JMJD5, a Jumonji C (JmjC) Domain-containing Protein, Negatively Regulates Osteoclastogenesis by Facilitating NFATc1 Protein Degradation*

Background: NFATc1 is a necessary and sufficient transcription factor for osteoclastogenesis. Results: JMJD5 negatively regulates NFATc1 protein level through its hydroxylase activity. Conclusion: JMJD5 is a novel osteoclastogenic repressor that induces the degradation of NFATc1 protein. Significance: This study revealed a novel mechanism that regulates NFATc1 activity during osteoclastogenesis. Osteoclastogenesis is a highly regulated process governed by diverse classes of regulators. Among them, nuclear factor of activated T-cells calcineurin-dependent 1 (NFATc1) is the primary osteoclastogenic transcription factor, and its expression is transcriptionally induced during early osteoclastogenesis by receptor activation of nuclear factor κB ligand (RANKL), an osteoclastogenic cytokine. Here, we report the novel enzymatic function of JMJD5, which regulates NFATc1 protein stability. Among the tested Jumonji C (JmjC) domain-containing proteins, decreased mRNA expression levels during osteoclastogenesis were found for JMJD5 in RAW264 cells stimulated by RANKL. To examine the functional role of JMJD5 in osteoclast differentiation, we established stable JMJD5 knockdown cells, and osteoclast formation was assessed. Down-regulated expression of JMJD5 led to accelerated osteoclast formation together with induction of several osteoclast-specific genes such as Ctsk and DC-STAMP, suggesting that JMJD5 is a negative regulator in osteoclast differentiation. Although JMJD5 was recently reported as a histone demethylase for histone H3K36me2, no histone demethylase activity was detected in JMJD5 in vitro or in living cells, even for other methylated histone residues. Instead, JMJD5 co-repressed transcriptional activity by destabilizing NFATc1 protein. Protein hydroxylase activity mediated by the JmjC domain in JMJD5 was required for the observed functions of JMJD5. JMJD5 induced the association of hydroxylated NFATc1 with the E3 ubiquitin ligase Von Hippel-Lindau tumor suppressor (VHL), thereby presumably facilitating proteasomal degradation of NFATc1 via ubiquitination. Taken together, the present study demonstrated that JMJD5 is a post-translational co-repressor for NFATc1 that attenuates osteoclastogenesis.

Mature osteoclasts are giant multinucleated cells that are responsible for bone resorption in bone remodeling and function in a regulated balance with bone-forming osteoblasts (1,2). Commitment of hematopoietic stem cells to differentiate to osteoclasts is defined by two key factors, macrophage colonystimulating factor and receptor activation of nuclear factor B ligand (RANKL), 3 factors that are produced by osteoblast lineage cells in bone tissues (3)(4)(5)(6). During cell maturation, mononuclear precursor cells fuse to generate multinucleated osteoclasts. This multinucleation process is necessarily required for osteoclastic maturation to undergo efficient bone resorption (7,8). In reorganizing genetic networks during osteoclastogenesis, diverse transcription factors play pivotal roles, and among them, nuclear factor of activated T-cells calcineurin-dependent 1 (NFATc1) is the necessary and sufficient transcription factor (9,10). Osteoclastogenesis induced by RANKL is aborted in NFATc1Ϫ/Ϫ embryonic stem cells (9). Therefore, NFATc1 is considered to be an essential factor in osteoclastogenesis. Moreover, ectopic expression of NFATc1 in osteoclast precursor cells efficiently induces osteoclastogenesis even in the absence of RANKL stimulation (10). Supporting the importance of NFATc1 in osteoclastogenesis, NFATc1 is a transcriptional inducer of several osteoclast-specific genes (required for maturation and bone resorption activity of osteoclasts), including DC-STAMP (dendritic cell-specific transmembrane protein), OSCAR (osteoclast-associated receptor), Acp5 (acid phosphatase 5, tartrate-resistant), and Ctsk (cathepsin K) (9,11,12).
