Functional reconstitution of vacuolar H⁺-ATPase from V_o proton channel and mutant V₁-ATPase provides insight into the mechanism of reversible disassembly

Abstract

The vacuolar H⁺-ATPase (V-ATPase; V₁V_o-ATPase) is an ATP-dependent proton pump that acidifies subcellular compartments in all eukaryotic organisms. V-ATPase activity is regulated by reversible disassembly into autoinhibited V₁-ATPase and V_o proton channel subcomplexes, a process that is poorly understood on the molecular level. V-ATPase is a rotary motor, and recent structural analyses have revealed different rotary states for disassembled V₁ and V_o, a mismatch that is likely responsible for their inability to reconstitute into holo V-ATPase in vitro. Here, using the model organism Saccharomyces cerevisiae, we show that a key impediment for binding of V₁ to V_o is the conformation of the inhibitory C-terminal domain of subunit H (H_CT). Using biolayer interferometry and biochemical analyses of purified mutant V₁-ATPase and V_o proton channel reconstituted into vacuolar lipid-containing nanodiscs, we further demonstrate that disruption of H_CT's V₁-binding site facilitates assembly of a functionally coupled and stable V₁V_o-ATPase. Unlike WT, this mutant enzyme was resistant to MgATP hydrolysis-induced dissociation, further highlighting H_CT's role in the mechanism of V-ATPase regulation. Our findings provide key insight into the molecular events underlying regulation of V-ATPase activity by reversible disassembly.

Introduction

The vacuolar H⁺-ATPase (V-ATPase, V₁V_o-ATPase)³ is an ATP-dependent proton pump found on the endomembrane system of all eukaryotic organisms. This multisubunit nano-motor acidifies subcellular compartments and, in certain specialized tissues, the extracellular space. V-ATPase is essential for vital cellular processes such as pH homeostasis, protein sorting, autophagy, endocytosis, mTOR, and Notch signaling, as well as bone remodeling, urine acidification, hormone secretion, and neurotransmitter release (1). Although complete loss of V-ATPase function is embryonic lethal in mammals, aberrant activity has been associated with widespread human diseases including renal tubular acidosis (2), osteoporosis (3), neurodegeneration (4), diabetes (5), male infertility (6), and cancer (7), making V-ATPase a potential drug target (8, 9). However, because of its essential nature, global inhibition of V-ATPase is not a therapeutic option. Instead, there is a need for targeted modulation of the enzyme's activity, a goal that requires a detailed understanding of V-ATPase's catalytic and regulatory mechanisms.

V-ATPase is composed of two subcomplexes, a cytosolic ATPase called V₁, and a membrane integral proton channel termed V_o (Fig. 1A). In yeast, the subunit compositions for the V₁ and V_o are A₃B₃(C)DE₃FG₃H and ac₈c′c″def, respectively (10). The A and B subunits of V₁ are arranged in a hexamer (A₃B₃), with three catalytic sites at alternating AB interfaces. Located within the hexamer and extending from it in the direction of the membrane is subunit D that, together with F, provides the functional link between V₁ and V_o. The V_o is constituted by subunit a that can be divided into cytosolic N-terminal and membrane-integral C-terminal domains (a_NT and a_CT), the c, c′ and c″ subunits (“proteolipids”) that form a ring (c-ring), and subunit d that connects the c-ring with V₁ subunits D and F. V₁ and V_o are held together by three heterodimers of subunits E and G (peripheral stalks EG1–3) that link the catalytic hexamer and the single copy C and H subunits to the membrane integral a subunit by binding to a_NT.

Figure 1.

Purification and characterization of V_oND. A and B, schematic of V-ATPase regulation by reversible disassembly. C, purification strategy. Yeast vacuoles are isolated by flotation on a Ficoll gradient (panel i). Detergent-solubilized vacuolar proteins are mixed with biotinylated MSP (panel ii) and reconstituted into lipid nanodiscs followed by α-FLAG affinity capture of V_oND (panel iii). D, size-exclusion chromatography of V_oND. E, peak fractions were resolved using SDS-PAGE. F, negative stain EM of purified V_oND. Bar, 20 nm.

