A crucial role for reactive oxygen species in RANKL-induced osteoclast differentiation

Receptor activator of NF-κB (nuclear factor-κB) ligand (RANKL; also called TRANCE [tumor necrosis factor (TNF) activation-induced cytokine], ODF [osteoclast differentiation factor], and OPGL [osteoprotegerin ligand])^1-4  is a key factor stimulating the differentiation and activation of osteoclasts and, therefore, is essential for bone remodeling.⁵  The binding of RANKL to its receptor RANK leads to recruitment of TNF receptor-associated factor 6 (TRAF6) to the cytoplasmic domain of RANK, thereby resulting in the activation of distinct signaling cascades mediated by mitogen-activated protein (MAP) kinases, including c-Jun N-terminal kinase (JNK), p38 MAP kinase (p38), and extracellular signal-regulated kinase (ERK).⁶  It has been shown that JNK1-activated c-Jun signaling in cooperation with nuclear factor of activated T cells (NFAT) is key to RANKL-regulated osteoclast differentiation.⁷  In addition, stimulation of p38 results in the downstream activation of the mi/Mitf (microphthalmia/microphthalmia transcription factor), which controls the expression of the genes encoding tartrate-resistant acid phosphatase (TRAP) and cathepsin K, indicating the importance of p38 signaling cascades.⁶  Although our understanding of signaling pathways associated with osteoclast differentiation has advanced considerably recently, the mechanism of RANKL-mediated osteoclastogenesis, specifically the molecular linkage between TRAF6 and MAP kinases, is still unknown.

At high concentrations, reactive oxygen species (ROSs) cause oxidative stress that has been viewed as deleterious phenomena, including inflammatory response, apoptosis, or ischemia.⁸  Recent studies, however, indicate that small nontoxic amounts of ROS may play a role as a second messenger in the various receptor signaling pathways.^9-13  Osteoclasts have shown to be activated by ROSs to enhance bone resorption,¹⁴  but little attention has been given to the role of ROSs in differentiation of macrophages and monocytes into osteoclasts. Signaling molecules such as JNK and p38, which are known to be essential for osteoclast differentiation,^6,7  are sensitive to activation by ROSs.^11,12  Thus, we hypothesized that signaling cascade(s) can be modulated by ROSs in bone marrow monocyte-macrophage lineage (BMM) cells.

Here, we show that RANKL generates ROSs in BMM cells. Examination of the mechanism by which RANKL generates ROSs revealed the involvement of TRAF6, Rac1, and NADPH (nicotinamide adenine dinucleotide phosphate) oxidase 1 (Nox1). These data suggest that RANKL-mediated ROS production serves to regulate RANKL signaling pathways, including JNK and p38 activation required for osteoclast differentiation.

2′,7′-dichlorofluorescein diacetate (DCFH-DA) was purchased from Molecular Probes (Leiden, The Netherlands); all other chemicals and FLAG (five NH2-terminally deleted epitope-tagged) epitope (M2) were from Sigma-Aldrich (St Louis, MO); recombinant macrophage colony-stimulating factor (M-CSF) was from R&D Systems (Minneapolis, MN); RANKL was from PeproTech (Rocky Hill, NJ); and RANK-crystallizable fragment (Fc) was previously described.¹⁵  Polyclonal antibodies (Abs) specific for p38, phosphorylated p38 (Thr¹⁸⁰/Tyr¹⁸²), JNKs (p46 and p54), phosphorylated JNKs (Thr¹⁸³/Tyr¹⁸⁵), and monoclonal anti-phosphorylated p42/44 MAP kinase (ERKs) were purchased from New England Biolabs (Beverly, MA). TRAF6 (H-274) was from Santa Cruz Biotechnology (Santa Cruz, CA) and hemagglutinin (HA; 12CA5) was from Boehringer Mannheim (Mannheim, Germany). A monoclonal antibody (mAb) specific for Rac1 was from Transduction Laboratories (Lexington, KY).

