Molecular mechanism of Oxr1p mediated disassembly of yeast V-ATPase

Prior studies in yeast had shown that glucose withdrawal-induced V-ATPase disassembly requires catalytically active enzymes (Parra and Kane, 1998). On the other hand, more recent data in higher organisms suggested that loss of TLDc domain protein function negatively impacts V-ATPase activity (Merkulova et al, 2018). As these two observations appeared to be at odds with our finding that Oxr1p disassembled V-ATPase in vitro without the need for ATP, we speculated that Oxr1p may serve in quality control by selectively removing V₁ subcomplexes that were rendered inactive in vivo due to e.g., oxidative damage (Khan et al, 2022). At the onset of the current study, we therefore tested Oxr1p’s ability to release V₁ from vacuoles harboring inactive V-ATPases. For this, we purified vacuoles from a yeast strain deleted for the CYS4 gene, which has been shown to reduce V-ATPase activity by more than 50% (Oluwatosin and Kane, 1997) (Appendix Fig. S1A). As an alternative approach, we treated wild-type vacuoles with hydrogen peroxide (H₂O₂), which is known to inhibit V-ATPases due to oxidation of active site cysteine residues (Liu et al, 1997), and the V-ATPase-specific inhibitor, Concanamycin A (ConA) (Appendix Fig. S1A). However, when we tested Oxr1p’s impact under these conditions, we found that Oxr1p was equally efficient in releasing V₁ from vacuoles with inactive V-ATPases compared to wild-type vacuoles (Appendix Fig. S1B-D). Overall, this shows that whereas Oxr1p is able to release V₁ from inactive V-ATPases, it appears that Oxr1p does not prefer inactive over active enzymes, contrary to our initial hypothesis. We therefore asked whether Oxr1p’s ability to disassemble the V-ATPase in vitro plays a role in canonical V-ATPase regulation by reversible disassembly in vivo after all.

To elucidate Oxr1p’s physiological role in V-ATPase regulation, we deleted OXR1 from wild-type yeast to generate strain oxr1Δ. We then transformed oxr1Δ with a low-copy number plasmid (pRS316, CEN6, URA3) for expression of C-terminally HA-tagged Oxr1p (Oxr1p-HA) under its native promoter (strain oxr1^NP), or with a high-copy number plasmid (YEp352, 2 µ, URA3) for overexpression of Oxr1p-HA (strain oxr1^OE). Western blots of oxr1^NP and oxr1^OE lysates revealed a more than twofold overexpression of OXR1 in the oxr1^OE strain (Fig. 2A). Viability assays revealed that all three strains were able to grow at elevated pH in presence of Ca²⁺, a condition that requires functional V-ATPase (Nelson and Nelson, 1990), with growth somewhat impaired for oxr1Δ and oxr1^OE strains at elevated temperature (37 °C) (Fig. 2B). As a control, we used a yeast strain deleted for VMA1 (subunit A of V₁ subcomplex; vma1Δ), which displayed the typical V-ATPase loss of function, or Vma^- phenotype characterized by the inability to grow at elevated pH in presence of Ca²⁺ (Fig. 2B). To verify that Oxr1p’s activity is not impaired by the C-terminal HA tag, we bacterially expressed Oxr1p-HA and found it to be equally efficient in inhibiting the ATPase activity of purified vacuoles as the untagged protein (Appendix Fig. S2). Since the Vma⁻ phenotype is not obvious unless V-ATPase activity is reduced to less than ~30% of wild-type (Liu and Kane, 1996), we compared the ATPase activities of vacuoles isolated from the three mutant strains and wild-type. Whereas ATPase activities of oxr1Δ and wild-type vacuoles were similar, those from oxr1^NP and oxr1^OE strains were reduced to ~75 and ~50% of wild-type, respectively (Fig. 2C). To see whether this reduction in activity for the oxr1^OE vacuoles was due to differences in V-ATPase assembly, we determined the relative ratios of V₁ over V_o on purified vacuoles and found that whereas vacuoles from the oxr1Δ strain contained ~40% more V₁ compared to wild-type, oxr1^OE vacuoles had ~25% less (Fig. 2D). However, western blot analysis of total cell lysates revealed comparable amounts of V₁ subunit A (Vma1p) in all strains, suggesting that V₁ expression is not affected by deletion or overexpression of OXR1 (Fig. 2E). Taken together, and considering that Oxr1p causes V-ATPase disassembly in vitro (Khan et al, 2022), the data suggest that the higher and lower levels of V₁ on oxr1Δ and oxr1^OE vacuoles are due to decreased and increased V-ATPase disassembly, respectively.

V-ATPase reversible disassembly is a dynamic process, in which the balance between assembled and disassembled states is dictated by, for example, the glucose concentration (Parra and Kane, 1998). If Oxr1p is involved in the reversible disassembly process, this balance should be perturbed in its absence. To study the role of Oxr1p on V-ATPase disassembly in vivo, we generated a C-terminal fusion of V₁ subunit C with mNeonGreen (C-mNG) using homologous recombination in wild-type and oxr1Δ strains. The expectation here is that under conditions that favor assembly, most of the C-mNG will be part of holo V-ATPases, with fluorescence thus largely limited to the vacuole. On the other hand, in situations that favor disassembly, most of the C-mNG (and V₁) will be released into the cytoplasm, resulting in a diffuse fluorescence signal throughout the cell. To capture the reversible disassembly process, we collected fluorescence images from exponentially growing cells in three different growth conditions: (i) glucose-fed, (ii) glucose starved, and (iii) glucose restored. In all experiments, vacuolar membranes were visualized using the vacuole-specific dye FM4-64, and colocalization of the C-mNG and FM4-64 fluorescence intensity was quantified via Pearson’s correlation coefficients (Fig. 3). In wild-type cells, we observed the expected switch from C-mNG fluorescence on the vacuole under glucose-fed conditions, to diffuse fluorescence upon glucose withdrawal, and return of the signal to the vacuole upon glucose re-addition (Fig. 3A). Strikingly, the outcome of the experiment in the oxr1Δ cells was different: here the C-mNG fluorescence remained on the vacuole in all three conditions (Fig. 3B), indicating that Oxr1p is required for the enzyme disassembly step. To validate this result, we complemented the OXR1 deletion with a pRS316 plasmid expressing C-terminally HA-tagged Oxr1p and found that the switch of C-mNG fluorescence upon glucose withdrawal and re-addition was restored to wild-type levels (Fig. 3C). Combined with our earlier finding that Oxr1p causes V-ATPase disassembly in vitro (Khan et al, 2022), the above C-mNG fluorescence experiments provide strong evidence that Oxr1p is required for efficient V-ATPase disassembly in vivo. Next, we asked whether V-ATPase can disassemble in the oxr1Δ strain if we starve the cells of glucose longer than the 15 min used in our fluorescence microscopy assays. To answer this question, we collected fluorescence images of the oxr1Δ strain expressing C-mNG over a time period of 2 h at time points 0.5, 0.75, 1, 1.5, and 2 h following glucose deprivation (Fig. EV1). Whereas no significant disassembly was observed up to the 1 h time point, an increasing number of cells showed some level of cytosolic mNG fluorescence at 1.5 (23%) and 2 h (42%) (Fig. EV1B). The mechanism of this delayed (Oxr1p independent) disassembly in some of the cells, however, is currently not known and will require further investigation.

