5-Azacytidine-induced Protein 2 (AZI2) Regulates Bone Mass by Fine-tuning Osteoclast Survival*

Background: 5-Azacytidine-induced protein 2 (AZI2) is critical in GM-CSF-induced dendritic cell differentiation. Results: AZI2 deficiency enhances osteoclast survival, leading to decreased bone mass in vivo. Conclusion: AZI2 suppresses osteoclast survival by inhibiting c-Src activation. Significance: This is the first study showing the osteoprotective function of AZI2. 5-Azacytidine-induced protein 2 (AZI2) is a TNF receptor (TNFR)-associated factor family member-associated NF-κB activator-binding kinase 1-binding protein that regulates the production of IFNs. A previous in vitro study showed that AZI2 is involved in dendritic cell differentiation. However, the roles of AZI2 in immunity and its pleiotropic functions are unknown in vivo. Here we report that AZI2 knock-out mice exhibit normal dendritic cell differentiation in vivo. However, we found that adult AZI2 knock-out mice have severe osteoporosis due to increased osteoclast longevity. We revealed that the higher longevity of AZI2-deficient osteoclasts is due to an augmented activation of proto-oncogene tyrosine-protein kinase Src (c-Src), which is a critical player in osteoclast survival. We found that AZI2 inhibits c-Src activity by regulating the activation of heat shock protein 90 (Hsp90), a chaperone involved in c-Src dephosphorylation. Furthermore, we demonstrated that AZI2 indirectly inhibits c-Src by interacting with the Hsp90 co-chaperone Cdc37. Strikingly, administration of a c-Src inhibitor markedly prevented bone loss in AZI2 knock-out mice. Together, these findings indicate that AZI2 regulates bone mass by fine-tuning osteoclast survival.

Bone-degrading osteoclasts are large multinucleated cells of myeloid lineage origin (1). It has been established that osteoclast differentiation and activation are induced by receptor activator of NF-B ligand (RANKL) 3 expressed on bone-forming osteoblasts. After binding of RANKL to its receptor activator of NF-B, transcription factors such as nuclear factor of activated T-cells cytoplasmic 1 (NFATc1) (2,3), Jun dimerization protein 2 (Jdp2) (4,5), c-Fos (6), and NF-B (7) are activated, and osteoclast-specific genes are induced. Once the differentiation process is completed, osteoclasts undergo apoptosis. Recently, it has been suggested that the longevity of osteoclasts controls bone destruction. For example, osteoblast-derived cytokines such as macrophage colony-stimulating factor (M-CSF) (8) and osteopontin (OPN) (9) promote osteoclast survival. After the stimulation of osteoclast precursors by such survival factors, proto-oncogene tyrosine-protein kinase Src (c-Src), a non-receptor tyrosine kinase is activated and induces resistance to apoptosis (10,11). c-Src activation is orchestrated by interplay between several other proteins. A recent study suggests that heat shock protein 90 (Hsp90), a molecular chaperone required for the stability of target proteins, suppresses c-Src activation (12). Moreover, previous studies revealed that Cdc37, a co-chaperone of Hsp90, directly binds to Hsp90 to strengthen interactions between Hsp90 and its binding partners, leading to enhanced chaperone activity of Hsp90 (13). Importantly, Hsp90 inhibitors such as 17-AAG disrupt the Hsp90-c-Src association, leading to the bone-degrading activity of osteoclasts via activation of c-Src. Therefore, the Hsp90 mediated regulation of the c-Src pathway is thought to be a novel therapeutic target for bone destructive diseases (12). * This work was supported by a research fellowship from the Japan Society for the Promotion of Science (JSPS) for the Promotion of Science for Young Scientists, JSPS KAKENHI for a grant-in-aid for challenging exploratory research, a grant from the Osaka University MEET project, a grant from the Astellas Foundation for Research on Metabolic Disorders, a grant from the Naito Foundation, a grant from the SENSHIN Medical Research Foundation, a grant from the Japan Intractable Disease Research Foundation, and a visionary research grant from Takeda Science Foundation (to K. M.). 1 To whom correspondence may be addressed: Laboratory of Host Defense, WPI Immunology Frontier Research Center, Osaka University, 3-1 Yamadaoka, Suita, Osaka 565-0871, Japan. Regulation of osteoclast formation is also modulated by innate immune cytokines, such as IFNs (14). Type-I IFNs such as IFN-␤ are induced by RANKL and inhibit osteoclast differentiation (15). Virus-induced type-I IFN production is mediated by TNF receptor (TNFR)-associated factor family member-associated NF-B activator-binding kinase 1 (TBK1) and IFN regulatory factor 3 (IRF3). Among the TBK1-binding proteins, TNF receptor-associated factor (TRAF) family memberassociated NF-B activator (TANK), 5-azacytidine-induced protein 2 (AZI2), and TBK1-binding protein 1 (TBKBP1) have been