Expression constructs encoding FLAG-tagged wild-type RANK, HA-tagged TRAF6 (amino acids [aa] 289-530), and maltose-binding protein (MBP)–RANKcyt, which comprises the cytoplasmic tail of RANK fused to MBP, were described.¹⁶  To generate a tetracycline-regulated FLAG-tagged RANK expression plasmid (pcDNA/TO-RANK), FLAG-tagged RANK was amplified by polymerase chain reaction (PCR) and inserted into pBluescriptSK+, and then subcloned into the EcoRI and XbaI sites of pcDNA/TO vector (Invitrogen, Carlsbad, CA). A retroviral vector, pMX-puro-TRAF6ΔZ3 was provided by Dr J. Inoue (Keio University, Kanagawa, Japan).

Nonadherent bone marrow–derived monocyte-macrophage cells derived from C57BL/6 mice were seeded and cultured in α-minimal essential medium (α-MEM) (GIBCO-BRL, Grand Island, NY) with 10% fetal bovine serum (FBS) containing 10 ng/mL M-CSF. After 2 days, adherent cells were used as BMM cells after washing out the nonadherent cells, including lymphocytes. TRAF6-deficient mice were described previously,¹⁷  and monocyte-macrophage progenitor cells of TRAF6-deficient mice were derived from the spleen and similarly cultured with 10 ng/mL M-CSF for 2 days. These osteoclast precursor cells were further cultured in the presence of 50 ng/mL soluble RANKL and 10 ng/mL M-CSF to generate osteoclasts. After 5 days, cells were fixed and stained for TRAP. TRAP-positive multinucleated (> 3 nuclei) cells were counted as osteoclast-like cells. The cells were observed with a Zeiss Axiovert 200 microscope equipped with a Plan-Neofluor 20×/0.50 objective lens (Carl Zeiss, Oberkochen, Germany), and images were obtained with an AxioCam HR (Carl Zeiss) using AxioVision 3.1 software (Carl Zeiss).

Isolated BMM cells were extensively washed to remove exogenous growth factors, cultured in media with low serum (0.5% FBS, 6 hours), then stimulated by adding RANKL in the presence or absence of inhibitors as indicated. Inhibitors were added to cultures 60 minutes prior to addition of RANKL (50 ng/mL). After stimulation, the cells were washed in ice-cold phosphate-buffered saline, lysed, and subjected to Western blot analysis or immunoprecipitation as described in the next paragraph. 293 cells expressing inducible FLAG-tagged RANK were transfected by calcium phosphate precipitation.¹⁶  The cells were processed for analysis 24 hours after transfection. All cells were harvested and lysed in extraction buffer (20 mM HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid) [pH 7.9], 150 mM NaCl, 1% Triton X-100, 10% glycerol, 5 mM EDTA (ethylenediaminetetraacetic acid), 1 mM Na₃VO₄, 1 mM phenylmethylsulfonylfluoride, 1 μg/mL leupeptin, 0.1 U/mL aprotinin) and cleared by centrifugation to obtain whole-cell extracts. Immunoprecipitations were performed on whole-cell extracts with anti-FLAG mAb. In brief, 1 μg Ab coupled to 30 μL protein G–Sepharose was incubated with 400 μg whole-cell extracts for 4 hours. The immunoprecipitates were then subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting. Maltose beads were used to precipitate the MBP-RANK fusion proteins (1 hour, 4°C) and subjected to SDS-PAGE as described earlier in this paragraph.

Intracellular production of ROSs was assayed as described.¹⁰  In brief, at various times after stimulation with RANKL, dishes of confluent cells were washed with α-MEM lacking phenol red and then incubated in the dark for 5 minutes in Krebs-Ringer solution containing 5 μM DCFH-DA. Culture dishes were transferred to a Zeiss Axiovert 135 inverted microscope (Carl Zeiss) that was equipped with a 20 × Neofluor objective and Zeiss LSM 410 confocal attachment, and DCF (2′,7′-dichlorofluorescein) fluorescence was measured with an excitation wavelength of 488 nm and emission at 515-540 nm. To avoid photooxidation of DCF, the fluorescence image was collected by a single rapid scan (4-line average, total scan time of 4.33 seconds). After collection of the fluorescence image, the cells were imaged by digital interference contrast, and the mean relative fluorescence intensity for each group of cells was then measured by Carl Zeiss vision system (KS400, version 3.0).

