Structure of V-ATPase from mammalian brain

Abstract

In neurons, loading of neurotransmitters into synaptic vesicles uses energy from proton-pumping V-ATPases. These membrane protein complexes possess numerous subunit isoforms, complicating their analysis. We isolated homogenous rat brain V-ATPase through its interaction with SidK, a Legionella pneumophila effector protein. CryoEM allowed construction of an atomic model, defining the enzyme’s ATP:H⁺ ratio as 3:10 and revealing a homolog of yeast subunit f in the membrane region, which we tentatively identify as RNAseK. The c-ring encloses the transmembrane anchors for cleaved ATP6AP1/Ac45 and ATP6AP2/PRR, the latter being the (pro)renin receptor that, in other contexts, is involved in both Wnt signalling and the renin-angiotensin system that regulates blood pressure. The structure shows how ATP6AP1/Ac45 and ATP6AP2/PRR enable assembly of the enzyme’s catalytic and membrane regions.

Vesicular- or vacuolar-type ATPases (V-ATPases) are ATP-hydrolysis-driven proton pumps that are essential for acidification of endosomes, lysosomes, and the trans Golgi network, as well as for acid secretion by osteoclasts, kidney intercalated cells, and some tumour cells (1, 2). ATP hydrolysis in the V-ATPase catalytic V₁ region drives rotation of a central rotor subcomplex and leads to proton translocation through the membrane-embedded V_O region. V-ATPase activity is regulated by reversible separation of the V₁ and V_O regions, with ATP hydrolysis inhibited in the isolated V₁ complex and the V_O complex becoming impermeable to protons (3, 4). In neurons, V-ATPase activity energizes synaptic vesicle membranes, allowing transporters to load the vesicles with neurotransmitters (Fig. 1A) (1, 2). Fusion of synaptic vesicles with the presynaptic membrane requires separation of the V₁ and V_O regions but it is not known how these events are coordinated (5). The regulated release of neurotransmitters from synaptic vesicles into the synaptic cleft allows signal propagation from the axon terminal of a presynaptic neuron to the dendrite of a postsynaptic neuron. After release of neurotransmitter, subsequent endocytosis and regeneration of synaptic vesicles occurs via clathrin-independent and clathrin-mediated routes (6), with the formation of clathrin-coated vesicles temporarily blocking V-ATPase activity (7). In the Saccharomyces cerevisiae enzyme, which is the most thoroughly characterized to date, the V₁ region contains subunits A₃B₃CDE₃FG₃H, and the V_O region contains subunits ac₈c′c″def and Voa1p (8, 9). The V₁ regions of mammalian V-ATPases contain the same subunits as the yeast enzyme while mammalian V_O regions are thought to be composed of ac_xc″de as well as ATP6AP1, also known as Ac45, and ATP6AP2, also known as the (pro)renin receptor (10, 11). ATP6AP2/PRR is involved in several signalling pathways (12) including the renin-angiotensin system for regulating blood pressure and electrolyte balance (10, 11, 13) and Wnt signalling in stem cells and embryo development (14). The precise arrangement of subunits in the mammalian V_O region remains unclear. Further, mammals have multiple isoforms of some subunits in both V₁ and V_O that are expressed in a tissue-dependent and cellular-compartment-dependent way. These include two isoforms of subunit B, two of C, two of E, three of G, four of a, two of d, and two of e (15, 16). Mass spectrometry of purified rat-brain synaptic vesicles detected V-ATPase subunits A, B1, B2, C1, D, E1, F, G1, G2, a1, a4, c, d1, ATP6AP1/Ac45, and ATP6AP2/PRR (17).

Fig. 1. Overall structure of brain V-ATPase.

(A) Cycle of synaptic vesicle loading, docking and priming, fusion, and recycling. (B) SDS-PAGE of rat brain V-ATPase isolated with 3×FLAG SidK_1-278 and gel filtration chromatography. (C) Native mass spectrometry of V₁ region (left) and native mass spectrometry of dissociated subunit G (right) at a higher-energy collisional (HCD) voltage of 250 V. The charge state for one peak per subunit is indicated. The table shows the measured mass for each peak (± SD of fit) and the calculated mass depending on subunit composition (Table S5). The difference between calculated and measured masses is indicated. (D) Composite cryoEM map (left) and atomic model (right) of brain V-ATPase in rotational state 1. Scale bar, 25 Å.

