High-Speed Atomic Force Microscopy Reveals Rotary Catalysis of Rotorless F1-ATPase

See allHide authors and affiliations

Science  05 Aug 2011:
Vol. 333, Issue 6043, pp. 755-758
DOI: 10.1126/science.1205510


F1 is an adenosine triphosphate (ATP)–driven motor in which three torque-generating β subunits in the α3β3 stator ring sequentially undergo conformational changes upon ATP hydrolysis to rotate the central shaft γ unidirectionally. Although extensive experimental and theoretical work has been done, the structural basis of cooperative torque generation to realize the unidirectional rotation remains elusive. We used high-speed atomic force microscopy to show that the rotorless F1 still “rotates”; in the isolated α3β3 stator ring, the three β subunits cyclically propagate conformational states in the counterclockwise direction, similar to the rotary shaft rotation in F1. The structural basis of unidirectionality is programmed in the stator ring. These findings have implications for cooperative interplay between subunits in other hexameric ATPases.

F1-ATPase, a water-soluble portion of adenosine triphosphate (ATP) synthase (1), is a rotary motor protein. The α3β3γ subcomplex (referred to here as F1) suffices as the motor, in which the rotor γ subunit rotates in the stator α3β3 ring upon ATP hydrolysis (2). The concept of the “rotary catalysis” of F1 was proposed on the basis of biochemical studies (3). It was strongly supported by the first crystal structure reported (4) and directly proven by observations of rotating single molecules (5). In F1, the catalytic sites are located at the α-β interfaces, mainly on the β subunits. In the crystal structure (4), three catalytic sites are in different nucleotide-bound states; one binds to an ATP analog (αTPTP in Fig. 1E), another binds to adenosine diphosphate (ADP) (αDPDP), and the third is unbound (αEE). Both βTP and βDP assume the closed conformation, swinging the C-terminal domain toward γ, whereas βE assumes the open conformation, swinging the domain away from γ. Because these two general conformational states appear to push or be pushed by γ, respectively, it was proposed that interactions with γ control the conformational and catalytic states of individual β’s to sequentially generate torque (6). In fact, some biochemical studies are thought to suggest that the α3β3 ring alone does not possess intrinsic cooperativity, and γ mediates the interplay among β’s (79). This view was reinforced by studies showing that backward mechanical rotation of γ with external force reverses the chemical reaction toward ATP synthesis (10, 11), whereas forced forward rotation results in accelerated ATP binding (12).

Fig. 1

(A) Averaged AFM image of C-terminal side of the α3β3 subcomplex without nucleotide (movie S1). (B) C-terminal side of the crystal structure of the nucleotide-free α3β3 subcomplex [Protein Data Bank (PDB) code 1SKY] (21). The α and β subunits are colored in cyan and pink, respectively. The C-terminal DELSEED motif of β corresponding to the high protruding portions is highlighted in red. (C) Simulated AFM image of the α3β3 subcomplex constructed from the structure in (B). (D) Averaged AFM image of C-terminal side of the α3β3 subcomplex in 1 mM AMPPNP. (E) Atomic structure of the α3β3 subcomplex with bound nucleotides. This structure is obtained by removing γ from the crystal structure of F1 (PDB code 1BMF) (4). (F) Simulated AFM image constructed from the structure in (E). The brightness of all AFM images in this paper represents the sample height but is not linearly set to highlight the top surface structure (fig. S4).

Recently, however, this contention has been challenged by the finding that even when most interaction sites between β and γ are abolished, F1 retains catalytic power to rotate γ unidirectionally (13, 14). A few biochemical studies also suggest the intrinsic cooperativity of the α3β3 ring (15, 16). However, because conventional single-molecule optical microscopy requires attachment of a probe onto the rotary shaft for visualization (5), it does not allow direct examination of whether the intrinsic cooperativity in the α3β3 ring is the core feature responsible for sequential torque generation. We clarified this issue by directly imaging the ATP-driven conformational transition of β’s in the isolated α3β3 ring, using high-speed atomic force microscopy (AFM) (17, 18), a technique that can visualize proteins at work in real time without probes (19, 20).

