Tender Love and Disassembly: How a TLDc domain protein breaks the V-ATPase

Abstract

Vacuolar ATPases (V-ATPases, V₁V_o-ATPases) are rotary motor proton pumps that acidify intracellular compartments, and, when localized to the plasma membrane, the extracellular space. V-ATPase is regulated by a unique process referred to as reversible disassembly, wherein V₁-ATPase disengages from V_o proton channel in response to diverse environmental signals. Whereas the disassembly step of this process is ATP dependent, the (re)assembly step is not, but requires the action of a heterotrimeric chaperone referred to as the RAVE complex. Recently, an alternative pathway of holoenzyme disassembly was discovered that involves binding of Oxidation Resistance 1 (Oxr1p), a poorly characterized protein implicated in oxidative stress response. Unlike conventional reversible disassembly, which depends on enzyme activity, Oxr1p induced dissociation can occur in absence of ATP. Yeast Oxr1p belongs to the family of TLDc domain containing proteins that are conserved from yeast to mammals, and have been implicated in V-ATPase function in a variety of tissues. This brief perspective summarizes what we know about the molecular mechanisms governing both reversible (ATP dependent) and Oxr1p driven (ATP independent) V-ATPase dissociation into autoinhibited V₁ and V_o subcomplexes.

Graphical Abstract

The vacuolar H⁺-ATPase is regulated by reversible disassembly into autoinhibited V₁-ATPase and V_o proton channel subcomplexes. Here we discuss possible physiological roles of our recent discovery that disassembly of the yeast V-ATPase is induced by binding of Oxr1p, a TLDc domain containing protein implied in the protection against oxidative stress.

Introduction

The eukaryotic vacuolar H⁺-ATPase (V-ATPase; V₁V_o-ATPase) is a ubiquitous proton pump found on the endomembrane system of virtually all eukaryotic cells where the enzyme functions to acidify the lumen of subcellular compartments.^[1–5] In polarized cells such as renal α intercalated cells or bone lining osteoclasts, V-ATPases can also be localized to the plasma membrane to acidify the extracellular milieu. V-ATPase belongs to the family of rotary motor ATPases, which also includes the F-ATP synthases found in mitochondria, bacteria, and chloroplasts, the archaeal A-ATPases/synthases, and various A/V-like ATPases found mainly in eubacteria.^[6–8] Depending on the type of motor, its location, and the organism, rotary ATPases either function as ATP synthases that are powered by ion (sodium or proton) gradients generated during e.g. oxidation of food stuffs, or as ion-pumping ATPases, whose function is to establish and maintain transmembrane ion-motive force for secondary transport processes. Despite their diversity in function, all rotary ATPases share a common architecture and mechanism of energy coupling: synthesis or hydrolysis of ATP on the membrane extrinsic F₁-, A₁- or V₁-ATPases is coupled to the translocation of ions across the membrane integral F_o, A_o or V_o ion channels via rotation of a central rotor subcomplex. Most rotary ATPases are reversible because they can be driven by an ion gradient to synthesize ATP, or they can hydrolyze ATP to pump ions across the lipid bilayer. The eukaryotic V-ATPase, however, only operates in the direction of proton pumping under physiological conditions (Figure 1A).

Figure 1.

The eukaryotic vacuolar H⁺-ATPase and regulation by disengagement of V₁-ATPase and V_o proton channel. (A) Schematic of the architecture and mechanism of the yeast V-ATPase. (B) Regulation by reversible disassembly (left side) and Oxr1p mediated disassembly (right side).

V-ATPase is regulated by a unique process referred to as reversible disassembly, a mechanism that has so far not been found for any other rotary ATPase family member. First observed in baker’s yeast and insects, V-ATPase reversible disassembly is a mode of regulation wherein the holoenzyme dissociates into V₁ and V_o,^[9,10] with both subcomplexes becoming autoinhibited^[11–13] (Figure 1B, left side). In yeast, for example, reversible disassembly occurs in response to changes in nutrient availability or extracellular pH.^[14,15] As V-ATPases are major consumers of ATP, regulated disassembly into autoinhibited V₁ and V_o is thought to preserve both energy and transmembrane pH gradients. Evidence for reversible disassembly as a regulatory mechanism has also been found in higher organisms, including in human cells.^[16–19] Here, the process is fine-tuned to the need for proton pumping in specific tasks such as maturation of dendritic cells,^[20] TOR signaling,^[21,22] neurotransmitter loading into synaptic vesicles,^[23] and nutrient sensing.^[24] While the process of reversible disassembly has been extensively characterized on the cellular level in yeast,^[25] our understanding of the molecular events that trigger the disassembly or reassembly on the level of the enzyme is only now beginning to emerge.^[26,27] From these studies, we know that while disassembly requires active enzyme and intact microtubules,^[28,29] (re)assembly is dependent on the heterotrimeric chaperone complex, RAVE.^[30,31] More recently, a novel factor capable of inducing enzyme disassembly was discovered in yeast. Here, binding of Oxidation resistance 1 protein (Oxr1p), a poorly characterized protein that has been associated with resistance to oxidative stress, was shown to cause disassembly of the V-ATPase in vitro, with concomitant formation of a V₁-ATPase complex bound to Oxr1p^[32] (Figure 1B, right side). Importantly, unlike in vivo reversible disassembly, which requires active enzyme^[14], the Oxr1p-induced disengagement of V₁ from V_o in vitro was found to occur without added ATP.^[32] Yeast Oxr1p is a 273-residue globular protein that contains a C-terminal TLDc domain (TLDc stands for Tre2/Bub2/Cdc16 (TBC), lysin motif (LysM), Domain catalytic), which has several homologs in mammals, including nuclear receptor coactivator 7 (NCOA7), OXR1, TLDC2, TBC1D24, mammalian enhancer of AKT1–7 (mEAK7), and interferon induced protein 44 (IFI44).^[33,34] All of the mammalian TLDc domain-containing proteins except IFI44 have now been shown to directly interact with V-ATPase,^[35–39] and several have also been linked to V-ATPase function.^[40,41] In the following sections, we will briefly review the structure and mechanism of the V-ATPase followed by a summary of our current knowledge of the mechanisms of reversible and Oxr1p induced disassembly of the yeast V-ATPase.

