Structure of the c14 Rotor Ring of the Proton Translocating Chloroplast ATP Synthase*

The structure of the membrane integral rotor ring of the proton translocating F1F0 ATP synthase from spinach chloroplasts was determined to 3.8 Å resolution by x-ray crystallography. The rotor ring consists of 14 identical protomers that are symmetrically arranged around a central pore. Comparisons with the c11 rotor ring of the sodium translocating ATPase from Ilyobacter tartaricus show that the conserved carboxylates involved in proton or sodium transport, respectively, are 10.6–10.8 Å apart in both c ring rotors. This finding suggests that both ATPases have the same gear distance despite their different stoichiometries. The putative proton-binding site at the conserved carboxylate Glu61 in the chloroplast ATP synthase differs from the sodium-binding site in Ilyobacter. Residues adjacent to the conserved carboxylate show increased hydrophobicity and reduced hydrogen bonding. The crystal structure reflects the protonated form of the chloroplast c ring rotor. We propose that upon deprotonation, the conformation of Glu61 is changed to another rotamer and becomes fully exposed to the periphery of the ring. Reprotonation of Glu61 by a conserved arginine in the adjacent a subunit returns the carboxylate to its initial conformation.

ATP synthases found in the energy-transducing membranes of bacteria, mitochondria, and chloroplasts catalyze ATP synthesis and ATP hydrolysis coupled with transmembrane proton or sodium ion transport. The enzymes are multi-subunit complexes composed of an extra-membranous catalytic F 1 domain and an interconnected integral membrane F 0 domain. The hydrophilic F 1 domain consists of five different polypeptides with a stoichiometry of ␣ 3 ␤ 3 ␥␦⑀. Detailed structural information obtained with the mitochondrial enzyme (1)(2)(3) in combination with biochemical (4), biophysical (5), and single molecule studies (6 -9) revealed that synthesis or hydrolysis of ATP in the F 1 domain is accomplished via a rotary catalytic mecha-nism. In addition to information on the catalytic mechanism, structure analysis and single molecule studies of the mitochondrial or the chloroplast F 1 complex have also unraveled the molecular mechanism of several F 1 -specific inhibitors (10 -14). Less detailed information is available on the integral membrane F 0 domain, which consists of three different polypeptides (a, b, and c) and mediates the transfer of protons or sodium ions across the membrane. Subunits a and b were shown to reside at the periphery of a cylindrical complex formed by multiple copies of the c subunit (15)(16)(17)(18). The number of c subunits in the cylindrical subcomplex shows substantial variation in different organisms. Ten protomers are found in ATP synthases from yeast, Escherichia coli and Bacillus PS3 (19 -21), 11 in Ilyobacter tartaricus, Propionigenium modestum, and Clostridium paradoxum (22)(23)(24), 13 in the thermoalkalophilic Bacillus TA2.TA1 (25), 14 in spinach chloroplasts (26), and 15 in the cyanobacterium Spirulina platensis (27). The structure of isolated subunits a, b, and c from E. coli has been studied by mutagenesis analysis and by NMR spectroscopy in a mixed solvent that was suggested to mimic the membrane environment (28 -32). These studies showed that subunit a folds with five membrane-spanning helices. The fourth of these helices directly interacts with subunit c and contains a conserved arginine (Arg 210 ), which is thought to be involved in proton transfer (33). Subunit b, which is present in two copies in the intact F 0 , contains a single transmembrane helix. Cross-linking data support a direct interaction of the two copies of the b subunit (29). Subunit c was studied at two different pH values to obtain the protonated and deprotonated form of a conserved carboxylate (Asp 61 in E. coli) that was shown to be essential for proton transport (34). NMR spectroscopy revealed that the isolated c subunit consists of two long hydrophobic membrane spanning segments connected by a short hydrophilic loop (30,35). This loop is located close to the ␥ and ⑀ subunit on the F 1 side of the membrane (36,37). Low resolution x-ray crystallography, cryoelectron microscopy, and atomic force microscopy showed that the membrane-spanning helices of the multiple copies of subunit c in the intact F 0 complex are tightly packed in two concentric rings (19,22,26). Atomic resolution of the c ring was recently provided for the Na ϩ -translocating F-type ATPase from I. tartaricus (38) and the related Na ϩ -translocating V-type ATPase from Enterococcus hirae (39). Rotation of the c ring was demonstrated by cross-linking (18), fluorescence studies (40), and single molecule visualization (41,42). Based on the structural and biochemical information on F 1 and F 0 , different mechanical models have been proposed describing how the rotation of the c ring is coupled to the rotation of the F 1 rotor subunits. This rotation in turn drives sequential conformational shifts at the three catalytic ␤ subunits that result in ATP synthesis (43)(44)(45). Vice versa hydrolysis of ATP in the F 1 domain is thought to drive rotation of the ␥⑀c 10 -15 subcomplex and transports protons or sodium ions across the membrane.