A previous study had shown that C-terminally GFP-tagged Oxr1p localizes to mitochondria (Elliott and Volkert, 2004). However, algorithms for predicting mitochondrial targeting sequences (Fukasawa et al, 2015; Savojardo et al, 2020) did not detect such a sequence in Oxr1p’s N-terminus, and available large-scale yeast protein localization studies using C-terminal GFP fusions were inconclusive (Huh et al, 2003). Moreover, the evidence presented above that Oxr1p is directly involved in reversible disassembly at the level of the enzyme means that Oxr1p must be able to access the vacuolar membrane under conditions that promote V-ATPase disassembly. To revisit the cellular localization of Oxr1p, we generated a genomic C-terminal fusion of Oxr1p with mNeonGreen (Oxr1p-mNG) using homologous recombination. Of note, our recent structure of V₁(C)Oxr1p shows Oxr1p’s C-terminus partially buried in the V₁:Oxr1p interface (Khan et al, 2022), which means that a bulky C-terminal fusion such as a fluorescent protein (~27 kDa) could impair Oxr1p’s ability to bind the enzyme. We therefore transformed the oxr1Δ background with a plasmid expressing N-terminally mNG-tagged Oxr1p (mNG-Oxr1p). To visualize mitochondria, we used MitoTracker dye. The fluorescence microscopy images show diffuse mNG fluorescence throughout the cytoplasm for both strains, with some mNG fluorescence co-localizing to MitoTracker dye fluorescence as quantified using the Pearson’s coefficient (Fig. 4A). Thus, whereas the experiment suggests that some Oxr1p may be targeted to mitochondria, the images show that the bulk of mNG-tagged Oxr1p is cytosolic, consistent with Oxr1p’s ability to bind and disassemble the V-ATPase on the vacuolar membrane in vivo. Given its cytoplasmic localization, we then asked whether Oxr1p is recruited to the vacuolar membrane upon glucose withdrawal, analogous to the V-ATPase assembly chaperone, RAVE, which goes to the vacuole upon glucose re-addition (Jaskolka and Kane, 2020). To address this question, we collected fluorescence images of both strains expressing mNG-tagged Oxr1p in the three different growth conditions as described in the previous section: (i) glucose-fed, (ii) glucose starved, and (iii) glucose re-addition. However, unlike what was observed for RAVE, both constructs remained equally distributed in between vacuole and cytoplasm in all three conditions, with only minor changes for Oxr1p-mNG and mNG-Oxr1p localization upon glucose re-addition and withdrawal, respectively (Fig. 4B,C).

Prior research using live-cell imaging showed that glucose withdrawal-induced disassembly of yeast V-ATPase occurs on a timescale of ~5–10 min (Dechant et al, 2010), consistent with the live-cell microscopy data for wild-type and oxr1^NP strains (Fig. 3A,C). This is in stark contrast to our recent study, where we found that ~30% disassembly of in vitro reconstituted V-ATPase required a 16-h incubation with a twofold molar excess of Oxr1p (Khan et al, 2022). In vitro reconstitution of V-ATPase from individual subcomplexes (V₁, V_o, and C), however, requires replacing of the inhibitory H subunit with a chimeric version of H (H_chim) composed of yeast N-terminal (H_NT) and human C-terminal (H_CT) domains (Sharma et al, 2019). While the H_chim containing V-ATPase is active and fully coupled, it resists the slow ATP hydrolysis-induced disassembly observed for the wild-type enzyme (Sharma et al, 2019; Sharma and Wilkens, 2017). We therefore wondered whether the slower disassembly rate observed for the in vitro assembled enzyme was due to the presence of H_chim. To address this question, we affinity-purified wild-type V-ATPase and reconstituted the complex into lipid nanodiscs (Appendix Fig. S3). The purified enzyme was then incubated with a twofold molar excess of Oxr1p and measured the time-dependent inhibition of ATPase activity over 16 h. However, similar to the H_chim containing enzyme, wild-type V-ATPases retained around 70% and 30% of ATPase activity after 30 min and 16 h of incubations, respectively (Fig. 5A). Thus, whereas the live-cell fluorescence microscopy data presented in the previous sections provided strong evidence that Oxr1p plays a key role in reversible disassembly, the rate of the process observed in vivo vs in vitro is significantly faster, suggesting the involvement of additional factor(s) in the cell.