The eukaryotic V-ATPase

Unlike related ATP synthases and bacterial rotary ion pumps, eukaryotic V-ATPase reversibly disassembles into V₁-ATPase and V_o proton turbine in vivo, with the degree of assembly or disassembly governed by the physiological state of the cell.^[1,2] Regulated enzyme dissociation is a unique property of the eukaryotic V-ATPase and while certain structurally related bacterial rotary ATPases are sometimes referred to as “V-ATPases”, such as, for example, the enzyme from Enterococcus hirae,^[42,43] this article uses V-ATPase strictly in the context of the eukaryotic enzyme. Early biochemical studies of eukaryotic V-ATPases were conducted with enzymes from bovine brain, chromaffin granules, and plants such as lemon and carrot.^[44] However, due to ease of genetic manipulation and generation of biomass for enzyme purification, baker’s yeast (Saccharomyces cerevisiae) has emerged as a powerful model system for the study of the highly conserved eukaryotic V-ATPase. Moreover, whereas complete loss of V-ATPase in higher organisms is embryonic lethal,^[45] lack of a functional enzyme in yeast results in a conditional phenotype referred to as Vma-, as the cells can grow at acidic but not elevated pH in presence of calcium or zinc ions.^[46] Early studies of the structure of the holoenzyme, its V₁ and V_o subcomplexes, as well as individual subunits and subunit complexes were conducted using negative stain electron microscopy,^[47–49] X-ray crystallography,^[50–54] and NMR spectroscopy.^[55] More recent advances in cryo electron microscopy (cryoEM) have provided near-atomic or secondary structure resolution models of the V-ATPases from a variety of organisms, including the enzymes from yeast,^{[27,32,56–58]} rat^[59] and bovine brain,^[60] pig kidney,^[39] human cell culture,^[38,61] and lemon.^[62] Pioneering cryoEM structural work by the Rubinstein lab resolved populations of purified yeast V-ATPase that showed the central rotor subcomplex halted in three angular positions each 120° apart, with the three populations referred to as rotary states 1–3.^[56] The combination of the earlier biochemical and structural studies and the more recent cryoEM data has provided a wealth of information on the structure and mechanism of this highly dynamic molecular machine.

Structure of the V-ATPase

The cryoEM structure of lipid nanodisc reconstituted yeast V-ATPase has been determined at an overall resolution of ~4.2 Å for rotary state 1 (Figure 2A, left panel).^[32] The cytosolic headpiece is called the V₁-ATPase, or V₁, with the membrane integral proton turbine referred to as V_o proton channel, or V_o. The subunit composition of V₁ is A₃B₃CDE₃FG₃H, and V_o is made of the single copy subunits adef and ten highly hydrophobic proteolipids (c,c’,c”) that are arranged in a ring (c-ring; in yeast c₈,c’c” and in mammals c₉c”).^[6,63,64] Yeast V_o contains an additional polypeptide, the C-terminal transmembrane (TM) ⍺ helix of the assembly factor Voa1p (Figure 2A, middle panel).^[58] Two additional polypeptides are found in mammalian V_o: Ac45 (ATP6AP1) and pro-renin receptor (ATP6AP2), both anchored to the enzyme via their C-terminal TM ⍺ helices within the central cavity of the c-ring.^[59,61] Ac45 contains a glycosylated globular domain with b-prism fold that protrudes from the c-ring towards the lumen or extracellular space.^[61] While Ac45 has been implicated in assembly and trafficking of the complex in mammals,^[65] the role of the pro-renin receptor is unclear. In yeast, the C-terminal TM ⍺ helix of Voa1p occupies an equivalent position to that observed for mammalian Ac45. However, Voa1 is not considered a core subunit of the yeast enzyme as it is not essential for assembly or function.^[66] The catalytic hexamer of V₁ is composed of three subunit AB heterodimers that are arranged around a central cavity, with three catalytic ATP binding sites at alternating AB interfaces (Figure 2B). The central cavity of the catalytic hexamer is filled by part of the D subunit, which, together with subunit F forms the central rotor of V₁ (Figure 2A, middle panel). The bottom of the DF subcomplex protrudes by ~45 Å from the catalytic hexamer to bind subunit d of V_o, thereby coupling V₁’s central rotor to V_o’s c-ring (Figure 2A, middle panel). Each of the ten c-ring proteolipids contributes one essential glutamic acid residue that is located mid-membrane facing the lipid bilayer (Figure 2C). Subunit a of the V_o contains an N-terminal cytosolic (a_NT) and a membrane integral C-terminal domain (a_CT) (Figure 2A, right panel). a_CT features two aqueous half-channels, one for proton entry from the cytosol, and one for proton release into the lumen (or extracellular space for V-ATPases found in the PM of higher eukaryotes) (Figure 2C). a_CT’s interface to the c-ring is constituted by two tilted TM α helices that contain several polar amino acids, including two essential arginine residues. Besides the DF:d connection, V₁ and V_o are linked by three peripheral stators that are each constituted by a parallel heterodimer of E and G subunits (EG1–3) (Figure 2A, left panel), which are engaged in a right-handed coiled coil for much of their length starting from their N-termini.^[52] EG’s C-terminal globular domains are linked to the N-terminal domains of the B subunits at the top of V₁, with their N-termini connected to a_NT and subunit C. C and a_NT together form the collar at the V₁-V_o interface, with each having two globular domains separated by a coiled coil. a_NT is divided into proximal and distal domains (proximal is relative to a_CT), and C is organized into head (C_head) and foot (C_foot), with C_foot bound to a_NT’s distal domain (Figure 2A, right panel). Peripheral stators EG1 and EG2 link subunit a to the catalytic hexamer through a_NT’s proximal and distal domains, respectively, with EG3 connected to the collar by way of C_head. The H subunit is folded in an N-terminal α helical solenoid (H_NT), and a C-terminal α helical globular domain (H_CT) (Figure 2A, right panel). H_NT is bound to the middle part of EG1, and H_CT rests on the coiled coil of a_NT. The e and f subunits of the V_o each contain two α helical TMS that are packed against a_CT facing the lipid bilayer (Figure 2A, right panel).

