How release of phosphate from mammalian F1-ATPase generates a rotary substep.

The rotation of the central stalk of F1-ATPase is driven by energy derived from the sequential binding of an ATP molecule to its three catalytic sites and the release of the products of hydrolysis. In human F1-ATPase, each 360° rotation consists of three 120° steps composed of substeps of about 65°, 25°, and 30°, with intervening ATP binding, phosphate release, and catalytic dwells, respectively. The F1-ATPase inhibitor protein, IF1, halts the rotary cycle at the catalytic dwell. The human and bovine enzymes are essentially identical, and the structure of bovine F1-ATPase inhibited by IF1 represents the catalytic dwell state. Another structure, described here, of bovine F1-ATPase inhibited by an ATP analog and the phosphate analog, thiophosphate, represents the phosphate binding dwell. Thiophosphate is bound to a site in the α(E)β(E)-catalytic interface, whereas in F1-ATPase inhibited with IF1, the equivalent site is changed subtly and the enzyme is incapable of binding thiophosphate. These two structures provide a molecular mechanism of how phosphate release generates a rotary substep as follows. In the active enzyme, phosphate release from the β(E)-subunit is accompanied by a rearrangement of the structure of its binding site that prevents released phosphate from rebinding. The associated extrusion of a loop in the β(E)-subunit disrupts interactions in the α(E)β(E-)catalytic interface and opens it to its fullest extent. Other rearrangements disrupt interactions between the γ-subunit and the C-terminal domain of the α(E)-subunit. To restore most of these interactions, and to make compensatory new ones, the γ-subunit rotates through 25°-30°.

The ATP synthase in bovine mitochondria is a membrane-bound protein assembly of about 30 polypeptides of 18 different kinds with a combined molecular mass of about 650 kDa (1, 2). They are organized into a membrane extrinsic globular F1-catalytic domain and a membrane intrinsic domain joined together by central and peripheral stalks. The F1-domain is an assembly of three α-subunits and three β-subunits arranged in alternation in a spherical complex around an α-helical structure in the γ-subunit (3), and the three catalytic sites of the enzyme lie at three of the six interfaces between the α- and β-subunits. The γ-subunit extends from the α3β3-spherical structure to the membrane domain of the enzyme, where it is augmented by the δ- and ε-subunits in the region of contact with a membrane-embedded ring of eight c-subunits (4). The γ-, δ-, and ε-subunits form the central stalk, and together with the c-ring, it constitutes the rotor of the enzyme. During ATP synthesis, the rotor turns clockwise as viewed from the membrane domain of the complex. Rotation is driven by the potential energy of a transmembrane proton motive force generated by respiration, and the rotor transmits energy into the F1-domain, bringing about the conformational changes in the three catalytic sites that are required to produce ATP from ADP and phosphate.

The F1-domain has been studied extensively in isolation from the rest of the enzyme complex. It is a globular assembly that functions as an ATP hydrolase, and many structural studies of the eukaryotic F1-ATPase (3–11) and biophysical studies of bacterial enzymes (12–14) have illuminated how the hydrolysis of ATP at its three catalytic sites drives the anticlockwise rotation of the central stalk as viewed from the membrane domain of the intact enzyme. The catalytic sites are formed predominantly from residues in the β-subunits plus an essential contribution from a residue known as the “arginine finger” at position 373 in the adjacent α-subunits (15). During a 360° rotary cycle, each β-subunit passes through three states corresponding to the βE-, βTP-, and βDP-subunits observed in the “reference” structure of the enzyme, where E, TP, and DP correspond to the unoccupied, or “empty,” ATP binding and ADP binding conformations respectively (3).

