Structural study on the architecture of the bacterial ATP synthase F_o motor

Abstract

We purified the F_o complex from the Ilyobacter tartaricus Na⁺-translocating F₁F_o-ATP synthase and performed a biochemical and structural study. Laser-induced liquid bead ion desorption MS analysis demonstrates that all three subunits of the isolated F_o complex were present and in native stoichiometry (ab₂c₁₁). Cryoelectron microscopy of 2D crystals yielded a projection map at a resolution of 7.0 Å showing electron densities from the c₁₁ rotor ring and up to seven adjacent helices. A bundle of four helices belongs to the stator a-subunit and is in contact with c₁₁. A fifth helix adjacent to the four-helix bundle interacts very closely with a c-subunit helix, which slightly shifts its position toward the ring center. Atomic force microscopy confirms the presence of the F_o stator, and a height profile reveals that it protrudes less from the membrane than c₁₁. The data limit the dimensions of the subunit a/c-ring interface: Three helices from the stator region are in contact with three c₁₁ helices. The location and distances of the stator helices impose spatial restrictions on the bacterial F_o complex.

F-type ATP synthases are the major supplier of chemically bound energy in the form of ATP in all living cells. These enzymes use a transmembrane electrochemical ion gradient as an energy source to convert ADP and P_i to ATP. Two structurally and functionally distinct domains together form a macromolecular protein complex: the water-soluble F₁ complex (1), with a subunit stoichiometry of α₃β₃γδε, and the membrane-embedded F_o complex (2) (ab₂c_8–15). Both complexes can function as separate and independent rotary motors, fueled by either H⁺ or Na⁺ (F_o) or ATP (F₁). Structurally, they are connected by a static peripheral (δb₂) and a rotating central (γε) stalk. Ion translocation through the F_o complex induces torque, which leads to rotation (3, 4) of the enzyme’s rotor (γεc_n). The central stalk γ-subunit rotates inside the F₁ (αβ)₃ headpiece and elicits sequential conformational changes in the three catalytic β-subunits, finally leading to ATP synthesis (5). In the reverse direction, along the same operation principle, the enzyme can act as an ATP hydrolysis-driven ion pump (6, 7).

The mechanism of ion translocation and torque generation in the F_o complex involves the stator subunits a and b₂, as well as the rotor ring (c_n, c-ring). In bacteria, the outer stalk consists of a pair of b-subunits, each forming a long, membrane-penetrating, and amphiphilic α-helix, which is in contact with the α-, β-, and δ-subunits at the F₁ headpiece of the enzyme, and with the a-subunit and the c-ring in the membrane (8–10). Its coiled and stiff construction was proposed to make it act as a physical barrier against the F_o-generated torque (11, 12), but the b-subunit might also be involved in the ion translocation process itself (13, 14).

Several c-ring structures (and 1 K-ring from V-type ATPase) were solved by X-ray crystallography at high resolution and with unequivocal side-chain assignments (15–19). The synergistic picture that arises from the available c-ring structures poses functional restrictions but also suggests a conserved mechanism in rotary ATPases (20, 21). Thus, at least as regards the c-ring, an overall model for ion translocation has started to develop in the past few years.

The exact role of the stator a-subunit, however, remains unclear because structural information for this subunit is scarce. The Escherichia coli a-subunit is generally accepted to consist of five transmembrane α-helices (9, 22). It contains a conserved region at the predicted helix 4, including an arginine residue that has been shown to be essential for function (23–25). In addition to the models for the a-subunit derived from a large number of cross-linking and aqueous accessibility studies (9, 26–29), a subnanometer resolution structure of the V-type ATP synthase from the bacterium Thermus thermophilus has been reported recently (30). This cornerstone contribution suggests that the membrane stator complex of this rotary ATPase consists of eight transmembrane helices, forming two bundles of four helices each. Each bundle is in contact with different c-subunits and forms a half channel that provides ion access from either the periplasm (entrance pathway) or cytoplasm (exit pathway). These structural data appear fully in agreement with a previously proposed model of F_o function (3), as well as an F_o-mechanistic proposal based on high-resolution X-ray data of the c-ring (20). However, there are currently no structural data available for the rotor-stator interface (c-ring/a-subunit) in F-type ATP synthases. Clearly, structural data on the molecular architecture of the bacterial F_o stator are essential for a better understanding of the F_o operation mechanism.

