Molecular mechanisms of cutis laxa– and distal renal tubular acidosis–causing mutations in V-ATPase a subunits, ATP6V0A2 and ATP6V0A4

The a subunit is the largest of 15 different subunits that make up the vacuolar H+-ATPase (V-ATPase) complex, where it functions in proton translocation. In mammals, this subunit has four paralogous isoforms, a1–a4, which may encode signals for targeting assembled V-ATPases to specific intracellular locations. Despite the functional importance of the a subunit, its structure remains controversial. By studying molecular mechanisms of human disease–causing missense mutations within a subunit isoforms, we may identify domains critical for V-ATPase targeting, activity and/or regulation. cDNA-encoded FLAG-tagged human wildtype ATP6V0A2 (a2) and ATP6V0A4 (a4) subunits and their mutants, a2P405L (causing cutis laxa), and a4R449H and a4G820R (causing renal tubular acidosis, dRTA), were transiently expressed in HEK 293 cells. N-Glycosylation was assessed using endoglycosidases, revealing that a2P405L, a4R449H, and a4G820R were fully N-glycosylated. Cycloheximide (CHX) chase assays revealed that a2P405L and a4R449H were unstable relative to wildtype. a4R449H was degraded predominantly in the proteasomal pathway, whereas a2P405L was degraded in both proteasomal and lysosomal pathways. Immunofluorescence studies disclosed retention in the endoplasmic reticulum and defective cell-surface expression of a4R449H and defective Golgi trafficking of a2P405L. Co-immunoprecipitation studies revealed an increase in association of a4R449H with the V0 assembly factor VMA21, and a reduced association with the V1 sector subunit, ATP6V1B1 (B1). For a4G820R, where stability, degradation, and trafficking were relatively unaffected, 3D molecular modeling suggested that the mutation causes dRTA by blocking the proton pathway. This study provides critical information that may assist rational drug design to manage dRTA and cutis laxa.

Vacuolar H ϩ -ATPases (V-ATPases) 4 are conserved, multisubunit rotary proton pumps that play crucial roles in regulating the pH of cells and their intracellular compartments (1)(2)(3)(4)(5). They can be categorized as endomembrane or plasma membrane V-ATPases, based on their subcellular localization (6,7). Endomembrane V-ATPases are expressed in all eukaryotic cells in the membranes of acidic organelles like lysosomes, endosomes, and the Golgi apparatus, where they translocate protons to acidify the luminal compartments of the organelles (8). Plasma membrane V-ATPases traffic to the surfaces of some specialized cells, such as osteoclasts, kidney-intercalated cells, and metastatic cancer cells, where they secrete protons into the extracellular fluid (6, 9 -11).
The V-ATPase complex consists of 15 different subunits arranged into two major sectors, the cytoplasmic V 1 sector and the membrane-integrated V 0 sector. V 1 is responsible for ATP hydrolysis that provides the energy to rotate a central shaft that powers proton translocation (3). V 0 contains a coupled rotor that carries protons for transport through a proton channel pathway formed largely by the approximately 100-kDa a subunit (2,12). The a subunit is the largest V-ATPase subunit, and in mammals there are four isoforms, a1-a4. The N-terminal half of the protein (NTa) is hydrophilic and associates with subunits of the V 1 sector in the V-ATPase complex, and the C-terminal half (CTa) is an integral membrane domain consisting of 8 transmembrane ␣-helices (TMs) and a cytoplasmic (C-terminal) tail domain (CTD). Whereas a1 and a2-containing V-ATPase complexes are targeted to endomembranes, a3 and a4 complexes are targeted to plasma membranes in some specialized cells (13,14).
Human missense mutations of the a subunits are implicated in diverse diseases (1,15). For example, mutations affecting the function of a2 result in cutis laxa (wrinkled skin syndrome), where aberrant Golgi function results in glycosylation defects with consequent abnormal elastin processing that affects skin and internal organs (16 -18). Mutations that affect a3 function result in inability of osteoclasts to resorb bone, causing autosomal malignant osteopetrosis that is characterized by dense, brittle bone (19,20). Loss of a4 function due to mutation results in distal renal tubular acidosis (dRTA) with occasional hearing loss (21,22). Here we focus on the effect of human mutations on a2 traffic to Golgi and a4 traffic to the plasma membrane based on their proposed in vivo locations and functions.
Despite such important implications for a subunit functions in disease, the structures of human a subunit isoforms are still controversial because of a lack of high-resolution structural data. Recently, however, a 6.4-Å model of the membrane-integrated domain of the yeast a subunit (Vph1p) has been published (23,24). This model is based on a synthesis of data derived from cryo-EM 3D reconstruction, evolutionary covariance mapping of key residues, and low resolution X-ray crystallography. It confirms that the a subunit membrane domain consists of 8 TMs, as has been previously shown (2,12), with TM7 and TM8 highly tilted and forming an interface with the V 0 rotor c-ring that enables proton translocation at the a subunit/c-ring interface.
Despite such recent advances, knowledge of a subunit folding, targeting, and assembly into the V-ATPase holocomplex remains sparse. Considerably more investigation will be required to elucidate issues such as, for example, the mechanism of plasma membrane a subunit targeting, the resolution of which will be required before efforts at designing strategies for targeted therapeutic interventions can realistically be considered. To that end, we conjectured that human disease-causing missense mutations within a subunits could be used to identify critical domains essential for V-ATPase targeting, activity and/or regulation. As an approach to testing this, we have studied the molecular consequences of introducing the cutis laxacausing mutation, Pro-405 3 Leu (P405L) in a2, and the dRTA-causing mutations, Arg-449 3 His (R449H) and Gly-820 3 Arg (G820R) in a4, into epitope-tagged human a subunit constructs for expression and characterization in the HEK 293 mammalian expression system. We present here results of these studies with respect to subunit glycosylation, stability, degradation, incorporation into V-ATPase complexes, and subcellular localization.