In order to isolate V-ATPase for structural analysis, we developed a purification strategy based on the high-affinity interaction of the enzyme with the Legionella pneumophila effector protein SidK (18, 19). While procedures capable of obtaining highly-purified synaptic vesicles have been described (17, 20) they enhance purity at the cost of yield, which would complicate or preclude structure determination. Instead, a two-step differential centrifugation procedure was used to collect rat synaptic vesicles and clathrin-coated vesicles along with other cell membranes. Membranes were solubilized with detergent and SidK (residues 1 to 278) fused to a C-terminal 3×FLAG tag was used with M2-agarose to purify V-ATPase. With this approach, a single ~2 g rat brain provided ~200 μg of highly purified V-ATPase (Fig. 1B). Mass spectrometry of tryptic fragments (Fig. S1A, Tables S1 and S2) identified the bands on the SDS-PAGE gel as V-ATPase subunits A, B2, C1, D, E1, F, G2, a1, c, and d1. A low-intensity band on the gel corresponding to subunit G1 was also detected, which from gel densitometry appears to comprise 17±3 % of subunit G in the preparation (±SD, three independent purifications). Subunit G1 may be a component of some V-ATPase complexes in synaptic vesicles (17). Alternatively, subunit G1 may be from copurifying lysosomal V-ATPase (21). The glycoprotein ATP6AP1/Ac45 and the small and hydrophobic proteins ATP6AP2/PRR and subunit e2, which do not stain clearly with Coomassie, were detected after cleavage with trypsin or chymotrypsin (Fig. 1B, Fig. S1A, Table S3). The subunit isoforms identified are all consistent with the V-ATPase from synaptic vesicle (17). This homogeneity could be due to the synaptic vesicle V-ATPase being the predominant form of the enzyme in the brain, or to V-ATPases from other cellular compartments in the brain having the same isoform composition as the synaptic vesicle enzyme. RNAseK, a hydrophobic protein recently found to associate with mammalian V-ATPase (22), was also detected after cleavage with trypsin or chymotrypsin (Fig. 1B, Fig. S1A, Table S3). Subunit H, which dissociates from bovine brain V-ATPase when treated with oxidizing agents (23), was not detected by mass spectrometry despite the absence of these reagents from the preparation. Subunit H is part of the mammalian V-ATPase (7, 17) and is needed for full enzyme activity (24). Loss of subunit H during purification of mammalian V-ATPase is markedly different from the subunit’s behaviour in S. cerevisiae (25). This difference is noteworthy because the protein’s physiological role of mechanically blocking ATP hydrolysis in the isolated V₁ region likely requires a strong attachment to the enzyme (26, 27). While tryptic peptides from subunits B1, C2, a2, a3, and a4 were also detected (Table S4), integrated peak intensities for these peptides in extracted ion chromatograms were two to three orders of magnitude lower than for peptides from subunits B2, C1, and a1, indicating that their abundance is negligible for structural studies.