implicated in the production of type-I IFNs (14). However, our previous knock-out study revealed that TANK is dispensable for the production of IFNs (16), but indispensable for the suppression of excessive osteoclast formation via the inhibition of TRAF6 activation (17). We also generated AZI2-and TBKBP1-deficient mice and revealed that both genes were dispensable for IFN production (18). However, to our surprise, AZI2-deficient bone marrow cells exhibited impaired GM-CSF-mediated dendritic cell differentiation in vitro (18). GM-CSF is involved in the survival, maturation, proliferation, and differentiation of myeloid cells including dendritic cells (19). GM-CSF is also important in alveolar protein clearance because GM-CSF neutralizing autoantibodies are involved in the pathogenesis of pulmonary alveolar proteinosis (PAP) (20). Bronchoalveolar lavage (BAL) from PAP patients and GM-CSF-deficient mice contains lipoproteinaceous material consisting of surfactant proteins, and alveolar macrophages in BALs exhibit a large foamy appearance (20,21). Importantly, such macrophages are also severely deficient in ATP-binding cassette subfamily G member 1 (ABCG1), a group of transmembrane proteins that play pivotal roles in mediating the cellular efflux of lipids, and peroxisome proliferator-activated receptor (PPAR)␥, a critical regulator of lipid homeostasis (22). Furthermore, lung tissues from PAP patients and GM-CSF-deficient mice exhibited many periodic acid-Schiffpositive alveolar fillings, leading to respiratory distress (20,21,23).
Despite the importance of AZI2 in GM-CSF signaling in vitro, the roles of AZI2 in immunity, lung homeostasis, and their functions in bone metabolism remain to be studied in vivo.
Here we explored the role of AZI2 in host defense and bone homeostasis. We report that AZI2 is dispensable for GM-CSF signaling in vivo, but we found that AZI2-deficient mice have severe osteoporosis due to enhanced osteoclast survival.

EXPERIMENTAL PROCEDURES
Mice-AZI2-and TBKBP1-deficient mice were generated as described previously (18) and maintained in specific pathogenfree conditions. All animal experiments were performed in accordance with the approval of the Animal Research Committee of the Research Institute for Microbial Diseases (Osaka University, Osaka, Japan).
In Vivo Infection-Staphylococcus aureus was cultured in tryptic soy broth for 15 h at 37°C. Mice were infected intravenously with a PBS solution containing 5 ϫ 10 7 S. aureus.
Analysis of in Vivo Bone Phenotype-To check the bone formation rate, double calcein labeling was performed as described (26). For bone histomorphometric analysis, tibias were stained with Villanueva bone stain (Wako Chemicals) and embedded in methyl-methacrylate. Serial longitudinal sections (6-mm-thick) were generated, and each section was observed by a Histometry RT camera (System Supply Co., Ltd., Nagano, Japan). Distal portions of femurs were analyzed by three-dimensional microcomputed tomographic (CT) using Scan-Xmate RB080SS110 (Comscan Techno Co., Ltd., Sagamihara, Japan) and TRI/3D-Bon software (Ratoc System Engineering Co., Ltd., Tokyo, Japan). Bone microarchitectural parameters were quantified in the trabecular regions at 0.1-1.5 mm from the chondro-osseous junction.
Osteoclast and Osteoblast Culture-M-CSF-derived macrophages (MDMs) were prepared as described (27) and used as osteoclast precursors. Osteoclast differentiation was induced in the presence of 25 ng/ml M-CSF for various times and with various concentrations of RANKL. TRAP staining was performed as described (17). For the osteoclast survival assay, MDMs were cultured with RANKL for 84 h. Then, RANKL and M-CSF were removed, and osteoclasts were cultured for an additional 36 h. The survival rate was quantified by counting morphologically intact osteoclasts. For the osteoclast/osteoblast co-culture assay, calvarial cells (5 ϫ 10 5 ) and MDMs (3 ϫ 10 5 ) were cultured in the presence of prostaglandin E 2 (PGE 2 ) and 1␣,25-dihydroxyvitamin D 3 (1␣,25(OH) 2 D 3 ). For analysis of osteoclast resorptive activity, MDMs were plated on bone resorption assay plates (Iwai Chemical Co., Tokyo, Japan). After 5 days of RANKL stimulation, plates were immersed in 1 M NH 4 OH, and pit numbers were counted. Calvariae from newborn mice were digested in ␣-MEM containing 0.1% collagenase and 0.2% dispase at 37°C for 20 min. Mesenchymal stem cells from compact bone were prepared as described (28). To generate osteoblasts in vitro, calvarial cells or mesenchymal stem cells were cultured in an osteoblast inducer reagent (Takara). ALP and calcified nodules in osteoblast cultures were detected using a TRAP/ALP staining kit (Wako Chemicals) and calcified nodule staining kit (AK-21; Primary Cell Co., Ltd., Hokkaido, Japan), respectively. Calcium concentrations were measured using a kit (Metalloassay LS-MPR, AKJ Global Technology).
