The JNK-dependent CaMK pathway restrains the reversion of committed cells during osteoclast differentiation

Osteoclasts have the specialized function of dissolving bone matrix and they play key roles in the pathology of many bone-erosive disorders, such as osteoporosis, rheumatoid arthritis and Paget disease (Boyle et al., 2003; Tanaka et al., 2005; Teitelbaum, 2007; Zaidi et al., 2003). Osteoclasts are derived from hematopoietic stem cells (HSCs), specifically from the monocyte-macrophage lineage of myeloid cells. Two cytokines play fundamental roles in osteoclast generation from the progenitors: macrophage colony stimulating factor (M-CSF) and receptor activator of NF-κB ligand (RANKL) (Lacey et al., 1998; Yasuda et al., 1998). M-CSF induces expression of RANK, the receptor for RANKL, in early-stage precursors of osteoclast lineages in mouse bone-marrow cells (Arai et al., 1999). Subsequent ligation of RANK by RANKL initiates an osteoclastic differentiation program. Under normal conditions, the RANKL-RANK axis appears to be essential for osteoclastogenesis because both RANKL^–/– and RANK^–/– mice develop no osteoclasts (Dougall et al., 1999; Kong et al., 1999). The major signaling events triggered upon RANK ligation include recruitment of tumor necrosis factor receptor-associated factor (TRAF) adaptor proteins, activation of protein kinases (Src, Akt, TAK, IKKs and MAPKs), and elevation in the expression and/or activity of transcription factors NF-κB, Fos, AP-1 and NFATc1 (Boyle et al., 2003; David et al., 2002; Huang et al., 2006; Lee et al., 2002; Lee and Kim, 2003; Matsuo et al., 2000; Kobayashi et al., 2001; Takayanagi et al., 2002a; Takayanagi et al., 2002b; Wong et al., 1999). The activity of these transcription factors ultimately leads to the induction of genes encoding proteins related to osteoclast function, such as tartrate-resistant acid phosphatase (TRAP), cathepsin-K, MMP9 and calcitonin receptor (Ikeda et al., 2004; Kim et al., 2005; Matrumoto et al., 2004; Sharma et al., 2007).

Osteoclasts share the same progenitor cells with macrophages and dendritic cells (Akagawa et al., 1996; Udagawa et al., 1990). The molecular understanding of the signaling mechanisms responsible for the lineage diversification between osteoclasts, macrophages and dendritic cells is currently very poor. With mouse bone-marrow precursor cells, osteoclastic and dendritic cell differentiation has been shown to be reciprocally regulated by M-CSF and granulocyte monocyte colony stimulating factor (GM-CSF); osteoclastic differentiation driven by M-CSF plus RANKL was suppressed in the presence of GM-CSF, whereas dendritic cell differentiation driven by GM-CSF was inhibited by M-CSF (Miyamoto and Akashi, 2005). A prominent intracellular event that correlated with the GM-CSF-dependent inhibition of osteoclast differentiation was reduction in the Fos level (Miyamoto and Akashi, 2005), suggesting a crucial role of Fos in the fate determination of this common lineage of cells to osteoclasts versus dendritic cells. Other signaling events have also been described for lineage determination from HSCs. The MEK-ERK signaling pathway was recently reported to play a crucial role in granulocyte-macrophage lineage commitment from HSCs (Hsu et al., 2007). Also, the ERK pathway was shown to be specifically required at an early stage in retinoic-acid-dependent adipocyte-lineage commitment of embryonic stem cells (Bost et al., 2002). By contrast, inhibition of ERK was reported to stimulate adipogenesis from adult human mesenchymal stem cells (Jaiswal et al., 2000). Therefore, it appears that the same signaling pathway can have different roles in lineage determination depending on the differentiation signals and progenitor cell types.

