Affinity Purification and Structural Features of the Yeast Vacuolar ATPase Vo Membrane Sector*

Background: Vacuolar ATPase is regulated by reversible disassembly into V1-ATPase and the Vo proton channel. Results: Upon enzyme disassembly, subunit a cytoplasmic domain changes conformation to bind subunit d. Conclusion: The conformational change of subunit a may play a role in blocking passive proton flow through free Vo. Significance: The interaction of aNT and d in free Vo ensures the efficient reassembly of the holo enzyme. The membrane sector (Vo) of the proton pumping vacuolar ATPase (V-ATPase, V1Vo-ATPase) from Saccharomyces cerevisiae was purified to homogeneity, and its structure was characterized by EM of single molecules and two-dimensional crystals. Projection images of negatively stained Vo two-dimensional crystals showed a ring-like structure with a large asymmetric mass at the periphery of the ring. A cryo-EM reconstruction of Vo from single-particle images showed subunits a and d in close contact on the cytoplasmic side of the proton channel. A comparison of three-dimensional reconstructions of free Vo and Vo as part of holo V1Vo revealed that the cytoplasmic N-terminal domain of subunit a (aNT) must undergo a large conformational change upon enzyme disassembly or (re)assembly from Vo, V1, and subunit C. Isothermal titration calorimetry using recombinant subunit d and aNT revealed that the two proteins bind each other with a Kd of ∼5 μm. Treatment of the purified Vo sector with 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-rac-(1-glycerol)] resulted in selective release of subunit d, allowing purification of a VoΔd complex. Passive proton translocation assays revealed that both Vo and VoΔd are impermeable to protons. We speculate that the structural change in subunit a upon release of V1 from Vo during reversible enzyme dissociation plays a role in blocking passive proton translocation across free Vo and that the interaction between aNT and d seen in free Vo functions to stabilize the Vo sector for efficient reassembly of V1Vo.

endomembrane system of all eukaryotic organisms, where they function to acidify the lumen of intracellular organelles (1)(2)(3). In specialized cells of higher eukaryotes, V-ATPases are also found in the plasma membrane, where the enzyme is pumping protons into the extracellular milieu (1,4). Aberrant V-ATPase function has been shown to be associated with numerous widespread human conditions, including renal tubular acidosis (5), sensorineural deafness (6), osteoporosis (7), diabetes (8), microbial (9) and viral infection (10), infertility (11), and cancer (12). V-ATPase has therefore been deemed a valuable drug target (13)(14)(15). However, lack of detailed structural information has limited the application of structure-based drug discovery so far.
V-ATPase can be divided into an ATP-hydrolyzing catalytic headpiece, V 1 , and a membrane-embedded proton-translocating sector, V o . The subunit composition of the enzyme from the model organism Saccharomyces cerevisiae is A 3 B 3 (C)DE 3 FG 3 H for V 1 (16) and ac 8 cЈcЉde for V o (17,18). Crystal structures of the bacterial V 1 -ATPase from Enterococcus hirae show the three A and B subunits arranged in an alternating fashion around a central cavity within which are located the N-and C-terminal ends of subunit D (19). The proton channel is formed at the interface of the ring of the c, cЈ, and cЉ subunits ("proteolipid" ring) and the C-terminal domain of subunit a (a CT ) (17,20). V-ATPase is a member of the family of rotary molecular motor enzymes that, next to V-ATPase, includes F-ATP synthase, found in bacteria, mitochondria, and chloroplasts; archaeal A-ATP synthase; and bacterial A/V-like ATPase (21)(22)(23). In eukaryotic V-ATPase, ATP hydrolysis taking place at three catalytic sites located at the interface of the A and B subunits on the membrane extrinsic V 1 is coupled to proton translocation across the V o via a central rotor formed by the DF heterodimer of the V 1 and the subunit d-proteolipid ring subcomplex of the V o . Three peripheral stalks, formed by subunit EG heterodimers together with the single-copy H and C subunits, form the stator that links the catalytic sector to the membrane-embedded proton channel via a NT , and that functions to withstand the torque generated during rotary catalysis (Fig. 1A). However, unlike the related F-, A-and bacterial V-type motors, eukaryotic vacuolar ATPase is regulated by a reversible disassembly and reassembly mechanism employed by the organism to modulate the activity of the complex in response to, e.g., nutrient availability or developmental state (24,25). The mechanism of reversible disassembly has been studied extensively in the yeast system, and it is known that V-ATPase dissociation results in a cytoplasmic V 1 and a membrane-bound V o , with the activity of the two sectors silenced (26,27) (Fig. 1B). In isolated V 1 , magnesium ATPase activity is silenced by subunit H, possibly together with inhibitory magnesium ADP (26,28). The mechanism of blocking passive proton translocation across isolated V o , however, is less well understood, in part because of a lack of detailed structural information for the eukaryotic V-ATPase membrane sector.
Previously, we obtained three-dimensional reconstructions of free V o (29) and holo V-ATPase from bovine brain clathrincoated vesicles (30) and yeast vacuoles (31), and, by fitting available crystal structures into the EM density, we were able to generate a pseudo-atomic model of the enzyme from yeast (31). Recent intermediate-resolution cryo-EM reconstructions of yeast (18,32) and insect (33) V-ATPase provided additional structural detail and allowed a first view "inside" the membrane sector of the complex.