Calcium Imaging-Cells were plated on poly-L-lysine-coated glass-bottom dishes and cultured with Fura2/AM. Fluorescent activities were analyzed as described previously (29). PCR-RNA was extracted using TRIzol reagent (Life Technologies) and reverse-transcribed by ReverTra Ace (Toyobo Co., Ltd., Osaka, Japan). Quantitative real-time PCR (qPCR) was performed using an ABI PRISM 7500 real time PCR system. TaqMan Assays-on-Demand primers were purchased from Applied Biosystems (Foster City, CA).
Viral Gene Transfer/Knockdown-AZI2 was cloned into a retroviral vector and transfected into the packaging cell line PlatE. Retroviral gene transduction of MDMs was performed as described previously (30). After transduction, cells were stimulated with RANKL. Cdc37 shRNA lentiviral particles (sc-35043-V) and control particles (sc-108080) were purchased from Santa Cruz Biotechnology. For lentiviral gene transfer, virus was added to wells with Polybrene. After 20 h, cells were stimulated with RANKL to induce osteoclasts.
Statistical Analysis-Student's t test was used to evaluate statistical significance, with significance set at p Ͻ 0.05.

RESULTS
AZI2 Is Dispensable for in Vivo Myeloid Cell Differentiation and Function-Because AZI2 is critical for in vitro GM-CSF signaling, we hypothesized that AZI2-deficient mice would exhibit PAP. We first evaluated the population of myeloid and lymphoid cell lineages in AZI2-deficient mice, but hematopoi-esis was normal (Fig. 1A). Unexpectedly, the number of alveolar macrophages in AZI2-deficient mice was normal. Moreover, alveolar macrophages presented normal morphology, cytokine production, and lipid metabolism-associated gene expression (Fig. 1, B-E). Furthermore, alveolar structures of AZI2-deficient mice were normal, and periodic acid-Schiff-positive alveolar fillings were not detected in these mice (Fig. 1F). Because GM-CSF-induced AZI2-deficient dendritic cells have shown impaired Toll-like receptor (TLR) signaling and antigen-presenting activity (18), we evaluated the production of cytokine and acquired immune functions in AZI2-deficient mice. However, AZI2 deficiency had no impact on the production of proinflammatory cytokines or survival in response to LPS (Fig. 1G), S. aureus (Fig. 1H), and Candida albicans (data not shown). Additionally, the production of OVA-specific IgM and IgG1 in response to intraperitoneal injection of OVA plus alum was normal in AZI2-deficient mice (Fig. 1I). Furthermore, AZI2 deficiency had no impact on T cell activation by CD11c-positive splenocytes (Fig. 1J). Thus, these findings indicate that TBK1-binding protein AZI2 is dispensable for myeloid cell differentiation and functions in vivo.
AZI2-deficient Mice Are Osteoporotic-During the flushing of bone marrow from 11-week-old mice, we noticed that AZI2- deficient femurs were more fragile when compared with wildtype control mice. Thus, this observation prompted us to explore the role of AZI2 in bone homeostasis. Body weight and femur length were normal in AZI2-deficient mice, but CT analysis of the distal portion of femurs revealed dramatically impaired trabecular bone volume and number when compared with wild-type mice (Fig. 2, A-C). Bone histomorphometric analysis of proximal tibias from AZI2-deficient mice showed a significant reduction of the trabecular bone area accompanied by a dramatic increase of osteoclast number and eroded area (Fig. 2, D-G). In contrast, osteoblast parameters and boneforming rate were normal (Fig. 2, F and G). Consistent with the high number of osteoclasts, the bone mineral density in cortical  areas of femurs was significantly decreased and the serum bone resorption marker C-terminal telopeptides type I collagen (CTXI) was increased in AZI2-deficient mice (Fig. 2, H and I).