Despite a substantial amount of studies on the factors that stimulate osteoclast differentiation and on the intracellular signaling events triggered by the stimulators, when and how osteoclastic-lineage commitment from the common precursor cells is achieved are not known. A study suggested that macrophagic differentiation is the default pathway in the fate determination of the common precursor cells and that RANKL actively suppresses the default pathway for osteoclastogenesis (Arai et al., 1990). However, molecular mechanisms to explain the observation were not provided in the study. Dynamic and temporal regulation of transcription factors is pivotal to cell-fate determination during hematopoiesis (Iwasaki et al., 2006; Evans et al., 2003; Terskikh et al., 2003; Wang et al., 2006). In support of the importance of transcription factors in the commitment and maintenance of specific lineages, alterations in the level and timing of expression of transcription factors were shown to change the lineage of already committed cells (Akashi, 2005). Genetic studies have suggested that the transcription factors Fos and NF-κB are crucial to osteoclastogenesis by demonstrating that mice deficient in these transcription factors display an osteopetrotic phenotype (Franzoso et al., 1997; Grigoriadis et al., 1994). More recently, expression of Fos and NFATc1 were shown to be induced by RANKL during osteoclastogenesis (Matsuo et al., 2000; Takayanaki et al., 2002a).

Defining the cells that are committed to the osteoclastic lineage is still obscure because of a lack of unique marker(s) to distinguish between uncommitted and committed cells. In osteoclastogenic culture, two distinctive steps occur: conversion of TRAP-negative (TRAP^–) precursor cells to TRAP⁺ mononuclear cells (pre-osteoclasts; pOCs) and fusion of TRAP⁺ mononuclear pOCs to generate multinuclear TRAP⁺ mature osteoclasts. Although the precise point of osteoclastic-lineage commitment is not known, TRAP⁺ pOCs are likely to be at or beyond the commitment point. Thus, TRAP positivity is so far the best distinguishable feature between cells that have undergone osteoclastic commitment and those at the precursor stage. The induction of Fos and NFATc1 by RANKL occurs before the fusion point of TRAP⁺ mononuclear cells (Takayanaki et al., 2002a), suggesting the possible involvement of these transcription factors in osteoclastic-lineage commitment. NFATc1 binds to the TRAP promoter (Ikeda et al., 2004) and activates – probably as a complex with other transcription factors, coactivators and chromatin remodeling proteins (Sharma et al., 2007) – the transcription of the TRAP gene.

In this study, we show that the JNK pathway was specifically required for the maintenance of the osteoclastic-commitment status. The TRAP⁺ phenotype established by RANKL at the pOC stage was lost with concomitant gain of macrophagic features upon the blockade of JNK activity, even in the continuous presence of RANKL. The reversion of pOC by JNK inhibition was associated with depression of calcium/calmodulin-dependent protein kinase (CaMK) expression and subsequent decrease in NFATc1 level. This is the first report to demonstrate that osteoclastic-differentiation steps could be reversed at a certain stage by blockade of a specific signaling pathway.

Bone-marrow-derived macrophage (BMM) cells become morphologically rounder and cytochemically TRAP⁺ when cultured in the presence of the osteoclast differentiation factor RANKL. Under our experimental conditions, more than 80% of cells cultured for 2 days with RANKL were TRAP⁺ and mononuclear, with the presence of a few small TRAP⁺ multinuclear cells (MNCs). These TRAP⁺ mononuclear cells (pOCs) are considered committed to osteoclastic lineage. pOCs progressively undergo fusion to generate large multinucleated TRAP⁺ mature osteoclasts upon stimulation with RANKL or inflammatory cytokines such as IL1 and TNFα. In an effort to understand the mechanism for osteoclast-lineage commitment, we set up an experiment in which pOCs were treated with the NF-κB inhibitor SN50, the JNK inhibitor SP600125, the MEK inhibitor PD98059 or the p38 inhibitor SB203580, in the presence of one of the cytokines (RANKL, IL1α or TNFα) or combinations of the cytokines (Fig. 1A). M-CSF was included during the whole culture as a survival factor. When TRAP⁺ MNC generation was assessed after 48 hours of incubation with the inhibitors and cytokines, little effect of PD98059 was observed, whereas SN50 and SB203580 showed a significant suppression (Fig. 1B,C). SP600125 almost completely blocked TRAP⁺ MNC formation in all combinations of the cytokines (Fig. 1B,C). Strikingly, SP600125 also decreased the percentage of TRAP⁺ mononuclear cells (Fig. 1D,E) without reducing the number of total cells and the viability of cells, which was determined by measuring the metabolic activity of cells using CCK-8 (Fig. 1F). This was clearly distinct from the effects of SN50 and SB203580, which reduced MNC formation with no effects on the TRAP⁺ nature of pOCs (Fig. 1E). The reducing effect of SP600125 on the percentage of TRAP⁺ cells was dose dependent (Fig. 1G). Because the blockade in JNK activity caused the conversion of TRAP⁺ cells to TRAP^–, we examined the effect of SP600125 on the expression levels of TRAP and other genes related to osteoclastogenesis. The mRNA expression of TRAP, calcitonin receptor (CTR) and MMP9 were significantly reduced by SP600125 treatment at the pOC stage (Fig. 1H).