Figure 2.

Structure of the yeast vacuolar ATPase in lipid nanodisc. (A) Left panel: CryoEM structure of lipid nanodisc yeast reconstituted V-ATPase in rotary state 1^[32] (PDBID 7FDA). Middle panel: V-ATPase central rotor subcomplex. Two of the c subunits are not shown to allow a view inside the c-ring cylinder. Right panel: V-ATPase stator subcomplex. (B) Cross section of states 1–3 at the level of the phosphate binding loops (P-loops) of the V₁ A subunits as indicated with the dashed line in (A), left panel. The blue arrowheads point to the open catalytic sites, with the P-loops highlighted in blue. The closed and half-closed catalytic sites are indicated by black arrowheads. (C) Cross section of the V-ATPase V_o at the level indicated by the dashed line in (A), left panel. For details, see text.

V-ATPase subunit isoforms –

A complicating aspect of V-ATPase structure and function is the presence of subunit isoforms. In yeast, the 100 kDa a subunit of the V_o is expressed as two isoforms, Vph1p and Stv1p.^[67] The Stv1p and Vph1p polypeptides are ~50% identical and found in complexes retained in the Golgi or trafficked to the vacuole, respectively, with the signal peptide responsible for the differential trafficking found in the isoforms’ N-terminal domains.^[68] Biochemical experiments in yeast revealed that only the Vph1p containing V-ATPases in the vacuole are subject to reversible disassembly, but not the Stv1p containing enzymes in the Golgi. However, when Stv1p is overexpressed and pushed to the vacuole in a VPH1 deletion strain, some reversible disassembly is also observed for the Stv1p containing enzymes,^[69] indicating that the propensity of a given V-ATPase to undergo reversible disassembly may depend more on the biological membrane the enzyme is residing on than on isoform content. In higher organisms, a little more than half of the subunits are expressed as multiple isoforms, including the four isoforms of subunit a found in mammals (a1–4).^[70] While tissue enrichment is observed for some isoforms (such as a4 and B1 in the kidney, epididymis, and the olfactory epithelium), most are ubiquitously expressed.^[2] The synaptic vesicle V-ATPase, which has been shown to undergo reversible disassembly,^[23] contains subunit isoforms that are ubiquitous in nature. However, whether different isoform combinations lead to differential propensities for regulated enzyme disassembly in mammals is not known.

Mechanism of ATP hydrolysis driven proton pumping

V-ATPases couple hydrolysis of ATP on the membrane extrinsic V₁ to ion translocation across the membrane integral V_o by way of a central rotor constituted by V₁ and V_o subunits. V₁-ATPase is a stepping motor; ATP hydrolysis at the three catalytic AB interfaces drives rotation of the DF heterodimer in steps of 120°, resulting in rotary states 1–3 as resolved in cryoEM structures of the yeast enzyme^[32,56] (Figure 2B). The DF heterodimer, through its interaction with subunit d, then rotates the c-ring past a_CT’s two almost horizontal TM a helices, which contain several polar residues that are critical for proton pumping.^[71] The biochemical and cryoEM structural data suggest that protons are guided from the cytosolic half-channel towards a conserved glutamic acid residue (Glu721 in yeast Vph1p) from where they are transferred to a deprotonated glutamate on the c-ring (Figure 2C). Once protonated, the c-ring glutamic acid residue can now exit the a_CT:c-ring interface and enter the lipid bilayer, allowing another c subunit to exit the bilayer into the a_CT:c-ring interface and transfer its proton to a conserved glutamate (Glu789) at the entry site to a_CT’s lumen half-channel. The now deprotonated c-ring glutamate is then guided by a conserved histidine (His796) towards the two conserved arginine residues (Arg735,799) to form a salt bridge. Salt bridge formation results in an energy minimum, and for the c-ring to continue the rotation requires the driving force of the V₁-ATPase. Since the V-ATPase has three catalytic sites and ten proteolipids, this means that on average 3.3 protons are pumped for each ATP molecule hydrolyzed.

