Osteopetrosis Mutation R444L Causes Endoplasmic Reticulum Retention and Misprocessing of Vacuolar H+-ATPase a3 Subunit*

Background: The human V-ATPase a3 subunit mutation, R444L, causes infantile malignant osteopetrosis. Results: In mouse, the R444L equivalent, R445L, causes endoplasmic reticulum retention, misprocessing, and defective trafficking of a3 to the plasma membrane. Conclusion: Arginine 444/445 plays a critical role in mammalian a3 folding, or stability. Significance: R444L a3 infantile malignant osteopetrosis is a protein folding disease that may be amenable to protein rescue therapy. Osteopetrosis is a genetic bone disease characterized by increased bone density and fragility. The R444L missense mutation in the human V-ATPase a3 subunit (TCIRG1) is one of several known mutations in a3 and other proteins that can cause this disease. The autosomal recessive R444L mutation results in a particularly malignant form of infantile osteopetrosis that is lethal in infancy, or early childhood. We have studied this mutation using the pMSCV retroviral vector system to integrate the cDNA construct for green fluorescent protein (GFP)-fused a3R445L mutant protein into the RAW 264.7 mouse osteoclast differentiation model. In comparison with wild-type a3, the mutant glycoprotein localized to the ER instead of lysosomes and its oligosaccharide moiety was misprocessed, suggesting inability of the core-glycosylated glycoprotein to traffic to the Golgi. Reduced steady-state expression of the mutant protein, in comparison with wild type, suggested that the former was being degraded, likely through the endoplasmic reticulum-associated degradation pathway. In differentiated osteoclasts, a3R445L was found to degrade at an increased rate over the course of osteoclastogenesis. Limited proteolysis studies suggested that the R445L mutation alters mouse a3 protein conformation. Together, these data suggest that Arg-445 plays a role in protein folding, or stability, and that infantile malignant osteopetrosis caused by the R444L mutation in the human V-ATPase a3 subunit is another member of the growing class of protein folding diseases. This may have implications for early-intervention treatment, using protein rescue strategies.

Osteopetrosis is a genetic bone disease characterized by increased bone density and fragility. The R444L missense mutation in the human V-ATPase a3 subunit (TCIRG1) is one of several known mutations in a3 and other proteins that can cause this disease. The autosomal recessive R444L mutation results in a particularly malignant form of infantile osteopetrosis that is lethal in infancy, or early childhood. We have studied this mutation using the pMSCV retroviral vector system to integrate the cDNA construct for green fluorescent protein (GFP)-fused a3 R445L mutant protein into the RAW 264.7 mouse osteoclast differentiation model. In comparison with wild-type a3, the mutant glycoprotein localized to the ER instead of lysosomes and its oligosaccharide moiety was misprocessed, suggesting inability of the core-glycosylated glycoprotein to traffic to the Golgi. Reduced steady-state expression of the mutant protein, in comparison with wild type, suggested that the former was being degraded, likely through the endoplasmic reticulum-associated degradation pathway. In differentiated osteoclasts, a3 R445L was found to degrade at an increased rate over the course of osteoclastogenesis. Limited proteolysis studies suggested that the R445L mutation alters mouse a3 protein conformation. Together, these data suggest that Arg-445 plays a role in protein folding, or stability, and that infantile malignant osteopetrosis caused by the R444L mutation in the human V-AT-Pase a3 subunit is another member of the growing class of protein folding diseases. This may have implications for early-intervention treatment, using protein rescue strategies.
Mammalian V-ATPases contain 14 different subunits organized into a cytoplasmic V 1 sector (subunits A-H), and a membrane-embedded V 0 sector (subunits a, c, cЉ, d, e, and Ac45) (10). Several of the V-ATPase subunits have multiple isoforms that may be expressed in a tissue, cell-type, or organelle-specific manner. In yeast, for example, the a subunit, which plays a direct role in proton translocation, has two isoforms, Vph1p and Stv1p. Vph1p is localized to the vacuole, whereas Stv1p is found primarily in Golgi (11,12). The mammalian a subunit has four isoforms, a1-4; a1, a2, and a3 are ubiquitously expressed, but to different degrees in different tissues and organelles, whereas a4 expression appears to be specific to plasma membranes of renal intercalated cells (9,(13)(14)(15). Although ubiquitously expressed, a3 appears to be most highly enriched in osteoclasts (9). In actively bone-resorbing osteoclasts, V-ATPases containing the a3 subunit isoform are specifically targeted to the osteoclast ruffled border, where they are involved in acidifying the resorption lacuna to demineralize bone (15).
The importance of the a3 subunit in bone biology has been demonstrated in mouse models: a3 knock-out (16), the oc/oc truncation mutant (17), or point mutations, e.g. R740S, at the Arg residue critical for proton translocation (18), lead to severe osteopetrosis, a bone disease characterized by increased bone density and fragility due to the inability of osteoclasts to secrete acid to resorb bone. This illustrates that, despite its ubiquitous expression, the critical function of a3 is its role in proton transport within the osteoclast ruffled border. Other a3 functions, such as its involvement in lysosomal acidification * This work was supported in part by Operating Grant FRN-79322 from the Canadian Institutes of Health Research. 1  apparently can be complemented by V-ATPases with alternate a subunit isoform composition (19). In humans, type 1 infantile malignant osteopetrosis (OPTB1; OMIM #259700) is caused by autosomal recessive mutations in the TCIRG1 gene (ATP6V0A3; 11q13.2), which codes for the a3 subunit isoform of the V-ATPase complex. This particular form of autosomal recessive osteopetrosis (ARO) is a rare and severe disease caused by a defect in competence of osteoclasts for resorption of bone (20). It results in dense, brittle bone with severe encroachment of marrow cavities of the long bones, leading to anemia and thrombocytopenia with compensatory hepatosplenomegaly. Associated cranofacial bone abnormalities also lead to hydrocephalus, nasal obstruction, and nerve compression resulting in progressive deafness and blindness. Functions of monocytes and macrophages are also compromised and most children with this affliction die in infancy, or early childhood, often of infection, unless early intervention by bone marrow transplantation is implemented (21)(22)(23).
