Correlation between the conformational states of F₁-ATPase as determined from its crystal structure and single-molecule rotation

Abstract

F₁-ATPase is a rotary molecular motor driven by ATP hydrolysis that rotates the γ-subunit against the α₃β₃ ring. The crystal structures of F₁, which provide the structural basis for the catalysis mechanism, have shown essentially 1 stable conformational state. In contrast, single-molecule studies have revealed that F₁ has 2 stable conformational states: ATP-binding dwell state and catalytic dwell state. Although structural and single-molecule studies are crucial for the understanding of the molecular mechanism of F₁, it remains unclear as to which catalytic state the crystal structure represents. To address this issue, we introduced cysteine residues at βE391 and γR84 of F₁ from thermophilic Bacillus PS3. In the crystal structures of the mitochondrial F₁, the corresponding residues in the ADP-bound β (β_DP) and γ were in direct contact. The βE190D mutation was additionally introduced into the β to slow ATP hydrolysis. By incorporating a single copy of the mutant β-subunit, the chimera F₁, α₃β₂β(E190D/E391C)γ(R84C), was prepared. In single-molecule rotation assay, chimera F₁ showed a catalytic dwell pause in every turn because of the slowed ATP hydrolysis of β(E190D/E391C). When the mutant β and γ were cross-linked through a disulfide bond between βE391C and γR84C, F₁ paused the rotation at the catalytic dwell angle of β(E190D/E391C), indicating that the crystal structure represents the catalytic dwell state and that β_DP is the catalytically active form. The former point was again confirmed in experiments where F₁ rotation was inhibited by adenosine-5′-(β,γ-imino)-triphosphate and/or azide, the most commonly used inhibitors for the crystallization of F₁.

The F_oF₁-ATP synthase is widely found in biological membranes such as the mitochondrial inner membrane, thylakoid membrane, and bacterial plasma membrane. It catalyzes ATP synthesis from ADP and inorganic phosphate (P_i) at the cost of the electrochemical potential of proton or sodium ion across the membrane (1–6). The synthase is composed of a water-soluble part, F_1, and a membrane-embedded part, F_o. F₁ is a rotary motor driven by ATP hydrolysis. The subunit composition of bacterial F₁ is α₃β₃γδε. The minimum complex of the rotary motor is the α₃β₃γ subcomplex that is hereinafter referred to as F₁^αβγ. Hydrolyzing ATP, F₁ rotates the γ-subunit against the α₃β₃ stator ring in a counterclockwise direction when viewed from the F_o (7). F_o is also a rotary motor that is driven by proton translocation across the membrane, down the electrochemical potential (8, 9). The subunit composition of bacterial F_o is ab₂c_10–15 in which the c_10–15 oligomer ring rotates against the ab₂ stator. These 2 rotary motors are connected through the central and peripheral stalks. Under ATP synthesis conditions where the proton electrochemical potential is dominant, F_o forcibly rotates the γ-subunit of F₁ in the reverse direction (clockwise when viewed from F_o). This reverse rotation of F₁ results in the reverse chemical reaction of ATP hydrolysis, i.e., ATP synthesis. Conversely, under ATP hydrolysis conditions, F₁ rotates γ in the anticlockwise direction with the c_10–15 ring, forcing F_o to pump protons against the proton electrochemical potential.