Amino acid residues a2 Pro-405, a4 Arg-449, and a4 Gly-820 are highly conserved
Alignments of a subunit polypeptide sequence segments affected by the human mutations causing cutis laxa and dRTA that are under study in the present work are shown in Fig. 1, A and B. The mutated residues (highlighted in red) are identical in all four human and mouse a subunit isoforms, and also in the yeast a subunit isoform, Vph1p (highlighted in yellow). Fig. 1A shows a segment of the integral membrane domain of the a subunit, where human mutations in a2 Pro-405 (in TM1; TMs highlighted in blue) and a4 Arg-449 (in TM3) result in cutis laxa and dRTA, respectively. Fig. 1B shows alignments for a C-terminal segment of the a subunit comprising the CTD, where the human mutation in a4 Gly-820 results in dRTA. Thus, the three mutations under consideration here, a2 P405L , a4 R449H , and a4 G820R , all affect highly conserved amino acid residues.

Glycosylation and stability of cutis laxa mutant, a2 P405L
We have previously shown that all human a subunit isoforms are N-glycosylated and that N-glycosylation is required for their stability (25,26). We have also shown in previous work that in the case of the osteopetrosis mutation, a3 R444L , the a subunit is misfolded, unglycosylated, retained in the ER, and ultimately subjected to proteolytic degradation (20). It was of interest, therefore, to determine whether the cutis laxa and dRTA mutations have similar impacts on a2 and a4 subunits, respectively, using methods for assessing N-glycosylation and stability that were previously described (25). Briefly, glycosylation and stability were tested in HEK 293 cells by transient transfection and expression of FLAG-tagged wildtype and mutation-bearing a2 and a4 subunit constructs. Whole-cell lysates prepared 24 h post-transfection were treated with peptide N-glycosidase F (PNGase F) to assess whether mutant proteins, a2 P405L , a4 R449H , and a4 G820R , were N-glycosylated, and with endo-␤-N-acetylglucosaminidase H (Endo H) to determine whether any bound glycans were of the high mannose or hybrid type (27,28). Stability was assessed using the CHX chase method as previously described (25,26). Briefly, cells were treated, 24 h posttransfection, with CHX (10 g/ml) for up to 12 h, and wholecell lysates were prepared, immunoblotted, and quantified (GAPDH was used as a loading control; see "Experimental procedures"). Fig. 1C shows immunoblots of wildtype FLAGtagged a2 protein (WT a2-2FLAG), and the similarly epitopetagged cutis laxa mutant subunit, a2 P405L (a2 P405L -2FLAG) expressed transiently in HEK 293 cells, with and without Endo H treatment of the whole-cell lysates. WT a2-2FLAG was observed as a 110-kDa band, and upon Endo H treatment its relative mobility was reduced to 105 kDa, representing the deglycosylated a2-2FLAG. The mutant a2 P405L -2FLAG was also observed as a 110-kDa band, and upon Endo H treatment its relative mobility was reduced to 105 kDa, representing the deglycosylated a2 P405L -2FLAG.
In the same manner, protein stability of a2 P405L was assessed by transient expression of the mutant protein or its wildtype counterpart. After allowing 24 h of expression, the cells were incubated with or without 10 g/ml of CHX, and were harvested after the indicated times for whole-cell lysate preparation (see "Experimental procedures"). Glycans were removed from all proteins, wildtype and mutant, prior to immunoblotting, by treatment with PNGase F. Fig. 1D shows quantitative band analysis of the immunoblots used to assess stability of a2 P405L -2FLAG transiently expressed in HEK 293 cells. All band intensities were normalized to GAPDH as a loading control, and to zero time controls. These data showed that a2 P405L -2FLAG was degraded at a significantly faster rate than WT a2 (p Ͻ 0.05), the mutant protein having a half-life of 13.4 Ϯ 1.0 h compared with 23.8 Ϯ 4.3 h for WT a2-2FLAG (see supporting Tables S1 and S2 for data and statistics for all stability assays in the present work).

Functional domains of ATP6V0A2 and ATP6V0A4
Glycosylation and stability of dRTA mutants, a4 R449H and a4 G820R Transient expression of FLAG-tagged human WT a4 and dRTA mutants was performed as for the a2 constructs. On immunoblotting, as shown in Fig. 1E, WT a4 -2FLAG was observed as a 105-kDa band and, upon PNGase F and Endo H treatments, its relative mobility was reduced to 98 kDa, representing the deglycosylated a4 -2FLAG. Similarly, a4 R449H -2FLAG and a4 G820R -2FLAG were observed as 105-kDa bands, and upon PNGase F or Endo H treatments their relative mobilities were reduced to 98 kDa, representing deglycosylated a4 R449H -2FLAG and a4 G820R -2FLAG. Thus, both a4 R449H and a4 G820R appeared to be N-glycosylated with Endo H-sensitive glycans, consistent with what was observed for WT a4.
To determine stability, WT a4 and the mutant proteins a4 R449H and a4 G820R were compared using CHX chase experiments, as described above; whole-cell lysates were prepared at the indicated time intervals and analyzed. Glycans were removed from all proteins, wildtype and mutants, by PNGase F treatment of cell lysates prior to immunoblotting; Fig. 1F shows quantitative band analysis of the immunoblots. All band intensities were normalized, as described above. Analysis of the data graphed in Fig. 1F showed that stability of a4 G820R -2FLAG, with a half-life of 13.7 Ϯ 1.7 h, was reduced by only 20% (p Ͻ 0.05) relative to WT a4 -2FLAG (17.0 Ϯ 1.2 h); however, the half-life of a4 R449H -2FLAG (4.8 Ϯ 0.39 h) was greatly reduced, by over 70% (p Ͻ 0.01) relative to WT a4 -2FLAG. Red highlights indicate amino acids affected by human disease-causing mutations (noted above alignments). Yellow highlights indicate amino acids corresponding to the human mutations, within the subunit isoforms and species shown. B, alignments as in A, but of sequences from the end of TM8 to the C terminus, encompassing the cytoplasmic CTD. C, HEK293 cells were transfected with either WT a2-2FLAG (WT a2), or mutant a2 P405L -2FLAG (a2 P405L ), and lysates treated with (ϩ) or without (Ϫ) Endo H. D, same constructs as in C, but treated with 10 M CHX for the time indicated; plots show band intensities quantified from immunoblots of post-CHX chase. Data were normalized to GAPDH and zero time control. E, same as C, except HEK 293 cells were transfected with WT a4 -2FLAG (WT a4), mutant a4 R449H -2FLAG (a4 R449H ), or mutant a4 G820R -2FLAG (a4 G820R ), and lysates were treated with or without PNGase F or Endo H. F, same as D, except HEK 293 cells were expressing WT a4, mutant a4 R449H , or mutant a4 G820R . Data are representative of three independent biological experiments; error bars indicate Ϯ S.D.