Here we describe the crystal structure of the chloroplast c 14 rotor, which is the first structure of an isolated c ring rotor from a proton driven ATPase. The structure was solved by molecular replacement using a tetradecameric search model that was generated from a monomer taken from the I. tartaricus c 11 structure. The imposition of noncrystallographic symmetry restraints during refinement substantially improved electron density and structure determination.
Analytical Methods-Protein concentrations were determined using the bicinchoninic acid method (Pierce) using bovine serum albumin as standard. The proteins were analyzed by SDS-PAGE on 15% polyacrylamide gels by the Laemmli system (48). After electrophoresis, the proteins were detected by silver staining according to Heukeshoven and Dernick (49).
Crystallization and Data Collection-The crystals were grown by micro batch or hanging drop vapor diffusion at 15°C in a crystallization buffer containing 30% (v/v) polyethylene glycol 400, 100 mM sodium acetate, pH 4.6, 100 mM cadmium chloride, and 100 mM lithium chloride (50). ADP at 1 mM was added to the protein solution in an attempt to stabilize the F 1 F 0 complex. In a typical experiment, 2 l of the purified CF 1 F 0 holoenzyme at 10 mg/ml were mixed with an equal volume of the crystallization buffer. In micro batch trials mixed droplets were covered by mineral oil, and in hanging drop trials droplets were equilibrated against 400 l of reservoir solution containing the crystallization buffer. A precipitate formed shortly after set-up of crystallization trials. Initial crystals appeared after 4 days and grew to the size of 0.1 ϫ 0.1 ϫ 0.1 mm. The crystals belonged to space group C2 (␤ ϭ 104.7°) with unit cell dimensions a ϭ 128.6 Å, b ϭ 90.0 Å, c ϭ 124.9 Å, and a Matthews coefficient of 3.64 Å 3 /Da corresponding to a solvent plus detergent content of 66% (v/v). The crystals were harvested in nylon loops and immediately transferred to polyethylene glycol 400 as a cryo-protectant. The crystals were flash frozen in a stream of nitrogen at a temperature of 100 K. The data were collected at the ID14-2 Beamline at the European Synchrotron Radiation Facility (Grenoble, France). The crystals showed anisotropic diffraction up to 3.3 Å. The native data were collected to 3.8 Å resolution using a wavelength of 0.933 Å and an oscillation range of 0.5°.
Data Integration, Scaling, and Structure Determination-Programs of the CCP4 suite (51) were used for data reduction and structure determination. The data were integrated with MOSFLM (52), scaled, and merged with SCALA (53), and the amplitudes were estimated using TRUNCATE (54). The data collection statistics are summarized in Table 1. The details of the structure determination are given below.
Model Building and Refinement-Model building and refinement were done using COOT (55) and REFMAC (56), respectively. To reflect the tetradecameric noncrystallographic symmetry (NCS), appropriate restraints were applied, and all of the chains were treated equally. The complete ring was used as a single group for TLS refinement. SFCHECK (57) was used for evaluation of the agreement of data and atomic model.
Electrostatic Potentials Calculations-We manually expanded the final crystallographic model to a complete structure with all amino acids. All previously omitted side chains were added in the most plausible rotamer conformation. To relax any clashes, we refined this model in REFMAC (r ϭ 33% and R free ϭ 36%). The web-based program Hϩϩ (58) was used to calculate pK a values Glu 61 with the following parameters: salin-ity, 0.15 M; pH 4.6; internal dielectral coefficient, 6; and external dielectric coefficient, 80.
Images-The images of protein structures were generated with COOT and PYMOL (59).

