How the regulatory protein, IF₁, inhibits F₁-ATPase from bovine mitochondria

Abstract

The structure of bovine F₁-ATPase inhibited by a monomeric form of the inhibitor protein, IF₁, known as I1–60His, lacking most of the dimerization region, has been determined at 2.1-Å resolution. The resolved region of the inhibitor from residues 8–50 consists of an extended structure from residues 8–13, followed by two α-helices from residues 14–18 and residues 21–50 linked by a turn. The binding site in the β_DP-α_DP catalytic interface is complex with contributions from five different subunits of F₁-ATPase. The longer helix extends from the external surface of F₁ via a deep groove made from helices and loops in the C-terminal domains of subunits β_DP, α_DP, β_TP, and α_TP to the internal cavity surrounding the central stalk. The linker and shorter helix interact with the γ-subunit in the central stalk, and the N-terminal region extends across the central cavity to interact with the nucleotide binding domain of the α_E subunit. To form these complex interactions and penetrate into the core of the enzyme, it is likely that the initial interaction of the inhibitor with F₁ forms via the open conformation of the β_E subunit. Then, as two ATP molecules are hydrolyzed, the β_E-α_E interface converts to the β_DP-α_DP interface via the β_TP-α_TP interface, trapping the inhibitor progressively in its binding site and a nucleotide in the catalytic site of subunit β_DP. The inhibition probably arises by IF₁ imposing the structure and properties of the β_TP-α_TP interface on the β_DP-α_DP interface, thereby preventing it from hydrolyzing the bound ATP.

A variety of ions and natural products inhibit the rotary mechanism of ATP synthase in mitochondria by binding to specific sites in its F₁ catalytic domain (1, 2). The antibiotics, the efrapeptins and the aurovertins, inhibit both ATP synthesis and hydrolysis by preventing, respectively, the clockwise and anticlockwise rotation of the central rotor (as viewed from the mitochondrial membrane) (3, 4). They do so by obstructing the closure and opening of catalytic interfaces between the three β-subunits and the three α-subunits. The efrapeptins bind to a site extending from the rotor, across the central cavity of the enzyme, into a catalytic site in a specific β-subunit (the β_E-subunit), and two molecules of aurovertin bind simultaneously to equivalent clefts between the nucleotide binding and C-terminal domains of two β-subunits (the β_TP- and β_E-subunits). Polyphenolic inhibitors, such as resveratrol, quercetin, and piceatannol, bind to a common site between the tip of the rotary central stalk and the β_TP-subunit in the annular sleeve of the “bearing” formed by loop regions below the “crown” of β-strands in the N-terminal domains of the α- and β-subunits, and prevent the rotor from turning in either direction (5).

Azide enhances the binding of an ADP molecule to one of the three catalytic sites in the β-subunits (6) and is a potent inhibitor of ATP hydrolysis by ATP synthase, but it does not inhibit ATP synthesis (7). Similarly, IF₁, the natural inhibitor protein found in mitochondria, prevents the hydrolysis, but not the synthesis of ATP by the ATP synthase (8). The protein fold of the 84-aa chain of bovine IF₁ is dominated by an α-helix 90 Å in length, and the active inhibitor is a homodimer where residues 49–81 form an antiparallel α-helical coiled-coil, leaving each inhibitory N-terminal region accessible for interaction with an F₁ catalytic domain (9). This dimeric state is favored by pH values <7.0, and at higher values the dimers associate to form higher oligomers, thereby occluding the inhibitory region and masking its inhibitory activity (9, 10). During cellular ischaemia, glycolysis becomes the only source of ATP, and so the pH of the cytosol and the mitochondrial matrix decrease (8, 11). It is thought that this acidification of the mitochondrial matrix promotes the formation of the active dimeric state of IF₁, the binding of the dimeric IF₁ to the F₁ domains of two ATP synthase complexes, and the inhibition of ATP hydrolysis (12).