We next investigated whether the reversion of TRAP⁺ cells to TRAP^– characteristics by JNK inhibition is restricted up to a certain stage. The JNK inhibitor SP600125 was added at different days of the osteoclastogenic culture of BMMs. Addition of SP600125 at day 2 of the culture elicited reversion of TRAP⁺ cells to TRAP^–, whereas addition at day 3 did not reduce the percentage of TRAP⁺ cells (Fig. 2A). Similar results were obtained with the RAW264.7 cell line, which undergoes osteoclastic differentiation upon culturing in the presence of RANKL alone. The effect of SP600125 was also observed in these cells when the inhibitor was added at the pOC stage (day 2), but not at the MNC-forming stage (day 3) (Fig. 2B). Next, we blocked JNK activity by overexpressing a dominant-negative form of JNK (JNK-DN) in RAW264.7 cells to further examine the role of JNK in maintaining the TRAP⁺ pOC characteristics. Transfection of JNK-DN at day 2 caused a prominent reduction in TRAP⁺ cell percentage, whereas wild-type JNK (JNK-WT) transfection did not have a significant effect (Fig. 2C). JNK-DN transfection at day 3 had no effect on the percentage of TRAP⁺ cells (Fig. 2C). The data shown in Figs 1 and 2 suggest that JNK activity is specifically required to keep the TRAP⁺ mononuclear cells in the osteoclast-lineage-committed status.

Because the TRAP⁺ mononuclear pOCs derived from BMMs returned to TRAP^– status following JNK blockade, we next inquired whether this transition is also accompanied by an increase in macrophage characteristics. To determine this probability, the macrophage cell-surface markers CD11b, CD68, CD14 and F4/80 were assessed by flow cytometry. As expected, BMMs expressed these markers (Fig. 3). CD11b-positive cell counts remained high in both JNK-inhibitor (SP600215)-treated and untreated cells (Fig. 3A). The number of CD68-, F4/80- and CD14-positive cells decreased as BMMs became pOCs (Fig. 3B-D). Upon SP600125 treatment of pOCs, the number of cells positive for either CD68, F4/80 or CD14 reached a level that was comparable to or higher than that found in BMMs (Fig. 3B-D). The mean fluorescence intensity (MFI) of the stained cells also showed a similar pattern of changes (Fig. 3E). Thus, SP600215 treatment of pOCs led to an increase in cell population that express the macrophage markers, suggesting that the cells have reversed to macrophages from pOCs.

To further characterize the nature of the cells that returned from pOCs to a TRAP^– status in response to JNK blockade, we tested the phagocytic activity of the cells. Cells were incubated with FITC-labeled zymosan microspheres, vigorously washed to remove free microspheres and stained with propidium iodide to visualize each cell in fluorescence microscopic fields. The phagocytosed microspheres showed as yellow in the merged images (Fig. 4A). It has been reported that phagocytic activity decreases when BMMs differentiate into osteoclasts (Mochizuki et al., 2006). In agreement with previous reports, BMMs showed a higher percentage of phagocytic cells and higher amounts of phagocytosed microspheres per cell than pOCs in our study (Fig. 4A). Analysis after 24 hours of incubation of pOCs with RANKL in the absence of SP600125 found the number of phagocytic cells and the extent of phagocytic activity per cell to be further decreased (Fig. 4A). This decrease in phagocytic-cell number and activity level was blocked by the presence of SP600125 during the incubation period (Fig. 4A). The phagocytic activity was also evaluated by flow cytometric analysis. The percentage of phagocytic pOCs decreased to 49.16% from the 80.97% of BMMs (Fig. 4B,C). The culture of pOCs for another day in the absence of SP600125 further decreased the number of cells displaying phagocytic activity to 19.24%, whereas incubation of pOCs with SP600125 increased the number of phagocytic cells to 67.64% (Fig. 4B,C).