Thus, these data clearly indicate that AZI2-deficient mice are osteoporotic. In contrast, femurs and osteoclasts from TBKBP1-deficient mice were normal (data not shown). APRIL 10, 2015 • VOLUME 290 • NUMBER 15

JOURNAL OF BIOLOGICAL CHEMISTRY 9383
Osteoblast Functions Are Normal in AZI2 Deficiency-Because the osteoporotic phenotype of AZI2-deficient mice could be due to impaired osteoblast functions, we explored the role of AZI2 in osteoblasts. First, we found that the expression level of AZI2 in osteoblasts considerably decreases during differentiation (Fig. 3A). Next, we cultured calvarial cells with osteoblast-inducing medium and analyzed their calcification potential. However, AZI2 deficiency did not affect bone nodule formation (Fig. 3, B and C). Furthermore, the expression levels of osteoblast differentiation markers such as ALP, RANKL, OPG, and Runx2 were normal (Fig. 3, D and E). Recently, it has been reported that osteoblast-derived factors such as EphB4 and Sema3a are involved in osteoclast suppression. Because these osteoblast-osteoclast communicating factors are critical for bone remodeling, we further checked the expression of these genes, but AZI2 deficiency had no impact on gene expression (Fig. 3E). Collectively, these findings suggest that AZI2 does not affect osteoblast differentiation and function both in vitro and in vivo.
Increased Survival of AZI2-deficient Osteoclasts-AZI2 is broadly expressed in various organs, especially bone marrow, kidney, spleen, and brain (Fig. 4A). This suggests that AZI2 is predominantly expressed in hematopoietic and neural systems. We also analyzed the AZI2 expression levels in various cells and found that AZI2 expression is lower in osteoclasts when compared with lymphocytes, myeloid cells, osteoblasts, and synovial cells (Fig. 4B). Importantly, osteoclasts from old mice exhibited significantly lower expression of AZI2 when compared with young mice, and the expression level of AZI2 in MDMs gradually decreased during RANKL-induced osteoclastogenesis (Fig. 4, C and D). To explore the role of AZI2 during osteoclastogenesis, we analyzed the differentiation of AZI2 knock-out mouse-derived MDMs into osteoclasts. Three days after RANKL-induced osteoclastogenesis, the number of osteoclasts generated was comparable between AZI2-deficient MDMs and wild-type MDMs (Fig. 4, E and F). However, 5 days after RANKL stimulation, wild-type cells began to die and the number of TRAP-positive multinucleated cell was lower when compared with day 4 (Fig. 4, E and F). In contrast, AZI2-deficient cells exhibited a high number of TRAP-positive multinucleated cells, even at day 5 (Fig. 4, E and F). Confirming the improved osteoclast activity of AZI2-deficient osteoclasts, the number of bone resorption pits generated by AZI2-deficient osteoclasts was significantly higher when compared with wildtype osteoclasts. However, this improvement of osteoclast activity was not due to an enhanced osteoblast-supporting function (Fig. 4, G-I). Importantly, osteoclast survival assay showed that AZI2-deficient osteoclasts have a significantly slower apoptosis (Fig. 4, J and K). Confirming this slower apoptosis rate, AZI2-deficient osteoclasts have a lower concentration of cleaved caspase-3, which is a critical executioner of apoptosis (Fig. 4L). Furthermore, slower apoptosis level in AZI2-deficient osteoclasts was partially rescued by AZI2 expression (Fig. 4M). Nevertheless, the expression levels of osteoclastic genes (Fig. 5A), calcium oscillation, and activation of osteoclast-inducing transcription factors such as NFATc1 and NF-B (Fig. 5, B and C) were similar in both AZI2-deficient osteoclasts and wild-type osteoclasts. Together, these observations clearly show that AZI2-deficient osteoclasts are defective for apoptosis pathways, but exhibit normal differentiation.