Many ARO mutations in TCIRG1 are in regulatory motifs or splice sites, or result in large deletions or truncations; however, the missense mutation R444L has been reported in the Costa Rican population (24). Study of this mutation could shed light on the underlying cause of a3 R444L -mediated ARO, but also has more general implications for understanding the role of the highly conserved Arg-444 in the structure and function of V-ATPase a subunits. Indeed, a homologous mutation in the kidney-specific a4 isoform (R449H) is also disease causing, resulting in autosomal recessive distal renal tubular acidosis (25). Thus, we constructed the homolog of the R444L mutation (R445L) in a mouse expression system to further elucidate its effect on a subunit protein structure and function, and the consequent cell biological mechanisms leading to disease.
Mammalian Primary Cells, Cell Lines, and Constructs-Bone marrow mononuclear (BMM) cells obtained from femurs of 2-month-old male (C3H ϫ FVB)F1 mice were plated at 10 6 cells/dish (10 5 /ml) in 100-mm tissue culture dishes, in ␣-MEM without nucleosides (Invitrogen, 12561) supplemented with 10% FBS, 10 g/ml of penicillin G, 50 g/ml of gentamicin, 30 ng/ml of amphotericin B, and 50 ng/ml of macrophage colony stimulating factor (Calbiochem) for 48 h. The medium was then changed and included, additionally, 200 ng/ml of RANKL. Cells were cultured further for 4 days prior to use. RAW 264.7 (RAW) cells and their virally transduced derivative cell lines were grown in high-glucose DMEM (Invitrogen, 11965) supplemented with 10% FBS, 100 units/ml of penicillin, and 100 g/ml of streptomycin. For differentiation into osteoclasts, RAW cell lines were plated in 100-mm tissue culture dishes at 7 ϫ 10 5 cells/dish and cultured for 5 days in the presence of 100 ng/ml of recombinant, soluble RANKL. For protein rescue experiments RAW cells were grown to 85% confluence, then the medium was replaced with regular medium (control), medium containing 10% glycerol, or medium containing 5% DMSO. Cells were then cultured for an additional 24 h at 37°C (control and 10% glycerol), or at 26°C (5% DMSO), prior to protein extraction.
The mouse a3 cDNA-containing vector, pcDNA3.1-a3, was a kind gift of Dr. Beth S. Lee (Ohio State University). The insert was subcloned into the pEGFP-N1 vector (Clontech) between EcoRI and SacII sites, generating the construct, pEGFP-N1-a3. Site-directed mutagenesis was used to introduce the R445L mutation into a3, yielding the construct, pEGFP-N1-a3 R445L . PCR was subsequently used to generate inserts from these constructs, with wild type, or mutant, cDNA inserts fused with GFP cDNA and flanked by XhoI and EcoRI restriction sites. These PCR products were subcloned into XhoI/EcoRI-cut pMSCVpuro vector (Clontech). The control GFP-expressing vector, pMSCV-EGFP, was a kind gift of Drs. Helen Sarantis and Scott D. Gray-Owen (University of Toronto). To construct this, EGFP was amplified from pEGFP-N1 with flanking BamHI and EcoRI sites. The PCR product was subcloned into BamHI/ EcoRI-cut pMSCVpuro. Construct inserts and flanking regions were confirmed by DNA sequencing.
Retroviruses were generated by cotransfecting GP-293 (Clontech) cells (30 -40% confluent in 6-well tissue culture plates) with 2 g of pVSV-G (Clontech) and 2 g of pMSCV-a3-GFP, pMSCV-a3 R445L -GFP, or pMSCV-GFP, a plasmid DNA with FuGENE HD (Roche Applied Science). The resulting cell supernatants were used to infect RAW cells (passage 3), which were selected in 7 g/ml of puromycin, 48 h post-infection. This resulted in establishment of the stable cell lines, RAW-a3-GFP, RAW-a3 R445L -GFP, and RAW-GFP, which were maintained in medium containing 4 g/ml of puromycin. Cloning primers, constructs, and cell lines are listed in Table 1.
Reverse Transcription and PCR-RNA was extracted from RAW-derived osteoclasts using TRIzol reagent (Invitrogen). The purified total RNA was digested with DNase I (Invitrogen) at a concentration of 1 unit per microgram of RNA and reverse transcription was performed using RevertAid H Minus reverse transcriptase (Fermentas). PCR was performed using HotStar Taq DNA polymerase (Qiagen) with primers for GFP and GAPDH (Table 1).
Whole Cell and Membrane Protein Extraction-To obtain whole cell lysates from retrovirally transduced RAW cell lines and osteoclasts derived from them, cells were washed twice with PBS and lysed in RIPA buffer (Cell Signaling Technologies, 9806) according to the supplier's instructions. To obtain whole cell lysates from BMM-derived osteoclasts, the cells were washed twice with ice-cold PBS and lysed in buffer containing 300 mM NaCl, 5 mM EDTA, 0.1% Triton X-100, 50 mM Tris-HCl, pH 7.4, at 25°C, with protease inhibitors, phosphatase inhibitors, and 1 mM PMSF (from 100 mM stock in anhydrous ethanol), added prior to use.