Gene expression can be regulated during transcriptional, post-transcriptional, translational, and post-translational processes (13). Most previous studies of NFATc1-related gene expression have been focused on uncovering the molecular mechanisms at the transcriptional level (9, 10, 14 -18). Nfatc1 was cloned as a RANKL-induced gene, and its expression is transcriptionally induced by transcription factors NFB and c-Fos, which are activated by RANKL/RANK signaling (9,16). Once the Nfatc1 gene is induced, an autoregulatory loop is established in which it binds to its own promoter region, leading to its robust induction (16). On the other hand, it was recently reported that NFATc1 protein is degraded during late stages of osteoclast differentiation by Cbl protein-mediated ubiquitination (19). Therefore, it is conceivable that NFATc1 activity is controlled by its protein stability through post-translational regulation.
The importance of epigenetic modification of histone tails has emerged in studies of the regulation of cellular differentiation, and a histone methylation signature generally appears upstream in epigenetic regulation of transcription. Among various histone methylation modifiers, Jumonji C (JmjC) domaincontaining proteins have recently been reported to bear histone demethylase activity. Indeed, several kinds of JmjC domaincontaining proteins have been identified as histone demethylases, including JHDM1 for the lysine 36 residue in histone H3 (H3K36), JHDM2 for H3K27, and so on (20 -23). On the other hand, two JmjC domain-containing proteins, factor inhibiting hypoxia-inducible factor 1 (HIF1) (FIH1) and Jumonji domain-6 (JMJD6), were characterized as hydroxylases for nonhistone proteins (24 -27). FIH1 post-translationally hydroxylates HIF1␣ on its asparagine residue, and hydroxylated HIF1␣ was shown to be a transcriptionally inactive form because of inhibition of cognate co-activator interaction (24). Likewise, JMJD6 serves as a lysine hydroxylase, modulating the function of RNA splicing factors, suggesting its specific roles in the regulation of alternative RNA splicing (27). Taken together, it is now believed that JmjC domain-containing proteins exert enzymatic activity as protein hydroxylases besides histone demethylases. However, the functions of JmjC domain-containing proteins during osteoclastogenesis remain to be studied.
In this study, we show that one of the epigenetic regulators, a JmjC domain-containing protein, JMJD5, negatively regulates osteoclastogenesis by reducing the stability of NFATc1 protein via its hydroxylase activity. Moreover, JMJD5 was found to induce the association of NFATc1 with E3 ubiquitin ligase Von Hippel-Lindau tumor suppressor (VHL) in HEK293T cells. Thus, the present study demonstrates that the enzymatic function of a JmjC domain-containing protein acts as a post-translational co-repressor of a primary transcription factor during osteoclastogenesis.

EXPERIMENTAL PROCEDURES
Plasmids-Full-length cDNA of mouse JMJD5 was obtained by RT-PCR from total RNA of mouse heart and cloned into the pcDNA3-FLAG or the pcDNA3.1-Myc mammalian expression plasmids (Invitrogen). To generate an expression plasmid of ZsGreen-tagged JMJD5, cDNA for ZsGreen protein was amplified from pZsGreen1-DR plasmid (Clontech) by PCR. Amplified cDNA was subcloned into the C terminus of the previously generated pcDNA3-FLAG-JMJD5 plasmid. The point mutant FLAG-JMJD5-K334A plasmid was generated from the wildtype plasmid by site-directed mutagenesis (28). GST-fused JMJD5 and mutant JMJD5-K334A expression plasmids were generated by subcloning into the pGEX-4T-1 plasmid. Fulllength cDNAs of mouse NFATc1␣ and NFATc1␤ were obtained by RT-PCR from mouse spleen and cloned into the pcDNA3-FLAG or the pcDNA3.1-Myc mammalian expression vector. To generate GST-fused truncated NFATc1␣ constructs, amino acids 1-300, 301-596, and 589 -827 were partially amplified by PCR from pcDNA-FLAG-NFATc1␣ plasmid and then subcloned into the pGEX-4T-1 plasmid. A full-length cDNA of mouse FIH1 was obtained by RT-PCR from mouse testis and cloned into the pGEX-4T-1 plasmid. A full-length cDNA of mouse p105 was obtained by RT-PCR from total RNA of mouse spleen and cloned into the pGEX-4T-1 plasmid. For the GST-fused p105-ARD expression plasmid, amino acids 537-809 were partially amplified by PCR from a full-length plasmid and then subcloned into the pGEX-4T-1 plasmid. Sequences of all cDNAs were verified by automated sequencing. For the NFATc1-response element (RE) reporter plasmid, the mouse Acp5 promoter (Ϫ485 to ϩ525) was amplified from genomic DNA of RAW264 cells and cloned into the upstream region of the luciferase gene derived from pGL4.10 (luc2) vector (Promega). Oligonucleotide short hairpin RNA (shRNA) targeting mouse JMJD5 (shRNA 1; 5Ј-GGAAGGAAGTA-CATCCGATTG-3Ј) was inserted into the pSUPER-Retro vector (OligoEngine) according to the manufacturer's instructions.
