Coupling H⁺ Transport to Rotary Catalysis in F-Type ATP Synthases: Structure and Organization of the Transmembrane Rotary Motor

H⁺-transporting F₁F_o-type ATP synthases utilize a transmembrane H⁺ potential to drive ATP formation by a rotary catalytic mechanism. The simplest F₁F_o-type enzymes, as in the case of Escherichia coli, are composed of eight types of subunit with a biologically unique stoichiometry of α₃β₃γδε for the F₁ sector at the periphery of the membrane and a₁b₂c₁₂ for the transmembrane F_o sector (Fillingame, 1997; Fillingame et al., 1998). From the atomic resolution X-ray structure of the α₃β₃γ portion of bovine F₁, the three α and three β subunits are known to pack alternately around a centrally located γ subunit, with the γ subunit interacting asymmetrically with the catalytic β subunits (Abrahams et al., 1994). Subunit γ was subsequently shown to rotate with respect to the three β subunits during catalysis (Duncan et al., 1995; Sabbert et al., 1996; Noji et al., 1997). Rotation of subunit γ is thought to change the binding affinities in alternating catalytic sites to promote tight substrate binding and product release during catalysis (Boyer, 1997). During ATP synthesis, the rotation of subunit γ must be driven by proton translocation through F_o.

The F₁ sector and proton-translocating F_o sector are connected by a central stalk formed by the γ and ε subunits. The γ and ε subunits are both known to contact the c subunit at the surface of F_o (Zhang and Fillingame, 1995b; Watts et al., 1995, 1996; Hermolin et al., 1999), and the 12 copies of the c subunit are thought to form an oligomeric ring that is proposed to rotate relative to a stationary a₁b₂ complex (Fig. 1). Protonation/deprotonation of Asp61 of subunit c at the center of the membrane is proposed to drive rotation of the c oligomer, with c-rotary movement being coupled to rotation of subunit γ by connections between the c, γ and ε subunits (Vik and Antonio, 1994; Vik et al., 1998; Junge et al., 1997; Engelbrecht and Junge, 1997; Elston et al., 1998). Low-resolution electron microscopic and atomic force microscopic images initially suggested a ring-like arrangement for the c oligomer, with subunits a and b lying at the periphery of the ring (Birkenhäger et al., 1995; Singh et al., 1996; Takeyasu et al., 1996). An atomic resolution structure of monomeric subunit c, the smallest subunit in F_o, has been solved using nuclear magnetic resonance (NMR) (Girvin et al., 1998). A detailed description of the organization of the c oligomer in F_o, obtained only recently using cross-linking approaches and molecular modeling (Fillingame et al., 1998; Jones et al., 1998; Jones and Fillingame, 1998; Dmitriev et al., 1999a), is given below. Cross-linking studies support the proposed placement of subunits a and b at the outside of the c-oligomeric ring (Fillingame et al., 1998; Jiang and Fillingame, 1998; Jones et al., 1998), as initially suggested by the electron and atomic force microscopic studies.

In the model shown in Fig. 1, the a₁b₂ subcomplex is proposed to play the role of a stator, holding the α₃β₃ subunits of F₁ fixed to the stationary F_o subunits as the c₁₂–γε subunits rotate as a unit. Recent electron micrographs now indicate a second stalk at the periphery of the F₁F_o interface which is presumed to represent a dimer of b₂ subunits extending from F_o to F₁ (Wilkens and Capaldi, 1998; Böttcher et al., 1998; Walker, 1998). The cytoplasmic domain of subunit b binds to subunit δ of F₁ in solution (Dunn and Chandler, 1998), and interactions between subunit b and subunits δ and α at the top of the F₁ molecule have been demonstrated in F₁F_o (Rodgers and Capaldi, 1998). To reach the top of the F₁ molecule, subunit b is estimated to extend 11 nm from the surface of the membrane (Rodgers and Capaldi, 1998). Subunit b is anchored in the membrane via a single transmembrane helix (TMH) at the N terminus, the structure of the segment having been determined recently using NMR (Dmitriev et al., 1999b). Subunit a is a highly hydrophobic protein which was recently shown to span the membrane with five transmembrane helices (Valiyaveetil and Fillingame, 1998; Long et al., 1998: Wada et al., 1999). It is presumed to be the major component of the alternate access channels shown in Fig. 1, but the placement of these channels in the protein is not yet easily reconciled with our current knowledge of the subunit. The interactions and role of aTMH-4, which contains the conserved and essential residue Arg210, will be considered in some detail below. In this essay, we will be emphasizing recent work that has helped to define details of the structural organization and interactions of subunits in F_o and discuss this new information in the context of the rotary motor hypothesis.

