Conformational ensemble of Yeast ATP Synthase at Low pH Reveals Unique Intermediates and Plasticity in F₁-F_o Coupling

Abstract

Mitochondrial ATP synthase utilizes the proton gradient across the inner mitochondrial membrane to synthesize ATP. Structural and single molecule studies conducted mostly at neutral or basic pH have provided details of the reaction mechanism of ATP synthesis. However, pH of the mitochondrial matrix is slightly acidic during hypoxia and pH-dependent conformational changes in the ATP synthase have been reported. Here we use single-particle cryo-EM to analyze the conformational ensemble of the yeast (Saccharomyces cerevisiae) ATP synthase at pH 6. Of the four conformations resolved in this study, three are reaction intermediates. In addition to canonical catalytic dwell and binding dwell structures, we identify two unique conformations with nearly identical positions of the central rotor but remarkably different catalytic site conformations. These structures provide new insights into the catalytic mechanism of the ATP synthase and highlight elastic coupling between the catalytic and proton translocating domains.

Introduction.

Adenosine triphosphate (ATP) synthases utilize the proton motive force (pmf) across biological membranes to catalyze the conversion of adenosine diphosphate (ADP) and inorganic phosphate (Pi) to ATP¹. The enzyme is composed of a soluble F₁-ATPase and a membrane-integral proton channel (F_o). F₁-ATPase consists of a hexamer with alternating α and β subunits and a central stalk composed of γ, ε, and δ subunits that acts as a rotor coupled to F_o. F_o is made of a proteolipid ring composed of ten c-subunits (in Saccharomyces cerevisiae), and one copy of subunit-a. F₁ and F_o are structurally and functionally coupled by the central rotor, ancillary subunits, and a peripheral stator composed of subunits b, d, e, f, g, i/j, k, F6, OSCP (oligomycin sensitivity conferring protein) and 8².

The pmf-driven translocation of protons through F_o induces rotation of the c-ring and central rotor^1,3, which leads to conformational changes in the catalytic α/β pairs, resulting in the synthesis of ATP from ADP and Pi as first proposed in the binding change mechanism⁴. During rotary catalysis, the peripheral stator holds F₁ in place, ensuring the conversion of mechanical energy to chemical energy⁵. In the absence of an electrochemical gradient, ATP synthase can hydrolyze ATP, leading to rotation of the central rotor and the c-ring in the opposite direction and thereby reversing the flow of protons. Under hydrolyzing conditions, the F₁F_o-ATP synthase is a coupled system that operates without the loss of energy⁶. With three catalytic sites in F₁ and ten c-subunits in F_o, three molecules of ATP synthesized or hydrolyzed correlate with ten protons translocated across the membrane. To accommodate this “symmetry mismatch” and enable elastic coupling between F₁ and F_o, the central rotor and/or the peripheral stator must be flexible^7–10.

In the absence of exogenous factors, purified ATP synthase exists in three distinct conformations with the central rotor occupying rotary positions separated by 120°. Early cryo-EM maps of bovine mitochondrial F₁F_o identified the three rotational states and named them States 1, 2 and 3¹¹. Based on single molecule studies visualizing the rotation of the central rotor during ATP hydrolysis¹², and cryo-EM structures of ATP Synthase, the rotational states proceed from State 1 to State 3 to State 2 to State 1^9,11.

Single molecule studies have further divided the 120° rotation steps into sub-steps of 80° and 40° corresponding to the ATP binding stroke and the ATP hydrolysis stroke. The two strokes are separated by waiting periods called the “binding dwell”, when the enzyme awaits ATP binding, and the “catalytic dwell”, the wait before ATP hydrolysis^13–15. Recently, Sobti et al. employed a combination of time- and temperature-resolved cryo-EM to determine the structures of the Bacillus PS3 F₁-ATPase in the catalytic dwell and binding dwell conformations¹⁶. Similarly, Guo et al. conducted cryo-EM of the yeast ATP synthase in presence of ATP to resolve the role of the peripheral stalk in the storage of elastic energy and observed the catalytic and binding dwell conformations⁹. Considering the arrangement of the catalytic sites with the rotational position of the central rotor, the catalytic dwell is similar to the ground state of the enzyme as observed in the crystal structure of F₁-ATPase¹⁷, with two closed catalytic sites that enclose ATP or ADP and a third open catalytic site¹⁸.

To limit the conformational heterogeneity of the yeast ATP synthase, we genetically fused the δ-subunit of the central rotor with the F6-subunit of the peripheral stator using T4 lysozyme, resulting in the high-resolution cryo-EM structure of the enzyme in one rotational state (State 2)². In the present study, we conducted cryo-EM of the rotor-stator fused construct in presence of 1 mM ATP and observed two distinct rotational conformations of the enzyme corresponding to States 1 and 2. The experiment suggested that while being conformationally limited, the central rotor of our fusion construct is capable of rotating at least up to 120°, and therefore can be used to study reaction intermediates during the synthesis/hydrolysis of ATP.

Since protonation/deprotonation reactions are central to the reaction pathway of the ATP synthase, we examined the effect of pH on the conformation of the enzyme. Several studies have suggested that low pH can induce conformational changes related to the reaction pathway. First, the specific activity of yeast F₁-ATPase decreases by ~55% from pH 8.5 to 6.0¹⁹ indicating that protonation of residues in F₁-ATPase alters the rate limiting step in ATP hydrolysis. Second, cross-linking experiments in E. coli have suggested pH-dependent conformational changes in subunit-a with implications in regulating the proton pathway^20,21. Third, single molecule studies using E. coli ATP synthase show increased stalling of the enzyme in transient dwells, at low pH (pH 5-6.5)²². Fourth, another single molecule study has demonstrated pH-dependent rotation of the E. coli c-ring by 11° in the direction of ATP synthesis²³. Therefore, protonation of residues in the ATP synthase alters its structure and kinetics, potentially stabilizing conformations not observed at neutral or basic pH.

To resolve new putative intermediates, we conducted cryo-EM of the yeast ATP synthase at pH 6. Four distinct conformations of the enzyme were observed with the predominant conformation identified as the binding dwell. The relative position of the central rotor with respect to the conformation of the catalytic sites in two of the four conformations have not been previously described. The remarkably different conformational landscape of ATP synthase at pH 6 as compared to pH 8² suggests a role for the protonation of residues in driving conformational changes in the reaction pathway, even in the absence of exogeneous ATP. Furthermore, comparison of the conformations provides evidence for elastic coupling between F₁ and F_o by the central rotor.