Osteoclast Differentiation-RAW264 cells were maintained in ␣-minimum Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (JHR Biosciences), 1% antibioticantimycotic solution (Invitrogen), and 1% minimum Eagle's medium non-essential amino acids (Invitrogen). For osteoclastogenesis, RAW264 cells were seeded in 10-cm dishes at a density of 1 ϫ 10 5 cells/dish or in 6-well plates at a density of 3 ϫ 10 4 cells/well and were treated with 235 ng/ml GST-RANKL (Oriental Yeast Co., Ltd.) for 6 days (29). The medium was changed every 2nd day. Differentiated cells were fixed with 3.7% formaldehyde for 10 min at room temperature and visualized with an Olympus IX70 light microscope equipped with Zeiss AxioVision 4.6.
Establishment of Stable Cell Lines-RAW264 cells were transfected with the pSUPER-mJMJD5-shRNA plasmid by Lipofectamine 2000 (Invitrogen). After 24 h, 3 g/ml puromycin (Sigma) was added, and cultivation was continued for approximately 3 weeks to select stably transfected cells (29). Culture media were changed every 2nd day. After detecting colony formation, several colonies were individually picked up, and their knockdown efficiency was confirmed by RT-PCR.
RNA Analysis-Total RNA was extracted with TRIzol reagent (Invitrogen) according to a previous report (30). For real time RT-PCR, 500 ng of total RNA was reverse transcribed into first strand cDNA with oligo(dT) and random hexamers using a PrimeScript RT Reagent kit (Takara Bio, Inc.) according to the manufacturer's instructions. Real time PCR was performed with SYBR Premix Ex Taq (Takara Bio, Inc.) and the Thermal Cycler Dice Real Time System (Applied Biosystems). For an internal control, the Gapdh gene was used. The primers were purchased from Takara Bio, Inc. For RT-PCR, 1 g of total RNA was reverse transcribed into first strand cDNA with oligo(dT) using SSIII reverse transcriptase (Invitrogen), and cDNA was amplified with the indicated primer sets. All primer sets used are indicated in the supplemental Experimental Procedures.
Luciferase Assay-HEK293T cells in 12-well plates were transfected with the NFATc1-RE reporter plasmid (500 ng) together with each expression vector (700 ng). Transfection efficiency was normalized by co-transfection with pGL4.75 (hRluc/CMV) (5 ng). After 36 h, a luciferase assay was performed with the Dual-Luciferase Reporter assay system (Promega) according to the manufacturer's instructions.