V-ATPase is a rotary motor enzyme and employs a catalytic mechanism that is shared with the F-, A-, and A/V-type ATPases (11). In V-ATPase, ATP hydrolysis–driven rotation of the DFdc-ring central rotor is coupled to proton translocation at the interface of a_CT and the c-ring. During catalysis, the three peripheral stalks, in conjunction with C, H, and a_NT, resist the rotary torque to keep the A₃B₃ hexamer static against a_CT for efficient energy coupling. However, unlike F-, A-, and A/V-type enzymes, eukaryotic V-ATPase is regulated by a unique mechanism referred to as “reversible disassembly,” wherein V₁ detaches from V_o in response to e.g. nutrient shortage (12–14) (Fig. 1B), with concomitant silencing of V₁'s MgATPase (15, 16), and V_o's proton transport activities (17, 18). First described in yeast and insect, reversible disassembly is now emerging as an important and conserved regulatory mechanism, having been observed in mammalian systems as well (19–21). However, although we have some understanding of the process at the cellular level (22), less is known about the mechanism of reversible disassembly at the molecular scale. Biochemical studies in yeast have shown that the single copy C and H subunits, which are unique to eukaryotic V-ATPase, play key roles in enzyme regulation. Both C and H are two-domain proteins, with C composed of “foot” (C_foot) and “head” (C_head), and H of N- (H_NT) and C-terminal (H_CT) domains. H has a dual role because it is required for both coupling MgATPase to proton-pumping activities and for stabilizing the autoinhibited state of membrane detached V₁; C functions to stabilize the V₁–V_o interface in the holo enzyme but dissociates from the complex upon regulated disassembly. Recent structural studies have revealed that in assembled V₁V_o, H_CT is bound to a_NT (23), an interaction required for energy coupling (24) (Fig. 1A). Upon disassembly of V₁ from V_o, H_CT undergoes a 150° rotation to wedge an inhibitory loop between the B subunit of an open catalytic site and subunits DF of the central rotor (16) (Fig. S1, A–D, red spheres). At the same time, a_NT moves from its peripheral position near C_foot and EG in V₁V_o (Fig. 1A) toward a more central position in free V_o to bind subunit d (18, 25–27) (Fig. S1, E–G). Moreover, cryoEM models of three distinct rotary states of holo V-ATPase (states 1–3) (23), along with the structures of autoinhibited V₁ (16) and V_o (25, 27), revealed that although V₁ is halted in state 2, V_o adopted state 3 (10) (Fig. S1). We hypothesized that the mismatch of rotational states observed in autoinhibited V₁ and V_o, together with the large conformational changes of H_CT and a_NT that accompany enzyme dissociation, explain why V₁ does not readily rebind free V_o under physiological conditions in vitro (10, 28), a safety mechanism that likely evolved to prevent spontaneous reassembly in vivo when the disassembled, inactive state is required.

We recently introduced biolayer interferometry (BLI) of purified V-ATPase in biotinylated and native lipid containing nanodiscs to analyze MgATP-dependent enzyme dissociation kinetics (29). Here, we have expanded on this approach to probe the interaction of V₁ and V_o. V_o sector was reconstituted into vacuolar lipid containing nanodiscs (V_oND) and immobilized on BLI sensors to screen V₁ mutants for their ability to bind V_o. In line with available literature (28), WT V₁ did not bind to V_oND, presumably because of the “state mismatch” observed in the structures of free V₁ and V_o (10). Previously, we generated a chimeric H subunit containing the yeast N-terminal and human C-terminal domains (H_chim) that does not inhibit free V₁ because human H_CT lacks an inhibitory loop that links an open catalytic site and the central rotor (16), and consequently, V₁ containing H_chim is not restrained in any particular rotational state. We show that replacement of endogenous H in yeast V₁ with H_chim (V₁H_chim) (16) permits binding to V_oND, and formation of a coupled holo V-ATPase (V₁H_chimV_oND) with catalytic properties similar to the ones of the recently characterized WT V₁V_oND (29). However, V₁H_chimV_oND was more resistant to ATP hydrolysis-induced disassembly compared with WT, highlighting the importance of H_CT's conformational switch in driving V-ATPase disassembly. The in vitro data presented here thus provide key insight into the molecular steps that accompany V-ATPase regulation by reversible disassembly.

Results

Purification and characterization of native lipid nanodisc reconstituted V_o (V_oND) and V₁ mutants

V_o was extracted from yeast vacuoles using the “reconstitution before purification” strategy as described for V₁V_o (29) (Fig. 1C, panels i–iii). The resultant V_oND complex consisted of V_o embedded in endogenous vacuolar lipid containing nanodiscs encircled by biotinylated membrane scaffold protein (MSP; Fig. 1C, panel iii). The purified complex was monodisperse and contained all V_o subunits plus MSP based on gel filtration and SDS-PAGE (Fig. 1, D and E). Examination of purified V_oND using negative stain EM showed single particles of V_oND with the typical size and appearance as described previously (26) (Fig. 1F).

Four different V₁ mutants were tested for their ability to bind V_oND and form coupled holo V-ATPase (Fig. 2A): WT V₁ (containing subunits A₃B₃DE₃FG₃H), V₁ purified from a yeast strain deleted for subunit H (V₁ΔH), V₁ΔH reconstituted with recombinantly expressed WT H (V₁H_wt), and V₁ΔH reconstituted with chimeric H (V₁H_chim) (16). Although WT V₁ has no measurable MgATPase activity (Fig. 2B, dark green trace), V₁ΔH had an initial specific activity of 15.7 ± 1.7 units/mg, consistent with previous reports (16, 30) (Fig. 2, B and C, blue trace and bar). Although MgATPase activity is measured in an ATP-regenerating system, the activity of V₁ΔH decreases over time as MgADP gets trapped in a closed catalytic site, leading to the MgADP-inhibited state (15). Although V₁H_wt has identical subunit composition to WT V₁, V₁H_wt distinguishes itself from WT by having only ∼0.4 instead of 1.3 mol/mol ADP in catalytic sites (16), and the complex therefore exhibits an initial MgATPase activity of ∼4.95 ± 0.55 units/mg before becoming MgADP-inhibited (Fig. 2, B and C, light green). V₁ΔH reconstituted with H_chim to yield V₁H_chim (Fig. 2, D–F) showed an initial MgATPase activity of 8.2 ± 0.4 units/mg that, as for the other V₁ mutants, declined over time because of MgADP inhibition (Fig. 2, B and C, pink).