The atomic-level structure of F₁ was first described in 1994, using a crystal of bovine mitochondrial F₁ (MF₁) prepared in the presence of ADP, adenosine-5′-(β,γ-imino)-triphosphate (AMP-PNP), and sodium azide (NaN₃) (10). This structure, referred to as the “reference structure,” revealed that 3 α-subunits and 3 β-subunits are alternately aligned to form a hetero hexamer ring and the γ-subunit is set into the central cavity of the α₃β₃ ring. The catalytic sites are located at each αβ interface, mainly on the β-subunit. Each β is in a different conformational state depending on the bound substrate; one binds to AMP-PNP (β_TP), another to ADP (β_DP), and the third to none (β_empty). Both β_TP and β_DP are in the closed conformation where the C-terminal domain swings toward the nucleotide-binding domain to close the cleft between these domains. As a result, these β-subunits wrap the bound nucleotide tightly. In contrast, β_empty adopts an open conformation to weaken the affinity to the nucleotide. These structural features agree well with the binding change mechanism (5), which assumes that each catalytic site is in a different catalytic state and the interconversion of catalytic states drives the rotary motion of the γ-subunit, although there are some inconsistencies (11). Since this work, many crystal structures of MF₁ with different chemical inhibitors have been reported. The crystal structure that differs most from the reference structure is that which has ADP·AlF₄⁻, where β_empty binds to ADP to adopt the half-closed conformation and the γ is twisted by −20° (12). However, all other MF₁ structures have very similar conformations to the reference structure. Thus, the crystal structures of MF₁ essentially represent a certain stable conformational state of F₁ except for the ADP·AlF₄⁻-bound MF₁ structure.

In contrast, single-molecule studies on the γ rotation of F₁ have revealed that F₁ has 2 distinct stable conformations. Since the observation that F₁ performs a 120° step rotation of γ upon hydrolysis of 1 ATP, intensive attempts have been made to resolve the 120° step into multiple substeps to allow better understanding of how the elementary catalytic steps are coupled with the mechanical rotation. In many of these studies including this one, F₁^αβγ from thermophilic Bacillus PS3 (TF₁) have been used because TF₁ is stable under the harsh conditions of the single-molecule experiments and can be genetically modified. Thus far, 2 substeps have been found: the 80° and 40° substeps. High-speed imaging of the rotation (13) and a study of a mutant F₁ having a noticeably low ATP hydrolysis rate (14) revealed that the 80° substep is induced by ATP binding and that the 40° substep was initiated after hydrolysis of bound ATP. Hereinafter, the 2 conformational states before the 80° or 40° substeps are referred to as the “binding dwell state” and “catalytic dwell state,” respectively. Recent studies have suggested that ADP release and P_i release occur at the binding dwell angle and catalytic dwell angle, respectively (15, 16). More recently, a temperature-sensitive reaction was found at the binding dwell angle in a rotation assay at low temperature (16).

Thus, single-molecule studies have revealed that F₁ has 2 stable conformational states, whereas F₁ adopts essentially 1 specific stable conformation in the crystal structure. Which conformational state do the crystal structures of F₁ represent, binding dwell state, catalytic dwell state, or another new state? On this issue, there are some interesting results. The analysis of fluorescence resonance energy transfer (FRET) between fluorescent probes attached to β and γ suggests that the crystal structure probably corresponds to the catalytic dwell state (17). However, FRET measurement provides the relative distance between the fluorophores, but not the precise position in the 3D structure. Therefore, this result is not conclusive. The reference structure was recently revised to show that β_DP binds to N₃⁻ at the γ-phosphate binding position (18), implying that F₁ in the crystal structure is in the azide-stabilized form, while the feature of this structure was later revealed to be almost consistent with that of the catalytic ground state MF₁ (19). Taking into account that F₁ pauses at the catalytic angle when in the ADP-inhibited form (20), it seems plausible that the crystal structure is in the catalytic dwell state. However, there are no experimental results that clarify this point. In fact, a common view on the nature of crystal structure of the F₁ has not been established yet: some postulate it to be in the binding dwell state (21, 22) and others think it is the catalytic dwell state (23, 24).