To isolate membrane proteins from transduced RAW cells in 100-mm tissue culture dishes, cells were washed twice with ice-cold PBS and scraped on ice in membrane protein collection buffer. The latter buffer consisted of 50 mM Na 2 HPO 4 , 10 mM KCl, 1.5 mM MgCl 2 , 2 mM DTT, 10 mM Tris-HCl, pH 7.4, at 25°C, with addition of mammalian protease inhibitor mixture (1:200, v/v) and 1 mM PMSF prior to use. The cell suspension was homogenized by passage 15-20 times through a 27.5-gauge syringe needle and then centrifuged at 5,000 ϫ g for 10 min. The resulting supernatant was collected and centrifuged at 100,000 ϫ g for 1 h and the membrane pellet was resuspended in 150 l of membrane protein collection buffer.
Deglycosylation, SDS-PAGE, and Immunoblotting of a3 and a3-GFP-Glycoproteins were deglycosylated with PNGase F, or Endo H, according to the supplier's instructions. Briefly, membrane pellets or whole cell lysates (15-20 g of total protein) were diluted to a final volume of 30 l in glycoprotein denaturing buffer. Samples were denatured at 65°C for 10 min, then 0.1 volume each of 10% Nonidet P-40 and ϫ10 G7 reaction buffer were added, followed by 1,000 units of PNGase F. The final volume was adjusted to 40 l with distilled water. The reaction mixture was incubated for 1 h at 37°C. Proteins were then solubilized by addition of 0.25 volume of 5ϫ SDS-PAGE gel-loading buffer, resolved by SDS-PAGE, and immunoblotted. Anti-GFP antibodies were used at a dilution of 1:500, and anti-GAPDH at 1:10,000, followed by appropriate HRPconjugated secondary antibody. Blots were developed with Western Lightning ECL detection solution (PerkinElmer Life Sciences) and images were acquired using the Bio-Rad Molecular Imager ChemiDoc XRS system.
Immunostaining, Confocal Microscopy, and Image Analysis-Cells expressing wild type a3-GFP, mutant a3 R445L -GFP, or control GFP, were cultured on glass coverslips for 48 h and then fixed in 2% paraformaldehyde in microtubule stabilizing buffer (MTSB; 127 mM NaCl, 5 mM KCl, 1.1 mM Na 2 HPO 4 , 0.4 mM K 2 HPO 4 , 2 mM MgCl 2 , 5.5 mM glucose, 20 mM PIPES, pH 7.4, at 25°C) for 20 min. Cells were permeabilized with 0.1% saponin in MTSB containing 100 mM glycine for 20 min, then blocked in 5% FBS and 0.05% saponin in MTSB at room temperature for 1 h. Coverslips were incubated with anti-LAMP2 (1:200), or anti-calnexin (1:100) antibodies in blocking buffer overnight, then washed with 0.05% saponin in MTSB and incubated with fluorescent second antibodies (1:500). The cells were then incubated in PBS with 0.1 g/ml of DAPI for 10 min for nuclear staining. Images were obtained with a Leica DMIRE2 inverted fluorescence microscope equipped with a Hamamatsu Back-Thinned EM-CCD camera and spinning disk confocal scan head.
Quantitative Confocal Image Analysis-Quantitative image colocalization analysis was performed with Volocity 5.2 software (PerkinElmer Life Sciences). Values shown in the present work are Pearson's correlation coefficients (r). A threshold of r Ͼ 0.6 ("better than moderately positive") was used to define colocalization, as described (26). Data were compared using two-tailed t tests to judge significance of differences, as appropriate.
Limited Proteolysis-Membrane fractions harvested as described above were lyophilized and then reconstituted in 100 l of proteolysis buffer (50 mM Na 2 HPO 4 , 10 mM KCl, 1.5 mM MgCl 2 , 10 mM Tris-HCl, pH 7.4, at 25°C). A final concentration of 2.5 g/ml, or 5 g/ml, of TPCK-trypsin (prepared in 20 mM CaCl 2 , 1 mM HCl, pH 3) was added to 15 g of membrane protein. The solution was incubated for 1 h at 37°C and trypsinolysis was stopped by adding 0.1 volume of 100 mM N ␣ -tosyl-L-lysine chloromethyl ketone hydrochloride (prepared in 1 mM HCl, pH 3). Trypsinolysed proteins were deglycosylated using PNGase F, as described above.
Quantitative Protein Band Analysis-Protein bands in immunoblot images were quantified using Bio-Rad Quantity One 4.6.9 software. Background was subtracted using the rolling disk method, and band intensities were normalized to GAPDH. Protein band intensities obtained after trypsinolysis and deglycosylation were normalized to control (untreated) bands and expressed as percentages. Band intensities were compared using two-tailed t tests to judge significance of differences, as appropriate.

Aberrant Expression of Mouse a3 R445L In Macrophages and
Osteoclasts-To investigate the functional effects of the human R444L mutation on the V-ATPase a subunit, the equivalent mutation, R445L, was engineered into a mouse a3 subunit expression system. As osteoclasts are the cells primarily affected by the R444L mutation, the mouse RAW macrophage cell line was chosen as the expression host. RAW cells can be differentiated into mature osteoclasts in the presence of RANKL, thereby providing an established in vitro osteoclastogenesis model (27). Murine wild type a3 and mutant a3 R445L constructs were expressed in RAW cells as C-terminal fusions with GFP to allow them to be distinguished from the native, endogenously expressed V-ATPase a3 subunit. Stable expression was obtained by pMSCV-derived retroviral transduction, followed by puromycin selection (see "Experimental Procedures"). Details of constructs and cell lines are summarized in Table 1.
