Crystal Structure of Subunits D and F in Complex Gives Insight into Energy Transmission of the Eukaryotic V-ATPase from Saccharomyces cerevisiae*♦

Background: Subunits D and F are essential in ATP hydrolysis and reversible disassembly of V1VO-ATPases. Results: First crystallographic structure of eukaryotic V-ATPase subunit D and entire subunit F is presented. Conclusion: Structural elements in subunits D and F enable the DF assembly to become a central motor element in the engine V-ATPase. Significance: Mechanistic understanding of the subunit DF assembly and its key roles in enzyme catalysis and regulation are presented. Eukaryotic V1VO-ATPases hydrolyze ATP in the V1 domain coupled to ion pumping in VO. A unique mode of regulation of V-ATPases is the reversible disassembly of V1 and VO, which reduces ATPase activity and causes silencing of ion conduction. The subunits D and F are proposed to be key in these enzymatic processes. Here, we describe the structures of two conformations of the subunit DF assembly of Saccharomyces cerevisiae (ScDF) V-ATPase at 3.1 Å resolution. Subunit D (ScD) consists of a long pair of α-helices connected by a short helix (79IGYQVQE85) as well as a β-hairpin region, which is flanked by two flexible loops. The long pair of helices is composed of the N-terminal α-helix and the C-terminal helix, showing structural alterations in the two ScDF structures. The entire subunit F (ScF) consists of an N-terminal domain of four β-strands (β1–β4) connected by four α-helices (α1–α4). α1 and β2 are connected via the loop 26GQITPETQEK35, which is unique in eukaryotic V-ATPases. Adjacent to the N-terminal domain is a flexible loop, followed by a C-terminal α-helix (α5). A perpendicular and extended conformation of helix α5 was observed in the two crystal structures and in solution x-ray scattering experiments, respectively. Fitted into the nucleotide-bound A3B3 structure of the related A-ATP synthase from Enterococcus hirae, the arrangements of the ScDF molecules reflect their central function in ATPase-coupled ion conduction. Furthermore, the flexibility of the terminal helices of both subunits as well as the loop 26GQITPETQEK35 provides information about the regulatory step of reversible V1VO disassembly.

During catalysis the central stalk subunit F is proposed to undergo structural alterations by interacting with subunits A, B, D, and d in a nucleotide-dependent manner (22,23). Most recently, we determined the crystallographic structure of the N-terminal 94 amino acids domain (ScF(1-94)) of the 118-residue S. cerevisiae subunit F (18). ScF  has an elliptical shape with a size of 30 ϫ 16 ϫ 38 Å (Fig. 1). It contains fourparallel ␤-strands, which are intermittently surrounded by four ␣-helices forming an egg-shaped structure. Unknown so far are the structures of the C-terminal domain of ScF as well as the structure of the entire subunit D. Both are described as the key elements in coupling ATPase activity in the A 3 B 3 headpiece via a rotary-like mechanism with ion conduction in the V O part (24). Furthermore, subunit F has been proposed to transmit the signal for V 1 V O disassembly by interacting with stalk subunit H and the nucleotide-binding subunits A and B in the V 1 part (4,11,18).
Understanding the enzymatic and regulatory functions of the two stalk subunits D and F requires knowledge of their high resolution structure. Here, we present the first structural models of the S. cerevisiae DF assembly (ScDF) at a resolution of 3.1 Å. The structural models of ScDF show the molecular interactions between both subunits as well as the catalytic A 3 B 3 headpiece of the V-ATPase, and they provide important insight into an interface essential for the structural integrity of the enzyme. Finally, the new structures shed light onto the coupling events of ATP-dependent ion pumping and reversible disassembly.

EXPERIMENTAL PROCEDURES
Biochemicals-Chemicals of analytical grade were obtained from Biomol (Hamburg, Germany), Merck, Sigma, or Serva (Heidelberg, Germany). Nickel-nitrilotriacetic acid chromatography resin and chemicals for SDS-PAGE were obtained from Qiagen (Hilden, Germany) and Bio-Rad, respectively.