We performed a study on the F_o complex from the bacterium Ilyobacter tartaricus. The isolated F_o complex was in native stoichiometry, and it was suitable as a target for biochemical and structural investigations. A projection map obtained from EM of 2D crystals at a resolution of 7.0 Å shows electron densities next to the c₁₁ rings. They belong to the stator a-subunit, and perhaps the b-subunit, and are also visible on atomic force microscopy (AFM) topographic images. On the basis of these observations, we propose a model for the localization and arrangement of the bacterial F_o stator subunits at the c-ring and discuss an ion translocation model, including the data presently available in the literature.

Results

Purification of the I. tartaricus F_o Complex.

We developed a purification protocol of the I. tartaricus F_o complex [ITF_o (ab₂c₁₁)] in amounts suitable for 2D crystallization experiments. An expression system was used that allows the production of His-tagged I. tartaricus F₁F_o-ATP synthase (ITF₁F_o) in E. coli DK8 cells. To purify ITF_o, we stripped off the F₁ complex using urea. Finally, a highly pure ITF_o was obtained by metal chelate affinity chromatography (Fig. 1A). Using the same expression plasmid, we also purified ITF₁F_o to compare it with the purified ITF_o (Fig. 1A). The final ATPase activity was 8.8 U/mg, and both protein subunits a and b, as well as the c₁₁ ring, showed an identical migration pattern. Silver stain signals on the SDS gel indicated that all F_o subunits were intact and apparently also kept in the expected stoichiometric amounts during the isolation procedure. The isolated ITF_o was further analyzed by size exclusion chromatography, resulting in an ITF_o-containing peak (e.g., in the presence of 0.5% Cymal-5) (Fig. S1). The a-subunit in isolated F_o showed a single signal on the Western blot, indicating that the a-subunit was intact and not degraded during purification (Fig. 1A). The final yield of pure F_o complex varied significantly from preparation to preparation and was in the range of 0.4–1.5 mg per 6-L × 2-L shaker culture setup.

Fig. 1.

Purification and MS analysis of the ITF_o. (A) Silver-stained SDS/PAGE (13.1%) with purified ITF₁F_o and ITF_o. Western blot of purified ITF_o against anti–a-His₁₂ (aH₁₂). The molecular masses (kDa) and F₁F_o protein subunits are indicated. (B) LILBID-MS of the ITF_o. At low laser intensity (Upper), a single charged (1−) peak of ITF_o has a mass of 168 kDa. At high laser intensity (Lower), the ITF_o was completely dissociated into single subunits (aH₁₂, b₁, and c₁) and to various ITF_o subunit oligomers (c₂, c₃, c₄, and c₁b₁).

Laser-Induced Liquid Bead Ion Desorption MS of the Purified ITF_o.

Laser-induced liquid bead ion desorption (LILBID) MS allows one to analyze masses and stoichiometries of macromolecular membrane protein complexes in detergent solution at physiological buffer conditions. The method has a proven track record to be applicable on, for example, ATP synthases (31), as well as their c-rings (32). We therefore analyzed the assembly and stoichiometry of the isolated ITF₁F_o and ITF_o by LILBID-MS and compared the results with the ITF₁F_o, which we isolated from WT I. tartaricus cells (Fig. 1B and Fig. S2). All theoretically calculated and measured masses from this experiment are summarized in Table S1. At soft laser intensity, the WT and heterologously expressed ITF₁F_o both showed a mass of 548 ± 5 kDa for the complete F₁F_o complex. At high laser intensity, all eight structural ATP synthase subunits were identified in the spectra; all measured masses corresponded with the expected masses for both ITF₁F_o complexes. The purified ITF_o analyzed at soft laser intensity shows, in addition to higher charged peaks (2− to 5−), the singly charged peak of the F_o complex at 168 ± 2 kDa, corresponding exactly to the calculated mass of the ab₂c₁₁ complex. At high laser intensity, the ITF_o dissociated into protein subunits a-His₁₂ (34 ± 0.3 kDa), c (8.8 ± 0.1 kDa), and b (18.6 ± 0.2 kDa), also in agreement with the literature values. Further, several oligomers were detected in the mass spectra (c₂, c₃, c₄, and c₁b₁). Taken together, LILBID-MS showed that the measured masses of the complexes (WT ITF₁F_o, recombinant ITF₁F_o, and ITF_o) correlate well with the calculated masses. Hence, the purified protein complexes show a native subunit composition and stoichiometry: α₃β₃γδεab₂c₁₁ for ITF₁F_o and ab₂c₁₁ for ITF_o.