As discussed earlier, efficient V-ATPase disassembly in vivo requires enzymatic activity (ATP hydrolysis) (Liu and Kane, 1996; Parra and Kane, 1998). However, which step of the disassembly process is ATP-dependent is not known. We recently reported that the dissociation of mutant V₁ subcomplexes from immobilized subunit C is greatly accelerated in the presence of ATP (Sharma et al, 2019). We therefore asked whether ATP might affect the rate of Oxr1p-mediated disassembly. To address this question, we incubated wild-type V-ATPase with a twofold molar excess of Oxr1p or ATP individually or in combination for 30 min, and measured ATPase activities. Whereas Oxr1p alone resulted in ~30% loss of activity, inhibition was close to complete in presence of Oxr1p and ATP (Fig. 5B), with negative stain EM confirming virtually complete V-ATPase disassembly (Fig. 5C,D). Thus, while Oxr1p alone is able to slowly disassemble V-ATPase, the presence of ATP makes the process much faster. To better understand the kinetics of Oxr1p-mediated V-ATPase disassembly, we carried out biolayer interferometry (BLI) experiments using purified V-ATPase immobilized on streptavidin biosensors via biotinylated membrane scaffold protein (bio-MSP) that was used to reconstitute V-ATPase. V-ATPase-loaded sensors were exposed to Oxr1p, or nucleotides, or combinations thereof, and the release of V₁ was monitored via the time-dependent decrease in BLI signal. Consistent with ATPase activity and negative stain EM analysis (Fig. 5A–D), Oxr1p or ATP alone caused only partial release of V₁ over the duration of the experiment, with their combination reaching almost 70% disassembly within 5 min (Fig. 5E), a rate similar to the one observed in vivo (Dechant et al, 2010). In contrast, the non-hydrolysable ATP analog, AMP-PNP, did not accelerate V₁ release in the presence of Oxr1p, consistent with the previously stated observation that ATP hydrolysis is required for efficient V-ATPase disassembly in the cell. Overall, these results provide strong evidence that whereas either Oxr1p or ATP alone promotes V-ATPase disassembly, the process is much faster and more efficient when Oxr1p and ATP work together.

While we find here that efficient Oxr1p-mediated disassembly of V₁ from purified wild-type V-ATPase requires the presence of ATP, the process appeared much faster with purified vacuoles even in the absence of added nucleotides (Khan et al, 2022). One explanation for this discrepancy is that the molar ratio of Oxr1p to V-ATPase was likely much higher in these experiments, as the ratio was based on total vacuolar protein, of which V-ATPase is only 10–25% based on specific ConA sensitive ATPase activities of vacuoles (~1–2.5 µmol × (min × mg)⁻¹) compared to purified enzyme (~10 µmol × (min × mg)⁻¹) (Appendix Fig. S3). To validate this explanation, and to test whether nucleotide could make the process even more efficient with vacuoles as we observed with purified enzyme, we incubated vacuoles with Oxr1p at ratios of 1:10 or 0.3:10 (w/w) over total vacuolar protein in presence and absence of ATP for 30 min and measured ATPase activities. While incubation with ATP alone caused a slight increase of the activity, a ratio of 1:10 (w/w) Oxr1p over vacuolar protein resulted in an almost complete loss of activity (but not vacuolar V₁, see below) as observed in our previous study (Khan et al, 2022) (Fig. EV2A). However, when Oxr1p was added at lower concentration (0.3:10 (w/w) over vacuolar protein), vacuoles retained close to 50% of the activity (Fig. EV2A). Strikingly, when combining 0.3:10 (w/w) Oxr1p with ATP, we saw an almost complete loss of activity after 30-min incubation (Fig. EV2A), indicating that Oxr1p is equally efficient at disassembling the V-ATPase on vacuoles than for the purified enzyme in lipid nanodiscs.

We previously found that whereas treatment of vacuoles with Oxr1p resulted in almost complete loss of ATPase activity, release of V₁ was incomplete (Khan et al, 2022). Here, using western blot analysis, we expanded on these earlier experiments and found that neither increasing the amount of added Oxr1p nor the incubation time resulted in more V₁ release (Fig. EV2B,C), suggesting that a subset of the V-ATPases were resistant to Oxr1p-induced disassembly. To test whether nucleotides could make enzyme disassembly from vacuoles more efficient, as seen for the purified enzyme (Fig. 5D,E), we incubated vacuoles with a 0.3:10 (w/w) ratio of Oxr1p over vacuolar protein alone or in the presence of ATP or AMP-PNP. The western analysis showed that whereas Oxr1p alone caused only partial release of V₁ subcomplexes from the membrane, in the presence of ATP, but not AMP-PNP, almost all of the V₁ was found in the supernatant fraction (Fig. EV2D). To summarize, ATP enhances Oxr1p’s ability to disassemble the enzyme on vacuoles, mirroring the effect of ATP on Oxr1p’s ability to disassemble the purified enzyme as measured using negative stain EM (Fig. 5C,D) and BLI (Fig. 5E).