2-OG Decarboxylation Assay-Decarboxylation assays of 14 C-radiolabeled 2-OG were performed as reported previously (26). Briefly, 80 M enzyme, 80 M substrate, 24 M 14 C-radiolabeled 2-OG, 576 M 2-OG, 8 mM sodium ascorbate, 200 M (NH 4 ) 2 Fe(SO 4 ) 2 , and 2 mM DTT were mixed in a total volume of 100 l with 50 mM Tris-Cl (pH 7.5). A 500-l Eppendorf tube containing 200 l of hyamine hydroxide (Fisher Scientific) was added to each tube, and tubes were closed with a rubber septum. After incubation at 37°C for 30 min, the reaction was stopped by setting tubes on ice for 30 min. Then Eppendorf tubes containing the hyamine hydroxide were transferred to scintillation vials and mixed with OptiPhase Hisafe 2 Liquid Scintillation Mixture (Fisher Scientific). Total released 14 C counts were quantified using an LSC-5100 scintillation counter (Aloka). Data were analyzed by two-tailed Student's t test, and p Ͻ 0.05 was accepted as significant.

JMJD5 Represses Osteoclastogenesis by Reducing Expression of Several
Osteoclast-specific Genes-To clarify epigenetic regulation during osteoclastogenesis, we focused on JmjC domain-containing proteins, which are histone demethylases. Their physiological roles in bone metabolism have not LacZ-specific shRNA was used for a negative control. C, stable clones were treated with RANKL for osteoclast formation. Differentiated cells were fixed at 3.5 days and assessed with an Olympus IX70 light microscope. Black arrowheads indicate multinucleated osteoclasts. D, during RANKL treatment, cells were harvested at the indicated time points, and total RNAs were prepared. Expression levels of osteoclast-specific genes (Ctsk, DC-STAMP, and Acp5) were analyzed by real time RT-PCR. been characterized. We first observed their expression profiles during osteoclastogenesis. RAW264 cells were differentiated to mature osteoclasts by RANKL treatment, and total RNA was prepared from each differentiation stage to assess the expression levels of JmjC domain-containing proteins. Through a real time RT-PCR analysis, we detected their various expression patterns. The expression levels of JMJD1a/2a transcripts increased, whereas those of JMJD5/3 decreased (Fig. 1A and supplemental Fig. S1). On the other hand, the expression level of JMJD4 transcripts was unaltered during osteoclastogenesis.
Although various factors, such as NFATc1, promote osteoclastogenesis and are induced during osteoclastogenesis (9), little is known about intracellular negative regulators of osteoclastogenesis. Because JMJD5/3 expression levels decreased during osteoclastogenesis, we reasoned that their functions were related to negative regulation. To test this idea, stable cell lines constitutively expressing JMJD5-or JMJD3-specific shRNAs with the puromycin-resistant gene were established from RAW264 cells by puromycin selection. Each cell line was then treated with RANKL for osteoclast formation. The shRNA knockdown was confirmed by attenuated expression levels of either JMJD3 or JMJD5 transcripts in the stable lines ( Fig. 1B and supplemental Fig. S2). The stable RAW264 JMJD3 knockdown line differentiated normally into osteoclasts and displayed normal expression levels of osteoclast-specific genes Acp5, DC-STAMP, and Nfatc1 (supplemental Fig. S2). However, when JMJD5 was knocked down, osteoclastic maturation was accelerated ( Fig. 1C and  supplemental Fig. S3). The same results were observed in bona fide osteoclasts derived from murine bone marrow macrophages (supplemental Fig. S4). Consistent with the accelerated formation of osteoclasts, the induced expression of Ctsk, DC-STAMP, and Acp5 genes was also observed (Fig.   FIGURE 2. Overexpressed JMJD5 did not affect methylation levels on tested histone H3 lysine residues. A, HEK293T cells were transfected with JMJD5-ZsGreen. After 36 h of transfection, cells were fixed and analyzed by immunocytochemistry. JMJD5 was visualized by the tagged ZsGreen protein (green). Histone H3 lysine methylation levels were observed by staining with the indicated primary antibodies (red). IgG antibody was used for the negative control. DAPI was used to stain nuclei (blue). B, HEK293T cells were transfected with FLAG-mJMJD5. After 48 h of transfection, chromosomal extracts were prepared and analyzed by immunoblotting with the indicated primary antibodies. Histone H3 was used as loading control for immunoblotting. 1D). These results implied that JMJD5 is a negative regulator for osteoclast differentiation.