Native mass spectrometry further demonstrated the homogeneity of the enzyme preparation (Fig. 1C, red, and Fig. S1B and C). Spectra show the V₁ region, presumably due to dissociation of the complex during analysis. The V₁ region has a native mass of 683369 ± 144 Da, consistent with a subunit composition of A₃B2₃C1DE1₃FG2₃SidK₃ (Fig. 1C and S1B, Table S5). Fragmentation of the V₁ region using a higher-energy collisional dissociation (HCD) voltage of 50 to 250 V confirmed the isoform composition of the V₁ region as B2, C1, and E1 (Fig. S1B). Specifically, spectra for subunit G showed a native mass of 13578 ±1 Da (Fig. 1C, right), consistent with subunit G2 (13578 Da) but not G1 (13621 Da), both of which are N-terminally acetylated (Fig. S1C). Lower abundance peaks were also seen for the V₁ region missing subunit C1 (Fig. 1C, green), and even with two copies of C1 (Fig. 1C, blue), the latter possibly being attributed to the loss of subunit H. The enzyme preparation had an ATP hydrolysis rate that decreased during hydrolysis, consistent with the complex lacking subunit H (24, 26) (Fig. S2). Further, the preparation did not demonstrate full coupling of ATPase activity in the V₁ region to proton transport through the V_O region: bafilomycin, which blocks the V_O region, only partially inhibited ATPase activity. This lack of coupling could be due to free V₁ complexes that remain active because subunit H is not present, or to the detergent-solubilized V-ATPase disassembling during ATP hydrolysis, as seen with the Manduca sexta enzyme (28). The extent of SidK inhibition on ATP hydrolysis activity could not be quantified because only the SidK-bound form of the enzyme was available, although previous work showed that SidK inhibits S. cerevisiae V-ATPase by ~40 % (19).

CryoEM of the preparation yielded three 3D maps, corresponding to ~120° rotations of the rotor subcomplex between states (25). These rotational states of the enzyme had overall resolutions of 3.9, 4.0, and 3.9 Å (Fig. S3 and S4). Focused refinement of rotational state 1 was able to improve the resolution to 3.8 Å resolution for the membrane region and 3.6 Å for the catalytic region of the complex. Together, these maps allowed construction of an atomic model for most of the complex, with a few components, including parts of subunits E1 and G2, the soluble N-terminal domain of subunit a1, subunit C1, and several luminal loops in the membrane region, modelled as backbone with truncated side chains (Fig. 1D, Table S6, Fig. S5). Similar to the yeast V_O structure (8, 29), no density was apparent for the loop between residues 667 and 712 in subunit a1. Owing to the averaging that occurs during cryoEM image analysis, the minor population of complexes in the preparation possessing subunit G1 rather than G2 could produce a map that shows a weighted average of both isoforms. Alternatively, images of complexes containing subunit G1 may be excluded during 2D and 3D image classification if they do not average coherently with the major population of complexes to produce high-resolution map features. Where the map shows high-resolution features for subunit G, it accommodates subunit G2 better than G1 (Fig. S6), consistent with most of the V-ATPase complexes containing subunit G2. Interpolating between the three rotational states produced a movie (Movie S1) that shows the conformational changes in the enzyme that couple ATP hydrolysis in the V₁ region to proton pumping through the V_O region. These changes illustrate the flexibility of the enzyme, particularly in the peripheral stalks, subunit C1, and N-terminal domain of subunit a1.