Purification of the S. cerevisiae V-ATPase DF Assembly-Cloning, production, and purification of the recombinant DF heterodimer of the S. cerevisiae V-ATPase (ScDF) has been described previously (18). The selenomethionine DF assembly was produced in E. coli strain BL21 (DE3) cells by growing 5-ml culture cells overnight at 37°C in Luria-Bertani (LB) medium, which contained 100 g/ml carbenicillin. The cells were pelleted by centrifugation (6,000 ϫ g), resuspended in M9 minimal media, pelleted (1,000 ϫ g), and resuspended twice. The cell suspension was used to inoculate 1 liter of pre-warmed (37°C) M9 minimal media and grown at 37°C. When the A 600 of 0.6 was obtained, the minimal media were supplemented with 100 mg/liter lysine, phenylalanine, and threonine, 50 mg of isoleucine, leucine, and valine, and 25 mg of selenomethionine and incubated for 15 min. Thereafter, protein production was induced by the addition of isopropyl ␤-D-thiogalactopyranoside to a final concentration of 0.8 mM, and growth was continued at 37°C for 4 h. The selenomethionine-substituted ScDF was purified as described earlier (18), except that 1 mM dithio- threitol (DTT) was used in all buffers to avoid oxidation of selenium.
Crystallization of the DF Heterodimer-Crystallization of the native ScDF was done using the vapor diffusion method and commercially available screens. Hanging drops were set up by mixing 1 l of the purified ScDF (5 mg/ml and 8 mg/ml) in buffer, 50 mM Tris/HCl, pH 8.5, 250 mM NaCl, 5 mM EDTA with 1 l of the precipitant solution and incubated at 18°C against 500 l of mother liquor. Initial crystals were obtained in Hampton research crystal screen 1, condition 11 (0.1 M sodium citrate tribasic dehydrate, pH 5.6, 1.0 M ammonium phosphate monobasic). Good diffraction quality crystals were obtained in optimized condition of 0.1 M sodium citrate tribasic dehydrate, pH 5.6, 1.2 M ammonium citrate monobasic. The crystals were flash-frozen in liquid nitrogen at 100 K in crystallization buffer containing 25% (v/v) glycerol. For the selenomethionine ScDF, the final buffer was exchanged with 50 mM Tris/HCl, pH 8.5, 250 mM NaCl, 5 mM EDTA, and 2 mM tris(2-carboxyethyl)phosphine (TCEP). Further optimization was done using additives from Hampton research, and the crystals, grown in the presence of 3-(1-Pyridino)-1-propane sulfonate (NDSB), were finally used for structural studies.
Data Collection and Processing of X-ray Diffraction Data-X-ray diffraction data for the native ScDF crystal was collected at the National Synchrotron Radiation Research Centre, Taiwan. Initial attempts to get a solution by molecular replacement, using the subunit DF structures of related Enterococcus hirae A-ATP-synthase (3AON and 3VR4 (25,26)) and Thermus thermophilus (3W3A (27)), as well as the S. cerevisiae mutant structure, ScF I-94 (4IX9) (18), were unsuccessful. Therefore, selenomethionine crystals of ScDF were produced as described above, and multiwavelength anomalous dispersion (MAD) data were collected at the synchrotron beamline 13B1 of the National Synchrotron Radiation Research Centre in Taiwan using the Q315 detector. Three datasets were collected to a 3.1 Å resolution from a single crystal at the appropriate inflection, peak, and remote wavelengths that were selected from a selenium absorption spectrum. Data were collected as a series of 0.5 oscillation images covering a crystal rotation range of 200 on cryo-cooled crystals at 100 K. The MAD dataset was indexed, integrated, and scaled using the HKL2000 suite of programs (28). The data were cut off at 3.1 Å based on the CC 1/2 statistics (29) (correlation coefficient between two random halves of the data set where CC 1/2 Ͼ10%) to determine the high resolution cutoff for our data. Phenix (30) was used to compute the CC 1/2 (78.7% for the highest resolution shell and 99.9% for the entire data set), supporting our high resolution cutoff determination.
