslip was coated with Ni-NTA (21). The biotinylated V₁ or F₁ in buffer A (50 mM Tris·HCl, pH 8.0, 100 mM KCl, and 2 mM MgCl₂) was applied to the flow cell, followed by incubation for 5 min. Unbound V₁ or F₁ was washed out with 20 μl of buffer A containing 0.5% BSA. The suspension (10 μl) of 0.02% (wt/vol) polystyrene carboxylate beads (φ = 209 or 340 nm, Polysciences), coated with streptavidin, as described in the manufacturer's instructions, suspended in buffer A containing 0.5% BSA, was infused into the flow cell, followed by incubation for 15 min. Unbound beads were removed by washing with 40 μl of buffer A. Finally, observation of rotation was initiated after infusion of 40 μl of buffer A supplemented with a range of concentrations of MgATP and an ATP-regenerating system (0.1 mg/ml pyruvate kinase and 2 mM phosphoenol pyruvate). For ATPγS-driven rotation, buffer A supplemented with a range of MgATPγS and an ADP-quenching system (0.1 mg/ml ADP-dependent glucokinase from Pyrococcus furiosus and 10 mM glucose) was used. Basically, rotation of the bead was observed with a phase-contrast microscope (IX70, Olympus) at ×1,000 magnification, and images were recorded with a charge-coupled device camera (300-RCX, Dage–MTI, Michigan City, IN) at 30 frames per second (fps). For rapid recording, we acquired images of the rotating bead with a dark-field microscope (IX70, Olympus) equipped with a mercury lamp and with a complementary metal oxide semiconductor (CMOS) camera (Hi-DcamII, NAC Image Technology, Tokyo) at 1,000 fps. Analysis of rotation was performed by using custom software, as described in refs. 4 and 5. Basically, time-averaged rotation speed was calculated over >10 consecutive revolutions. One exception is the data at 200 μM ATPγS, which was calculated over >5 consecutive revolutions, because V₁ rarely turns more than 10 revolutions under this condition.

Results

Visualization of Rotation. To investigate the rotation mechanism, we used the V₁ complex of T. thermophilus H⁺-ATPase, a bacterial homologue of eukaryotic V-ATPases. The isolated V₁ of T. thermophilus is capable of hydrolyzing MgATP in contrast to the eukaryotic V₁, which is inactivated when detached from the V₀ part, probably by an action of a regulatory H subunit (22) not present in prokaryotic V-ATPases. The T. thermophilus V₁ is, itself, an ATP-driven rotary motor with a presumed subunit composition of A₃B₃D₁F₁, in which a rotor shaft composed of the D and F subunits rotates relative to the A₃B₃ hexamer (2). To visualize the rotation of the D subunit, V₁ was immobilized onto a Ni-NTA-coated glass surface by a His-10-tag introduced to the N terminus of the A subunit, and a duplex of streptavidin-coated beads was attached to the biotin-labeled D subunit (see Fig. 5, which is published as supporting information on the PNAS web site). The motion of the beads was observed under a microscope. In the absence of ATP, the beads exhibited either no movement or Brownian motion. In contrast, in the presence of ATP, some beads showed unidirectional rotation. The rotation was always counterclockwise, as viewed from the membrane side. The time-averaged speed of rotation depended on ATP concentrations and obeyed simple Michaelis–Menten kinetics (Fig. 1a). At saturating ATP concentrations, the rotation was almost continuous, and no obvious steps in the rotation were observed.

Fig. 1.

ATP-driven rotation of V₁. Rotation was visualized under a microscope by attaching a duplex of 340-nm beads to the D subunit. (a) ATP dependence of rotation speed and ATPase rate. Time-averaged rotation speed of the D subunit of single-molecule V₁ (blue circle) and one-third of bulk-phase ATPase rate (red circle) are plotted against MgATP concentration. By using the dependence of the rates observed on the concentration of ATP with the Michaelis–Menten equation, bead rotation (blue dotted line) occurred with a V_max of 8.1 Hz and a K_m of 107 μM, and bulk-phase ATP hydrolysis (red dotted line) occurred with a V_max of 24.5·s^–1 and a K_m of 333 μM. (b) Stepwise rotation of the D subunit at 1 μM MgATP recorded at 30 fps. (Insets b and c) The centroid of the rotating bead. (c) Stepwise rotation of the D subunit at 4 μM MgATP recorded at 30 fps. (d) Histogram of dwell time between successive steps at 1 μM MgATP (n = 306) is fitted with a single exponential: k_on = (4.17 ± 0.13) × 10⁵ M^–1·s^–1 (mean ± SE). (e) Histogram of dwell time between successive steps at 4 μM MgATP (n = 542) is fitted with a single exponential: k_on = (3.19 ± 0.39) × 10⁵ M^–1·s^–1 (mean ± SE).