Given Oxr1p causes release of wild-type V₁ from V_o, and that wild-type V₁ does not co-purify with Oxr1p (Kitagawa et al, 2008) we wondered, what happens to Oxr1p following V-ATPase disassembly? The endogenous Oxr1p-bound mutant V₁ subcomplex (V₁(C)Oxr1p) resolved by cryoEM was derived from a yeast strain in which V₁ subunit H was deleted (vma13Δ) (Khan et al, 2022) (Figs. 1B and 6A). To determine the fate of Oxr1p in yeast strains expressing wild-type V-ATPase, we performed immunoprecipitation (IP) experiments by pulling down Oxr1p or subunit A (Vma1p) from oxr1^OE and oxr1Δ strain lysates and testing for co-precipitation of subunit A and Oxr1p, respectively. Surprisingly, V₁ and Oxr1p did not Co-IP from the oxr1^OE lysate (Fig. 6B), suggesting that either Oxr1p does not remain bound to wild-type V₁ after enzyme dissociation, or that the V₁-Oxr1p interaction is broken upon preparation of the cytosolic lysate. To resolve this discrepancy, we tested whether recombinant biotinylated Oxr1p (bio-Oxr1p) can pull down purified wild-type V₁ or V₁ that is contained in cytosolic lysate from the oxr1Δ strain. However, in both cases, we were not able to detect an interaction between Oxr1p and wild-type V₁ (Fig. EV3A,B). Prior to the above experiments, we tested bio-Oxr1p’s ability to inhibit V-ATPase on purified vacuoles and found it to be equally effective as the non-biotinylated protein (Appendix Fig. S2). In summary, whereas Oxr1p inhibits the MgATPase activity of mutant V₁ subcomplexes and forms a stable complex with V₁ΔH and C as seen by cryoEM (Khan et al, 2022), above experiments show that Oxr1p does not bind wild-type V₁ under the conditions tested. Next, since the V₁(C)Oxr1p complex solved by cryoEM was isolated from a subunit H deletion strain (vma13Δ) (Khan et al, 2022), we determined whether we could Co-IP Oxr1p and V₁ΔH from vma13Δ lysate. To do so, we deleted OXR1 from the vma13Δ strain (resulting in oxr1Δvma13Δ) and then complemented oxr1Δ with Oxr1p-HA via the pRS316 plasmid to generate strain oxr1^NPvma13Δ. Subsequent experiments using cleared lysate of the oxr1^NPvma13Δ strain showed that whereas a pulldown of V₁ΔH co-precipitated Oxr1p-HA (Fig. 6C), IP of Oxr1p via its C-terminal HA tag failed to pull down V₁ΔH. As mentioned above, Oxr1p’s C-terminus is partially buried in the V₁:Oxr1p interface, and it is possible that the α-HA antibody therefore only binds the excess of Oxr1p-HA that is not attached to V₁ΔH. To verify this suspicion, we performed pulldown and BLI experiments with V₁ΔH and N-terminally tagged Oxr1p (bio-Oxr1p). To get an accurate measure of V₁ΔH’s affinity for Oxr1p, we purified V₁ΔH from the oxr1Δvma13Δ double-deletion strain (hereafter referred to as V₁ΔH^(ΔOxr1)) to avoid contamination with endogenous Oxr1p. V₁ΔH^(ΔOxr1) is active and has a SEC elution profile similar to V₁ΔH purified from the vma13Δ strain (Fig. EV3C). Consistent with our explanation that IP of Oxr1p-HA does not co-precipitate V₁ΔH due to steric hindrance, we find that bio-Oxr1p is able to pull down purified V₁ΔH^(ΔOxr1) as analyzed by Coomassie blue stained SDS-PAGE (Fig. EV3D). The interaction between Oxr1p and V₁ΔH^(ΔOxr1) was further analyzed using biolayer interferometry (BLI). Here, bio-Oxr1p was immobilized on streptavidin sensors and the sensors were then dipped into solutions of V₁ΔH^(ΔOxr1) in the absence and presence of subunit C (Fig. 6D) as subunit C was previously shown to enhance Oxr1p’s ability to inhibit MgATPase activity of mutant V₁ subcomplexes (Khan et al, 2022). From the BLI experiments, we find that bio-Oxr1p binds V₁ΔH^(ΔOxr1) both in absence and presence of subunit C, with K_ds of ~67 ± 33 nM and ~7 ± 1 nM, respectively (Fig. 6D), consistent with our earlier ATPase inhibition assays and the stability of the V₁(C)Oxr1p complex (Khan et al, 2022).

The above experiments indicated that Oxr1p’s ability to bind V₁ is controlled by the presence of subunit H, which is known to be required for autoinhibition of membrane detached wild-type V₁ (Parra et al, 2000). Previous structural studies had shown that enzyme disassembly involves a conformational switch of subunit H’s C-terminal domain (H_CT) from a binding site on the N-terminal domain of V_o subunit a in holo V-ATPase to its autoinhibitory binding site on V₁ (Oot et al, 2016; Vasanthakumar et al, 2022). Since we find that Oxr1p binds V₁ΔH, but not wild-type autoinhibited V₁, we hypothesized that following Oxr1p-induced V-ATPase disassembly, Oxr1p remains bound to V₁ until the complex adopts its autoinhibited conformation. To test this prediction, we incubated purified wild-type vacuoles with bio-Oxr1p to induce V-ATPase disassembly and then isolated the resulting V₁ subcomplex for pulldown (using bio-Oxr1p as bait) and western blot analysis. In support of our hypothesis, we found that following Oxr1p-induced V-ATPase disassembly, Oxr1p remains bound to V₁. Unexpectedly, the experiment showed that both subunits C and H were also part of the disassembled V₁ subcomplex next to Oxr1p (Fig. 6E), an observation that appeared to contradict our earlier observation that Oxr1p does not bind autoinhibited wild-type (H containing) V₁. The presence of subunit C in the membrane detached V₁, however, indicated that this complex must represent a disassembly intermediate, V₁(C,H)Oxr1p, as subunit C is known to dissociate from V₁ during enzyme disassembly in vivo (Liu and Kane, 1996). To resolve this seeming contradiction, we surmised that V₁(C,H)Oxr1p is not (yet) autoinhibited, and therefore still able to hydrolyze ATP. We previously reported that the interaction between immobilized C subunit and an active mutant V₁ subcomplex (V₁H_chim) is disrupted upon ATP hydrolysis (Sharma et al, 2019), and that Oxr1p’s ability to inhibit mutant V₁ subcomplexes (V₁ΔH and V₁H_chim) is C dependent (Khan et al, 2022). This means that if V₁(C,H)Oxr1p is capable of hydrolyzing ATP, the accompanying conformational changes could then facilitate release of subunit C and Oxr1p and allow H_CT to assume its autoinhibitory conformation. To test this prediction, we again incubated wild-type vacuoles with bio-Oxr1p and treated the resulting V₁(C,H)Oxr1p disassembly intermediate with ATP or AMP-PNP before pulldown. The western blot revealed that ATP, but not AMP-PMP, disrupted the interaction between Oxr1p and V₁ (Fig. 6E). Whether subunits H and/or C were also released under these conditions, however, was not clear. To answer this question, we carried out reverse pulldowns with V₁ as bait and found that in the presence of ATP, both Oxr1p and C were released from the disassembly intermediate, but not H (Fig. 6F), consistent with H being required for V₁ autoinhibition. In the recent cryoEM structure of autoinhibited V₁, H_CT is seen in contact with the N-termini of EG2 (Vasanthakumar et al, 2022), a conformation that requires less bending of EG2 than seen in the structure of the V₁(C)Oxr1p complex (Khan et al, 2022) (Appendix Fig. 4A). Since the autoinhibitory conformation of subunit H is mediated by H_CT, this suggests that presence of Oxr1p and C interferes with H_CT binding to V₁. Thus, EG2 can either support binding of Oxr1p + C in V₁(C)Oxr1p (and V₁(C,H)Oxr1p), or H_CT in autoinhibited V₁, but not both, which explains why Oxr1p does not bind autoinhibited V₁. To further support our model, we tested the interaction of immobilized subunit H with V₁ΔH in the absence and presence of Oxr1p and subunit C using BLI. The data shows that binding of V₁ΔH to wild-type H is significantly impaired by the presence of Oxr1p and C (Fig. 6G, left panel). However, when the experiment was carried out using H_chim, the interference was minimal (Fig. 6G, right panel). H_chim, which is not autoinhibitory but supports in vitro assembly of functional V-ATPase (Oot et al, 2016; Sharma et al, 2019), has been proposed to bind V₁ via H_NT only, unlike wild-type H that binds via both H_NT and H_CT (Oot et al, 2016; Sharma et al, 2018; Vasanthakumar et al, 2022) (Fig. 6A). Overall, these results support a model in which Oxr1p-mediated V-ATPase disassembly first produces a V₁ subcomplex that has both Oxr1p and subunit C bound, and that subsequent ATP hydrolysis by this disassembly intermediate allows H_CT to adopt its autoinhibitory conformation, with concomitant release of Oxr1p and subunit C (see also Movie EV1).