The structure of ScDF was phased by the single wavelength anomalous diffraction method, using the peak dataset at 3.1 Å of the selenomethionine MAD data set. The Shelx program (31) and the Autorickshaw server were used to identify 5 out of 14 selenium sites in the protein (32,33). The resulting phases were channeled into Pirate for density modification (34), and automated model building was done using Buccaneer of CCP4 suite (35). Manual correction of the model was done with COOT (36), and refinement was carried out using REFMAC of the CCP4 suite (37). A final check on the stereochemical quality of the final model was assessed using PROCHECK (38) and Mol-probity (39), and any conflicts were addressed. Molprobity analysis indicated that the overall geometry of the final model was ranked in the 100th percentile (Molprobity score of 1.69). The clash score for all atoms was 2.87 corresponding to a 100th percentile ranking of structures of comparable resolution. All the figures were drawn using the program PyMOL (40). The structure refinement statistics are given in Table 1. The atomic coordinates and structure factors have been deposited in the Protein Data Bank with accession code 4RND.
Solution Small Angle X-ray Scattering of Subunit F-SAXS data of recombinant subunit F of the S. cerevisiae V-ATPase (ScF), purified according to Basak et al. (18), were measured with the Bruker NANOSTAR SAXS instrument, equipped with a MetalJet x-ray source and Vantec 2000 detector system (41). The x-ray radiation was generated from a liquid gallium alloy with microfocus x-ray source (K ␣ ϭ 1.3414 Å with a potential of 70 kV and a current of 2.857 mA). The x-rays are filtered through Montel mirrors and collimated by the two-pinhole system. The sample to detector distance was set at 67 cm, and the sample chamber and x-ray paths were evacuated. Protein concentrations of 2.0 and 4.2 mg/ml of ScF have been measured at 15°C with a sample volume of 40 l in a vacuum tight quartz capillary. The buffer containing 50 mM Tris, pH 7.5, 200 mM NaCl, and 5 mM EDTA was measured twice. For each measurement, a total of six measurements at 5-min intervals were recorded. The data were flood-field, spatially corrected, and processed using the inbuilt SAXS software. The data were tested for possible radiation damage by comparing the six data sets, and no changes were detected. The scattering of the buffer was subtracted, and the difference curves were scaled for the concentration. All the data processing steps were performed automatically using the program package PRIMUS (42). The experimental data obtained for different concentrations were analyzed for aggregation and folding state using Guinier (43) and Kartky plots (44), respectively. The forward scattering I(0) and the radius of gyration R g were evaluated using the Guinier approximation (43). These parameters were also computed from the entire scattering patterns using the indirect transform package GNOM (45), which also provided the distance distribution function p(r), which gives the maximal particle diameter, D max . The hydrated volume Vp, used to estimate the molecular mass of globular proteins, was computed using the Porod invariant. Empirically, it was found that the hydrated volume in cubic nanometers should numerically be between 1.5 and 2 times the molecular mass in kilodaltons (46). Low resolution models of subunit F were built by the program DAMMIN (47). Ab initio solution shapes of subunit F were obtained by superposition of 10 independent model reconstructions with the program package SUBCOMP (48) and building an averaged model from the most probable one using the DAMAVER program (49).