The hydrolysis of ATP by F1-ATPase from mitochondria, but not from bacteria or chloroplasts, can be inhibited by the natural inhibitor of F1-ATPase, IF1 (16). Bovine IF1 is 84 aa long (17), and the active form is a homodimer held together by an antiparallel coiled-coil of α-helices from residues 49–81 (18). Its N-terminal region from residues 1–45 provides the inhibitory part of the protein. In the free inhibitor, this region is intrinsically disordered (19), and it becomes structured progressively as it binds to the enzyme (20) and inhibits ATP hydrolysis. A monomeric derivative of bovine IF1 containing residues 1–60 is also an effective inhibitor, and it was found to be bound to the C-terminal domains of both the α- and β-subunits in one, two, or all three catalytic interfaces of F1-ATPase (21, 22).

The rotation of the central stalk of F1-ATPase has been studied by direct observation of the tethered enzyme. In the F1-ATPase from the Bacillus stearothermophilus strain PS3, each 360° rotary cycle has been resolved at high concentrations of ATP (2 mM) into three 120° steps (14). The three intervening pauses are referred to as the “catalytic dwells,” where the enzyme is considered to be poised to carry out, or to be carrying out, the hydrolysis of ATP. At lower concentrations of ATP (2 μM), a second pause can be observed 40° after the catalytic dwell. It is known as the “ATP binding dwell,” where the enzyme is awaiting the binding of the substrate, ATP (14). The rotary 360° cycle of the human F1-ATPase also contains three catalytic dwells separated by 120°, and 30° after each catalytic dwell, three ATP binding dwells (23). Unlike the bacterial enzyme, both dwells in the human enzyme can be observed at both high (4 mM) and low (50 μM) concentrations of ATP. Inhibition of the human enzyme with human IF1 halts the rotary cycle at a position corresponding to the catalytic dwell (23). Thus, structures of the closely related bovine enzyme inhibited by bovine IF1 (21, 22) can be considered to be structural representations of the catalytic dwell. Moreover, upon addition of the ATP analog adenosine 5′-(β,γ-imido)triphosphate (AMP-PNP) and the phosphate analog monothiophosphate (referred to as “thiophosphate”), the rotating human enzyme stalls at an additional intermediate position 25° before the catalytic dwell (23). Although this intermediate stall position has been referred to as the “phosphate release dwell,” it differs from the catalytic and ATP binding dwells in that it has not been observed during each 120° step in an active catalytic cycle, and it is manifest only when rotation of the enzyme has been stopped with the inhibitors AMP-PNP and thiophosphate. It is reasonable to assume that this state corresponds to a point in catalysis after the cleavage of the β-γ phosphate bond of ATP, where phosphate is about to be released.

Therefore, as described here, we have determined a structure of bovine F1-ATPase with crystals grown in the presence of AMP-PNP and thiophosphate. To ascertain whether the bovine F1-ATPase inhibited with IF1 is capable of binding thiophosphate, we have also redetermined its structure with crystals of the inhibited complex grown in the presence of thiophosphate. Because these structures correspond to the prephosphate release and the catalytic dwell states of the enzyme, respectively, they provide the basis of a molecular explanation of the mode of binding of phosphate to F1-ATPase at the attendant dwell, and how the subsequent release of phosphate from the enzyme is coupled to the generation of the 25° rotary substep between the phosphate release and catalytic dwells.

Results

Structure Determination.

