MgATP hydrolysis destabilizes the interaction between subunit H and yeast V1-ATPase, highlighting H's role in V-ATPase regulation by reversible disassembly

Vacuolar H+-ATPases (V-ATPases; V1Vo-ATPases) are rotary-motor proton pumps that acidify intracellular compartments and, in some tissues, the extracellular space. V-ATPase is regulated by reversible disassembly into autoinhibited V1-ATPase and Vo proton channel sectors. An important player in V-ATPase regulation is subunit H, which binds at the interface of V1 and Vo. H is required for MgATPase activity in holo-V-ATPase but also for stabilizing the MgADP-inhibited state in membrane-detached V1. However, how H fulfills these two functions is poorly understood. To characterize the H–V1 interaction and its role in reversible disassembly, we determined binding affinities of full-length H and its N-terminal domain (HNT) for an isolated heterodimer of subunits E and G (EG), the N-terminal domain of subunit a (aNT), and V1 lacking subunit H (V1ΔH). Using isothermal titration calorimetry (ITC) and biolayer interferometry (BLI), we show that HNT binds EG with moderate affinity, that full-length H binds aNT weakly, and that both H and HNT bind V1ΔH with high affinity. We also found that only one molecule of HNT binds V1ΔH with high affinity, suggesting conformational asymmetry of the three EG heterodimers in V1ΔH. Moreover, MgATP hydrolysis–driven conformational changes in V1 destabilized the interaction of H or HNT with V1ΔH, suggesting an interplay between MgADP inhibition and subunit H. Our observation that H binding is affected by MgATP hydrolysis in V1 points to H's role in the mechanism of reversible disassembly.

Typically, the V-ATPase is localized on the endomembrane system where the enzyme acidifies intracellular compartments, a process essential for pH homeostasis, protein trafficking, endocytosis, hormone secretion, mTOR signaling, and lysosomal degradation (1). The V-ATPase is also present on the plasma membrane of certain specialized cells such as osteoclasts, renal cells, the vas deferens, and the epididymis where the enzyme acidifies the extracellular environment. V-ATPase's proton pumping activity has been linked to numerous human diseases including osteoporosis and -petrosis (2, 3), renal tubular acidosis (4), male infertility (5), neurodegeneration (6), diabetes (7), viral infection (8), and cancer (9), making the enzyme a valuable drug target (10,11).
The V-ATPase shares a similar architecture and catalytic mechanism with the F-ATP synthase such that it consists of a water-soluble ATP-hydrolyzing machine (V 1 ) and a membrane-integral proton channel (V o ), which are structurally and functionally coupled via a central stalk and multiple peripheral stalks (12)(13)(14). The subunit composition of the V-ATPase from yeast is A 3 B 3 CDE 3 FG 3 H for the cytosolic V 1 (15) and ac 8 cЈcЉdef for the membrane-integral V o (16,17). The subunit architecture of the V-ATPase has been studied by electron microscopy (EM) and several low-to intermediate-resolution reconstructions of bovine, yeast, and insect V-ATPase are available, which, together with X-ray crystal structures of individual subunits and subcomplexes of yeast V-ATPase and bacterial homologs, have provided a detailed model of the subunit architecture of the eukaryotic V-ATPase complex (17)(18)(19)(20)(21)(22) (see Fig. 1A). V-ATPase is a rotary-motor enzyme (14). ATP hydrolysis in the catalytic A 3 B 3 hexamer is coupled to rotation of the proteolipid ring (c 8 cЈcЉ) via a central rotor made of subunits D, F, and d with concurrent proton translocation at the interface of the proteolipid ring and the C-terminal domain of subunit a (a CT ). During rotary catalysis, the motor is stabilized by a peripheral stator complex consisting of three peripheral stalks constituted by heterodimers of subunits E and G (hereafter referred to as EG1-3) that connect the A 3 B 3 hexamer to the N-terminal domain of the membrane-bound a subunit (a NT ) via the singlecopy "collar" subunits H and C (19,21) (see Fig. 1A). Three intermediates (referred to as rotational states [1][2][3], in which the central rotor is spaced 120°relative to the catalytic hexamer and subunit a, have been visualized in the yeast enzyme by cryo-EM (21).
Unlike the related F-ATP synthase, eukaryotic V-ATPase is regulated by a unique mechanism referred to as reversible disassembly wherein, upon receiving cellular signals, V 1 dissociates from V o , and the activity of both sectors is silenced (22)(23)(24)(25)(26) (see Fig. 1B). Reversible disassembly was first observed in yeast (27) and insects (28), but the process has recently also been observed in higher animals including human (29 -32). Studies in yeast have shown that, during enzyme disassembly, subunit C is released into the cytosol by an unknown mechanism and reincorporated during reassembly (27). Because of regulation of the V-ATPase by reversible disassembly, the peripheral stator subunit interactions at the V 1 -V o interface draw particular attention as they must be sufficiently strong to withstand the torque generated during ATP hydrolysis, but at the same time they must be vulnerable enough to allow breaking on a timescale for reversible disassembly to occur efficiently.
We previously characterized the interaction of the collar subunit C with EG and a NT and found that although the head domain of C (C head ) binds EG with high affinity (C head -EG3; see Fig. 1A), its foot domain (C foot ) and EG both interact weakly with a NT , resulting in a high-avidity ternary interface (EG2a NT -C foot ) in holo-V-ATPase (33,34). Another collar subunit at the V 1 -V o interface is subunit H, and although deletion of C prevents stable assembly of V 1 and V o (35), deletion of H results in the formation of a V 1 V o (⌬H) complex that lacks MgATPase and proton-pumping activities (36,37). Moreover, although C is released into the cytosol upon disassembly of V 1 V o , H remains stably associated with V 1 (Fig. 1B). The crystal structure of H revealed that it consists of a larger N-terminal (H NT ) and smaller C-terminal domain (H CT ) connected by a short linker (38). Previous functional characterization of H NT and H CT suggested that, although H NT is required for MgATPase activity in V 1 V o , H CT has a dual function in that it is required for both coupling of V 1 's ATPase activity to proton pumping in V 1 V o (37) and inhibition of MgATPase activity in membranedetached V 1 (39). The dual role and functional separation of H NT and H CT along with differences in regulatory function compared with C are not well understood and prompted the analyses of the interactions of H, H NT , and H CT with its binding partners in V 1 and V o . We therefore used recombinant H, H NT , and H CT for quantification of their interactions with purified EG, a NT , and V 1 lacking subunit H (V 1 ⌬H) using isothermal titration calorimetry (ITC) and biolayer interferometry (BLI). We found that H NT binds no more than one of the three EGs on V 1 ⌬H and that the affinity of this interaction is ϳ40-fold higher than that between H NT and isolated EG, suggesting that H NT prefers a particular conformation of EG on V 1 . We further found that full-length H interacts with V 1 ⌬H with an ϳ70-fold higher affinity than H NT , indicative of a significant contribution of H CT to the binding energy. Furthermore, we show that MgATP hydrolysis-driven conformational changes in the catalytic A/B pairs, the central rotor (DF), and the peripheral stalks (EG) destabilize the V 1 -H interaction until inhibitory MgADP is trapped in a catalytic site. The findings are discussed in context of the mechanism of V-ATPase regulation by reversible disassembly.

