The structure of the membrane extrinsic region of bovine ATP synthase

Abstract

The structure of the complex between bovine mitochondrial F₁-ATPase and a stator subcomplex has been determined at a resolution of 3.2 Å. The resolved region of the stator contains residues 122–207 of subunit b; residues 5–25 and 35–57 of F₆; 3 segments of subunit d from residues 30–40, 65–74, and 85–91; and residues 1–146 and 169–189 of the oligomycin sensitivity conferral protein (OSCP). The stator subcomplex represents its membrane distal part, and its structure has been augmented with an earlier structure of a subcomplex containing residues 79–183, 3–123, and 5–70 of subunits b, d, and F₆, respectively, which extends to the surface of the inner membrane of the mitochondrion. The N-terminal domain of the OSCP links the stator with F₁-ATPase via α-helical interactions with the N-terminal region of subunit α_E. Its C-terminal domain makes extensive helix–helix interactions with the C-terminal α-helix of subunit b from residues 190–207. Subunit b extends as a continuous 160-Å long α-helix from residue 188 back to residue 79 near to the surface of the inner mitochondrial membrane. This helix appears to be stiffened by other α-helices in subunits d and F₆, but the structure can bend inward toward the F₁ domain around residue 146 of subunit b. The linker region between the 2 domains of the OSCP also appears to be flexible, enabling the stator to adjust its shape as it passes over the changing profile of the F₁ domain during a catalytic cycle. The structure of the membrane extrinsic part of bovine ATP synthase is now complete.

The molecular structure of the ATP synthase complex in mitochondria is being built up from high-resolution structures of components within the framework of a 32-Å resolution overall structure determined by electron cryomicroscopy (1). The best-studied component is the F₁ catalytic domain, which protrudes into the mitochondrial matrix. It consists of a spherical region about 100 Å in diameter made up of 3 α-subunits and 3 β-subunits arranged alternately around a central stalk composed of single copies of subunits γ, δ, and ε. Ground state (2) and ground state and transition state analogue structures of the bovine F₁ domain have been solved to a resolution of 1.9, 2.0, and 2.2 Å, respectively (3), and other structures of the enzyme inhibited in various ways have been described (3), as has the ground state structure of the yeast F₁ domain (4). In the intact complex, the F₁ domain is attached to a ring of hydrophobic c-subunits in the membrane domain by the foot of the central stalk. In Saccharomyces cerevisiae, the ring has 10 c-subunits, whereas the composition of the bovine c-ring is not known (5). The F₁ domain is also attached to the membrane domain by the peripheral stalk or stator, an assembly of single copies of subunits b and d, oligomycin sensitivity conferral protein (OSCP), and F₆ (6). The peripheral stalk interacts with the top of the F₁ domain via the OSCP and extends along the periphery of the F₁ domain into the membrane domain of the enzyme, where it probably interacts with subunit a (and possibly with other membrane components) via transmembrane α-helices in subunit b. The chemical synthesis of ATP from ADP and phosphate in the F₁ domain is driven by the rotor, an ensemble of the central stalk and the c-ring, using energy from the proton-motive force (pmf) across the inner membrane of the mitochondrion. The generation of rotation requires the translocation of protons through the membrane into the matrix via a channel or channels at the interface between the c-ring and subunit a (or ATPase-6). The stator plays a crucial role in coupling the mechanical rotation to chemical synthesis by linking subunit a to the external surface of the α₃β₃-subcomplex in the F₁ domain, thus keeping them both static relative to the turning of the rotor by resisting the rotational torque. The molecular structures of parts of the stator are known. The N-terminal domain of the OSCP and its mode of interaction with the N-terminal region of 1 of 3 α-subunits have been described (7), and a large segment of the structure of the remaining matrix exposed region has been determined and docked approximately into the electron microscopy map (1, 8). Important missing elements from the structure of the stator are the structure of the C-terminal part of the OSCP and its mode of interaction with subunit b and other stator components.

As described below, a complex containing F₁-ATPase and the whole of the OSCP and the regions of other interacting subunits has been assembled in vitro and crystallized, and its structure has been solved to a resolution of 3.2 Å. This structure connects the known structural elements of the stator and F₁-ATPase and allows a structural model to be assembled of all the regions of ATP synthase that project from the inner membranes of mitochondria into the matrix. The structures of the membrane domain of subunit b, subunit a, and 4 small membrane proteins associated with subunit a but with no known roles in catalysis remain to be determined.