With an aim to addressing this issue, we generated a mutant F₁^αβγ from Bacillus PS3 in which cysteine residues were introduced at βE391 and γR84, respectively (25). In the crystal structures of bovine MF₁, corresponding residues in the ADP-bound β (β_DP) and the γ are in direct contact (18, 19, 22, 26). The residue of γR84 is a part of the “ionic track” (25, 27), which is the distinctive zonation of positively-charged residues around the axis of γ. It is postulated that β bends and unbends its conformation, tracing the ionic track with the negative charges of βD394 and βE395 (for MF₁) so as to convert the bending motion of β into the rotary motion of γ (27). The importance of the ionic track is supported by the analysis of mutant F₁^αβγ in which βE391 and γR84 are substituted with cysteine (25). Thus, the direct contact between βE391 and γR84 is the representative β_DP–γ interaction in the conformational state of the crystal structure. In this study, F₁^αβγ with βE391C and γR84C mutations was investigated in a single-molecule rotation assay to observe where it pauses when the introduced cysteine residues form a disulfide bond, thereby fixing the motor in the conformational state corresponding to the conformational state of the crystal structures. In addition to the cross-link experiment, the correlation between the crystal structures and the single-molecule experiment was studied by investigating at which angle F₁^αβγ stops rotation in the presence of AMP-PNP and/or NaN₃, the most commonly used chemicals for the crystallization of F₁.

Results

A Chimera F₁^αβγ, α₃β₂β(E190D/E391C)γ(R84C).

A mutant F₁^αβγ from Bacillus PS3 was prepared in which cysteine residues were introduced at βE391 and γR84, which correspond to the βE395 and γR75 of bovine MF₁, respectively. In the crystal structure of bovine MF₁, these 2 residues are in direct contact (Fig. 1). Although the whole structure of TF₁ has not been solved, it is most likely that TF₁ structure is very similar to MF₁, and that βE391 and γR84 of TF₁ are also in a very close proximity when TF₁ takes the conformational state corresponding to the crystal structure of MF₁. This is because F₁ from other species shows structural features essentially identical to that of MF₁ even if the amino acid sequence is not highly homologous to that of MF₁, as seen in the γ subunit of F₁ from Escherichia coli (EF₁): The structure of the γ of EF₁ is very similar to that of MF₁ (28), although the sequence homology against the γ of MF₁ is low (only ≈30%) the same as the γ of TF₁. Furthermore, the sequences around βE391 and γR84 are highly conserved in many species, suggesting the high structural conservation of these regions (Figs. S1 and S2). When βE391 and γR84 of F₁^αβγ from Bacillus PS3 were replaced with cysteine residues, they efficiently (≈90%) formed a disulfide bond under oxidizing conditions within only a few minutes (25). Thus, it is reasonable to assume that TF₁ has the essentially identical structural features to MF₁. In the MF₁ crystal structure, the β that has direct contact with γR75 adopts the β_DP form. The γ-carbons of these residues that are in equivalent positions to the sulfur atoms of cysteine residue are only 5.9 Å away (Table 1), which is sufficiently close for disulfide bond formation. In contrast, the distances in the β_TP– or β_empty–γ pairs are 11 and 23 Å, respectively. Thus, the disulfide bond would be formed in the β_DP–γ pair. In the single-molecule rotation assay, γ was actually cross-linked to β only when the β was in a specific catalytic state, revealed to be the catalytic dwell state (see below). Furthermore, E190D mutation was introduced to the β-subunit in addition to E391C mutation. The mutant β(E190D/E391C)* was reconstituted with the wild-type β to build a chimera F₁^αβγ, α₃β₂β(E190D/E391C)γ(R84C), which has a single copy of β(E190D/E391C) (Fig. 2A). Here, the βE190D mutation was used as an angular position marker to determine the pause angle of cross-linked F₁^αβγ. Shimabukuro et al. (14) showed that the βE190D mutation severely slows the hydrolysis step and causes the long pause (τ = 320 ms) at the catalytic angle. When a single copy of β with E190D mutation is incorporated in F₁^αβγ, the reconstituted chimera F₁^αβγ exhibits a transient pause at the catalytic dwell angle of βE190D that is +200° from the angle at which the βE190D subunit binds to ATP (Fig. 2B) (29). Thus, a chimera F₁^αβγ with a single β(E190D) enables us to determine the pause angle of the cross-linked F₁^αβγ relative to the catalytic angle of β(E190D). It should be noted that a chimera F₁^αβγ with 1 β(E190D) was reported to exhibit another short pause at +120° from the catalytic dwell angle of β(E190D) (29). In this study, the short pause is neglected because the time constant of the short pause is too short (τ ≈15 ms) for the recording rate in this study (33 ms per frame).