Pioneering cryoEM structural work by the Rubinstein lab showed that purified yeast V-ATPase vitrified in the absence of substrate can be sorted into three major classes distinguished by three angular positions of the central rotor separated by 120° (Zhao et al, 2015). The three classes are referred to as rotary states 1–3, with state 1 being the most and state 3 being the least populated. Our recent cryoEM structure of V₁ΔH bound to Oxr1p and subunit C (V₁(C)Oxr1p) showed the complex to be halted in rotary state 1, with Oxr1p interacting with the C-terminal domain of a B subunit of the open catalytic site (Khan et al, 2022). How Oxr1p binds holo V-ATPase, however, is not known. To address this question, we incubated lipid nanodisc reconstituted yeast V-ATPase with a twofold molar excess of Oxr1p and imaged the resulting mixture using cryoEM. We used 3-D classification to obtain two major classes of particles, with the final maps determined at ~8.2 Å resolution. Strikingly, one of the maps showed a prominent density above the C subunit next to EG2 that resembled Oxr1p (Fig. 7A, see the pink dashed circle in the bottom left panel). The second class on the other hand had no significant density at that position (Fig. 7B, see pink dashed circle in bottom right panel). Comparison of the two maps to our recent models of the yeast enzyme in states 1–3 (Khan et al, 2022) showed that Oxr1p-bound V-ATPase is in rotary state 1 (~31% of the dataset) (Fig. 7A), with the second class matching rotary state 2 (~60%) (Fig. 7B) (the remaining ~9% of the particles did not allow unambiguous assignment to either class). This suggests that Oxr1p-induced V-ATPase disassembly starts with the enzyme in state 1, and that subsequent ATP hydrolysis (at least one 120° rotary step) allows the resulting V₁(C,H)Oxr1p disassembly intermediate to adopt the autoinhibited state 2 conformation with concomitant release of Oxr1p and C (see “ATP hydrolysis breaks the Oxr1p-V₁ interaction” above). We also obtained a dataset of V-ATPases vitrified in the absence of Oxr1p. Here, we obtained two major classes, with the corresponding 3-D maps resolved at 7.3 and 7.4 Å resolution, respectively (Appendix Fig. S4B). For the Oxr1p free dataset, the two classes matched rotary states 2 (42%) and 1 (35%), respectively (Fig. 7B) (the remaining 23% of the particles did not align to a particular rotary state). In summary, the analysis shows that Oxr1p-bound holo V-ATPase is in rotary state 1, much like the V₁(C)Oxr1p complex isolated from the vma13Δ strain. Moreover, the observations that Oxr1p selectively binds (and probably disassembles) holoenzymes that are in rotary state 1 (Fig. 7A), and that approximately half of the wild-type complexes purified from vacuoles and vitrified in the presence or absence of Oxr1p are halted in state 2 (Fig. 7B; Appendix Fig. S4B) provide an explanation for why V₁ release from vacuoles is incomplete unless ATP is added to replenish the pool of state 1 enzymes.

Yeast vacuoles were isolated and purified as described previously (Khan et al, 2022; Sharma and Wilkens, 2017). Briefly, yeast cells were harvested at an OD of ~1.0, spheroplasted by zymolyase treatment and lysed with 10-12 strokes in a Dounce homogenizer in buffer A (10 mM MES-Tris pH 6.9, 0.1 mM MgCl₂, 12% Ficoll 400). Following ultracentrifugation (71,000 × g, 4 °C, 40 min), vacuoles were recovered from the top of the tube (floating white layer), resuspended in buffer A and overlaid with buffer B (10 mM MES-Tris pH 6.9, 0.1 mM MgCl₂, 8% Ficoll 400) and ultracentrifuged again. Vacuoles were again collected from the top layer and resuspended in 1.5 mM MES-Tris pH 7, 4.8% glycerol, and either used right away or stored at -80 °C. Of note, specific ATPase activities of purified vacuoles vary significantly depending on yeast strains (wild type with and without affinity tags, mutant strains) and the growth media (YEPD, synthetic complete (SC), synthetic dropout (SD)). Typical specific activities for wild-type in SC were 3.5–4.5 µmol × (min × mg)⁻¹; for wild-type in YEPD, 4–8 µmol × (min × mg)⁻¹; for wild-type with purification tag in YEPD, 1–2.5 µmol × (min × mg)⁻¹.

V-ATPase was purified and reconstituted into lipid nanodiscs as described (Sharma and Wilkens, 2017). Briefly, vacuoles from a yeast strain deleted for subunit G (SF838-5Aα-vma10Δ) and complemented with pRS315 expressing N-terminally FLAG-tagged G were supplemented with protease inhibitors and solubilized in n‐dodecyl β‐d‐maltopyranoside (DDM) at 1.2 mg/mg of vacuolar membrane protein for 1 h at 4 °C. Insoluble material was removed by ultracentrifugation (100,000 × g, 30 min) and the supernatant was mixed with biotinylated membrane scaffold protein (MSP1E3D1) (Sharma and Wilkens, 2017) at 1:50 molar ratio and rotated at 4 °C for 1 h followed by detergent removal with 0.4 g/ml Bio-beads SM2 for 2 h. The sample was then passed over 2 ml α-FLAG affinity resin, and the column washed with 10 column volumes (CV) of TBSE buffer (20 mM Tris pH 7.2, 150 mM NaCl, 0.5 mM EDTA) before eluting with 5 CV of 0.1 mg/ml FLAG peptide in TBSE. Fractions containing V-ATPase were identified by SDS-PAGE, pooled and concentrated using a 100 kDa MWCO centrifugal concentrator.