Previous high-resolution insight into the structure of the eukaryotic V₁ region is limited to cryoEM of the intact yeast V-ATPase at ~7 Å resolution (19, 25, 29) and a 6.2 to 6.7 Å resolution crystal structure of the autoinhibited yeast V₁(-C) complex (27). These maps only allowed visualization of α-helices in the structure. The current structure of the V₁ region (Fig. 2A) enables comparison to atomic models of prokaryotic V/A-ATPase catalytic regions (30–32) as well as numerous atomic models of ATP synthase F₁ regions (33). ATP synthases possess three catalytic and three non-catalytic nucleotide binding sites, found at the interfaces between subunits α (corresponding to V-ATPase subunit B) and β (corresponding to subunit A) (33). In contrast, and consistent with structures of the bacterial V/A-ATPase (31, 34), the mammalian V₁ region lacks non-catalytic nucleotide binding sites and shows only a single bound nucleotide in one of the catalytic sites (Fig. 2B and 2C-i). This nucleotide could be modelled in the density as ADP, consistent with biochemical analysis of the M. sexta V-ATPase (28). The three pairs of catalytic subunits in F- and V-type ATPases interchange between three different conformations, originally described for the F₁-ATPase as ATP-bound, ADP-bound, and empty (33). In the present structure, each catalytic AB pair similarly adopts one of three conformations (Fig. 2B and C). Comparison of the conformation of catalytic AB pairs to crystal structures of a prokaryotic V₁/A₁ region suggests that the ADP-bound site is in a post-hydrolysis state (Fig. 2B and C-i) (34) and that the site in an open conformation (clockwise from the ADP-bound site when viewed from V₁ towards V_O) is in a state with high affinity for ATP (31) (Fig. 2B and C-ii). The conformation of the third pair of AB subunits does not appear to correspond to previous structures but its nucleotide-binding pocket is occluded (Fig. 2B and C-iii), suggesting a low affinity for nucleotide (31). Therefore, the enzyme imaged here appears to be poised to bind ATP. Three copies of SidK_1-278 are bound to the three A subunits (Fig. S7). The structural consequences of SidK binding to the mammalian V₁ region are not known, but SidK perturbed the conformation of the yeast V₁ region only subtly (19). Although lower-resolution structures of the yeast enzyme suggested that subunit G does not contact the rest of the V₁ region (25, 27), the present structure shows that it does indeed participate in linking the EG heterodimers of the peripheral stalks to the catalytic A₃B₃ subcomplex in the mammalian enzyme (Fig. 2D). This connection relies on residues from the N terminus of subunit B2 and C-terminal residues of subunits E1 and G2 (Fig. 2D, asterisks) and includes a structure where β-strands from both subunit B2 and E1 form a single β-sheet (Fig 2E, purple arrowhead), also seen in a recent cryoEM map of a prokaryotic V/A-ATPase (32). Numerous mutations in subunits A, B1, B2, and E1 are linked to disease and can be mapped onto the structure (Fig. S8).

Fig. 2. Structure of the V₁ region.

(A) Surface representation. Scale bar, 25 Å. (B) Cross-section through the V₁ region viewed from V₁ towards V_O. ADP is shown in green. (C) Superposition of the catalytic AB pairs (left) and closeup of the conformations of the nucleotide binding sites (right). Scale bar, 5 Å. (D) Interaction of subunits B2, E1, and G2. N and C termini of models are indicated by ‘*’. (E) A continuous β-sheet is formed by E1 and B2 (purple arrowhead).

The V_O region contains subunits a1, d1, e2, f, ATP6AP1/Ac45, ATP6AP2/PRR, and the c-ring (Fig. 3A). While previous high-resolution structures of S. cerevisiae V_O regions were determined from auto-inhibited complexes separated from their V₁ regions (8, 9, 29), the structure presented here is within the context of an assembled V-ATPase. Consequently, the soluble N-terminal domain of subunit a1 is found in its non-inhibitory conformation and does not make contact with subunit d1 (Fig. 3B, green). The membrane-embedded C-terminal domain of subunit a1 includes eight transmembrane α-helices (Fig. 3C) and creates the two offset half-channels that allow proton translocation (8). The subunit starts with a V-shaped insertion into the lipid bilayer formed by two short α-helices that do not entirely cross the membrane (α1 and α2), followed by four transmembrane α-helices (α3 to α6) and two highly tilted transmembrane α-helices (α7 and α8) that create the surface that contacts the c-ring. This fold closely follows yeast Vph1p and Stv1p (8, 29), with RSMDs between C_α atoms of 1 and 0.8 Å, respectively (Fig. S9A). The major difference between these structures is the insertion of a structured linker with α-helices connecting transmembrane α3 and α4 in subunit a1 (Fig. 3C and Fig. S9B). The c-ring contains nine copies of subunit c and one of subunit c″ (Fig. 3B, pink and purple). As expected from the lack of a homologous gene in mammals, the ring does not contain subunit c′ found in yeast. Conserved Glu residues from protomers of the c-ring are each capable of carrying a single proton during proton translocation (Fig. 3B, red spheres). As with the yeast V_O region, subunit c″ breaks the pattern of conserved proton-carrying Glu residues occurring on alternating outer α-helices of the c-ring (Fig. 3B, red arrow) (8). The presence of ten protomers in the c-ring sets the ATP:H⁺ ratio at three ATP molecules hydrolyzed for every ten protons translocated. With a Gibbs free energy for ATP hydrolysis of approximately -60 kJ/mol (35), this ratio limits the transmembrane proton motive force established by the brain V-ATPase, $\Delta {\tilde{\mu }}_{H^{+}}$, to ~18 kJ/mol, which is equivalent to ~3 pH units or ~180 mV (35). Similarity between the mammalian and yeast V-ATPase structures suggests they share the same mechanism for rotation-driven proton pumping (8) (Fig. 3D). In this mechanism, rotation of the c-ring drives abstraction of a proton from the cytoplasmic half-channel to neutralize the charge on a conserved c-ring Glu residue as it enters the lipid bilayer. Rotation brings a protonated Glu residue close to Arg741 of subunit a1, with formation of a salt bridge between the Arg and Glu causing release of the proton into the luminal half-channel that begins near Glu795 of subunit a1. While structures of ATP synthases (32, 36–38) have resolved the gap between the two tilted α-helices of subunit a where protons must pass, this opening cannot be seen here (Fig. 3E, grey arrow) or in other V_O region structures (8, 9, 29). This difference is either because of limited resolution or because the opening only forms transiently in V-ATPases that, unlike ATP synthases, pump protons against a proton motive force. The surface of subunit a1 that contacts the c-ring has the pattern of a positive charge at the conserved Arg residue and negative charges at the mid-membrane termini of the two half channels (39) that has now been seen in all other rotary ATPases (Fig. S9C).