Figure 2.

Purification and characterization of V₁ mutants. A, schematic representation of the V₁ mutants. B, time-dependent MgATPase activities of the V₁ mutants measured in an ATP regenerating assay. C, specific activities of the V₁ mutants ± S.E. from at least two independent purifications per mutant. D, size-exclusion chromatography of H_chim (blue trace) and V₁ΔH reconstituted with H_chim (V₁H_chim, red trace). E and F, SDS-PAGE of column fractions of H_chim (E) and V₁H_chim (F).

Binding of C subunit to V₁

Experiments in yeast have shown that upon deletion of the gene encoding the C subunit, V₁ does not stably/functionally associate with V_o. Moreover, biochemical analysis revealed that C_head binds isolated EG heterodimer with moderate affinity, whereas both C_foot and EG bind a_NT weakly (31, 32). From these data we concluded that V₁ and V_o are held together by multiple weak interactions, resulting in an overall high-avidity interface, and that destabilization of one of these interactions by a cellular response to starvation would result in enzyme dissociation (10, 32). More recently we showed that although H_NT binds isolated EG with a K_d of ∼0.2 μm, the affinity of the interaction is increased 40-fold when EG is part of V₁ (30). Moreover, when we analyzed binding of H (and H_NT) to V₁ΔH, we found that MgATP hydrolysis destabilized the V₁–H interaction, an effect likely caused by the cyclic conformational changes at the catalytic AB interfaces to which the EG heterodimers are bound (30). We therefore wished to determine whether C binding to EG on V₁ is also enhanced compared with isolated EG, and if so, (i) what the affinity of the interaction is, and (ii) whether the interaction is also destabilized during ATP hydrolysis. Because V₁ isolated from starving yeast has varying levels of substoichiometric amounts of C bound (15, 16, 33, 34), we purified V₁ from a yeast strain in which C was deleted (V₁ΔC) (16) to test for C binding. Using a BLI setup similar to the one we recently employed to analyze binding of H (and H_NT) to V₁ΔH (30), we found that V₁ΔC binds C with a K_d of ∼0.7 nm (Fig. S4), indicating that one of the EG heterodimers bound to V₁ is in a conformation that is more favorable for C binding compared with the isolated heterodimer. However, to test whether C binding is also destabilized as a result of ATP hydrolysis, we could not use V₁ΔC because it contains the H subunit and so has no MgATPase activity, and we therefore used the catalytically active V₁H_chim instead. As seen for the V₁–H interaction, dissociation of V₁H_chim from MBP-C loaded sensors was greatly accelerated only when the sensors were dipped into wells containing MgATP. Fitting the subunit C release in MgATP to two exponentials revealed a fast off rate of ∼0.012 ± 2.3 × 10⁻⁵ s⁻¹ and a slower off rate of 5.1 × 10⁻⁴ ± 2 × 10⁻⁶ s⁻¹, values similar to those observed for MgATP hydrolysis induced subunit H release as reported earlier (30) (Fig. S5). This suggests that the in vivo dissociation of the C subunit from the vacuolar membrane that occurs as a result of starvation is a direct result of MgATP hydrolysis–induced conformational changes of the EG heterodimer that is bound to C_head in the assembled enzyme (EG3) (23, 35). In summary, as for subunit H, binding of C to EG is significantly enhanced when EG is part of V₁, and the V₁–C interaction is greatly destabilized upon ATP hydrolysis.

V₁H_chim and C subunit associate with V_oND to form coupled V-ATPase in vitro

The ability of the V₁ mutants depicted in Fig. 2A to interact with V_o was tested using BLI. V_oND reconstituted with biotinylated MSP was immobilized on streptavidin-coated BLI sensors (Fig. 3A, step 1), which were then dipped in wells containing V₁ mutants and subunit C (Fig. 3A, step 2). We found that of the four V₁ mutants, only V₁H_chim showed significant association with V_oND (Fig. 3A, red trace). The sensors were then dipped in buffer to measure dissociation rates (Fig. 3A, step 3). However, no significant dissociation was observed, indicating stable assembly of V₁H_chim with V_o. Without C, none of the V₁ mutants showed significant binding (Fig. S3), consistent with studies in yeast that showed that deletion of C prevents assembly of V₁V_o (36). As a control, the sensors were then dipped in PreScission protease to cleave and release any remaining complex (Fig. 3A, step 4). Further, we conducted BLI experiments in which V_oND-coated sensors were first dipped into wells containing a mAb (10D7) against a_NT, which recognizes a cryptic epitope only available for binding in free V_o (12), before dipping the sensors into V₁H_chim plus C containing wells. Under these conditions, the observed on-rate (k_obs) of V₁H_chim was significantly (∼60-fold) reduced, indicating that the observed BLI signal upon dipping the sensors into V₁H_chim plus C was indeed due to binding of V₁H_chim to immobilized V_oND (Fig. S2).

Figure 3.