Results and Discussion

Reconstitution and Characterization of an F₁-Stator Complex.

A bovine F₁-stator complex was reconstituted in vitro from F₁-ATPase; the OSCP; and a subcomplex between subunits b, d, and F₆ that had been truncated systematically in subunits b and d (residues 99–214 and 1–118 remaining, respectively) so as to produce suitable crystals. The reconstituted F₁-truncated stator complex is known as F₁-stator_T. The molecular masses of the stator_T components corresponded to their calculated values. When methionine was replaced by selenomethionine, there was no evidence of either partial retention of methionine or oxidized products [supporting information (SI) Table S1].

Structure Determination.

The structure of F₁-stator_T (Fig. 1) was solved by molecular replacement with data to a resolution of 3.2 Å. The asymmetrical unit contains 2 copies of the complex. The electron density for one copy was slightly better than that for the other, but their structures are very similar with an rmsd in main-chain atoms of 0.15 Å; thus, they are not distinguished below. Data processing and refinement statistics are summarized in Table 1, and regions of the electron density map are shown in Fig. S1.

Fig. 1.

Structure of the bovine mitochondrial F₁-stator_T complex is shown in ribbon representation and contains the complete F₁ catalytic domain made up of 3 α-subunits and 3 β-subunits (red and yellow, respectively) and single copies of subunits γ, δ, and ε (blue, magenta, and green, respectively). The stator_T subcomplex contains residues 1–146 and 169–189 of the OSCP (teal); residues 122–207 of subunit b (pink); residues 5–25 and 35–57 of F₆ (pea green); and residues 30–40, 65–74, and 85–91 of subunit d (orange).

Table 1.

Data collection and refinement statistics for the complex between F₁-ATPase and selenomethionine substituted stator_T

Space group	P212121
Unit cell dimensions a, b, c, Å	158.2, 231.2, 286.4
Resolution range, Å	41.78–3.20
High-resolution bin, Å	3.37–3.20
No. unique reflections	172,437
Multiplicity	3.7 (3.8)
Completeness, %	99.7 (100)
Rmerge*	0.119 (0.836)
〈I/σ(I)〉	8.4 (1.5)
B factor, from Wilson plot, Å2	86.6
R factor†, %	21.25
Free R factor‡, %	27.08
rmsd of bonds, Å	0.016
rmsd of angles, °	1.87
Multiplicity of anomalous data	1.9 (1.9)
Completeness of anomalous data, %	98.2 (99.3)

Statistics for the highest-resolution bin are in parentheses.

*R_merge = Σ_hΣ_i|I(h) − I(h)_i|/Σ_hΣ_iI(h)_i, where I(h) is the mean weighted intensity after rejection of outliers.

^†R = Σ_hkl||F_obs| − k|F_calc||/Σ_hkl|F_obs|, where F_obs and F_calc are the observed and calculated structure factor amplitudes, respectively.

^‡R_free = Σ_hkl⊂T||F_obs| − k|F_calc||/Σ_hkl⊂T |F_obs|, where F_obs and F_calc are the observed and calculated structure factor amplitudes, respectively, and T is the test set of data omitted from refinement (5% in this case).

Electron density was not observed for all stator_T, and in some regions, the side-chain density was either indistinct or absent. Structural assignments in these regions were made by docking into the density the known structures of the N-terminal region of the OSCP (residues 11–115) (7) and that of the stator subcomplex BDF_sol (8), consisting of residues 79–183, 3–123, and 5–70 of subunits b, d, and F₆, respectively. The structure of the N-terminal domain of the OSCP fitted the density well, and the only deviations were at its N- and C-terminal ends, which were rebuilt. The C-terminal end of the model was extended to residue 146, but there was little or no density for residues 147–168, and they were not interpreted in the final model. The predicted secondary structure for this region is predominantly extended (Fig. S2). The C-terminal α-helix of the OSCP (residues 172–185) was assigned from the observed electron density and selenomethionine signals, secondary structure predictions, and biochemical data. However, the orientation could not be determined unambiguously, and the sense of the helix may be inverted. The register of the sequence of the OSCP model was assigned or confirmed by selenomethionine data (see Materials and Methods and Fig. S2). Residues 146–183 of subunit b, 5–25 and 35–57 of F₆, and 65–74 of subunit d from BDF_sol were placed readily in tubular electron density, and the position was confirmed by selenomethionine peaks at residues 164, 165, and 173 in subunit b and residues 50 and 56 in subunit F₆. Residues 122–145 of subunit b were modeled by applying a rigid body rotation of ca. 9^° to this region of the model of BDF_sol and extended to residue 207 by building into the electron density. Finally, residues 30–40 and 85–91 of subunit d were built into the density. The readily interpretable regions in the current map of subunits b, d, and F₆ are α-helices, and the uninterpreted regions are extended.