The α3β3 subcomplex (fig. S1) was covalently immobilized on a mica surface and observed with high-speed AFM with a frame capture time of 80 ms unless otherwise mentioned (Fig. 1 and figs. S2 and S3). In the absence of nucleotide, the α3β3 showed a pseudo-sixfold symmetric ring in which three alternately arranged subunits were elevated relative to the other three (Fig. 1A and fig. S4). The simulated AFM image of the C-terminus side of the ring constructed from the crystal structure of the nucleotide-free α3β3 subcomplex (21) well reproduced the observed image (Fig. 1C). This indicates that the N-terminus side was selectively attached to the mica, although a smaller ring corresponding to the N-terminus side was occasionally observed (fig. S5). The three β’s, which all assumed an open conformation in the crystal structure of α3β3, gave three protruding peaks in the simulated image as bright spots. When a nonhydrolyzable ATP analog AMPPNP was added, the ring became triangular and the central hole was obscured (Fig. 1D). Although the three α’s with lower protrusions retained the same conformation as those under the nucleotide-free condition, two of three β’s retracted toward the center and simultaneously lowered their protrusions. Consequently, the ring showed a single high protrusion. A simulated image of α3β3 with bound nucleotides was constructed using a structure in which γ was removed from the crystal structure of F1 (Fig. 1E) (4). The simulated image (Fig. 1F) also showed an asymmetric ring very similar to that of the observed image. The excellent agreement indicates that only two β’s can assume the closed conformation, even in saturating AMPPNP. This feature is consistent with the observation that three β’s do not assume the closed conformation simultaneously (22).

When ATP was added, β showed distinct conformational dynamics; each β underwent a conformational transition between the outwardly extended high state (open) and the retracted low state (closed) (Fig. 2A and table S1); the outwardly extended and retracted conformations correlated well with the high- and low-protrusion states, respectively (fig. S6). The most prominent features are that only a single β assumes the open state, as in the presence of AMPPNP, and that when the open-to-closed transition occurs at one β, the opposite closed-to-open transition occurs simultaneously at its counterclockwise neighbor β in most cases. Thus, the high and outwardly extended conformation propagates in the counterclockwise direction (Fig. 2, A and B).

Fig. 2

(A) Successive AFM images showing the conformational change of β’s in 2 μM ATP (movie S3). The highest pixel in each image is indicated by the red circle. Frame rate, 12.5 frames/s. (B) Time evolution of the cumulated angle of the highest pixel. The inset shows a trajectory, superimposed on an AFM image, of the highest pixels corresponding to the high protrusions of open β’s (412 frames, movie S6). The center of rotation is defined by the averaged x and y positions of the highest pixels, and the cumulated angles are calculated relative to the first frame. (C) Correlation coefficient histograms calculated for each β designated in (A) (n = 220). An open β is used as a reference for the analysis of each β. (D) Time courses of correlation coefficients. The white and gray backgrounds show periods of O and C states, respectively. Solid lines show the mean correlation coefficient for each period. (E) Time evolution of the cumulated number of counterclockwise shifts of the CCO state. Circles, x’s, and crosses correspond to CCO, COO, and other irregular states (OOO and CCC), respectively. The increase in the cumulated number indicates that the open β in CCO shifts counterclockwise.

For a more quantitative analysis of this cooperativity, the conformational states of the three β’s in each image frame were determined from their correlation coefficients, using respective reference images of β’s in the open state (fig. S7). The correlation coefficients are distributed around two distinct peaks at ~0.995 and ~0.96 (Fig. 2C). The peaks with larger and smaller coefficients correspond to the open (O) and closed (C) states, respectively. On the other hand, the correlation coefficient analysis for the nucleotide-free and AMPPNP-bound α3β3 showed static conformations of β (figs. S8 and S9 and table S1). Figure 2D shows typical time courses of the two conformational states for individual β’s. In all images observed in the presence of 2 to 4 μM ATP (n = 8746 frames), three β’s at a given frame dominantly showed the CCO state (82% of total). The COO state was also observed (14.5%). Other states such as CCC (3%) and OOO (0.5%) were rare. We analyzed the rotary propagation of the O and C states by counting the number of counterclockwise shifts of the CCO state [Fig. 2E and supporting online material (SOM)]. The propagation is unidirectional, rotating counterclockwise with an efficiency of 83% (in 2 μM ATP) over the total shift events examined (n = 371) (fig. S10). Considering that two consecutive counterclockwise shifts occurring occasionally within the frame capture time (80 ms) would be counted as a clockwise shift, this value of efficiency is probably underestimated by ~6% (Fig. 3D, fig. S11, and SOM).