All yeast strains were generated by transforming either plasmid DNA or PCR-amplified product using standard lithium-acetate procedures as described (Khan et al, 2022). Transformants were selected on synthetic dropout media or drug-containing YEPD plates as appropriate. In case of genomic integration, positive colonies were identified using colony PCR (Phire Plant Direct PCR Master Mix, ThermoScientific, catalog no. F160S), and the PCR products confirmed by DNA sequencing. For deletion of OXR1, a hygromycin cassette was amplified from a vector using Oxr1del FWD and Oxr1del REV primers and the PCR product was transformed into SF838-5Aα (Stevens et al, 1986) and BY4742-vma10Δ-vma13Δ (containing pRS315-FLAG-VMA10) (Diab et al, 2009) strains to generate SF838-5Aα-oxr1Δ (called oxr1Δ) and BY4742-oxr1Δ-vma10Δ-vma13Δ (pRS315-FLAG-VMA10) (called V₁∆H^(∆Oxr1)) strains, respectively. To tag Oxr1p C-terminally with mNeonGreen (mNG), the mNG coding sequence was amplified using Gen-Oxr1CT-mNG F and Gen-Oxr1CT-mNG R primers and the PCR product was transformed into SF838-5Aα to generate strain SF838-5Aα-OXR1-mNG. To generate SF838-5Aα-VMA5-mNG and SF838-5Aα-oxr1Δ-VMA5-mNG strains, the mNG sequence was PCR-amplified using Scarletvma5CTF and mNeonvma5CTR primers and the PCR product was transformed into SF838-5Aα and SF838-5Aα-oxr1Δ strains, respectively. All other strains were generated by transformation with the respective plasmids: (1) oxr1^NP: pRS316-OXR1-HA plasmid was transformed into oxr1Δ; (2) oxr1^OE: YEp352-OXR1-HA plasmid was transformed into oxr1Δ; (3) SF838-5Aα-oxr1Δ-VMA5-mNG strain expressing Oxr1p (oxr1^NPC-mNG): pRS316-OXR1-HA plasmid was transformed into SF838-5Aα-oxr1Δ-VMA5-mNG; (4) V₁∆H expressing C-terminally HA-tagged Oxr1p (oxr1^NPvma13Δ): pRS316-OXR1-HA plasmid was transformed into V₁∆H^(∆Oxr1); (5) mNG-Oxr1p: pRS316-mNG-OXR1 plasmid was transformed into oxr1Δ. Genotypes of yeast strains and primer sequences are provided in Appendix Tables S1 and S2, respectively.

All plasmids were constructed using restriction-free cloning (Bond and Naus, 2012). A pair of hybrid primers were used to amplify the gene of interest and the amplified product was then used as a megaprimer to insert the gene into the target vector. The insertion of the gene was then confirmed by restriction digestion and DNA sequencing. C-terminally HA-tagged Oxr1p was first generated in a pMECS vector in two steps. First, the Oxr1p coding sequence (excluding the stop codon) plus its 5′ UTR (333 nucleotides) was amplified from the yeast genome using Oxr1UTR_pMECS_FWD and Oxr1UTR_HA_pMECS_REV primers and placed at the N-terminus of the HA epitope coding sequence present in the pMECS vector, with the Oxr1p coding sequence and the HA tag separated by a linker peptide (SSAAA). Second, the 3′ UTR (333 nucleotides) of Oxr1p was amplified from the yeast genome using Oxr1-3UTR F and Oxr1_3UTR R primers and placed after the stop codon present at the C-terminus of HA to construct a pMECS-OXR1-HA plasmid. From the pMECS-OXR1-HA plasmid, OXR1-HA with or without OXR1’s UTRs were transferred to pRS316 and YEp352 (2 µ) vectors using Oxr1UTR_pRS316 F and Oxr1UTR_pRS316 R, and Oxr1_HA_YEp352_F and Oxr1_HA_YEp352_R primers, respectively, to generate pRS316-OXR1-HA and YEp352-OXR1-HA vectors. The pRS316-mNG-OXR1 plasmid was generated in two steps: first, plasmid pRS316-OXR1-HA was used to construct plasmid pRS316-mCherry-OXR1, and then mCherry was replaced with mNG. In plasmid pRS316-OXR1-HA, the Oxr1p sequence was replaced with mCherry using primers mCher_OxUTR_F and mCher_OxUTR_R, and then HA was replaced with Oxr1p sequence using primers Oxr1_mCherry_F and Oxr1_mCherry_R to generate plasmid pRS316-mCherry-OXR1. The mNG coding sequence was amplified using primers mCher_OxUTR_F and mNG_Oxr1_R, and the PCR-amplified product was then used as a megaprimer to replace mCherry in the vector to construct pRS316-mNG-OXR1. The pET28a-OXR1-HA and pET28a-BAP-OXR1 (BAP= biotin acceptor peptide or Avi tag) plasmids were constructed using a modified pET28a vector (pET28a-BAP-MSP) containing a 7x His, Avi tag and Prescission protease cleavage site (Sharma and Wilkens, 2017). The OXR1-HA coding sequence was PCR-amplified from the pMECS-OXR1-HA vector using Oxr1_ETF and Oxr1_HA_ETR primers, and the PCR product was then used to replace BAP-MSP in the pET28a vector to construct the pET28a-OXR1-HA plasmid. To construct the pET28a-BAP-OXR1 plasmid, Oxr1p’s coding sequence was amplified from yeast genomic DNA using Oxr1_pET28a_F and Oxr1_pET28a_R primers and used as a megaprimer for insertion into the modified pET28a vector. Primer sequences are provided in Appendix Table S2.