Subunit c is a hydrophobic protein of 79 residues that folds through the membrane in a hairpin-like structure with two membrane-traversing α-helices and a more polar loop region exposed to the F₁ binding side of the membrane (Fillingame, 1997). Asp61, centered in the second transmembrane helix, is known to protonate and deprotonate during H⁺ transport. The binding of the loop region to subunits γ and ε of F₁ is proposed to force rotation of subunit γ as proton transport drives rotation of the c₁₂ oligomer (Fig. 1). A high-resolution structure of monomeric subunit c has been solved by NMR, and the structure closely resembles the folding predicted for the native subunit in F_o (Girvin et al., 1998). The protein folds as a helical hairpin in solution with residue–residue interactions as predicted from genetic analysis of the folded protein in vivo. These interactions include the close proximities of residues Ala24 and Ile28 in TMH-1 to Asp61 in TMH-2 (Miller et al., 1990; Fillingame et al., 1991) and of Ala20 in TMH-1 to Pro64 in TMH-2 (Fimmel et al., 1983; Zhang and Fillingame, 1995a).

Extensive cross-linking studies with membranous native F_o indicate an oligomeric ring of 10–12 subunits with TMH-1 inside and TMH-2 outside (Jones et al., 1998). The cross-links between genetically introduced cysteine residues were generated in part to test, and now do support, the NMR model. The stoichiometry of 12 c subunits per oligomeric ring is based upon cross-linking studies with genetically fused dimers and trimers of subunit c (Jones and Fillingame, 1998). A spacer loop was inserted between the C terminus of the first monomeric unit and the N terminus of the next on the basis of the precedent of V-ATPase subunit c, which is thought to have arisen by gene duplication (Nelson, 1992). The genetically fused c₂ dimers and c₃ trimers proved to be functional, suggesting that the stoichiometry was a multiple of 2 and 3. On introduction of cysteine residues into the N- and C-terminal helices of the dimers and trimers, and oxidation in the membrane, multimeric ladders extending to the equivalent of c₁₂ were observed in both cases. Previous values for stoichiometry based upon quantification of biosynthetically incorporated radiolabel had ranged from 9 to 12 (Foster and Fillingame, 1982; von Meyenburg et al., 1982), i.e. a range within experimental error of the value of 12 derived from the cross-linking studies.

The structure of the c₁₂ oligomer has been modeled using molecular dynamics and energy minimization calculations from the solution structure of monomeric subunit c and 21 inter-subunit distance constraints derived from cross-linking of subunits in native F₁F_o (Dmitriev et al., 1999a). In the c₁₂ oligomeric structure, the subunits pack to form a compact hollow cylinder with an outer diameter of 5.5–6.0 nm and an inner space with a minimal diameter of 1.1–1.2 nm (Fig. 2A,B). Phospholipids are presumed to pack in the inner space in the native membrane. The transmembrane helices pack in two concentric rings with TMH-1 on the inside and TMH-2 on the outside. The H⁺-transporting Asp61 residue packs towards the center of the four transmembrane helices of two interacting subunits (Fig. 2C). The packing supports the suggestion that the proton-binding site is formed at the packed interface of two subunits, with Asp61 at the front face of one subunit interacting with Ala24, Ile28 and Ala62 at the back face of a second subunit. The packing can explain the interchange of the functional carboxyl group from position 61 to position 24 in the functional cA24D/D61G mutant (Miller et al., 1990), since the carboxyl would end up at approximately the same position between subunits whether attached to TMH-2 or TMH-1. The packing would also explain the slowed reaction of dicyclohexylcarbodiimide (DCCD) with Asp61 in the cA24S and cI28T DCCD-resistant mutants (Fillingame et al., 1991). In the Propionigenium modestum F₁F_o, which transports Na⁺, a Gln residue at the position of E. coli Ile28 and a Ser residue at the position of E. coli Ala62 provide additional liganding groups for Na⁺ binding to a Glu residue at the position of E. coli Asp61 (Kaim et al., 1997). The Na⁺-binding site would thus be expected to be formed by the front-to-back packing of two subunits. The positioning of the Asp61 carboxyl in the center of the interacting transmembrane helices, rather than at the periphery of the cylinder, suggests that the helices must rotate or swivel to open the proton-binding site to subunit a during proton transport.