Here we developed a procedure for purifying the yeast V o sector amenable for biochemical and biophysical characterization. Electron microscopy of V o single particles and two-dimensional crystals showed a ring-like structure with additional protein densities at the periphery and cytoplasmic side of the ring. A comparison of the structure of free versus holo V-ATPase-bound V o revealed that enzyme regulation by reversible disassembly involves a large structural rearrangement of a NT from a peripheral position seen in V 1 V o (where a NT interacts with subunit C and the peripheral stator EG2) to a position in free V o where a NT binds subunit d. We speculate that the conformational change in a NT that accompanies V-ATPase dissociation plays a role in activity-silencing in the isolated V o sector and that the interaction between subunit d and a NT seen only in free V o stabilizes the membrane sector to ensure efficient reassembly of the holo enzyme.
Yeast Strains and Growth-Yeast strain YSC1178-7502926 with a TAP tag fused at the C terminus of subunit a (a TAP ) was from Open Biosystems. To disrupt the subunit B (Vma2p) gene in YSC1178-7502926, a fragment of genomic DNA containing the NAT1 fragment was amplified by PCR from genomic DNA of strain BY4741 vma2⌬::nat1 using oligonucleotides vma2-840 5Ј-GAATCGGCTAGAGATTACAAC-3Ј and vma2-c4 5Ј-CAT-GTTCTTCGAGACCGGGTTG G-3Ј. The resulting 1.2-kb product was used to transform YSC1178-7502926. Transformed colonies were selected on yeast extract (10 g/liter), peptone (20 g/liter), dextrose (20 g/liter) plates supplemented with 100 g/ml ClonNat (nourseothricin, Werner BioAgents). Colonies were further selected on the basis of their ability to grow on yeast extract, peptone, dextrose buffered to pH 5.0 but not on yeast extract, peptone, dextrose buffered to pH 7.5 ϩ 60 mM CaCl 2 . Western blot analysis of whole-cell lysates using anti-TAP and anti-subunit B antibodies were performed to confirm the presence and absence of a TAP and subunit B, respectively. The resulting strain was grown in yeast extract, FIGURE 1. Yeast V-ATPase subunit architecture and regulation by reversible disassembly. A, schematic of yeast V-ATPase subunit architecture. B, V-ATPase is regulated by nutrient-dependent reversible disassembly, resulting in V 1 and V o sectors that have no magnesium ATPase and passive proton translocation activity, respectively. The rotor is shown in green, the stator in red/orange, the catalytic core in cyan/blue, and subunit H in purple.
peptone, dextrose supplemented with 50 mM KH 2 PO 4 and 50 mM succinic acid (pH 5). For large-scale biomass production, cells from 8 -10 liters of flask culture (A 600 ϳ7) were collected by centrifugation, transferred to a 10-liter fermenter, and grown to the second diauxic log phase. Cells were harvested by centrifugation, washed once with distilled water, and stored at Ϫ80°C until use. Final cell weight was 8 -10 g/liter of culture.
Isolation of Membranes-All steps were performed at 4°C unless noted otherwise. Cells were resuspended in lysis buffer (20 mM Tris-HCl, 150 mM NaCl (pH 7.4) (TBS) supplemented with 8% sucrose, 2% sorbitol, and 2% glucose), and an inhibitor mixture was added to a final concentration of 2 g/ml leupeptin, 2 g/ml pepstatin A, 0.5 g/ml chymostatin, and 1 mM PMSF. 1 mM EDTA was added before disrupting cells in a homemade bead beater using 0.5 mm Zirconia beads (BioSpec), keeping the temperature below 14°C inside the chamber. Cell debris was removed by low-speed centrifugation (1200 ϫ g, 10 min), and crude membranes were collected by ultracentrifugation at 130,000 ϫ g for 1 h and washed once in lysis buffer. The final membrane pellet was resuspended in the presence of the inhibitor mixture mentioned above. Protein concentration was measured, and membranes were frozen at Ϫ80°C until use.
V o Purification-Isolated membranes were diluted to a final concentration of 10 mg/ml in lysis buffer, and inhibitor mixture was added. Extraction was carried out by adding DDM from a 20% stock solution in water to a final concentration of 2 mg of detergent/1 mg of protein, followed by gentle stirring for 1 h. Extracted membranes were cleared by ultracentrifugation at 106,000 ϫ g for 1 h, and the pellet was discarded. The supernatant was collected carefully, avoiding the upper lipid layer, and CaCl 2 was added to a final concentration of 4 mM. The mixture was incubated with 4 ml of Calmodulin beads for 1 h at 4°C under gentle agitation. The beads were collected in a chromatography column and washed with 20 column volumes of 10 mM Tris-HCl (pH 8), 10 mM ␤-mercaptoethanol (BME), 2 mM CaCl 2 , 0.1% DDM, 150 mM NaCl, and 20 column volumes of the same buffer without NaCl. The column was eluted with 10 mM Tris-HCl (pH 8), 10 mM BME, 0.5 mM EGTA, and 0.1% DDM. Fractions were analyzed by 13% SDS-PAGE, and fractions containing V o were pooled and concentrated in a 100-kDa Vivaspin concentrator (Sartorius Stedim Biotech).
Glycerol Gradient Centrifugation and Removal of Subunit d-1 mg of purified V o was applied to the top of a discontinuous glycerol gradient (15-35% (v/v), 10 mM Tris-HCl (pH 8), 10 mM BME, 0.5 mM EGTA, and 0.01% phosphatidylcholine:phosphatidic acid (19:1)) and centrifuged at 200,000 ϫ g for 16 h at 4°C. For removal of subunit d, 0.05% LPPG was included in the gradient. Otherwise, 0.5% CHAPS was used. Fractions were collected from the bottom and analyzed by SDS-PAGE.