Pathways for degradation of unstable mutant proteins, a2 P405L and a4 R449H
The stability of the a4 G820R mutant subunit was not greatly different from wildtype, but the a2 P405L and a4 R449H mutants were clearly unstable. It was of interest to further characterize whether degradation of the latter two mutants was via the proteasomal pathway or the lysosomal pathway. After expression of a2 P405L -2FLAG and a4 R449H -2FLAG in HEK 293 cells, CHX chase experiments were done with and without either an inhibitor of proteasomes (10 M MG132), or lysosomes (25 mM NH 4 Cl), as previously described (25). Fig. 2 shows quantitative band analyses for the immunoblots loaded with WT a2-2FLAG and a2 P405L -2FLAG, or WT a4 -2FLAG and a4 R449H -2FLAG, before and after MG132 treatment. Analysis of data in Fig. 2A showed that stability of the a2 P405L -2FLAG construct (half-life 13.4 Ϯ 1.4 h) was 64% that of WT a2-2FLAG (half-life 21.0 Ϯ 1.6 h); however, after proteasomal inhibition, the degradation rates of a2 P405L -2FLAG (half-life 17.7 Ϯ 0.69 h) and WT a2-2FLAG (half-life 17.8 Ϯ 1.2 h) were indistinguishable (p ϭ 0.89). Data for Fig. 2B showed that without proteasomal inhibition, the half-life of the mutant a4 R449H -2FLAG (5.6 Ϯ 0.12 h) was 26% that of WT a4 -2FLAG (21.3 Ϯ 3.7 h; p Ͻ 0.05). After proteasomal inhibition, there was a highly significant decrease (p Ͻ 0.01) in the degradation rate of a4 R449H -2FLAG (half-life 5.6 Ϯ 0.12 h before treatment, 21.5 Ϯ 1.5 h after), with restoration of stability to levels exceeding that of WT a4 -2FLAG with the same treatment (half-life 17.2 Ϯ 1.0 h). Data for Fig. 2C showed that lysosomal inhibition partially restored stability of a2 P405L -2FLAG (half-life 7.8 Ϯ 0.51 h before and 11.4 Ϯ 1.0 h after treatment, p Ͻ 0.01), by about half (56%) of the difference between untreated mutant levels and treated wildtype levels. Finally, data from Fig. 2D showed that lysosomal inhibition had no significant effect (p ϭ 0.10) on the degradation rate of a4 R449H -2FLAG (half-life 5.1 Ϯ 0.15 h before treatment, 5.6 Ϯ 0.34 h after treatment). Taken together, this suggested that degradation of a2 P405L occurs both in the proteasomal and lysosomal pathway, whereas the degradation of a4 R449H predominantly occurs in the proteasome.

a2 Pro-405 is required for Golgi trafficking, and a4 Arg-449 for ER exit
The apparently significant degradation of both a2 P405L and a4 R449H suggests that the mutant subunits fail to assemble into the V-ATPase complex; therefore, we conducted immunofluorescence localization experiments to establish whether there is colocalization of these mutant proteins with ER and/or Golgi compartment markers. Fig. 3, A and B, show colocalization studies of WT a2-2FLAG and a2 P405L -2FLAG with calnexin (ER marker) and syntaxin 6 (Golgi marker). Fig. 3A shows representative fluorescence photomicrography images of HEK 293 cells transfected with empty vector (left-most panel), WT a2-2FLAG (middle panel), and a2 P405L -2FLAG (right-most panel), probed with anti-FLAG antibody (green) and antibodies to the ER marker, calnexin (red). These images showed that a2 P405L -2FLAG (green) colocalized with calnexin at a rate similar to that seen for WT a2-2FLAG (p ϭ 0.073). A similar experiment is shown in Fig. 3B, but using the Golgi marker protein, syntaxin 6 (red). The a2 P405L -2FLAG mutant protein appeared to colocalize with the Golgi marker at a rate lower than was apparent for WT a2-2FLAG (p Ͻ 0.05). Fig. 3C shows representative fluorescence photomicrography images of control, empty vector-transfected cells (left-most panel), WT a4 -2FLAG (second from left), a4 R449H -2FLAG (second from right), and a4 G820R -2FLAG (right-most panel) probed with anti-FLAG (green) and anti-calnexin (red) antibodies. The data suggested that a4 R449H -2FLAG colocalized e Figure 2. a2 P405L is degraded in the proteasomal pathway with some lysosomal contribution, whereas a4 R449H is degraded only in the proteasomal pathway. A, plot of quantified bands from anti-FLAG antibody-probed immunoblots of whole-cell lysates from WT a2-2FLAG and a2 P405L -2FLAG-transfected HEK 293 cells. Cells were treated with CHX (10 g/ml) for the times indicated, with and without proteasome inhibitor (designated MG) as indicated. B, same as panel A, but cells were transfected with WT a4 -2FLAG and a4 R449H -2FLAG constructs. C, same as panel A, but cells were treated with lysosomal inhibitor (designated Am) rather than proteasomal inhibitor. D, same as panel B, but cells were treated with lysosomal inhibitor. Data were normalized to GAPDH and zero time control and are representative of three independent biological experiments; error bars indicate Ϯ S.D.