Stepwise Rotation of V₁ at Low ATP Concentrations. When a duplex of 340-nm beads was used as a probe, the rotation speed was almost proportional to the ATP concentration between 1 and 20 μM, indicating that binding of ATP to V₁ is rate-limiting under these conditions (Fig. 1a). Below 4 μM ATP, the bead attached to the D subunit rotated stepwise, pausing every 120° (Fig. 1 b and c), very similar to F₁, which also takes 120° steps at lower ATP concentrations (5). Each 120° step of F₁ is triggered by alternate binding of ATP to the three β subunits (23). One ATPase cycle of F₁ is terminated in one 120° rotation. The dwell time between successive 120° steps in V₁ at lower ATP concentrations corresponds to the time that V₁ is waiting for binding of ATP, because the binding of ATP is rate-limiting under these conditions. Fig. 1 d and e shows histograms of the dwell times at 1 and 4 μM ATP, respectively. The lengths of each dwell time were stochastically distributed. The histograms were well fitted with a single exponential equation, indicating that one, not two or three, ATP molecule binds to V₁ in one dwell. Thus, V₁ consumes one ATP per 120° rotation, like F₁. In the recently solved structure of proteolipid subunits of vacuole-type Na⁺-pumping ATPase from Enterococcus hirae, there are 10 Na⁺-binding sites, indicating that E. hirae V-ATPase transports 10 Na⁺ across the membrane per rotation (24). Assuming that H⁺-pumping V-type ATPases also have 10 H⁺-binding sites, the H⁺/ATP ratio is ≈3.3, although the number of H⁺-binding sites remains controversial. By fitting the histograms of dwell times with a single exponential equation, k_on for ATP was estimated at 1 and 4 μM ATP to be 4.2 × 10⁵ and 3.2 × 10⁵ M^–1·s^–1, respectively (Fig. 1 d and e). These values are ≈100-fold lower than those for F₁ (2.5 × 10⁷ M^–1·s^–1) (see ref. 5). One-third of the bulk-phase ATPase rate of V₁ without the bead measured in solution was roughly equal to the rotation rate at micromolar ATP (Fig. 1a), also supporting the suggestion that one ATP is consumed per 120° rotation. At higher ATP concentrations, bead rotation is much slower than that expected from bulk-phase ATPase (Fig. 1a), because rotation of the bead is impeded by viscous friction.

Torque. The maximum rotation rate of the bead is determined by viscous friction and torque generated by V₁. Conversely, torque (N) can be calculated from the frictional load (ξ) and angular velocity (ω) of the rotating bead by using the equation N = ξω. We measured the angular velocity of the single 120° step rotation at low ATP concentration, which is driven by the hydrolysis of one ATP molecule, to exclude the effect of the static state of rotation from the torque calculation (Fig. 2a). The torque produced by V₁ was estimated from this experiment to be 35 pN nm (Fig. 2b). The torque of the α₃β₃γ subcomplex of Bacillus PS3 F₁-ATPase estimated from the same experiment was 46 pN nm (Fig. 2b), consistent with the value reported in ref. 5. The estimated energy dissipated by the bead per 120° step of V₁ was 73 pN nm, ≈25% less than that of F_1, but is much larger than the work done in a single step of myosin or kinesin (25). Thus, like F₁, V₁ is a highly efficient motor.

Fig. 2.

Estimation of torque. (a) Magnification of step. To better resolve stepping rotations, rotation of a duplex of 340-nm beads was recorded with a high-speed camera at 1,000 fps. The concentration of MgATP used was 10 and 0.5 μM for V₁ and F₁, respectively. Steps distinguished from pauses by eye (red circle for step and black circle for pause) are fitted with linear segments (shown by thick blue line) to estimate the average angular velocity. (b) Torque generated in a step by V₁ and F₁. The torque in each step was calculated as ξω, where ξ is the frictional load of the bead, and ω is stepping angular velocity, estimated as in a. Frictional load ξ was calculated as described in ref. 21. Here, we assumed that the center of rotation is at the center of one of two beads, and that bead duplex is horizontal to the glass surface. A set of nine successive steps was chosen for each molecule. The box represents torque averaged over five molecules (45 steps) ±SEM.