Reconstitution in Liposomes-200 g of V o (in CHAPS) or subunit d-depleted V o (V o ⌬d) (in LPPG) was mixed with 15 mg of phosphatidylcholine:phosphatidic acid (19:1 v/v) and adjusted with CHAPS to 6%. In some experiments, 9% ergosterol was included in the reconstitution mix. Samples were applied to a Sephadex G50 column (50 cm ϫ 1.6 cm) and eluted with high-potassium buffer (20 mM HEPES (pH 7), 2 mM BME, 0.2 mM EGTA, 10% glycerol, 100 mM K 2 SO 4 , and 0.5 mg/ml fatty acid-free BSA) at a flow rate of 0.5 ml/min. The eluate was collected in 1-ml fractions, and turbid fractions were analyzed by 13% SDS-PAGE and silver staining.
Passive Proton Translocation Assay-Proton translocation assays were performed as described by Qi and Forgac (34). Briefly, assays were conducted in a 3-ml cuvette. 30 l of each fraction was preincubated in high-sodium buffer (20 mM HEPES (pH 7), 2 mM BME, 0.2 mM EGTA, 10% glycerol, 150 mM NaCl, and 0.5 mg/ml fatty acid-free BSA) for 5 min at 30°C in the presence of 2 M 9-amino-6-chloro-2-methoxyacridine. After 300-s incubation, the process was started by addition of 1 M valinomycin, followed by 1 M carbonyl cyanide p-trifluoromethoxyphenylhydrazone. Different fractions were probed after 1, 4, and 10 days. Six independent preparations of V o and V o -d were analyzed.
Flotation Assays-To verify lipid vesicle reconstitution of V o and V o -d, fractions from the G50 column (fraction 29 for V o and fraction 28 for V o -d) were adjusted with sucrose to 53% and placed in the bottom of an 11-ml tube. A sucrose step gradient (40,20, and 0% (w/v) in 20 mM HEPES (pH 7), 2 mM BME, 0.2 mM EGTA, 10% glycerol, and 150 mM NaCl) was layered on top of the V o sample and centrifuged at 200,000 ϫ g for 16 h at 4°C. Fractions were collected from the top of the tube and analyzed by 13% SDS-PAGE and silver staining.
Mass Spectrometry of V o Subunits-The V o sector was precipitated with 1% trichloroacetic acid, and the centrifuged pellet was washed with water. The pellet was extracted with a 1:1 mixture of water and trifluoroethanol, and the soluble fraction was analyzed by electrospray ionization mass spectrometry using a Q-TOF Micro mass spectrometer (Waters, Inc.) in positive ion mode. Charge envelopes between 800 -2500 m/z were deconvoluted using MaxEnt2 as implemented in MassLynx4.1. Calibration of the instrument was carried out with phosphoric acid and sodium/cesium iodide. Analysis of gel bands by peptide sequencing was done at the Upstate Medical University mass spectrometry core facility using a Thermo LTQ Orbitrap mass spectrometer.
Small-angle X-ray Scattering Analysis-Small-angle x-ray scattering (SAXS) data were collected at the Cornell High-energy Synchrotron Source (MacCHESS) F2 beam line operating at a wavelength of 1.2524 Å at 4°C. For SAXS data collection, V o was purified using UnDM instead of DDM. V o was diluted into 10 mM Tris-HCl (pH 8), 10 mM BME, 0.5 mM EGTA, and 0.05% UnDM to 1, 2, 4, 6, 8, and 10 mg/ml. 30-l samples were exposed twice for 180 s without an obvious decay in signal. Signal averaging, buffer subtraction and Guinier analysis were done in Bioxtas RAW (35). Molecular weight was estimated using lysozyme as the standard. Thirty data points were used for Guinier analysis for each tested concentration (qRg was ϳ1.1 in all cases).
Two-dimensional Crystallization and EM Analysis-Purified V o was diluted to 3 mg/ml in 10 mM Tris-HCl (pH 6.5), 10 mM BME, 0.5 mM EGTA, 10% glycerol, and 1 mM DTT, and sonicated DOPC was added to reach a lipid-to-protein ratio of 0.3 (w/w). After 24 h, the detergent was removed by stepwise addition (every 3 days) of equal amounts of polystyrene beads (Bio Beads SM2, Bio-Rad) for a total of 10 days so that the final ratio of beads to liquid was ϳ1:1. Samples were kept at 4°C, and 1 mM sodium azide was added to inhibit bacterial growth. V o two-dimensional crystals were spotted on glow dischargetreated, carbon-coated copper grids and stained with 1% uranyl acetate. Micrographs were recorded on a 4096 ϫ 4096 chargecoupled device (TVIPS F415MP) at ϫ20,000 -40,000 electron optical magnification and an underfocus of 1.5 m. The quality of the crystalline areas was assessed from calculated power spectra, and areas showing isotropic reflections to ϳ20-Å resolution were excised and analyzed with the 2dx package of programs (36) and/or by correlation averaging as implemented in IMAGIC 5 (37,38).