Comparison of the experimental scattering curve with the theoretical scattering curves calculated for the monomer of subunit F with both conformations, parallel and perpendicular, were performed with CRYSOL (50). The dimer of subunit F in parallel or perpendicular conformation and a dimer model containing both conformations in one were modeled based on the truncated crystal structure of S. cerevisiae F(1-94) (Protein Data Bank code 4IX9 (18)), and their theoretical scattering curves were computed. Analysis showed that the dimer (parallel ϭ 1.85; perpendicular ϭ 2.05; both ϭ 1.93) had a better fit than the monomer (parallel ϭ 2.20; perpendicular ϭ 2.08). The models were then used in OLIGOMER (42) to find the best fit to a multicomponent mixture of proteins. A fit of ϭ 1.62 was achieved when both conformations of the dimer were present in equal amounts (perpendicular to parallel conformation of 51:49%), and no monomers were present. A dimer model was generated using CORAL (52) allowing flexibility for the loop (95-105) between the globular domain and the C-terminal helix (52).

Crystallographic Data of the ScDF Heterodimer-
The complex of the subunits D (256 residues) and F (118 residues) of the S. cerevisiae V-ATPase was crystallized in the hexagonal space group P6 1 , and its structure was determined by selenium SAD method and refined to 3.1 Å resolution. Assuming two ScDF molecules in the asymmetric unit, the solvent content was 80.27% and the V m was 6.09 Å 3 (53). The two molecules of ScDF (called ScDF1 and ScDF2) are arranged in a head to tail orientation in the asymmetric unit ( Fig. 2A). The final asymmetric unit consists of two molecules of ScDF and 59 molecules of water and 3 glycerol molecules ( Fig. 2A). The final R-factor and R-free (calculated with 5% of reflections that were not included in the refinement) were 0.20 and 0.22, respectively. In the ScDF1 and ScDF2 molecules, the residues 27-204 and 37-203 of the 256 amino acid subunit D were assigned, respectively. In the case of ScDF1, all 118 residues of subunit F were resolved in the structure, whereby amino acids 1-116 could clearly be assigned in the ScDF2 molecule. The side chain densities for all the amino acids are well resolved. Ramachandran restraints were turned on during real space refinement in Coot (36). Validation with the ERRAT server (54) gave an overall quality factor of 94%.
Overall Structure of ScDF-The overall structure of S. cerevisiae DF assembly is shown in Fig. 2B. Subunit D of the S. cerevisiae V-ATPase consists of a long pair of ␣-helices (110 Å), connected by a short helix ( 79 IGYQVQE 85 ) as well as a ␤-hairpin region, which is flanked by two flexible loops. The long pair of helices is composed of the N-terminal ␣-helix, which commences from Tyr-29 to Thr-75, and the C-terminal helix, containing residues Gly-128 to Arg-202. The remaining portion containing residues 76 -127 is composed of a short ␤-hairpin region formed by residues Phe-92 to Ile-113, two flexible loops, and a short ␣-helix. The N-terminal helix consists of both hydrophobic residues and hydrophilic residues and is slightly basic. The short helix ( 79 IGYQVQE 85 ) is slightly acidic (Fig.  2B). The ␤-hairpin region contains hydrophobic and hydrophilic residues. The C-terminal helix mostly contains acidic residues. There is a slight bend in the C-terminal helix near the hinge region where a conserved Pro-179 is located. In comparison, a clear break can be seen in the same position of this helix near the proline-induced hinge region of the ScDF2 molecule. The 118-residue-long subunit F of the S. cerevisiae V-ATPase reveals four parallel ␤-strands, which are enveloped by five ␣-helices (Fig. 2C). The four ␤-strands (␤1-␤4) are arranged parallel to each other and are connected by five ␣-helices (␣1-␣5). In both molecules, the conformation is similar except for the ␣5-helix. The four ␤-strands (␤1, Leu-7 to Ala-12; ␤2, Phe-37 to Val-39; ␤3, Ile-64 to Asn-70, and ␤4, Ala-90 to