An alternative model for a subunit c oligomer in which TMH-2 is packed in the inner ring and TMH-1 in the outer ring has been suggested by Groth and Walker (1997). However, as described by Dmitriev et al. (1999a), energy calculations from the cross-linking data strongly favor a structure with TMH-2 at the periphery of the ring and TMH-1 inside the ring. Groth et al. (1998) have constructed single Trp substitutions in residues 61–72 of helix 2 in an attempt to probe the arrangement of c subunits. Interpretation of the effects of most of the substitutions, where activities were severely compromised or abolished, is complicated and hence may not provide good evidence to distinguish between models. For example, some activity was observed with substitutions centered around Pro64, i.e. at residues 62, 63 and 65, which could reflect a required structural flexibility in this region related to proton binding and release from Asp61. The effects of substitutions at positions 69–72, which are expected to be located near the hydrocarbon–headgroup interface of the phospholipid bilayer, i.e. the usual position of Trp in membrane proteins, are quite compatible with the model presented by Dmitriev et al. (1999a), the expected positions of the bulky Trp side chains having been discussed more thoroughly by Dmitriev et al. (1999a). In addition, the Groth and Walker (1997) model is not easily reconciled with the extensive pattern of cross-linking observed between cysteine residues in TMH-2 of subunit c and cysteine residues in TMH-4 of subunit a, as discussed below.

Additional cross-linking analyses have established transmembrane interactions between subunits a, b and c (Fig. 3). Cys-substituted subunit b can be cross-linked to form homodimers, preferentially with Cys substitutions at residues 2, 6, 10 and 14. The structure of the N-terminal region has been determined using NMR, and the structure of the dimer was modeled on the basis of these cross-links (Dmitriev et al., 1999b). The N-terminal segments of the two copies of subunit b in F_o are suggested to cross the membrane with the transmembrane helices in close association, with the direction of packing changing at a ‘hinge’ region near the cytoplasmic surface. A b–c dimer can be formed from bCys2 to Cys74, Cys75 and Cys78 in the C-terminal region of subunit c (P. C. Jones, W. Jiang and R. H. Fillingame, unpublished results). The b–c dimer can be formed from bCys2, but not from cysteine residues lying deeper in the transmembrane regions of subunit b and c (i.e. not from bCys4, 5, 7 or 8 to cCys71 or from bCys7, 8, 11 or 12 to cCys67). The lack of cross-linking between these residue pairs may indicate that the transmembrane helices of subunits b move away from the c-oligomeric cylinder as they traverse the membrane at an angle relative to the more perpendicularly oriented c-oligomeric cylinder. Cys in TMH-4 of subunit a can be cross-linked to Cys in TMH-2 of subunit c over a span of 19 residues in each helix (Jiang and Fillingame, 1998), i.e. from residues 55 to 73 in cTMH-2 and from residues 207 to 225 in aTMH-4 (Fig. 3). The cross-linking pattern brings Asp61 of subunit c and Arg210 of subunit a close together in the transmembrane region, as would be predicted if they interact functionally. Cys residues at positions 227 and 228 in the loop region between aTMH-4 and aTMH-5 form strong cross-links with bCys2 and cCys78 (Fig. 3; W. Jiang and R. Fillingame, unpublished results). The short loop between aTMH-4 and aTMH-5 is quite precisely placed at the periplasmic surface of the membrane by multiple Cys substitutions from residues 230 to 232, which preferentially react from the periplasmic surface (Fig. 3; for key results, see Valiyaveetil and Fillingame, 1998; Wada et al., 1999). On the basis of this and other information, the depth of placement of the helical hairpin of subunit c in the membrane phospholipid bilayer can be estimated fairly precisely (Fig. 4). The placement conforms well with predictions made from hydropathy analyses.

The Cys residues in TMH-2 of subunit c which cross-link with TMH-4 of subunit a fall on both the ‘front’ and ‘back’ faces of subunit c in the NMR model (Jiang and Fillingame, 1998). Given the arrangement of the c₁₂ oligomer discussed above, with the front face of one subunit packed against the back face of a second subunit, the cross-linking pattern is difficult to understand. Two explanations were suggested (Jiang and Fillingame, 1998). In the first, TMH-4 of subunit a was suggested to insert between c subunits of the oligomer in order for Arg210 to interact with the carboxyl of cAsp61 during its deprotonation (Fig. 5). In such a model, the c subunits would swivel about the centrally located TMH-1s to provide the gap for helix insertion, and the packing interactions between TMH-1 and TMH-2 within a single subunit c would remain unperturbed. In the second explanation, TMH-2 of the c subunit at the a–c stator interface was suggested to swivel relative to cTMH-1, not only to make the cross-linkable residues accessible to aTMH-4 but also to expose Asp61 at the periphery of the rotor. This idea, which is described in greater detail in Fig. 6, now seems plausible on the basis of a new NMR structure for subunit c performed at pH 8, where Asp61 should be fully ionized (Rastogi and Girvin, 1999). The previous structure was solved at pH 5, where Asp61 should be protonated. Ionization of the carboxyl group was known from previous studies to change the structure significantly (Assadi-Porter and Fillingame, 1995), particularly in the region of Asp61, in the polar loop and at the C- and N-terminal ends of the protein. In the new pH 8 structure, the protein folds over the N-terminal segment to Arg41, essentially as at pH 5, but the loop then unravels somewhat and twists such that TMH-2 reorients by 120 ° as it packs with TMH-1. It is not clear whether the twisting of the loop causes reorientation of the helices or vice versa. The reorientation of helices places Asp61 closer to the periphery of the ring, where both an interaction with Arg210 and the cross-linking between the c and a subunits described previously could take place (Jiang and Fillingame, 1998; see Fig. 6).