Structural features of autoinhibited V₁

Whereas V₁ is a highly active ATPase when coupled to V_o (~300 ATP × s⁻¹),^[72] V₁ that is dissociated from the membrane no longer hydrolyzes MgATP with a measurable rate.^[12] Biochemical experiments in yeast showed that subunit H is responsible for inhibiting V₁’s MgATPase activity.^[12] However, while V₁ purified from a yeast strain from which H was deleted (V₁ΔH) readily hydrolyzes MgATP, the activity quickly decays due to retention of inhibitory MgADP in a catalytic site¹.^[12,53] MgADP inhibition, which is a conserved feature of the catalytic headpieces of rotary ATPases, is a stable, off-catalytic pathway conformation that is distinct from competitive product (ADP) inhibition. To obtain a more detailed picture of the mechanism of autoinhibition, we obtained a crystal structure of V₁ purified from a yeast strain deleted for the C subunit (V₁ΔC).^[53] The structure showed that autoinhibited V₁ is halted in rotary state 2, with H_CT rotated from its position on a_NT to bind the rotor subunit D and the C-terminal domain of one of the B subunits. The structure thus suggested that autoinhibition is due to interfering with the rotation of the central rotor, reminiscent of the mode of inhibition by the C-terminal α helical domain of the ε subunit in the bacterial F-ATPase.^[73] Supporting biochemical data showed that while V₁ΔH is essentially nucleotide free, autoinhibited V₁ has at least one tightly bound ADP in a catalytic site. Taken together, this indicated that the H_CT:V₁ interaction also served to stabilize inhibitory MgADP in a catalytic site. The interaction of H_CT and D was seen to be mediated by a loop in H_CT that is not conserved in higher organisms including mammals. Indeed, replacing yeast H_CT with that from the human enzyme to generate a chimeric H subunit (H_chim, containing yeast H_NT and human H_CT) restored V₁’s ability to hydrolyze MgATP.^[53] Moreover, since H_CT’s binding site on a_NT in the holoenzyme is conserved, H_chim was not only able to complement deletion of H in yeast,^[53] but replacing wild type H with H_chim also allowed in vitro assembly of holo V-ATPase from V₁ and V_o, a process that does not occur at a significant rate with wild type V₁.^[74]

However, the interpretation of H_CT’s binding site as seen in the crystal structure of (V₁ΔC) was recently disputed based on single particle cryoEM structures of wild type V₁ and V₁ΔC, which showed H_CT in contact with the N-termini of EG2^[27] (Figure 3A). Here, the same site in H_CT that is bound to a_NT in V₁V_o (Figure 3B) is seen interacting with EG2’s N-termini in membrane detached and autoinhibited V₁ (Figure 3A, see dashed circle)². The authors of that study suggested that this arrangement of H_CT seen in the cryoEM structure is responsible for autoinhibition, an interpretation supported by mutagenesis of a residue in the H_CT:EG2 interface that restored V₁’s MgATPase activity. On the other hand, yet another H_CT conformation can be seen in the more recent cryoEM structures of citrus V-ATPase that resembled the conformation as seen in the crystal structure of V₁ΔC,^[62] suggesting that H_CT is able to adopt and sample multiple conformations and binding sites. More work will be needed to understand the physiological significance of the different conformations of H_CT and their roles in V-ATPase regulation.

Figure 3.

Structure of autoinhibited yeast V₁-ATPase. Upper panel: CryoEM structure of autoinhibited V₁ in state 2 (PDBID 7TMM),^[27] highlighting the interaction between H_CT and the N-termini of EG2. Lower panel: Region of the cryoEM structure of nanodisc bound holo V-ATPase in state 2 (PDBID 7FDB),^[32] highlighting the binding site of H_CT on a_NT.

Structural features of autoinhibited V_o

The structure of the isolated V_o has been determined by cryoEM,^[57,58] with the highest resolution structure obtained for the lipid nanodisc reconstituted complex.^[75] The structures show that autoinhibited V_o is halted in rotary state 3, with the essential glutamates of c” and the neighboring c subunit forming salt bridges with the two conserved arginines in a_CT’s tilted TM α helices. The overall structure of isolated V_o resembles the structure of the V_o in the state 3 holoenzyme^[32,56] (Figure 4A, left panel), except for the conformation of a_NT. In the holoenzyme, a_NT is seen at a peripheral position where it binds C_foot and EG2 (Figure 4B). Once V₁ disengages from V_o, a_NT moves towards the central rotor to bind subunit d (see arrow in Figure 4B, right panel), thereby making a structural link between the rotor (c-ring) and the stator (subunit a) (Figure 4A, right panel). Unlike the F-ATPase F_o proton channel, which is a leaky proton pore in isolation,^[76] the V_o proton channel is sealed to protons once V₁ disengages.^[13] It has been suggested that inhibition of proton leakage across free V_o is caused by the interaction between a_NT and d.^[77] However, while one study using purified yeast V_o showed that selective removal of d did not open the channel,^[78] experiments with V_o from plant V-ATPase indicated some restoration of proton flow upon d removal.^[79] Bafilomycin sensitive channel function was also partially restored by exposing reconstituted bovine V_o to low pH, however, the subunit composition of the acid treated V_o was not further analyzed.^[80] Taken together, while the interaction between a_NT and d that is only observed in isolated V_o is clearly inhibitory, there may be additional features that restrict c-ring rotation in absence of V₁.^[75]

Figure 4.