Ile-94) and the four ␣-helices (␣1, Glu-14 to Ala-23; ␣2, Lys-47 to Glu-60; ␣3, Gln-71 to Glu-75, and ␣4, Arg-78 to Ser-83) form the N-terminal globular domain. This N-terminal domain is connected via a flexible loop of 12 residues (Pro-95 to Asp-106) to the C-terminal ␣-helix (␣5, residues Ser-107 to Leu-115). As confirmed by amino acid sequence alignment, the subunit F structure contains a unique loop between ␣1 and ␤2 ( 26 GQITPETQEK 35 ) and a highly conserved loop with a 60 ERDD 63 motif between ␣2 and ␤3, which are present only in eukaryotic V-ATPases (Fig. 2D). The C-terminal helix ␣5 consists of mostly positively charged residues ( 107 SVLKRVRKL 115 ). The r.m.s.d. (C␣) between the 115 residues of subunit F of the ScDF1 and ScDF2 molecule is 3.5 Å. In comparison, the r.m.s.d. of the N-terminal 94 C␣ atoms of subunit F in both molecules (ScDF1 and ScDF2) is 0.3 Å, reflecting that the maximum differences are prevailing in the C-terminal region. As shown in Fig. 4A, the flexible loop (Pro-95 to Asp-106) of subunit F in the ScDF1 molecule is bent at residue Lys-105, causing a perpendicular orientation of its C-terminal helix ␣5 relative to the Nand C-terminal helices of subunit D. In comparison, in the ScDF2 molecule the C-terminal helix ␣5 of subunit F is almost parallel to the N-and C-terminal helices of subunit D.
Interaction of Subunits D and F-As revealed by Fig. 4A, the S. cerevisiae subunit F is bound to the lower middle part of subunit D. The hydrophobic residues from the C-terminal helix ␣3 of subunit D and ␣1 as well as ␤4 of subunit F, respectively, generate the major contacts of the hydrophobic interface of both subunits. A second point of interaction is formed by the 26 GQITPETQEK 35 loop, which is in close proximity with the loop residues 85 ESVSTARFK 94 of subunit D, which span the short helix ( 79 IGYQVQE 85 ) and the ␤-hairpin (Fig. 4B). In addition, the C-terminal helix ␣5 of subunit F of the ScDF1 and ScDF2 molecule is interacting with the N-and C-terminal helix of subunit D. In the case of the ScDF1 molecule, the interaction of subunit D and F occurs via the amino acids Asp-51, Asp-160, and Lys-54 and Arg-111, Arg-113, and Asp-102 of subunits D and F, respectively (Fig. 4B). In comparison, residues Asp-160 and Thr-47 interact with Arg-111 and Arg-113 of subunits D and F for the ScDF2 molecule (Fig. 4C). The existence of these two conformations of helix ␣5 of subunit F relative to the termini of subunit D indicates a very flexible conformational switching between an extended and slight compact conformation of subunit F.
Subunit F in Solution Studied by SAXS-The two conformations of the C terminus of the ScF structures may also represent intermediate conformations that arise during movement about the hinge regions. To understand whether subunit F is able to undergo structural alterations in solution, SAXS experiments were performed. SAXS patterns of recombinant subunit F of the S. cerevisiae V-ATPase (ScF) were recorded at 2.0 and 4.2 mg/ml, which showed similar final composite scattering curves ( Fig. 5A and Table 2). The Guinier plots at low angles are linear and revealed good data quality with no indication of protein aggregation (Fig. 5A, inset). The Kratky plot (I(s) ϫ s 2 versus s) (Fig. 5B, inset) shows a bell-shaped peak at low angles indicating a well folded protein. The radius of gyration (R g ) values from the Guinier approximation were consistent at the two concentrations measured with a value of 19.3 Ϯ 0.6 Å. The distance distribution functions (p(r)) were similarly shaped for the concentrations used (Fig. 5B), and the maximum particle dimension is D max ϭ 67.5 Ϯ 3 Å. The gross shape of the subunit F was ab initio reconstructed and had a good fit to the experimental data in the entire scattering range and had a discrepancies of 2 ϭ 0.951. Ten independent reconstructions produced similar envelope with normalized spatial discrepancy of 0.65 Ϯ 0.01, and the average structure is shown in Fig. 5D.