Fig. 1.

Spatial positions of βE395 (light gray spheres) and γR75 (dark gray spheres) of MF₁ in the crystal structure (Protein Data Bank ID code 1e79) (26). These residues correspond to βE391 and γR84 of TF₁. The protruding part of the γ is the F_o binding side. The α-subunit is shown in blue, the β-subunit is green, and the γ-subunit is red. To show the βE395 and γR75 residues, R33, D74, D110, R113, R133, R134, and P135 residues of the γ were removed. The figure was produced with Pymol.

Table 1.

Distance between γ-carbons of βE395 and γR75 in crystal structures of MF₁

Protein Data Bank ID code	Ref.	Distance, Å
		βTP	βDP	βempty
1e79	26	11.2	6.1	25.1
1h8e	12	11.1	6.0	17.8
1w0j	22	11.2	6.0	24.8
2ck3	18	10.8	5.7	ND
2jdi	19	11.0	5.8	ND
		11.0 ± 0.2*	5.9 ± 0.2*	22.6 ± 4.1*

ND, not determined because E395 in β_empty was not visible in these structures.

*Average ± SD.

Fig. 2.

The chimera F₁^αβγ, α₃β₂β(E190D/E391C)γ(R84C). (A) Schematic image of the chimera F₁^αβγ, α₃β₂β(E190D/E391C)γ(R84C) used in the cross-link experiment. Short black prongs denote cysteine residues at βE391 and γR84 for disulfide bond formation. (B) Circular diagram of the reaction and rotation of the chimera F₁^αβγ. At 0°, the mutated β, β(E190D/E391C), binds ATP (red dotted line) and makes the long catalytic dwell at 200° (blue dotted line) casued by slowed ATP hydrolysis by the βE190D mutation (29). The wild-type β-subunits bind ATP at 120° or 240° (black dotted lines). The catalytic dwells by the wild-type β-subunits at 80° or 320° were too short (only 1 ms) to detect with the present recording system (33 ms per frame) and are not shown.

Pausing Position of the Cross-Linked F₁^αβγ.