V₁H_chim and C associate with V_oND to form coupled V₁V_o-ATPase. A, V_oND was immobilized on streptavidin-coated BLI sensors via biotinylated MSP (step 1). Sensors were then dipped into 0.4 μm of V₁ mutants in presence of 1 μm C (association; step 2) followed by buffer (dissociation; step 3). Association with V_oND was most efficient with V₁H_chim (red trace). Sensors were then dipped in PreScission protease to verify that the BLI signal was not due to nonspecific binding (step 4). Inset shows an enlarged view of the association and dissociation steps. B, equimolar amounts of V₁H_chim and V_oND, and a 2-fold molar excess of C subunit were incubated at 22 °C, and the ConA-sensitive MgATPase activity was measured as a function of time. Each point represents the mean ± S.E. of two separate reconstitutions from two individual purifications. Inset, specific MgATPase activities of reconstituted V₁H_chimV_oND and V₁H_wtV_oND (± S.E. from two independent purifications) compared with purified V₁V_oND (29). C, following association of the V₁H_chimV_oND complex, sensors were dipped in wells containing buffer (green) or buffer + 1 mm MgATP (blue) for dissociation rate measurement. The dissociation phase of WT V₁V_oND in buffer (red) and buffer +1 mm MgATP (orange) is included for comparison (data from Ref. 29).

Whereas the BLI experiment showed slow, but stable, association of V₁H_chim and C with V_oND, it was not clear whether functional V-ATPase was formed under these conditions. To address this question, we monitored MgATPase activity of a 1:1:2 mixture of V₁H_chim, V_oND, and C that is sensitive to the V-ATPase specific inhibitor, concanamycin A (ConA), as a function of time (Fig. 3B). Although V₁H_chim has MgATPase activity on its own, ConA binds to the V_o complex and prevents c-ring rotation, so ATPase activity that is abolished by treatment with ConA is evidence of a functionally coupled V-ATPase complex. The experiment demonstrated that binding of V₁H_chim and C to V_oND resulted in the formation of a coupled V₁H_chimV_oND complex and that the reconstitution under these conditions was complete in ∼2 h, with a final specific activity of 7.2 ± 1.09 units/mg, similar to what is reported for purified WT V₁V_oND (6.9 ± 0.6 units/mg) (29) (Fig. 3B, inset, pink and blue bars, respectively). The ability of V₁H_chim to form a functional complex with V_oND is consistent with the previous observation that H_chim can complement deletion of the native H subunit in yeast cells (16). ConA-sensitive ATPase activity was also measured with V_oND, C, and V₁H_wt and produced only 0.26 ± 0.2 units/mg of coupled activity (Fig. 3B, inset, orange bar). Therefore, reconstitution of V₁H_wt with V_o and C is highly inefficient under these conditions, consistent with earlier in vitro studies (28) and the real-time BLI experiments presented here (Fig. 3A).

V₁H_chimV_oND is more stable in presence of MgATP compared with V₁V_oND

Using BLI, we have previously demonstrated that V₁'s dissociation from V_o is negligible under non-ATP hydrolyzing conditions but that in presence of MgATP, the complex undergoes spontaneous dissociation with an off-rate of 1 × 10⁻³ ± 3.3 × 10⁻⁶ s⁻¹ (29). We conducted a similar experiment using V₁H_chimV_oND, wherein after association of V₁H_chim and C with V_oND on BLI sensors, we dipped the sensors in wells containing buffer or 1 mm MgATP. Significantly, unlike WT V₁V_oND, V₁H_chimV_oND showed very little to no dissociation in the presence of MgATP, indicating that the assembled V₁H_chimV_oND complex is inherently more stable than the WT complex (Fig. 3C). The experiment thus highlights the importance of H_CT's conformational switch (from energy coupling in the holo enzyme to autoinhibition of membrane detached V₁) in driving V-ATPase disassembly.

Structural and functional characterization of V₁H_chimV_oND

To determine the efficiency of V₁H_chimV_oND complex formation, equimolar amounts of V₁H_chim and V_oND with a 2-fold molar excess of C were incubated for up to 16 h at 22 °C, and the reconstitution mixture was resolved by glycerol density gradient centrifugation. SDS-PAGE of the gradient fractions showed that the majority of V₁ and V_o subunits co-migrated to fractions 7–10, similar to what was observed for purified WT V₁V_oND (29), with the excess C subunit remaining in lighter fractions (Fig. 4A). Negative stain EM of the peak fractions showed single particles of V₁H_chimV_oND, with the typical dumbbell-shaped appearance of holo V-ATPase reported previously (29, 37, 38) (Fig. 4B). A more detailed analysis indicated a good match between averages obtained by reference free alignment and classification of a small data set of V₁H_chimV_oND and corresponding projections of a cryoEM map of yeast V₁V_o (23) (Fig. 4C). Taken together, these data show that reconstitution of V₁H_chim with V_oND and C result in a stable and coupled holo V-ATPase that is structurally similar to the WT enzyme.

Figure 4.

Structural and functional characterization of the V₁H_chimV_oND complex. A, reconstituted V₁H_chimV_oND was subjected to glycerol gradient centrifugation, and the gradient fractions were analyzed by silver-stained SDS-PAGE. B, negative stain EM of V₁H_chimV_oND showing homogeneous and monodisperse dumbbell-shaped molecules. Inset in the top right shows 2× zoomed area highlighted in the bottom left. C, a data set of ∼5800 particle projections was subjected to reference-free alignment and classification, and selected class averages were overlaid with projections of the cryoEM model of yeast V₁V_o (Protein Data Bank code 3J9U). Bars in B, 50 nm.