The final model of the F₁-stator_T complex contains the following residues: α_E, 1–510; α_TP, 23–401 and 410–510; α_DP, 27–510; β_E, 9–474; β_TP, 9–474; β_DP, 9–475; γ, 1–61, 70–96, and 101–272; δ, 15–145; ε, 1–47; the OSCP, 1–146 and 169–189; b, 122–207; F₆, 5–25 and 35–57; and d, 30–40, 65–74, and 85–91. The second copy of the complex contained the same residues, except that residues 122–160 of subunit b; the entire subunit d; and residues 5–16, 35–39, and 49–57 of F₆ were not resolved. In both copies, the nucleotide binding site in the β_DP- and β_TP-subunits contained Mg-ADP and Mg-AMP-PNP, respectively, and the β_E-subunit had no bound nucleotide, but there was electron density in the vicinity of the P-loop (residues 157–163) interpreted as bound phosphate (or sulfate), as in other structures of F₁-ATPase. The nucleotide binding sites in the α-subunits contained Mg-AMP-PNP.

The Structure of F₁-Stator_T.

The structure of the F₁ domain (Fig. 1) is very similar to that of the ground state structure (2), except that the central stalk is rotated clockwise slightly as viewed from the membrane. Therefore, the C-terminal regions of α-subunits differ slightly, probably because of differences in crystal contacts.

Stator_T has an elongated structure oriented along the noncatalytic α_DPβ_TP-interface of the F₁ domain. It consists of most of the OSCP (residues 1–146 and 169–189) and residues 122–207 of subunit b, with the 2 subunits interacting via their C-terminal regions. Subunit b is augmented by 2 extensive α-helical regions in subunit F₆ (residues 5–25 and 35–57) and by 3 α-helical regions in subunit d (residues 30–40, 65–74, and 85–91).

The OSCP sits on top of the F₁ domain, in agreement with earlier experiments (9). Its N-terminal domain (residues 1–113) contains the bundle of 6 α-helices described for residues 1–120 (7). The buried surface area of the interface between the OSCP and the F₁-ATPase is 2 × 1,300 Ų. Helices 1 and 5 provide the binding site for residues 6–17 of subunit α_E, largely via hydrophobic interactions and 2 possible charge–charge interactions (R15 and E91 and E7 and R94 from subunits α and the OSCP, respectively; Fig. 2). In support of the latter interaction, the mutation R94Q or R94A reduced the binding of OSCP to rat mitochondrial F₁-ATPase (10). There is another region of interaction involving residues 1–14 of the OSCP with the N-terminal β-barrel domains of especially the β_DP-subunit and also the α_DP- and α_E-subunits. The high-temperature factors of the model of the OSCP in this region do not permit the identification of specific side-chain interactions with certainty, but polar interactions are likely to be dominant. In the Escherichia coli enzyme, the equivalent interactions have been estimated to be capable of resisting the torque of up to 50 kJ/mol generated by the rotary motor of the enzyme (11). The α_E-subunit–OSCP interaction in F₁-stator_T is similar to the interaction determined with synthetic peptides from the N-terminals of mitochondrial and bacterial α-subunits (7, 12). Electron density for the N-terminal region of subunit α_DP was poorly defined, but its proximity to the C-terminal domain of the OSCP suggests that they may interact. There was no electron density corresponding to the N-terminal region of subunit α_TP, but it could interact with the OSCP around residue 50. The N-terminal regions of the 3 β-subunits (residues 1–9) do not appear to contribute to the binding of the OSCP to F₁-ATPase. Bovine OSCP binds to F₁-ATPase at 1 high-affinity site (K_d = 80 nM) and at 2 low-affinity sites (K_d = 6–8 μM) (13). The interaction between residues 6–17 of subunit α_E and helices 1 and 5 of the OSCP probably accounts for the high-affinity site, and the possible interactions of the OSCP with the N-terminal regions of subunit α_DP or α_TP could provide the low-affinity sites.

Fig. 2.

Interaction between the OSCP and the N-terminal region of subunit α_E. The interaction is viewed parallel (A) and toward the inner mitochondrial membrane (B). The interaction involves α-helices 1 and 5 of the OSCP (teal) and residues 6–17 of subunit α_E (red).