In the structure of the dimeric IF₁ associated with two F₁-ATPase moieties [known as (F₁-IF₁)₂], the N-terminal inhibitory region of IF₁ is bound to a catalytic interface between the C-terminal regions of the α_DP- and β_DP-subunits, but because of disorder in the crystals, the precise details of the interaction between the inhibitor and the enzyme could not be described (12). The coiled-coil region is not required for inhibitory activity, and residues 14–47 have been defined by deletion analysis as the minimal inhibitory sequence (13). Therefore, as presented in this article, bovine F₁-ATPase has been inhibited by an active monomeric N-terminal fragment of bovine IF₁ consisting of residues 1–60 plus a C-terminal histidine₆ sequence, known as I1–60His. The complex between I1–60His and F₁-ATPase has been crystallized and solved by x-ray crystallography at 2.1-Å resolution, thereby revealing the precise mode of binding of the inhibitor to F₁-ATPase in the inhibited complex.

Results and Discussion

Characterization of Inhibitor Proteins.

The molecular masses measured by mass spectrometry of bovine IF₁, and of the monomeric forms I1–60 and I1–60His, agreed with the values calculated from their sequences [see supporting information (SI) Table 2]. The inhibitory potency of I1–60 determined with F₁-ATPase at pH 6.7 was very similar to that of the native inhibitor IF₁ (see SI Fig. 5), and it was retained at higher pH values, whereas the potency of the native inhibitor declined with increasing pH. As expected, the complex between F₁-ATPase and I1–60His had a similar apparent molecular mass (370 kDa) to F₁-ATPase, whereas the complex between F₁-ATPase and the native inhibitor was estimated to be 700 kDa (SI Fig. 6). These values are consistent with the former complex being monomeric, consisting of one molecule of I1–60His associated with one F₁-ATPase complex, and the latter being a dimeric complex with two F₁ complexes bound to the dimeric inhibitor, as described (12).

Structure of the F₁-I1–60His Complex.

The diffraction properties of crystals of the F₁-I1–60His complex were improved by shrinking the unit cell, especially in the a dimension, by changing the relative humidity from 98.5% to 98.0–97.5% by controlled dehydration. Similar benefits derived from controlled dehydration of crystals of bovine F₁-ATPase grown in the presence and absence of azide have been described (6, 14). Quaternary structural changes relative to the reference (14, 15) and (F₁-IF₁)₂ structures can be attributed to increased lattice contacts (see SI Text).

The structure of the complex (Fig. 1) was solved by molecular replacement (see Materials and Methods) using data to 2.1 Å. Data processing and refinement statistics are summarized in Table 1. The final model of the F₁-I1–60His complex contains the following residues: α_E, 24–510; α_TP, 24–401 and 410–510; α_DP, 23–404, 410–483, and 494–510; β_E, 10–474; β_TP, 9–474; β_DP, 9–477; γ, 1–59, 65–96, and 101–272; δ, 15–145; ε, 1–47; and IF₁, 8–50. The nucleotide binding sites in each of the three noncatalytic α-subunits contained mainly Mg-ATP. In the α_E- and α_DP-subunits there was evidence from the refined temperature factors for the presence of some Mg-ADP, and the occupancy of the γ-phosphate moiety in these subunits was estimated to be 75%. The β_DP- and β_TP-subunits both contained Mg-ADP, and the β_E-subunit had no bound nucleotide, but, as in many previous structures (16–20), there was electron density in the vicinity of the P-loop residues 157–163 that was interpreted as bound phosphate (or sulfate).

Fig. 1.

The structure of the bovine F₁-I1–60His complex. (A) Overall view of the complex. The α-, β-, γ-, δ-, and ε-subunits are shown in ribbon form in red, yellow, dark blue, magenta, and green, respectively. Residues 8–50 of I1–60His are shown in light blue solid representation. (B) View upward (away from the foot of the central stalk), along the axis of the γ-subunit showing the orientation of the long α-helix of I1–60His relative to the C-terminal domains of the α- and β-subunits. The N- and C-terminal ends of the I1–60His are labeled N and C, respectively.

Table 1.

Data processing and refinement statistics for the complex between bovine F₁-ATPase and I1–60His

Space group	P212121
Unit cell dimensions a, b, c, Å	262.5, 103.3, 135.6
Resolution range, Å	69.7–2.1
No. of unique reflections	212,416 (30,284)
Multiplicity	3.5 (3.1)
Completeness, %	99.2 (97.7)
Rmerge,*	10.6 (42.1)
<I/σ(I)>	9.1 (2.0)
B factor, from Wilson, Å2	31.0
Water molecules	1,923
R factor†, %	19.0
Free R factor‡, %	24.5
rmsd of bonds, Å	0.01
rmsd of angles, °	1.2

Statistics for the highest resolution bin (2.21–2.10 Å) are shown in parentheses.