Expression, purification, and biophysical characterization of H, H NT , H CT , and a NT (1-372)
To understand the role of the V 1 -V o interface in the mechanism of reversible disassembly, our laboratory has previously characterized the interactions among C head , C foot , EG, and a NT (33,34). Interactions involving subunit H, however, are yet to be quantified. Pulldown and yeast two-hybrid assays have shown that H is able to bind the N-terminal region of subunit E (40). In addition, EM and crystal structures of V 1 V o and V 1 , respectively, show H NT bound to one of the three EG peripheral stalks (EG1; see Fig. 2, A and B), whereas H CT is seen to either rest on the coiled-coil middle domain of a NT (in V 1 V o ; Fig. 2A) or at the bottom of the A 3 B 3 hexamer (in autoinhibited V 1 ) (Fig. 2B) (21,22). To analyze the interactions of H within the enzyme in more detail, we expressed H, H NT , and H CT separately and quantified their interactions with recombinant EG, a NT , and V 1 ⌬H purified from yeast.
Full-length H, H NT (residues 1-354), H CT (residues 352-478), and a NT (residues 1-372) were expressed in Escherichia   (21), H NT (dark green) is bound to EG1 (blue/orange), and H CT (light green) is in contact with a NT (purple). B, in membrane-detached and autoinhibited V 1 (Protein Data Bank code 5D80) (22), the contact between H NT and EG1 is preserved, but H CT undergoes a ϳ150°rotation to bind the bottom of the A 3 B 3 hexamer and the rotor subunit D. The large conformational change in H CT is depicted by the positions of the C-terminal ␣-helix (magenta) and the inhibitory loop (red) in holo-V 1 V o versus autoinhibited V 1 -ATPase.

Subunit H interactions at the V 1 -V o interface
coli as N-terminal fusions with maltose-binding protein (MBP). After amylose affinity capture, fusions were cleaved, and MBP was removed by ion exchange and size exclusion chromatography, resulting in purified subunits and subunit domains (Fig.  3A). All proteins eluted near their expected molecular masses on size exclusion chromatography except a NT (1-372), which exists in a dimer-monomer equilibrium as described previously (26,34) (Fig. 3, B and C). Consistent with available structural information, circular dichroism (CD) spectroscopy revealed a high degree of ␣-helical secondary structure, suggesting proper folding of the recombinant polypeptides (Fig. 3D).

Interaction of H NT with EG
We previously established that binding of C (or C head ) to isolated EG occurs with high affinity and that the interaction greatly stabilizes EG (33). To further characterize the interactions at the V 1 -V o interface, we set out to determine the affinity of the H NT -EG interaction using ITC. Titration of H NT into EG was exothermic, and the binding curve was fit to a single-binding-site model, revealing a K d of the interaction of 187 nM. The binding enthalpy (⌬H) and entropy (⌬S) were Ϫ8 kcal/mol and 2.5 cal/(mol⅐K), respectively, with a concomitant free energy change (⌬G) of ϳϪ36 kJ/mol (Fig. 4A). Consistent with the ITC titration, size exclusion chromatography of a mixture of EG and an excess of H NT resulted in the formation of a ternary H NT -EG complex (Fig. 4, B and C), and taken together, the data show that H NT forms a stable complex with the EG heterodimer. Previously, we found that the EG's N-terminal right-handed coiled coil is thermally labile with a T m of ϳ25°C (Fig. 4D, blue trace) (29) and that the T m of EG is increased by about 10°C upon complex formation with C (head) (33). To test whether H NT binding has a similar stabilizing effect on EG, thermal unfolding of the individual proteins and EGH NT complex was monitored by recording the CD signal at 222 nm as a function of temperature. The data show that isolated H NT unfolds with an apparent T m of ϳ63°C (Fig. 4D, green trace). The thermal unfolding curve of the EGH NT complex showed two transitions, one at 25°C and one at 64°C, suggesting that the stability of EG is not increased upon H NT binding (Fig. 4D, black trace). Moreover, as also shown previously (33), isolated EG heterodimer dissociates during native agarose gel electrophoresis, whereas in presence of C head the three proteins migrate as a heterotrimeric complex in the electric field. However, consistent with the thermal unfolding data, a complex of EGH NT did not comigrate on the native gel but ran as three separate species (Fig. 4E). Therefore, although both C head and H NT form a  These data were fit using a one-site binding model resulting in a K d of 187 nM with ⌬H and ⌬S values of Ϫ8 kcal/mol and 2.5 cal/(mol⅐K), respectively, resulting in a ⌬G of the interaction of ϳϪ36 kJ/mol. A representative of three separate titrations is shown. B, the ITC cell content was subjected to gel filtration on a 1.6 ϫ 50-cm Superdex 200 column (black trace). The individual gel filtration elution profiles of H NT and EG using the same column are shown in green and blue, respectively. C, SDS-PAGE of gel filtration fractions from the ITC cell content. EGH NT elutes at higher molecular mass with excess H NT in a well separated peak. The shift of the EGH NT peak (compared with EG or H NT ) toward higher molecular weight indicates complex formation. The inset in B shows the EGH NT peak fraction (fraction 55) from the SDS-PAGE gel of the ITC cell content shown in C. D, thermal unfolding of H NT , EG, and EGH NT monitored by recording the CD signal at 222 nm with increasing temperature. H NT shows highly cooperative unfolding with an apparent T m of ϳ63°C (green). EG has an apparent T m of ϳ25°C (data taken from Ref. 33). The EGH NT complex shows two unfolding transition with T m values similar to the those observed for the individual proteins, suggesting that H NT binding to EG does not stabilize the EG heterodimer. E, native agarose gel electrophoresis of EG heterodimer, H NT , and EGH NT . 30 g of purified EG, 30 g of purified H NT , and a mixture of equimolar amounts of H NT and EG (total 60 g) were loaded. Unlike binding of C head to EG (33), binding of H NT does not appear to stabilize EG under these conditions. F, isothermal titration calorimetry of the interaction between a NT and H. H was titrated into a NT (top panel, lower trace) or buffer (top panel, top trace), and the heat associated with the interaction was measured at 10°C in 20 mM Tris-HCl, 0.5 mM EDTA, 1 mM TCEP, pH 8. Subtracting the heat of dilution of titration of H into buffer revealed an endothermic binding reaction. Fitting the data (bottom panel) to a one-site binding model with a fixed n ϭ 1 allowed an estimate of the K d of ϳ130 M. A representative titration from two repeats is shown. mAU, milliabsorbance units.