The structure of the C-terminal domain of OSCP, which is unstructured in the isolated protein, consists of a β-hairpin followed by 2 α-helices, H7 and H8 (Fig. 3). There was no interpretable electron density for residues 147–168. The C-terminal helix of the OSCP forms a 3-helix bundle with the N-terminal α-helix of F₆ and a segment of the very long α-helix in subunit b. This bundle is capped by the C-terminal helix of subunit b, which lies at almost 90° to the long helix. It was not present in the construct used in the structure determination of the isolated peripheral stalk (8). The C-terminal helix of subunit b also packs against helix H7 of the OSCP, resulting in an extended 5-helix bundle to which all 3 subunits contribute (Fig. 3). These helix–helix interactions produce an extensive, and probably rather stable, interface between the OSCP and subunits b and F₆. In contrast, there are few interactions of residues in the OSCP β-hairpin with either the extended helical bundle or the N-terminal domain of the OSCP, suggesting that this region could represent a point of flexure in the peripheral stalk. The C-terminal α-helix of F₆ (residues 35–57) and the 3 α-helical regions of subunit d lie alongside the long α-helix of subunit b and appear to be stiffening it (8).

Fig. 3.

Subunit interactions at the junction in stator_T between subunits OSCP, b, and F₆. The interactions involve residues 133–146 and 169–189 of the OSCP (blue), residues 196–207 and 170–185 of subunit b (magenta), and residues 8–16 in helix 1 of F₆ (green) and the C-terminal α-helix of the OSCP. The possible location of the extended region joining residues 146 and 169 of the OSCP is shown as a dashed line.

The interaction between the C-terminal regions of bovine subunits OSCP and b and the role of subunit F₆ in stabilizing this interaction were established by in vitro reconstitution and characterization of subcomplexes (6). Fragments of subunit b and the OSCP with intact C-terminal regions form somewhat unstable heterodimeric complexes that are stabilized by F₆. Deletion of residues 185–214 of subunit b prevents any association between the OSCP and the rest of the stator (6), and deletion of the C-terminal region of the yeast OSCP uncouples pmf partially from ATP synthesis (14). Also, biotin attached to the C-terminus of the yeast OSCP was detected at the junction between the crown and the nucleotide binding domains of F₁ (15), which is consistent with the position of the C-terminal region of the OSCP in the bovine F₁-stator_T complex.

In F-ATPases from eubacteria and chloroplasts, the stator is composed of the δ-subunit (the homologue of mitochondrial OSCP) and either 2 copies of subunit b or single copies of 2 related proteins, b and b′ (subunits I and II in chloroplasts). The N-terminal regions of the OSCP and the δ-subunit from the E. coli ATP synthase have the same fold (7, 16), and their sequences suggest that the structures of C-terminal regions are related also. The C-terminal region of E. coli subunit b is required for it to bind to the δ-subunit (17), and although originally thought to be interacting in an end-to-end manner, recent data (18) are consistent with an arrangement very similar to that described here for the mitochondrial enzyme.

Structure of the Membrane Extrinsic Region of ATP Synthase.

The structures of stator_T and BDF_sol were superimposed, demonstrating that in stator_T, the segment of subunit b before residue 146 and α-helical regions in subunit d have moved inward toward the F₁ domain by ca. 9° (Fig. S3). In this region, the stator consists of a single α-helix in subunit b that has bent, bringing the lower membrane proximal regions of the peripheral stalk inward toward the F₁ domain. Augmented models of the peripheral stalk were constructed by combining the structures of stator_T and BDF_sol both with and without the 9° kink at residue 146 in subunit b. These models were docked into the envelope of the overall structure of ATP synthase (Fig. S4), and the one without the 9° inward bend fitted the map better. Therefore, the inward bend and the resulting contact between the d and β_TP-subunits appear to be artifacts produced by packing constraints in the crystal lattice (see Fig. S5). However, the structure does demonstrate that there is a region of flexion (a hinge) around residue 146 of subunit b. A second region of localized contact involving residues 43–50 of the F₆ subunit and residues 247 and 301 of the β_TP-subunit is not relieved by removing the inward bend. However, because no complex was formed between F₁-ATPase and the b′dF₆ subcomplex in reconstitution experiments (6), there are unlikely to be any significant interactions between the F₁ domain and subunits b, d, and F₆.