The rotation assay of α₃β₂β(E190D/E391C)γ(R84C) was carried out under a substrate-limiting condition, at 200 nM ATP. Because the other kinds of chimera F₁^αβγ, which carry 2, 3, or 0 of β(E190D/E391C) contaminated in the rotation assay, α₃β₂β(E190D/E391C)γ(R84C) was identified based on 2 criteria. (i) The existence of obvious pauses in every turn at the 3 binding angles, 120° apart from each other. (ii) The existence of the clear pause in every turn at a catalytic angle, which divides 1 of 3 120° steps into the 80° and 40° substeps in this order (not 40° and 80° substeps) (Fig. 3 A and C) (29). In the selection based on the first criterion, >95% of molecules were omitted from the further analysis because of ambiguous stepping. Among selected molecules, 50% showed the clear pausing at a catalytic angle and were identified as the chimera, α₃β₂β(E190D/E391C)γ(R84C). The remaining were mostly α₃β₃γ(R84C), of which all β-subunits were wild type, because the other chimera F₁^αβγ molecules carrying 2 or 3 mutant β-subunits were difficult to find out because of their slow rotary motion and the reconstitution efficiency of the mutant β was lower than that of the wild type as reported (29). Fig. 3 A–F shows a single dataset of the cross-link experiment. First, the free rotation of a α₃β₂β(E190D/E391C)γ(R84C) molecule was observed to determine the angular position of the catalytic dwell of β(E190D/E391C) and the 3 binding dwell angles (Fig. 3 A and D). Then, an oxidizing buffer containing 3-carboxy-4-nitrophenyl disulfide (DTNB) was injected to the flow chamber to promote a disulfide bond formation between βE391C and γR84C. After the buffer exchange, which took several minutes, a significant fraction of the molecules (≈70%) stopped rotation (Fig. 3 B and E). The remaining would be molecules in which cysteine residue was modified with a biotin that blocks the disulfide bond formation. After observing a pause, a pausing F₁^αβγ molecule was manipulated with magnetic tweezers to confirm the β–γ cross-linkage (Fig. 4). Unlike ADP-inhibited F₁^αβγ, cross-linked F₁^αβγ never resumes rotation, even when forcibly rotated more than +80°, which is sufficient to reactivate ADP-inhibited F₁^αβγ with nearly 100% efficiency (Fig. 4A) (30). Instead, cross-linked F₁^αβγ behaved as a twisted spring: when forcibly rotated and released, it just returned to the original pausing position. The angular velocity of return was always very fast (3.8 revolutions per s) and comparable with the ATP-driven rotation velocity at high ATP concentration (≈5 revolutions per s). In rare cases, F₁^αβγ exhibited irregular behaviors such as large fluctuations at an irregular position (Fig. 4 B–D). Such irregular behaviors were observed in some cases where cross-linked F₁^αβγ was rotated more than ±120° with the magnetic tweezers. This phenomenon is probably caused by the partial unfolding of the γ- or β-subunits. When the buffer was exchanged with a reducing one, the disulfide bond was cleaved and F₁^αβγ molecules resumed active rotation (Fig. 3 C and F). Molecules that did not resume rotation or significantly changed their binding dwell angles were omitted from our data analysis. The angle differences of the cross-link angle (Fig. 3E) from the nearest binding angle on the clockwise side (Δθ₁ in Fig. 3D) or the catalytic angle of β(E190D/E391C) (Δθ₂ in Fig. 3D) were determined. These values were also determined by the comparison with the binding and catalytic dwell angles after reduction (Δθ₃ and Δθ₄ in Fig. 3F). For this analysis, the set of the experiments was repeated a total of 72 times with 36 molecules. The histograms of Δθ₁ and Δθ₂ showed a single peak and gave mean values of 82.7 ± 15.9° for Δθ₁ and 2.0 ± 11.0° for Δθ₂ (Fig. 3G). The histograms of Δθ₃ and Δθ₄ also gave essentially the same values, 81.3 ± 17.0° for Δθ₃ and 1.4 ± 13.5° for Δθ₄, showing the reproducibility of the experiment (Fig. 3H). These results show that the cross-linked F₁^αβγ pauses at the catalytic dwell angle of β(E190D/E391C). This means that the reference structure of MF₁ represents the conformational state of F₁ in the catalytic dwell state, and that β_DP conformation represents the catalytically active state that executes ATP hydrolysis. Considering that the standard deviations are much smaller than the magnitude of the 40° substep, cross-linked F₁^αβγ will not be stable at the binding angle.

Fig. 3.

β-γ cross-linking experiment in the single-molecule rotation assay. (A–C) The centroid traces of the rotation of a α₃β₂β(E190D/E391C)γ(R84C) molecule. Data are derived from 60-s recordings. (A) The rotation at 200 nM ATP with 3 binding dwells by all β-subunits and 1 catalytic dwell caused by the β(E190D/E391C). (B) Pause of rotation after cross-linking by disulfide bond between βE391C and γR84C. (C) Rotation resumed after the reduction of the disulfide bond. (Scale bars: 100 nm.) (D–F) Histograms of rotary angle of traces shown in A–C. The angular positions of the pauses were determined by fitting the data with Gaussian curves (orange, light blue, and yellow green lines in D, E, and F, respectively). Then, the angular distance of the cross-link pause (dashed line in E) from the binding dwell angle on the clockwise side (Δθ₁) or the catalytic dwell angle of βE391C (Δθ₂) of the rotation before cross-link were determined. These distances were again determined by comparison with the rotation after reduction as Δθ₃ and Δθ₄. (G and H) Histograms of the angular distances (Δθ₁ to Δθ₄). n = 72 (36 molecules). The means ± SD for Δθ₁, Δθ₂, Δθ₃, and Δθ₄ were determined by Gaussian curve fitting (red and blue lines) to be 82.7 ± 15.9°, 2.0 ± 11.0°, 81.3 ± 17.0°, and 1.4 ± 13.5°, respectively.

Fig. 4.