*R_merge = Σ_hΣ_i|I(h) − I(h)_i|Σ_hΣ_iI(h), 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.

^‡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 and T is the test set of data omitted from refinement (5% in this case).

In this structure, the resolved region of I1–60His from residues 8–50 consists of an extended structure from residues 8–13, followed by two α-helices from residues 14–18 and residues 21–50, linked by a turn from residues 19–20. Residues 1–7 and 51–60 plus the His tag were not resolved. In the structure of the dimeric IF₁, in which residues 19–83 were resolved (9), the α-helical region from residues 22–50 continues unbroken beyond residue 50 up to residue 81, and the two monomers are dimerized by residues 49–81, which form an antiparallel α-helical coiled-coil.

The longer helix (residues 21–50) of I1–60His extends for 42 Å. Residues 47–50 lie outside of the F₁ domain and are not in contact with it, whereas the rest of the structure of I1–60His lies within it. Its binding site in F₁-ATPase is complex and involves five subunits of F₁-ATPase. The most significant contributors are the β_DP-, α_DP- and γ-subunits, and the β_TP- and α_E-subunits provide additional minor contributions. All of the residues in the β_DP-, α_DP-, and β_TP-subunits that are involved in forming the binding site lie within their C-terminal domains, whereas the regions involved in the α_E-subunit are in the nucleotide binding domain. The long helix of I1–60His extends in a C- to N-terminal direction from the external surface of the F₁ domain to the central cavity surrounding the γ-subunit. Its path rises steeply from the outside at an angle of ≈45° to the central axis of the γ-subunit. Near to the external surface of the F₁ domain, the binding site consists of a deep groove beneath the C-terminal domain of the β_DP subunit. The entrance to the binding groove is somewhat constricted by the ε-subunit in the vicinity of residue 47 in its C-terminal region. Lys-48 is widely conserved (although not in fungi), but residues 48–50 at the C terminus of the ε-subunit were not resolved (as in previous structures), and there is no evidence of any direct contact between I1–60 and the C-terminal region of the ε-subunit. The binding groove itself is formed from the C-terminal helix (or helix 6) and helix 2 of the β_DP-subunit, which interact at right angles near to their N- and C-terminal ends, respectively, and the loop (residues 447–453) linking helices 4 and 5 (Fig. 2A). Then the binding site continues inward via a cleft between the β_DP- and α_DP-subunits. This cleft is formed from the C-terminal end of helix 1 in the β_DP-subunit and the loop linking it to helix 2, and the C-terminal ends of helices 1 and 2 in the α_DP-subunit (Fig. 2B). From the cleft, the N-terminal end of the long helix emerges into the central cavity of the α₃β₃γ-subcomplex, making contacts with the C-terminal ends of helix 1 in both the β_DP- and β_TP-subunits, and with the N-terminal helix of the γ-subunit (Fig. 2 C and D). At this point, the chain turns abruptly right (as viewed from the side of the F₁ particle), at almost 90°, allowing the linker (residues 19–20) between the short and long helices, and the short helix itself (residues 14–18), to make more contacts with the N-terminal helix of the γ-subunit. The unstructured region (residues 8–13) preceding the short helix snakes around the N-terminal α-helix of the γ-subunit in an anticlockwise direction (as viewed from the mitochondrial membrane), emerges into the central cavity, and crosses it, making contacts with the nucleotide binding domain of the α_E-subunit (Fig. 2D). This description of the fold of IF₁ is similar to that described before in the (F₁-IF₁)₂ complex, but in the earlier complex, because of the difficulties of interpretation associated with statistical disorder in the crystals, the register of the sequence of IF₁ was incorrect (12); in the current structure, the sequence is shifted by 8 aa toward the N terminus relative to the earlier structure. In a comparison of the β_DP-subunit of the F₁-I1–60His complex with the same subunit of the ground-state structure, the only significant change in conformation is in the region of residues 382–398 at the C-terminal end of helix 1 and the loop joining helices 1 and 2. This region provides many of the interactions with the inhibitor protein and shows displacements of up to 2.9 Å, and the equivalent region of the α_DP-subunit, helices 1 and 2, and the joining loop (residues 389–417) is displaced by up to 5 Å.