Finally, the structure of the c-ring from S. cerevisiae (5) was added to produce the current mosaic model of ATP synthase (Fig. 4). The structures of subunits a, A6L, e, f, and g and the membrane domain of subunit b are lacking from this model.

Fig. 4.

Mosaic structure of mitochondrial ATP synthase. The structure of F₁-stator_T without the inward 9° bend was augmented with BDF_sol (4), and the c-ring was derived from the yeast F₁-c₁₀ subcomplex (5). The mosaic structure was docked into a structure of the intact bovine enzyme determined by electron cryomicroscopy. The region occupied by F₁c₁₀ is blue, and the peripheral stalk and the membrane subunits a, e, f, g, and A6L are green.

Mechanistic Implications.

Previously undescribed information about the stator has emerged from the structure of F₁-stator_T. The most significant interaction between stator_T and the external surface of F₁ is between the N-terminal regions of subunit α_E and the OSCP. Various experiments have confirmed the proximity of subunit b in the E. coli enzyme to the external surface of F₁, but they did not demonstrate direct interaction (9, 19). In addition, stator_T follows a specific noncatalytic subunit interface, α_DPβ_TP, along the surface of the F₁ domain. During ATP synthesis, this α_DPβ_TP-interface will convert in turn to the α_TPβ_E- and α_Eβ_DP-interfaces and then back to the α_DPβ_TP-state. During ATP hydrolysis, this cycle is reversed (20). Therefore, in a solution of F₁-stator_T in the absence of ATP or in an enzyme inhibited by AMP-PNP (as in the crystallization of F₁-stator_T), it is reasonable to anticipate that structural states with stator_T associated with all 3 noncatalytic interfaces would be present. If so, the process of crystallization seems to have selected the single state that was observed. Another possible explanation is that on formation of the F₁-stator_T complex, the stator associates preferentially with the α_DPβ_TP-interface because it is energetically favorable. Because the complex was assembled in the presence ADP and azide, this preferred interaction would be maintained and a single (or at least dominant) state would be present in the crystallization medium. One noticeable feature of the noncatalytic interfaces is that they have a negative surface potential (Fig. S6) as does the dimple in the crown region of the F₁ domain in which the OSCP binds, whereas the facing surface of stator_T is positively charged. Experiments carried out in other species are inconsistent or partly supportive of the observations with the bovine enzyme. In the E. coli enzyme, the stator has been proposed to lie along a noncatalytic interface (19), although in the absence of catalytic turnover, 50% of the stators occupied the position furthest from the ε-subunit, equivalent to the mitochondrial δ-subunit, which corresponds to the α_DPβ_TP-interface (21). In the resting state of the chloroplast enzyme, it has been proposed that the stator lies along the surface of the α_TP-subunit (22).

Other significant findings concern the flexibility of the stator. From the structure of BDF_sol, it was proposed that this part of the stator is rather stiff (8), differing from proposals that the stator in the E. coli resembles a flexible rope (23). The current work shows that under the conditions imposed by the crystal lattice, the bovine stator can bend inward around residue 146 of subunit b, with the regions above and below it in the long α-helix remaining undeformed and apparently rigid. Whether the stator bends in this way during catalysis is not known. Another feature concerns the C-terminal domain of the OSCP. Although the interpretation of the electron density of this domain was difficult, 2 features relating to the rigidity of the stator are apparent. The interface linking the OSCP to subunit b, which is essential for the stator integrity, is extensive and buttressed by interactions with subunit F₆, probably making this region inflexible. In contrast, the β-hairpin of the OSCP makes few interactions with the N-terminal and remainder of the C-terminal domains of the OSCP, and this region may be a flexible joint in the stator, which would allow it to adjust to the changing shape of the surface over which it passes during catalysis. Once the stator becomes severed from its membrane anchor, as in the stator_T complex, the joint in the linker makes the residual stator mobile, as evident from the high values of the B-factors (Fig. S7).

The coupling of pmf to ATP synthesis depends on an integral stator. To achieve this coupling, a number of roles for the stator have been proposed. One is that the stator maintains the integrity of the proton channel and the coupling of the pmf to ATP synthesis by clamping subunit a to the rotating c-ring (8, 24). This action would be provided by the membrane domain of subunit b binding to subunit a and would require a rigid or elastic stator to provide an inward force to keep subunit a and the c-ring together. The presence of a flexible linker between the N- and C-terminal domains of the OSCP impairs or prevents the ability of the otherwise seemingly rigid stator from exerting the required inward force; thus, this clamping feature now seems to be implausible. In the current model of ATP synthase (Fig. 4), the long helix of subunit b enters the membrane about 10 Å from the edge of the c-ring, suggesting that its 2 transmembrane α-helices could interact with both subunit a and the c-ring. Another proposed role for the elasticity of the stator is to store energy transiently during the 120° rotational steps of the rotor during catalysis. This transient energy storage could be shared by the central stalk and the stator, but it seems that the central stalk alone fulfills this role (25). Therefore, the only remaining role of the stator appears to be to help the α₃β₃-domain to resist the rotational torque of the rotor.