The structures of the two bovine F1-ATPase complexes, F1-ATPase inhibited with AMP-PNP and thiophosphate (known as F1-ThioP) and F1-ATPase inhibited with I1-60His-K39A and ATP () and the crystals grown in the presence of thiophosphate (known as F1-I3-ThioP), were determined by molecular replacement with data to 3.1 Å and 3.3 Å resolution, respectively. The asymmetric units of both crystals contain one F1-ATPase, and in the case of F1-I3-ThioP, three inhibitor proteins bound to the enzyme in catalytic interfaces. The data processing and refinement statistics are summarized in Table S1. The final model of F1-ThioP contains the following residues: αE, 22–510; αTP, 23–401 and 410–510; αDP, 23–510; βE, 9–387 and 396–474; βTP, 9–474; βDP, 9–475; γ, 1–49, 67–96, 107–148, 159–173, and 196–272; δ, 15–53 and 57–145; and ε, 1–47. An Mg-AMP-PNP complex is bound to the nucleotide binding sites in each of the three noncatalytic α-subunits and in the catalytic βDP- and βTP-subunits, and a thiophosphate ion, but no nucleotide, is bound to the βE-subunit. Likewise the final model of F1-I3-ThioP contains αE, 24–510; αTP, 16–402 and 411–510; αDP, 23–404 and 410–510; βE, 8–477; βTP, 9–477; βDP, 9–477; γ, 1–44, 73–91, 108–153, 164–173, and 205–272; I1-60E, 29–50; I1-60TP, 23–50; and I1-60DP, 10–50. However, the inhibited complex was formed by inhibiting the active F1-ATPase with IF1, and therefore in this structure, the nucleotide binding site in each of the three noncatalytic α-subunits is occupied by an Mg-ATP complex, both the βDP- and βTP-subunits contain a bound Mg-ADP moiety, and neither a nucleotide nor a thiophosphate ion is bound to the βE-subunit. Attempts to crystallize bovine F1-ATPase in the presence of phosphate and nucleotides were unsuccessful because the protein aggregated and precipitated.

Structure of the Bovine F1–ThioP Complex.

In many respects, the structure of F1-ThioP (Fig. 1) is similar to previous structures of bovine F1-ATPase. It resembles especially the ground state structure, where the βE-subunit is fully open and unoccupied by either substrate or hydrolysis products (6) and the superimposed structures have an rmsd of 0.6 Å for all atoms. In the initial refinement of the model of F1-ThioP, a strong positive electron difference density peak (>8σ) was revealed in the region of the catalytic site of the βE-subunit. Because the crystals were grown in the presence of thiophosphate, and in the absence of phosphate or sulfate, this density probably represents a bound thiophosphate ion, and its shape is consistent with a phosphorous atom, with three equivalent oxygen atoms and a sulfur atom at the fourth apex of an asymmetric tetrahedron facing toward βE-R189 (Fig. 2). Its three negatively charged oxygen atoms form ionic interactions with the guanidinium of βE-R260, the amido moiety of βE-N257, and the ε-amino group of βE-K162, and its sulfur atom makes a fourth ionic interaction with the guanidinium of residue βE-R189 (Fig. 2). In addition, there is a hydrogen bond between a thiophosphate oxygen atom, which is probably protonated, and the carboxylate group of βE-D256.

Fig. 1.

Structure of the bovine F1-ThioP complex. Cross-sectional views of the F1-ATPase from above and toward the membrane domain of the intact ATP synthase of the C-terminal domains of the α- and β-subunits (residues 380–510 and 364–474, respectively), arranged alternately around the γ-subunit (residues 1–47 and 206–270) (A) and from the side showing the αDP-subunit; the γ-, δ-, and ε-subunits; and the βE-subunit (B). The α-, β-, γ-, δ-, and ε-subunits are colored red, yellow, blue, magenta, and green, respectively. Bound nucleotides and a thiophosphate ion are colored black and orange, respectively.

Fig. 2.

Thiophosphate bound to the βE-subunit of the F1-ThioP complex. (A) Electron density of the 2FO-FC map of the thiophosphate binding site (contour level = 1.0 σ). The electron density (gray mesh) extends for a radius of 1.6 Å around the residues of the binding site. (B) Side chains of residues contributing to the thiophosphate binding site and their distances (in angstroms) from the bound thiophosphate. The side chains of residues of the βE-subunit and the thiophosphate are colored yellow and orange, respectively. Oxygen, nitrogen, and sulfur atoms are colored red, blue, and gold, respectively.

Structure of the Bovine F1–I3–ThioP Complex.