Results

Conformations of ATP synthase in presence of ATP.

In the absence of ATP, the cryo-EM structure of rotor-stator fused ATP synthase occupied in a single rotational state (State 2)². To test whether this construct would allow sufficient rotation of the central rotor to enable the study of reaction intermediates, we conducted cryo-EM of modified yeast ATP synthase reconstituted in lipid nanodiscs, at pH 8 and in presence of 1 mM Mg:ATP (Extended Data Fig. 1). Three-dimensional (3D) classification of the particle images generated maps in which the central rotor occupied two distinct rotational positions, suggesting the presence of at least two conformations in the dataset (Extended Data Fig. 1C). The first conformation corresponding to ~20% of the particle images was refined to 4.3-Å resolution. This map aligned with our previously observed conformation in State 2² (Extended Data Figs. 1F and 1G).

A second conformation of ATP synthase was observed marked by a distinct rotational position of the central rotor, but poorly resolved density for the peripheral stalk (Extended Data Fig. 1C). Increased motion of the peripheral stalk under ATP hydrolyzing conditions is consistent with recent cryo-EM studies of the yeast ATP synthase⁹. For the second conformation, classification focused on the F₁ region resulted in an F₁ map at 3.2-Å resolution. Image processing focused on the peripheral stalk significantly improved its resolution and resulted in a complete map of F₁F_o at 4.2-Å resolution (Extended Data Fig. 1C). Using the F₁ and F₁F_o maps, a complete model for F₁F_o was built (details described in Methods) (Extended Data Fig. 1I). This second and predominant (~80% of particle images) conformation of F₁F_o was not reported in the absence of exogenous ATP previously², and it aligns with rotational State 1 of bovine ATP synthase¹¹ (Extended Data Fig. 1J). Taken together, our results indicate that the rotor-stator fusion construct, although conformationally limited, can undergo rotation of the central rotor by at least 120° and is therefore suitable for resolving reaction intermediates during the hydrolysis/synthesis of ATP.

Cryo-EM of ATP synthase at pH 6.

Prior studies have indicated that shifting the pH of the medium from neutral to acidic is sufficient to cause conformational changes that possibly represent reaction intermediates^19–21. To investigate this, the rotor-stator fused ATP synthase was purified from yeast mitochondria, reconstituted in lipid nanodiscs², the buffer replaced with buffer at pH 6.0, and subjected to single particle cryo-EM analysis. The flexibility of the central rotor and peripheral stalk induces motion between the F₁ and F_o domains⁷. As a result, the initial two-dimensional (2D) class averages showed that either F₁ or F_o was well resolved (Extended Data Fig. 2B). To improve the resolution of both F₁ and F_o, the cryo-EM images were processed using multiple strategies (Extended Data Fig. 2C, D and E), resulting in high-resolution cryo-EM maps of the yeast ATP synthase in four distinct conformations (Table 1, Extended Data Figs. 3 and 4).

Table 1.

Cryo-EM data collection, refinement and validation statistics

	#1 MgATP (pH 8) F1Fo (EMDB-29270) F1 (EMDB-29278) (PDB 8FL8)	#2 Conf-0 (pH 6) F1Fo (EMDB-28685) F1: (EMDB-28687) Fo: (EMDB-28689)	#3 Conf-1 (pH 6) F1Fo (EMDB-28809) F1 (EMDB-28811) Fo (EMDB-28813) Fo+CR (EMDB-28814) (PDB-8F29)	#4 Conf-2 (pH 6) F1Fo (EMDB-28835) F1 (EMDB-28836) Fo+CR (EMDB-28837) (PDB-8F39)	#5 Conf-3 (pH 6) F1Fo (EMDB-29250) Fo+CR (EMDB-29251) (PDB-8FKJ)
Data collection and processing
Magnification	130,000x	75,000x	75,000x	75,000x	75,000
Voltage (kV)	300	300	300	300	300
Electron exposure (e–/Å2)	51.07	52	52	52	52
Defocus range (μm)	−0.9 to −2.5	−0.7 to −1.5	−0.7 to −1.5	−0.7 to −1.5	−0.7 to −1.5
Pixel size (Å)	1.06	1	1	1	1
Symmetry imposed	C1	C1	C1	C1	C1
Initial particle images (no.)	362,671	1,128,548	1,128,548	1,128,548	1,128,548
Final particle images (no.)	F1Fo: 65,559 F1: 212,237	F1Fo: 110,313 F1: 110,313 Fo: 110,313	F1Fo: 146,826 F1: 146,826 Fo: 146,826 Fo+CR: 97393	F1Fo: 99,780 F1: 99,780 Fo+CR: 37,000	F1Fo: 15,189 Fo+CR: 15,189
Map resolution (Å)	F1Fo: 4.2 F1: 3.2	F1Fo: 4.5 F1: 2.9 Fo: 3.3	F1Fo: 4 F1: 3 Fo: 3.3 Fo+CR: 3.8	F1Fo: 3.5 F1: 2.8 Fo+CR: 3.7	F1Fo: 4.3 Fo+CR: 4.1
FSC threshold	0.143	0.143	0.143	0.143	0.143
Map resolution range (Å)	3.5 to 7.5	3.5 to 7.5	2.5 to 6.0	2.5 to 6.5	3.5 to 7.5
Refinement
Initial model used (PDB code)	6CP6		6CP6	6CP6	6CP6
Model resolution (Å)	4.3		4.1	4.0	7.5
FSC threshold	0.5		0.5	0.5	0.5
Model resolution range (Å)
Map sharpening B factor (Å2)	−100		−80	−50	−50
Model composition
Non-hydrogen atoms	38784		38906	38767	25001
Protein residues	5098		5116	5098	5088
Ligands	10		11	9	0
B factors (Å2)
Protein	−205.54		−179.67	−160.42	−160.82
Ligand	−170.04		−134.29	−124.07	−160.82
R.m.s. deviations
Bond lengths (Å)	0.003		0.003	0.003	0.003
Bond angles (°)	0.724		0.785	0.740	0.655
Validation
MolProbity score	2.1		2.1	2.04	1.93
Clashscore	16.12		18	14	6.82
Poor rotamers (%)	0.05		3.9	1.4	0
Ramachandran plot
Favored (%)	94.15		95.3	94.6	90.02
Allowed (%)	5.65		4.66	5.18	9.86
Disallowed (%)	0.2		0.04	0.22	0.12