Fig. 2.

The binding site for I1–60His in bovine F₁-ATPase. (A and B) Ribbon and solid representations, respectively, of the binding site groove formed from α-helices and loops between them in the C-terminal domains of the β_DP-, α_DP-, and β_TP- subunits, occupied by the long helix (residues 21–50) of I1–60His (light blue). The N- and C-terminal ends of the I1–60HIS are labeled N and C, respectively, in A. (B) The orange side chains are those of the strictly conserved residues Lys-24, Arg-25, and Glu-26 of I1–60His that do not interact with F₁-ATPase. (C and D) Interactions between residues 8–46 and F₁-ATPase. (C) View from the side of the central stalk showing the orientation of I1–60His relative to the γ-subunit. The N- and C-terminal ends of the I1–60HIS are labeled N and C, respectively. (D) View down along the axis of the γ-subunit showing the interaction of the short helix with the γ-subunit and the interaction between the extended region formed by residues 10–12 of I1–60His and side chains in the nucleotide binding domain of the α_E-subunit.

The detailed interactions between I1–60His and F₁-ATPase are both extensive and complicated. Among the most significant are hydrophobic interactions between the β_DP-subunit and residues Tyr-33 and Phe-34 in the inhibitor protein (see Fig. 3A), and a network of polar and ionic interactions involving residues Glu-29, Glu-30, Arg-32, and Arg-35 of the inhibitor protein with residues mainly in the β_DP-subunit, but also in the α_DP-subunit (Fig. 3B). There are salt bridges between Glu-29 and Arg-32, and between Glu-31 and Arg-35, which may help to orient the inhibitor protein, but the residues are not strictly conserved. Nine residues in the long α-helix are absolutely conserved (Fig. 3 and SI Fig. 7). Four of them, Phe-22, Ala-28, Leu-42, and Leu-45, sit in hydrophobic pockets and contribute to the binding of I1–60 to F₁-ATPase. In contrast, four polar residues, Lys-24, Arg-25, Glu-26, and Glu-40, make no direct contacts whatever with the F₁-ATPase, and their side chains are exposed in the aqueous phase. Their roles are somewhat obscure, but they could be helping to orient the long α-helix with respect to its binding pocket, or they may have roles in the earlier stages of binding the inhibitor protein to F₁-ATPase during a rotary catalytic cycle (see Formation of the Inhibited Complex). Only one totally conserved polar residue, Glu-30, appears to be contributing significantly to the binding of the inhibitor via a series of polar interactions, already described above. One additional absolutely conserved residue, Gly-13, is in the extended region immediately preceding the short α-helix (residues 14–17). The possible significance of the region from residues 8–21, which are not part of the minimal inhibitory sequence, will be discussed in the next section.

Fig. 3.

The structure of bovine I1–60His and its interactions with bovine F₁-ATPase. The resolved region from residues 8–50 is shown in light blue. (A) All of the residues that interact with F₁-ATPase are shown. The purple residues are those that are strictly conserved and interact with F₁-ATPase. (B) Polar interacting side chains of I1–60His and the β_DP-subunit (yellow) and α_DP-subunit (red). The purple residue (E30) is strictly conserved in all known sequences of IF₁. (C) Conserved interacting side chains are purple, and those that are strictly conserved but do not interact with F₁-ATPase are orange.

Formation of the Inhibited Complex.