EM Analysis of Reconstituted and 2D Crystalline ITF_o.

The ITF_o was subjected to several 2D crystallization attempts, in which the various parameters, such as the protein concentration, lipid type, lipid-to-protein ratio (LPR), pH, dialysis time and temperature, and salt conditions, were tested. The samples were screened by EM (SI Results and Figs. S1 and S3 provide more information about ITF_o stability and lipid analysis, respectively). Vesicles with tightly packed proteins but no 2D crystals were found. Some samples containing densely reconstituted protein were tested by immunogold labeling, and the presence of the a-subunit was confirmed at pH 8.0 and 8.5 (Fig. 2 A and B), whereas in control experiments with empty vesicles or reconstituted c₁₁ rings only, no gold particles were found (Fig. 2 C and D).

Fig. 2.

Two-dimensional crystallization and EM analysis of the ITF_o. ITF_o purified in dodecyl maltoside was reconstituted for 2D crystallization at an LPR of 1.2 and a protein concentration of 1.1 mg/mL. To investigate whether the His-tagged a-subunit is present in the samples, immunogold labeling was performed. Electron micrographs of an immunogold-labeled (anti–a-His₁₂, arrows) vesicle from reconstituted ITF_o at pH 8.0 (A) and pH 8.5 (B) are shown. As a dialysis buffer, we used 50 mM Tris⋅HCl (at both pHs), 120–150 mM KCl, 3 mM NaN₃, and 1 mM DTT. The samples were dialyzed at 4 °C. As controls, we used empty lipids at pH 8.5 (C) and reconstituted c₁₁ proteoliposomes (D) as done in the study by Pogoryelov et al. (20), which did not show any bound gold label. (E) Electron micrograph of a negatively stained p12₁ 2D crystal from ITF_o crystallization setup. (F) Projection map at a resolution of 7.0 Å with symmetry p12₁ applied. Continuous lines represent density above the mean, and dotted lines represent negative contours. The rectangular unit cell is shown with a = 252 Å and b = 81 Å. Screw axes parallel to the membrane plane are indicated. One unit cell contains two asymmetrical units with three c₁₁ rings in each (marked as A, B, and C); oppositely oriented c₁₁ rings are indicated in red and blue. Densities of F_o stator subunits are visible in the map adjacent to c₁₁ rings (black).

Separation of the rotor and stator components in the F_o part during crystallization procedures of ATP synthase is a known feature of the enzyme (e.g., it also occurred in 3D crystallization trials of the yeast mitochondrial ATP synthase, yielding F₁c₁₀ subcomplexes) (33). Similarly, the inherent tendency of I. tartaricus c₁₁ rings to form 2D crystals hampered us in the crystallization of the entire ITF_o; in some cases, 2D crystals of c₁₁ rings in the previously described space group p1 (34–36) appeared. However, under some conditions, 2D crystals in plane group p12₁ were formed. This crystal form was observed when reconstituting ITF_o in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC)/1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) (POPG)/1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) mixture at an LPR of 0.94–1.25 (protein concentrations from 0.95-1.5 mg/ml) at 20–25 °C and with a pH of 8.0 and 150–250 mM KCl. These crystalline areas, however, were scarce and were parts of small sheets with a size of ∼0.3–0.6 μm (Fig. 2E).