Wild-type and mutant V₁ subcomplexes were purified as described (Khan et al, 2022). Briefly, cells expressing N-terminally FLAG-tagged subunit G were grown to ~3.5 OD in YEPD or synthetic dropout (SD-Leu) media, harvested and resuspended in TBSE buffer before storing at –80 °C for later use. Frozen cells were thawed, supplemented with 5 mM β-mercaptoethanol, protease inhibitors (leupeptin, pepstatin, and PMSF) and lysed by passing through a microfluidizer at 18,000 PSI. The lysate was cleared by two subsequent centrifugations at 4000 × g and 13,000 × g, respectively. The cleared lysate was then passed over 5 ml α-FLAG affinity resin pre-equilibrated with TBSE, washed in the same buffer, and eluted with FLAG peptide (0.1 mg/ml in TBSE). Fractions containing V₁ subcomplexes were then pooled, concentrated and applied to a Superose 6 Increase HiScale (16 mm × 400 mm) size-exclusion chromatography (SEC) column attached to an ÄKTA FPLC. Fractions were analyzed by SDS-PAGE and V₁ containing fractions were pooled and concentrated using a 50 kDa MWCO centrifugal concentrator.

In total, 50–75 µg vacuolar protein was incubated with or without recombinant Oxr1p for 30 min at room temperature (a portion kept aside served as input on the blot). Vacuoles were then resuspended in buffer C (5 mM MES-Tris pH 6.9, 2.5 mM MgCl₂, 12.5 mM KCl) and ultracentrifuged at 37,000 × g for 20 min at 4 °C. The pellet fractions from each sample were resuspended in an equal amount of hot cracking buffer (8 M Urea, 5% SDS, 1 mM EDTA, 50 mM Tris-HCl pH 6.8, 10 mM DTT) (pellet). The supernatant was TCA precipitated and washed in cold acetone before resuspending in an equal amount of hot cracking buffer (sup). Samples were then heat treated (at 95 °C for 10 min for subunit A and Oxr1p, or at 65 °C for 15 min for subunits a and C) and separated by SDS-PAGE before analysis by western blot.

Analysis of the assembly state of holo V-ATPase by negative stain EM was carried out as described (Khan et al, 2022). Briefly, 5 µl protein sample at a concentration of ~30 µg/ml from each of the four conditions (V₁V_o only, V₁V_o + Oxr1p, V₁V_o + 4 mM MgATP, and V₁V_o + Oxr1p + 4 mM MgATP) were applied to glow discharged carbon-coated copper grids. After 1 min of incubation, grids were briefly washed with water and then stained with 5 µl of 1% uranyl acetate solution for another minute. Micrographs were collected in SUNY Upstate Medical University’s TEM core facility on a JEOL JEM-1400 transmission electron microscope equipped with Gatan Orius SC1000 CCD camera at ×200,000 magnification and 80 keV. Images were manually inspected and assembled and disassembled particles were counted from 20 micrographs for each condition.

For cryoEM analysis, 2.5 µl protein samples at ~1–2 mg/ml (V-ATPase in lipid nanodisc with or without a twofold molar excess of Oxr1p) were vitrified on AuFlat 1.2/1.3 grids using a Leica EM GP2 with settings 6 °C, 85% humidity, 5 s blot time. Grids were mounted in a Gatan 914 side entry cryo holder and movies containing 25 frames (2 e^-/frame) were recorded at 40,000× on a JEOL JEM-2100FS equipped with Gatan OneView and K2 Summit cameras using SerialEM (Schorb et al, 2019). Movies were combined using MotionCorr as implemented in Relion4 (Kimanius et al, 2021), and the CTF was corrected using ctffind-4.1.14 (Rohou and Grigorieff, 2015). Motion-corrected micrographs of the V-ATPase only sample were then imported into EMAN2 (Tang et al, 2007) and particles were picked in e2boxer.py using the neural net training option. Particle coordinates were then transferred to Relion4 and particle images were extracted as 480 × 480 pixel boxes (0.96 Å/pixel) and binned 3× to give 160 × 160 pixel boxes (2.88 Å/pixel). Extracted particles were subjected to several rounds of 2-D classification to remove bad particles. Good 2-D averages/classes were then used for training and particle picking in Topaz (Bepler et al, 2019) as implemented in Relion4. Particles were extracted using a FOM of -3 and bad particles were removed by 2-D classification. The dataset was then 3-D classified using EMD-31538 (Khan et al, 2022) as a starting reference, and 3-D classes were refined using Refine3D. Particles from the images of the Oxr1p-containing sample were picked in Topaz using the training model obtained from the V-ATPase (no Oxr1p) dataset and analyzed using the same strategy. Maps and models were displayed in UCSF Chimera 1.17.3 and ChimeraX 1.6.1 (Pettersen et al, 2004) for image rendering.

Co-immunoprecipitation of V₁ and Oxr1p was carried out as described (Jaskolka and Kane, 2020). Briefly, ~100 OD₆₀₀ of cells were harvested from an exponentially growing culture and stored at −80 °C until use. Cells were lysed using glass beads and cytosolic lysate collected by ultracentrifugation as described in “Pulldown”. After measuring the protein concentration, 0.25–0.30 mg protein was TCA precipitated to serve as input, and 2.5–3.0 mg protein was used for Co-IP. The cytosolic lysate was first pre-cleared using protein A sepharose beads followed by incubation with either α-Vma1p (8B1F3 mAb) or α-HA antibody (ThermoFisher, catalog no. 26183) for 2 h at 4 °C while mixing gently. The antibody–lysate mixture was then incubated overnight at 4 °C while rotating in the presence of sepharose protein A beads. Centrifuged beads were then washed, and proteins were eluted using a hot cracking buffer at 95 °C for 10 min. Eluted proteins were analyzed by SDS-PAGE followed by western blot.

Whole-cell lysate was prepared as described (Zhang et al, 2011) with the following modifications. Briefly, exponentially growing yeast cells were harvested and stored at −80 °C until use. Cells were thawed, resuspended in 2 M lithium acetate, and incubated for 5 min on ice before centrifugation. Pellets were resuspended in 0.4 M NaOH, and incubated for 5 min on ice again. Pellets were then resuspended in hot cracking buffer and incubated at 95 °C for 5 min. Cellular debris was pelleted by centrifugation at 16,000 × g for 10 min, and supernatants containing whole-cell extracts were used immediately or stored at −80 °C until use.

BLI experiments were performed using an Octet RED384 system as described (Khan et al, 2022; Sharma and Wilkens, 2017). All experiments were carried out in black, flat-bottom 96-well plates with wells containing 200 µl buffer with or without protein samples at the specified concentrations. Before dipping into the sample wells, all biosensors were pre-equilibrated in buffer for at least 10 min. Experiments were carried out at 23 °C with plates shaking at 1000 rpm, and measurements were taken at a rate of 5 s⁻¹. In all cases, individual controls for each condition were subtracted from the experimental data prior to analysis by the Octet Red HT data analysis v10 software.