The extensive binding site of I1–60His in F₁-ATPase has a buried surface area of 2,200 Ų, and the region of interaction with the C-terminal domain of the β_DP-subunit is the major contributor (1,100 Ų). In the ground-state structure of F₁-ATPase, the conformations of the C-terminal domains of all three β-subunits are very similar. The main difference is in the region of residues 384–399, which interact with the γ-subunit differently from one β-subunit to another. This region also interacts with the inhibitor protein in the structure of F₁-I1–60His. Therefore, the binding site for the inhibitor protein is largely present in all three β-subunits, and modeling shows that the conformational changes observed on inhibitor binding to the β_DP-subunit can be accommodated in the β_E-subunit without introducing any steric clashes with the γ-subunit. However, the IF₁ binding sites in the β_DP-α_DP and the β_TP-α_TP interfaces of the reference structure are buried almost entirely, and it is unlikely that these interfaces will be perturbed by large-scale structural fluctuations. Therefore, the structural evidence does not support the proposal that, during inhibition of ATP hydrolysis by IF₁, the inhibitor must bind first at the β_TP-α_TP interface of F₁-ATPase (21, 22). The only binding site accessible to the long helix of I1–60His is that which lies in the β_E-α_E interface (helices 1, 2, and 6 in subunit β_E; helices 1 and 2 in subunit α_E), and it is more likely that the initial binding of IF₁ takes place at the β_E-α_E interface, as proposed (12). At this early stage of binding, few or no interactions will be likely to have formed between IF₁ and the α_E-subunit, and the interactions between IF₁ and the β_E-subunit will be limited to residues 21–46 of the inhibitor. It is not known whether the N-terminal region of IF₁ (residues 1–20) will have formed interactions with F₁-ATPase, and it may be disordered, as in the structure of dimeric IF₁. The low affinity of IF₁ in the absence of ATP (21, 22) and the limited nature of the interactions between F₁-ATPase and IF₁ at this stage of binding are consistent with this proposal. Partial closure of the β_E-subunit after ATP binding may increase the affinity of the inhibitor protein.

At the next stage, ATP hydrolysis and rotation of the γ-subunit by 120° will convert the β_E-α_E interface to the β_TP-α_TP interface. During this transition, it is likely that the unstructured N-terminal region of IF₁ will wrap around the γ-subunit and become structured, as observed in the F₁-I1–60His structure. However, to accommodate IF₁, the β_TP-α_TP interface must remain partially open so as to resemble the β_DP-α_DP interface in the F₁-I1–60His structure, and a second potential IF₁ binding site will become exposed in the newly formed β_E-subunit. However, it is likely that a second IF₁ will be prevented from binding tightly as modeling suggests that the position of the γ-subunit will block the binding site for the N-terminal region of IF₁. In the next step of the inhibition process, a second ATP molecule is hydrolyzed, resulting in a second 120° rotation of the γ-subunit, and the β_TP-α_TP interface will be converted to the β_DP-α_DP interface, with IF₁ bound at the latter site. This scheme assumes that IF₁ bound at the β_TP-α_TP interface, resulting in the incomplete closure of this catalytic interface (as confirmed by modeling), does not prevent the second hydrolysis step, and that the normal catalytic sequence is followed up to the state observed in the crystal structure. As the kinetic pathway from initial binding to the final inhibited state has not been characterized, other pathways (involving, for example, a rotation of the γ-subunit without hydrolysis of ATP) remain possible.

This proposal does not specify on which of three β-subunits ATP hydrolysis occurs. Two possible schemes of ATP hydrolysis by F₁-ATPase have been described (17), and the related schemes of inhibition of ATP hydrolysis by the inhibitor protein are shown in Fig. 4. In Fig. 4A, which is supported by single-molecule experiments (23), ATP hydrolysis occurs at the catalytic interface where the inhibitor is bound to progress from state II to state III. However, the structure suggests that this transition is unfavorable as the catalytic residue αArg-373 is not correctly positioned, and therefore it is likely to be slow. Fig. 4B has no such requirement, and therefore it might be considered to be preferable. The two schemes differ by ADP being bound to the β_DP-subunit in A and ATP in B. However, ATP hydrolysis is likely to occur in the β_DP-subunit on the time scale of the crystallization process, even with the inhibitor protein bound, as has been observed in F₁-ATPase inhibited with efrapeptin (3). Therefore, the structure of F₁-I1–60His does not help to determine which scheme is preferrable.

Fig. 4.