The 293 cell line expressing FLAG-tagged RANK protein in an inducible manner was generated with use of a tetracycline-regulated gene expression system (Tet-ON gene expression system, T-REX system; Invitrogen). 293 cells were cotransfected with a regulatory plasmid, pcDNA6/TR (Blasticidin resistant) and the inducible expression plasmid encoding RANK, pcDNA/TO-RANK (Zeocin resistant). Cells were selected in culture medium containing 125 μg/mL Zeocin and 3 μg/mL Blasticidin. Zeocin- and Blasticidin-resistant clones were examined for expression of the FLAG-tagged RANK by Western blotting using anti-FLAG Ab.

An adenovirus containing human catalase (Adcat) or green fluorescent protein (GFP; AdGFP), a generous gift of S. G. Rhee (National Institute of Health, Bethesda, MD), was prepared as previously described.¹⁸  The recombinant adenoviral-expressing vector encoding the catalase and the control vector were used to infect BMM cells. The infection efficiency was monitored by the expression of GFP.

To generate infectious retroviral particles, Plat-E cells (a kind gift from T. Kitamura, University of Tokyo, Tokyo, Japan) were transfected with the retroviral vectors, and supernatant collected from 24 to 48 hours after transfection was used as the viral stocks. For retrovirus infection, RAW264.7 or RAW-RacN17 cells (provided by Dr J. Kim, Korea University, Seoul, Korea), stably expressing a dominant-negative form of Rac1, were incubated with retrovirus stock with polybrene (4 μg/mL) for 6 hours. Two days after the exposure to virus, samples of the infected cells were assayed for infection efficiency, and the rest of the cells were further cultured in the presence or absence of RANKL for the osteoclast formation assay.

Reverse transcription (RT)–PCR was performed as described.²  Nucleotide sequences of PCR primers are as follows: Nox1 forward, 5′-AAGTGGCTGTACTGGTTGG-3′; Nox1 reverse, 5′-GTGAGGAAGAGTCGGTAGTT-3′; Nox2 forward, 5′-ACTTCTTGGGTCAGCACTGG-3′; Nox2 reverse, 5′-ATTCCTGTCCAGTTGTCTTCG-3′; Nox3 forward, 5′-CAGGCTCAAATGGACGGAAAGG-3′; Nox3 reverse, 5′-CCATGCCAATGCGGAACCCAGA-3′; Nox4 forward, 5′-CCTTGAACTGAATGCAGCAA-3′; Nox4 reverse, 5′-ACCACCTGAAACATGCAACA-3′; β-actin forward, 5′-TCACCCACACTGTGCCCATCTAC; and β-actin reverse, 5′-GAGTACTTGCGCTCAGGAGGAGC-3′.

Duplex, small interfering (si) RNA oligonucleotides specific for murine Nox1 or Nox2 and a nonspecific control were synthesized using Silencer siRNA construction kit (Ambion, Austin, TX) with the following siRNA-encoding DNA oligonucleotides: Nox1 antisense, 5′-AACAACAGCACTCACCAATGCcctgtctc-3′; Nox1 sense, 5′-aaGCATTGGTGAGTGCTGTTGcctgtctc-3′; Nox2 antisense, 5′-AAGATGCCTGGAAACTACCTAcctgtctc-3′; Nox2 sense, 5′-aaTAGGTAGTTTCCAGGCATCcctgtctc-3′; control antisense, 5′-AAGACTCTTCTCTGTTCCAAGcctgtctc-3′; and control sense, 5′-aaCTTGGAACAGAGAAGAGTCcctgtctc-3′ (the 8 italicized nucleotide sequences at the 3′ end are complementary to the T7 promoter primer provided with the Silencer siRNA construction kit, and siRNA-producing sequences are capitalized). siRNAs were transfected into BMM cells using X-tremeGENE siRNA transfection reagent (Roche, Indianapolis, IN) according to the manufacturer's recommendations. After 48 hours, cultures were used for RT-PCR analysis. Transfection efficiency (usually > 95%) was assessed in parallel wells by using Silencer siRNA labeling kit (Mirus, Madison, WI).

Cell lysates (150 μg) obtained from either RAW264.7 or RAW-RacN17 cells with or without stimulation with RANKL were incubated with 15 μg recombinant GST-PBD (glutathione S-transferase–p21-binding domain; human Pak1 aa 67-150) for 1 hour at 4°C, and then Rac activity was assayed as described.¹⁹ 

Data are presented as means plus or minus SDs from at least 3 independent experiments. Statistical significance was determined using 1-way analysis of variance (ANOVA) followed by the Student t test (*P < .05, **P < .005).