JMJD5 Represses Transcriptional Activity of NFATc1 by Reducing Level of Its Protein-JMJD5 is reportedly an H3K36me2 demethylase (34). To verify its histone demethylase activity, we measured the methylation level of histone H3K36 in HEK293T cells overexpressing JMJD5 using immunofluorescence and immunoblotting analyses. However, we could not detect the expected alteration in the level of H3K36me2 ( Fig.  2A). Although other methylated histone lysine residues were tested with several methylation-specific histone antibodies, we did not detect altered methylation levels on any tested lysine residues of histone H3 (Fig. 2, A and B). Thus, from these observations, we assumed that JMJD5 is not a histone demethylase.
To assess other enzymatic functions of JMJD5, we examined the cellular distribution of JMJD5 protein during osteoclastogenesis based on its solubility differences in lysis buffers. JMJD5 was present in the chromatin fraction as expected, and its distribution appeared to be unaffected by RANKL stimulation (supplemental Fig. S5). JMJD5 was also abundant in the soluble nuclear fraction, suggesting that JMJD5 serves as a soluble transcriptional co-regulator in the nucleus rather than a histone demethylase bound to chromatin.
To test this idea, we examined the association of JMJD5 with the primary transcription factors for osteoclastogenesis, i.e. NFATc1, NFB, and AP-1. We expressed and purified recombinant affinity-tagged full-length proteins from E. coli and then performed in vitro pulldown assays. His-tagged JMJD5 was an interactant with tested GST-fused transcription factors NFATc1␣/␤, c-Jun, and NFB (p105/50) but not c-Fos (supplemental Fig. S6). To further address their association, we tested the in vivo interactions within cells. Total cellular extracts were prepared from HEK293T cells expressing FLAG-JMJD5 and Myc-NFATc1, and these proteins were immunoprecipitated by anti-FLAG antibody. The bound proteins were eluted with FLAG peptides and observed by immunoblotting with anti-Myc and anti-NFB (p65/50) antibodies. Among the tested factors, only NFATc1 was detected (Fig. 3A). The interaction of JMJD5 with NFATc1 was still significant with a catalytically inactive mutant form of JMJD5 (JMJD5-K334A) in which the ␣-ketoglutarate (2-OG)-binding residue Lys-334 was substituted by alanine (Fig. 3A, last lane).
NFATc1 is a primary transcription factor for osteoclastogenesis, and it regulates the expression of osteoclast-specific genes (9,16). From its interaction with JMJD5, we hypothesized that JMJD5 co-regulated the transcriptional activity of NFATc1. To examine this possibility, we constructed an NFATc1-RE reporter plasmid by inserting the Ϫ485 to ϩ525 region of the Acp5 gene in front of the luciferase gene and measured the transcriptional activity of NFATc1 when JMJD5 was overexpressed. The luciferase activity was increased ϳ15-fold when NFATc1 was expressed, whereas it was significantly decreased by co-expression of JMJD5 (Fig. 3B). To characterize the repressive function of JMJD5, we then analyzed the lysates used in the luciferase assay by immunoblotting, testing whether JMJD5 regulated an event after transcription. Unexpectedly, the NFATc1 protein level was significantly reduced by expression of JMJD5, indicating that decreased NFATc1 transcriptional activity is due to the reduced protein level of NFATc1 (Fig. 3C). Such JMJD5 action was not seen for NFB (p50/65) (supplemental Fig. S7, A and B). Importantly, the suppressive actions of JMJD5 on the transcriptional activity and protein level of NFATc1 were not observed when the JMJD5-K334A mutant was expressed (Fig. 3C, last lane), suggesting that the repressive function of JMJD5 against NFATc1 depended on its 2-OG-dependent activity.
Post-translational Destabilization of NFATc1 Protein by JMJD5-To determine whether JMJD5 was capable of modulating the stability of NFATc1 protein, the level of exogenous NFATc1 protein was monitored during the knockdown of JMJD5. Human JMJD5-specific siRNA was introduced into HEK293T cells, and the level of exogenously expressed NFATc1 was measured by immunoblotting. As anticipated, reduced expression of JMJD5 was seen, and the NFATc1 protein level was significantly elevated (Fig. 4A). Furthermore, when JMJD5 was overexpressed in these knockdown cells, the NFATc1 protein level was indistinguishable from that in the control cells. The JMJD5-K334A mutant was not effective in regulating the protein level of NFATc1 (Fig. 4A).