The molecular mass as determined by the I 0 value, Porod hydrated volume, and from the SAXS MoW server (55) ranged from 21 to 24 kDa, indicating the presence of a dimer or a mixture of both monomer and dimer in solution. Comparison of the theoretical scattering curves for the monomer and dimer in both parallel and perpendicular conformations in CRYSOL confirmed the presence of dimer formation with the fit being improved when both conformations of the dimer were present in equal amounts. Based on the above observation, a dimer model was generated using CORAL (52), which allows flexibility for the loop with the residues 95-105 between the globular domain and the C-terminal helix ␣5 of ScF. This program performs modeling of complexes against the experimental SAXS data using a combined rigid body for the atomic structures of the domains and the ab initio modeling of the unknown loop regions. The CORAL model fits with ϭ 1.24 and reveals that the two conformations (parallel and perpendicular) co-exists in the same dimer with the C terminus being more compact in one and more extended in another molecule of the dimer (Fig. 5, C  FIGURE 3. Surface electrostatic and hydrophobicity of DF assembly and sequence comparison of subunit D. A, surface electrostatic potential of the ScDF complex with ScF shown as schematic and the DF interface labeled by an arrow is shown in the center. Surface electrostatic potential of ScD reveals that the interacting surface with ScF is mostly hydrophobic, whereas the exposed N and C termini of ScD are composed of basic and predominantly acidic residues (left). The interacting surface of ScF is predominantly hydrophobic (right). The electrostatic potential surfaces were calculated using APBS (56) and mapped at contouring levels from Ϫ3 kT (blue) to 3 kT (red). B, sequence alignment of subunit D from different V-ATPases and ATP synthases, respectively. The secondary structure elements of subunit D of the E. hirae A-ATP synthase and the S. cerevisiae V-ATPase are shown. The conserved sequence in the eukaryotic V-ATPases is highlighted in red. FEBRUARY 6, 2015 • VOLUME 290 • NUMBER 6

JOURNAL OF BIOLOGICAL CHEMISTRY 3189
and D). These data reveal that the C-terminal region of subunit F can undergo dynamic movements in solution.

DISCUSSION
The crystallographic structure of subunit F of the ScDF1 and  ScF in the ScDF1 molecule as well as the 116-amino acid structure of subunit F in the ScDF2 molecule give insights of the important C terminus with its flexible loop (Pro-95-Asp-106) and helix ␣5. The comparison of both subunit F structures (Fig.  4A) reveals differences mainly in the C terminus of both molecules indicated by the high r.m.s.d. value (3.58 Å). The Pro-95-Asp-106-loop with its bent feature at Lys-105 enables structural rearrangements, leading to a perpendicular or parallel orientation of helix ␣5 relative to the N-and C-terminal helix of subunit D (Fig. 4A). The solution x-ray scattering data of S. cerevisiae V-ATPase subunit F confirm the flexibility of the Pro-95-Asp-106 loop and demonstrate that this loop and helix ␣5 are able to undergo a rearrangement from a compact sub-unit F to an extended C-terminal formation (Fig. 5D). These data extend the recent findings of 15 N-( 1 H) heteronuclear NOE studies on ScF, demonstrating that helix ␣5 with the adjacent Pro-95-Asp-106 loop is flexible in solution (18). As demonstrated by the structure of the ScDF1 and ScDF2 molecules, the flexibility of the C-terminal segment of ScF leads to different interactions with the rotary subunit D, in which the C terminus of subunit F is either in proximity to the subunit D residues Asp-51, Lys-54, and Asp-160 (ScDF1 molecule) or with the subunit D amino acids as Thr-47 and Asp-160 (ScDF2 molecule; Fig. 4C). We showed most recently that subunit F of the eukaryotic S. cerevisiae V-ATPase activates ATP hydrolysis of the related archaea A 1 complex, in which ScF was reconstituted  FEBRUARY 6, 2015 • VOLUME 290 • NUMBER 6 with the A 3 B 3 D complex of the archaeal Methanococcus mazei Gö1 A-ATP synthase. 5 In the same set of experiments, it was demonstrated that no A 3 B 3 DF complex could be formed when the C-terminal truncated ScF(1-94) was used, which underlines the need of the C terminus of subunit F for proper assembly and enzyme activity.