Discussion

In our experiments, only V₁H_chim (in presence of C) shows significant binding to V_o. As mentioned earlier, human H_CT lacks the inhibitory loop found in yeast H_CT (16) (Fig. S1, A–D, red spheres), consistent with H_chim's inability to inhibit V₁ΔH's MgATPase activity (Fig. 2, B and C). This lack of inhibition by H_chim is likely due to reduced binding of human H_CT to the open catalytic site at the bottom of the A₃B₃ hexamer, with the remaining binding interaction between H_chim and V₁ being mediated by H_NT's interaction with one of the EG peripheral stalks. Therefore, we conclude that for V₁ to reconstitute with V_o, H_CT must be released from its inhibitory position on V₁, so that H_CT is available for binding to V_o's a_NT. It is known that upon disassembly of V₁ from V_o, a_NT moves from a peripheral position in V₁V_o to a more central position in autoinhibited V_o, where it binds subunit d (18, 25–27) (Fig. S1, E–G). The observation that V₁H_chim, C, and autoinhibited V_oND are sufficient to form a structurally and functionally coupled V-ATPase suggests that the release of H_CT from its autoinhibitory position on V₁ is necessary and sufficient for efficient reassembly of V₁ and V_o. Our finding is consistent with the fact that reassembly of V₁ with V_o on vacuoles is not inhibited by ConA, an inhibitor of c-ring rotation in the V_o sector (14).

H_NT and H_CT occupy specific binding sites on free V₁, with H_NT bound to EG1 and H_CT bound to the bottom of the A₃B₃ hexamer, with its inhibitory loop wedged between the B subunit of an open catalytic site and the central stalk (Fig. S1, A–D) (16). The specific interaction of H_CT with an open catalytic site maintains inhibitory MgADP in the adjacent closed catalytic site, locking autoinhibited V₁ in rotational state 2. We have recently observed that transient MgATP hydrolysis on V₁H_wt, which, unlike WT V₁, is not in the MgADP-inhibited state (Fig. 2B), lowers H's affinity for V₁, and we reasoned that this destabilization of the V₁–H interaction is caused by MgATP hydrolysis driven conformational changes at the catalytic sites and the central (DF) and peripheral stalks (EG1–3) (30). We propose that an allosteric structural change at the open catalytic site that is driven by release of inhibitory MgADP from the closed catalytic site in autoinhibited, WT V₁ by a yet unknown mechanism leads to the detachment of H_CT from its inhibitory position so that it can bind a_NT on V_o.

From studies in yeast it was shown that although C is required for binding of V₁ to V_o on yeast vacuolar membranes (36), deletion of H allows assembly of a labile but inactive complex (39). The requirement of C for association of V₁ with V_o is supported by our in vitro BLI experiments, wherein none of the V₁ mutants reconstituted with V_o in absence of C (Fig. S3). However, unlike in vivo, the presence of C along with V₁ΔH, WT V₁, or V₁H_wt is not sufficient for reconstituting V₁V_o in vitro. This discrepancy is not due to a reduced affinity of C for V₁, because we obtained a K_d of ∼0.7 nm for the interaction between C and V₁ΔC (Fig. S4). A high-affinity interaction between C and V₁ΔC is consistent with substoichiometric amounts of C remaining associated with purified V₁ΔH and V₁ (16), as well as the reported K_d of ∼42 nm for the EG–C interaction (31). However, tight and stable binding of C to V₁ΔC would be inconsistent with the observed release of C into the cytosol upon disassembly of V₁ from V_o (14), but, as shown here, ATP hydrolysis by V₁ leads to the rapid release of the V₁–C interaction, which likely explains why catalytically inactive enzyme does not disassemble upon glucose withdrawal (14). Although it has been reported that in the presence of the microtubule depolymerizing drug benomyl, C does not dissociate from V₁V_o upon glucose removal (40), it is possible that C under these conditions quickly rebinds EG3 once V₁ is in the autoinhibited state. In addition, a direct interaction between C and tubulin has been observed (40, 41), suggesting the possibility that, upon disassembly, C is sequestered by microtubules, preventing its reassociation with V₁ and/or V_o.

Although reincorporation of C upon glucose addition does not require the microtubule network (40), efficient (re)assembly of holo V-ATPase requires a heterotrimeric chaperone complex referred to as RAVE (regulator of H⁺-ATPase of vacuolar and endosomal membranes). It has been proposed that upon receiving the signal for reassembly, RAVE recruits C and V₁ to V_o on vacuolar membranes by directly interacting with C, EG (as part of V₁), and a_NT (as part of V_o) (42). Under the in vitro conditions employed here, it takes ∼2 h for a 1:1:2 mixture of V₁H_chim, V_oND, and C to complete reconstitution of V₁H_chimV_oND, a relatively slow process compared with the kinetics of reassembly observed in vivo (∼5 min) (43). It is possible that the RAVE complex, by increasing the proximity of V₁, C, and V_o, facilitates the otherwise low-affinity interactions at the V₁–V_o interface (32), thereby accelerating reassembly.