Next, we asked whether the level of endogenous NFATc1 protein was indeed regulated by JMJD5. Using the stable cell lines expressing JMJD5-specific shRNA, total RNA and cellular extracts were prepared for RT-PCR and immunoblotting. Although the level of NFATc1 transcripts was unaltered (Fig.  4B), the level of NFATc1 protein was greater than that in the control knockdown cells (Fig. 4C). Taken together, it appears that the mechanism of down-regulating the level of NFATc1 protein by JMJD5 is post-translational.
JMJD5 Hydroxylates NFATc1 Protein through Its Enzymatic Activity-Recently, some members of JmjC domain-containing proteins have been shown to serve as protein hydroxylases, for example JMJD6 and FIH1 (24,27). Because JMJD5 exhibits sequence homology with JMJD6 and FIH1, it is conceivable that JMJD5 is a protein hydroxylase. This idea is further supported by the present findings that a catalytically inactive JMJD5 mutant was impaired in the regulation of the NFATc1 protein level.
To address this idea, we performed an in vitro "2-OG decarboxylation assay" with GST-fused recombinant proteins (25,26). Recombinant NFATc1 deletion mutant proteins were expressed and purified from E. coli (Fig. 5A), and each mutant was incubated with the purified GST-tagged JMJD5 to induce 14 CO 2 formation. After confirming FIH1 activity for the known hydroxylase substrate (p105) (Fig. 5B, lanes 7 and 8) (26), we tested whether NFATc1 is a hydroxylase substrate for JMJD5. The C-terminal domain of NFATc1 was found to serve as a substrate, although its 2-OG decarboxylation was less than that for p105 (Fig. 5B, lanes 5 and 6). To confirm that the observed decarboxylation activity of JMJD5 mediated by the JmjC domain, the JMJD5-K334A mutant was also used for this assay. As anticipated, a reduction in 14 CO 2 formation was seen (Fig.  5C). These findings suggest that JMJD5 has hydroxylase activity and that NFATc1 is its substrate.
Destabilization of NFATc1 by JMJD5 Mediates Ubiquitinmediated Proteasomal Degradation-Here, we showed that JMJD5 destabilizes NFATc1 protein, presumably through its hydroxylase activity. To determine whether this process was related to protein degradation, we first examined the effect of a proteasome inhibitor (MG132). In HEK293T cells expressing exogenous NFATc1, NFATc1 protein did indeed accumulate in a time-dependent manner during MG132 treatment (Fig. 6A,  lanes 4 -6). Moreover, the reduced protein level of NFATc1 caused by co-expression of JMJD5 was completely reversed by MG132 treatment (Fig. 6A, lanes 7 and 8). We also confirmed in osteoclasts derived from bone marrow macrophages that endogenous NFATc1 protein accumulated during MG132 treatment to the same level as that seen when JMJD5 was knocked down (supplemental Fig. S8).
Next, we examined whether the proteasomal degradation of NFATc1 was ubiquitin-dependent. Immunoprecipitation by anti-NFATc1 antibody was performed in HEK293T cells exogenously expressing NFATc1 and JMJD5 after which the ubiquitinated NFATc1 was detected with an anti-ubiquitin antibody. Ubiquitinated NFATc1 was barely detectable when MG132 was used (Fig. 6B, upper panel, lane 4), whereas it was After 48 h of siRNA transfection, cells were also transfected with human JMJD5-specific siRNA together with FLAG-NFATc1 and FLAG-JMJD5-WT or FLAG-JMJD5-K334A. After 48 h, total cell lysates were prepared and analyzed by immunoblotting with the indicated antibodies. ␤-Actin was used as a loading control for immunoblotting. An asterisk indicates a nonspecific band. B and C, total RNAs and cell lysates were prepared from two stable clones constitutively expressing JMJD5 shRNA in RAW264 cells. Expression levels of the Nfatc1 gene were detected by real time RT-PCR (left). Relative mRNA levels are represented by values that were normalized to Gapdh transcript levels. The error bars represent S.D. values from at least triplicate experiments. The p value for statistical difference was calculated using a two-tailed Student's t test. Also, the NFATc1 protein level was detected by immunoblotting with an anti-NFATc1 antibody (right). ␤-Actin was used as a loading control for immunoblotting. The signals obtained from the immunoblotting were quantified with the ImageJ program. Ctrl, control.