Initial 3D classification of particle images identified two conformations with distinct rotational positions (Extended Data Fig. 2C). The first conformation (Extended Data Fig. 2C, green arrows) aligned well with the State 2 structure of yeast ATP synthase resolved at pH 8.0². This State 2 structure was resolved in three different datasets of our rotor-stator fusion construct collected under different conditions: previously reported at pH 8², in presence of exogenous ATP (described above) and reported now at pH 6. Therefore, we refer to this conformation as the ground state for our construct, or Conf-0 (Fig. 1 C and D). Relion multibody refinement²⁴ of Conf-0 resulted in maps for F₁ and F_o at 2.9- and 3.3-Å resolution, respectively (Extended Data Fig. 2C, green arrows). The structure of ATP synthase determined at pH 8 (pdb id: 6CP6)² fits well in the Conf-0 maps, and both align well with the cryo-EM map of bovine ATP synthase in State 2 (Extended Data Fig. 5A). The conformations of the three catalytic sites and the rotational angle of the central rotor in Conf-0 align well with the catalytic dwell conformations of Bacillus PS3 F₁-ATPase (pdb id: 7L1R)¹⁶ and yeast ATP synthase from a separate study (pdb id: 7TKR)⁹ (Extended Data Fig. 5B, C and D), indicating that Conf-0 represents the catalytic dwell. The catalytic dwell is the most commonly observed conformation and is reported in the crystal structures of the F₁-ATPase as well as cryo-EM structures of F₁ and F₁F_o^{2,10,11,17,25,26}.

Figure 1. Structure of yeast ATP synthase in Conf-1 and its comparison with Conf-0.

A, Cryo-EM map of yeast ATP synthase in Conf-1 with individual subunits and domains color coded. B, Sectional view of the catalytic F₁ domain in Conf-1 as seen from the membrane with cryo-EM density for nucleotides enclosed in the catalytic sites depicted with arrows. C, Sectional view of the catalytic F₁ domain in Conf-0 (pdb ID 6CP6) as seen from the membrane with cryo-EM density for enclosed nucleotides depicted with arrows. D, Cryo-EM map of yeast ATP synthase in Conf-0 with individual subunits and domains color coded E, Atomic model for Conf-1 F, Comparison of the central rotor in Conf-1 (solid) and Conf-0 (transparent). Same residues in two conformations have been colored in blue in γ or depicted as spheres in ε and δ, to show the extent of conformational change. G, Cross-sectional view of the γ subunit in Conf-1 (dark green) and Conf-0 (light green). H, Sectional view of F₁ from the membrane side showing the extent of rotation of central rotor from Conf-0 to Conf-1. “PS” refers to peripheral stalk.

Binding dwell is the predominant conformation at pH 6.

A second conformation, referred to here as Conf-1 (Extended Data Fig. 2C, red arrows), displays a different rotational position of the central rotor compared to Conf-0 and is the predominant conformation observed at pH 6 (Fig. 1 A and B, Extended Data Fig. 2C). Multibody refinement of Conf-1 generated F₁ and F_o maps at 3- and 3.3-Å resolution, respectively. To improve the resolution of the central rotor, 3D classification focused on F_o and the F_o-proximal region of the central rotor (F_o+CR) was conducted resulting in a map at 3.8-Å resolution (Extended Data Fig. 2C, red arrows). The high-resolution maps for F₁, F_o, and F_o+CR along with the consensus map for F₁F_o were used to build a complete model for ATP synthase in Conf-1 (details in methods) (Fig. 1E).

Comparison of the structures of ATP synthase in Conf-0 and Conf-1 revealed rotation of the central rotor by 70° in the counterclockwise direction from Conf-0 to Conf-1, as viewed from the membrane, towards F₁ (Figs. 1F–H, Supplementary Movie 1). In F_o, the interface between subunit-a and the c-ring appear very similar in Conf-0 and Conf-1 (Extended Data Fig. 6). The simplest route in going from Conf-0 to Conf-1 is for the c-ring to rotate by two c-subunits (72°) to reach an equivalent position as Conf-0, with the central rotor rotating with it (Extended Data Fig. 6, Supplementary Movie 1). The rotation of the central rotor switches the position of γ (without significant distortion), with concomitant changes in the catalytic sites. These conformational changes proceed in the direction of ATP hydrolysis, from State 2 to State 1.

Large-scale conformational changes are observed in the F₁ domain in Conf-1, as compared to Conf-0. In Conf-0, two of the catalytic sites are in the closed conformation with bound Mg:ADP, whereas the third site is open and empty (Fig. 1C). In contrast, the F₁ domain of Conf-1 exhibits one closed catalytic site with bound Mg:ADP, a second open and empty catalytic site, and a third half-open catalytic site with Mg:ADP and phosphate (Fig. 1B). The presence of phosphate in the catalytic site is likely the result of the hydrolysis of ATP to ADP and phosphate. The conformations of the catalytic sites in Conf-1 are similar to the binding dwell conformation in Bacillus PS3 F₁ (pdb id:7L1Q)¹⁶ and yeast ATP synthase (pdb id: 7TKM)⁹ (Fig. 2 A, E and I). Particularly, all three structures have a half-open catalytic site with Conf-1 and Bacillus PS3 F₁ containing bound Mg:ADP and phosphate and a similar arrangement of amino acid residues in the catalytic site (Fig. 2B, F, J, M and N). The second catalytic site is open and empty in Conf-1 and yeast ATP synthase (pdb id: 7TKM), whereas it is half-closed with Mg:ATP bound in Bacillus PS3 F₁ (Fig. 2C, G and K). The third catalytic site is closed in all three structures, with Mg:ADP bound in Conf-1, while Mg:ATP is bound in Bacillus PS3 and yeast ATP synthase (Fig. 2D, H and L). This discrepancy in nucleotide occupancy in catalytic sites 2 and 3 is potentially because while the structure of Conf-1 was determined at pH 6 and in absence of exogenous Mg:ATP, cryo-EM of Bacillus PS3 F₁ and yeast ATP synthase were conducted at basic pH and in presence of Mg:ATP. Therefore, based on the comparison of the catalytic sites of Conf-1 with existing binding dwell structures, Conf-1 most likely corresponds to the binding dwell.

Figure 2. Comparison of Conf-1 with the binding dwell of F₁-ATPase from Bacillus PS3.