Subunit H interactions at the V 1 -V o interface
stable complex with EG, the nature of the two interactions are strikingly different.

Interaction of H with a NT
Prior work from our laboratory has shown that the EG2a NT -C foot junction at the V 1 -V o interface (Fig. 1A) is formed by multiple low-affinity interactions, and we reasoned that the sum of these interactions provides a high-avidity binding site between V 1 and V o that could be targeted for regulated disassembly (34). Another interaction that is seen in EM reconstructions of the intact V-ATPase, and that must be broken and reformed during reversible disassembly, is between H and a NT (Figs. 1A and 2A). To estimate the affinity between H and a NT , we performed ITC experiments by titrating H into a NT (1-372) (Fig. 4F). Subtracting the heat generated from diluting H into buffer from the heat generated from titrating H into a NT (1-372) revealed a weak endothermic binding reaction between the two proteins. Fitting the data to a one-site binding model revealed a K d of 135 M and ⌬H, ⌬S, and ⌬G values of 4 kcal/ mol, 36.1 cal/(mol⅐K), and Ϫ26 kJ/mol, respectively. Consistent with our ITC data, a mixture of H and a NT eluted at the same volumes as the individual proteins on size exclusion chromatography (Fig. S1). The low-affinity interaction between H and a NT supports our existing model that V 1 binds V o via several low-affinity interactions that, taken together, result in a highavidity interface in assembled V 1 V o .

Interaction of V 1 ⌬H with H, H NT , and H CT characterized by BLI
Previous experiments showed that H remains bound to V 1 even at the low concentrations used in enzyme essays (e.g. ϳ15 nM) (22,25,39) and under the conditions of electrospray ionization used for native MS (15). Although the data so far have shown that the affinity of H NT for EG as measured using ITC is moderately high, the observed K d of ϳ0.2 M (Fig. 4A) could not explain the above observations, which means that the interaction of H with V 1 has to be much stronger (39). We therefore wished to determine the affinity of full-length H as well as H NT and H CT for V 1 ⌬H. The interaction between V 1 ⌬H and MBPtagged H, H NT , and H CT was quantified using BLI. MBP-tagged proteins were immobilized on anti-mouse Fc capture (AMC) biosensors using an anti-MBP antibody, and the rate of association and dissociation of V 1 ⌬H was measured hereafter. The slow dissociation of MBP-tagged H, H NT , and H CT from the anti-MBP antibodies was subtracted from the V 1 ⌬H dissociation rates for analysis of the kinetic data. BLI experiments for measuring association and dissociation kinetics between V 1 ⌬H and MBP-H/H NT were conducted at five different V 1 ⌬H concentrations, and the resulting association and dissociation curves were fit to a global single-site binding model (Fig. 5, A  and B). Analysis of the data for MBP-H and MBP-H NT revealed K d values of ϳ65 pM (Fig. 5A) and ϳ4.5 nM (Fig. 5B), respectively. We also tested the binding of V 1 ⌬H to MBP-H CT , but we were not able to determine a K d as there was no measurable association at low V 1 ⌬H concentrations (Ͻ100 nM), and higher V 1 ⌬H concentrations (e.g. 1 M) resulted in nonspecific binding to the BLI sensors (data not shown). Overall, the interaction of H NT with EG as part of V 1 ⌬H was ϳ40-fold tighter when compared with the interaction between H NT and isolated EG (as measured using ITC; Fig. 4A), suggesting that the conformation of EG on V 1 ⌬H is more favorable for H NT binding than the conformation(s) of isolated EG. In addition, although we could not detect an interaction between H CT and V 1 ⌬H under the conditions of BLI, a ϳ70-fold higher affinity of V 1 ⌬H for H as compared with H NT suggests a significant contribution of H CT to the V 1 -H interaction. From our ITC and BLI experiments, we infer that the binding interaction between H NT and EG allows H CT to switch conformations so that it can either bind a NT (in V 1 V o ) or subunits B and D (in V 1 ) to efficiently carry out its dual role in reversible disassembly.