RAW cell lines derived by this method were grown in the presence of RANKL to obtain mature osteoclasts. That fulllength a3 was being transcribed was demonstrated by reverse transcription of mRNA extracted from the RAW cell lines, followed by PCR using GFP primers, as shown in Fig. 1A. Significant differences in fused or control GFP expression among the cell lines were not apparent, relative to housekeeping GAPDH PCR products. Membrane-associated protein expression was confirmed by preparing microsomal membrane pellets and analyzing them by SDS-PAGE and immunoblotting (Fig. 1B). Immunostaining blots specifically for GFP precluded background from endogenous cellular a3 subunit expression. Furthermore, native a3 has a predicted polypeptide size of 93.4 kDa, whereas the predicted size for a3-GFP is 121 kDa. In Fig.  1B a sharp band was observed at 134 kDa in proteins derived from the RAW-a3-GFP cell line. A much more intense, diffuse band was observed at 152 kDa. We have shown in other work that the mouse V-ATPase a3 subunit is glycosylated at two sites on a single luminal loop of the membrane domain, 3 and Lee et al. (28) have shown that the mouse a1 subunit is also glycosylated. Based on these observations, the diffuse 152-kDa band has been identified here as the mature glycoprotein, whereas the 134-kDa band is the core-glycosylated protein, a biosynthetic intermediate that is expected to be present at lower steady-state levels in the endoplasmic reticulum (ER). For the RAW-a3 R445L -GFP cell line, expression of the mutant protein was observed only as the 134-kDa band, with no diffuse band apparent at 152 kDa. The mutant 134-kDa band was also present at a much lower intensity than was seen for the combined wild type bands.
Because the effects of the human R444L mutation are manifested in osteoclasts, it was of interest to determine whether the fate of the mutant a3 R445L -GFP protein was affected by osteoclastogenesis in the RAW cell system. Thus, whole cell lysates were collected from the three transduced cell lines after 5 days of culture in the presence, or absence, of RANKL. Immunoblots probed with anti-GFP antibodies are shown in Fig. 1C, which demonstrated again that wild type a3-GFP protein was expressed as 134-and 152-kDa bands, both in undifferentiated cells and osteoclasts. In RAW-a3 R445L -GFP-derived osteoclasts, however, protein was observed only at 134 kDa, as in undifferentiated cells. An additional band was observed in all cases at 128 kDa that may be the unglycosylated protein; lower bands likely are degradation products. It was also observed that wild type a3-GFP expression was up-regulated in osteoclasts, compared with undifferentiated RAW cells (as was GFP; 27-kDa band not shown). This phenomenon likely resulted from the fact that the RANKL-induced signal cascade can activate Sp1-family transcription factors, which may bind the GCrich LTRs of the retrovirally integrated cDNA constructs. The FIGURE 1. a3 R445L -GFP is expressed and membrane associated, but is degraded in osteoclasts. A, PCR products representing detection of GFP (upper row), relative to GAPDH (lower row). cDNA from total RNA of osteoclasts that were RANKL-differentiated from RAW-a3-GFP, RAW-a3 R445L -GFP, or RAW-GFP (Table 1) cells, or extracted from untransduced control cells was amplified with primers for GFP or GAPDH detection (Table 1) and PCR products were run on agarose gels. B, 20 g of microsomal membrane protein per lane, from cells described for panel A, was separated by SDS-PAGE and immunoblotted with anti-GFP antibody. Specific bands were mature, processed glycoprotein at 152 kDa and immature, core-glycosylated protein at 134 kDa. Smaller bands were ubiquitous and nonspecific. LTRs normally serve as promoters that constitutively drive expression of the insert. Despite this up-regulation, the mutant protein was observed at reduced levels. This was quantified by normalizing band intensities to GAPDH levels, obtained by staining the same blots with anti-GAPDH antibodies. Fig. 1D graphically confirms the observations of Fig. 1C. The amount of a3 R445L -GFP protein was significantly decreased in mature osteoclasts (ϩ RANKL) compared with that in undifferentiated cells (Ϫ RANKL) (p Ͻ 0.05; n ϭ 3), despite the fact that a3-GFP and GFP expression were significantly up-regulated (both p Ͻ 0.05; n ϭ 3).
Misprocessing of the a3 R445L -GFP Glycoprotein-There are numerous post-translational modifications that can alter protein mobility in SDS-PAGE. To verify that the shifts in mobility of a3 R445L -GFP protein were due to differences in glycosylation, deglycosylation experiments were done with PNGase F, an amidase that cleaves between the initial oligosaccharide N-acetylglucosamine (GlcNAc) and the polypeptide Asn residue of N-linked glycoproteins. Additional experiments were done with Endo H, which cleaves mannose-rich, N-linked oligosaccharides between the initial two GlcNAc residues, leaving a single GlcNAc attached to the polypeptide Asn of the glycosylation site.