Functional domains of ATP6V0A2 and ATP6V0A4
with the ER marker, calnexin, at a higher rate than the WT a4 -2FLAG (p Ͻ 0.05), whereas a4 G820R -2FLAG was similar to WT a4 -2FLAG in this respect (p ϭ 0.081). Fig. 3D shows representative micrographs of the same cell series as in Fig. 3C, but probed with anti-FLAG (green) and anti-syntaxin 6 (red) antibodies. The mutant protein, a4 R449H -2FLAG, colocalized with syntaxin 6 at a lower rate than the WT a4 -2FLAG (p Ͻ 0.05), whereas a4 G820R -2FLAG was again similar to the WT a4 -2FLAG in this respect (p ϭ 0.090). Fig. 3E shows colocalization analysis of images represented in Fig. 3, A and B, revealing that a2 P405L -2FLAG colo-calized with calnexin in the ER, the same as WT a2-2FLAG. The localization of a2 P405L -2FLAG to Golgi (syntaxin 6), however, was reduced with reference to the wildtype (p Ͻ 0.001; r ϭ 0.5-0.8). Similarly, Fig. 3F shows colocalization analysis of images represented in Fig. 3, C and D, revealing significant retention of a4 R449H -2FLAG in the ER, and significantly lower association with the Golgi marker, compared with WT a4 (p Ͻ 0.001; r ϭ 0.5-0.8). The a4 G820R mutant, on the other hand, was indistinguishable from wildtype in these respects (p ϭ 0.081 for calnexin, p ϭ 0.090 for syntaxin 6). Ordinate is Pearson's correlation coefficient (r). Results show that a2 P405L colocalized with calnexin at a rate similar to that of WT a2, but there was significantly less colocalization with syntaxin 6 compared with WT a2. F, quantitative colocalization analysis of data in panels C and D. Results show that a4 R449H is mostly retained in the ER. Images are representative of 20 images (10 -15 cells/image) each from three independent biological experiments.

Functional domains of ATP6V0A2 and ATP6V0A4
Defective cell-surface expression of a4 R449H As shown above, a4 R449H was unstable relative to WT a4, was retained in the ER, and was ultimately degraded in the proteasome. To further its characterization, it was of interest to determine whether any of the mutant protein was able to traffic to its normal location at the cell surface. To assess cell-surface expression (Fig. 4), a4 was tagged with both HA in extracellular loop II (ELII) and FLAG at the end of the C terminus. We have shown ELII is the site of N-glycosylation within subunit a1-a4 (25,26) indicating that ELII is luminal/extracellular. In contrast, we and others have shown that the C-terminal domain is cytoplasmic (12,25,26). Comparing the accessibility of either epitope in permeabilized versus non-permeabilized cells can determine whether a4 is expressed on the cell surface. In permeabilized cells, one would expect that both cytoplasmic and extracellular epitopes would be assessable to fluorescently-labeled antibodies; in non-permeabilized cells, only HA on the extracellular EL2, would be available. Fig. 4, A-D, shows representative fluorescence micrographs of HEK 293 cells transfected with WT a4 -3HA-2FLAG, a4 R449H -3HA-2FLAG, or a4 G820R -3HA-2FLAG, double-stained with anti-HA (red) on non-permeabilized cells followed by cell permeabilization and staining with anti-FLAG (green). Total protein expression is represented by anti-FLAG (green) staining, and cell-surface expression by anti-HA (red) staining. Fig. 4A shows empty vector-transfected (control) cells stained, Fig. 4B shows intracellular as well as cell-surface expression for cells transfected with WT-a4 -3HA-2FLAG, and Fig. 4C shows only intracellular expression in cells transfected with WT-a4 R449H -3HA-2FLAG, with no cell-surface expression detected. Fig. 4D shows intracellular, as well as cell-surface expression for cells transfected with WT-a4 G820R -3HA-2FLAG, a mutant that has a half-life similar to that of WT a4.
To confirm the above findings for a4 subunits, which are expected to traffic ultimately to the plasma membrane, cellsurface proteins of intact cells were biotinylated, and the biotinylated proteins were then affinity purified for further assessment (see "Experimental procedures"). Fig. 4E shows an immunoblot of the whole-cell lysates and the cell-surface fraction from cells that were transfected with WT a4 -2FLAG, a4 R449H -2FLAG, or a4 G820R -2FLAG. WT a4 -2FLAG and a4 G820R -2FLAG were expressed on the surface, as expected, but there was no cell-surface expression of a4 R449H -2FLAG. This result confirms the immunofluorescence findings in Fig. 4, A-D, suggesting that a4 R449H is largely retained in the ER.

a4 R449H shows increased association with VMA21
As demonstrated above, a2 P405L -2FLAG and a4 R449H -2FLAG had substantially shorter half-lives and defective Golgi and ER localization, as compared with their wildtype counterparts. It was of interest to characterize the effect of these mutations on their incorporation into the V-ATPase complex. The assembly of human V-ATPase is not well characterized, but studies in yeast have revealed that biosynthesis of V 0 in the ER is dependent on three assembly factors, Vma12p, Vma21p, and Vma22p (7). Due to the high homology between mammalian a subunit and the yeast ortholog, Vph1p, a similar biosynthetic mechanism was expected. VMA21, the human ortholog of yeast Vma21p, is the only characterized human V-ATPase assembly factor (29). VMA21 is required for incorporation of the a subunit into the V 0 subcomplex, but the dissociation of the a subunit from V 0 is required for further V 1 -V 0 assembly. Thus, prolonged association of VMA21 with V 0 inhibits the formation of the V-ATPase holocomplex (30,31). Fig. 5, A-C, show representative immunoblots loaded with immunoprecipitated fractions that were pulled down with anti-FLAG antibody from lysates of HEK 293 cells transfected with WT a2-2FLAG, a2 P405L -2FLAG, WT a4 -2FLAG, a4 R449H -2FLAG, or a4 G820R -2FLAG and immunoblotted with either anti-VMA21 or anti-B1 antibodies. Protein band quantification