V 1 ⌬H binds no more than one H NT
Because V 1 ⌬H contains three EG heterodimers, we wished to determine whether all three or only one of the EGs can bind H NT . Purified V 1 ⌬H was mixed with a 5-fold molar excess of H NT followed by size exclusion chromatography. Under these conditions, some H NT coeluted with V 1 ⌬H with the excess H NT eluting from the column as a separate, lower molecular weight peak (Fig. 5, C and D). The V 1 (⌬H)H NT complex was concentrated, and approximately equal amounts of V 1 ⌬H and V 1 (⌬H)H NT were resolved using SDS-PAGE. The staining of H NT in the purified V 1 (⌬H)H NT complex was similar to that of single-copy subunits in the V 1 complex (for example subunit D), indicating that no more than one copy of H NT bound to V 1 ⌬H (Fig. 5E). Therefore, although there are three EGs in V 1 ⌬H, only one of these is in a conformation that is able to bind H NT with high affinity, highlighting the conformational asymmetry of the peripheral stalks.

The interaction of H with V 1 ⌬H is destabilized upon MgATP hydrolysis
The preference of H NT for one of three EGs suggested that the asymmetry of the peripheral stalks originates in the catalytic core (A 3 B 3 DF) of V 1 . Upon MgATP hydrolysis, however, the conformational changes of the catalytic sites from open to loose to tight drive counterclockwise rotation of the central rotor along with cyclic structural changes in the peripheral stalks from EG1 to EG3 to EG2 (41). In addition, based on the structure and nucleotide occupancy of the autoinhibited V 1 sector, our laboratory suggested that H CT inhibits V 1 -ATPase activity by preferentially binding to an open catalytic site, consequently maintaining inhibitory MgADP in the adjacent closed catalytic site (22). Taken together, H NT 's preference for EG1, as well as H CT 's role in MgADP inhibition, indicated a potential interplay between the nucleotide occupancy of the catalytic sites and the interaction of V 1 ⌬H with H. To probe the effect of nucleotides on the interaction of H with V 1 ⌬H, we again used BLI. V 1 ⌬H was bound to immobilized MBP-H, and the sensor was then dipped in wells containing buffer or buffer with 1 mM MgATP, MgADP ϩ P i , or MgAMPPNP. Interestingly, in the presence of MgATP, a biphasic dissociation curve was observed with an initial dissociation rate that was ϳ6 times faster than the rate in buffer alone (Fig. 6, A and B). However, only ϳ25% of the bound V 1 ⌬H dissociated with a fast rate with the remaining 75% coming off the sensor at a rate similar to the dissociation rate in buffer (Fig. 6, A and B). In contrast, a relatively slower rate of

Subunit H interactions at the V 1 -V o interface
dissociation was observed in the presence of MgADP ϩ P i and MgAMPPNP.
The destabilization of the V 1 -H interaction upon MgATP hydrolysis came as a surprise to us as the H subunit is known to inhibit V 1 -ATPase activity (22,25,39). To confirm that the fast dissociation of V 1 ⌬H from MBP-H in wells containing MgATP was specifically due to MgATP binding to V 1 's catalytic sites, we conducted a similar BLI experiment using V 1 ⌬H treated with N-ethylmaleimide (NEM). It is known that NEM modification of a catalytic-site cysteine residue prevents binding of nucleotides (42). NEM-treated and untreated V 1 ⌬H were bound to MBP-H immobilized on sensors and then dipped in wells containing MgATP, MgADP ϩ P i , or buffer (Fig. 6, C and D). We found that NEM-treated V 1 ⌬H no longer showed a fast dissociation rate when dipped in MgATP-containing wells, suggesting that MgATP binding to the catalytic sites caused destabilization of the V 1 -H interaction. However, if the above mentioned fast dissociation rate was a result of only MgATP binding, but not hydrolysis, we should have observed fast dis-sociation in the presence of MgAMPPNP, the nonhydrolyzable ATP analog. MgAMPPNP, however, had no effect on the V 1 ⌬H dissociation rate (Fig. 6A, green trace). Taken together, the BLI experiments with NEM-modified V 1 ⌬H and in the presence of MgAMPPNP suggest that it is MgATP binding and hydrolysis that destabilize the V 1 -H interaction. Because both H NT and H CT contribute to the interaction of H with V 1 ⌬H, we also measured the off-rate of V 1 ⌬H from immobilized MBP-H NT and found that the H NT -V 1 ⌬H interaction was also destabilized upon MgATP hydrolysis as seen for H and V 1 ⌬H (Fig. 6, E  and F).
To verify that V 1 ⌬H bound to H was capable of transient turnover, we purified V 1 ⌬H, incubated it with an excess of H, and resolved the mixture using size exclusion chromatography (Fig. S2). We found that V 1 ⌬H reconstituted with H (V 1 (⌬H)H) showed ϳ4.9 Ϯ 0.55 units/mg of MgATPase activity, which was ϳ30% of the activity of V 1 ⌬H (15.75 Ϯ 1.7 units/mg) (Fig. 6G and Ref. 22). Considering the high-affinity interaction between V 1 ⌬H and H with a K d of ϳ65 pM, we expected stoichiometric The data were globally fit (traces in red) to reveal a K d of ϳ65 pM. B, a similar experiment was conducted to analyze the interaction between H NT and V 1 ⌬H. H NT was dipped in 10, 20, 40, 80, and 160 nM V 1 ⌬H followed by buffer to generate association and dissociation curves, respectively (blue). The data were globally fit (red traces) to reveal a K d of ϳ4.5 nM. C, V 1 ⌬H was incubated with a 5-fold excess of H NT , and the mixture was resolved on a 1.6 ϫ 50-cm Superdex 200 column. D, SDS-PAGE of gel filtration fractions. The higher molecular weight peak showed V 1 ⌬H in complex with H NT with the lower molecular weight peak corresponding to excess H NT . E, approximately equimolar amounts of V 1 ⌬H and V 1 ⌬H in complex with H NT were resolved by SDS-PAGE. The staining intensity of the H NT compared with the single-copy subunit D band in the V 1 (⌬H)H NT complex suggests that V 1 ⌬H binds no more than one copy of H NT with high affinity. mAU, milliabsorbance units.