From the here presented data, we conclude that the detachment of H_CT from V₁ and the presence of C subunit are required for the reassembly of V₁ with V_o (Fig. 5, A–C). In our in vitro reconstitutions, the association between V₁H_chim and V_oND is driven by the H_CT–a_NT interaction (Fig. 5B), but in vivo, the chain of events that leads to reassembly of autoinhibited V₁ and V_o are probably different, because H_CT is in its inhibitory conformation on V₁ (Fig. 5A). We propose that in yeast, upon receiving cellular signals, autoinhibited V₁ and C are first recruited to V_o (Fig. 5D), a process that is likely the rate-limiting step for reassembly. Our reason for this hypothesis is that even with the requirements for reassembly being met in our in vitro reconstitution of V₁H_chim and C with V_o, the rate of reassembly was slow. In vivo, recruitment of V₁ and C to V_o is facilitated and probably accelerated by the RAVE complex (42), but efficient (re)assembly in vivo requires additional factors such as the glycolytic enzymes aldolase (44) and phosphofructokinase (43), whose function in the process is currently not known. Once V₁ and C are recruited to V_o at the vacuolar membrane, inhibitory MgADP is released upon opening of the closed catalytic site by a yet unknown mechanism (Fig. 5E). The release of inhibitory MgADP allows MgATP hydrolysis to resume, with concomitant conformational changes at the catalytic sites and rotation of the central stalk (DF), structural changes that result in detachment of H_CT from V₁ (30) (Fig. 5F). The proximity of a_NT to V₁-detached H_CT facilitates the H_CT–a_NT interaction, a requirement for coupling of V₁ to V_o in the holo-enzyme (24). The H_CT–a_NT interaction stabilizes the peripheral conformation of a_NT such that the C_foot–EG2–a_NT ternary complex can be formed, thus completing functional (re)assembly (Fig. 5C).

Figure 5.

Model for reassembly of autoinhibited V₁ and V_o. A–C, our in vitro experiments have shown that although WT H containing V₁ does not readily bind V_oND (A), V₁H_chim spontaneously associates with V_oND (B) to form a structurally and functionally coupled V-ATPase, albeit at a slow rate. C, in vivo, however, V₁ exists in the autoinhibited conformation (A), and the rate of assembly with V_o is significantly faster (within 5 min). D–F, for in vivo (re)assembly, we propose that the following steps occur: step 1, recruitment of V₁ and subunit C to the vacuolar membrane (D); step 2, release of inhibitory MgADP (E); step 3, detachment of H_CT from its inhibitory position on V₁; and step 4, H_CT binding to a_NT (F). For further details, see text.

V-ATPase regulation by reversible disassembly, originally discovered in lower eukaryotes, has been confirmed to be conserved in higher animals, including humans (19–21). Reconstitution of V₁ with V_o has been investigated for mammalian V-ATPase from bovine brain clathrin-coated vesicles and in one study, chaotropically removed V₁ reassembled with V_o on coated vesicle membranes upon dialysis, thereby regenerating ∼80% of the initial MgATPase activity (45). In another study, in vitro reconstitution of coated vesicle V-ATPase from V₁ and V_o was shown to require the mammalian H subunit homolog SFD (sub-fifty-eight dimer) (46). However, in both cases, resulting V-ATPase complexes were not further characterized for subunit composition and structural integrity. Curiously, unlike yeast V₁, removal of SFD from bovine coated vesicle V₁ did not create a MgATP hydrolyzing V₁-ATPase, suggesting the presence of other regulatory mechanisms in mammalian V₁ (46). One of the likely reasons that few biochemical studies have focused upon the molecular mechanism of reversible disassembly in higher organisms is because mammalian V-ATPase is extraordinarily heterogeneous, with most subunits expressed as multiple isoforms or splice variants (including subunits H and a) (47), and to our knowledge, no in vitro system comparable to the one described here for the yeast V-ATPase has been reported for the mammalian enzyme. Yeast contains only one subunit with multiple isoforms (subunit a), and the two V-ATPase populations resulting from this single subunit difference appear to have different propensities to undergo dissociation, and only one of them requires the RAVE complex for (re)assembly (48, 49). It is likely that different isoform-containing enzymes in mammalian systems are subjected to differential regulatory mechanisms, resulting in variable propensities to dissociate. Because the human H_CT does not silence yeast V₁ but does facilitate efficient functional coupling in V₁V_o, it will be of interest to explore the mechanism of regulation by reversible disassembly as a function of subunit isoform composition of the mammalian system in greater detail by using the tools developed and presented here for the yeast enzyme.