Macrophages have the potential to differentiate to dendritic cells in the presence of GM-CSF and interleukin-4 (IL4) (Alnaeeli et al., 2006). We examined the response of pOCs treated with SP600125 in comparison with BMMs and vehicle-untreated pOCs when the cells were cultured in the presence of GM-CSF plus IL4. SP600125-treated cells behaved like BMMs, showing generation of CD11c⁺ F4/80^– dendritic cells (Fig. 4D,E). Collectively, the results shown in Figs 3 and 4 suggest that the cells converted to TRAP^– from pOCs upon SP600125 treatment became macrophage-like.

We next investigated whether the conversion of committed TRAP⁺ pOCs to TRAP^– cells by JNK blockade could also be induced in human cells. Human peripheral blood monocytes (PBMCs) were cultured in osteoclastogenic medium until the percentage of TRAP⁺ cells reached ∼80% (Fig. 5A). The addition of SP600125 to the TRAP⁺ pOCs for 24 hours reduced the TRAP⁺ cell percentage to ∼30% (Fig. 5B). As was the case with BMM-derived pOCs (Fig. 4), the number of PBMC-derived pOCs with phagocytic activity decreased (Fig. 5C,D). SP600125 treatment of human pOCs for 24 hours caused an increase in the phagocytic cell population from 24.31% to 35.29% (Fig. 5C,D). These data demonstrate that the phenotypic reversion of pOCs by JNK inhibition is conserved between mouse and human osteoclast-differentiation programs.

Several transcription factors are crucial for osteoclast differentiation. Among them, NFATc1 has been suggested to play a master-switch role in osteoclast differentiation (Takayanagi et al., 2002a). A previous report showing that cooperative activity between Jun-Fos complexes and NFATc2 is necessary for NFATc1 expression (Ikeda et al., 2004) suggests possible JNK involvement in NFATc1 regulation. We found that SP600125 treatment blocked the RANKL-induced transcriptional activity of NFATc1 in pOCs (Fig. 6A). In line with the reduced activity, both the mRNA and protein levels of NFATc1 were downregulated by SP600125 treatment of pOCs (Fig. 6B,C). These data suggest that NFATc1 is a key downstream player in the phenotypic reversion by JNK blockade. To support this notion, we knocked down the NFATc1 expression level in pOCs by using retrovirus for NFATc1 siRNA and assessed the percentage of TRAP⁺ cells. When NFATc1 expression was reduced by siRNA (Fig. 6D), TRAP⁺ cell percentage was substantially decreased (Fig. 6E,F). Furthermore, the retroviral infection of active NFATc1 prevented the decrease in TRAP⁺ cells by SP600125 (Fig. 6G,H). Therefore, the expression of NFATc1 controlled by JNK activity at the pOC stage is required for maintaining the TRAP⁺ phenotype for further differentiation into mature osteoclasts.

We next analyzed the gene-expression-profile changes in pOCs caused by JNK inhibition by using Affymetrix human GeneChip. pOCs were generated from three different human PBMC preparations in osteoclastogenic medium and further cultured in the presence (SP+ cells) or absence (SP– cells) of SP600125 for 24 hours. The gene-expression levels in the SP+ and SP– cells were compared together and with those in PBMCs (SP0 cells). The genes for which the expression level increased or decreased in SP– cells compared with that in SP0 cells with the change being attenuated by SP treatment by more than 1.5-fold in all three different samples were collected (see Table S1 and S2 in the supplementary material). Of special interest, CamKIV (CamK4) mRNA decreased by 1.73-fold in response to SP600125 treatment in pOCs (Fig. 7A). We then sought to delineate the relationship between CaMK and the JNK-blockade-induced reversion of pOCs. The expression of CaMK-family kinases was confirmed by real-time PCR in BMMs, pOCs and pOCs further cultured in the presence or absence of SP600125. The mRNA expression of CamKIIα (CamK2a), CamKIIδ (CamK2d) and CamKIV increased in pOCs compared with the level in BMMs (Fig. 7B). Treatment of pOCs with SP600125 decreased CamKIIδ and CamKIV mRNA, whereas CamKIIα was not affected (Fig. 7B). Thus, we hypothesized that CaMKIIδ and CaMKIV are responsible for JNK-mediated maintenance of osteoclastic commitment. To gain evidence for this notion, we knocked-down CaMKII or CaMKIV expression in pOCs using siRNA. The respective CamKIIδ or CamKIV siRNAs were effective and specific (Fig. 7C). After CaMKII or CaMKIV knockdown, the percentage of TRAP⁺ cells was substantially reduced (Fig. 7D). In addition, a significantly higher percentage of pOCs that were transduced with CamKIIδ and CamKIV siRNA showed phagocytic activity compared with the cells transduced with control siRNA (Fig. 7E).