ATPase activity assays of vacuoles and purified V-ATPase were carried out as described (Khan et al, 2022). Briefly, 1 ml assays (50 mM HEPES pH 7.5, 25 mM KCl, 0.5 mM NADH, 2 mM phosphoenolpyruvate, 5 mM ATP, 30 units each of lactate dehydrogenase and pyruvate kinase) were prewarmed at 37 °C for 10 min and supplemented with 4 mM MgCl₂. The reaction was initiated by adding 2.5–10 µg protein into the assay, and the drop in absorbance at 340 nm was monitored in a Varian Cary 100 Bio UV–Visible Spectrometer in kinetics mode. After establishing a linear rate of ATP hydrolysis, 200 nM of the specific V-ATPase inhibitor Concanamycin A (ConA) was added into the assay. The assay was maintained at 37 °C using a temperature-controlled cuvette holder, and the rate of ATP hydrolysis was determined using the Kinetics Application as implemented in the Cary WinUV software package version 3.

Oxr1p, HA-tagged Oxr1p (Oxr1p-HA) and biotinylated Oxr1p (bio-Oxr1p) were expressed and purified as described for Oxr1p (Khan et al, 2022). Briefly, Escherichia coli BL21DE3 cells expressing N-terminally 7×His tagged proteins were harvested, lysed, sonicated and cleared of the debris. Lysates were then passed over a Ni-NTA column before being further purified by a Superdex-75 SEC column (16 mm × 500 mm). MBP-H_WT, MBP-H_chim, and subunit C were expressed and purified using established protocols (Oot and Wilkens, 2010; Sharma et al, 2019; Sharma et al, 2018). Briefly, E. coli Rosetta2 cells expressing N-terminally MBP-tagged H_WT, H_chim and C were harvested, lysed, sonicated, and cleared of the debris. MBP-tagged proteins were then captured by amylose resin and eluted in 10 mM maltose before being passed over a Superdex-75 SEC column (16 mm × 500 mm) to polish MBP-H_WT and MBP-H_chim. For purification of subunit C, the MBP tag was cleaved using Prescission protease, and the sample was dialyzed to adjust the pH and salt concentration for subsequent removal of MBP using a diethylaminoethyl (DEAE) sepharose column. The flow-through of the column was dialyzed and concentrated before being further purified by a Superdex-75 SEC column (16 mm × 500 mm). Biotinylated membrane scaffold protein (MSP1E3D1) was expressed and purified from E. coli BL21DE3 cells as described (Sharma and Wilkens, 2017). Briefly, cells were grown in the presence of D-biotin, harvested, enzymatically lysed, and sonicated. Cellular debris and large organelles were removed by centrifugation and protein was captured by Ni-NTA resin. Eluted protein from the column was dialyzed, concentrated and stored at −80 °C until use.

Western blot analysis was performed as described (Khan et al, 2022). Briefly, proteins were separated by SDS-PAGE and transferred to a low-fluorescence polyvinylidene fluoride (LF-PVDF) membrane. Blots were blocked for 1 h in TBST (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.05% Tween-20) plus 5% milk before incubating overnight with primary antibodies at 4 °C with gentle agitation. After washing three times with TBST buffer, goat α-mouse IgG Alexa Fluor Plus 488 (ThermoFisher, catalog no. A32723) or goat α-rabbit IgG H&L (IRDye® 800CW) (Abcam, catalog no. ab216773) secondary antibody was added at a final dilution of 1:2000 or 1:10,000, respectively, and incubated for 60-90 min at room temperature. The blot was washed again, dried, and imaged in a Bio-Rad ChemiDoc MP Imaging System. All images were analyzed using the Bio-Rad Image Lab software. Primary antibodies used in this study include- mouse monoclonal IgG penta-His antibody (Qiagen, catalog no. 34660), mouse α-Vma1p (8B1F3), mouse α-Vph1p (10D7), mouse α-Vma5p (7A2), rabbit α-Vma13p (polyclonal), mouse α-HA (ThermoFisher, catalog no. 26183) and mouse α-GAPDH (ThermoFisher, catalog no. MA5-15738).

The growth phenotype of OXR1 mutant strains was determined using a spot test as described (Khan et al, 2022). Briefly, yeast cells were grown to an OD₆₀₀ of ~1 in liquid synthetic complete (SC) or dropout (SD-URA) media. ~1 OD₆₀₀ cells were harvested, washed, and resuspended in 1 ml of sterile water. Cells were then serially diluted up to 10⁻⁴. In total, 6 µl from each dilution was spotted on YEPD, YEPD pH 5 and YEPD pH 7.5 + 60 mM CaCl₂ plates, and incubated at 25, 30, and 37 °C for 3–4 days before being imaged.

The localization of mNeonGreen (mNG) tagged Oxr1p and subunit C (Vma5p) in different growth conditions was carried out as described (Jaskolka and Kane, 2020). Briefly, cells were grown to exponential phase in SC or SD-URA media, stained with 8 μM FM4-64 (ThermoFisher, catalog no. T3166) in YEPD before recovery in SC or SD-URA media. Fluorescence images of cells were then collected using a Zeiss fluorescence microscope in DIC, TexasRed and GFP channels under (i) glucose-fed (+) condition, (ii) 15 min (30, 45, 60, 90, and 120 min for the time course study) after glucose starvation (−), and (iii) 15 min after glucose re-addition (−/+) to the starved cells. Images were then processed in Fiji ImageJ (Schindelin et al, 2012) and Adobe Photoshop. Pearson’s coefficient was calculated using the Fiji ImageJ plugin “Just Another Colocalization Plugin” (JACoP) (Bolte and Cordelieres, 2006). To determine the cellular localization of mNG-tagged Oxr1p, exponentially growing cells were stained with MitoTracker™ Red CMXRos dye (ThermoFisher, catalog no. M7512) for 10 min and washed twice with SC media before collecting fluorescence images using DIC, TexasRed, and GFP channels. Overlay images were created by stacking green and red images at 50% opacity over the DIC image, then brightness was increased to 50%.