A, Sectional view from the membrane of F₁ in Conf-1. The individual α/β pairs with respect to the γ subunit have been shown in B (half-open), C (open) and D (closed). E. Sectional view from the membrane of F₁ from Bacillus PS3 (pdb id. 7L1Q). The individual α/β pairs with respect to the γ subunit have been shown as F (half-open), G (half-closed), H (closed). I, Sectional view from the membrane of F₁ from yeast ATP synthase (pdb id. 7TKM). The individual α/β pairs with respect to the γ subunit have been shown as J (half-open), K (open) and L (closed). M, MgADP + Pi enclosed in the half open catalytic site (B) in Conf-1. N, MgADP + Pi enclosed in the half open catalytic site (F) in Bacillus PS3 F₁ (pdb ID 7L1Q). “PS” refers to peripheral stalk.

Conf-2 highlights flexibility of the central rotor.

Given the large structural differences in the F₁ domain between Conf-0 and Conf-1, we looked for the presence of additional conformations based on F₁. Indeed, 3D classification of particle images focusing on F₁ identified another conformation, referred to as Conf-2, with the central rotor position similar to Conf-1 (Fig. 3A, Extended Data Fig. 2D, blue arrows). The F₁ portion of Conf-2 was refined to 2.8-Å resolution and using the same image processing strategy, the F₁ region of Conf-1 was refined to 2.7-Å resolution. To improve the resolution of the F_o domain, the particle images for Conf-2 were subjected to multibody refinement that generated a 3.7-Å resolution map for F_o+CR (Extended Data Fig. 2D, blue arrows). A complete model for Conf-2 was built using the high-resolution maps for F₁, F_o+CR and the consensus map for F₁F_o (details in Methods) (Fig. 3E).

Figure 3. Conf-2 and its comparison with Conf-1.

A, Cryo-EM map of yeast ATP synthase in Conf-2 with individual subunits and domains color coded. The corresponding model is shown in E. B, Sectional view of the F₁ domain (as observed from the membrane) in Conf-2 showing the catalytic sites in closed, closed and open conformations. C, Sectional view of F₁ in Conf-2 (solid) and Conf-1 (transparent). Changes in conformation of the β/γ interface showing a half-open (Conf-1) to open (Conf-2) transition of the catalytic site is highlighted in F (Conf-1) and G (Conf-2). Similarly, changes in conformation of the β/γ interface showing an open (Conf-1) to close (Conf-2) transition of the catalytic site is highlighted in H (Conf-1) and I (Conf-2). D. Comparison of the central rotor in Conf-2 (solid) with Conf-1 (transparent). Changes observed in the N-terminal helix and C-terminal helix of γ are highlighted in J and K respectively. “PS” refers to peripheral stalk.

The Conf-2 structure appears to be the earliest reported intermediate emerging after the binding dwell. While the rotational position of the central rotor in Conf-2 is similar to Conf-1 (Fig 3D, Extended Data Fig. 5 H and I), the catalytic subunits in Conf-2 undergo significant conformational changes with respect to Conf-1. In common, both structures have one closed catalytic site with bound Mg:ADP and one open and empty catalytic site (Fig. 3B and C). However, there is a transition in the active sites from Conf-1 to Conf-2. The empty catalytic site (β-subunit, chain E) in Conf-1, is closed with bound Mg:ADP in Conf-2, while the half-open catalytic site (β-subunit, chain D) in Conf-1, is fully opened and devoid of nucleotide in Conf-2 (Fig. 3C). These conformational changes are expected in the catalytic sites when they transition out of the binding dwell and are typically associated with the binding of nucleotide and the rotation of the central rotor in the direction of ATP hydrolysis. However, between Conf-1 and Conf-2 this transition occurs without any significant rotation of the central rotor (Supplementary Movie 2). In fact, novel changes in the conformation of the central rotor are observed (described below) which likely account for the changes in the active site conformations. It is also possible that Conf-2 represents a state identified by Martin et al.²⁷, using single molecule studies, when ADP is released from the catalytic dwell with rotation of the central stalk by about 60°. If that is the case, then Conf-2 would be within the normal catalytic pathway.

The conformational changes in the catalytic sites of F₁ between Conf-1 and Conf-2 are caused by the displacement of the N-terminal and C-terminal helices of the γ-subunit by 3 and 6 Å, respectively (Fig. 3D, J and K). The displacement of the N-terminal helix of γ changes the half-open catalytic site in Conf-1 to open in Conf-2 (Fig. 3C, F and G). Similarly, displacement of the C-terminal helix of γ changes the open catalytic site in Conf-1 to closed in Conf-2 (Fig. 3C, H and I). In the half-open catalytic site of Conf-1, the N-terminal helix of the γ-subunit interacts with the conserved β-DELSEED region²⁸ via hydrophobic and salt-bridge interactions. In Conf-2, the conformational change from half-open to open, increases the distance between the γ- and β-subunits, thereby breaking these interactions (Fig. 3F and G). Similarly, the conformational change of the catalytic site from open in Conf-1 to closed in Conf-2, brings the β-DELSEED region closer to the C-terminal helix of the γ-subunit to restore interactions between the γ- and β-subunits (Fig. 3H and I). While in Conf-1, the C-terminal helix of γ is in contact with the β-DELSEED region of the closed catalytic site (β-subunit, chain F), in Conf-2, the same region of γ is in contact with the β-DELSEED region of the newly closed catalytic site (β-subunit, chain E). Therefore, large conformational changes in the catalytic sites are induced by distortion of the N- and C-terminal helices of γ-subunit.

Conf-3 represents a second binding dwell.

We next searched for conformational differences within F_o as protonation reactions are central to the reaction mechanism. As such, we performed image processing focusing on F_o+CR, a strategy that improved the map quality of F_o in our previous work²⁹. This processing scheme generated two cryo-EM maps of F_o+CR at 3.6- and 4.1-Å resolution (Extended Data Fig. 2E). While the first map (Extended Data Fig.2E, green arrows) aligns well with Conf-0, superimposition of the second map (Extended Data Fig. 2E, pink arrows) with those for Conf-0 and Conf-1, indicates that its rotational conformation is unique and lies between Conf-0 and Conf-1.

To accurately determine the rotational conformation of F₁F_o, the particle images contributing to the F_o-centered maps were re-centered based on F₁F_o and reconstructed to generate complete maps of ATP synthase at overall resolutions of 4.2-Å for the first conformation and 4.3-Å for the second conformation (Extended Data Fig. 2E). The ground state structure (pdb id: 6CP6)² aligned well with the F₁F_o map of the first conformation, further supporting that the map represented Conf-0.