Atomic Structure of Subunits D and F of Eukaryotic V-ATPase
The question arises to how the Pro-95-Asp-106-loop with helix ␣5 of ScF connects the activation of ATP hydrolysis in the catalytic A 3 B 3 headpiece of the enzyme. When the presented ScDF1 and ScDF2 molecules, respectively, were docked into the nucleotide-bound A 3 B 3 headpiece of the related E. hirae A-ATP synthase (26), no interaction between the compact form of ScF (ScDF1 molecule) and the nucleotide-binding subunits A or B was observed. In comparison, Lys-114 of the extended ␣5 helix of subunit F (ScDF2 molecule) comes in close proximity (3.3 Å) to Asp-480 of the AMP-PNP-bound catalytic A subunit (Fig. 6A), allowing a direct coupling between the catalytic A subunit and the central stalk subunit F. The subunit A-F arrangement confirms the requirement for the very C terminus of subunit F, as described by the C-terminal truncated ScF(1-94) above. 5 As shown for the related bacterial F-ATP synthase, the central stalk subunits ␥⑀ in ␣ 3 ␤ 3 ␥⑀ complexes, which are proposed to be homologues to the DF assembly, and the A 3 B 3 headpiece of V-ATPases, rotate in 80°and 40°substeps (57-60), or as most recently shown for the human mitochondrial F 1 -ATPase subunit, ␥ rotates in three 120°steps with sub-steps of 0 -65, 65-90, and 90 -120° (61). Here, we moved the DF assembly inside the A 3 B 3 shaft by 80°and observed that amino acid Lys-114 of the extended helix ␣5 of ScF comes in the neighborhood to Glu-391 of subunit B, belonging to the C-terminal region 388 VLGESALSDI 397 of subunit B (Fig. 6, B and C), which occupies a similar position to the so-called DELSEED region of the nucleotide-binding subunits ␣ and ␤ of F 1 F 0 -ATP synthases (62). At the same time the subunit F residues Arg-80, Arg-83, and Thr-51, Asp-52, Asn-55, Glu-59, Glu-60 of helix 2 and helix 4, respectively, are close to subunit A residues Ile-584, Val-585, Ser-586, and Asn-543, Glu-544, Glu-547, Gly-548, respectively (Fig. 6B). These interactions of subunit F with several subunit A and B residues may change the nucleotide-binding and ATP cleavage steps in the catalytic subunit A. The 80°movement of the DF assembly also brings the bended C-terminal helix of ScD in the interface of the nucleotide-free subunits A and B (Fig.  6D), where the amino acids Asn-190 and Glu-199 come in proximity to the residue Thr-439 and Lys-335 of subunit A and B, respectively. In the future, single molecule rotation studies of the S. cerevisiae A 3 B 3 DF complex may provide more details about sub-steps of the central stalk subunits DF inside the V 1 -ATPase. This will be of great interest, because subunit F and ⑀ reveal no structural similarity, and subunits D and ␥ are similar in respect to their N-and C-terminal helices. When the globular domain of the E. coli subunit ␥ (Ec␥) was superimposed with ScF, a similar fold of both domains could be observed with respect to the central parallel ␤-sheet (␤1-␤3), which is flanked by the ␣-helices ␣1 and ␣2 on both sides of its plane (Fig. 7). The sequence similarity of the globular domain of Ec␥ and ScF is 18.9%. Future studies may show whether subunit ␥ is a fused form of subunits D and F of eukaryotic V-ATPases as well as the related and evolutionary older A-ATP synthases, respectively (5).