AZI2 Regulates c-Src Activation-To gain insight into the mechanism underlying the aberrant osteoclastogenesis in AZI2-deficient mice, we analyzed RANKL and M-CSF signaling. First, because AZI2 is implicated in the production of IFN-␤, and because RANKL can also induce IFN-␤, we quantified the expression of IFN-␤ induced by RANKL in AZI2-deficient MDMs (Fig. 6A). However, AZI2-deficient MDMs exhibited normal IFN-␤ expression in response to RANKL (Fig. 6A). Next, because M-CSF and OPN promote osteoclast survival via activation of c-Src, we stimulated AZI2-deficient osteoclasts with M-CSF or OPN and measured the level of phosphorylated c-Src level (Tyr(P)-416 Src) by Western blotting (Fig. 6B). Intriguingly, the activation of c-Src was enhanced in AZI2-deficient cells, in response to M-CSF or OPN stimulation (Fig.  6B), and a c-Src inhibitor AZD0530 suppressed excessive osteoclast generation in AZI2-deficient cells (Fig. 6, C-E). To clarify the molecular mechanisms leading to an enhanced c-Src activation in AZI2-deficient cells, we measured the expression levels of Hsp90, which is a potent inhibitor of c-Src activation (Fig.  6F). Unexpectedly, Hsp90 expression in AZI2-deficient osteoclasts was normal (Fig. 6F). Notably, a Hsp90 inhibitor, 17-AAG, significantly increased osteoclast numbers in wildtype cells, but AZI2-deficient cells did not respond to this inhibitor (Fig. 6, G and H). Moreover, the expression levels of Cdc37, which is a critical binding partner of Hsp90 for the inhibition of c-Src, were comparable between wild-type and AZI2 knock-out cells (Fig. 6I). Because the Hsp90 inhibitor had no effect on osteoclast formation in AZI2-deficient cells, we predicted that AZI2 may be a component of the functional Hsp90-Cdc37 complex. To confirm this hypothesis, we examined the associ- ation of AZI2 with Cdc37. As expected, an association between Hsp90 and Cdc37 was observed by immunoprecipitation of mouse primary MDMs (Fig. 6J). Intriguingly, an association between AZI2 and Cdc37 was also identified (Fig. 6J). Furthermore, retroviral expression of AZI2 significantly rescued excessive osteoclast formation in AZI2-deficient cells, but not in Cdc37 knockdown AZI2-deficient cells (Fig. 6, K-M). Thus, these findings clearly indicate that the Hsp90-Cdc37-AZI2 complex is critical for the suppression of c-Src activation in osteoclasts. Finally, oral administration of a c-Src inhibitor, AZD0530, significantly rescued the osteoporotic phenotype of AZI2-deficient mice (Fig. 6, N-P). Based on these data, we concluded that AZI2 regulates bone mass by fine-tuning c-Src activation in osteoclasts.

DISCUSSION
Previous studies demonstrated that among the TBK1-binding proteins, TANK, AZI2, and TBKBP1 regulated the production of type-I IFNs in vitro (14). In contrast, our knock-out mouse studies revealed that TANK, AZI2, and TBKBP1 are dispensable for IFN production in vivo. In addition, we have demonstrated that AZI2 is indispensable for in vitro GM-CSFinduced dendritic cell differentiation (18), and it has been reported that GM-CSF signaling is critical for alveolar protein clearance by macrophages (20,21). Here, to our surprise, we could not observe any immunological or alveolar abnormalities in AZI2-deficient mice. Although GM-CSF is predominantly used as an inducer of dendritic cells in vitro, GM-CSF-deficient mice exhibit normal dendritic cell populations (20,21). Thus, we speculate that in vivo GM-CSF signaling is dispensable for dendritic cell differentiation, but indispensable for proper functions of lung macrophages. Most likely, additional studies are needed to clarify the molecular functions of AZI2 in dendritic cells, but our data clearly indicate that the in vitro signaling defects of GM-CSF in AZI2-deficient cells are compensated by unknown mechanisms in vivo.
Although we did not detect any immunological or alveolar abnormalities in AZI2-deficient mice, we discovered that they are osteoporotic. We found that c-Src activation is enhanced in AZI2-deficient osteoclasts, when they are stimulated by osteoclast survival factors such as M-CSF or OPN. Recently, it was reported that disruption of Hsp90-c-Src binding by 17-AAG stimulates c-Src activation (12), and previous studies also suggested that Cdc37 is critical for the Hsp90 inhibitory function on c-Src (13). Here, we demonstrate that AZI2 regulates the c-Src repressive activity of Hsp90 by directly binding to Cdc37. Moreover, because the expression of AZI2 gradually decreases during osteoclast differentiation, whereas the expression of Hsp90 and Cdc37 increases during this process, AZI2 expression levels might regulate the c-Src repressive activity of the Hsp90-Cdc37 complex.
In our previous knock-out study, we found that TANK negatively regulates osteoclast formation by suppressing the activation of TRAF6 (16). In this study, we show that AZI2, but not TBKBP1, also functions as a negative regulator of osteoclast formation by suppressing the activity c-Src. Therefore, we suggest that these TBK1-binding partner proteins are indispensable for proper bone homeostasis.
Although further experiments are required to understand the precise molecular mechanisms of AZI2 in osteoclastogenesis, our data clearly indicate that AZI2 plays a pivotal role by fine-tuning osteoclast survival. Moreover, our findings provide a basis for AZI2-stimulating therapeutic strategies to treat osteoporosis and rheumatoid arthritis.