Initial experiments were done using lysates of mouse BMMderived osteoclasts to observe carbohydrate cleavage from endogenous, native a3. Immunoblot analysis using anti-a3 antibody is shown in Fig. 2A, where the mature glycosylated a3 subunit is seen as a diffuse 116-kDa band. PNGase F treatment yielded an a3 band of ϳ94 kDa. This corresponds well to the predicted polypeptide size of the subunit, confirming that the observed mobility shift for untreated a3 is due to N-linked glycosylation. A weak band was also seen in the untreated lane at 102 kDa. This likely represents the core-glycosylated glycoprotein, the biosynthetic intermediate in the ER.
To learn more about the effect of the R445L mutation on a3, lysates were collected from transduced RAW cells, expressing GFP fusion proteins, and subjected to PNGase F digestion. As seen in Fig. 2B, fusion proteins from RAW-a3-GFP cells immunoblotted with anti-GFP antibody gave the expected bands at 134 and 152 kDa, whereas RAW-a3 R445L -GFP lysates yielded a major band only at 134 kDa. Upon PNGase F treatment, both constructs yielded bands only at 128 kDa, the unglycosylated fusion protein. To test whether the 134-kDa band indeed represents core-glycosylated a3, lysates were treated with Endo H. Endo H cannot cleave complex oligosaccharide that has been processed in the Golgi; thus, it can be used to discriminate ER-retained, mannose-rich, core-glycosylated glycoproteins from glycoproteins that have trafficked to the Golgi (29, 30). Fig. 2C shows immunoblots where a3-GFP and a3 R445L -GFP lysates were treated with Endo H, followed by immunoblotting with anti-GFP antibodies. At first glance, the mature glycoprotein appears to be unaltered by Endo H treatment, whereas the 134-kDa band is eliminated and a corresponding increase is seen in the 128-kDa band. The latter observation reiterates that the 134-kDa band is the core-glycosylated intermediate and the 128-kDa band is the unglycosylated protein. Closer examination of the 152-kDa band, however, reveals a reproducible reduction in size by 2 kDa, to 150 kDa (n ϭ 3). The observation that the 152-kDa band was reduced slightly in size closely resembles what has been seen for Endo H treatment of CD4, which also has two glycosylation sites. Shin et al. (29) have explained this by suggesting that one of the oligosaccharide chains in the mature CD4 molecule is biantennary and not converted to the complex type in the Golgi, whereas the other site is complex and therefore Endo H-resistant. A similar explanation may hold for the a3 subunit, but the nature of its oligosaccharide chains has yet to be investigated in detail. The 128-kDa deglycosylated protein bands are reasonably close in size to the . Right lane, upon PNGase F treatment, the major 116-kDa band was absent and a new, sharp band was observed at 94 kDa, equivalent in size to the predicted, unglycosylated a3 subunit. ␤-Actin provided a loading standard (lower panels). B, 20 g per lane of cell lysate proteins obtained from RAW-a3-GFP and RAW-a3 R445L -GFP cells was immunoblotted with anti-GFP and anti-GAPDH antibodies. The major diffuse band observed for wild type fusion protein was 152 kDa (glycosylated) and was reduced to 128 kDa (deglycosylated) after PNGase F treatment. The faint 134-kDa band (core-glycosylated) was also absent in the PNGase F-treated lane. For a3 R445L -GFP the major specific band was at 134 kDa (core-glycosylated), and this was also reduced to 128 kDa (deglycosylated) after PNGase F treatment. GAPDH staining provided a loading standard (lower panels). All lysates were subjected to the same digestion protocol (see "Experimental Procedures"), with or without PNGase F addition. C, as in panel B, but cell lysates were incubated with (ϩ) and without (Ϫ) Endo H, instead of PNGase F. The mature glycosylated 152-kDa a3-GFP band was only partially Endo H sensitive (see "Results"), and the 134-kDa band was reduced to 128 kDa. For a3 R445L -GFP, only the core-glycosylated 134-kDa band was seen, which was Endo H sensitive, yielding a 128-kDa band (far right lane). The sharp band at ϳ140 kDa is nonspecific and was observed in all lanes in all experiments. Images in this figure are representative of three independent experiments. AUGUST 3, 2012 • VOLUME 287 • NUMBER 32 predicted polypeptide size of 121 kDa, given that hydrophobic membrane proteins, and fusion proteins with strikingly different domains, often migrate with anomalous mobility in SDS-PAGE.

V-ATPase a3 R444L Misfolding Causes Osteopetrosis
Taken together, these observations suggest that the 134-kDa band is the mannose-rich, core-glycosylated form of the a3 subunit fusion protein, which would go on to be processed in the Golgi to the mature 152-kDa glycoprotein. Lack of the latter band in a3 R445L -GFP-expressing cells supports the notion that the mutation causes misfolding or instability in the nascent subunit, leading to retention of the immature, core-glycosylated intermediate in the ER, without further trafficking and processing. Direct proof of this was sought in cellular localization experiments.
Retention of a3 R445L -GFP within the ER-In macrophages, V-ATPases containing the a3 subunit are targeted to lysosomal membranes (15). GFP fusion protein constructs of a3 can be localized by virtue of their innate fluorescence, so it was of interest to determine the fates of wild type and mutant constructs by fluorescence microimaging. Fig. 3A shows RAW-a3-GFP macrophages illuminated to show GFP fluorescence (green) and Alexa Fluor-labeled anti-LAMP2 fluorescence (red). LAMP2 is predominantly a lysosomal marker, and the merged image reveals a strong colocalization of a3-GFP with LAMP2-positive compartments. Fig. 3B shows RAW-a3 R445L -GFP cells, where a3 R445L -GFP was observed in the perinuclear region and, in contrast with the wild type protein, it did not colocalize with LAMP2-positive compartments. Fig. 3C demonstrates that wild type a3 does not colocalize with calnexin, a marker for ER, and Fig. 3D, in contrast, shows a strong colocalization of mutant a3 and calnexin. Quantification of LAMP2 and calnexin colocalization with either a3-GFP or a3 R445L -GFP (Fig. 3E) showed that there was a significantly greater (p Ͻ 0.001) association of wild type a3-GFP with lysosomes (rather than ER), and a significantly greater (p Ͻ 0.001) association of the mutant a3 R445L -GFP with ER (rather than lysosomes). Values of r Ͼ 0.6 generally indicate a moderate to high degree of colocalization. These observations support the notion that the primary defect of the mutant a3 R444L protein lies in its misfolding and consequent retention in the ER.