Subunit H interactions at the V 1 -V o interface
amounts of H in the V 1 (⌬H)H reconstituted complex. However, to exclude the possibility that the observed MgATPase activity in the V 1 (⌬H)H complex was due to substoichiometric binding of H, we performed a pulldown experiment in which a 10-fold molar excess of MBP-H bound to amylose resin was used to capture V 1 ⌬H (Fig. S3). Although some MBP-H and V 1 ⌬H appeared in the supernatant and washes of the amylose resin, most of the MBP-H eluted in 10 mM maltose along with stoichiometrically bound V 1 ⌬H. Elution fraction E1 (Fig. S3A) exhibited significant MgATPase activity (Fig. S3B), indicating that a stoichiometric complex of V 1 ⌬H with MBP-H was capable of hydrolyzing MgATP.
A consistent feature of the dissociation curve of V 1 ⌬H from H in MgATP was its biphasic nature (Fig. 6A, orange trace), indicating that MgATP hydrolysis-dependent destabilization was transient. We found that, by using different concentrations of MgATP during dissociation, we were able to regulate the fast phase of the dissociation rate and consequently the duration of destabilization (Fig. 7). Not only does this experiment confirm MgATP hydrolysis as being the cause of V 1 -H destabilization, Figure 6. The interaction between subunit H and V 1 ⌬H is destabilized upon MgATP hydrolysis. A, sensorgrams of V 1 ⌬H dissociation from immobilized H subunit in the absence (blue) and presence of 1 mM nucleotides (MgADP ϩ P i , purple; MgAMPPNP, green; MgATP, orange). A representative from at least two independent experiments is shown. B, off-rates determined from fitting the sensorgrams from A to dual-exponential equations. C, sensorgrams of association and dissociation of NEM-inhibited V 1 ⌬H to and from immobilized H subunit in the absence or presence of 1 mM nucleotides (association of NEM-modified V 1 ⌬H in buffer/dissociation in MgADP ϩ P i , purple; association of NEM-modified V 1 ⌬H in buffer/dissociation in MgATP, red; both association and dissociation of NEM-modified V 1 ⌬H in buffer, blue; association of unmodified V 1 ⌬H in buffer/dissociation in MgATP, orange). A representative from at least two independent experiments is shown. D, observed on-rates (k obs ) and off-rates obtained from fitting the sensorgrams in C to single-and dual-exponential equations, respectively. E, sensorgrams of V 1 ⌬H dissociation from immobilized H NT in the absence (blue) and presence of 1 mM nucleotides (MgADP ϩ P i , purple; MgAMPPNP, green; MgATP, orange). A representative from at least two independent experiments is shown. F, off-rates determined from fitting the sensorgrams from E to dual-exponential equations. G, average specific activities of V 1 ⌬H and V 1 (⌬H)H plotted ϮS.E. (error bars) from two independent purifications. Inset, raw data for activity measurement using an ATP-regenerating assay (22). V 1 -ATPase activity is determined by monitoring a decrease in A 340 as a function of time. Traces for WT V 1 , V 1 ⌬H, and V 1 (⌬H)H are shown in dark green, blue, and light green, respectively. N/A, not applicable.

Subunit H interactions at the V 1 -V o interface
it also explains why the destabilization is transient. Using P i release-based ATPase assays, it has been shown that MgATPase activity of V 1 ⌬H subsides rapidly with time (25). This rapid decrease in activity, which is also observed in an ATP-regenerating assay (Fig. S4), has been attributed to the trapping of MgADP in a catalytic site, a phenomenon termed MgADP inhibition. We have observed that MgADP inhibition of V 1 ⌬H is more efficient under high Mg 2ϩ (and by extension high MgATP) concentrations and that decreasing the initial concentration of MgATP results in delayed MgADP inhibition (Fig.  S4). In the BLI experiment shown in Fig. 7, decreasing the concentration of MgATP to 100 M (Fig. 7, red trace) and consequently decreasing the rate of MgATP hydrolysis led to a delay in the culmination of the fast dissociation phase. Taken together, these data suggest that MgATP hydrolysis causes transient destabilization of the V 1 -H interaction until MgADP inhibition sets in.

Discussion
V-ATPase is regulated by reversible disassembly, a process that involves the breakage and reformation of several proteinprotein interactions at the interface of V 1 and V o . These interactions are mediated by the central rotor of V 1 (DF) binding to V o 's subunit d and the three peripheral stalks (EG1-3) that link the collar subunits H and C to a NT (see Fig. 2A). Both H and C are two domain proteins, and we previously found that C head binds EG with high affinity, whereas C foot and EG bind a NT weakly. Here, we analyzed binding of H and H NT to isolated EG, a NT , and V 1 ⌬H purified from yeast. We found that the majority of the binding energy between H and V 1 is contributed by the interaction between H NT and EG and that binding of H NT to EG is much stronger when EG is part of V 1 compared with isolated EG. However, only one copy of H NT binds V 1 ⌬H with high affinity, indicating that the three EGs on V 1 are in different conformations and that only one of these conformations (EG1) is competent for H binding. The three different conformations of the EGs are evident from the crystal structure of autoinhibited V 1 (22) as well as the cryo-EM structures of V 1 V o (20,21). The observation that H NT binds only EG1 as part of V 1 with high affinity suggests that H NT 's preference is a result, and not the cause, of the conformational asymmetry of the peripheral stalks, which most likely originates in the catalytic core of V 1 (A 3 B 3 DF). Although we were not able to detect an interaction between isolated H CT and V 1 ⌬H, the observation that intact H binds V 1 ⌬H with significantly higher affinity compared with H NT suggests that the contact between H CT and A 3 B 3 DF seen in the V 1 crystal structure (22) contributes to the avidity of the V 1 -H interaction. In addition, much like the C foot -a NT -EG low-affinity (but high-avidity) ternary junction, we found that H and a NT interact weakly. Taken together, the data support and extend our earlier findings that V 1 and V o are held together by multiple weak interactions that allow rapid breaking and reforming in response to cellular needs.