Manipulation of cross-linked F₁^αβγ with the magnetic tweezers. (A) Time course of the manipulation. Blue dots represent the period for the manipulation. The molecule was clamped at the indicated angle for 3 s in the forward direction and released. (B–D) F₁^αβγ was rotated at the rate of 36° per s for near or >120° in both directions. When twisted in the backward direction, F₁^αβγ showed irregular responses. Some resisted the external magnetic field, inversed the magnetic moment of the beads instantaneously (at the point indicated by * in B and D), and then rotated back to the inversed angle of the original magnetic field (B and D), or some paused at irregular positions (indicated by a mark in C) after being released from the field.

Other chimera F₁^αβγ molecules, α₃β(E190D)₂β(E391C)γ(R84C) and α₃β(E190D)β(E391C)₂γ(R84C), were also examined to confirm that γR84C can form a disulfide bond with βE391C, which does not have the βE190D mutation. The rotation of α₃β(E190D)₂β(E391C)γ(R84C) was rarely seen because of the low reconstitution efficiency of β(E190D) as described above. The cross-linked chimera paused at either of the 2 catalytic angles that did not correspond to the catalytic angle of βE190D (Fig. S3). Thus, it was verified that βE190D mutation does not affect the pause position of cross-linked F₁^αβγ.

Pause Angle of F₁^αβγ Inhibited by AMP-PNP or N₃⁻.

To crystallize F₁, chemical inhibitors are often used to stabilize it in a specific conformational state. Another complementary method to correlate the crystal structure of F₁ and the substeps found in the single-molecule rotation assay is to analyze the pausing angle of F₁^αβγ inhibited by these inhibitors in the rotation assay. AMP-PNP and N₃⁻ are the chemical inhibitors most often used for the crystallization of F₁. Therefore, AMP-PNP and N₃⁻ were used to stop the rotation of F₁^αβγ in this experiment (Fig. 5). As in the cross-linking experiment, the 120° stepping rotation of a wild-type F₁^αβγ molecule was first observed under substrate-limiting conditions (60 nM ATP) to determine the 3 binding dwell angles. Then, the buffer containing an inhibitor (1 μM AMP-PNP or 1 mM NaN₃) was introduced into the flow chamber with 60 nM ATP for the AMP-PNP inhibition or 200 μM ATP for N₃⁻ inhibition. When the motor exhibited a long pause (>3 min), it was verified that the pause was not caused by ADP inhibition by forcibly rotating the motor through +80° with the magnetic tweezers; when inhibited by AMP-PNP or N₃⁻, F₁^αβγ was never activated by this manipulation. However, when AMP-PNP- or N₃⁻-inhibited F₁^αβγ was forcibly rotated well over +80°, for example +180°, some molecules resumed rotation and stopped after a few turns. Such harsh manipulation would repel the tightly bound AMP-PNP- or N₃⁻-ADP from F₁^αβγ. After determining the pause position of inhibited F₁^αβγ, the buffer was replaced with the inhibitor-free ATP buffer to confirm the molecule was still active by observing the resumption of ATP-driven rotation. Fig. 5A shows the pause position under AMP–PNP inhibition. The red bars represent the angle distribution during the 120° stepping rotation, and blue bars represent the angle distribution when inhibited by the inhibitor mixture. The angular distance (Δθ) of the AMP-PNP pause from the nearest binding angle on the clockwise side was determined (Fig. 5B) as in the above cross-link experiment. The mean angular distance was 84.2° ± 16.3°. This position corresponds to the catalytic dwell angle, consistent with the results of the cross-link experiment. The same result was obtained in the N₃⁻ inhibition experiments, and the mean angular distance of 81.9° ± 16.3° was obtained (Fig. 5 C and D). The pause angle of F₁^αβγ inhibited by 200 μM AMP-PNP, 5 μM ADP, and 1 mM NaN₃, which mimics the crystallization buffer of the reference structure, was also examined. The mean angular distance was 80.2° ± 14.5° (Fig. 5 E and F), essentially the same as the above experiments. Thus, many lines of experiments confirmed that the crystal structure of F₁ represents the catalytic dwell state found in the rotation assay.