Experimental procedures

Strains

The yeast strain SF838–5Aα deleted for the vma2 gene (B subunit) vma2Δ::Nat was a kind gift from Dr. Patricia Kane, SUNY Upstate Medical University. A plasmid containing the FLAG tag with a KanMX6 marker, pFA6a-6xGly-FLAG-kanMX6 was a gift from Dr. Mark Hochstrasser (50) (Addgene plasmid no. 20751). The primers vph1CTFlagFWD (gct gtt gct agt gca agc tct tcc gct tca agc GGG GGA GGC GGG GGT GGAA) and vph1CTFlagREV (cct gga tgt gga ttt cga ttc taa cgt tac ccc aag gca aat gat ggt cac tgg GAA TTC GAG CTC GTT TAA AC) were used to amplify the FLAG tag and KanMX marker from pFA6a-6xGly-FLAG-kanMX6. The ∼1.8-kb product was gel purified and used for homologous recombination to insert the FLAG-KanMX cassette in the C terminus of vph1 in the yeast strain SF838–5Aα vma2Δ::Nat using the same primers as above. Colonies were selected for growth on YPD G418 plates, and the insertion of the FLAG tag at the C terminus of vph1 was confirmed by sequencing. The construction of chimeric H subunit (H_chim) encoding the N-terminal domain from Saccharomyces cerevisiae (residues 1–352) and the C-terminal domain (349–483) of the human H subunit into the yeast pRS316 vector has been discussed in Ref. 16. From the pRS316 vector, using the primers MalChimF (TTA GCC GGT ACC GGG AGC AAC GAA GAT ATT AAT GGA C) and MalChimR (TTA CCA AAG CTT TTA GCT TCG GGC GGC AG), H_chim was amplified. The primers additionally introduced the restriction sites 5′ KpnI and 3′ HindIII, which were used to insert the amplicon into a pMal vector. The resultant vector encoded MBP-tagged H_chim separated by a PreScission protease cleavage site, as confirmed by sequencing.

Purification of V_o and its reconstitution into endogenous vacuolar lipid

V_o was purified from yeast vacuoles and reconstituted into endogenous vacuolar lipid containing nanodiscs as described for V₁V_oND in Ref. 29. The steps are briefly described as follows.

Purification of biotinylated MSP

Biotinylated MSP was purified as described in Ref. 29. Briefly, BL21 (DE3) cells were co-transformed with the plasmids pHBPMSP1E3D1 and pBirAcm (encoding the BirA gene). The cells were grown in rich broth supplemented with 0.1 mm d-biotin, 34 μg/ml chloramphenicol, and 30 μg/ml kanamycin to an A₅₉₅ of ∼0.5 at 37 °C followed by induction using 0.5 mm isopropyl β-d-thiogalactopyranoside for 3–4 h. Harvested cells were purified as described in Ref. 26. Briefly, the cells were lysed by sonicating three times for 30 s. The lysate was cleared by centrifugation at 13,000 × g and passed over a nickel–nitrilotriacetic acid affinity column. The column was washed with 10 column volumes of the each of the three buffers: 40 mm Tris-HCl, 300 mm NaCl, and 1% Triton X-100, pH 8; 40 mm Tris-HCl, 300 mm NaCl, 50 mm sodium cholate, and 5 mm immidazole, pH 8; and 40 mm Tris-HCl, 300 mm NaCl, and 10 mm immidazole, pH 8. MSP was eluted with a 10–column volume gradient of the elution buffer (40 mm Tris-HCl, 300 mm NaCl, and 100 mm immidazole, pH 8). Purified biotinylated MSP was dialyzed into 25 mm Tris, 150 mm NaCl, 0.5 mm EDTA, pH 7.2; concentrated to ∼5 mg/ml; snap frozen in liquid nitrogen; and stored at −80 °C until use.

Isolation of yeast vacuoles

Yeast vacuoles were isolated by flotation on a Ficoll density gradient as described in Ref. 51. Briefly, SF838–5Aɑ vma2Δ::Nat with a Flag tag on the C terminus of vph1 (a subunit) was grown to an A₅₉₅ of ∼1.0 in YPD pH 5. 12 liters of cells were harvested by centrifugation at 5000 × g for 30 min. The pellet was washed and resuspended in 100 ml of 1.2 m sorbitol with ∼15 mg of zymolyase to form spheroplasts. The spheroplasts were recovered in 100 ml each of 2.4 m sorbitol and 2× YPD and then resuspended in buffer containing 12% Ficoll 400. The suspension was homogenized in a Dounce homogenizer and centrifuged at 71,000 × g for 40 min. Vacuole wafers from the top of the gradient were extracted, homogenized in buffer containing 8% Ficoll, and centrifuged at 71,000 × g for 40 min. The final vacuole wafers were resuspended in 1.5 mm Mes-Tris, pH 7.0, 5% glycerol, and 1 mm β-mercaptoethanol. Vacuolar protein concentration was measured using a modified BCA assay (18), and the vacuoles were frozen in liquid nitrogen until further use.

Extraction of V_o and reconstitution into lipid nanodiscs

Three batches of purified vacuoles (12 liters each) were typically used for one extraction as described in Ref. 29. Briefly, thawed vacuoles were combined, supplemented with protease inhibitors, and solubilized using 1.2 mg of n-dodecyl β-d-maltopyranoside/1 mg of vacuolar protein. To the detergent-solubilized sample, purified biotinylated MSP was added in a molar ratio of 1:50 (vacuolar protein:MSP). The mixture containing vacuolar protein, vacuolar lipids, and MSP was incubated at 4 °C for 1 h followed by detergent removal using bio-beads. Reconstituted vacuolar membrane proteins in biotinylated and endogenous vacuolar lipid containing nanodiscs were subjected to anti-FLAG affinity chromatography to purify V_o-containing nanodiscs. The eluate from the FLAG column was then subjected to size-exclusion chromatography using a Superdex 200 1 × 30-cm column. Peak fractions from gel filtration were combined and concentrated using a Vivaspin 100,000 molecular weight cutoff concentrator.