Functional domains of ATP6V0A2 and ATP6V0A4
analysis (Fig. 5, D and E) showed a difference between the mutant protein, a4 R449H -2FLAG, and its wildtype counterpart. The mutant had a significantly higher association (p Ͻ 0.05) with VMA21 (representing a-V 0 assembly), and a lower association with B1 (representing V 1 -V 0 assembly) compared with wildtype a4 (Fig. 5E). Interestingly, there was no significant difference (p ϭ 0.12) between the association of a4 G820R -2FLAG or a2 P405L -2FLAG with either B1 or VMA21, compared with their wildtype counterparts.
a4 Gly-820 resides within the putative proton pathway As shown above, the a4 G820R -2FLAG, compared with WT, showed only a small or insignificant difference in terms of protein stability, localization in the secretory pathway, or cell-surface expression. Therefore, it remained of interest to determine the mechanism by which the a4 G820R mutation causes dRTA. In an attempt to address this, we constructed a homology model for the CTa domain of the human a4 subunit, based on a recent model for the CTa domain of yeast Vph1p. The latter was built based on low-resolution X-ray crystallography, high-resolution cryo-EM, mutagenesis studies, and analysis of evolutionary covariance (23). This model showed the locations of highly conserved, key functional residues within the proton translocation pathway, or proton channel. In a similar manner, we reconstructed the same residues within our human a4 model, and found that the a4 Gly-820 residue was located within the putative interface of the proton translocation pathway (Fig. 6, A and  B). We created a second homology model for the a4 G820R mutant protein (Fig. 6C) and showed that the positivelycharged side chain of the mutant a4 Arg-820 residue possibly interferes with the proton pathway by forming a salt bridge (3.2 Å) with the adjacent negatively-charged residue, Glu-729. The latter amino acid has been previously recognized as an important residue for proton translocation (32).

a2 Pro-405, a4 Arg-449, and a4 Gly-820 are conserved and crucial for function
Mutation of the V-ATPase a2 subunit amino acid residue Pro-405 results in cutis laxa, and a4 mutations in residues Arg-449 and Gly-820 result in dRTA. In an effort to understand how these missense point mutations can lead to disease, we first conducted multiple amino acid alignments, which revealed that the residues of interest were highly conserved (Fig. 1, A and B). The a2 Pro-405, a4 Arg-449, and a4 Gly-820 residues reside within TM1, TM3, and the CTD, respectively, which are highly conserved domains in species ranging from human to yeast. By characterizing the effects that these mutations have on a subunit glycosylation, structural stability, trafficking, and assembly, we hoped to elucidate their disease mechanisms and also add to the as yet limited understanding of structural/functional domains within human V-ATPases, ultimately to provide a basis for rational drug design.

Human a2 P405L and a4 R449H are N-glycosylated, but are unstable and a4 R449H degraded predominantly in the proteasomal pathway
We have previously shown that human a1-a4 subunits are N-glycosylated and that this is important for subunit stability (25,26). In the present study, results of Fig. 1, C and E, show that

Functional domains of ATP6V0A2 and ATP6V0A4
mutant proteins, a2 P405L , a4 R449H , and a4 G820R , were all N-glycosylated, and all with Endo H-sensitive high-mannose or hybrid glycan moieties. Moreover, a2 P405L and a4 R449H , but not a4 G820R , showed a much higher rate of turnover (i.e. decreased stability) relative to their respective wildtype subunit (Fig. 2). These results also showed that turnover rates of a2 P405L and a4 R449H could be restored to wildtype levels by treatment with the proteasomal inhibitor, MG132. Treatment with the lysosomal inhibitor, NH 4 Cl, had no significant effect on the turnover rate of a4 R449H and modestly reduced the turnover rate of a2 P405L . This suggested that a4 R449H was degraded predominantly in the proteasomal pathway, which is activated in response to the presence of misfolded proteins in the ER (33), with some degradation of the former occurring also in the lysosomal pathway.
Within this study, we tagged both WT and mutant subunits with C-terminal epitopes. We, and others, have shown that a variety of different epitope types and sizes inserted at the extreme C-terminal domain of the mammalian V-ATPase a subunit does not appear to affect activity or stability (25, 26, 34 -36). In yeast, we were able to show that introducing green fluorescent protein, a 238-amino acid, 26.9-kDa polypeptide, to the C-terminal of Vph1p, the yeast V-ATPase a subunit, did not affect subunit stability, assembly, function, and trafficking with respect to endogenous Vph1p (12).

a2 Pro-405 is required for Golgi trafficking, and a4 Arg-449 for ER exit
The relatively high degradation rates of both a2 P405L and a4 R449H in the proteasomal pathway suggested that these sub- Figure 6. a4 Gly-820 resides within the putative proton translocation pathway. A, homology model for C-terminal integral membrane (CTa) domain of the human a4 subunit. This model was constructed based on the recent high-resolution cryo-EM structure and evolutionary covariance analysis for the yeast a subunit, Vph1p (23). The indicated residues are highly conserved and essential for proton translocation. The red dashed line shows the hypothetical proton channel from the cytoplasmic side of the membrane to the luminal space. Cyan dashed box indicates the subregion where amino acid residue Gly-820 is located. B, shows the close proximity of Gly-820 and the highly conserved Glu-729, a residue thought to be key in proton translocation (10,32,38). C, illustrates how the G820R mutation may result in a salt-bridge interaction (red asterisk) between Glu-729 and Arg-820, possibly distorting or blocking the proton channel, resulting in inhibition of proton translocation, and providing a causative explanation for the role of the a4 G820R mutation in dRTA.

Functional domains of ATP6V0A2 and ATP6V0A4
units fail to assemble into the V-ATPase complex and therefore fail to traffic to their normal destinations. Despite the higher turnover rate of a2 P405L relative to the wildtype (Fig. 2, A and C), however, quantification of colocalization of a2 P405L with calnexin showed no significant difference in association of a2 P405L with calnexin, compared with WT a2 (p ϭ 0.073). Additionally, however, a2 P405L showed significantly less association (p Ͻ 0.05) with the Golgi marker, syntaxin 6 ( Fig. 3E), which suggested that the a2 P405L mutation results in misprocessing that leads to defective Golgi trafficking, but not ER retention. In contrast, quantification of colocalization analysis for a4 R449H and a4 G820R with the ER-resident marker, calnexin (Fig. 3, C  and F), revealed a significantly higher colocalization of a4 R449H with calnexin (p Ͻ 0.05), suggesting ER retention of a4 R449H , but not of a4 G820R , which was not different from wildtype a4 in that respect (p ϭ 0.081). However, the exact mechanism of a4 R449H ER retention remained to be investigated. Taken together, these observations suggest that the a2 Pro-405 and a4 Arg-449 residues within TM1 and TM3, respectively, are essential for human a2 and a4 stability, and for their trafficking in the secretory pathway.