significantly increased by co-expression of JMJD5 (Fig. 6B,  upper panel, lane 6). In addition, when the immunoprecipitants were probed with anti-FLAG antibody, the band shifts representing ubiquitinated NFATc1 were made more evident by JMJD5 co-expression (Fig. 6B, middle panel, lanes 5 and 6). These findings suggest that JMJD5 facilitates the ubiquitin-dependent proteasomal proteolysis of NFATc1.
Because JMJD5 does not contain a specific domain to ubiquitinate, it appeared likely that ubiquitin ligases were required for NFATc1 degradation by JMJD5. Among them, we focused on VHL (a component of the multimeric E3 ubiquitin ligase) because this factor is known to directly bind to HIF1␣ protein through a hydroxylated proline residue on HIF1␣ for subsequent ubiquitination and proteasomal degradation (35)(36)(37). Therefore, we asked whether degradation of NFATc1 by JMJD5 was mediated by VHL. Total cellular extracts of HEK293T cells overexpressing NFATc1 and JMJD5 were immunoprecipitated with anti-FLAG antibody and probed with an anti-VHL anti-body. VHL was visible when the cells were treated with MG132, indicating that it interacted with NFATc1 (Fig. 6C, lane 4). Furthermore, the interaction of NFATc1 with VHL was potentiated when JMJD5 was overexpressed (Fig. 6C, lane 6). Taken together, these results imply that JMJD5 facilitates the association of NFATc1 with VHL E3 ligase, and as a result, NFATc1 is prone to undergo ubiquitin-mediated proteasomal degradation (Fig. 6D).

DISCUSSION
JMJD5 Attenuates Osteoclastogenesis-In the present study, we showed that JMJD5 functions as a novel repressor for osteoclastogenesis. JMJD5 was found to serve as a protein hydroxylase for NFATc1, thereby facilitating NFATc1 protein degradation through association with VHL E3 ubiquitin ligase. This is a novel mechanism of regulating NFATc1 activity during osteoclastogenesis at a post-translational level. Interestingly, the expression level of JMJD5 was decreased the 1st day after RANKL treatment in contrast to NFATc1 expression during osteoclastic differentiation (Fig. 1A). It appears that JMJD5 functions to maintain the undifferentiated state of osteoclastic progenitor cells. Because JMJD5 appears to be a co-regulator of NFATc1 at the post-translational level, this factor represents a previously uncharacterized class of regulators in osteoclastogenesis. The present study also implies the possible presence of other co-regulators of NFATc1 at other levels. Additionally, it is quite conceivable from recent reports of epigenetic co-regulators of DNA-binding transcription factors that co-regulators of other important transcription factors directing osteoclastogenesis remain to be identified.