In case of the second conformation, the 4.1-Å resolution map for F_o+CR and the complete map for F₁F_o at 4.3-Å resolution were used to build a model for F₁F_o (Fig. 4B). In this model, the central rotor displays a clockwise rotation, when viewed from the membrane side towards F₁, relative to Conf-1 by 35°, and counterclockwise rotation with respect to Conf-0 by 35° (Fig. 4C and D). Thus, F_o-centered image processing revealed a conformation of ATP synthase in which the central rotor is positioned between Conf-0 and Conf-1. We refer to this conformation as Conf-3 (Fig. 4A). The arrangement of catalytic sites and rotational angle of the central rotor in Conf-3 are similar to the binding dwell conformations of Bacillus PS3 F₁ (pdb ID: 7L1Q)¹⁶ and yeast ATP synthase (pdb ID: 7TKM)⁹ (Extended Data Fig. 5E, F and G). Counterclockwise rotation of the central rotor in Conf-3 by 35° without apparent changes in the catalytic sites, results in Conf-1 (Extended Data Fig. 5E and H). Therefore, Conf-3 and Conf-1 are similar in the F₁ region, such that the arrangement of catalytic sites in both conformations is consistent with the binding dwell (Extended Data Fig. 7), but they differ in rotation of the c-ring by 35°, which roughly corresponds to one c-subunit (Extended Data Fig. 8 and Supplementary Movie 3). When viewed in relationship to Conf-0, the central rotor is rotated by 35° in the direction of ATP hydrolysis – the same direction as that from Conf-0 to Conf-1 (Extended Data Fig. 5B and E). Given that both Conf-1 and Conf-3 represent the binding dwell, this indicates that the transition from catalytic dwell to binding dwell only requires the central rotor to rotate by 35°, a finding that is consistent with recent observations^9,16,27,30. While the conformations of the catalytic sites in Conf-3 align well with Conf-1 (Extended Data Fig. 7), the F₁ region in Conf-3 does not show sufficient resolution to assign nucleotides in the catalytic sites.

Figure 4. Conf-3 and its comparison with Conf-1 and Conf-0.

A, Cryo-EM map of yeast ATP synthase in Conf-3 with individual subunits and domains color coded. B, Corresponding model for Conf-3. C, Comparison of the central rotor in Conf-0, Conf-1 and Conf-3. D, Sectional view of F₁ (as observed from the membrane) showing the angular rotation of the central rotor in Conf-0, Conf-1 and Conf-3 with respect to each other. “PS” refers to peripheral stalk.

Discussion.

We previously reported the high-resolution cryo-EM structure of ATP synthase using a modified version in which the δ-subunit of the central rotor was linked to the F6 subunit of the peripheral stalk by a T4-lysozyme molecule². As a result of our rotor-stator fusion, the enzyme predominantly occupied the State 2 conformation, as defined originally for bovine ATP Synthase¹¹. Here we continue to use this approach to resolve conformational intermediates at pH 6.

We came across several unexpected findings in this study. First, despite the fusion of the rotor and stator, hydrolysis of ATP at pH 8 caused rotation of the central stalk by 120°, going from State 2 to State 1 (Fig. 5, Table 2). Thus, the fusion linkage was flexible enough to allow rotation of the central rotor by 120°. However, the major conformation remained the catalytic dwell, albeit in two different rotational states (State 1 and State 2). We believe that the combined flexibility of the fusion construct and the peripheral stalk allows this rotation. In hindsight, this is not surprising since the peripheral stalk has been shown to change conformations during the hydrolysis of ATP, thus acting to store elastic energy⁹. It should be noted that the designation of the conformations as the catalytic dwell or binding dwell is based on the conformations and nucleotide content of the catalytic sites, and not on the position of the central rotor.

Figure 5. Conformations of ATP Synthase resolved in this study, in context of the reaction cycle.

Sectional views as seen from the membrane towards F₁ are depicted. Conf-0 (bottom) is the ground state (State 2, catalytic dwell). Counterclockwise rotation of the central rotor by 35° and concomitant conformation changes in the catalytic sites, results in Conf-3 (State 2, binding dwell). Additional counterclockwise rotation of the central rotor by 35° with no apparent change in the catalytic sites results in Conf-1 (binding dwell). Distortion of γ without apparent rotation of the central rotor leads to conformational changes in the catalytic sites, resulting in Conf-2 (State 1, catalytic dwell). The transition to State-2 might be an off-cycle phenomenon, that is observed more readily at low pH. Counterclockwise rotation of the central rotor in Conf-0 by 120° with concomitant changes in the catalytic sites leads to the State 1 catalytic dwell (top). The conditions for structure determination are labelled and the rotational angle of γ-subunit is indicated with grey arrows. Blue and green arrows denote the direction of ATP hydrolysis and ATP synthesis, respectively (not to suggest that ATP synthesis occurs at pH 6).

Table 2. Summary of the structures resolved in this study.

The “Rotational State” is in reference to the States defined for the bovine enzyme11. “Rotation” refers to the rotation of the central rotor in reference to State 2, with a negative value in the direction of ATP hydrolysis. “Sites” are defined as O (open), C (closed), and HO (half open). “Catalytic State” refers to the reaction coordinate and is defined by the conformations and nucleotide occupancy of the active sites. ND: Not determined as the resolution was not sufficient to determine the identity of the nucleotides bound.

Conformation	Rotational state	Rotational angle	Catalytic sites	Catalytic state	Nucleotides
Conf-0	State 2 Catalytic dwell	0	O,C,C	Catalytic dwell	ADP x2
Conf-3	State 2 Binding dwell	−35°	O,C,HO	Binding dwell	ND
Conf-1	State 2 Binding dwell	−70°	O,C,HO	Binding dwell	ADP, ADP+Pi
Conf-2	State 1 Catalytic dwell	−70°	C,C,O	Catalytic dwell	ADP x2
With ATP:Mg	State 1 Catalytic dwell	−120°	C,C,O	Catalytic dwell	ADP x2

Second, at pH 6 but in the absence of exogeneous ATP, we were able to see changes in the rotational position of the central stalk relative to Conf-0, by 35° (Conf-3) and additionally by 70° (Conf-1 and Conf-2), both in the direction of ATP hydrolysis (Fig. 5, Table 2). In all cases, the peripheral stalk was less resolved than that in Conf-0, suggesting that the added tension resulted in its conformational heterogeneity.