RESULTS
Structure Determination-Analysis of the crystals by SDS gel electrophoresis ( Fig. 1) showed that they contain the membrane integral c ring rotor of the chloroplast ATP synthase, but F 1 subunits and subunits a, b, and bЈ of the membrane integral F 0 domain are lost during the crystallization process. We solved the structure of the chloroplast c 14 ring rotor at 3.8 Å resolution by molecular replacement. Initial trials using c monomers from I. tartaricus (PDB code 1yce) or E. coli (PDB codes 1a91 and 1c99) as search molecules were unsuccessful. The self-rotation function of the diffraction data calculated with MOLREP (60) (supplemental Fig. S1) and atomic force microscopy studies (26) suggest that the chloroplast c ring rotor consists of 14 identical copies. Thus we built hypothetical c 14 rings from the backbone of the Ilyobacter monomer (1yce) (38) and used these as search models. We tested a large number of initial models of tetradecameric rings generated by shell scripts varying the radii and the relative orientations of poly-Ala monomer models derived from the Ilyobacter structure (38). Residues to the C-terminal side of the kinked area (corresponding to Ser 71 in 1yce) were removed, because this region varies between the structures from Ilyobacter and Enterococcus. Each multimeric starting model was used in molecular replacement with MOL-REP, followed by rigid body and restrained refinement steps with REFMAC, while monitoring the contrast value from MOL-REP and the R free from REFMAC. To test the suitability of a model for model building, monomers in the complex were randomly deleted, and PHASER (61) was used to replace them. The optimum values obtained for the radius and angles were used to generate a tetradecameric ring with side chains of the spinach chloroplast c subunit included according to the Chainsaw procedure. The structure was improved in several rounds of model building with COOT and refinement with REFMAC. All of the monomers were treated equally using NCS-averaged electron density maps in COOT and by applying strict NCS restraints in REFMAC. This way, we averaged the structures of the 14 monomers and improved the structural precision by a factor of 3.7 (which is the square root of 14). The refinement statistics are summarized in Table 1.
As already indicated in the analysis of the crystals by SDS-PAGE, the electron density maps also showed no indication that any further subunits of the F 1 F 0 ATP synthase are present in the crystal. Furthermore, the crystal packing does not accommodate any subunit of the F 1 subcomplex. Crystallization of the isolated c ring starting from intact F 1 F 0 has also been described recently by Varco-Merth et al. (62), even though no crystallization conditions were given in this reference. However, cell parameters suggest that conditions were similar to those obtained earlier by our group (50).
Structure of the Chloroplast c 14 Ring-The final model of the chloroplast c ring rotor consists of 14 identical protomers that are symmetrically arranged around a central pore (Fig. 2). Each protomer consists of two membrane spanning ␣-helices that are connected by a short loop. The N-terminal helices are tightly packed on the inside, whereas the C-terminal helices are located at the periphery of the ring. The overall density of the c 14 rotor corresponds to a barrel-shaped complex with an external diameter of 58 Å and internal diameters of 25 Å (top) and 38 Å (bottom), respectively. The cylinder has a waist at the con-   (Fig. 3). Localization of lipids or detergents at this part of the rotor ring is further supported by the hydrophobic character of the residues lined up in this area of the c 14 -cylinder (Ala 12 -Val 26 ). Nevertheless, precise interpretation of this density was not possible at the present resolution of the diffraction data. Final parameters of refinement and model stereochemistry are summarized in Table 1. 61 -Mutagenesis and inhibitor studies revealed that a conserved carboxylate in the C-terminal helix (Asp 61 in E. coli and Glu 61 in spinach) that is thought to undergo cycles of protonation and deprotonation plays an essential role in proton transfer across the integral membrane domain of H ϩ -translocating F-ATPases. Models of the c ring that are based on the solution NMR structure of the E. coli c subunit monomer have placed the conserved carboxylate in the periphery of the ring but in a shielded position where it packs between the helices of adjacent protomers pointing toward the N-terminal transmembrane helix at the interior of the c ring rotor (35,65). Reversible deprotonation of the conserved carboxylate during proton transport is thought to introduce a transiently charged residue in the c ring. Hence, a shielded location of the carboxyl group in the complex seems essential to avoid an energetically unfavorable exposure of the charge toward the hydrophobic phase of the membrane. Cysteine-cysteine crosslinking studies supported the buried location of the carboxylate in the c ring rotor (66). However, in our structure of the chloroplast c 14 ring, the conserved carboxylate points toward the periphery. This conformation is in agreement with the structure of the Na ϩ -transporting F-ATPase from I. tartaricus. To remove model bias in the structure of the chloroplast c 14 ring caused by using the atomic model of the I. tartaricus monomer for the construction of the search model, the side chains of the  conserved Glu 61 and Tyr 66 were deleted in the atomic structure. This structure was then randomized with a maximum shift of 0.2 Å and re-refined, and electron density maps were calculated. The resulting omit map that is shown in Fig. 4 underlines the correct position of both residues in the chloroplast structure. Fig. 5 shows the proton binding pocket around Glu 61 as seen from the outside of the c 14 ring. Residues identical to I. tartaricus are colored in blue, and chloroplast-specific residues are shown in green. The side chain of Gln 28 is given in faint blue as clear density was only found in 2 of the 14 protomers, and it was not used for model building and refinement. Similar to the Ilyobacter structure, the peripheral position of the chloroplast Glu 61 ␥-carboxylate is stabilized by potential hydrogen bonding with adjacent side chains. This stabilization is mainly provided by hydrogen bonding to the side chain of Tyr 66

DISCUSSION
Here we present the crystal structure of the chloroplast c 14 ring rotor. This first proteolipid ring structure of a proton translocating F-ATPase was solved by molecular replacement using ring-shaped tetradecameric model structures generated from a c-monomer of the sodium translocating ATP synthase from Ilyobacter. Even though the resolution of the chloroplast structure is limited to 3.8 Å, most side chains could be clearly determined because of the 14-fold noncrystallographic symmetry applied. Because of the 14-fold averaging, the resulting model corresponds to an optical resolution of 2.8 Å. The structural model obtained allows analysis of the potential protonbinding site at the essential carboxylate Glu 61 and a detailed comparison of proton and sodium-binding sites in F-ATPases. It represents a step toward a better understanding of the mechanism of proton translocation.