Because NFATc1 also mediated the JNK-blockade-dependent reversion of pOCs (Fig. 6), we inquired whether CaMK activity would be linked to NFATc1 regulation. First, the effect on NFATc1 expression of CaMK downregulation by siRNA was assessed. Introduction of CamKIIδ and CamKIV siRNA in pOCs resulted in a decrease in the NFATc1 level (Fig. 8A). Next, the effect of CaMK-activity inhibition with the pharmacological reagent KN93 was examined. KN93 treatment also reduced NFATc1 expression (Fig. 8B), suggesting that the catalytic activity of CaMK is required for NFATc1 expression in pOCs. Consistently, KN93 treatment of pOCs decreased the percentage of TRAP⁺ cells (Fig. 8C) and increased phagocytic activity (Fig. 8D). These data suggest that JNK modulates the expression of CaMK and subsequently that of NFATc1 to maintain the commitment status for successful completion of osteoclastic differentiation.

Both lineage commitment from HSCs and subsequent differentiation in the bone microenvironment are essential steps for osteoclast generation. However, the osteoclastic-lineage commitment has been ill-defined because of lack of markers specific for cells committed to osteoclastogenesis. In this regard, it was reported that mouse bone-marrow monocytic cells rendered to become RANK⁺ by treatment with M-CSF, which were designated as late-stage precursors, could still differentiate into macrophage cells (Arai et al., 1999). In that study, it was proposed that the stage of osteoclastic commitment is late in the differentiation process (Arai et al., 1999). In another study, mouse BMMs exposed to RANKL together with M-CSF and TGFβ for 24 hours, under culture conditions in which full osteoclastic differentiation is completed by 72 hours, were considered as cells committed to osteoclastogenesis (Mochizuki et al., 2006). These `committed' cells exhibited moderate levels of RANK and NFATc1 and a low level of TRAP, and could not differentiate to dendritic cells in the presence of GM-CSF (Mochizuki et al., 2006). The TRAP<sup>+</sup> pOCs used in our study are probably at the state comparable to the `committed' cells in the latter study (Mochizuki et al., 2006) or at a point a little further progressed in the osteoclastic differentiation process. Regardless of the precise differentiation status of pOCs, it is clear that JNK inhibition caused loss of the TRAP⁺ phenotype of pOCs even in the presence of RANKL in our study. However, the cells already fused did not lose their TRAP⁺ characteristics. Therefore, JNK activity is required until the fusion stage during osteoclastogenesis to maintain the status of osteoclastic commitment. Our study is distinct from previous genetic and pharmacological studies (David et al., 2002; Ikeda et al., 2004) that investigated the effects of a JNK blockade that was present from the beginning of the differentiation, which can not distinguish the role of JNK for the initiation of commitment itself from the role of JNK after the commitment has been established.