a4 Arg-449 is crucial for cell-surface expression and a-V 0 association
We have previously shown experimentally that the exogenously expressed WT a4 is able to traffic to the plasma membrane of HEK 293 cells (25). In the current study we have used the same strategy to determine the effect of the mutations in a4 R449H and a4 G820R on cell-surface expression. Fig. 4, B and D, showed that both WT a4 and a4 G820R were able to traffic to the cell surface, whereas a4 R449H showed defective cell-surface expression (Fig. 4C). The same findings were subsequently confirmed by cell-surface biotinylation (Fig. 4E).
It was of interest also to determine the effect of the mutations under investigation on formation of the V-ATPase holocomplex. To that end, we specifically characterized the association of a2 P405L , a4 R449H , and a4 G820R with the only characterized human V-ATPase assembly factor, VMA21. Protein band quantification of co-immunoprecipitates (Fig. 5E) revealed that a4 R449H had a significantly higher association (p Ͻ 0.05) with VMA21. In yeast, the assembly chaperone Vma21p assembles with V 0 -associated a subunits, and dissociates only after V 0 exits the ER (37); dissociation of Vma21p from V 0 is required for V 1 -V 0 assembly, and prolonged Vma21p-V 0 association reduces V 1 -V 0 assembly. We propose that the significantly higher association observed between a4 R449H and VMA21 indicates a prolonged association of a4 R449H -V 0 with VMA21 that leads to failure of V 1 -V 0 assembly, ER retention of a4 R449H , and ultimately its proteasomal degradation, resulting in defective cell-surface expression.

a4 Gly-820 is a functional residue residing in the putative proton pathway
The a4 Gly-820 residue is highly conserved among species (Fig. 1B). Due to the lack of a mammalian model, the mechanism of the a4 G820R dRTA-causing mutation has been studied previously only in yeast. One of these studies reported that the a4 G820R mutation in the yeast homolog, Vph1p, did not affect pump assembly or targeting but decreased V-ATPase hydrolytic and proton pumping activities by 83-85% (10). Another study in the yeast a subunit showed that the a4 G820R homologous mutation (Vph1p G812R ) was associated with severe loss of proton translocation (by 78%) and a moderate decrease in ATPase activity (by 36%). This study also showed that the a4 Gly-820 residue lies within the domain that interacts with the glycolytic enzyme, phosphofructokinase-1 and that the a4 G820R equivalent mutation inhibited this interaction (38). In the present work we used exogenous expression in HEK 293 cells to investigate the role of this mutation in protein stability, glycosylation, and trafficking in the secretory pathway and the plasma membrane, and our results showed that the stability of a4 G820R was only mildly affected (Fig. 1F) and trafficking to the Golgi and plasma membrane were not discernably altered (Figs.  3D and 4D).
In an attempt to obtain further insights into how a4 G820R might impact V-ATPase function, we used the recently published atomic model synthesized from studies of Thermus thermophilus and Saccharomyces cerevisiae V-ATPase and bovine F-ATPase (23) as a template to construct a 3D human a4 C-terminal domain model (Fig. 6A). The model revealed that the a4 Gly-820 residue interfaces with the proposed proton transport pathway. Furthermore, swapping arginine for glycine (a4 G820R) resulted in a putative salt bridge (3.2 Å) with the adjacent negatively charged residue a4 Glu-729 (compare Fig. 6, B with C), which is also highly conserved and is thought to be important for proton translocation (32). Therefore, we proposed that the a4 G820R mutation likely causes dRTA by forming a salt bridge that sterically interferes with the structure of the proton channel and consequently with proton translocation. It might do this directly, by blocking the proton channel physically (i.e. in its immediate vicinity), or allosterically by altering the conformation of the CTD more extensively. Some conformational change must be occurring to be consistent with the previous observation that the mutation also inhibits the interaction of the a subunit with phosphofructokinase-1 (38), the binding of which must occur at a cytoplasmically accessible site; however, this does not preclude the possibility that the putative salt bridge resulting from the G820R mutation both blocks the proton pathway directly and causes extended conformational changes in the CTD.

Conclusion
Characterization of highly conserved residues implicated in diseases has been successfully used by others as a strategy for determining protein domain function and to inform targeted drug discovery (39 -41). For example, deletion of the highly conserved residue Phe-508 (⌬F508) in the human cystic fibrosis transmembrane-conductance regulator (CFTR) leads to cystic fibrosis. The ⌬F508 mutation results in protein misfolding, misprocessing, and aberrant trafficking (42). Characterization of the molecular mechanism of ⌬F508 CFTR disease causation has led to the development of molecular chaperone approaches to correct CFTR folding and promote its trafficking to its normal functional destination, yielding a promising approach for treatment of ⌬F508 cystic fibrosis (41).

Functional domains of ATP6V0A2 and ATP6V0A4
V-ATPase a isoforms are potential targets for therapeutics directed toward a number of diseases (1). Thus, a further understanding of the structural domains affecting a subunit folding, trafficking, membrane targeting, function, and regulation will enhance our ability to target specialized V-ATPases. We previously showed that N-glycosylation is required for a subunit stability, assembly, and trafficking to the plasma membrane (25,26). In the present work we showed that a2 P405L and a4 R449H resulted in cutis laxa and dRTA through interfering with protein stability, and subsequent ER retention and degradation. a4 R449H was degraded predominantly in the proteasomal pathway, whereas a2 P405L was degraded in both proteasomal and lysosomal pathways. In summary, we have proposed a model for how we believe that the N-glycosylated a4 subunit is assembled, trafficked in the secretory pathway, and delivered to the plasma membrane (see Fig. 7). Our data also suggest routes to drug discovery such as screening for chemical chaperons to rescue a subunit folding to allow ER exit for treatment of cutis laxa and dRTA.