The rotation of TMH-2 to the peripheral position shown in Fig. 6 could occur by a 120 ° clockwise movement or conversely by a counterclockwise rotation of 240 ° as shown in Fig. 6A. The 240 ° counterclockwise movement could simultaneously promote deprotonation of Asp61 as it moves past Arg210 to the position shown in Fig. 6B, and possibly force movement of the entire rotor by 30 ° by a rolling of the a–c surfaces against each other. Reprotonation of Asp61 could occur by one of two means. A continued counterclockwise rotation of TMH-2 by 120 ° would complete a 360 ° rotation cycle for TMH-2 and return the unprotonated Asp61 to its original position, where it would be reprotonated from the inlet half-channel. Conversely, the ionized carboxylate could be reprotonated from the inlet channel at the position shown in Fig. 6B, and on reprotonation return to its original position by a clockwise rotation of 240 °. The 30 ° movement of the rotor could also be driven at this step. The portion of subunit a making up the inlet channel remains to be defined. It could be formed, at least partially, by polar residues Asn214, Glu219, Asn238 and His245 on the periplasmic sides of transmembrane helices 4 or 5, or perhaps by using the polar residues which are enriched on the periplasmic sides of transmembrane helices 1, 2 and 3.

Rotatability of cTMH-2 relative to cTMH-1 would provide an elegant explanation for the cross-linking observed by Jiang and Fillingame (1998) between the surfaces of aTMH-4 and cTMH-2 shown in Fig. 6, and also the homodimer formation observed with some single Cys substitutions in cTMH-2 (e.g. from positions 58 to 60 and 62 to 66, as described by Jones et al., 1998). Homodimer formation for mutants with Cys substitutions in cTMH-2 cannot be explained without rotation of TMH-2. The model described in Fig. 6 therefore has attractive aspects. However, as shown, it cannot explain the observed function of the cA24D/D61G and cA24D/D61N aspartyl interchange mutants, where the essential carboxyl group would be anchored to the inner helix (Miller et al., 1990; Zhang and Fillingame, 1994). Additional experiments on the organization of the c-oligomer in this mutant are planned.

The polar loop of subunit c was shown to interact directly with the γ and ε subunits of F₁ by Cys–Cys cross-linking studies (Zhang and Fillingame, 1995b; Watts et al., 1995, 1996) in experiments designed to test interactions suggested by suppressor mutation analysis (Zhang et al., 1994). In the most recent analysis (Hermolin et al., 1999), an extensive series of cross-links was interpreted by using the solution NMR structures of subunits c and ε (Girvin et al., 1998; Wilkens et al., 1995). Cysteine residues in the continuous span of residues 26–33 of subunit ε were shown to cross-link to Cys at positions 40, 42 and 44 in the polar loop of subunit c. Residues 26–33 of subunit ε form a turn of antiparallel β-sheet extending from the bottom of the subunit as a well-defined lobe (Fig. 7A). The side chains of residues 42 and 44 project from opposite sides of the loop of subunit c, i.e. the ‘front’ and ‘back’ face, respectively (Fig. 7B), which indicates that the cross-linkable domain of subunit ε must pack between the loops of adjacent subunits c in the c₁₂ oligomer for cross-linking to occur from both faces. The interaction between subunits has been modeled on the basis of cross-linking distance constraints and places the residue 26–33 lobe of subunit ε between loops of subunit c in a well-packed structure (Fig. 7C). The modeling indicates that the cross-linking of subunit γ with these same residues in subunit c (Watts et al., 1995, 1996) must take place by interaction with a separate pair of c subunits, perhaps the adjacent pair, in the oligomer. An interesting question posed by the changes in polar loop structure seen in the pH 8 NMR model of subunit c is whether the binding interactions in the loop change during the deprotonation–protonation interactions of subunit c with subunit a (discussed in the previous section).

The work described in the authors’ laboratory was supported by US Public Health Service Grant GM23105 and a grant from the Human Frontiers Science Organization.