The cells converted to TRAP^– from TRAP⁺ pOCs by JNK inhibition appeared macrophage-like. Several lines of data support the macrophage-likeness of the converted cells. First, FACS analyses showed that CD68, CD14 and F4/80 expression became lower in pOCs and further decreased in untreated pOCs compared with BMMs, whereas it was comparable between the SP600125-treated pOCs and BMMs (Fig. 3). Second, the phagocytic activity of cells gradually decreased as the BMM or PBMC osteoclast precursor cells differentiated to osteoclasts (Figs 4, 5). Treatment with the JNK inhibitor SP600125 at the pOC stage reversed the direction of phagocytic activity change, i.e. untreated cells showed lower phagocytic activity than pOCs, whereas SP600125-treated cells had higher activity than pOCs (Figs 4, 5). Furthermore, pOCs treated with SP600125 showed a behavior more like BMMs in conditions in which dendritic differentiation was driven by GM-CSF and IL4 (Fig. 4D,E). In addition, removal of SP600125 at the pOC stage after 24 hours of treatment allowed normal osteoclast formation (data not shown), suggesting that treated cells retained osteoclastogenic potential. Thus, JNK blockade in pOCs seems to return the committed cells to the precursor-like state, reversing the differentiation program.

In our study, the molecular mechanism underlying the JNK-inhibition-induced reversal of osteoclastic commitment was attributable to CaMK and NFATc1. JNK inhibition reduced the expression of both NFATc1 and CaMK (Figs 6, 7). NFATc1 is a key transcription factor for the expression of TRAP and of other osteoclastogenesis-associated genes (Ikeda et al., 2004; Kim et al., 2005; Matsumoto et al., 2004; Sharma et al., 2007). The upregulation of NFATc1 by RANKL requires an auto-amplification process in which active NFATc1 binds to its own promoter and enhances transcription (Asagiri et al., 2005). JNK inhibition also decreased specifically the expression level of CaMKIIδ and CaMKIV in pOCs (Fig. 7A,B). This downregulation of CaMK by JNK inhibition appeared to involve AP-1, the complex of Fos and Jun family members, because SP600125 reduced AP-1 activity in promoter reporter assays (data not shown). It is likely that Jun phosphorylation that otherwise enhances the transcription activity of AP-1 was blocked by SP600125. In support of this notion, promoter-sequence analyses revealed two and one AP-1 binding sites in CaMKIIδ and CaMKIV, respectively (data not shown). Despite several documentations describing the involvement of CaMK in the regulation of JNK activity, there has been no report showing CaMK regulation by JNK in any cell system. To our knowledge, our study is the first to show JNK-mediated regulation of CaMK. Interestingly, although the overexpression of NFATc1 in SP600125-treated pOCs resulted in the restoration of the TRAP⁺ population, it could not facilitate the generation of mature, multinuclear osteoclasts (Fig. 6G,H). Thus, it is likely that the JNK-CaMK-NFATc1 pathway described in the present study is not sufficient to promote the differentiation of osteoclasts from pOCs. These results not only suggest that NFATc1 downstream of JNK is important for the maintenance of osteoclastic commitment, but also indicate that other JNK-dependent signaling pathways are required to fully support osteoclast differentiation.

In our study, both gene knockdown of CamKIIδ and CamKIV, and treatment of cells with the CaMK inhibitor KN93 elicited the reversion of pOCs to TRAP^– cells while elevating the phagocytic activity (Figs 7, 8). During the progression of our study, Sato et al. showed that KN93 blocked Fos induction by RANKL in BMMs and suggested that CaMKIV regulates Fos transcription through CREB phosphorylation (Sato et al., 2006). In pOCs, gene knockdown of either CamKIIδ or CamKIV led to a reduction in Fos level (data not shown). Fos, in turn, is one of the key factors required for NFATc1 induction by RANKL (Matsuo et al., 2004; Takayanagi et al., 2002a). Consistently, the addition of both CaMK siRNA and KN93 to pOCs decreased NFATc1 levels (Fig. 8A,B). One interesting observation was that the addition of Fos siRNA reduced CaMKIIδ and CaMKIV levels (data not shown). Because the promoters of those CaMK proteins contain AP-1-binding sites (see above), it is plausible that Fos upregulation by CaMK feeds back to elevation of CaMKIIδ and CaMKIV expression. The NFATc1 auto-amplification and CaMK-Fos positive feedback might reinforce the profound and sustained expression of NFATc1 in the late phase of osteoclastogenesis. In our present study, SP600125 decreased NFATc1 levels in pOCs even in the continuous presence of RANKL, and introduction of NFATc1 siRNA converted TRAP⁺ cells to TRAP^– (Fig. 6). Therefore, NFATc1 is the ultimate point to which the JNK signaling reaches through CaMK for the maintenance of osteoclastic-commitment status (Fig. 9). For activation of both CaMK and NFATc1, Ca²⁺ is required. pOCs manifest an oscillation response in the intracellular Ca²⁺ level (Takayanagi et al., 2002a). Therefore, in addition to the expression induction of CaMK and NFATc1, RANKL signaling might stimulate their activity by regulating intracellular Ca²⁺ in pOCs to sustain the osteoclastic-commitment state.