We first examined whether ROS production is detectable in osteoclast precursors upon RANKL treatment. Intracellular production of ROSs in the cells was measured with the cell-permeant, oxidation-sensitive dye DCFH-DA by using laser-scanning confocal microscopy. Stimulation of both RAW264.7 cells and BMM cells resulted in an increase in the intensity of DCF fluorescence (Figure 1A), indicating oxidation by hydroxyl radicals such as H₂O₂.²⁰  The addition of soluble RANK-Fc significantly inhibited RANKL-stimulated ROS production in BMM cells (Figure 1B), confirming that RANKL generates ROSs through its cognate receptor. RANKL induced ROS production in a dose-dependent manner in BMM cells (Figure 1C), and the amount of ROS rapidly increased to its maximum level around 10 minutes after RANKL addition and thereafter decreased toward its basal level (Figure 1D). Taken together, these data indicate that RANKL generates ROSs in osteoclast precursors.

In view of previous results demonstrating that the binding of TRAF6 to the cytoplasmic domain of RANK is critical for RANKL signaling and osteoclast differentiation,^16,21,22  it is an interesting issue whether ROS production is linked to TRAF6 binding. To answer this question, we generated a 293 cell line termed 293-RANK inducibly expressing FLAG epitope-tagged RANK. Expression of FLAG-RANK protein was detected by Western blotting in 293-RANK, 8 hours after the addition of tetracycline, but not in the uninduced 293-RANK cells (Figure 2A, top). RANKL treatment after FLAG-RANK induction in this clone increased the intensity of DCF fluorescence (Figure 2A, bottom). However, overexpression of TRAF6.DN (aa 289-530), a dominant-interfering mutant of TRAF6, significantly decreased the level of ROS induction in a dose-dependent manner (Figure 2B). IFN-γ is known to induce a rapid degradation of TRAF6 (Figure 2C, bottom, lane 3) by activation of the ubiquitin-proteasome system.²³  Consistently, treatment of RANKL in IFN-γ–treated cells prevented ROS production (Figure 2C, top), which might be caused by a decreased level of TRAF6 (Figure 2C, bottom, lane 2 versus 4). To further explore the importance of TRAF6 in ROS production, we compared ROS levels in TRAF6-null osteoclast precursors and wild-type cells. The ROS levels induced by RANKL were significantly reduced in TRAF6-null osteoclast precursors (Figure 2D). Collectively, these findings indicate that TRAF6 plays a key linkage role in ROS production by RANKL.

To correlate ROS production with RANKL responses, including MAP kinase activation and osteoclast differentiation, we first tested whether H₂O₂ alone activates MAP kinases in BMM cells. Activation of JNK, p38, or ERK was observed by treatment with H₂O₂ in a dose-dependent manner (Figure 3A). We next tested the effect of NAC, a chemical oxidant scavenger, on the RANKL responses. As expected, when cells were treated with N-acetylcysteine (NAC), RANKL-stimulated DCF fluorescence was significantly reduced in a dose-dependent manner (Figure 3B). In addition, the increasing amount of NAC exerted an inhibitory effect on RANKL-mediated ERK, JNK, or p38 MAP kinase activation (Figure 3C). We next examined the effect of a noncytotoxic concentration of NAC (30 mM) on osteoclast differentiation. NAC treatment strongly inhibited the formation of TRAP-positive osteoclasts with high numbers (> 3) of nuclei compared with that in the absence of NAC (Figure 3D). Taken together, these data show that the effect of NAC onROSs paralleled its effects on MAP kinase activation and osteoclast differentiation and imply that the elevated levels of ROSs might be the cause of MAP kinase activation, which in turn leads to stimulation of osteoclastogenesis.

To further confirm the relevance between ROS production and osteoclast differentiation, BMM cells were infected with catalase-expressing adenovirus or with a control virus. As expected, no effect on ROS production was observed with the control virus (data not shown). In contrast, expression of catalase in BMM cells blocked RANKL-induced ROS production (data not shown) and inhibited the formation of TRAP-positive osteoclasts in a dose-dependent manner (Figure 4), showing the involvement of ROSs in osteoclast differentiation.