Third, the drop in pH had a remarkable shift in the conformational landscape of the ATP synthase. For the same rotor-stator fused construct of ATP synthase, a single conformation (Conf-0) was observed at pH 8, but three additional conformations (Conf-1, Conf-2 and Conf-3) have been observed at pH 6. In addition, while the catalytic dwell was the predominant structure (Conf-0) at pH 8, the binding dwell conformation (Conf-1) was most prevalent at pH 6.

Fourth, although both Conf-1 and Conf-2 have the central rotor rotated by 70° in the direction of ATP hydrolysis as compared to Conf-0, unlike Conf-1 representing the binding dwell, Conf-2 is in the catalytic dwell conformation (Fig. 5, Table 2). The differences in the catalytic site conformations between Conf-1 and Conf-2, are caused by distortion of the γ-subunit (Supplementary Movie 2), rather than the expected rotation of the central rotor. This is consistent with the report by Martin et al.³⁰ which showed direct evidence that elastic energy stored in the coil-coiled of the γ-subunit was utilized during the catalytic cycle. In Conf-2, the γ-subunit is bent from its canonical conformation, altering its interactions with the core of F₁, including the “catch regions”^31,32 which accounts for conformational changes in the catalytic sites. Conf-1 and Conf-2 correspond to the binding dwell of State 2 and the catalytic dwell of State 1, respectively (Fig. 5, Table 2). To the best of our knowledge, this is the first report of the State 1 catalytic dwell conformation observed at 70° counterclockwise rotation from the State 2 catalytic dwell (Conf-0). It is possible that our observation is specific to low pH and that Conf-2 is not stable enough to resolve at pH 8. Alternatively, this could be akin to a transition state that is extremely short during a regular ATP synthesis/hydrolysis cycle and observed here because of the low pH coupled with the linkage between the central rotor and the peripheral stalk. Based on our observations, we hypothesize that at pH 6, the −70° rotational position is a transition point between the binding dwell of State 2 and the catalytic dwell of State 1. Taken together, comparison between Conf-1 and Conf-2 suggests that the γ-subunit stores elastic energy that allows the observed conformational switch which is consistent with the elastic coupling mechanism of Martin et al.³⁰.

Fifth, as seen in Conf-3, the counterclockwise rotation of the central rotor by 35° with respect to Conf-0, was sufficient to switch conformations from catalytic dwell to binding dwell, with both conformations in State 2. This observation is consistent with the results from single molecule^13–15,27 and cryo-EM studies^9,16. Additionally, we observed the State 2 binding dwell conformation (Conf-1) at a 70° counterclockwise rotational angle from Conf-0, the catalytic dwell of State 2 (Fig. 5, Table 2). To the best of our knowledge, this is the first report of a binding dwell conformation observed at −70° from the corresponding rotational state’s catalytic dwell.

The established cycle in the direction of ATP hydrolysis, starting from the ground state (State 2), is State 2 catalytic dwell, to State 2 binding dwell, to State 1 catalytic dwell, to State 1 binding dwell, with rotations of the central stalk by 40° for transition out of the catalytic dwell and 80° for transition out of the binding dwell, within each reaction cycle^{9,14,16,27,30}. Fig. 5 and Table 2 summarize the conformations we observed, set in the context of the reaction cycle. The rotations and configurations we observed are State 2, catalytic dwell (0°), State 2, binding dwell (−35°), State 2, binding dwell (−70°) ↔ State 1, catalytic dwell (−70°), State 1, catalytic dwell (−120°). While some of these steps are seemingly new, single molecule studies using the E. coli enzyme show that at pH 6, 11° sub-steps can be observed throughout the reaction cycle²³. Thus, the rotational conformations we observe can be accounted for. The specific insight from our current study is that the transition of the active sites from catalytic dwell to binding dwell and then to the catalytic dwell of the next State, are not necessarily fixed to 40° and 80° rotation of the central rotor.

We were surprised to see such dramatic changes in the rotational state of the central stalk with change in pH. There are at least two possible explanations for the large movement of the rotor without the apparent input of energy. We hypothesize that the fusion of the central rotor and the peripheral stalk, by restricting rotation of the c-ring, results in a strained conformation that stores elastic energy. The lower pH might have simply reduced the energy differences between the various conformations allowing movement driven by the stored elastic energy. Another source of energy could be a result of the mismatch between the c-ring and the position of the central stalk of F₁. In this case, the energy would be stored in the peripheral stalk as indicated by its distortion under ATP hydrolyzing conditions (Extended Data Fig. 1C), also supported by earlier studies⁹. A less likely reason might be that small amounts of ATP might have co-purified with the enzyme. At pH 8, most of this ATP gets hydrolyzed and ATP synthase reaches its ground state of Conf-0. However, at pH 6, the enzyme has a higher tendency to get stalled in intermediate conformations²², potentially due to generation of new energy minima caused by protonation of residues. It is possible that in the process of hydrolyzing co-purified ATP, the enzyme gets stalled in these low energy states, thereby allowing the resolution of Conf-1, 2 and 3.

To accommodate the symmetry mismatch between F₁ containing three catalytic sites and F_o with ten c-subunits, the central rotor and/or the peripheral stator are required to be flexible^8,9. A comparison of the structures of F₁F_o in Conf-1 and Conf-2 enable the visualization of how subtle bending of the N- and C-terminal helices of γ, without changing the rotational position of the central rotor, causes large-scale conformational changes in the catalytic sites. In contrast, comparison of Conf-1 and Conf-3 indicates that rotation of the central rotor by 35° does not elicit a conformational change in the catalytic sites. It has been proposed that the central rotor stores the torque generated by rotation of the c-ring, until sufficient energy for the generation of ATP is accumulated^7,8,33, while other studies give this role to the peripheral stalk⁹ or to its components, such as the OSCP¹⁰. The conformational changes between Conf-1, Conf-2 and Conf-3 support the first hypothesis such that, rotation by 35° accumulates energy, that is released by bending of the N- and C-terminal helices of γ. Taken together, these observations highlight how the flexibility of γ facilitates efficient coupling between the 36° stepping motion of the c-ring and the 3-fold symmetry of the F₁ catalytic sites.