The structural differences observed in the C terminus of subunit F in the ScDF1 and ScDF2 molecules of our presented structure depicts not only two mechanistically relevant conformations but also shows the stabilizing effect in the DF complex via the formation of salt bridges at the interface of the C termini of both subunits. The perpendicular orientation of ScF helix ␣5 relative to the N-and C-terminal helix of subunit D leads to the interaction of Asp-51 and Asp-160 of ScD (ScDF1 molecule) with amino acids Arg-111 and Arg-113 of subunit D, respectively (Fig. 4B). This may stabilize the coiled coil formation during rotary steps of the central DF assemblage. As shown for the E. hirae A-ATP synthase, depletion of the C-terminal residues of subunit F led to insolubility of subunit D, indicating the stabilizing role of the C terminus of subunit F in DF formation (26).
As described recently and confirmed here, subunit F of eukaryotic V-ATPases contains the conserved 26 GQITPETQEK 35 loop (Fig. 4B), which faces the C-terminal serine residue Ser-381 of subunit H (18), revealed to be involved in F-H crosslinking formation (Fig. 1) (18, 63). It has been proposed that in the process of V 1 and V O dissociation, the flexible C-terminal domain of subunit H moves slightly closer to its nearest neighbor, the exposed 26 GQITPETQEK 35 loop of subunit F, where it causes conformational changes, leading to an inhibitory effect of ATPase activity in the V 1 -ATPase (18,63). The present ScDF structure shows for the first time that the unique 26    actions between the guanidine side chain of subunit D residue Lys-93 and the carbonyl side chain of ScF residue Gln-27 as well as the one between Gln-84 of ScD and Lys-35 of ScF may play significant roles in this interaction. Interestingly, amino acid Lys-93 is highly conserved in subunit D of eukaryotic V-ATPases and is located at the beginning of the ␤-hairpin (Fig.  4B), which is reported to stimulate ATPase activity 2-fold in the bacterial E. hirae A-ATP synthase (25). Furthermore, amino acid Glu-85 is located at the end of the short helix ( 79 IGYQVQE 85 ) of subunit D. This short helix is discussed to be involved in energy coupling and to play an important role in reversible disassembly (18). 15 N 1 H heteronuclear NOE studies on the S. cerevisiae subunit F revealed a rigid core formed by ␤-strands, ␤1 to ␤4, and ␣2 to ␣4. In comparison, the N-and C-terminal helices ␣1 and ␣5 with their adjacent loops 26 GQITPETQEK 35 and 94 IPSKDHPYD 102 , respectively, are more flexible in solution (18). The N-terminal helix ␣1 of subunit F and the bottom segment of subunit D form the neighborhood with subunit d (18). It has been proposed that this area is important during the process of dissociation and reassembly of the V 1 and V O sectors. In this scenario, the higher flexibility of ␣1 in subunit F would allow transmission of a change in subunit d during dissociation from the DF heterodimer and also the movement of subunit H closer to F, via the neighboring 26 GQITPETQEK 35 loop (18). In the latter case, the signal of V 1 V O disassembly may be communicated by the 26 GQITPETQEK 35 loop to the neighboring 85 ESVSTARFK 93 loop of ScD and in turn to the short helix ( 79 IGYQVQE 85 ) as well as the ␤-hairpin of subunit D. In addition, the 26 GQITPETQEK 35 loop would transfer the H to F movement via the second flexible 94 IPSKDHPYD 102 loop, which is adjacent to helix ␣5. Subsequently, helix ␣5 moves from its extended position away from the catalytic A subunit to the perpendicular orientation, leading to reduction of ATPase activity.