The formation of ROSs in both phagocytic and nonphagocytic cells involves membrane-localized NADPH oxidases (Noxs).²⁴  To determine whether an enzyme functionally similar to the NADPH oxidases is involved in RANKL-induced ROS production, we treated BMM cells with diphenylene iodonium (DPI), a specific inhibitor for flavoprotein that is a constituent of the NADPH oxidase complex. The addition of DPI abolished the rise in DCF fluorescence by RANKL treatment in a dose-dependent manner (Figure 5A). Likewise, DPI treatment blocked activation of JNK, p38, or ERK mediated by RANKL (Figure 5B) as well as RANKL-dependent osteoclastogenesis of BMM cells (Figure 5C). The fact that treatment of osteoclast precursors with DPI blocked RANKL responses supports the view that ROSs produced by an NADPH oxidase is required for osteoclast differentiation.

To find which Nox isozyme is responsible for the RANKL responses, we examined the expression of Nox isozymes in BMM cells with semiquantitative RT-PCR. As expected, Nox2, gp91phox, was found to be the main isotype expressed in BMM cells (Figure 6A). Interestingly, expression of Nox1 was also detectable at a low level in BMM cells, whereas expression of the other Nox members such as Nox3 and Nox4 was undetectable. To investigate the role of Nox1 in ROS production and osteoclastogenesis, we performed loss-of-function experiments using siRNAs. Nox1 was effectively knocked down by Nox1-specific siRNAs, as shown by RT-PCR (Figure 6B). Remarkably, the silencing of Nox1 in BMM cells resulted in a significant decrease in ROS production in response to RANKL stimulation (Figure 6C). Such reduction was not observed in IL-1β stimulation of BMM cells which also generates ROS production (Figure 6C). Likewise, siRNA of Nox1 blocked RANKL-mediated osteoclastogenesis (Figure 6D) as well as activation of JNK or p38 mediated by RANKL (Figure 6E). However, Nox2 knockdown had no effects on the ROS production and osteoclastogenesis induced by RANKL (data not shown). These results suggest that Nox1 is the critical mediator of the RANKL-mediated responses in BMM cells.

The small guanosine triphosphatase (GTPase) Rac1, predominant in macrophages and monocytes,²⁵  is a cytosolic component of NADPH oxidase complex and is responsible for the activation of NADPH oxidases.²⁶  To examine the involvement of Rac1 in RANKL-mediated ROS production and osteoclast differentiation, RAW264.7 and RAW-RacN17 cells stably expressing a dominant-negative form of Rac1 were used. As shown in Figure 7A, RANKL activated endogenous Rac1 in RAW264.7 cells but not in RAW-RacN17 cells, and this activation was greatly reduced by expression of a dominant-negative form of TRAF6 (T6ΔZ3), suggesting that the Rac1 activation by RANKL is dependent on TRAF6. Moreover, both ROS production (Figure 7B) and osteoclastogenesis (Figure 7C) were inhibited in RAW-RacN17 cells compared with its parental RAW264.7 cells. Collectively, these results indicate that RANKL-mediated ROS generation and osteoclastogenesis require Rac1.

Although the production of ROSs by phagocytes such as macrophages and neutrophils has been mainly studied in the context of bacterial killing,^8,24  our data imply that RANKL, which stimulates osteoclast differentiation, induced the production of ROSs for physiologic responses through RANKL-TRAF6 axis that is known to play a central role in osteoclastogenesis.^6,21,22  We also found that the small GTPase Rac1 and Nox1, functioning downstream of TRAF6, are essential for RANKL-mediated ROS production and osteoclastogenesis. Moreover, our current findings suggest that RANKL-induced ROSs play a significant role in the activation of MAP kinases. Thus, these results suggest, besides their destructive properties in the microbicidal role, that ROSs produced in macrophages and monocytes through RANKL-TRAF6-Rac1-NADPH oxidase–dependent pathways may be involved in mediation of osteoclast differentiation.

We have shown that RANKL activates Rac1 in osteoclast precursors. Importantly, expression of a Rac1 dominant-negative mutant (Rac1N17) blocked ROS production and osteoclast differentiation induced by RANKL. It has been reported that the small GTPase Rac1 is a cytosolic component of NADPH oxidase complex and is responsible for the activation of an NADPH oxidase.²⁶  In addition, Rac1 binds to Nox1 to stimulate its NADPH oxidase activity in a growth factor-dependent manner.²⁷  Our findings are consistent with those reports in that RANKL signaling is linked to Rac1-Nox1 complex to generate ROSs. It is possible that similar mechanisms may control distinct biologic outcomes by growth and differentiation factors, despite their use in overlapping signaling cascades.