From cryoelectron micrographs of crystals embedded in trehalose, structure factors up to 5 Å were obtained after correction for lattice distortions (Fig. S4). The five best lattices were merged to 7.0 Å (Fig. S5) to calculate a projection electron density map (Fig. 2F). The rectangular unit cell has the cell dimensions a = 252 Å and b = 81 Å. One unit cell contains two asymmetrical units with three c₁₁ rings in each. The c₁₁ rings are present in an alternating “up” and “down” orientation, as also observed in the crystals of plane group p1. In both plane groups, each ring establishes crystal contacts with neighboring rings oriented in opposite cytoplasmic/periplasmic directions. Strikingly, however, the p12₁ crystal maps showed, in addition to the c-ring densities, unique prominent densities that were not present in any of the previously obtained I. tartaricus c₁₁ projection maps (34–36). They were located in the proximity of the c-rings at three different positions in the asymmetrical unit. These three densities build irregular structures in proximity to two, three, or four rings (Figs. S6 and S7). The unique features are partly resolved into strong density peaks that may represent transmembrane helices, perpendicular to the membrane. Up to seven helices may be present in total, but their appearance is ambiguous in some cases.

AFM.

Two-dimensional crystalline ITF_o samples were investigated by high-resolution AFM (Fig. 3). Two different packings were observed. The p1 crystals, as seen also in c-ring crystals (34), were the most abundant (Fig. 3A). The c₁₁ rings showed two orientations, either visible from the periplasmic side and filled with phospholipids (37) or visible from the cytoplasmic side with no central mass inside (Fig. 3 A, D, and E). In some of the topographs, the crystalline areas showed a different packing, in which one ring was in contact with four neighboring rings oppositely oriented in the membrane. Despite an intensive search, 2D crystals of the p12₁ packing type were not found.

Fig. 3.

AFM images of reconstituted ITF_o samples. A crystal composed of c₁₁ rings (A) and paracrystalline arrays of reconstituted ITF_o with visible c₁₁ rings (C) are shown. (B) Gallery shows the presence of additional densities between the c₁₁ rings (white arrowheads) and broken c-rings (red arrowheads). (D) Averages of the cytoplasmic (Left; n = 100) and periplasmic (Right; n = 50) interfaces of c₁₁ from the crystals: nonsymmetrized (Upper) and symmetrized (Lower). (E) Average of the c₁₁ rings from the paracrystalline arrays (Upper; n = 100) and their symmetrization (Lower). (Scale bars: A and C, 10 nm; B, 5 nm; D and E, 3 nm.)

In regions of densely packed protein, close to the 2D crystal borders, c₁₁ rings with and without a central plug were visible. The latter were better resolved (Fig. 3 A and C), and a height profile analysis performed on cross-sections of selected regions in the topographs (Fig. S8) revealed that they protruded 0.7 ± 0.1 nm (n = 20) from the membrane surface level. Interestingly, between two neighboring c-rings, additional objects sometimes could be seen (Fig. 3B, white arrows). They protruded only 0.4 ± 0.1 nm (n = 23) from the membrane. Each protrusion most likely represents an attached a-subunit with an average width and length of 1.5 ± 0.3 nm (n = 21). Based on the height difference in the unprocessed AFM topographs, these objects could be distinguished from c₁ (38) or c_n oligomers occasionally attached at the outside of the c-rings (Fig. 3B, red arrows). The presence or absence of the b-subunit could not be determined with certainty. However, in addition to the described densities, highly flexible parts protruding higher than the c-rings and presumably belonging to the b-subunit were present in the samples, but these caused artifacts during imaging.

Noncrystallographic Symmetry Averaging of the ITF_o.