Fig. 3. Structure of the V_O region.

(A) Surface representation with cryoEM density for subunits f, ATP6AP1/Ac45, and ATP6AP2/PRR. Scale bar, 25 Å. (B) Viewed from V₁ with conserved proton-carrying Glu residues as red spheres. The direction of ATP-hydrolysis-driven rotation of the ring is indicated. (C) Cartoon representation viewed parallel to the plane of the lipid bilayer. (D) Proton path through the V_O region. (E) Surface representation of a1_CTD viewed parallel to the plane of the lipid bilayer. The grey arrow indicates the expected location of an opening between α7 and α8 leading towards the luminal half-channel. (F) Closeup view of the luminal terminus of the luminal half-channel, showing the interaction of subunits f, e2, a1, and the unidentified density. Scale bar, 10 Å.

Subunit e2 is adjacent to transmembrane α-helices 3 and 4 of subunit a1 and consists of an α-helical hairpin with a C-terminal tail, similar to the corresponding yeast protein Vma9p (8) (Fig. 3A and C, blue, and Fig. S10). However, unlike yeast Vma9p, subunit e2 has an extended C-terminal sequence that terminates in a third short α-helix (α3) encircled by the linker between α3 and α4 of subunit a1 and the luminal loop of subunit f (Fig. 3C and S10, blue arrow). The function of subunit e2 is unknown (40), but deletion of Vma9p causes the Vma^-V-ATPase deficiency phenotype, where yeast can grow on medium buffered to pH 5.5 but not pH 7.5, are sensitive to extracellular calcium and a variety of heavy metals, and cannot grow on medium with typical concentrations of non-fermentable carbon sources (41). Recent analysis of both the vacuolar and Golgi forms of the S. cerevisiae V-ATPase identified the hypothetical protein YPR170W-B as a transmembrane α-helical hairpin subunit, named subunit f (8, 29). YPR170W-B is highly conserved in fungi but deletion of the gene did not cause the Vma^-phenotype (8). The structure of the brain V-ATPase revealed density for a similar transmembrane α-helical hairpin in the V_O region in a position corresponding to the yeast subunit f (Fig. 3, red). Bioinformatic analysis suggests RNAseK, a conserved metazoan protein (42), as a homolog of S. cerevisiae subunit f, sharing 32 % sequence identity and 52 % sequence similarity (Fig. S11A). Immunoprecipitation of RNAseK previously revealed that it is associated with V-ATPase (22) and mass spectrometry done here shows that RNAseK is present in this enzyme preparation (Tables S1 to S3). Consistent with a role in V-ATPase activity, RNAseK is necessary for viral endocytosis and replication (22, 43). The cryoEM density for subunit f, although lacking the necessary resolution to unambiguously identify the protein as RNAseK, is consistent with the expected size of RNAseK and accommodates most of its bulky side chains (Fig. S5B). Together, this evidence tentatively identifies RNAseK as subunit f in mammalian V-ATPases. Whether or not RNAseK functions as a ribonuclease within the V-ATPase is not known, but we note that the in vitro assays that established this protein and its homologs as ribonucleases (42, 44) were performed in the absence of detergents or other membrane mimetics despite clear transmembrane α-helices in hydropathy analysis of the protein’s amino acid sequence (Fig S11B). An additional unidentified protein-like density previously found in both the vacuolar and Golgi forms of the S. cerevisiae V-ATPase (9, 29) is also present in the mammalian complex (Fig. 3C to F, purple). All four of the membrane-embedded components from this part of the enzyme interact at the opening of the luminal proton half-channel, with the linker between α3 and α4 of subunit a1 acting as a scaffold for packing the luminal loop connecting the transmembrane α-helices of subunit f, the C-terminal sequence of subunit e2, and part of the unknown component (Fig. 3F). These four components are all elongated relative to their yeast counterparts in order to accommodate their interaction (Fig. S9 to S11).