Possible schemes of the inhibition of F₁-ATPase by the inhibitor protein. F₁-ATPase is depicted as viewed from the foot of the central stalk. For simplicity, only the catalytic β-subunits and the γ-subunit are shown. (A) In state I, the inhibitor protein and ATP bind to the β_E-subunit. ATP binding promotes the anticlockwise rotation of the γ-subunit, leading to hydrolysis of ATP bound to the β_TP-subunit as it converts into the β_DP-subunit. Both ATP binding to the β_E-subunit and ATP hydrolysis on the β_TP-subunit that is converting to the β_DP-subunit are required for a 120^o rotation of the γ-subunit. In state II, ADP and phosphate are released, and ATP binds to the newly formed β_E-subunit. ATP binding drives the anticlockwise rotation of the γ-subunit, and this rotation is linked to hydrolysis of ATP on the β_TP-subunit as it converts to the β_DP-subunit. The second 120^o rotation of the γ-subunit is generated to give state III, the fully inhibited form. (B) ATP hydrolysis takes place on the β_DP-subunit after binding of ATP to the β_E-subunit. As in A, two rounds of ATP binding and hydrolysis are required to give the fully inhibited form shown in state III. The crystal structure of F₁-I1–60His represents state III in this scheme, and the reference (or ground state) structure corresponds to state I.

Mechanism of Inhibition by IF_1.

In the F₁-I1–60His complex, as in (F₁-IF₁)₂, the overall conformation of the β_DP-α_DP catalytic interface resembles the β_TP-α_TP catalytic interface in the reference structure (14, 15). Several pieces of evidence indicate that in the active enzyme, ATP hydrolysis takes place at the β_DP catalytic site and not the β_TP catalytic site (6, 17). For example, in the ground-state analog structure, BeF₃⁻.F₁ (17), the guanidinium of the catalytically essential “arginine finger” residue αArg-373 (15), and consequently the attacking nucleophilic water molecule, are significantly further away from the analog of the γ-phosphate of ATP in the β_TP-subunit than in the β_DP- subunit, hence disfavoring ATP hydrolysis in the β_TP-subunit. In the various structures of F₁-ATPase, the side chain of αArg-373 adopts either an “inward pointing” position or an “outward pointing” position. In both the β_DP and β_TP catalytic sites of the reference structure, of the ADP-azide inhibited enzyme (6, 14, 15), and of ground-state and transition-state analog structures formed with metallo-fluorides (17–19), αArg-373 points in toward the catalytic site with its guanidinium group interacting with the β- and γ-phosphates (or their γ-phosphate analogs, BeF₃⁻ or AlF₄⁻, or with azide). This position is consistent with its proposed role in stabilizing the pentavalent intermediate in ATP hydrolysis (15). In contrast, in the β_TP-α_TP catalytic interface in the structures of F₁-ATPase inhibited by dicyclohexyl-carbodiimide (20), where ADP is bound in the β_TP catalytic site, and in a form of the enzyme with ADP bound to both β_TP and β_DP catalytic sites (17), the side chain of α_TPArg-373 is oriented away from the phosphates of the bound ADP molecules. Hence, the inward and outward pointing positions correlate with the presence of a nucleotide with or without a γ-phosphate moiety, respectively. In the F₁-I1–60His structure, α_DPArg-373 points outward, with its guanidinium group stacked against that of β_DPArg-191. Hence, the outward pointing conformation of this residue correlates with ADP being bound to the β_DP catalytic site. When ATP is present, as indicated by the (F₁-IF₁)₂ structure, where AMPPNP was bound to the β_DP-subunit, α_DPArg-373 will adopt the more usual inward pointing conformation. Modeling α_DPArg-373 in the inward pointing conformation in the F₁-I1–60His structure places the guanidinium group ≈1 Å further away from the active site than in the reference structure, disfavoring ATP hydrolysis at this site. A structure of F₁-ATPase inhibited by I1–60His with a nonhydrolysable analog of ATP bound to the β_DP catalytic site would help to clarify this point further.

Two possible explanations for mechanism of inhibition of F₁-ATPase by IF₁, corresponding to schemes A and B (Fig. 4) suggest themselves. In scheme A, hydrolysis of ATP bound to the β_DP-subunit is required to reach state III, despite the inhibitor protein being bound at that catalytic interface.

Then the final state III represents a fully inhibited form because the extensive interactions between the inhibitor protein and the α_DP- and β_DP-subunits stabilize the α_DP-β_DP interface, so that ATP binding to the β_E-subunit is no longer able to generate the rotation of the γ-subunit, as partial opening of the interface is required for rotation. Alternatively, in scheme B, the inhibitor protein prevents the hydrolysis of ATP bound to the β_DP-subunit required for the next 120° rotation of the γ-subunit. Thus, the inhibition arises from the disruption of the catalytic site. Further experiments are required to distinguish between these two possibilities. While the inhibited state with bound ATP has similarities to a “prehydrolysis state” (12), it resembles most closely a “dead-end” product (24, 25). The reversal of rotation of the central stalk in the ATP synthase leads to the reactivation of the enzyme and the expulsion of the inhibitor protein by the conversion of the β_DP-α_DP interface to the β_E-α_E interface.