Purification of the chimeric H subunit (H_chim)

Escherichia coli Rosetta2 (Novagen) cells expressing N-terminal MBP-tagged H_chim were grown to an A₆₀₀ of ∼0.5 (in LB, 0.2% glucose, 50 μg/ml carbenicillin, and 34 μg/ml chloramphenicol), and expression was induced with 0.5 mm isopropyl β-d-thiogalactopyranoside at 30 °C for 4 h. Protein was purified using amylose affinity chromatography, and the MBP tag was cleaved with PreScission protease as previously described (31). The pH of the cleavage product was adjusted to 7 by overnight dialysis in 25 mm sodium phosphate, pH 7, 0.5 mm EDTA, and 5 mm β-mercaptoethanol. At pH 7, H_chim has a predicted charge of +3.5, whereas the predicted charge of MBP is −9 (Protein Calculator v3.4), allowing separation of the two proteins using cation exchange (carboxymethyl) chromatography. The cleaved MBP came off the carboxymethyl column in the flow through and wash steps, whereas pure H_chim was eluted in dialysis buffer supplemented with 100 mm NaCl. The preparation was subjected to a final step using a Superdex 200 1.6 × 50-cm size-exclusion chromatography column.

Purification of V₁ mutants and subunit C

WT V₁, V₁ΔH and V₁ΔC were purified from vma10Δ::KanMX (29), vma13Δ::KanMX, vma10Δ::Nat (34), and vma5Δ::Nat, vma10Δ::URA3 (16), respectively, as described in Ref. 16. In all cases, yeast strains were transformed with a pRS315 vector containing N-terminally FLAG tagged vma10 (G subunit) for affinity purification on αFLAG agarose (33). Cells were grown to an A₅₉₅ of ∼4 in synthetic dropout medium without leucine (SD −Leu) and harvested by centrifugation at 4000 × g for 15 min. The cells were lysed by ∼15 passes through a microfluidizer (Microfluidics M-110L). Unbroken cells were pelleted by centrifugation at 4000 × g for 30 min, and the resultant supernatant was cleared by centrifugation at 13,000 × g for 40 min. Cleared lysate was subjected to affinity chromatography using αFLAG resin. The eluate from the αFLAG column was concentrated and subjected to size-exclusion chromatography using a Superdex 200 1.6 × 50 cm column.

For preparation of V₁H_chim and V₁H_wt, V₁ΔH eluted from the αFLAG column was incubated for 1 h at 4 °C with a ∼5-fold molar excess of either H_chim or H_wt (purified as in Ref. 30) to form the V₁H_chim and V₁H_wt complexes, respectively. V₁ bound to H_chim or H_wt was then separated from the excess of H_chim or H_wt by size-exclusion chromatography using a Superdex 200 1.6 × 50-cm column. Subunit C was purified as previously described (31).

Biolayer interferometry

Interaction of V_o with the purified V₁ mutants was screened using BLI, a light interference–based technique, similar to surface plasmon resonance. An Octet-RED system with streptavidin coated biosensors (FortéBio, SA biosensors, catalog no. 18-5019) were used for the experiments. All BLI experiments were conducted using 25 mm Tris-HCl, pH 7.2, 150 mm NaCl, 0.5 mm EDTA, 1 mm β-mercaptoethanol, 0.5 mg/ml BSA, except for experiments analyzing V₁H_chim release from immobilized MBP-C, which required 10 mg/ml BSA because of an increased propensity of chimeric H containing V₁ to bind nonspecificically to the BLI sensors. The temperature was maintained at 22 °C, with each biosensor stirred in 0.2 ml of sample at 1000 rpm and a standard measurement rate of 5 s⁻¹. Streptavidin-coated biosensors were prewetted in BLI buffer and then dipped in wells containing 3 μg/ml of biotinylated V_oND. A buffer control was included to show that none of the buffer components interacted with the sensors. Details of individual experiments have been described in the respective figure legends. The affinity of interaction between V₁ΔC and MBP-C was measured using anti-mouse IgG Fc capturing biosensors (FortéBio, AMC biosensors catalog no. 18-5088) as described (30).

ATPase activity assay

MgATPase activity of purified V₁ mutants and reconstitution mixtures (V₁ mutants + V_oND + subunit C) was measured using a coupled enzyme assay as described in Ref. 16. Briefly, 10 μg of the V₁ mutant was added to an assay mixture containing 1 mm MgCl₂, 5 mm ATP, 30 units/ml each of lactate dehydrogenase and pyruvate kinase, 0.5 mm NADH, 2 mm phosphoenolpyruvate, and 50 mm HEPES, pH 7.5, at 37 °C. The decrease of absorbance at 340 nm corresponding to the decline of NADH in the system was measured in the kinetics mode on a Varian Cary Bio100 spectrophotometer. In case of reconstitution mixtures, 20 μg of V₁ mutant with equimolar amounts of V_oND, and a 2× molar excess of C subunit was added to an assay containing 4 mm MgCl₂.

Author contributions

S. S., R. A. O., and S. W. conceptualization; S. S., R. A. O., and M. M. K. investigation; S. S. writing-original draft; R. A. O. and S. W. writing-review and editing; S. W. funding acquisition.