Structure of autoinhibited yeast V_o proton channel. (A) CryoEM structure of autoinhibited V_o in lipid nanodisc (PDBID 6M0R)^[75] as seen parallel to the membrane (left panel) and from the cytoplasm towards the bilayer (right panel). The interaction of a_NT and d is highlighted by the dashed pink circles. (B) CryoEM structure of yeast holo V-ATPase in state 3 (PDBID 7FDC)^[32] seen as in the panels in (A). The conformational change of a_NT from a peripheral location in V₁V_o to its binding site on d in free V_o is highlighted by the pink arrow in the view from the cytoplasm towards the membrane.

How Oxr1p binds V₁ and induces V-ATPase dissociation

We recently obtained cryoEM structures of yeast V-ATPase assembled in vitro from V₁ΔH, V_o in lipid nanodisc (V_oND), and recombinant subunits H_chim and C^[32] (Figure 2). Besides holoenzyme, the analysis revealed a population of V₁ΔH complexes bound to subunit C and a hitherto unseen component that we identified as Oxr1p. Yeast Oxr1p is a poorly characterized 273 residue polypeptide that is composed of a C-terminal TLDc domain plus a 60 residue N-terminal segment predicted to be largely disordered. The structure shows that V₁ is halted in rotary state 1, with Oxr1p wedged between the C-terminal domains of a non-catalytic AB heterodimer, the N-termini of EG2, and subunit C (Figure 5A).^[32] Sequence and structure analysis showed that many of the residues that form the V₁-Oxr1p binding site are conserved on both Oxr1p and V₁ subunits.^[32] Subsequent in vitro biochemical experiments revealed that recombinant Oxr1p inhibited V₁DH’s MgATPase activity in a subunit C-dependent manner, blocked in vitro assembly of the holoenzyme, and caused rapid release of V₁ from the vacuolar membrane.^[32] None of these inhibitory activities by Oxr1p required MgATP, suggesting that it is the greater stability of the resulting V₁(C)Oxr1p complex that provides the required driving force for breaking the protein-protein interactions at the V₁-V_o interface. A closer analysis of the Oxr1p binding site revealed that much of the interaction is mediated by the C-terminal domain of the B subunit that is part of the open catalytic site (Figure 5A). Since inhibition of V₁ΔH’s MgATPase activity by Oxr1p in absence of C is weak,^[32] this suggests that the interaction with C and EG2 is required for the formation of a high-affinity binding site. However, to engage with Oxr1p that is bound to B, both C and EG2 have to undergo significant shifts from their binding sites in V₁V_o;^[32] these structural changes appear to be sufficient for breaking of the V₁-V_o interface. We and others have previously shown that the V₁-V_o interface is constituted by multiple weak interactions except for the high-affinity interface between EG3’s N-termini and C_head.^[26,81,82] Moreover, from a crystal structure of the yeast heterotrimeric EGC_head complex we speculated that disassembly is initiated by destabilizing the ternary junction between the distal domain of a_NT, the N-termini of EG2, and C_foot.^[52] However, what was missing in our model was the identity of the trigger that would destabilize this ternary junction. Our recent structural and biochemical data suggest that one such trigger may be Oxr1p binding.

Figure 5.

Structure of yeast V₁ in complex with Oxr1p and the C subunit and comparison to mEAK7 bound human V-ATPase. (A) Two views of the cryoEM structure of V₁(C)Oxr1 (PDBID 7FDE),^[32] highlighting Oxr1p’s binding site formed by EG2, subunit C, and the B subunit of the open catalytic site. V₁(C)Oxr1 is halted in rotary state 1. (B) CryoEM structure of human V-ATPase in complex with mEAK7 (PDBID 7UNF).^[38] Left panel, same view as in left panel 5A. Right panel, like yeast Oxr1p, mEAK7 binds the N-termini of a EG heterodimer and the C-terminal domain of a B subunit with the open catalytic site conformation. However, note that mEAK7 binds to peripheral stalk EG3 rather than EG2 as Oxr1p. (C) Zoomed in views of the Oxr1p binding site (upper panel) and comparison to the same site in state 1 V₁V_o without Oxr1p bound (PDBID 7FDA). (D) Zoomed in view of the mEAK7 binding site (upper panel), and comparison to the same site in state 2 V₁V_o (PDBID 6WM3).