To investigate whether there are conserved elements in the three extra densities of the asymmetrical unit of the p12₁ crystals, noncrystallographic symmetry (NCS) averaging based on single-particle averaging was performed. The three c₁₁ rings in the asymmetrical unit with the same up/down orientation and the extra densities next to them were averaged using translational and rotational alignment in IMAGIC (39). Because it was not clear which extra density belongs to which ring, all six possibilities were taken into consideration (Fig. S6). A single ring with all surrounding densities was contoured, excluding the neighboring rings (Fig. S7 A–C), and then aligned and averaged with the two others (Fig. S7 D–I). In all six possible arrangements, the c-ring densities were enhanced by the alignment, but only in one were the extra densities significantly increased relative to the background (Fig. S7D). The selected regions are shown in Fig. 4 A–C. Each of the extra densities contains a four-helix bundle (red in Fig. 4). In the average (Fig. 4D) of the selected three regions, the four-helix bundle can also be seen, although one helix is very weak because of a slight difference in its position within the three regions. Two of these four helices interact with one or two outer α-helices from the c₁₁ ring. The distances measured between the highest density peaks of the extra densities and the c₁₁ peaks are shown in Fig. 4E (lines a–g). Apart from the four-helix bundle, only 11.4 ± 0 Å apart, a fifth prominent helix (brown in Fig. 4) interacts with two adjacent outer c₁₁ ring helices (nos. 1 and 11 in Fig. 4E) at a very close distance, 10.6 ± 1.8 Å and 9.5 ± 1.7 Å, respectively. The packing in this region (g) is so tight that the density center of helix no. 11 seems to shift toward the c-ring center by a few angstroms (Fig. 4E, red line). It is not clear whether this shift is induced by the various ITF_o crystal contacts or represents a native conformation of the protein complex. Further, up to two more helices are present at each position in the asymmetrical unit. Two prominent helices are visible in Fig. 4 B and C next to the bundle (blue in Fig. 4), in a position that is occupied by a neighboring c-ring in Fig. 4A. It is thus possible that the densities in Fig. 4A are in a different conformation than the other two because of crystal contacts. In pairwise averaging (Fig. S7 J–L), it can be seen that the extra densities in Fig. 4 B and C are most similar to each other. The interaction of the c-ring and the neighboring density covers a length of ∼21 Å and involves three α-helices of the c-ring and up to three α-helices of the stator.

Fig. 4.

NCS averaging, ITF_o map, and model interpretation. (A–C) Three selected c-rings with their adjacent regions derived from the projection map (Fig. 2F) were cut out and aligned to each other. Ring B was rotated 33.9°, and ring C was rotated 32.9° counterclockwise relative to ring A. (D) Average of A–C. Densities that are present in all three positions are indicated with red and brown colors; two densities only seen in B and C are shown in blue. (E) Map interpretation using conserved densities and distance determination. Outer helices from c₁₁ that make contact with the extra densities (1, 10, and 11) are circled in yellow. One density (11) is not in c₁₁-symmetrical position; a red line indicates the shift. Densities outside c₁₁ are circled in red (four-helix bundle), blue, and brown. The vorticity of the rings is visible and shows that the view (here and red rings in Fig. 2) is from the periplasm (35). A network of helix-to-helix distances (measured between the highest density peaks) is indicated: a = 11.5 ± 0.8 Å, b = 10.5 ± 0.8 Å, c = 10.1 ± 1.5 Å, d = 11.4 ± 0 Å, e = 10.6 ± 1.8 Å, and g = 9.5 ± 1.7 Å. The subunit a/c-ring interface has a length of ∼21 Å (f); it involves three helices from the stator (a-subunit and possibly b-subunit) and maximally three α-subunits of the c₁₁ ring. (F) Model interpretation. According to the model for the E. coli a-subunit, the four-helix bundle consists of helices a2–a5. The other densities could represent the b-subunit, a0, or a1. The possible helix assignment of the four-helix bundle is indicated. Cytoplasmic and periplasmic connections are indicated as in Fig. S9C. The color code and view are the same as in E. Rotation of the c-ring in the ATP synthesis direction is indicated by an arrow. The blue spheres represent Na⁺ on the c₁₁ ion binding site.

Discussion

This work provides structural data of the stator components of the F_o complex from a bacterial F-type Na⁺-ATP synthase at a resolution of 7.0 Å. The map shows electron densities that can be unequivocally attributed to the 22 α-helices of the I. tartaricus c₁₁ ring. In addition, up to seven electron density peaks are visible next to the c-rings. These must represent the transmembrane domains of the F_o stator components, namely, the a-subunit and perhaps also the b-subunit. The presence of the b-subunit could not be explicitly confirmed. These densities were weaker than the helix densities of the c-rings, which may mean that the helices of the stator are shorter or more highly tilted than the c-ring helices, or that the stator components are present at a lower occupancy in the crystal. AFM studies showed that these features protrude less from the membrane than the c-rings. LILBID-MS analysis, gel filtration, and electrophoresis experiments show that all F_o subunits were present and in native stoichiometry (ab₂c₁₁) before the 2D crystallization setups were made. It is, however, possible that some of the stator components dissociate during the crystallization process and that they are present only in the noncrystalline parts of the vesicles.