Lastly, for the structures obtained at pH 6, we see little evidence for aberrant changes in the ATP synthase that might be due to mildly acidic conditions. That is not to say that the novel structures are in the normal reaction pathway for ATP synthesis, at least at pH 8. Indeed, it is possible that Conf-2 is off cycle and potentially a result of bridging the rotor and stator coupled with altered reaction kinetics due to low pH. However, Conf-1 and probably Conf-2 are likely present under acidic conditions in the cell. In vivo, ATP synthase is exposed to acidic pH under hypoxic conditions³⁴. Under acidic conditions, the natural inhibitor protein (IF1) undergoes a conformational change to inhibit F₁-ATP hydrolysis³⁵ with bovine IF1³⁶ and yeast IF1³⁷ most inhibitory at acidic pH. The pH dependence of IF1 suggests that the matrix pH is acidic under stressful conditions, such as hypoxia. In addition, the tissue pH in regional ischemia in canine heart undergoes rapid acidification with concomitant decrease in oligomycin sensitive ATPase activity^38,39. Therefore, the structures reported here are consistent with being physiologically important.

Methods

Purification and reconstitution of ATP synthase into lipid nanodiscs.

The methods used for the expression, purification, and reconstitution of the yeast ATP synthase into nanodiscs are as reported earlier². Briefly, yeast submitochondrial particles were suspended in buffer containing 1% dodecyl-β-D-maltoside (DDM) and F₁F_o was purified using Ni-affinity chromatography. F₁F_o was reconstituted in lipid nanodiscs using membrane scaffold protein (MSP1E3D1) and POPC:cardiolipin (7.5:1 w/w). Reconstituted F₁F_o in lipid nanodiscs was subjected to size exclusion chromatography using a Superose-6 column (GE Healthcare) in buffer containing 20mM Tris, 50mM sucrose, 2mM MgSO₄, 1 mM EDTA, 150mM NaCl, 5mM benzamidine, 5 mM 6-aminocaropic acid, 0.25mM ADP, 0.05% DDM. The buffer was changed to pH 6.0 buffer (20 mM MES (4-Morpholineethanesulfonic acid), pH 6, 150 mM NaCl, 0.5 mM EDTA, 1 mM MgCl₂, 0.25 mM ADP) using a centrifuge column with Biogel P6 resin⁴⁰. The sample was flash frozen in liquid nitrogen and shipped on dry ice.

Cryo-EM data acquisition.

To collect the cryo-EM dataset of ATP Synthase in lipid nanodiscs at pH 6, 2.5 μl of the enzyme at a concentration of 0.5 mg/ml was applied to graphene oxide (GO) coated grids. The procedure for GO grid preparation was adapted from⁴¹. Briefly, GO (Sigma 763705) at a concentration of 0.2 mg/ml was sonicated in a water bath for 1 minute and centrifuged at 1000 g to remove aggregates. Quantifoil R2/1 on 400 mesh copper grids were glow discharged at 40 mA for ~120 seconds and incubated with 3 μl of GO for 1 minute. Subsequently, the grid was blotted, washed with water, and dried for 5 to 10 minutes. 2.5 μl of the protein sample was applied to the GO coated grid and blotted for 5.5 seconds with 100% humidity before plunge freezing in liquid ethane using a Vitrobot Mark IV system (Thermo Fisher Scientific). Cryo-EM images were collected using a Titan Krios G3i operated at 300 kV, equipped with a Falcon IV detector (Thermo Fisher Scientific) at the Stanford-SLAC Cryo-EM Center (S²C²) using EPU (version 2.9). Images were recorded at a nominal magnification of x75,000, corresponding to a pixel size of 1 Å, with the total dose of 52 e/Ų.

To collect the cryo-EM dataset of ATP Synthase in lipid nanodiscs in presence of Mg:ATP, the purified enzyme at a concentration of 1 mg/ml at pH 8 was incubated with 1 mM ATP, 1mM MgCl₂ at room temperature for 10 min. After incubation, 2.5 μl mixture was applied to a glow-discharged Quantifoil holey carbon grid (1.2/1.3, 400 mesh), and blotted for 3 seconds with 90% humidity before plunge-freezing in liquid ethane using a Cryoplunge 3 System (CP3, Gatan). Cryo-EM images were collected using a Titan Krios (Thermo Fisher Scientific) operating at 300 kV, equipped with a Gatan K2 Summit direct electron detector at the National Center for Cryo-EM Access and Training (NCCAT) using SerialEM (version 3.7)⁴². Images were recorded at a nominal magnification of x130,000, corresponding to a pixel size of 1.06 Å, with a total dose of 51.07 e/Ų.

Image processing.

Cryo-EM data was analyzed as previously described⁴³ with some modifications. Dose-fractionated movie stacks were motion-corrected using MotionCor2⁴⁴ and defocus was calculated using CTFFIND4⁴⁵. For both datasets, initial image processing was conducted using Simplified Application Managing Utilities for EM Labs (SAMUEL v21.01)⁴⁶. Particle images were selected using “samautopick.py” and 2D classification of selected particles was conducted using “samclasscas.py”, “samtree2dv3.py” or RELION-3.0⁴⁷. Image processing in 3D, including classification, domain masking, multibody refinement²⁴, signal subtraction and global and local refinement was conducted using RELION-3.0 (details in Extended Data Figs. 1 and 2). Overall resolution of cryo-EM maps was computed according to the gold standard Fourier Shell Correlation (FSC) method, with the determination of resolution based on the correlation between independent half-maps at FSC=0.143. Local resolution was determined using postprocessing in RELION-3.0 (Extended Data Fig. 3).

Model Building and Validation.

The cryo-EM structure of yeast ATP Synthase (pdb ID 6CP6)² was fit into the map obtained for Conf-0 and served as a reference for building the model for Conf-1 and Conf-2. The corresponding domain of the model was fitted as a rigid body in the high-resolution F₁-focused maps, F_o-focused map and F_o+CR-focused maps followed by manual adjustments in Coot (version 0.9.1)⁴⁸. Models for individual domains were then fit into the consensus maps for Conf-1 and Conf-2 respectively. The complete models for F₁F_o were refined against the consensus maps using Phenix (version 1.20.1)⁴⁹. The model for Conf-1 was fitted as a rigid body into the F_o+CR and F₁F_o maps for Conf-3, manually adjusted using Coot and refined using Phenix. The model was converted to poly-alanine chains using the pdb modifications tool in Phenix. The model for F₁F_o in Conf-0 (pdb ID 6CP6)² was used as a reference for building the model for ATP synthase in presence of exogeneous ATP. Models were fit into the F₁-focused map and consensus map followed by manual adjustment using Coot and refinement using Phenix. Maps and models were visualized using UCSF Chimera (version 1.12)⁵⁰ with distances and rotational angles being measured using the structure analysis tool. Overall map quality and fit of the corresponding model is demonstrated in Extended Data Fig. 4. Figures and movies were generated using UCSF Chimera and ChimeraX (version 1.4)⁵¹.