Altered protein conformation of a3 R445L -GFP-Misfolding of mutant protein has been elucidated in other systems by using the method of limited proteolysis to demonstrate altered protease susceptibility (31). To confirm the conclusions of the above work, it was of interest to determine directly if the R445L mutation had an effect on the protein conformation of a3 R445L -GFP. To test this, limited trypsinolysis was performed on microsomal membrane proteins derived from RAW-a3-GFP and RAW-a3 R445L -GFP cells. Intact membranes containing a3-GFP or a3 R445L -GFP protein were treated with 2.5 and 5 g/ml of trypsin and then detergent-denatured and deglycosylated with PNGase F. Fig. 4A shows immunoblots of these digests, probed with anti-GFP antibodies.
The a3 R445L -GFP mutant protein was found to be significantly more susceptible to trypsinolysis in comparison with a3-GFP, and also to have some alterations in banding pattern. A closer comparison of band patterns is shown in Fig. 4B, revealing new 26-and 60-kDa bands for the mutant protein (lane h). Fig. 4A, comparing ratios of the 152-kDa bands for untreated and trypsinized wild type protein (lanes b/a) and 134-kDa bands for mutant protein (lanes f/e). There were significant differences (p Ͻ 0.05) for both 2.5 and 5 g/ml trypsin digests. The topology models for the a3 subunit membrane domain argues that Arg-445 is buried in a transmembrane helix 3 (32). Because trypsinolysis was per- formed on intact membranes, this potential cleavage site would not normally be accessible in the wild type protein, so its loss in the mutant protein should not affect the tryptolytic fragment pattern directly. Greater protease susceptibility, and more fragments, were observed for the mutant protein, which cannot be reconciled simply with the loss of a potential trypsin cleavage site at Arg-445. Thus, the data strongly suggest that altered trypsinolysis of a3 R445L -GFP in comparison with the wild type a3-GFP must be the result of more global conformational changes in the mutant protein.

Fig. 4C shows quantification of bands in
Partial Rescue of a3 R445L with Osmolyte-Similar observations have been made for a number of disease-causing mutations in other proteins, most notably in cystic fibrosis transmembrane conductance regulator (CFTR), where the deletion of Phe-508 (⌬F508) results in cystic fibrosis. The ⌬F508 CFTR is retained in the ER due to misfolding and is misprocessed in a manner very similar to what we describe here for the a3 R445L mutant in the mouse V-ATPase system (33). In experimental systems ⌬F508 CFTR can be partially rescued at low temperature, or in the presence of DMSO or glycerol (34). It would be of great interest to know if, in principle, the human a3 R444L mutant could be similarly rescued. We show in Fig. 5 that in the experimental mouse system such rescue is possible. Although, unlike what is seen for CFTR, low temperature and DMSO, even in combination do not appear to rescue a3 R445L , the osmolyte, glycerol, at 10% concentration is able to achieve partial rescue of the protein. As has been shown for ⌬F508 CFTR expression in C127 cells (34), higher or lower concentrations of glycerol were not effective (data not shown), 10% being optimal. These observations suggest that protein rescue with chemical chaperones might be worth considering as a therapeutic strategy. . ER-retained a3 R445L -GFP has an altered protein conformation, as determined by limited proteolysis. A, 15 g of membrane proteins obtained from RAW cells stably transduced to express a3-GFP or a3 R445L -GFP were treated with trypsin (2.5 or 5.0 g/ml) for 1 h. Subsamples were then treated with PNGase F to deglycosylate tryptic fragments (untreated control samples were subjected to the same incubation conditions, but without enzyme; see "Experimental Procedures") and immunoblotted with anti-GFP antibody. Wild type a3-GFP showed a major, uncleaved band at 152 kDa (glycosylated) in the absence of enzyme treatment (lane a, black arrowhead), but on trypsinolysis the band intensity was reduced substantially (lane b). PNGase F deglycosylation shifted the band to 128 kDa (lane c, white arrowhead), as expected, and the trypsin banding was somewhat altered (lane d).