Electron Density Map Interpretation.

A remarkable feature of the map is the presence of a four-helix bundle (red in Fig. 4). The distances between pairs of helices in this bundle are almost equal (10.1 ± 1.5 Å), and two of those helices contact the c-ring. Concurrent with our data, a four-helix bundle has been proposed for the E. coli a-subunit (40). Two four-helix bundles are also visible in the membrane-embedded stator region of the T. thermophilus ATP synthase, as recently visualized by cryo-EM (30). Our data hence suggest that such bundles are conserved elements of the stator a-subunits, and thus important for F_o stator function. At the side of the bundle, a fifth helix is in close contact with the c-ring (brown in Fig. 4). The map indicates that this rotor-stator helix arrangement is packed so tightly that the usual symmetry of the c-ring apparently is disturbed at this position and one of the outer helices (no. 11 in Fig. 4) shifts slightly inward. This remarkable observation might have consequences for the F_o mechanism because it could indicate functionally important conformational changes that can occur during the ion exchange process. However, this helix is involved in crystal contacts; any mechanistic interpretation needs further careful structural investigation.

In addition to these five helices, there are up to two more density peaks present, but their position differs depending on the reference ring. In two cases (blue in Fig. 4 B and C), there are two peaks between the four-helix bundle and a c-ring in close proximity to the fifth helix. In one case (Fig. 4A), a neighboring c-ring occupies this position, but two helices are visible at the side of the four-helix bundle facing away from the c-ring and the fifth helix. Apparently, the c-rings are more rigid than the densities next to them, and the stator subunits might be only weakly bound to the c-ring. Therefore, some of the stator helices perhaps adopt a conformation that simply fits best in the crystal lattice. The inherent flexibility of the rotor-stator region remains a pertinent problem for structural investigations. A covalent linking of the rotor with the stator (25) seems a valid approach for future studies with a stabilized F_o complex.

The interpretation of the map shows that the rotor-stator interface harbors three stator transmembrane helices, which are in contact with three outer (C-terminal) c₁₁ ring helices over a length of ∼21 Å. Although our data suggest that the visible helices are rather straight, a limitation of the projection map is that their 3D path cannot be seen. In the V-type ATP synthase, there are indications (30) that some a-subunit helices approach the ion binding site in the middle of the membrane, at the narrowest part of the hour glass-shaped c₁₁ ring (15, 35).

Helix Assignment.

A secondary structure and topography analysis of the I. tartaricus a-subunit suggests six α-helices (SI Results and Fig. S9), but a complicating factor for the helix assignment in our map was the presence or absence of the b-subunit(s) after 2D crystallization. Cross-link experiments performed with the ITF₁F_o based on the E. coli and I. tartaricus enzymes (26, 27) show that helix a4, harboring the functionally important a-subunit arginine (25), and helix a5 (41) are in contact with outer c-ring helices (c2). Biochemical studies proposed that the helices a2–a5 in E. coli form a four-helix bundle (40). There is also evidence that both b-subunit helices can interact with the same c-monomer (10) and two a-subunit loop regions (a1–a2 and a2–a3) can be cross-linked to the b-subunit (9), suggesting the close proximity of helix a2 to the N-terminal part of the b-subunit. The four-helix bundle (red in Fig. 4) could therefore represent helices a2–a5. The assignment of the fifth density (brown in Fig. 4) as well as the two remaining densities (dashed blue lines in Fig. 4F), which are missing in one of the three densities (Fig. 4A), is less clear. They all could represent one or two b-subunits as well as one or both a-subunit helices (a0, a1). It is also possible that as a result of crystal contacts, some of the density in Fig. 4A results from a different conformation than the density observed in Fig. 4 B and C.