Enzymes and reagents
Restriction enzymes, Endo H (catalog number P0702S), and PNGase F (number P0704S) were from New England Biolabs (Whitby, Canada). Octaethylene glycol mono-n-dodecyl ether (C 12 E 8 ) was from NIKKO Chemicals (Barnet Products, Englewood Cliffs, NJ). Bradford protein assay reagent (500-0006) was from Bio-Rad (Mississauga, Canada), 4Ј,6-diamidine-2Ј-phenylindole dihydrochloride (DAPI; 10236276001) was from a Figure 7. Model for human a4 trafficking in the secretory pathway and to the plasma membrane. Steps 1-7 and 6Ј-8Ј suggest two putative mechanisms of mammalian V-ATPase assembly. First, within the ER, the VMA21 assembly factor facilitates the assembly of subunit a into the V 0 subcomplex (steps 1-3). The assembled V 0 subcomplex is subsequently trafficked to Golgi (step 4), assembled with the V 1 subcomplex within the Golgi (steps 5 and 6), with the fully assembled V-ATPase complex targeted to the plasma membrane (step 7). Alternatively, the V 0 subcomplex itself could traffic to the plasma membrane (7Ј) and only assemble with the V 1 subcomplex at the plasma membrane (steps 6Ј-8Ј). A, an unglycosylated mutant, a4 N489D , which was described in a previous study (25), is unable to assemble into a V 0 subcomplex. It is retained in the ER and is targeted to the ERAD pathway for proteolysis. In contrast, the glycosylated a4 R449H mutant described here assembles within the V 0 complex; however, it is ultimately degraded in the proteasome, and thus also fails to reach the plasma membrane. B, the glycosylated a4 G820R assembles within the V 1 V 0 complex and is trafficked to the plasma membrane but, unlike the wildtype complex, functional proton translocation appears to be inhibited by the mutation. Red asterisks symbolize the a4 G820R mutation; a dark green circle symbolizes the a4 R449H mutation; red bars indicate blockade of a pathway; V 1 subunits are indicated by uppercase letters; V 0 subunits are indicated by lowercase italic letters.

Cell culture and transfection
Liquid nitrogen-stored HEK 293 cells were rapidly thawed in a water bath at 37°C followed by incubation in 75-cm 2 tissue culture flasks containing 17 ml of DMEM, supplemented with a 10% fetal bovine serum and 1% penicillin/streptomycin mixture, in a humidified 5% CO 2 incubator for 4 days at 37°C. The cells, at 70 -80% confluence, were trypsinized with 1 ml of 1ϫ trypsin/EDTA and seeded into 6-well plates at a density of 4 -7 ϫ 10 5 cells/well and incubated for 24 h. Cells were subsequently transiently transfected with 1 g/well of plasmid construct in a transfection complex containing GenJet reagent and plasmid DNA in a 3:1 ratio. The transfection complex was diluted to 200 l of final volume with serum-free DMEM and incubated for 10 min prior to transfection. Post-transfection cells were incubated for 24 h and then harvested for protein expression analysis. There was no significant difference (p Ͻ 0.05) in cell viability between HEK 293 cells transfected with WT, mutant, or empty vectors (data not shown).

Protein expression analysis and assessment of glycosylation
Whole-cell lysates were prepared as previously described (25). Briefly, cells were harvested in 0.2 ml/well of lysis buffer (PBS containing 1% C 12 E 8 , 1 mM PMSF, and 1:100 (v/v) Protease Inhibitor Mixture) and incubated on ice for 30 min. Lysates were then centrifuged at 15,000 ϫ g for 30 min at 4°C, and supernatants were collected for further analysis. Protein concentrations of the supernatants were quantified using the Bradford protein assay.
Protein glycosylation was assessed by treatment of samples with either PNGase F or Endo H. Briefly, 30 g of whole-cell lysate was denatured in 3 l of 10ϫ glycoprotein denaturation buffer (5% sodium dodecyl sulfate, 0.4 M dithiothreitol; New England Biolabs), the reaction mixture was adjusted to 20 l and incubated at 65°C for 10 min, then 2 l of 10ϫ Glyco Buffer was added (for PNGase F, 0.5 M sodium phosphate, pH 7.4, at 25°C; for Endo H, 0.5 M sodium citrate, pH 7.5, at 25°C). Subsequently, 2 l of 10% (w/v) Nonidet P-40 (New England Biolabs) and 2,000 units of PNGase F, or Endo H, were added. The final volume was adjusted to 40 l with distilled H 2 O, incubated for 1 h at 37°C, and then analyzed by immunoblotting.

Immunoblotting
Immunoblotting was conducted as previously described (25). Briefly, 30 g of whole-cell lysate was loaded per well and subjected to 7% SDS-PAGE. Proteins were then transferred to nitrocellulose membrane and incubated overnight at 4°C with 1:2,000 -1:3,000 diluted primary antibodies (anti-FLAG, anti-B1, or anti-VMA21). 1:5,000 diluted anti-GAPDH was used in some experiments to provide loading controls. The blots were then incubated for 1 h at room temperature with 1:5,000 HRPlabeled secondary antibody and bands were developed with chemiluminescent substrate reagent.