The structure of F1-I3-ThioP (Fig. S1) is essentially identical to the structure of F1-(I1-60His K39A)3 (21); the rmsd for all atoms of the two superimposed structures is 0.4 Å. Both complexes consist of F1-ATPase with three inhibitor proteins bound to the C-terminal α-helical domains of α- and β-subunits in the αEβE-, αTPβTP-, and αDPβDP-catalytic interfaces of the enzyme. As observed previously (21), in both complexes, the extent of the secondary structure of the inhibitor proteins resolved at each catalytic interface differs (Fig. S1). The least extensively defined inhibitor is the one at the αEβE-catalytic interface, where residues 29–50 were resolved. A more extensive section of the inhibitor protein from residues 23–50 was resolved at the αTPβTP-catalytic interface, and the inhibitor at the αDPβDP-catalytic interface had the most extensive secondary structure, extending from residues 10–50. There was no evidence of a thiophosphate ion bound to the βE-subunit of F1-I3-ThioP, and no bound anion (phosphate or sulfate) was found previously bound in the βE-subunit of F1-(I1-60His K39A)3 (21).

Discussion

Phosphate Binding Sites in F1-ATPase.

The thiophosphate binding site in F1-ThioP is equivalent to the site where sulfate has been observed to be bound previously in one of the three F1-ATPase complexes (“molecule” II) in the asymmetric unit of crystals of the ground state structure of F1-ATPase from Saccharomyces cerevisiae (8) (Fig. S2) and in a structure of bovine F1-ATPase inactivated covalently with dicyclohexylcarbodiimide (known as F1-DCCD) (7) (Fig. S3). In the structure of bovine F1-ATPase inhibited with ADP and aluminum fluoride (known as F1-AlF4), ADP and sulfate are bound to the half-closed βE-subunit, with the sulfate occupying the same position as the sulfate in F1-DCCD (15) (Fig. S3). In all of the other structures of bovine F1-ATPase, there is no evidence for either phosphate or sulfate being bound at a site equivalent to the site where thiophosphate is bound in F1-ThioP. However, in the structures of bovine F1-ATPase in the ground state (3), in the enzyme inhibited with beryllium fluoride (9) or azide (5), and in the complex with F1-ATPase and the peripheral stalk subcomplex (10), electron density in the βE-subunit adjacent to the phosphate binding loop (P-loop), was interpreted as either a phosphate or a sulfate ion (Fig. S3). The P-loop is a conserved feature of many NTPases, and is so-named because it interacts with phosphate moieties of bound NTP or NDP molecules (24). In these structures of bovine F1-ATPase, the anion binding site in the βE-P-loop is about 8 Å from where the γ-phosphate of the substrate ATP is bound in the catalytically active βDP-subunit, and from where phosphate is presumably released following scission of the bond between the β- and γ-phosphates. Currently, there is no experimental evidence supporting the involvement of a phosphate ion bound in the vicinity of the βE-P-loop of F1-ATPase in the mechanism of hydrolysis of ATP (8).

Changes in the Phosphate Binding Site in F1-ThioP and F1-I3-ThioP.

In the structure of F1-I3-ThioP, the site in the βE-subunit that is equivalent to the thiophosphate binding site in the F1-ThioP complex differs subtly from that site. In F1-ThioP, the thiophosphate interacts with βE-K162, βE-R189, βE-D256, βE-N257, and βE-R260 (Fig. 2), and residue βE-E188 is positioned directly above the bound thiophosphate in a loop (residues 187–189) at the αEβE-interface (Fig. 3). In contrast, in F1-I3-ThioP, the side chain of βE-E188 points “downward” (Fig. 3); its carboxylate group has moved by 7 Å and now occupies the position occupied by the guanidinium of βE-R260 in F1-Thio-P (Fig. 3). In addition, in F1-I3-ThioP, both βE-R189 and βE-R260 have moved away from the thiophosphate binding site by 1.5–2.0 Å relative to their positions in F1-ThioP (Fig. 3). This rearrangement has prevented thiophosphate from binding to F1-I3-ThioP, and is associated with a concerted movement of residues βE-Q221, βE-N223, and βE-E188, and the extrusion of the loop (residues βE-G187 to βE-R189), toward the αEβE-interface (Fig. 4 and Movie S1). As a result, overall, the αEβE-interface is more open in F1-I3-ThioP than in F1-ThioP (accompanying identification of structural features is provided in Fig. S4 and Movie S2), although residues αE130-αE136 in F1-I3-ThioP are closer to the βE-subunit than in F1-ThioP. The rmsd for all atoms for the αEβE dimers of the F1-ThioP and F1-I3-ThioP complexes is 2.1 Å.