Whereas Oxr1p is the only TLDc domain containing protein in yeast, so far six TLDc domain-containing homologs have been identified in the human genome: OXR1, NCOA7, mEAK7, TBC1D24, TLDC2 and IFI44³. While most of these proteins have N- or C-terminal extensions, it has been shown for some of them that their TLDc domains alone are sufficient for function.^[33,41,83] In fact, the shortest splice variants of mammalian NCOA7 and OXR1, like yeast Oxr1p, are composed almost entirely of their TLDc domains.^[84,85] Notably, both NCOA7 and OXR1 were identified as high-scoring binding partners in an interactomics study of the kidney V-ATPase.^[35] More recently, a direct interaction was observed between the V-ATPase and mammalian NCOA7, OXR1, TBC1D24, mEAK7, and TLDC2.^[36] Further, mEAK7 has been identified in V-ATPase preparations from human cells^[38,86] and pig kidney.^[39] The accompanying cryoEM structures showed the enzymes halted in rotary state 2, with mEAK7 bound to the C-terminal domains of the non-catalytic subunit AB pair located behind peripheral stator EG3 (Figure 5B).^[38,39,86] Unlike Oxr1p, mEAK7 has a short C-terminal a helix that inserts between the catalytic site to contact subunit D. Thus, while both mEAK7 and Oxr1p bind subunit B and a peripheral EG stator, mEAK7 contacts EG3 whereas Oxr1p sits next to EG2 (compare left panel in Figure 5A to views in Figure 5B). Moreover, since Oxr1p binds EG2 and subunit B in the state 1 enzyme, its TLDc domain binds the B subunit of the open catalytic site (Figure 5C). In contrast, mEAK7 binds EG3 and the B subunit that is part of a closed catalytic site in the state 2 enzyme (Figure 5D, bottom panel), however, due to mEAK7 binding, the B subunit is now in a more open conformation (Figure 5D, top panel), possibly stabilized by mEAK7’s C-terminal a helix. Thus, both Oxr1p and mEAK7 bind B that is in the open conformation. However, because mEAK7 binds next to EG3 its interaction does not involve subunit C. So, why does mEAK7 not bind the same site as Oxr1p? mEAK7 has a large N-terminal extension, which would likely clash with the C subunit. The exclusion of subunit C from the binding interface may explain why mEAK7 does not lead to V-ATPase dissociation, contrary to Oxr1p’s interaction with the yeast enzyme. Consistent with this, and also contrary to Oxr1p, binding of mEAK7 appeared to have no significant effect on V-ATPase activity.^[38] As mentioned, a recent study showed that five mammalian TLDc family members were able to co-immunoprecipitate with the V-ATPase.^[36] Thus, the TLDc domain can be classified as a V-ATPase interaction module, with the various N- or C-terminal extensions likely mediating different effects on enzyme function that favor similar but distinct binding sites on the enzyme, as seen for mEAK7.

V-ATPase regulation by reversible disassembly

Early data from yeast and insects showed that V-ATPases readily dissociate under certain metabolic conditions - molting (starvation) in larvae of the tobacco moth, Manduca sexta,^[10] and glucose withdrawal in baker’s yeast.^[9] It was further found that when nutrients are added back, V₁ returns to the membrane to restore holoenzyme function, and the process is therefore referred to as reversible disassembly (or dissociation) (Figure 1B, left side). As described above, the activities of V₁ and V_o subcomplexes resulting from this V-ATPase regulation were found to be silenced.^[12,13,87] This suggested that the assembly state of the V-ATPase is linked to energy metabolism and glucose signaling, with the concomitant activity silencing of the V₁ and V_o serving to preserve energy (MgATP) and existing transmembrane proton gradients, respectively. Subsequent studies, however, have painted a more nuanced picture as several environmental conditions were identified that had no obvious link to glucose or ATP levels under which the degree of assembly was significantly altered, such as elevated extracellular pH or salinity.^[15,88] V-ATPase reversible disassembly was subsequently found to be conserved in higher animals including mammals, with reversible disassembly observed in specific enzyme populations and/or processes such as dendritic cell maturation,^[20] TOR signaling,^[21,22] synaptic vesicle loading,^[23] and glucose metabolism.^[24] Of note, conditions that triggered enzyme disassembly in yeast (glucose starvation)^[14] were in some cases conditions that led to the opposite result in mammalian cells,^[24] pointing to the convergence of diverse signaling pathways at the level of the enzyme. However, despite extensive characterization of the physiological causes and effects of V-ATPase regulation by reversible disassembly, the molecular mechanisms of the (re)assembly and disassembly steps at the level of the enzyme complex remain poorly understood. The following sections briefly summarize what we know about the processes of regulated disassembly and reassembly.