Materials and Methods

Production, Purification, and Characterization of Proteins.

The bovine stator_T complex was produced by protein overexpression in E. coli from a plasmid encoding all 3 proteins in an operon. Bovine OSCP was produced similarly by bacterial overexpression and refolded. The proteins were mixed together, and the stator_T complex was purified by ion exchange and gel filtration chromatography. Its purity was assessed by SDS/PAGE analysis. The molecular masses of the proteins were verified by mass spectrometry. The stator_T complex was assembled with bovine F₁-ATPase and purified by gel filtration chromatography. For details of these procedures, see SI Text and Fig. S8.

Crystallization of the F₁-stator_T Complex.

The purified F₁-stator_T complex was concentrated to 0.25 mL on a Vivaspin membrane (Sartorius Stedim Biotech) with a molecular mass cutoff of 100,000 Da (4,000 × g, 45 min), and the buffer was exchanged on a Micro Bio-Spin 6 column (Bio-rad) to the F₁-stator purification buffer made in D₂O. The protein was centrifuged (160,330 × g, 30 min), and its concentration was adjusted to 12.5 mg/mL. Crystals were grown by microbatch under light paraffin oil in 72-well microbatch plates (Nalgene Europe). The drops contained a solution (3 μL) of 15–22% (wt/vol) polyethylene glycol 8000, 300 mM ammonium sulfate, 100 mM 2-(N-morpholino)ethanesulfonic acid (pH 6.5), 1 mM AMP-PNP, and an equal volume of protein solution. Plates were covered with filtered liquid paraffin (5 mL; BDH Laboratory Supplies) and were kept at 23 °C for 14 days. Four crystals were harvested, washed 3 times in the buffer, and then dissolved in water (5 μL) and sample loading buffer (5 μL) containing lithium dodecyl sulfate instead of SDS. Analysis of the crystals by SDS/PAGE showed that the subunits of F₁-ATPase and of the stator had remained intact during crystal growth, and there was no indication of proteolytic degradation (Fig. S9). Crystals were harvested with micromeshes (Hampton Research) and passed through cryoprotection solutions [10 mM Tris-HCl (pH 8.0), 100 mM NaCl, 25 mM sucrose, 1 mM MgSO₄, 2.5 mM DTT, 0.5 mM EDTA, 0.01% NaN₃, 0.0005% (wt/vol) phenylmethylsulfonyl fluoride, 20 μM ADP, 19–22% (wt/vol) polyethylene glycol 8000, 150 mM (NH₄)₂SO₄, 50 mM 2-(N-morpholino)-ethanesulfonic acid buffer (pH 6.5), and 0.5 mM AMP-PNP] containing 5–20% (vol/vol) glycerol increased in 5% steps. The crystals were then vitrified in liquid nitrogen.

Data Collection and Structure Determination.

X-ray diffraction data were collected from the flash-frozen cryoprotected crystals at the European Synchrotron Radiation Facility (Grenoble, France). The data were collected at a wavelength of 0.9790 Å on beamline ID23eh1 using a Q315r charge-coupled detector from Area Detector Systems Corporation. Diffraction images were integrated with MOSFLM (26), and data were reduced with SCALA (27) and TRUNCATE (28). Molecular replacement with the structure of azide-inhibited F₁-ATPase (PDB ID code 2CK3) was carried out with Phaser (29). The N-terminal domain of the OSCP (PDB ID code 2JMX) and the core structure of the stator (PDB ID code 2CLY) were fitted manually into the density. Manual model building and remodeling were performed with COOT (30). Alternate rounds of rebuilding and refinement, including noncrystallographic symmetry restraints, were carried out with either REFMAC5 (31) or PHENIX (32). Anomalous difference Fourier maps calculated with data to a resolution of 5 Å clearly indicated the positions of all 10 selenomethionine residues included in the model (peak heights between 3.5 and 8.4 σ). They were used to confirm the sequence register of the model. Images of 3D structures and electron density maps were generated with PyMOL (33).