Fig. 3.

Comparison of the structures of the thiophosphate binding site in the F1-ThioP complex and of the equivalent sites in complexes of bovine F1-ATPase inhibited with the natural inhibitor protein, IF1. F1-ThioP (A) and F1-I3-ThioP (B) are shown. The residues of the surrounding βE-subunit, the Arg finger residue αER373, and the thiophosphate are colored yellow, pink, and orange, respectively. Oxygen, nitrogen, and sulfur atoms are colored red, blue, and gold, respectively. The distance of the salt bridge interaction between βE-E188 and βE-R260 is shown in angstroms.

Fig. 4.

Comparison of the structures of a loop region in the vicinity of the thiophosphate binding site in F1-ThioP with the equivalent loop in F1-I3-ThioP. The loop, residues 187–189 of the βE-subunit, contains residue βE-E188. The αE- and βE-subunits and their structural elements are depicted in cartoon representation in red and yellow, respectively. (A) View toward the interface between the αE- and βE-subunits with the bound thiophosphate shown in orange and, in the box, an expanded view of the thiophosphate binding site. (B) Equivalent expanded view of the site in F1-I3-ThioP. In A and B, residues βE-E188 and βE-R260 are shown in stick representation, with oxygen and nitrogen atoms shown in red and blue, respectively.

This interpretation of the differences in the phosphate binding site in F1-ThioP and F1-I3-ThioP is supported by the structures of bovine F1-ATPase inhibited by two or three copies of the inhibitor protein (21). In all three structures, the side chain of βE-E188 is in the “downward” conformation. However, in the historically first determined structure of bovine F1-ATPase inhibited with residues 1–60 of bovine IF1 (22), the concerted side-chain movements of residues βE-E188, βE-R260, βE-Q221, and βE-N223 were not observed. This difference may be accounted for by increased crystal lattice contacts arising from the controlled dehydration and shrinkage of the unit cell of the crystals of this particular complex before collection of X-ray diffraction data. This dehydration procedure was not used in any of the subsequent structure determinations of F1-inhibitor protein complexes.

Changes in the γ-Subunit in F1-ThioP and F1-I3-ThioP.

In the structures of F1-ThioP and the ground state structure of F1-ATPase (6), the γ-subunit is found in the same position. In contrast in F1-I3-ThioP and all of the other structures of F1-ATPase with a bound inhibitor protein, including the first structure of the F1-inhibitor complex (22), the γ-subunits are in a different common position where, relative to F1-ThioP, the subunit has rotated by ∼30° in an anticlockwise direction as viewed from the membrane domain of the intact ATP synthase (Table S2). This rotation is accompanied by changes in the mode of interaction of the γ-subunit with the α- and β-subunits close to where the γ-subunit emerges from the α3β3-domain (Tables S3 and S4; see below). However, other interactions, between the γ-subunit and the “bearing” region in the “crown” domain involving the N-terminal regions of α- and β-subunits are unchanged.

Release of Phosphate and Generation of the 25° Rotary Substep in Mammalian F1-ATPase.

The sequences of human and bovine α-, β-, and γ-subunits are 99%, 99%, and 93% identical, respectively. Thus, bovine F1-ThioP and bovine F1-I3-ThioP are structural representations of F1-ATPase in the phosphate release dwell and the postphosphate release states of bovine and human F1-ATPases, respectively, and they provide a molecular mechanism of the coupling of phosphate release to the 25° rotary substep. The observation that the γ-subunit in the structure of F1-I3-ThioP and related structures is 30° from its position in F1-ThioP (as viewed from the membrane domain of the intact ATP synthase; Table S2) is consistent with this interpretation, and the 30° step observed in the structures is considered to be equivalent to the 25° substep observed in rotational experiments.