Unlike the fungal V-ATPase, three rather than two transmembrane α-helices are within the lumen of the c-ring (Fig. 4A). The centremost α-helix corresponds to the N-terminal α-helix of subunit c″, also seen in the yeast V-ATPase (8, 9, 29). The two other transmembrane α-helices are from ATP6AP1/Ac45 and ATP6AP2/PRR (Fig. 4A and S5C). ATP6AP1/Ac45 mutations in humans can lead to immunodeficiency, cognitive impairment, liver dysfunction, and abnormal protein glycosylation (45) while mutations in ATP6AP2/PRR can result in neurodegeneration as well as X-linked parkinsonism and epilepsy (12). Both proteins are frequently mutated in granular cell cancers (46). Immature ATP6AP1/Ac45 contains two soluble N-terminal domains on the luminal side of the membrane separated by a furin cleavage site (47) (Fig. 4B, bottom). In the map, the soluble region of APT6AP1/Ac45 appears as a single low-resolution domain ~30×20×30 Å in size (Fig. 4C, light blue), consistent with only the second luminal domain being in the complex. Further, although mass spectrometry detected peptides from uncleaved ATP6AP1/Ac45, it suggested that predominantly the cleaved form is present (Fig. S12A and C). The C-terminal transmembrane α-helix of ATP6AP1/Ac45 is homologous with that of the yeast V-ATPase subunit Voa1p (45) and is found in an equivalent position inside the c-ring (9). This transmembrane α-helix stretches approximately half-way across the thickness of the lipid bilayer terminating as a C-terminal tail that interacts with two of the subunits in the c-ring and subunit d1 before entering the cytoplasm (Fig. 4C, light blue). ATP6AP2/PRR is known to associate with mammalian V-ATPase (10, 11) and is essential for biogenesis of active V-ATPase (48). The strong density for its transmembrane α-helix in the map suggests that every complex includes the subunit (Fig. S5C). Consistent with an emerging role for V-ATPases in cell signalling (49), the presence of ATP6AP2/PRR in the complex has tied V-ATPases to the renin-angiotensin system for regulation of blood pressure and electrolyte balance (10, 11, 13), Wnt signalling (14), and other pathways (12). The gene for ATP6AP2/PRR encodes an N-terminal extracellular/luminal soluble domain and a transmembrane anchor (Fig. 4B, top). The soluble domain increases the angiotensin I generating activity of renin and can function both in a membrane-bound form and, when released by proteolysis, in a soluble form (50,51). In the structure, the renin-activating domain of ATP6AP2/PRR is missing (Fig. 4D, yellow). The transmembrane anchor consists of a long α-helix and a short α-helical turn connected by an extended linker with N- and C-terminal tails. The location of ATP6AP2/PRR’s transmembrane region, captured within the stable c-ring, dictates that after its incorporation into the c-ring the protein must remain associated with the V-ATPase. Mass spectrometry detected some intact ATP6AP2/PRR in the preparation but, consistent with the cryoEM density, the cleaved transmembrane region alone was much more abundant (Fig. S12B and D). In contrast to the renin-angiotensin system, Wnt signalling relies on ATP6AP2/PRR remaining membrane-anchored, with interaction of proteins with the extracellular/luminal part of ATP6AP2/PRR leading to signalling in the cytoplasm (14). The absence of the soluble domain of ATP6AP2/PRR from the structure suggests that ATP6AP2/PRR’s function in Wnt signalling involves either a subpopulation of intact ATP6AP2/PRR molecules that are not associated with V-ATPase, or a different population of V-ATPase or V_O complexes containing intact ATP6AP2/PRR.