It is currently believed that epigenetic regulation through histone methylation/demethylation plays a central role in gene regulation during differentiation of many types of cells, presumably also bone cells. In particular, such enzymes (histone methyltransferases and demethylases) would transcriptionally co-regulate the function of transcription factors governing osteoclastic differentiation. For example, the JmjC domain-containing protein NO66 reportedly acts as a negative regulator of Osterix, an essential transcription factor in osteoblastogenesis (38). Consistent with biochemical characterization of NO66 as a direct interactant with Osterix, NO66 co-repressed the Osterix-mediated gene cascade through demethylation of active histone methylation marks (H3K4 and H3K36). Gene regulation by epigenetic marks is also evident during osteoclastic differentiation. Likewise, promoter activity of the Nfatc1 gene appears to be under negative control by methylation of histone H3K4me3 and H3K27me3 at an undifferentiated stage of osteoclasts (39). In contrast, the inactive methylation markers of H3K27me3 are removed for Nfatc1 gene induction along with RANKL-induced osteoclastogenesis (39). Therefore, it is quite conceivable that JmjC domain-containing proteins are critical for proper regulation of nuclear events during osteoclastic differentiation through histone demethylation. After 48 h of transfection, total cell lysates were prepared and analyzed by immunoblotting with anti-FLAG antibody. ␤-Actin was used as a loading control for immunoblotting. The signals obtained from the immunoblotting were quantified with the ImageJ program. B and C, HEK293T cells were co-transfected with FLAG-NFATc1 and Myc-mJMJD5 and then treated with MG132 (10 M) for 16 h. After 48 h of transfection, total cell lysates were prepared and subjected to immunoprecipitation (IP) with an anti-NFATc1 antibody (B) or anti-FLAG M2-agarose (C). Bound proteins were detected by immunoblotting with the indicated antibodies. GAPDH was used as loading control for immunoblotting. D, suggested model. JMJD5 Lacks Histone Demethylase Activity-It was recently reported that JMJD5 is a histone demethylase (34). However, in our experimental settings, no histone demethylase activity was detected for methylated H3K36 and the known methylated histone lysine residues (Fig. 2). Based on the abundant distribution of JMJD5 in the nucleoplasm, we next raised the possibility that its associating protein, NFATc1, is a substrate of JMJD5 demethylase. However, methylated lysine residues were not evident on NFATc1 protein when tested by immunoblotting with anti-Lys methylation antibody (data not shown). From the fact that JMJD5 is a member of the same JmjC domain-containing protein group as FIH1 and JMJD6 protein hydroxylases (24 -27, 40), we next asked whether JMJD5 functioned as a hydroxylase of NFATc1. Indeed, hydroxylation of NFATc1 by JMJD5 was detected with an in vitro 2-OG decarboxylation assay (Fig. 5), suggesting that JMJD5 bears uncharacterized enzymatic function as an NFATc1 hydroxylase. As JMJD5 is found in the soluble nuclear fraction (supplemental Fig. 5), other non-histone proteins may be hydroxylation substrates for JMJD5.
These previous and present observations suggest that JMJD5 has dual functions, and the functions are very dependent on its cellular locations, functioning as a histone demethylase in chromatin and functioning as a protein hydroxylase in soluble nucleus. In our experimental system, although distribution of JMJD5 was detected in chromatin, it was more abundantly located in soluble nucleus (supplemental Fig. 5). Thus, we think that we could easily detect its function as a hydroxylase but not its histone demethylase activity, which may occur at a level below our detection.
Hydroxylation Is One Post-translational Modification Directing NFATc1 Functions-Post-translational modifications are important to regulate the function or expression of many kinds of proteins. It is well known that NFATc1 is modified by phosphorylation (41)(42)(43). In its phosphorylated form, NFATc1 is present in the cytoplasmic compartment and is imported into the nucleus when dephosphorylated by calcineurin. Recently, it was reported that NFATc1 is ubiquitinated in the cytoplasm for its degradation during late stages of osteoclastogenesis (19). However, it remained to be uncovered whether post-translational modifications regulated NFATc1 during osteoclastogenesis. In the present study, we provide evidence of a novel regulatory system for NFATc1 protein stability by post-translational modification. Specifically, NFATc1 is hydroxylated by JMJD5 and degraded via a ubiquitin-mediated proteasomal pathway. Hypoxia-specific transcription regulator HIF1␣ is constitutively synthesized at a high level even under normoxia. However, its level is repressed through hydroxylation on specific proline residues by prolyl hydroxylases and subsequent proteasomal degradation until hypoxia is induced. In this respect, it will be interesting to understand at the molecular level how JMJD5 enzymatic activity is controlled during osteoclastogenesis.