Ion Translocation.

An ion translocation model through the F_o complex involves two ion pathways, which both end at the ion binding site of the c-ring in the middle of the membrane (3). The two pathways have been proposed to be physically separated by the conserved arginine in the a-subunit with a dual role: (i) acting as an electrostatic separator between the two access/release pathways (25) and (ii) catalyzing reversible ion exchange at the rotor’s (c-ring) ion binding site (15, 20). A bundle of four or sometimes three helices was observed in ion (H⁺ and Na⁺) conducting membrane proteins as sufficient to provide a functional ion conducting pathway (42, 43). Our data show that the bacterial F_o stator component has one four-helix bundle and, in addition, up to three more helices that belong to the a-subunit or perhaps to the b-subunit. They clearly do not form a second helix bundle, but they can still be involved in ion translocation. Apparently, despite a conserved ion translocation mechanism in different rotary ATPase types (20, 21), the rotor-stator interface architecture varies and involves two four-helix bundles in a V-type ATPase (30) instead of one, as seen here.

A total number of eight (I. tartaricus) or seven (E. coli and most other bacteria) (Fig. S9) helices are, in principle, compatible with the requirement of three to four helices per ion pathway. Although our data do not yet allow interpretations about the locations of these access and exit ion pathways, they appear to be compatible with the contemporary model of ion translocation (20). A feature of this model is, for example, that the F_o stator region provides a hydrated region in which the ion binding glutamate of the c-ring can change its conformation from locked to open, enabling ion exchange. The rotor-stator interface we found is large enough to provide such an environment; our work imposes previously undescribed spatial restrictions on the F_o complex in bacterial ATP synthases.

Materials and Methods

Expression and Purification of ITF₁F_o and the F_o Subcomplex.

ITF₁F_o and ITF_o were expressed and purified as described in detail in SI Materials and Methods. Briefly, for ITF_o purification, the membranes were washed with 3.5 M urea before solubilization to separate F₁ from F_o; subsequently, ITF_o was purified via affinity chromatography.

Two-Dimensional Crystallization of ITF_o.

Lipids were dissolved in 1% decyl-β-d-maltoside at 4 mg/mL. ITF_o purified in dodecyl-β-d-maltoside was mixed with POPC, POPE, and POPG (1:1:1 mixture) at a protein concentration of 0.95–1.5 mg/mL and at LPRs of 1–1.25 (wt/wt) and incubated at 4 °C overnight. The samples were dialyzed in dialysis buttons (Hampton Research) against 50 mM Tris⋅HCl at pH 8.0, 150–250 mM KCl, 3 mM NaN₃, and 1 mM DTT. The crystals in plane group p12₁ formed at 20 °C to 25 °C after 8 wk of dialysis.

EM, AFM, and LILBID-MS.

Negatively stained samples [1.5% (wt/vol), uranyl acetate] were screened using a Philips CM120 microscope at a magnification of 45,000×. Samples containing crystals were frozen by the back-injection method in 4.5% (wt/vol) trehalose on carbon-coated copper grids, and data collection was as described by Matthies et al. (44). Images were processed using the software 2dx (45) to correct lattice distortions and the contrast transfer function (46). Five lattices were merged using the CCP4 suite (47) to a resolution of 7.0 Å to generate a projection map. An isotropic temperature factor (B = −200 Ų) was applied to compensate for the resolution-dependent degradation of image amplitudes. Immunogold on grid labeling, AFM of reconstituted ITF_o, and LILBID-MS were performed as described in SI Materials and Methods.

NCS Averaging.

Real-space NCS averaging of the c₁₁ rings and extra densities was performed using IMAGIC (39). An image was created of each of the three rings in the asymmetrical unit in the same in-plane orientation with its surrounding densities. The other rings were masked out manually. Six possible arrangements of rings with associated extra density were identified. The rings were prealigned to superimpose the extra densities for each possible arrangement. Then, the 11 subunits of the rings were accurately aligned translationally and rotationally. The aligned images were averaged, and the averages were examined for enhanced densities outside the rings. This was the case for only one of the six possibilities.