Mutant a3 R445L -GFP showed a major uncleaved band at 134 kDa (core-glycosylated) in the absence of enzyme (lane e, asterisk), but on trypsinolysis the band intensity was reduced almost to background (lane f). After PNGase F treatment the major band shifted to 128 kDa (lane g, white arrowhead), as expected, and the trypsin banding was somewhat altered (lane h). Comparing lanes d and h showed that two novel bands appeared in lane h (small leftpointing arrowheads at right margin, indicating bands at 26 and 60 kDa). Note the corresponding right-pointing small arrowheads at lane d, indicating differences at the same positions). B, vertical tranches were cropped from panel A (lanes d and h) and aligned precisely, for side-by-side comparison. No alterations to relative lengths of lanes were made; only vertical alignment was adjusted to account for gel "smiling." Although some band intensities are different, the only novel bands are those of lane h at 26 and 60 kDa, as indicated. C, deglycosylated 128-kDa protein bands were quantified from scans of blots shown in panel A. Rates of degradation of mutant a3 R445L -GFP bands were found to be significantly greater than those for wild type a3-GFP (p Ͻ 0.05). Immunoblots show typical results from one of two independent experiments. FIGURE 5. The misfolded a3 R445L -GFP protein can be rescued with osmolyte. 20 g of membrane protein obtained from RAW cells stably transduced to express a3-GFP, or a3 R445L -GFP was loaded per lane, as indicated, and immunoblotted with anti-GFP antibody. Treated cells were exposed to medium containing 10% glycerol for 24 h at 37°C, or medium containing 5% DMSO at 26°C for 24 h (see "Experimental Procedures"). Wild type a3-GFP showed a mature glycosylated band at 152 kDa and a core-glycosylated band at 134 kDa. The mutant protein, a3 R445L -GFP from untreated cells showed no mature band, only a core-glycosylated band at 134 kDa and an unglycosylated band at 128 kDa. Cells exposed to 10% glycerol showed some rescue of the 152-kDa mature glycoprotein, whereas those exposed to both 5% DMSO and low temperature (26°C) showed no detectable rescue.

DISCUSSION
The R444L Mutation in a3 Confers a Protein Misfolding/Misprocessing Phenotype-We have shown elsewhere that wild type mouse a3 is N-glycosylated at two sites, Asn-484 and Asn-504, located on a single luminal loop within its membrane domain. 3 Consistent with this, in the present work, mouse a3, with a predicted size of 93.4 kDa, was observed in immunoblots as an Endo H-sensitive, immature, core-glycosylated band migrating at 102 kDa, and an Endo H-insensitive, mature, glycosylated band at 116 kDa. The processing and maturation of the a3 glycoprotein is indicative of its biosynthetic pathway, involving ER biosynthesis and assembly into the V-ATPase complex, trafficking to the Golgi for processing, and ultimately trafficking to lysosomes and, in active osteoclasts, to the plasma membrane.
It was of interest to determine whether the disease-associated R444L mutant protein is similarly processed. To investigate the fate of the homologous mutant protein, a3 R445L , in the mouse system, a retroviral vector was used to integrate GFPtagged wild type or mutant cDNAs into the genomic DNA of RAW macrophage cells. Stable expression of a3 R445L -GFP in this system demonstrated that the primary phenotypic defect resulting from the mutation was the retention of immature a3 R445L -GFP glycoprotein in the ER, presumably due to misfolding, and its ultimate degradation. That this is not simply an artifact of the heterologous expression system was demonstrated by the apparently normal processing and trafficking of the wild type a3-GFP fusion protein. Further evidence of correct targeting of the C-terminal GFP-tagged a subunits, which requires functional assembly into the V-ATPase complex, has been demonstrated also in our work with the Vph1p homolog in yeast, 3 and by others in mouse macrophages (35) and in the slime mold, Dictyostelium (36).
In the context of the dysfunctional osteoclasts seen in a3 R444L ARO patients, the above observations would imply that the mechanism behind the dysfunction is that the mutant a3 R444L subunit is incapable of being incorporated into the V-ATPase complex and subsequently trafficking to the plasma membrane of the osteoclast ruffled border. V-ATPases that specifically incorporate the a3 subunit isoform need to be localized to the osteoclast ruffled border to acidify the external resorption lacunae; without this process, bone cannot be resorbed (37,38).
Four lines of evidence presented here support the notion that a3 R444L is misfolded and retained in the ER. First, the mutant fusion protein was observed only as 128-and 134-kDa bands, lacking the 152-kDa mature glycosylated band of the wild type fusion protein (PNGase F treatment reduced all bands to 128 kDa, i.e. the deglycosylated protein). Lack of a 152-kDa a3 R445L -GFP band, and reduced intensity of the 128-and 134-kDa bands (compared with the combined 134-and 152-kDa bands of a3-GFP), suggested misprocessing of the mutant glycoprotein, due to lack of trafficking to the Golgi, and degradation of the retained protein by the endoplasmic reticulum-associated degradation (ERAD) pathway of the ER quality control machinery (39), respectively. Second, we showed by colocalization, after fluorescence immunostaining of LAMP2, or calnexin, that a3-GFP colocal-ized with LAMP2, not calnexin, suggesting that it was associated primarily with the lysosomal compartment. In contrast, a3 R445L -GFP colocalized with calnexin, not LAMP2, suggesting that it was associated with the ER. This supports the notion that, like native a3, a3-GFP can traffic normally to the lysosome, whereas the a3 R445L -GFP mutant protein does not leave the ER.
Third, upon limited proteolysis, the mutant protein was observed to be more sensitive to protease degradation, and novel proteolytic fragments were obtained, strongly suggesting increased protease accessibility due to global conformational changes caused by misfolding. Finally, the observation that the mutant protein was amenable to osmolyte rescue by growth of cells in 10% glycerol strongly suggests similarities with other systems, such as ⌬F508 CFTR, where protein misfolding, misprocessing, and aberrant trafficking have been well characterized (33).