Materials and Methods

Crystallization of the F₁-I1–60His Complex.

The conditions for growth of crystals were based on those used for crystallizing the dimeric F₁-IF₁ complex (12). An ammonium sulfate precipitate of F₁-ATPase was dissolved in a minimum volume of buffer prepared in D₂O containing 100 mM Tris·HC1 (pH 8.2), 40 mM MgCl₂, 2 mM EDTA, and 0.004% (wt/vol) PMSF. This solution was passed through a Micro Bio-Spin P-6 column (BioRad Laboratories, Hemel Hempstead, U.K.), and the protein concentration of the eluate was adjusted to 10 mg·ml⁻¹. This solution was mixed with a 4-fold molar excess of purified Il-60His in the same buffer, and the solution was kept at 23°C for 60–90 min. Then 1 mM ATP was added from a 100-mM neutralized stock solution. After an additional 15–20 min, the inhibition of F₁-ATPase was complete, and NaCl (300 mM) and spermidine (5 mM) were added from neutralized stock solutions. This solution was centrifuged (4,000 × g, 5 min, 23°C), and then dispensed into 72-well microbatch plates (Nalge Europe, Hereford, U.K.) coated with decane. The drops were made of the protein solution (3 μl; 10 mg·ml⁻¹) to which was added 3 μl of either PEG 4000 from 21.25% to 24% (wt/vol) or PEG 6000 from 18% to 20.75% (wt/vol). Both polyethylene glycols were dissolved in D₂O. The plates were covered with filtered liquid paraffin (5 ml; BDH Laboratory Supplies, Poole, U.K.) and kept at 23°C. Small crystals had formed within a few hours, and their growth was complete in 2 weeks. Analysis of the crystals by SDS/PAGE showed that the subunits of F₁-ATPase and I1–60His had remained intact during crystal growth, and there was no indication of proteolytic degradation.

A harvest solution, identical to the buffer in the crystallization drops, but containing an additional 2% (wt/vol) of the polyethylene glycol, was added to the drops, and the crystals of the F₁-Il-60His complex were harvested with micromesh cryoloops (MiTeGen, Ithaca, NY). They were introduced into a free-mounting system (Proteros Biostructures, Martinsried, Germany) where any excess mother liquor was wicked away. The crystals were dehydrated gradually by controlled reduction of the relative humidity, and their x-ray diffraction properties were monitored. Reduction of the relative humidity from the initial value of 98.5% to 98–97.5% improved both the resolution limit and mosaic spread of the diffraction data significantly. Then the crystals were covered in perfluoropolyether (PFO X125/03; Alfa Aesar, Lancashire, U.K.), plunged into liquid nitrogen and stored at 100 K.

Data Collection.

Diffraction data were collected to 2.1-Å resolution on a Q210 charge-coupled detector from Area Detector Systems (Poway, CA) on beam line ID29 (λ = 0.979 Å) at the European Synchrotron Radiation Facility, Grenoble, France. They were processed with MOSFLM (26) and other programs from The Collaborative Computational Project 4 suite (27).

Solution and Refinement of the Structure.

The structure of F₁-ATPase inhibited with I1–60His was solved by molecular replacement with AMoRe (28). The starting model was one (complex A) of the two F₁ complexes from the dimeric F₁-IF₁ structure (Protein Data Bank ID code 1OHH) without IF₁. No water molecules had been modeled into this dimeric structure. Refinement was carried out with REFMAC (29), alternating with manual rebuilding with COOT (30). TLS parameters were refined in the final stages. Some water molecules were built with ARP/warp (31). The stereochemistry of the model was assessed with PROCHECK (32), and figures were produced with PyMol (33).

See SI Text for cloning and overexpression of I1–60His, purification and characterization of proteins, assay of inhibition of F₁-ATPase, and estimation of molecular weights of complexes.