Fig. 5.

Pausing position of F₁^αβγ stalled by chemical inhibitors. (A) Histogram of the angular position from a single experimental dataset of the AMP-PNP inhibition experiment. After observing active rotation of a molecule at 60 nM ATP, 1 μM AMP-PNP was infused into the reaction chamber with 60 nM ATP to stop the rotation. Red bars represent the angular position of free rotation; blue bars represent AMP-PNP inhibition. (B) Histogram of the angular deviation of the position of AMP-PNP inhibition from the binding dwell angle on the clockwise side (Δθ in A). The mean was 84.2 ± 16.3°. n = 107 (15 molecules). (C and D) The angular position of azide (N₃⁻) inhibition. The rotation was inhibited by infusing 1 mM NaN₃ with 200 μM ATP. The mean value of Δθ was 81.9 ± 16.3°. n = 129 (29 molecules). (E and F) The angular position when inhibited by 200 μM AMP-PNP, 5 μM ADP, and 1 mM NaN₃. The mean value of Δθ was 80.2 ± 14.5°. n = 54 (9 molecules).

Discussion

The crystal structure of F₁ was experimentally shown to represent the conformation of the catalytic dwell state found in single-molecule rotation assay. This means each β-subunit is in the β_TP, β_DP, or β_empty conformations during the catalytic dwell state. The β-γ cross-link using a chimera F₁^αβγ carrying 1 copy of β(E190D/E391C) showed that the catalytic angle of β(E190D/E391C) coincides with the pause position of F₁^αβγ where β_DP is cross-linked with γ. This finding means that the β_DP corresponds to the conformational state of the β that executes ATP hydrolysis reaction, consistent with theoretical works on quantum mechanics and molecular mechanics (31) or free-energy difference simulations (32). The crystal structure of BeF₃⁻-F₁, which is thought to mimic the catalytic intermediate state, also supports this result (22).

Fig. 6 shows the present reaction scheme of F_1, in which there are 2 chemical states during the catalytic dwell: prehydrolysis (state B) and posthydrolysis states (state C). Both states have the corresponding crystal structures. The recently reported structures (19, 33), the so-called “ground state” structures, in which both β_TP and β_DP bind with AMP-PNP correspond to the prehydrolysis state. However, the ADP·AlF₃-bound structure (34) would represent the posthydrolysis state. The azide-bound structures (10, 18) would also correspond to this state. Considering that β_DP represents the state that hydrolyzes ATP, β_DP corresponds to the β before and after executing hydrolysis (the β at the right bottom in states B and C). Consequently, β_TP conformation represents the ATP-bound state that was the ATP-waiting state in the prior binding dwell sate (state A in Fig. 6), and β_empty corresponds to the state after ADP release.

Fig. 6.

The proposed reaction scheme of F₁ and correlation with the crystal structure. State A represents the binding dwell state. After ATP binding and ADP release, F₁ makes an 80° substep. Thereafter, F₁ hydrolyzes the tightly bound ATP (denoted by *) in the state transition from B to C. These states correspond to the crystal structure of F₁. After releasing P_i, F₁ makes a 40° substep to complete a cycle of ATP hydrolysis reaction coupled with 120° rotation. State D is the next binding dwell state.

Thus, this study established the correlation between the conformational states of F₁ found in the crystal structures and the single-molecule rotation assay. However, there is one uncertain point in the reaction scheme, the timing of P_i release. The present model assumes that β_empty is the P_i-bound state, and after β_empty releases P_i, the 40° substep is triggered (state C to state D). This assumption is based on the recent study of the crystal structure of yeast MF₁ (33) in which P_i binds to β_empty. The P_i-binding residues in the structure are consistent with those that biochemical studies have identified as the P_i binding site (35), supporting our reaction model. However, considering the other crystal structures that do not show an obvious electron density of P_i in the putative P_i-binding site, an alternative model is still possible; P_i is released from the β_DP immediately after hydrolysis. A clarification of this issue is the next challenging task in the study of the mechanochemical-coupling mechanism of F₁-ATPase.