Recent studies have suggested that presenilin 1 is a mandatory chaperone required for N-linked glycosylation of the a1 subunit, working in conjunction with oligosaccharyltransferase and the Sec61␣ subunit of the translocon complex (28). If a3 glycosylation has the same requirements, we can conclude that the interaction between presenilin 1 and a3 R445L -GFP, and the subsequent core glycosylation step by oligosaccharyltransferase is not impaired, as we have shown that a mannose-rich, biosynthetic-intermediated state of glycosylation is achieved for a3 R445L -GFP. It is possible that interaction with calnexin, or Bip, is ultimately responsible for retention of a3 R445L -GFP, as both are known to bind V-ATPase during assembly in plant cells (40), but the precise mechanism of retention remains unknown. In yeast, the a subunit homolog, Vph1p, is rapidly degraded if it fails to assemble into the V 0 complex within the ER. Hill and Cooper (41) have suggested that it is degraded via an alternative ER quality control pathway. In mammalian cells, entry into the ERAD pathway (42) must come into play, as significant degradation of the retained mutant protein was observed, and trafficking to Golgi and lysosomes clearly does not occur. The significance of the increased degradation in mature osteoclasts is presently unclear, but many cellular processes are induced during osteoclastogenesis, some of which might impact protein turnover. Whether the heterologous expression of misfolded mutant a3 R445L -GFP protein might induce up-regulation of ERAD as part of the unfolded protein response (39,43) remains to be determined.
The Structural Significance of Arg-444-The current structural topology model for the yeast V-ATPase a subunit, Vph1p, suggests that the Arg-462 residue is found within a hydrophobic transmembrane ␣-helix, TM3 3 (32). The topology model places the Arg residue within 1-2 helical turns of the cytoplasmic interface. Because Vph1p and human a3 are highly conserved orthologs, sharing 36% identity, it is likely that the homologous human Arg-444 and mouse Arg-445 are similarly placed within putative TM3 helices. In the yeast system, the R462L mutation has a relatively mild phenotype (44), suggesting that, unlike Arg-735, where any missense mutation completely disrupts V-ATPase activity (45), Arg-462 is not involved directly in proton transport.
Arginine in hydrophobic membrane domains of proteins is not uncommon, and can have important implications for protein structure and disease (46). Asymmetric distribution of basic amino acids, with respect to the lipid bilayer, is a well known factor influencing membrane topology, as has been generalized in the "positive inside rule" of von Heijne (47), and may assist TM membrane insertion (48). Furthermore, Vostrikov et al. (49) have shown that Arg, Gly, and aromatic residues near the ends of membrane-spanning ␣-helices may be determinants of helical tilt within the membrane bilayer. Arginine is the most effective, determining tilt direction and inducing tilt angles as high as 24°in synthetic peptides incorporated into artificial membrane systems. It is intriguing, in this regard, that the sequence, GRYL, which contains all of the aforementioned residues (R being the human Arg-444), is highly conserved at the cytoplasmic end of the putative TM3 of the a subunit. Thus, one possible effect of Arg-444 mutation may be failure to maintain a critical tilt angle or tilt direction for TM3 in the membrane domain of the a3 subunit, resulting in misalignment with other membrane-spanning helices and consequent misfolding or instability of folded a3.
Conclusion-In summary, in this study it has been demonstrated that the R444L point mutation responsible for a type 1 infantile malignant osteopetrosis is caused by protein misfolding that results in retention of the V-ATPase a3 subunit in the ER. The misfolded protein consequently does not traffic to the plasma membrane, where it needs to be to make its functional contribution to the process of bone resorption. Furthermore, we have shown experimentally that conditions that are known to rescue misfolded proteins can partially rescue the a3 R445L mutant glycoprotein.
Hematopoietic stem cell transplantation (HSCT) is performed in severe cases of ARO (23), but there are associated high risks (25%) of disease progression and poor 5-year survival (24%) for recipients of HLA-haplotype-mismatched hematopoietic stem cell transplantation. Furthermore, preservation of vision requires intervention prior to 3 months of age (22). Radiological prenatal diagnosis can be obtained at 25 weeks gestation (50), but already encroachment of marrow cavities is obvious, suggesting that earlier fetal DNA testing is required. Pharmacological intervention, if available, would likely have to begin in neonates, if not in utero, to prevent development of severe manifestations of disease.
The severe ARO phenotype seen with the R444L a3 point mutation resembles what is seen in oc/oc mice where death occurs at 3 weeks of age (51). The oc/oc model results from a naturally occurring deletion within the Tcirg1 gene, in the N-terminal domain of the protein, which prevents the expression of the a3 subunit. The osteoclasts of these mice lack a ruffled border, and cannot secrete acid. The R445L point mutation causes ER retention and degradation of mouse a3, and this prevents the localization of a3-containing V-ATPases to lysosomal and plasma membranes. The functional outcome for critical plasma membrane expression of V-ATPase, for the human R444L mutation, would likely be the same as is seen in oc/oc mice; however, unlike the oc/oc model, some steady-state level of intact, core-glycosylated protein is observed in the experimental mouse system.
In another protein misfolding disease, ⌬F508 CFTR-mediated cystic fibrosis, the ER-retained ⌬F508 CFTR can under some circumstances be induced to traffic to the plasma membrane in mammalian cells, where it is observed to have significant function despite the mutation (52). Thus, much effort has been expended in attempting to find a clinically viable pharmacological means of achieving trafficking of the mutant protein to the plasma membrane as a potential cure for a large percentage of cystic fibrosis cases (53,54). In a similar manner, a pharmacological chaperone that can rescue a3 R444L , especially if usable during early infancy, or in utero, might be of benefit to patients afflicted with a3 R444L -mediated ARO. This remains a tall order, but the work described here, elucidating the disease mechanism of the a3 R444L mutation and showing the potential for protein rescue, is a first step in addressing this need.