Comparison with c Ring Rotors of Other Organisms-High resolution crystal structures of intact c ring rotors have recently been reported for two Na ϩ -transporting ATPases: the F-type ATPase from I. tartaricus (38) and the V-type ATPase from E. hirae (39). Structural information on H ϩ -translocating ATPases on the other hand was limited up to now to a backbone  model of the yeast mitochondrial F-ATPase (19). Comparison of the crystal structure of the chloroplast c 14 complex to the rotor rings of the Na ϩ -transporting ATPases reveals that all rotor complexes form a barrel-shaped cylinder. However, the diameters of these rotary rotors differ because of the varying number of protomers in the cylindrical complexes. The undecameric ring of I. tartaricus shows an external diameter of 50 Å at its boundaries and of 40 Å in the middle at the Na ϩbinding site. The decameric ring of E. hirae consists of protomers with four membrane-spanning helices and has an external diameter of 80 and 68 Å at the conserved Glu 139 in helix 4. Although the I. tartaricus ring has a pronounced waist in the middle of the complex, the E. hirae rotor complex is more like the chloroplast c 14 ring with a less pronounced concave curvature of the cylindrical outer surface. In a similar way, the backbone structure of the yeast mitochondrial decameric c ring features only a slight curvature. This structure has an external diameter of 34 Å at the conserved carboxylate residue (Asp 61 ) that is located approximately in the middle of the complex. Both the Na ϩ -and the H ϩ -transporting ATPases have their conserved carboxylate side chains close to the outer surface of the cylinder in the helices forming the peripheral ring of the complex. Analysis of the conserved residues in the Na ϩ -transporting F-ATPase from I. tartaricus and in the H ϩ -translocating F-ATPases from spinach chloroplasts and yeast mitochondria shows that the carboxylates are 10.6 -10.8 Å apart in all c ring rotors. Providing that this value reflects an intrinsic constant of the rotor complexes, we propose that the diameter of any c-multimer can be calculated from the number of c subunits in the complex or the stoichiometry of any unknown c ring complex can be estimated from its diameter. The identical distance of the conserved carboxylates in adjacent protomers of Na ϩ -and H ϩ -transporting ATPases implies that despite their different stoichiometries, they all may have the same gear distance. Furthermore, this figure might reflect the (electrostatic) boundary to which the transport of a single sodium ion or proton can dislocate a single protomer in the complex.
Comparison of Proton-and Sodium-binding Sites in F-ATPases-Comparison of the Na ϩ -binding site of the I. tartaricus ATPase to the H ϩ -binding site at the conserved carboxylate of the spinach chloroplast ATPase suggests that a shift in the ion specificity might be related to the hydrophobicity or to the hydrogen bonding potential of side chains adjacent to the conserved carboxylate (supplemental Table S1). In the chloroplast enzyme residues Leu 57 , Phe 59 , Ala 62 , and Leu 63 provide a hydrophobic shell around the conserved Glu 61 carboxylate. Alanine in position 62 is found in the c subunit of all H ϩ -translocating F-ATPases, whereas Na ϩ -translocating enzymes hold a polar serine or threonine in the equivalent position. Similarly, the hydrophobic side chain of Leu 63 in the chloroplast c ring or various hydrocarbon chains found in the same position in other H ϩ -transporting enzymes are substituted by an invariant threonine in Na ϩ -translocating ATPases. Furthermore, leucine or hydrocarbon side chains at position 57 of H ϩ -ATPases are also substituted by polar residues (Asp, Gln, and Ser) in most c subunits transporting sodium ions. In addition, substitution of valine, which is found in most Na ϩ -transporting c rings at a position equiv-alent to residue 59 in the chloroplast enzyme, by a more bulky hydrophobic residue like leucine or phenylalanine seems advantageous to promote proton translocation.