V-ATPase disassembly —

Studies in yeast showed that glucose withdrawal induced V-ATPase disassembly occurs on a timescale of about ~5 minutes,^[89] and that the process requires catalytically competent V-ATPases, as either mutations of residues essential for ATP hydrolysis and proton pumping, or presence of the specific inhibitor Concanamycin A result in hyperassembly.^[14] While enzyme dissociation occurs spontaneously in vitro upon MgATP hydrolysis, the process is much slower compared to in vivo disassembly.^[90] Besides inhibition of disassembly due to lack of enzyme activity, several instances have been described where disassembly kinetics is altered significantly despite presence of active V-ATPases. For example, mutations in peripheral stalk subunits at positions Arg25 in G^[91] and Glu44 in E^[92] result in V-ATPases that are highly active but resistant to disassembly. It is likely that these mutations interfere with one of the molecular steps that take place during regulated disassembly, but what these steps may be is not known. Taken together, these observations show that enzyme activity is a necessary, but insufficient prerequisite for efficient disassembly. Moreover, efficient disassembly depends on the presence of intact microtubules.^[29,93] While there is evidence for a physical interaction of tubulin with the C subunit, the molecular role of this interaction is not understood. As mentioned above, cryoEM analysis of V-ATPases from yeast and mammalian organisms revealed populations of the enzyme halted in three rotary states 1–3, with state 3 being the least populated.^{[32,56,59,61]} This suggests that state 3 has the highest energy, which led to the proposal that disassembly occurs while the enzyme pauses in that state;^[27,56] a speculation supported by the observation that autoinhibited V_o is halted in state 3.^[57,58,94] However, it can also be argued that disassembly has to occur in the most abundant state (state 1) for maximum efficiency, and that the conformations observed for autoinhibited V₁ (state 2) and V_o (state 3) are adopted subsequent to separation of V₁ from V_o. Support for this notion comes from a recent cryoEM study of in vitro assembled yeast V-ATPase, which revealed a similar distribution of rotary states compared to holoenzyme purified from yeast.^[32] Since the V_o used for in vitro assembly was in state 3, this indicates that the central rotor becomes mobile once autoinhibition of V_o is relieved and the holoenzyme is formed, even in absence of MgATP.

V-ATPase reassembly —

Along with having a role in V-ATPase biosynthesis, efficient reassembly of yeast V-ATPase requires a chaperone called Regulator of the H⁺-ATPases of the Vacuolar and Endosomal membranes (RAVE).^[95] RAVE is a heterotrimeric protein complex composed of three subunits, Rav1p, Rav2p and Skp1p.^[30] Biochemical studies showed that RAVE has multiple binding sites for both V₁ and V_o subunits,^[96] and it has been suggested that RAVE’s role is to recruit V₁ to V_o in a glucose dependent manner.^[97] Chaperone complexes with the function of RAVE are also found in higher organisms, including mammals, where they are celled rabconnectins; though the structure of rabconnectins is more diverse due to tissue specific isoforms.^[98] In yeast, RAVE is only required for assembly of the Vph1p containing V-ATPases on the vacuole, but not for the Stv1p containing enzymes in the Golgi; thus, loss of the RAVE complex results in only a partial Vma- phenotype.^[99] Biochemical experiments showed that RAVE binds the cytosolic domain of Vph1p and that the glucose dependent recruitment of RAVE to the vacuolar membrane depends on a highly conserved amino acid segment near the middle of Rav1p.^[97]

A comparison of the structure of the holoenzyme to the structures of the isolated V₁ and V_o subcomplexes revealed the conformational changes in the V₁-V_o interface that accompany V-ATPase disassembly. As shown for the yeast enzyme, autoinhibited V₁ and V_o are in different rotational states that are stabilized by interactions not present in intact V₁V_o (H_CT:A₃B₃D/EG2 in V₁ and a_NT:d in V_o)^[26,27] (Figures 3,4). Besides halting fruitless MgATPase activity by isolated V₁, and preventing collapse of proton gradients through V_o, these autoinhibited conformations likely serve to prevent spontaneous reassembly when enzyme function is not desired. Indeed, purified V₁ and V_o do not assemble with a measurable rate when mixed at physiological pH in vitro.^[74,100] For reassembly to occur, these inhibitory conformations have to be reversed. Since in vitro assembly of V_o with active V₁(H_chim) is a spontaneous reaction,^[32,74] this suggests that the rate limiting step is (i) release of H_CT from its inhibitory binding site on V₁ and (ii) reversal of the MgADP inhibited conformation. Enzyme assembly, unlike disassembly, does not require catalytically competent V₁ or V_o,^[14] but whether the structures of inactive mutant enzymes resemble that of the active V-ATPase is not known. As mentioned above, efficient (re)assembly in yeast requires presence of the RAVE complex. However, biochemical experiments have shown that while RAVE accelerates assembly of V_o with V₁(H_chim), the chaperone does not facilitate in vitro assembly of V_o with wild type V₁.^[101] Taken together, this suggests that additional factors are required to overcome V₁ autoinhibition. For example, the glycolytic enzymes aldolase^[102] and phosphofructokinase^[103] have been shown to play a role in reassembly, and while there is evidence that aldolase exerts its role via a physical interaction with V-ATPase’s peripheral stator,^[102] the role of phosphofructokinase at the level of the enzyme is not known.