MgATP hydrolysis-dependent destabilization of the V 1 -H interaction
Studies in yeast have shown that membrane-detached V 1 has no measurable MgATPase activity, a property of WT V 1 that has been attributed to the presence of the inhibitory H subunit (25). Therefore, our BLI data showing that the V 1 -H interaction is destabilized in the presence of MgATP came as a surprise as we did not expect V 1 (⌬H)H to be catalytically active. In contrast, previous biochemical studies had shown that V 1 ⌬H retained ϳ20% MgATPase activity upon addition of an excess of recombinant H, an observation that, at the time, was attributed to substoichiometric binding of H (39). However, using pulldown assays, we here show that a stoichiometric complex of V 1 ⌬H with H has indeed transient MgATPase activity, indicating that WT V 1 isolated from yeast is not equivalent to reconstituted V 1 (⌬H)H. One striking difference between V 1 and V 1 (⌬H)H is that V 1 contains ϳ1.3 mol/mol of tightly bound ADP, whereas V 1 ⌬H, and by extension V 1 (⌬H)H, has only ϳ0.4 mol/mol of ADP (22). This suggests that V 1 's ATPase activity is inhibited by tightly bound ADP and that the lack of ADP in V 1 (⌬H)H allows transient MgATP hydrolysis with the associated conformational changes leading to destabilization of the V 1 -H interaction on the BLI sensor.
MgADP inhibition is a conserved feature of the catalytic headpiece of rotary ATPases wherein, under ATP-regenerating conditions, the rate of MgATP hydrolysis decreases due to retention of tightly bound MgADP at a closed catalytic site. The MgADP-inhibited state is a conformation off-pathway from the catalytic cycle and associated with a structural change in the catalytic site (43). MgADP inhibition has been observed in both the F 1 -ATPase (e.g. F 1 from bovine heart (44)) and Bacillus PS3 (45)) and the cytosolic A 1 /V 1 sector from Thermus thermophilus (46). Parra et al. (25) reported a decrease in MgATPase

Subunit H interactions at the V 1 -V o interface
activity of purified yeast V 1 ⌬H using P i release assays, and we have observed a similar decrease in MgATPase activity of V 1 ⌬H using an ATP-regenerating assay system (22) (Fig. S4).

Interplay of MgADP inhibition with the V 1 -H interaction
The structure of autoinhibited V 1 revealed an inhibitory loop in H CT (amino acids 408 -414) that mediates important contacts with V 1 . First, H CT binds to the C-terminal domain of subunit B, thereby stabilizing the corresponding catalytic A-B interface in its open conformation. Second, it interacts with two ␣-helical turns in the central rotor subunit D (residues 38 -45) (22) (Fig. 2B). At any given rotational state of V 1 , only one of the catalytic sites is in the open state with the two ␣-helical turns from the central rotor facing the open site (22,47). It is therefore evident that H CT preferentially binds to the open catalytic site with the central rotor in a particular conformation. Our data suggest that the peripheral stalks exhibit a conformational asymmetry, which most likely originates from the conformations of the catalytic core of the enzyme (A 3 B 3 DF). H NT preferentially interacts with EG1, the peripheral stalk associated with the open catalytic site. Under the conditions of our BLI experiments, when V 1 ⌬H is bound to subunit H on the sensor, both H NT and H CT are associated with their binding partners at the open catalytic site, resulting in overall tight binding of H as evident from the observed slow dissociation of V 1 ⌬H from the BLI sensor (Fig. 8A). When sensors containing V 1 ⌬H bound to MBP-H are dipped into MgATP, the nucleotide binds to the open catalytic site and is hydrolyzed. Subsequent MgATP hydrolysis results in central stalk rotation, which destabilizes the interaction with H CT . The conformational changes are also propagated to the peripheral stalks, resulting in destabilization, and ultimately breaking, of the interaction between H NT and EG1 (Fig. 8B). However, MgATP hydrolysis on V 1 (⌬H)H is transient and stops once inhibitory MgADP gets trapped in a tight catalytic site. All the V 1 (⌬H)H complexes that withstood transient MgATP hydrolysis are now bound with high affinity because the binding site for H CT is restored once the MgADPinhibited conformation is obtained (Fig. 8C). Therefore, we conclude that the lack of inhibitory MgADP in V 1 (⌬H)H allows transient MgATP hydrolysis and destabilization of the V 1 -H interaction and that high-affinity binding of H is restored once inhibitory MgADP is trapped in a catalytic site.
Considering the here observed MgATP hydrolysis-dependent destabilization of the interaction between H (and H NT ) with V 1 raises the question: how is this interaction maintained in V 1 V o during rotational catalysis? Between the three rotational states of the enzyme observed by cryo-EM (21), minor conformational differences are observed for the peripheral stalk bound to H (EG1 in V 1 V o ). In V 1 V o , besides providing binding sites for both H NT and H CT , a NT also interacts with and probably stabilizes the N termini of EG1 (Figs. 1A and 2A). The N-terminal region of the peripheral stalks have been described as unstable and flexible based on experiments conducted with isolated EG (33) and EG as part of A 1 /V 1 (48) and as seen in the V 1 crystal structure (22). With EG1's N termini unsupported, as in membrane-detached V 1 , expected conformational changes associated with rotary catalysis would be larger than those observed in V 1 V o . Hence, although multiple interactions with a NT maintain the conformation of EG1 in V 1 V o , the lack of these interactions in V 1 enable rotary catalysis-driven conformational changes in EG1 with concomitant destabilization of the H NT -EG1 interaction.