Data availability:

Three-dimensional cryo-EM density maps of the yeast mitochondrial ATP synthase in nanodiscs have been deposited in the Electron Microscopy Data Bank under accession numbers EMD-29270, EMD-29278, EMD-28685, EMD-28687, EMD-28689, EMD-28809, EMD-28811, EMD-28813, EMD-28814, EMD-28835, EMD-28836, EMD-28837, EMD-29250, EMD-29251. The corresponding atomic coordinates for the atomic models have been deposited in the Protein Data Bank under accession numbers 8FL8, 8F29, 8F39 and 8FKJ (Table 1). PDB ID 6CP6 was used as an initial model for model building.

Extended Data

Extended Data Fig. 1. Cryo-EM of ATP synthase in presence of Mg:ATP.

A, Representative micrograph (from a dataset of 4219 images) of ATP synthase in lipid nanodiscs under continuous ATP hydrolysis. B, Two-dimensional class averages of F₁F_o. C, Image processing workflow of the dataset (see Methods for details). D, Local resolution (left) and angular distribution (right) of the consensus and focused maps. E, Fourier shell correlation (FSC) between the half maps with resolution at FSC=0.143 indicated. F, Model for yeast ATPase in the rotational state 2 (pdb ID 6CP6) fit into the map for the minor conformation (20% of particle images), with sectional view showed in G. Consensus map (H) and model (I) of the major conformation. J, Sectional view (as observed from the membrane) of the major conformation (ribbon) overlaid with the structure of bovine ATP synthase in rotational state 1 (pdb ID 5ARA) depicted as transparent surface. In panels G and J, the rotational angle of the central rotor is designated by a black arrow. “CR” refers to central rotor.

Extended Data Fig. 2. Cryo-EM image processing workflow.

A, Representative micrograph (from a dataset of 19,744 images) of ATP synthase in lipid nanodiscs at pH 6. B. 2D class averages of F₁F_o C, Image processing workflow resulting in cryo-EM maps for Conf-0 (green), Conf-1 (red). D, Image processing workflow focused on F₁ resulting in cryo-EM maps for Conf-1 (red) and Conf-2 (blue). E, F_o-centered image processing workflow resulting in cryo-EM maps for Conf-0 (green) and Conf-3 (pink). “CR” refers to central rotor.

Extended Data Fig. 3. Analysis and Statistics of Cryo-EM maps.

Local resolution (left) and angular distribution (right) of Conf-1 consensus map (A), Conf-1 F₁ focused map (B), Conf-1 F_o+CR (central rotor) focused map (C) and Conf-1 F_o-focused map (D). E, Fourier shell correlation (FSC) between the half-maps for Conf-1 with resolution at FSC=0.143 indicated. Local resolution (left) and angular distribution (right) of Conf-2 consensus map (F), Conf-2 F₁- focused map (G) and Conf-1 F_o+CR focused map (H). I, Fourier shell correlation (FSC) between the half-maps for Conf-2 with resolution at FSC=0.143 indicated. Local resolution (left) and angular distribution (right) of Conf-3 consensus map (J), and Conf-3 F_o+CR focused map (K). L, Fourier shell correlation (FSC) between the half-maps for Conf-3 with resolution at FSC=0.143 indicated.

Extended Data Fig. 4. Representative density from high-resolution maps with the corresponding model fit into the density.

for Conf-1 (A) and Conf-2 (B)

Extended Data Fig. 5. Comparison of observed conformations at pH 6 with existing catalytic and binding dwell structures.

A, Sectional view across F1 of Conf-0 (pdb ID 6CP6) fit into the map of bovine ATP synthase in State 2 (EMD-3166). B, Sectional view across F1 of Conf-0 showing catalytic site conformations and rotational angle of the central rotor (indicated with gray arrow) similar to catalytic dwell structures of Bacillus PS3 (pdb ID 7LIR) (C) and Sc (pdb ID 7TKR) (D). E. Sectional view across F₁ of Conf-3 showing catalytic site conformations and rotational angle of the central rotor similar to binding dwell structures of Bacillus PS3 (pdb ID 7L1Q) (F) and Sc (pdb ID 7TKM) (G). H, Sectional view across F₁ of Conf-1 with catalytic sites conformations consistent with binding dwell but rotational angle of the central rotor is not. I, Sectional view across F1 of Conf-2 with catalytic sites conformations consistent with catalytic dwell but rotational angle of the central rotor is not. Rotational angle of the central rotor in each structure is indicated by gray arrows and the rotation from catalytic dwell to binding dwell to Conf- 1 and Conf-2, is indicated in blue. Sc refers to Saccharomyces cerevisiae.

Extended Data Fig. 6. Comparison of the a/c interface between Conf-1 and Conf-2.

Cryo-EM map and model from F_o domain’s a/c interface in Conf-1 (A) and Conf-0 (pdb ID 6CP6 fit into the map for Conf-0) (B). The pdb chain IDs of individual c-subunits have been labeled (letters K-T). C, Overlay of Conf-1 (blue) and Conf-0 (green).

Extended Data Fig. 7. Structure of the F1 domain from Conf-1 docked into the map for Conf-3.

A, Cryo-EM map for conf-3. B, Model for F₁ domain and peripheral stalk in Conf-1 docked into the corresponding cryo-EM density for Conf-3 (transparent), shown here as a sectional view.

Extended Data Fig. 8.

Comparison of the model from F_o domain’s a/c interface between conf-1 (A) and conf-3 (B). The pdb chain IDs for individual c-subunits have been denoted as letters (K-T).

Supplementary Material

Movies

Extended Data Movie 1. Conformational transition from Conf-0 to Conf-1. The animation shows a morph from Conf-0 to Conf-1 with subunits color coded as the manuscript figures.

Extended Data Movie 2. Conformational transition from Conf-1 to Conf-2. The animation shows a morph from Conf-1 to Conf-2 with subunits color coded as the manuscript figures.

Extended Data Movie 3. Conformational transition from Conf-1 to Conf-3. The animation shows a morph from Conf-1 to Conf-3 with subunits color coded as the manuscript figures.