Cryo EM and Single-particle Image Analysis-The solubilized V o sector was vitrified at 1-2 mg/ml on glow-discharged, holey, carbon-coated copper grids (C-flat, 2/2 m). Grids were mounted in a Gatan 626 cryoholder and imaged in a JEOL JEM-2100 transmission electron microscope operating at 120 kV. Micrographs were recorded on a 4096 ϫ 4096 charge-coupled device (TVIPS F415MP) at an electron optical magnification of ϫ40,000 and an underfocus of between 1.5-2.5 m. The calibrated pixel size on the specimen level was 2.62 Å. A total of 12,035 particles was extracted as 144 ϫ 144 pixel images using the "boxer" program of the EMAN1.9 software package (39). The dataset was contrast transfer function-corrected using "ctfit" as implemented in EMAN1.9. All subsequent image analysis was done with the IMAGIC 5 package of programs (37). Images were bandpass-filtered to remove low (Ͻ6.4 ϫ 10 Ϫ3 Å Ϫ1 ) and high (Ͼ0.15 Å Ϫ1 ) spatial frequencies, and a soft-edge circular mask was applied before subjecting the images to reference-free alignment (40). Averages from the reference-free alignment were then used in subsequent multireference alignment, and the multireference alignment was iterated to obtain averages of the most abundant projections. Three-dimensional reconstruction was initiated by one round of projection-matching using the low pass-filtered three-dimensional reconstruction of the bovine V-ATPase membrane sector (29) as a reference model. Because the bovine V o sector subunit Ac45 is not found in yeast V o , the density corresponding to Ac45 was removed with the volume eraser tool as implemented in the visualization software Chimera (41). Cycles of projectionmatching alignment and three-dimensional reconstruction were iterated with increasing numbers of references until no further improvement was observed. The resolution of the final model was estimated using the 0.5 Fourier shell correlation criterion (42). The final EM density was fitted manually with crystal structures of bacterial homologs for the yeast V o subunits, including the a NT homolog from Meiothermus ruber (I NT , PDB code 3RRK), the subunit d homolog from Thermus thermophilus (subunit C, PDB code 1R5Z), and the c subunit ring homolog from E. hirae (K 10 , PDB code 2BL2). Although the primary sequence conservation between the yeast and bacterial subunits is limited (11%, 16%, and 26% for a NT and I NT , d and C, and c and K, respectively), their secondary and tertiary structure is highly conserved, as evidenced by the fact that the yeast subunit structures can be modeled on the basis of the bacterial A/V-ATPase subunit crystal structures using the Phyre2 server (43) with 100% confidence.
Expression and Purification of a NT  and Subunit d-Plasmid pRS316 containing the open reading frame for subunit d (Vma6p) was as a gift from the laboratory of Dr. Karlett Parra (University of New Mexico). The coding sequence for subunit d was PCR-amplified using primers GCTCAGGT ACCGATGG-AAGGCGTGTATTTCAATATT (forward) and CGAGTCC-TGCAGTCAATCAATAAACG GAAATATAATT (reverse), and the resulting PCR product was ligated into pGEM T-easy. Subsequent cloning of the subunit d coding sequence into a modified plasmid pMAL-c2e (New England Biolabs, enterokinase cleavage site replaced by the human rhinovirus 3C site) for bacterial protein expression was done by BioBasic, Inc. (Markham, Ontario). The resulting construct consisted of subunit d with an N-terminal fusion of maltose binding protein (MBP) separated by a protease cleavage site (human rhinovirus 3C protease) for removal of MBP. pMAL-c2E harboring MBP subunit d was expressed in Escherichia coli strain Rosetta2. Cells were grown in rich broth (Lennox broth plus 0.2% glucose) to an A 600 of 0.6 and induced with 500 M isopropyl 1thio-␤-D-galactopyranoside for 4 h at 37°C. Purification was done following the recommended protocol for MBP-tagged proteins (New England Biolabs). The PreScission Proteasecleaved (GE Healthcare) fusion was dialyzed against 25 mM Tris-HCl, 1 mM EDTA, 1 mM tris(2-carboxyethyl)phosphine (pH 7), followed by anion exchange chromatography on a 1-ml mono Q-Sepharose column attached to an AKTA FPLC (GE Life Sciences). Under these buffer conditions, subunit d bound to the column and was eluted using a 0 -500 mM sodium chloride gradient in the same buffer. Protein-containing fractions were pooled, concentrated to 1-2 ml, and subjected to size-exclusion chromatography over Superdex S75 (16 ϫ 500 mm). An expression construct for the N-terminal domain of subunit a (Vph1p) consisting of residues 1-372 (a NT(1-372) ) was generated as described previously (44). The cleavable MBP tag resulted in an N-terminal extension for both subunit d and a NT(1-372) constructs with the amino acid sequence GPKVP. Constructs were confirmed by DNA sequencing. A detailed biochemical and biophysical characterization of recombinant subunit d and a 1-372 will be presented elsewhere.
Isothermal Titration Calorimetry (ITC)-ITC measurements of the interaction of a NT(1-372) with subunit d using a Microcal VP-ITC isothermal titration calorimeter were done as described previously (44,45), with the following modifications. Prior to the titration, both proteins were dialyzed (in the same container) against 2 liters of 25 mM Tris-HCl (pH 7), 0.5 mM EDTA, and 1 mM tris(2-carboxyethyl)phosphine. a NT(1-372) was concentrated to 375 M and titrated into 25 M subunit d at 10°C using a total of 30 injections with 10.7% saturation/injection. A heat of dilution titration of 375 M a NT(1-372) into dialysis buffer was subtracted from the a NT(1-372) into subunit d titration. A second titration was carried out with 320 M a NT  in the syringe using, again, 25 M subunit d in the ITC cell. Both titrations produced very similar results. Protein concentrations were determined from A 280 using calculated extinction coefficients. ITC data were fitted to a one-site model using the VP-ITC programs in Originlab.
Other Methods-Membrane protein concentrations were measured by BCA method (Thermo Scientific) and improved by TCA precipitation as in Lowry-TCA (46) using fatty acidfree BSA as the standard.