Protein stability and protein band quantification
Protein stability was evaluated using the CHX chase assay. Briefly, HEK 293 cells were transfected with WT and mutant cDNA constructs and 24 h post-transfection the cells were Functional domains of ATP6V0A2 and ATP6V0A4 treated with 10 g/ml of CHX with or without proteasomal inhibitor (10 M MG132), or lysosomal inhibitor (25 mM NH 4 Cl), for up to 12 h. The cells were subsequently harvested and whole-cell lysates were prepared for immunoblotting with anti-FLAG, and anti-GAPDH as a loading control.
Protein band quantification of CHX immunoblots was performed using Bio-Rad Quantity One 4.6.9 software. Briefly, band intensities were quantified after subtracting the background signals from band signals using the rolling-disc method. Relative protein levels were estimated after normalizing band intensities relative to GAPDH loading controls and zero time controls. Glycoproteins tend to run in SDS-PAGE as diffuse bands so, for more accurate comparison of unglycosylated bands with the more similar deglycosylated protein bands, whole-cell lysates were treated with PNGase F prior to immunoblotting to remove glycan moieties, yielding uniformly sharp protein bands.
Statistical analysis of the CHX chase data were done using GraphPad Prism 5 software. Non-linear curve fitting was done assuming a simple exponential one-phase decay model. Prior subtraction of background was accommodated by modifying the default model equation in GraphPad Prism as follows: exponential/one-phase decay model, Y ϭ (Y0) ϫ exp (ϪK ϫ X). Automated curve fitting and non-linear regression analysis provided half-life times (h). For time 0, mean Ϯ S.D. were obtained from data normalized to GAPDH only, then applied proportionately to the zero points normalized for GAPDH and zero time (i.e. 1.0). Data in figures were plotted point-to-point rather than as fitted exponential curves to preserve clarity of the original data. Mean Ϯ S.D. values were derived from three independent experiments, and p values, representing significance of differences for comparisons, were derived from unpaired, two-tailed Student's t tests. Data analyses, including raw data, correlation coefficients (R 2 ) of fit of the one-phase exponential decay model, and half-life values and their standard deviations, are tabulated in supporting Table S1. The derived p values for comparisons are tabulated in supporting Table S2.

Co-immunoprecipitation
HEK 293 cells were transfected with WT and mutants cDNA constructs and whole-cell extracts were prepared in IP buffer (150 mM NaCl, 25 mM Tris HCl, pH 7.2, at 25°C, containing 1% C 12 E 8 , 1:100 (v/v) Protease Inhibitor Mixture and 1 mM PMSF), as previously described (25). Co-immunoprecipitation of WT and mutants was conducted by treating 50 g of whole-cell lysate with 5 g of anti-FLAG antibody and incubating overnight at 4°C with agitation. Antigen-anti-FLAG antibody immunocomplexes were pulled down by incubation with 100 l (50% packed volume) of protein A-agarose beads for 2 h at room temperature with agitation. The antigen-coated beads were then incubated 5 min with SDS-PAGE sample buffer at 95°C to elute the antigens. Antigen-containing supernatants were collected after centrifugation at 2,500 ϫ g for 3 min and then immunoblotted with anti-FLAG, anti-B1, and anti-VMA21 antibodies.

Immunofluorescence and colocalization analysis
HEK 293 cells were grown on glass coverslips and transiently transfected with WT and mutant cDNA constructs. The cells were washed with DPBS and fixed with 3.7% (w/v) paraformaldehyde for 15 min at room temperature. Subsequently, cells were permeabilized with DPBS containing 0.2% Triton X-100 at room temperature for 15 min. Cells were then blocked with DPBS containing 5% bovine serum albumin for 1 h at room temperature, followed by immunostaining with anti-FLAG (1:1,000), anti-calnexin (1:500), or anti-syntaxin 6 (1:500) antibodies in DPBS containing 5% bovine serum albumin for 45 min at room temperature. Cells then were washed 3 times with DPBS and immunostained with fluorescent second antibodies (1:500) for 45 min at room temperature. Nuclei were stained with 0.1 mg/ml of DAPI in DPBS for 10 min and cells were mounted with ProLong Gold Antifade Reagent (Fisher Scientific). Photomicrography images were acquired using a Quorum Spinning Disk Confocal System equipped with a Hamamatsu C9100-13 EM-CCD, Yokogawa CSU X1 scan head, and Improvision Piezo focus drive (Imaging Facility, Hospital for Sick Children, Toronto, Canada).
Colocalization quantification of 20 images (10 -15 cells/image) each from three independent experiments was conducted using Volocity version 6.3 3D image analysis software (PerkinElmer Life Sciences, Woodbridge, Canada). Colocalizations of two fluorescent signals (red and green) were quantified and expressed as Pearson's correlation coefficients (r). Significance of differences between WT and mutants were estimated using two-tailed Student's t tests.

Cell-surface biotinylation
Cell-surface labeling was performed using EZ-Link NHS SS-Biotin reagent (Pierce 21328; Fisher Scientific), as described previously (25). Briefly, HEK 293 cells were transiently transfected and, 24 h post-transfection, the cells were incubated with 1 mg/ml of freshly prepared EZ-Link NHS-SS-Biotin for 1 h at 4°C with gentle agitation. The cells were then incubated with ice-cold quenching buffer (192 mM glycine, 25 mM Tris, pH 8.3, at 25°C) to remove excess biotin. Cells were harvested in 0.4 ml of ice-cold RIPA buffer (150 mM NaCl, 1% sodium deoxycholate, 0.1% SDS, 1% Triton X-100, 1 mM EDTA, and 10 mM Tris-HCl, pH 7.5, at 25°C) containing Protease Inhibitor Mixture (1:100 v/v) and 1 mM PMSF, and were incubated for 30 min on ice, then centrifuged at 15,000 ϫ g for 30 min at 4°C. Supernatants were collected, and biotinylated cell-surface proteins were affinity purified by incubating supernatants with 100 l (50% packed volume) of streptavidin-agarose beads (Pierce 20347; Fisher Scientific) for 2 h at 4°C. The eluted, biotinylated cell-surface proteins and total lysate proteins were analyzed by 7% SDS-PAGE and immunoblotted, as previously described (25).

Structural modeling of human a4 subunit
Homology modeling of the integral membrane domain of the human a4 subunit was generated by SWISS-MODEL, using the yeast a subunit ortholog, Vph1p, as a template (PDB code 5I1M) (23,43). Subsequently, the model was corrected and the 3D representation was generated using the 3D graphical

Functional domains of ATP6V0A2 and ATP6V0A4
YASARA interface (44). A 3D representation of a4 G820R was generated after substituting Gly-820 with Arg, using the YASARA FoldX plug-in.
Author contributions-S. E., N. K., R. A. F. R., and M. F. M. conceptualized, planned, and analyzed experimental work. S. E. performed all experiments except those represented in Fig. 1C, which were performed by J. W. K.; S. E. and Y. Y. designed and prepared constructs used in the study; and Y. Y. provided additional technical expertise. S. E. and N. K. wrote the manuscript and prepared figures.