Fig. 3

(A to C) Time evolutions of the cumulated number of counterclockwise CCO shifts at various concentrations of ATP ([ATP]). At each [ATP], four representative curves for the different molecules are indicated as in Fig. 2E. (D) [ATP] dependence of the initial rates of ATP hydrolysis determined by biochemical assay [circles; 2 μM, 1.6 ± 0.9 s−1; 3 μM, 2.5 ± 1.0 s−1; 4 μM, 3.3 ± 1.3 s−1 (mean ± SD, n = 6 individual measurements at each [ATP])] and the rates of conformational change of β determined by fitting the dwell time histograms of open (squares; 2 μM, 1.5 s−1; 3 μM, 3.0 s−1; 4 μM, 3.8 s−1) and closed (triangles; 2 μM, 1.5 s−1; 3 μM, 2.5 s−1; 4 μM, 3.2 s−1) states shown in fig. S11.

The rate of rotational propagation of the CCO state increased with increasing ATP concentration (Fig. 3, A to C), indicating that ATP binding is rate-limiting, consistent with the biochemically determined Michaelis constant of 12 μM for ATP hydrolysis (fig. S1). Histograms of the dwell time of the O state followed a single exponential function, and those of the C state were well fitted with the model that two consecutive events of ATP binding to the other two β’s trigger the transition of the β from the C to the O state (fig. S11). The rate constants of the counterclockwise shift of the CCO state were comparable with the initial rates of ATP hydrolysis measured biochemically (Fig. 3D). Thus, each ATP hydrolysis is well coupled with the open-to-closed transition of β. Just before transitioning to the next CCO state, a COO state occasionally appeared (~30% of total transitions) (x’s in Figs. 2E and 3, A to C). In this event, one of the two closed β’s positioned at the counterclockwise side opened in most cases, implying that ADP can be released from the closed β before ATP binds to the open β (23, 24).

The present results prove that the stator α3β3 ring alone possesses high cooperativity for sequential power stroking among three catalytic β’s. This was also indicated by the observations that the occasional subunit dissociation completely stopped the rotary propagation of the conformational state (fig. S12). Thus, the “γ-dictator” model (13), which proposes that only the interaction with γ determines the conformational and catalytic states of β’s (23, 24), is not valid. On the other hand, the ATP-binding rate and the efficiency of unidirectionality of the α3β3 subcomplex are distinctly lower than those of F1 (Fig. 3 and fig. S11). Thus, the interaction with γ is dispensable but still important for the rapid and precise rotary catalysis. Our findings are not inconsistent with the observations that the rates and equilibriums of the catalytic reactions are apparently under the control of the rotary angle of γ (1012). The intrinsic interplay among β’s would reinforce catalytic control by γ; even if γ tightly interacts with only one β, it still can act on all β’s through β-β interplay.

These results also have implications for the cooperativity of other structurally related hexameric ATPases such as RecA- and AAA+-family proteins, in which cooperativity among the separate catalytic sites is a central issue (25). These ATPases may also have intrinsic cooperativity. Comparative studies on these proteins should shed light on the common operating principles of hexameric ATPases.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Table S1

References (2628)

Movies S1 to S8

References and Notes

  1. Acknowledgments: We thank M. Tanigawara, H. Yamashita, R. Hasegawa, and D. Okuno for technical help and members of the Noji and Ando laboratories for valuable comments. This work was supported by Grants-in-aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology project ID nos. 18074005, 18201025, 21770168, 21681017, 2102301, 21107517, 22247025, and 20221006); the Knowledge Cluster Initiative; and the Japan Science and Technology Agency for Core Research for Evolutional Science and Technology.
View Abstract

Stay Connected to Science

Navigate This Article