Stepwise Rotation of V₁ by ATPγS. In general, one ATPase cycle is composed of four events: binding and cleavage of ATP, and release of ADP and phosphate. There should be four chemical transitions in each 120° rotation of both V₁ and F_1, triggered by the catalytic events (for example the transition from the ATP state to the ADP·Pi state caused by ATP cleavage). All or part of these chemical transitions should be directly related to mechanical rotation. Thus, it is necessary to distinguish the rotational positions associated with each chemical state (or the mechanical substep rotation associated with each catalytic event) for understanding the rotation scheme of these motors. However, all these chemical states, except the ATP-waiting state, are difficult to distinguish from each other during rotation, because the transitions between them are too fast to be detected under normal conditions. To prolong the pre-ATP cleavage state of V₁ sufficiently to distinguish from the other chemical states, we used as a substrate adenosine-5′-O-(3-thiotriphosphate) (ATPγS), which is a slowly hydrolyzing ATP analogue. Cleavage of ATPγS to ADP and thiophosphate by F₁, for example, requires a 60-fold longer time than that of ATP to ADP and phosphate (7). It has been confirmed, by using ATPγS, that F₁ cleaves ATP after an 80° substep rotation and before a further 40° substep rotation (7). Before and during the V₁ rotation assay, contaminant ADP in ATPγS (purchased ATPγS contained about 7% ADP) was depleted by using ADP-dependent glucokinase, which uses ADP instead of ATP to phosphorylate glucose (26, 27). It is necessary to remove this contaminant, because only small amounts of ADP inhibit V₁ activity. The ATPγS did not contain ATP when analyzed on HPLC (data not shown). When ATPγS was used as a substrate, rotation of the D subunit was also observed. The rotation was stepwise, pausing every 120°, even at high concentrations of ATPγS (Fig. 2b). If the V₁ rotated by using undetectable amounts of ATP rather than the ATPγS, then the observed pauses should correspond to the ATP-waiting state, and, thus, the rotation rate should be proportional to the concentration of ATPγS. However, the rotation speed of V₁ did not vary significantly (≈0.16 Hz) in 1–8 mM ATPγS (Fig. 2a), indicating that rotation of the D subunit was coupled to hydrolysis of ATPγS and cleavage of ATPγS is the rate-limiting step under these conditions. The rotation speed of the D subunit was decreased at 200 μM ATPγS to 0.06 Hz (Fig. 3 a and b). If V₁ utilizes the “80° and 40° scheme” of F₁, observation of 80° and 40° substeps is expected in this condition (see ref. 7). However, V₁ took only 120° steps, and no rotational substeps were observed at 200 μMATPγS (Fig. 3b). This result suggests that binding and cleavage of ATPγS occur at the same angle in V₁.

Fig. 3.

ATPγS-driven rotation of V₁. Rotation was visualized under a microscope by attaching a duplex of 209-nm beads to the D subunit and recorded at 30 fps. (a) Dependence of rotation speed on concentration of ATPγS. (b) Rotation of the D subunit over time at 4 mM (green) and 200 μM (red) ATPγS. (Inset) Position of the bead centroid. (c) Observation of a single rotating V₁ in different conditions. After recording the rotation of the D subunit at 4 mM ATPγS(Left), buffer containing 10 μM ATP was infused into the flow cell (see Movie 1, which is published as supporting information on the PNAS web site), followed by further recording of the rotation (Right). See Movie 2, which is published as supporting information on the PNAS web site). (Inset) Position of the bead centroid. (d) Angular position of ATPγS cleavage relative to that of ATP binding. Histograms of angular distribution of the D subunit at 4 mM ATPγS(Upper) and at 10 μM ATP (Lower) of a single V₁ molecule shown in c. The center of each dwell position was estimated by fitting to a Gaussian equation (red line).