Proton Translocation in the c Ring Rotor-Current models of proton transport in F-ATPases suggest that either a single access channel or two-half channels located in the a subunit or at the a-c interphase are engaged in proton transport to and from the conserved carboxylate in the middle of the c ring (44,45). During proton transfer the central carboxylate is thought to be transiently exposed to a conserved arginine in the a subunit stator located at the periphery of the c ring rotor. Exposure of the charged arginine to the carboxyl side chain will lower the pK a of the acidic residue resulting in the release of the proton from the conserved carboxylate. Crystals of the chloroplast c 14 ring were obtained at pH 4.6. Hence, the structure reflects the protonated form of the chloroplast c ring rotor, assuming a pK a of 8.5 for the acidic residue as described for the Glu 65 carboxylate in the cS66A mutant of the I. tartaricus enzyme (67). Protonation of Glu 61 in the crystal is further supported by the webbased program Hϩϩ, which computes pK values of ionizable groups in macromolecules on the basis of structural coordinates. Using this program an average value of 7.3 Ϯ 1.3 was calculated for Glu 61 in the c ring rotor. In the protonated conformation the conserved Glu 61 carboxylate is located at the periphery of the c ring but largely shielded from the external hydrophobic membrane phase. The Glu 61 side chain conformation is stabilized by hydrogen bonding with Tyr 66 , with the backbone carbonyl of Phe 59 in an adjacent protomer and with Thr 64 in the same protomer. We propose that upon exposure to the charged Arg 189 of the a subunit, the conformation of the Glu 61 side chain changes to another rotamer and becomes fully exposed to the periphery of the ring because of the deprotonation of the acidic group, which affects hydrogen bonding to the Phe 59 main chain carbonyl (see Fig. 6 for illustration). Reprotonation of Glu 61 by the conserved arginine returns the carboxylate to the initial conformation, which is largely shielded from the hydrophobic membrane. This mechanism does not require substantial reorientation of the C-terminal helix of subunit c to form contact with the arginine in the subunit a stator. This idea contradicts previous models based on the E. coli solution NMR structure and cysteine-cysteine cross-linking studies (35,65). But how are the protons transported to and from the conserved Glu 61 ? The structure of the chloroplast c 14 ring shows no apparent intrinsic proton transport channel within the protomers, neither from the thylakoid lumen (P-side) nor from the stroma (N-side), to provide access to the putative proton-binding site at the conserved carboxylate. Nevertheless, it cannot be ruled out that dynamic fluctuation of the protein might open a transient pathway that is not evident in the static crystal structure. However, based on accessibility studies with hydrophilic probes (68), it seems more likely that an aqueous access channel formed by the membrane-spanning helices 2-5 of subunit a provides access to the conserved arginine from the P-side of the membrane. Proton transport along this channel reprotonates the arginine in the a subunit, which in turn protonates the conserved carboxylate Glu 61 in the c subunit. Transfer of the protons from the proton-binding site at Glu 61 to the N-side of the membrane is supposed to occur via an aqueous access channel formed at the a-c interphase. Modification of genetically engineered cysteines by Ag ϩ ions and various thiolate-reactive reagents in the E. coli c ring rotor suggests that residues at positions 57, 58, 62, and 65 are involved in this pathway, and residues 63 and 64 were moderately affected (69). With the exception of the substitutions causing moderate effects, all residues identified in these studies are located at the periphery in the chloroplast c 14 rotor and are readily accessible from the external phase (supplemental Fig. S2). However, none of the residues identified in these studies is charged or polar and has the potential to provide side chain hydrogen bonding. Thus it seems reasonable to assume that a water wire for proton transport is formed at least in part at the backbone of the c subunit, but residues in the a subunit might also participate in this pathway. In this context it is interesting to note that residues Gly 51 and Leu 57 in the c 14 structure seem not to be involved in the hydrogen-bonding network of the peptide backbone, and water molecules might be bound to their peptide carbonyl groups as reported for other proteins (70). Access of bound water to the backbone carbonyl group of other residues in the helix might be blocked in a side chain-specific and rotamer-specific manner (71). Although the suggested putative involvement of the carbonyl backbones of Gly 51 and Leu 57 has to be taken with caution at the present resolution of the structure, it might provide a clue for further experimental analysis of the proton transport in F-type ATPases.