On the physiological role of the V-ATPase-TLDc domain interactions

Human OXR1 (and yeast Oxr1p) were first identified in a screen for proteins that protect against oxidative DNA damage in E. coli,^[104] and while there was some initial speculation that OXR1 has anti ROS enzymatic activity, no evidence to support this notion has been found. A follow-up study suggested that the protective activity was contained in a segment of OXR1 N-terminal to the TLDc domain.^[105] However, as the shortest splice variant of human OXR1 encodes only the TLDc domain, like in yeast, and the TLDc domain on its own has been shown to rescue a deletion phenotype in animals, this suggests that the V-ATPase interaction is of critical and central importance to the function of this protein family.^[41] Notably, the V-ATPase itself has also been linked to the oxidative stress response in yeast.^[106] How the stress response and V-ATPase regulation are linked by this family of proteins is an exciting new avenue of research. If mammalian TLDc domain proteins are able to disassemble the V-ATPase, this poses the question how is specificity achieved? Interestingly, in binding studies using mammalian TLDc domain containing proteins, it was found that some display a preference for binding specific B subunit isoforms. Whereas TLDC2 and TBC1D24 interacted only with V-ATPases containing isoform B1, Meak7 interacted with both B1 and B2 isoform-containing V-ATPases.^[36] Further, under what cellular circumstances TLDc proteins are employed for V-ATPase regulation is an open question. Several environmental stimuli have been linked to the assembly state of the enzyme, but how these various signals converge on enzyme disassembly remains unclear. Are the multiple inputs linked to a single factor that acts directly on the enzyme? Are there multiple unique factors, each linked to different stimuli or restricted to distinct cellular locations? Since deletion of OXR1 from the yeast genome has no obvious phenotype (aside from a slightly increased sensitivity to oxidative stress),^[104] this shows that Oxr1p is unlikely to play an essential role in RAVE catalyzed enzyme assembly. On the other hand, our recent structural and in vitro biochemical studies indicate that Oxr1p has the ability to rapidly dissociate V₁ from the vacuolar membrane. However, it remains to be determined in what physiological context Oxr1p’s activity is employed. Several recent studies in animals or mammalian cells revealed an interplay between TLDc proteins and V-ATPase, either through physical interaction, or impairment of V-ATPase function upon TLDc protein deletion or mutation.^[37,40,41] Thus, it appeared that the biochemical data obtained in yeast, which pointed to an inhibitory mechanism of Oxr1p, were in contradiction with the physiological data obtained with the animal V-ATPase. To reconcile this apparent inconsistency, we proposed that yeast Oxr1p (and possibly the mammalian TLDc proteins) function in V-ATPase quality control. It has been known for some time that V-ATPase is subject to inhibition due to oxidation of cysteine residues in the enzyme’s catalytic A subunits.^[107,108] Because release of V₁ from V_o via reversible disassembly requires active enzyme, we speculated that Oxr1p may serve to disassemble damaged V-ATPases by removing inactive V₁ that otherwise would occupy functionally competent V_o on the membrane, especially under conditions of oxidative stress. Another possibility is that Oxr1p is directly involved in the reversible disassembly step in vivo, which would pose the question why in vivo disassembly requires ATP hydrolysis on the enzyme; as this would contrast Oxr1p induced dissociation, which occurs in the absence of added nucleotide. Synaptic vesicle (SV) V-ATPase has been shown to disassemble once neurotransmitter loading is complete, and while it has been proposed that the luminal pH may play a role in the disassembly step,^[23] the molecular details of the process are not known. Recently, using single-enzyme proton pumping assays, SV-V-ATPase was shown to be regulated by mode-switching between active, leaky, and inactive states; but whether the inactive state was due to reversible disassembly was not explored.^[109] Unlike in yeast, where V-ATPase hyper-assembly is without obvious phenotype, preventing disassembly of the SV-enzyme led to inhibition of SV fusion with the presynaptic membrane.^[23] Moreover, several TLDc domain-containing proteins have been shown to have an important role in neurological function.^[33,34,41] For example, knockout of NCOA7 in mouse was shown to cause neurological defects, with the additional knockout of OXR1 being embryonic lethal.^[37] Taken together, these observations raise the possibility that some of the mammalian TLDc domain containing proteins such as NCOA7, OXR1 or TBC1D24 function in the disassembly of the SV-enzyme; a process that is critical for proper synaptic transmission.

Conclusions and future perspectives

Since its discovery more than 25 years ago, tremendous progress has been made in elucidating the molecular mechanism of V-ATPase regulation by reversible disassembly. A major contribution to this progress was from high-resolution structure determination of holo V-ATPases and their V₁-ATPase and V_o proton channel subcomplexes from yeast and higher organisms using X-ray crystallography and cryoEM. These studies revealed snapshots of stable intermediates in the “life cycle” of the highly dynamic V-ATPase; structural models that highlight the conformational changes that the enzyme is undergoing during reversible disassembly. The recent discovery of a new player, Oxr1p’s TLDc domain, in yeast V-ATPase regulation has opened a new avenue of V-ATPase regulation, a mechanism that is only beginning to emerge. Important questions include (i) whether binding of Oxr1p is reversible so that Oxr1p disengaged V₁ can be recycled to reassemble with V_o when needed, (ii) whether Oxr1p is able to disassemble both active and inactive V-ATPases, or whether one form is preferred, (iii) whether Oxr1p driven disengagement of V₁ from V_o is somehow linked to the reversible disassembly process, or whether Oxr1p’s in vitro activity is part of an entirely novel in vivo pathway, and (iv) whether some of the mammalian TLDc proteins share yeast Oxr1p’s ability to disassemble the V-ATPase. There is ample evidence that TLDc domain containing proteins are intimately involved in V-ATPase function and regulation in higher animals, including mammals, but how these proteins perform their task at the level of the enzyme is largely unknown. From comparing Oxr1p’s and mEAK7’s binding sites and the proteins’ effect on holoenzyme function and integrity, it seems clear that while different TLDc proteins likely share the general mode of interaction with the enzyme, their effect on V-ATPase regulation may vary depending on the nature of the TLDc proteins’ N- and C-terminal extensions and the isoform composition and intracellular location of the V-ATPase.

Data availability

Data sharing is not applicable to this article as no new data were created or analyzed in this study.