In this mechanism, the release of phosphate from the βE-catalytic site is accompanied by the movement of the side chains of βE-E188, βE-R189, and βE-R260; the Glu toward the phosphate binding site, and the Args away from that site (Fig. 4 and Movie S1). This rearrangement hinders the rebinding of the released phosphate and helps to drive the rotation forward. It is also associated with the extrusion toward the αE-subunit of the loop region from residues 187–189 in the βE-subunit (Fig. 4 and Movie S1). The movement of this loop is accompanied by a rotation of the nucleotide binding and C-terminal domains of the αE-subunit (by 7° and 11°, respectively) relative to the nucleotide binding domain of the βE-subunit, resulting in the opening of the αEβE-interface in F1-I3-ThioP (Movie S2). In the absence of the rotation of the γ-subunit, this rotation of the C-terminal domain of the αE-subunit would remove favorable packing (van der Waal’s) interactions between the helix–turn–helix region of the αE-subunit (residues 402–409) and the N-terminal α-helix of the γ-subunit (residues 18–30). However, in F1-ThioP, these interactions have been retained by a 30° rotation of the γ-subunit (Fig. 5 and Movie S3). The interaction area between the γ- and αE-subunits is 650 Å2 in both F1-ThioP and F1-I3-ThioP. However, if the γ-subunit in the F1-I3-ThioP complex is rotated artificially by 30° (modeled by using the αE-subunit from F1-I3-ThioP and the γ-subunit from F1-ThioP), the interaction area between the γ- and αE-subunits is reduced to 407 Å2. Thus, the binding energy gained from these new interactions, and from the formation of the salt bridge between βE-E188 and βE-R260 (Fig. 3), would help to compensate for the loss in binding energy associated with phosphate release.

Fig. 5.

Comparison of interactions between the γ-subunit and the α- and β-subunits in the F1-ThioP and F1-I3-ThioP complexes. The blue α-helical regions are residues 1–41 and 211–253 of the γ-subunit. (A–D) F1-ThioP and F1-I3-ThioP complexes, respectively; the residues in the γ- and αE-subunits involved in these contacts are shown as light blue and red spheres, respectively. Views from the side (A and B) and from below (C and D) are depicted.

Order of Release of the Products of Hydrolysis by F1-ATPase.

The order of the release of the products of ATP hydrolysis from F1-ATPase is an unresolved issue. It has been proposed on the basis of single-molecule rotation experiments conducted on a bacterial F1-ATPase that phosphate is released from the βE-subunit after release of ADP (25, 26). Similar conclusions (27, 28) have been reached from molecular dynamics calculations based on bovine structural data (6) and bacterial rotational data (14). However, the structural data, derived almost entirely from eukaryotic enzymes, are ambiguous. The current structure of F1-ThioP and the structure of “molecule” II of yeast F1-ATPase (8) can be interpreted as being consistent with the release of ADP and the magnesium ion before the release of phosphate from the βE-subunit. However, the structure, known as F1-PH, of bovine F1-ATPase crystallized in the presence of the magnesium ion chelator, phosphonate, contains an ADP molecule only, without an accompanying magnesium ion or phosphate, bound to a fully open βE-subunit (11), suggesting that the release of both phosphate and magnesium (in an unspecified order) precedes the release of ADP. This order is consistent with the order of product release from other P-loop–containing nucleotide hydrolases, such as kinesin (29), G proteins (30), and myosin (31). In F1-PH, the γ-subunit is rotated by 30.5° relative to the γ-subunit in F1-ThioP, to a position similar to the position of the γ-subunit in F1-I3-ThioP. Therefore, it appears from this structure that ADP can remain bound to bovine F1-ATPase after the release of phosphate. Further experiments are required to determine whether thiophosphate will bind to F1-PH and whether ADP will bind to F1-ThioP.