Results
Purification of the V o Proton Channel Sector-To isolate the yeast V o sector for structural and functional studies, we attempted several strategies, including C-and N-terminal histidine tags fused to subunits d and cЉ as well as a TAP tag fused to the C terminus of subunit a or c. In the end, the best results were obtained using only the second affinity step of the TAP procedure, where V o (containing TAP-tagged subunit a) is detergent-solubilized from vacuolar membranes and captured by a calmodulin column by way of the calmodulin binding peptide in the tag. To eliminate possible co-purification of (partially) assembled V-ATPase, the gene for V 1 -ATPase subunit B (VMA2) was disrupted. For large-scale purification of the vacuolar ATPase membrane sector, yeast was grown in a 10-liter fermenter. The yield of V o sector purified as described under "Experimental Procedures" was ϳ2-3 mg/150 g of cells. Fig. 2 summarizes the purification of the yeast V-ATPase V o sector and characterization of the protein by negative-stain transmission electron microscopy and SAXS. Fig. 2A shows SDS-PAGE of fractions 1-4 eluted from the calmodulin column. The Coomassie-stained gel shows bands for subunits a, d, cЉ, c, cЈ, and e.  Fig.  2, A and B, only produced subunit a-derived peptides, with no peptides from protein A being detected, indicating that the protein A moiety of the TAP tag was lost because of proteolytic degradation following cell lysis and detergent extraction of membranes. Loss of protein A was confirmed by immunoblot analysis (using an antibody directed against the C-terminal end of the calmodulin binding peptide) that showed that the apparent molecular mass of the subunit a band decreased in size from an initial ϳ130 kDa at the washed membrane stage to the final ϳ116 kDa after the elution from the calmodulin affinity column (data not shown). Electrospray ionization/TOF mass spectrom-   NOVEMBER 13, 2015 • VOLUME 290 • NUMBER 46 for subunit d. Possibly because of their large size or lower abundance, no peaks for subunits a (100,143 Da without protein A), cЈ (16,902 Da), and cЉ (22,464 Da) were observed in the deconvoluted charge series (data not shown). Fig. 2C shows negative-stain transmission electron microscopy analysis of detergent-solubilized V o . The image shows homogeneously sized particles with a diameter of ϳ10 -15 nm, indicating that the preparation contains intact V o sectors that are stable in the detergent used for purification (DDM). Furthermore, Guinier plots of small angle x-ray scattering profiles obtained from solutions of V o sector purified in UnDM (chosen here for its smaller micelle size) showed that the preparation is monodisperse at concentrations of up to 10 mg/ml. The molecular mass of V o as estimated by SAXS was 544 Ϯ 33 kDa, with a calculated radius of gyration of ϳ51 Ϯ 3.5 Å. The expected mass of V o (assuming a subunit ratio of ac 8 cЈcЉde, see next paragraph and "Discussion") is ϳ320 kDa, resulting in a difference between the measured and expected mass of ϳ244 kDa. Considering the average size of UnDM micelles of 35 kDa (micelle size reported by Anatrace) suggests that each V o sector binds six to seven detergent micelles. Taken together, the data show that highly purified, stable, and monodispersed yeast V o sector can be obtained via affinity chromatography using a calmodulin peptide fused to the C terminus of subunit a.

Molecular Architecture of the Yeast V-ATPase Proton Channel
Two-dimensional Crystallization of the Yeast V o Sector- Fig.  3 summarizes the transmission electron microscopy analysis of yeast V o domain two-dimensional crystals. The two-dimensional crystals were obtained by mixing the purified V o domain at 3 mg/ml with 1 mg/ml DOPC, followed by removal of detergent (DDM) using polystyrene beads over a period of 7-10 days. V o two-dimensional crystals were visualized by negative-stain transmission electron microscopy, and images showing crystalline areas (Fig. 3, A and B) with reflections in calculated power spectra extending to the first zero of the contrast transfer function (ϳ24 Å; Fig. 3, C and D) were processed as described under "Experimental Procedures." The crystals belong to plane group P1 with an alternating up and down orientation of the molecules, as evident from the final projection map shown in Fig. 3E. At the current resolution of ϳ24 Å, the projections obtained in negative stain show a ring-like structure with a diameter of 8.5 nm and an asymmetric mass at its periphery (arrow and arrowheads, respectively, Fig. 3E). We interpret the ring to represent the proteolipid ring (for comparison, see the low pass-filtered projection of the E. hirae K 10 ring (PDB code 2BL2 (48), shown in Fig. 3F) and the asymmetric mass a CT . As can be seen in Fig.  3, E and F, both rings are of equal size, consistent with the now established number of 10 proteolipids in the yeast enzyme (18). Taken together, the data show that the detergent-solubilized yeast V-ATPase V o sector can be lipid-reconstituted and crystallized in two dimensions, opening a path for high-resolution structure determination of the V-ATPase membrane sector in its native environment.  (36). E, projection map with unit cell outlined by the rectangle. The cell dimensions are 101 ϫ 236 Å with 2 molecules/cell. The crystals belong to plane group P1 with alternating up and down orientation of the molecules. At the current resolution of ϳ24 Å, the projections obtained in negative stain show a ring-like structure (arrow) with a two-domain asymmetric mass at the periphery (arrowheads). We interpret the ring to represent the proteolipid ring and the asymmetric mass a CT . F, cross-section of the E. hirae K 10 ring as seen perpendicular to the plane of the membrane, filtered to a resolution of 16 Å. Fig. 4 summarizes single-molecule cryo-transmission electron microscopy analysis of the detergent-solubilized V o sector. V o was vitrified at a concentration of 2 mg/ml, and charge-coupled device (CCD) images were recorded at a magnification of ϫ40,000 (Fig. 4A). Three enlarged raw images of particles and three averages of aligned projections in the typical "side" and "top" view orientations are shown next to the micrograph. A dataset of 12,035 single-particle images was used to generate a three-dimensional model, starting with projections of the low pass-filtered three-dimensional reconstruction of the bovine V o sector (29) as initial references, followed by multiple rounds of projection-matching refinement. The resolution of the final model (Fig. 4B) was estimated to be ϳ18 Å on the basis of the 0.5 Fourier shell correlation criterion using models calculated from half-data sets (data not shown). The yeast V o sector is composed of a ring of 10 proteolipids (likely c 8 cЈcЉ, see "Discussion"), the 100-kDa a subunit that is equally divided into a cytoplasmic a NT and a membrane integral a CT that is bound at the periphery of the proteolipid ring, the ϳ40 kDa d subunit that is bound at the cytoplasmic rim of the proteolipid ring (Fig.  4C), and subunit e, which is likely bound to a CT (49). Because there are no x-ray crystal structures available for any of the yeast polypeptides, the EM model was fitted with crystal struc-tures of the homologous subunits of related bacterial rotary ATPases: the K 10 ring from E. hirae (PDB code 2BL2 (48)), I NT from M. ruber (PDB code 3RRK (50)), and the subunit d homolog from T. thermophilus (PDB code 1R5Z (51)). A comparison with the negative-stain three-dimensional model of the bovine V o (29) revealed an overall similar architecture except for the presence of density for Ac45 on the luminal side of the bovine model (a homolog for Ac45 is not present in yeast) and the presence of density for the linker (or tether) connecting a NT and a CT in the yeast cryo-EM model (Fig. 4C). At the current resolution of ϳ18 Å, the proteolipid ring within the membrane domain is not clearly resolved, likely because of the small size of the relatively featureless complex and the lack of internal symmetry. Well resolved, however, are a NT and subunit d, which are situated on the cytoplasmic side of the membrane. Fig. 4C, fitting the crystal structures of the bacterial homologs of yeast a NT (PDB code 3RRK (50)) and subunit d (PDB code 1R5Z (51)) into the yeast V o EM density shows an interaction of the distal lobe of a NT with the d subunit (Fig. 4D, right panel, arrow), with subunit d binding slightly off-center at the edge of the proteolipid ring. Interestingly, a comparison with the three-dimensional reconstruction of holo yeast V-ATPase (  4D, left and center panels) reveals that a NT in the holo enzyme adopts a different conformation where the distal domain of a NT is bound to C foot (subunit domain nomenclature as in Refs. 50 and 52, respectively) and the N termini of one of the three EG heterodimer peripheral stalks (EG2; Fig. 4D, center panel,  arrow). In line with this observation, we have previously characterized the binding interactions between a NT , C foot , and the EG heterodimer and we found that these interactions are of moderate affinity (44). Taken together, the comparison of V o and holo V 1 V o suggests that a NT undergoes a large conformational rearrangement upon enzyme disassembly, going from a conformation in free V o that binds the d subunit to a more peripheral conformation that binds C foot and EG2 in holo V 1 V o .

Comparison of EM Reconstructions of the Isolated V o and V o Sector as Part of Holo V-ATPase-As can be seen in
In Vitro Interaction of Recombinant Subunit d and a NT  -To test whether the interaction between a NT and subunit d as seen in the EM reconstruction is specific and can be quantified in vitro, we performed ITC experiments with recombinant subunits. For these experiments, subunit d and a NT  were expressed in E. coli as N-terminal fusions with MBP and affinity-purified on amylose resin. MBP was cleaved, and the resulting subunits were further purified using ion exchange and size-exclusion chromatography. Fig. 5A shows SDS-PAGE of a NT(1-372) (lane 1) and subunit d (lane 2). Both proteins are stable and highly soluble at pH 7, and, although recombinant subunit d elutes with an apparent molecular mass of ϳ42 kDa from a S200 gel filtration column (expected, 40,267 Da), suggesting a globular monomeric protein, a NT(1-372) exists in a concentration-dependent monomer-dimer equilibrium, as already described for the shorter a NT(104 -372) construct (Ref. 44) and data not shown). Fig. 5B shows a representative ITC experiment in which 375 M a NT(1-372) was titrated into 25 M subunit d. As can be seen from the titration, complex formation between a NT(1-372) and subunit d was exergonic, and fitting the data to a one-site model revealed an N value of 0.98 (consistent with a 1:1 stoichiometry of complex formation), a K a of 2.1 ϫ 10 5 Ϯ 3.5 ϫ 10 4 M Ϫ1 (K d , ϳ4.8 M), a ⌬H of Ϫ4.2 Ϯ 0.22 kcal/mol, a ⌬S of 9.7 cal/(K⅐mol), and a ⌬G of Ϫ6.9 kcal/mol. After the titration, the ITC cell content was resolved by gel filtration (S200, 16 ϫ 500 mm), and fractions were analyzed by SDS-PAGE (Fig. 5C). As can be seen from the gel, a NT(1-372) and subunit d co-elute around fraction 31 (62 ml), corresponding to an apparent molecular mass of ϳ72 kDa (84 kDa expected for the a NT(1-372) -d complex; subunit d alone elutes at an apparent molecular mass of ϳ42 kDa, see above). Together, the ITC and gel filtration data suggest that subunit d and a NT  bind each other in a specific manner, albeit with moderate affinity.