Recently, novel gp91phox (Nox2) homologues, termed Nox1, Nox3, Nox4, and Nox5, were identified in various nonphagocytic cells.²⁴  Mature osteoclasts express an alternative oxidase, Nox4, also responsible for ROS generation.²⁸  Our data from RT-PCR demonstrated that Nox1 is expressed in BMM cells, suggesting that it is Nox1 rather than Nox2 or Nox4 that is important for RANKL-induced ROS production in BMM cells. Indeed, Nox1 siRNAs resulted in an inhibition of both RANKL-induced ROS formation and osteoclastogenesis. There was a clear difference in the level of inhibition of RANKL-mediated ERK activation by DPI or by Nox1 siRNA. Specifically, DPI blockage of RANKL-induced ROSs resulted in a decrease in ERK activity, whereas Nox1 siRNA had no effect. This likely results from the specificity of Nox1 inhibition, which was revealed by Nox1 siRNA application. On the contrary, the most widely used inhibitor of NADPH oxidase, DPI, is a broad one, including the neuronal (nNOS) and endothelial (eNOS) nitric oxide isoforms²⁹  and other flavo-containing proteins.³⁰  It is worth noting that IL-1β stimulates osteoclast differentiation in the presence of RANKL in a synergistic fashion.³¹  In addition, it has been shown that IL-1β treatment in monocytes led to a production of ROSs which requires an NADPH oxidase.³²  Unlike RANKL, IL-1β–mediated ROS production was not affected when Nox1-silenced BMM cells were stimulated by IL-1β, suggesting that distinct Nox isozymes regulate ROS production.

The mechanism by which ROSs activate MAP kinases is not completely understood. Note that the activation of MAP kinases by a number of stimuli, including growth factors and cytokines, appears to require ROSs,^33,34  suggesting that the activation of MAP kinases by ROSs might be a rather common, if not general, response to different stimuli. Recently, it has been reported that ROS oxidize cysteine residues in proteins, including protein tyrosine phosphatases (PTPs; eg, PTEN [phosphatase tensin homologue deleted on chromosome 10]), thus inactivating the proteins.³⁵  These results suggest that inhibition of tyrosine phosphatase activity by ROSs may account for another mechanism in triggering downstream signaling events. Therefore, it is possible that ROSs may regulate the activation of MAP kinases by inactivating a tyrosine phosphatase activity. In support of this possibility, we have preliminary results that RANKL stimulation induces diverse phosphotyrosyl proteins in RAW264.7 cells (data not shown), although there is no evidence that RANK has a functional tyrosine kinase domain. Pretreatment with NAC or DPI reduced RANKL-induced tyrosine phosphorylation, suggesting that the relationships between RANKL-induced ROSs and inhibition of PTP(s). Otherwise, we cannot rigorously rule out the possibility that serine-threonine phosphatases are regulated by ROSs, thereby affecting MAP kinases activation. Further study will be required to elucidate how RANKL-induced ROSs regulate the activation of MAP kinases directly or indirectly.

Recently, consistent with our data it has been reported that administration of NAC or ascorbate antioxidants decreased osteoclast number on bone surface, thereby abolishing ovariectomy-induced bone loss.³⁶  In this model, ROSs have been implicated as a pathologic agent as in inflammation and aging. However, our results show that ROSs may act as a physiologic second messenger in RANKL signaling essential for osteoclastogenesis. We want to emphasize that a mechanism similar to the role of ROSs in osteoclast precursors is also operational during plant cell growth³⁷  and invertebrate vulval development,³⁸  indicating that ROSs can control cell development or differentiation.

In conclusion, our results indicate that ROSs produced by RANKL in BMM cells stimulate osteoclast differentiation, whereas decreasing the level of ROSs reverses RANKL responses. Therefore, application of ROS scavengers or inhibitors of RANKL-induced ROSs generating pathways may be beneficial strategies for an alternative therapy of bone diseases.

We thank J. Kim for critical reading of the manuscript, J. Inoue for providing the pMX-TRAF6 vector, and T. Kitamura for providing the Plat-E cell line.