Another issue is the structure of the binding dwell state. Although the ADP·AlF₄⁻-bound MF₁ shows obvious structural differences, the direction of the twist in the γ is opposite to our expectations. Therefore, ADP·AlF₄⁻-bound MF₁ is not in the binding dwell state. Rather, it might represent the intermediate state between state A and B of our reaction scheme in Fig. 6. Thus, there are no structural data about the binding dwell state. Because it is crucial for the understanding of the mechanochemical coupling mechanism of F₁, a crystallographic study of F₁ in the binding dwell state is one of the most important tasks for the understanding of the mechanism of F₁-ATPase. However, the fact that the crystal structures identified so far are comparable with the reference structure implies that the crystallization of F₁ in the binding dwell state is very difficult. One possible way is to crystallize F₁ at low temperature (16) or F₁ inhibited by tentoxin (36), where it spends most of the catalytic turnover time pausing at the binding dwell angle.

Materials and Methods

Preparation of F₁^αβγ.

Throughout this work, α₃β₃γ subcomplex of F₁-ATPase from thermophilic Bacillus PS3 (TF₁) was used. For inhibition by AMP-PNP or N₃⁻, wild-type F₁^αβγ modified for the rotation assay α(His₆ at N terminus/C193S)₃β(His₁₀ at N terminus)₃γ(S108C/I211C) was used. For simplicity, this F₁^αβγ was referred to as wild-type F₁^αβγ or α₃β₃γ. For the β-γ cross-link experiment, cysteine residues were introduced into γR84 and/or βE391 of wild-type F₁^αβγ and α₃β(E190D)₃γ (14) to construct 4 mutants: α₃β₃γ(R84C), α₃β(E190D)₃γ(R84C), α₃β(E391C)₃γ(R84C), and α₃β(E190D/E391C)₃γ(R84C). Mutagenesis to construct the expression vectors of these mutants was performed as per the previous report on β-γ cross-linking (25). The mutants of F₁^αβγ were expressed in E. coli, purified, and biotinylated as reported (37). For the reconstitution of the chimera, α₃β₂β(E190D/E391C)γ(R84C), solutions of α₃β₃γ(R84C) and α₃β(E190D/E391C)₃γ(R84C) were mixed in a molar ratio of 2:1 and incubated for >2 days in the presence of 200 mM NaCl and 100 mM DTT at 4 °C and pH 7.0. The chimera α₃β(E190D)₂β(E391C)γ(R84C) was prepared by mixing solutions of α₃β(E190D)₃γ(R84C) and α₃β(E391C)₃γ(R84C) in a molar ratio of 2:1.

Rotation Assay.

The rotary motion of F₁^αβγ was visualized by attaching a magnetic bead (<0.2 μm; Seradyn) onto the γ-subunit of F₁^αβγ and immobilizing the α₃β₃ ring on a Ni-NTA-modified glass surface. Phase-contrast images of the rotating bead were obtained with an inverted optical microscope (IX-70; Olympus) equipped with magnetic tweezers (30). The image was captured with a CCD camera (FC300M; Takenaka) and recorded with a DV-CAM (DSR-11; Sony) at 30 fps. The recorded images were analyzed with image analysis software (Celery; Library) or a custom-made program (K. Adachi, Waseda University). The experimental procedures of the rotation assay for β-γ cross-linking were mostly the same as those reported (16), except for the content of the buffer. The basal buffer for the rotation assay contained 50 mM Hepes-KOH at pH 8.0, 50 mM KCl, 2 mM MgCl₂, 1 mM phospho(enol)pyrubate, 0.1 mg/ml pyrubate kinase, 5 mg/ml BSA, and 1 mM DTT. For β-γ cross-linking, 200 μM DTNB was added to the basal buffer, from which DTT and BSA were omitted. The rotation assay for AMP-PNP and/or NaN₃ inhibition was carried out in the buffer containing 50 mM Mops-KOH at pH 7.0, 50 mM KCl, 2 mM MgCl₂, 5 mg/ml BSA, and the indicated amount of nucleotides and NaN₃.