Angles for ATP Binding and Cleavage. To investigate the above idea, we compared the dwell positions of single V₁ molecules at 4 mM ATPγS and those at 10 μM ATP. After recording the stepwise rotation of V₁ at 4 mM ATPγS, ATPγS was replaced by 10 μM ATP by infusing buffer containing 10 μM ATP into the flow chamber. Buffer exchange was confirmed by increases in the rotation speed. The rotation speed of V₁ at 10 μM ATP is ≈1 Hz but much lower (≈0.16 Hz) at 4 mM ATPγS (Fig. 3c). The angular distribution of the bead centroid showed three distinct peaks, corresponding to three dwell positions in the rotation (Fig. 3d). The centers of the three dwell positions were determined by fitting the histogram to a Gaussian equation. The changes in dwell angles (Δq) after buffer exchange were only 5.8 ± 5.4° (mean ± SD, n = 12). Next, we performed a similar experiment using the V₁ in which the bead was attached to the F subunit instead to the D subunit to exclude the possibility that the dwell position was affected by steric hindrances caused by attachment of the bead to the D subunit. It is thought that part of the D subunit is buried in the V₁ molecule. In contrast, the F subunit is thought to bind peripherally to the D subunit (19), and, thus, binding of the bead to the F subunit may have less effect on rotation. After exchange from 10 μ MATP to 4 m MATPγS, the dwell position did not alter significantly [Δq = 8.2 ± 8.7° (mean ± SD), n = 15] (see Fig. 6a, which is published as supporting information on the PNAS web site), suggesting that the above observation is not an artifact due to steric hindrances. In addition, the above observation is not an artifact due to the substitution of two residues (Ser-232 to Ala and Thr-235 to Ser) in the P-loop region of the A subunit, because the dwell positions in the rotation of wild-type V₁ at 4 μM ATP and 4 mM ATPγS did not change (Δq = 4.7 ± 4.2° (mean ± SD), n = 9) (Fig. 6b).

Discussion

Assuming that V₁, like F₁, cleaves ATPγS at the same angle it cleaves ATP, the results described above indicate that the binding and cleavage of ATP take place at almost the same angle. However, it is unclear whether binding or cleavage occurs first. In any event, cleavage accompanying binding must occur at a previously loaded catalytic site. It has been suggested that the high mechanical efficiency of F₁ is accounted for only when the energy that drives rotation is derived from the binding energy of ATP (28). In this model, the energy of ATP binding produces a constant torque along with a specific rotation angle, and the energy from ATP hydrolysis is used for weakening interaction between the protein and the reaction products but not for rotation. The model is consistent with Boyer's model, which suggests that, during ATP synthesis, the energy derived from proton translocation is used to release tightly bound ATP from a catalytic site (29). As described above, V₁, like F₁ has high mechanical efficiency. In addition, the torque generated by V₁ seems constant, irrespective of the rotation angle θ of the D subunit (Fig. 2a). Thus, it is most likely that binding of ATP to an empty catalytic site of V₁ drives rotation through 120° without assistance from cleavage of ATP (Fig. 4).

Fig. 4.

Possible rotation model. Rotation model for V₁ (a) and F₁ (b). Three catalytic sites are represented by blue (V₁) or green (F₁) filled circles, and the orientation of the D subunit or γ subunit is represented by arrows. Of the three catalytic sites, two are always occupied by nucleotides in this model. Other models such that three catalytic sites are occupied by one or two nucleotides (bi-site model) or by two or three nucleotides (tri-site model) might also be possible.

In the model deduced here, coupling between mechanical rotation and chemistry in V₁ is distinct from that in F₁. The 80° and 40° scheme of F₁ is not a prerequisite for ATP-driven rotary motors. In addition, energy dissipation from the bead during V₁ rotation is ≈25% less than that during F₁ rotation, as described above. It may be possible to attribute these differences between the two motors to the differences between the D subunit of V₁ and the γ subunit of F₁, at least in part, because the D and the γ subunits do not share apparent sequence similarity. Furthermore the central γ subunit of F₁ interacts with three surrounding β subunits (30), an interaction that will change upon rotation of the γ subunit. Because γ prefers 0° and 80° positions in the ATP-waiting state and the pre-ATP cleavage state, respectively, there should be tighter interactions between γ and β at these angles as compared with other positions. In contrast, interaction between the A and D subunits of V₁ seems to be the strongest at the 0° position in both ATP-waiting and pre-ATP cleavage states.

By analogy to F₁, the reverse rotation of V₁ would result in synthesis of ATP from ADP and phosphate. Lower torque generated through an ATP hydrolysis reaction by V₁ might imply that V₁ can be rotated in the reverse (synthesis) direction by an external force lower than F₁. It is an interesting problem whether V₁ can synthesize ATP as efficiently as F₁ from the point of view of the energy conversion between chemical reaction and mechanical movement.