Preparation and Functional Analysis of the ac 8 cЈcЉe Subcomplex (V o ⌬d)-In living cells, V-ATPase activity is regulated by reversible dissociation into V 1 -ATPase and membrane integral V o sectors (Fig. 1B). Upon enzyme dissociation, the activity of both V 1 and V o is silenced so that V 1 loses the ability to hydrolyze magnesium ATP and V o becomes impermeable to protons. Considering the interaction described above between a NT and subunit d seen in free V o but not V 1 V o , we speculated that this interaction may contribute to the inhibition of passive proton translocation through isolated V o by blocking rotation of the c-ring past a CT . To assess the role subunit d might be playing in blocking proton translocation through isolated V o , we developed a procedure to selectively remove subunit d from V o to generate the ac 8 cЈcЉe subcomplex (V o ⌬d). Fig. 6, A and B, shows SDS-PAGE of glycerol density centrifugation of V o sector in presence of the ionic detergent LPPG and CHAPS, respectively. As can be seen from Fig. 6A, in the presence of LPPG, subunit d remains at the top of the gradient separated from ac 8 cЈcЉe, whereas, in CHAPS, subunit d migrates as part of intact V o . To determine whether removal of subunit d allows passive proton translocation through the resulting V o ⌬d, V o and V o ⌬d were reconstituted into liposomes in the presence of potassium chloride-containing buffer. Liposomes were collected by centrifugation and subjected to SDS-PAGE and silver staining (Fig. 6C). V o -and V o ⌬d-containing liposomes were subjected to a fluorescence-based assay to test for passive proton conductance. Fig. 6D shows a representative assay for passive proton conductance. As can be seen from the Fig. 6D, both V o -and V o ⌬dcontaining liposomes as well as control liposomes showed the same slow quenching upon addition of the potassium ionophore valinomycin and a sharp drop in the fluorescence signal following addition of the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone, which verified the presence of a proton gradient. This result indicates that removal of subunit d does not alleviate the inhibition of proton translocation across the V o , suggesting that the interaction of a NT and subunit d in free V o is not (solely) responsible for activity silencing in the isolated V-ATPase membrane sector.

Discussion
Transmembrane proton transport across the vacuolar ATPase V o sector involves rotation of the proteolipid ring past a CT . A major difference between eukaryotic vacuolar ATPase and the related F-and A-type motors is the mode of regulation of V-ATPase, which involves dissociation into free ATPase and proton channel sectors triggered by nutrient availability or developmental cues (Fig. 1). However, unlike F-and bacterial A/V-like ATPase ion channels, which, when detached from the ATPase, catalyze passive transmembrane proton transport (53,54), the eukaryotic V o sector becomes impermeable to protons ⌬d in CHAPS were reconstituted at a high lipid:protein ratio using gel filtration (Sephadex G50) in the presence of potassium chloride-containing buffer. Liposomes were collected by centrifugation and subjected to SDS-PAGE and silver staining. D, fluorescence-based assay for passive proton translocation. V o , V o ⌬d, and control liposomes prepared in high-potassium buffer were diluted into sodium-containing buffer in the presence of 9-amino-6-chloro-2-methoxyacridine (ACMA), and, after 5 min, a proton gradient was generated by the addition of the potassium ionophore valinomycin. The presence of a proton gradient was subsequently confirmed by addition of the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone. Both V o and V o ⌬d showed essentially the same behavior as control liposomes, indicating that removal of subunit d did not result in passive proton translocation activity. NOVEMBER 13, 2015 • VOLUME 290 • NUMBER 46 upon dissociation of the V 1 , preserving any existing proton gradient across the organellar membrane (55). In case of the F-AT-Pase ion channel, F o , it has been shown that rotation of the proteolipid ring relative to the a subunit is essential for passive proton conductance (56), and, because the mechanism of ion transport is highly conserved between F-and V-type motors, that means that rotation of the proteolipid ring in free V o appears to be blocked by an unknown mechanism. Previously, we generated an EM reconstruction of free V o sector from bovine brain V-ATPase that suggested an interaction between a NT and subunit d, although the linker connecting a NT and a CT was not resolved in the negative stain model (29). Later, we (57) and others (34) speculated that the interaction between a NT and d may serve to silence passive proton transport by linking the rotor and stator of the motor. To address the mechanism of activity silencing as well as other aspects of V o structure and function, we developed a protocol to isolate milligram amounts of yeast V o proton channel sector using affinity chromatography. Biochemical experiments show that the complex is stable in low and intermediate critical micelle concentration (cmc) detergents such as UnDM, DDM, and CHAPS, respectively, as evident from glycerol gradient centrifugation. Negative-stain and cryo-electron microscopy as well as small-angle x-ray scattering experiments show that detergent-solubilized yeast V o is monodisperse at concentrations up to several milligrams per milliliter, a prerequisite for structural studies.

Molecular Architecture of the Yeast V-ATPase Proton Channel
We next used transmission electron microscopy of two-dimensional crystals and single V o molecules to obtain structural information that may provide clues regarding the mechanism of activity silencing. Upon reconstitution into DOPC lipid bilayers, we were able to generate two-dimensional crystals of the V o , to our knowledge the first two-dimensional crystals of any eukaryotic V-ATPase proton channel sector. Projection maps calculated at a resolution of ϳ24 Å show a ring-like structure with an asymmetric mass bound at the periphery of the ring, consistent with current structural models obtained for the holo V-ATPase from single-particle reconstructions (18,(31)(32)(33). Although the relatively small size of the crystals obtained so far has limited our ability to use cryo-electron crystallography for structural analysis, there are several features that are noteworthy at the current resolution. The diameter of the ring is, with ϳ8.5 nm, virtually identical to the diameter of the K subunit ring of the related sodium V-like ATPase from E. hirae (K 10 , 8.3 nm (48)). This is consistent with the now established stoichiometry of 10 proteolipids for the yeast V-ATPase proteolipid ring (18), eight of which are c subunits (Vma3p), with each one copy of cЈ (Vma11p) and cЉ (Vma16p) (17) to give a complex of c 8 cЈcЉ. Another observation from the two-dimensional crystal projection is that a CT appears to be organized in two domains of slightly unequal size (Fig. 3E,  arrowheads). A recent cryo-EM model of another rotary motor enzyme, the dimeric F-ATPase from Polytomella mitochondria revealed two almost horizontal ␣ helices as part of the F o -a subunit (58). F-ATPase F o a subunits are predicted to contain five transmembrane ␣ helices, and it is possible that the larger of the two domains observed here for V o a CT in the two-dimensional crystals represents the structural and functional homologue of F o a, with the smaller of the two domains representing a part of V o a that is not present in F-ATPase. Almost horizontal transmembrane ␣ helices were also observed in a recent cryo-EM reconstruction of the holo yeast V-ATPase (18), suggesting that tilted ␣ helices in the interface between a CT and the proteolipid ring are a conserved feature in all rotary motor enzymes.