Fig. 4. Interaction of subunits within the membrane embedded rotor subcomplex.

(A) Subunits ATP6AP1/Ac45, ATP6AP2/PRR, and c″possess transmembrane α-helices in the centre of the c-ring. (B) ATP6AP2/PRR (top) and ATP6AP1/Ac45 (bottom) possess luminal domains that are absent from the structure. SP, signal peptide; Cyt, cytosolic. (C and D) Subunit d1 interacts with the C termini of ATP6AP1/Ac45 (C) and ATP6AP2/PRR (D). (E) Subunits c″, c₍₈₎, and ATP6AP2/PRR interact with the second luminal domain of ATP6AP1/Ac45. Scale bar, 10 Å. (F) Subunits c″, ATP6AP1/Ac45, ATP6AP2/PRR, and d1 create a network of interactions that connect the vesicle lumen and the cytoplasm. Scale bar, 25 Å.

The structure shows numerous interactions between ATP6AP1/Ac45, ATP6AP2/PRR, subunits d1, c″, and multiple c-subunits (Fig. 4E and F). Because these proteins are all part of the rotor subcomplex, the interactions persist during rotation and are also found in rotational states 2 and 3 (Fig. S13). In the lumen, the soluble domain of ATP6AP1/Ac45 interacts with the N-terminal tail of cleaved ATP6AP2/PRR, the linker that connects subunit c″ to its N-terminal α-helix in the middle of the c-ring, and the N terminus of subunit c₍₈₎ (Fig. 4E). Near the cytoplasmic surface of the V_O region, the C-terminal tail of ATP6AP1/Ac45 and the short C-terminal α-helix of ATP6AP2/PRR, including 12 of the 19 residues in its intracellular domain (14), are sandwiched between subunits of the c-ring and subunit d1 (Fig. 4C and D, arrow). V_O complexes assemble in the endoplasmic reticulum (ER), with the incorporation of subunit d allowing ER release (52, 53) and subsequent binding of the V₁ region (53). Free subunit d in the cytoplasm (54) does not bind the V₁ complexes that are preassembled there (55), suggesting that the conformation of subunit d changes upon incorporation into V_O, making it competent to interact with subunit D of V₁ (Fig. 1D). Indeed, subunit d adopts a more open conformation when it engages subunit D of the central rotor in the intact V-ATPase and a more closed conformation when V₁ is detached and subunits d and D are disconnected (Fig. S14). The C-terminal tails of ATP6AP1/Ac45 and ATP6AP2/PRR produce part of the surface onto which subunit d1 assembles (Fig. 4C, D, and F). This structure explains why mutations in ATP6AP1/Ac45 and ATP6AP2/PRR are associated with related disease phenotypes (12, 45, 46). It also explains why ATP6AP1/Ac45 and ATP6AP2/PRR appear to work together to allow subunit d binding, ER release, and subsequent V₁ assembly onto V_O in the mammalian V-ATPase (9, 56). It is tempting to speculate that ATP6AP1/Ac45 and ATP6AP2/PRR are also involved in the reverse process, with conformational changes within them altering the conformation of subunit d1, disrupting its interaction with subunit D, and triggering separation of the V₁ and V_O regions.

One Sentence Summary.

CryoEM and mass spectrometry reveal the subunit composition and how (pro)renin receptor works with Ac45 in V-ATPase assembly.