Implications for the mechanism of reversible disassembly
Experiments conducted with yeast spheroplasts, isolated vacuoles, and purified V 1 V o have established that efficient disassembly of V-ATPase requires a catalytically active complex (49 -51). From a structural comparison with the three rotational states of V 1 V o , we previously noted that upon disassembly of the holoenzyme autoinhibited V 1 and V o end up in dif-

Subunit H interactions at the V 1 -V o interface
ferent rotational states: V 1 in state 2 and V o in state 3 (22,52,53). This suggests that the MgATPase activity that is necessary for efficient disassembly serves to generate the rotational state mismatch associated with enzyme dissociation, a mismatch that likely functions to prevent rebinding of V 1 to V o under cellular conditions that favor the disassembled state. It is well established that enzyme dissociation is accompanied by a release of subunit C into the cytosol, and it is possible that the energy from ATP hydrolysis also serves to break the high-affinity EGC head interaction (33). Live cell imaging has captured V 1 on vacuolar membranes while C is released into the cytosol upon glucose removal, suggesting that C release may be one of the initial steps of disassembly. Our data suggest that, due to catalysis-driven conformational changes in EG, membrane-detached V 1 sector is incapable of binding H in a stable conformation while MgATP is being hydrolyzed at the catalytic sites. Therefore, it is possible that V 1 detaches from V o on vacuolar membranes after its MgATPase activity is completely silenced.  (22). Hence, inhibitory MgADP and subunit H synergize by stabilizing each other to ensure that free V 1 remains in the autoinhibited state to prevent wasteful ATP hydrolysis when V-ATPase's proton-pumping activity is down-regulated.
The autoinhibited state of V 1 (with MgADP in a closed catalytic site stabilized by H CT bound to an open catalytic site) is likely a low-energy state of V 1 . For reassembly to occur, V 1 needs to be "reactivated" to allow H CT to switch from its binding site on V 1 to bind a NT in V 1 V o . Based on our observations in this study, we speculate that release of inhibitory MgADP and subsequent MgATP binding/hydrolysis induce structural changes in V 1 that detach H CT , making it available to bind a NT , in turn coupling V 1 to V o . What then causes the required release of inhibitory MgADP from cytosolic V 1 ? Although this mechanism is currently not understood, it is possible that one of the protein factors that have been shown to be required for efficient reassembly, such as the regulator of the ATPase of vacuolar and endosomal membranes (RAVE) complex (54) or aldolase (55), plays a role in the release of inhibitory ADP, thereby allowing H CT to assume its binding site on a NT and restore MgATP hydrolysis-driven proton pumping.

Materials and methods
Plasmids encoding subunit H and its C-terminal domain (H CT ; residues 352-478) N-terminally tagged with a Prescission protease-cleavable maltose-binding protein (MBP-H and MBP-H CT encoded by a pMalPPase vector derived from pMAL-c2E), a yeast strain deleted for subunits H and G (39), and a pRS315 vector containing FLAG-tagged subunit G (56) were kind gifts from Dr. Patricia Kane, SUNY Upstate Medical University.

Plasmid construction
The plasmid expressing the N-terminal domain of subunit H (H NT ; residues 1-354) was made using the above MBP-H  pMalPPase vector as a template for QuikChange mutagenesis  to delete the nucleotide sequence coding for amino acids 355-478 using the following primers: H NT1-354 F, GGA AAT CCT  AGA AAA CGA GTA CCA AGA ATT GAC CTA AAA GCT  TGG CAC TGG CCG TCG TTT TAC AAC GTC G; H NT1-354  R, GAC GGC CAG TGC CAA GCT TTT AGG TCA ATT CTT GGT ACT CGT TTT CTA GGA TTT CCT TG. The construction of the pMalPPase plasmid encoding N-terminally MBP-tagged a NT (1-372) has been described (26).

Expression and purification of recombinant V-ATPase subunits
V-ATPase subunit constructs H NT , EG, H CT , H, and a NT were expressed in E. coli strain Rosetta2, grown to midlog phase in rich broth (LB Miller plus 0.2% glucose) supplemented with ampicillin (100 g/ml) and chloramphenicol (34 g/ml). Protein expression was induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside (except for expression of EG where 1 mM isopropyl ␤-D-thiogalactopyranoside was used). Expression was induced at 30°C for 6 h for H NT , 20°C for 6 h for H, 20°C for ϳ16 h for a NT , 25°C for ϳ16 h for H CT , and 30°C for 6 h for EG. Cells were harvested by centrifugation, resuspended in amylose column buffer (CB; 20 mM Tris-HCl, 200 mM NaCl, 1 mM EDTA, pH 7.4), and stored at Ϫ20°C until use. For purification, cells were treated with DNase (67 g/ml), lysozyme (840 g/ml), and PMSF (1 mM) before lysis by sonication. The lysate was then cleared by centrifugation at 12,000 ϫ g for 30 min, the supernatant was diluted 1:4 with CB and applied to an amylose affinity column at a rate of 1 ml/min, and nonspecifically bound material was removed by washing with 12 column volumes of CB followed by 15 column volumes of CB without NaCl. Bound protein was eluted using 25 ml of 10 mM maltose in CB without salt. The MBP tag was cleaved using Prescission protease as described (34). H, H CT , and a NT were separated from MBP by anion exchange chromatography (Mono Q) using a linear gradient of 0 -300 mM NaCl in 25 mM Tris-HCl, 1 mM EDTA, pH 7, for elution. Residual MBP was removed by a small amylose column, and the concentrated protein was subjected to size exclusion chromatography using a Superdex S75 column (1.6 ϫ 50 cm) for H and H CT and a Superdex 200 column of the same size for a NT . H NT was separated from MBP using a gravity DEAE (anion exchange) column. At a pH of 6.5, MBP was immobilized on the DEAE column, and the H NT flowed through. The flowthrough was collected, concentrated, and subjected to size exclusion chromatography using a Superdex S75 column (1.6 ϫ 50 cm). EG heterodimer was purified as described (33).