Because the current two-dimensional crystals are too small to generate a three-dimensional structural model from images of a tilted specimen, we used cryo-electron microscopy of detergent-solubilized single V o sectors to calculate a three-dimensional reconstruction of the complex using our earlier negative-stain, three-dimensional EM reconstruction of the bovine V o (29) as a starting model. Although the resolution of the yeast V o reconstruction presented here is limited to about 18 Å (likely because of the relatively small size of the V o (ϳ320 kDa), the presence of a featureless detergent belt, and the lack of overall symmetry), the EM density allows placing of the crystal structures of equivalent subunits from related bacterial enzymes, namely the E. hirae K 10 ring (48) and a NT and subunit d homologs from M. ruber (50) and T. thermophilus (51), respectively. The resulting pseudo-atomic model shows a NT and subunit d in close proximity, suggesting that the two polypeptides bind each other in free V o , consistent with what had already been described for the bovine complex (29,57). However, the cryo-EM model of yeast V o presented here provides more detail by showing the tether connecting a NT and a CT that was not resolved in the bovine model (29). Interestingly, a comparison of EM reconstructions of free V o and V o as part of holo V-ATPase (Fig. 4D) revealed that a NT must undergo a large conformational change during regulated enzyme disassembly, from a conformation in holo V-ATPase, where the distal domain of a NT binds C foot and EG2 (18,31,44), to a conformation in free V o , where a NT binds subunit d. As mentioned above, we initially reasoned that the a NT -d interaction may play a role in blocking passive proton conductance. However, as summarized in Fig. 6, removal of subunit d by the ionic detergent LPPG to produce V o ⌬d, followed by proton conductance assays, showed no difference in the behavior of the V o and V o ⌬d complexes. This result suggests that the interaction of d with a NT is not (solely) responsible for blocking proton flow across free V o , consistent with earlier experiments by Qi and Forgac (34) that showed that proteolytic removal of a NT on vacuolar vesicles did not render the membrane permeable to protons. Taken together, this means that there must be other (or additional) mechanisms that prevent proton leakage through free V o . As illustrated in Fig. 7, one possibility is that the conformational change in a NT upon enzyme dissociation is transmitted to a CT , thereby disrupting the path of protons along the interface between a CT and the proteolipid ring. Another mechanism for blocking passive proton transport could lie within the structure of the proteolipid ring itself. V-ATPase proteolipids have four transmembrane ␣ helices but only one essential proton carrying carboxylate (59), resulting in a larger distance between proton binding sites compared with F-ATP synthase. The large gap between proton binding sites (Fig. 7) could represent too high of a barrier to overcome without the driving force from ATP hydrolysis, resulting in kinetic inhibition of proton flow from the vacuole into the cytoplasm.
However, if the a NT -d interaction in free V o is not involved in blocking passive proton translocation, what then, if any, might its physiological role be? It has been shown that removing the tether linking a NT and a CT prevents assembly of holo V-ATPase (resulting in free cytoplasmic V 1 and vacuolar membrane-bound V o that lacks subunit d), a defect that can be partially rescued upon overexpression of subunit d (60). This finding suggests that the interaction of d with the proteolipid ring is relatively weak and that the additional interaction with a NT is needed to increase avidity for d during V o biogenesis and for retaining d upon regulated enzyme disassembly. In line with this model is the relatively weak affinity (K d , ϳ5 M) between a NT and d, as measured by ITC using recombinant subunits because this interaction must be readily reversible for enzyme reassembly. Interestingly, recent studies have shown that the vacuole-specific phosphoinositide PI(3,5)P 2 plays a role in regulating V-ATPase (re)assembly and that PI(3,5)P 2 is able to directly bind a NT (61). One possibility is that the PI(3,5)P 2 headgroups compete with subunit d for a NT binding, thereby helping to change the conformation of a NT from the free V o state to a more peripheral conformation in preparation for enzyme reassembly.
Currently, there is no high-resolution structure available for an intact membrane domain of any of the rotary motor enzymes, and this lack of structural information has limited our understanding of the mechanism of ion translocation and activity silencing in case of the eukaryotic V-ATPase. The protocol described here allows isolation of highly purified and stable V o , paving the way for obtaining a high-resolution structure of a rotary motor ATPase proton channel sector using crystallo-graphic or single-molecule techniques. Studies toward that aim are ongoing in our laboratory.
Author Contributions-S. C. C. and S. W. designed the study and wrote the manuscript. S. C. C. performed the V o purification and structural characterization with technical assistance from S. W. S. W. performed the ITC experiments, including recombinant protein purification. E. M. generated the subunit B deletion strain.