Another issue relating to product release from F1-ATPase, where there are conflicting data, concerns the identity of the subunit from which phosphate is released. The data presented here, together with rotational studies on the human F1-ATPase (23), show clearly that phosphate is released from the βE-subunit. In an alternative catalytic mechanism, it is proposed that phosphate is released from the βDP-subunit of the bacterial enzyme immediately following the hydrolysis of ATP and before the release of ADP (32). It is now apparent from rotational experiments that the catalytic cycles of human and bacterial F1-ATPases differ significantly, as indeed they differ from the catalytic cycle of the related V-type ATPases, where there are three 120° steps per cycle with no observable substeps (33). These differences are likely to underlie the difficulties encountered hitherto in reconciling structural data from the bovine enzyme with rotational data determined on bacterial enzymes. The current experiments demonstrate that the structural and rotational data from the same or closely related species are compatible. The remaining challenge is to provide molecular explanations of the generation of the rest of the 120° step in the catalytic cycle.

Materials and Methods

Purification of Proteins.

The bovine inhibitor protein I1-60His-K39A, comprising residues 1–60 of IF1 with a C-terminal hexahistidine sequence and containing the mutation K39A, and bovine mitochondrial F1-ATPase were purified as described previously (21).

Crystallization of the F1-ThioP and F1-I3-ThioP Complexes.

The growth of crystals, the dehydration procedure, and freezing conditions are described in .

Data Collection and Processing.

Diffraction data for the F1-ThioP complex were collected to 3.1 Å on a Pilatus 6M detector (Dectris) on beamline I02 (fixed wavelength of 0.979 Å and beam size of 20 μm by 70 μm) at the Diamond Light Source (Harwell, UK). Diffraction data for the F1–I3–ThioP complex were collected to 3.3 Å on a Mar/Rayonix 3 × 3 mosaic 225 detector (Rayonix) on beamline ID23-2 (fixed wavelength of 0.87 Å and beam size of 8 μm by 8 μm) at the European Synchrotron Radiation Facility (Grenoble, France). Diffraction images were integrated with MOSFLM (34), and the data were reduced with AIMLESS (35). The structure of F1-ThioP was solved by molecular replacement with PHASER (36) using the structure of bovine F1-ATPase inhibited by azide (Protein Data Bank ID code 2CK3) without any of the bound nucleotides as the starting model. The structures were remodelled manually with Coot (37), and alternate rounds of rebuilding and refinement were carried out with REFMAC5 (38). The stereochemistry of each model was assessed with Coot and MolProbity (39). Images of the structures and electron density maps were generated with PyMOL (40). Surface areas of interaction between different subunits of F1-ATPase were calculated with 'Protein interfaces, surfaces and assemblies' service PDBePISA (41).

Assessment of Positions of the γ-Subunit in Structures of Bovine F1-ATPase.

The azide-free ground state structure of bovine F1-ATPase determined at 1.9 Å resolution was aligned with the structures of bovine F1-ThioP, F1-I1-60His, F1-(I1-60His)2, and F1-(I1-60His K39A)3 via their “crown” domains, consisting of residues 30–80 and 15–75 in the α- and β-subunits, respectively. In these structures, residues 22–32 of the γ-subunit interact directly with the C-terminal domains of the α- and β-subunits, and the position of this segment of the γ-subunit acts as a rigid body (Fig. S5) and should not be influenced by contacts in the crystal lattice between adjacent F1-ATPase complexes, whereas the positions of the regions of the γ-subunit (residues 33–226), and the associated δ- and ε-subunits that, together, lie outside the α3β3-domain, may be subject to such influences. Therefore, the rotations of residues 22–32 of the γ-subunit about the pseudo-threefold axis of the α3β3-domain in the various aligned structures were measured relative to the position of the same segment in the azide-free ground state structure.

Generation of Movies.

Movies were made with eMovie (42) and PyMOL. Morphs were generated with RigiMOL (40).