Purification of V 1 ⌬H
V 1 -ATPase lacking subunit H was purified as described (22). Briefly, the yeast strain deleted for the genes encoding subunits H and G was transformed with a pRS315 plasmid encoding subunit G with an N-terminal FLAG tag (56). The cells were grown in synthetic defined medium lacking Leu to an OD of ϳ4.0 and harvested by centrifugation, and the cell pellets were resuspended in TBSE (25 mM Tris-HCl, pH 7.2, 150 mM NaCl, 0.5 mM EDTA) and stored at Ϫ80°C until use. Thawed cells

Subunit H interactions at the V 1 -V o interface
were supplemented with 5 mM ␤-mercaptoethanol, 1 mM PMSF, and 2 g/ml each of pepstatin and leupeptin before lysis by 15 passes through a microfluidizer with intermittent cooling on ice. The lysate was centrifuged at 4000 ϫ g for 25 min, and the resultant supernatant was centrifuged again at 13,000 ϫ g for 40 min. The cleared lysate was applied to a 5-ml FLAG column (Sigma) topped with Sephadex G50 and pre-equilibrated in TBSE. The column was washed with 10 column volumes of TBSE and eluted using 0.1 mg/ml FLAG peptide. V 1 ⌬H-containing fractions were pooled, concentrated, and resolved using a Superdex 200 1.6 ϫ 50-cm column attached to an ÄKTA FPLC (GE Healthcare). Fractions were analyzed by SDS-PAGE and concentrated to 10 mg/ml, and the activity of the complex was measured using a coupled enzyme assay as described below (22).

CD spectroscopy
CD spectra were recorded on an Aviv 202 spectropolarimeter using a 2-mm-path length cuvette. CD spectra were recorded between 250 and 195 nm in 25 mM sodium phosphate, pH 7, and protein stability was monitored by recording the CD signal at 222 nm as a function of temperature. For cysteinecontaining proteins, 0.3-1 mM TCEP was included in the buffer. Protein concentrations of H, H NT , H CT , and a NT were 2, 2.25, 9.2, and 2.36 M, respectively. The far-UV CD spectrum of 6.7 M H NT -EG complex was obtained with protein dissolved in 20 mM Tris-HCl, 1 mM TCEP, pH 8 buffer at 4°C followed by monitoring the temperature dependence of the CD signal at 222 nm.

Isothermal titration calorimetry
The thermodynamic parameters of the interaction between H NT and EG and between H CT and a NT were determined using a Microcal VP-ITC isothermal titration calorimeter. The interaction of H NT and EG was monitored by titrating a stock of 0.278 mM H NT into 0.0315 mM EG in 20 mM Tris-HCl, 0.5 mM EDTA, 1 mM TCEP, pH 8, at 10°C. A total of 30 injections with 5% saturation per injection was carried out. The average value of signal postsaturation (last eight titration points) was subtracted from the H NT into EG titration. Complex formation between a NT and H was analyzed by titrating 0.3 mM H into 0.017 mM a NT in 20 mM Tris-HCl, 0.5 mM EDTA, 1 mM TCEP, pH 8. A blank titration of H into buffer was subtracted from the a NT into H titration using the curve fit option in OriginLab. ITC data were analyzed using VP-ITC programs in OriginLab.

Biolayer interferometry
BLI was used to measure the association and dissociation kinetics of interaction between V 1 ⌬H and MBP-tagged H, H NT , and H CT . An Octed-RED system and AMC-coated sensors (FortéBio, AMC biosensors, catalogue number 18-5088) were used to monitor protein-protein binding and dissociation for determination of binding affinities. Anti-MBP antibody (New England Biolabs) at 1 g/ml was immobilized on the AMC biosensors. The anti-MBP antibody formed the bait for MBPtagged H, H NT , and H CT (used at 5 g/ml). Biosensors with immobilized H, H NT , or H CT were dipped in varying concentrations of V 1 ⌬H followed by buffer to measure association and dissociation rates. BLI buffer (20 mM Tris-HCl, pH 7.2, 150 mM NaCl, 1 mM EDTA, 0.5 mg/ml BSA) was used in all BLI experiments to reduce nonspecific binding to the biosensors except for experiments in the presence of nucleotides where the EDTA concentration was reduced to 0.5 mM. All steps were done at 22°C with each biosensor agitated in 0.2 ml of sample at 1000 rpm and a standard measurement rate of 5 s Ϫ1 . Control experiments were performed to check for any nonspecific binding interaction between the antibodies and the proteins used. Reference sensors were used in each experiment with immobilized MBP-H/H NT /H CT but no V 1 ⌬H. FortéBio data analysis software (version 6.4) was used for subtraction of reference sensors, Savitzky-Golay filtering, and global fitting of the kinetic rates of V 1 ⌬H binding with H, H NT , or H CT .

ATPase activity assays
MgATPase activity of V 1 , V 1 ⌬H, and V 1 (⌬H)H was measured using an ATP-regenerating assay as described (22). Briefly, 10 g of the V 1 mutant was added to an assay mixture containing 1 mM MgCl 2 , 5 mM ATP, 30 units/ml each of lactate dehydrogenase and pyruvate kinase, 0.5 mM NADH, 2 mM phosphoenolpyruvate, 50 mM HEPES, pH 7.5, at 37°C. The decrease of absorbance at 340 nm was measured in the kinetics mode on a Varian Cary Bio100 spectrophotometer.

Native gel electrophoresis
For native gel electrophoresis, purified V-ATPase subunits and subcomplexes were resolved using 2% agarose gels in 20 mM bis-Tris-acetic acid, pH 6, 1 mM TCEP. Gels were resolved for 1 h at 100 V, fixed in 25% isopropanol, 10% acetic acid for 30 minutes, rinsed in 95% ethanol, and dried on a slab dryer for 2 h at 80°C. The dried gel was stained with Coomassie G and destained in fixing solution.