In summary, we showed here for the first time that JNK is specifically involved in maintaining the commitment state during osteoclastogenesis by RANKL. We also provided evidence that this function of JNK is mediated by regulation of CaMK, which is crucial for maintaining high NFATc1 levels in committed cells. The JNK-CaMK pathway might be an efficient therapeutic target for blocking cells already committed for osteoclast formation.

Recombinant soluble human RANKL (310-01), murine RANKL (315-11), human M-CSF (300-25), murine M-CSF (315-02) and IL1α (211-11A) were purchased from Peprotech EC (London, UK). Recombinant murine TNFα (416-MT-50) was obtained from R&D Systems (Minneapolis, MN). Polyclonal antibodies (Abs) against NFATc1 (sc7294) and CD14 (sc-9150) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-actin monoclonal Ab (mAb) was purchased from Sigma-Aldrich (St Louis, MO). Rat FITC-conjugated anti-mouse CD11b mAb (557396), hamster PE-Cy7-labeled anti-CD11c mAb (558079), rat IgG2b-PE and mouse IgM-FITC were obtained from BD Pharmingen (San Jose, CA). Rat FITC-labeled anti-mouse F4/80 mAb (Cl:A3-1) was purchased from Serotec (Raleigh, NC). FITC-zymosan (Z2841) was purchased from Molecular Probe (Carlsbad, CA). SN50 (NF-κB inhibitor) (481480), SP600125 (JNK inhibitor) (420119), PD98059 (MEK inhibitor) (513000) and SB203580 (p38 inhibitor) (559389) were obtained from Calbiochem (San Diego, CA). Acid Phosphatase kit (387A) was obtained from Sigma-Aldrich. Cell-counting kit-8 (CCK-8) was purchased from Dojindo Laboratories (Kumamoto, Japan).

For osteoclastogenesis from mouse bone-marrow-derived precursor cells, 6- to 8-week-old ICR mice were used (Charles River Laboratories, Wilmington, MA). Animal experimental protocols were approved by the committee on the care and use of animals in research at Seoul National University. Mouse bone-marrow space of the tibiae and femora was flushed with α-minimal essential medium (MEM) containing 100 U/ml penicillin and 100 μg/ml streptomycin. After removing erythrocytes by lysing in ACK buffer (0.15 M NH₄Cl, 1 mM KHCO₃ and 0.1 mM EDTA, pH 7.2), the bone-marrow cells were plated on 100-mm culture dishes and cultured in α-MEM supplemented with 10% fetal bovine serum (FBS) for 16-24 hours in 5% CO₂ at 37°C. Non-adherent cells were collected, plated on 100-mm bacterial dishes and cultured for 3 days in the presence of 30 ng/ml M-CSF. The adherent cells were considered to be BMMs and were used as osteoclast precursor cells. BMMs were seeded on 48-well plates at 5×10⁴ cells/well and cultured with 100 ng/ml RANKL plus 30 ng/ml M-CSF. Cells were stained for TRAP using the Acid Phosphatase kit as per the manufacturer's instruction (Sigma-Aldrich). The percentage of TRAP⁺ cells among the total cells and the number of MNCs containing more than three nuclei were scored. After 2 days of BMM culture in the osteoclastogenic medium, more than 80% of total cells were TRAP⁺ and most of the cells were pOCs, with a few small MNCs. The maximum level of formation of fully mature osteoclasts was achieved after 4 days.

Cell viability was assessed using CCK-8 according to the manufacturer's protocol (Dojindo Laboratories). Cells were incubated with the CCK-8 reagent (100 nM) for 45 minutes and optical density was determined at 450 nm.

For osteoclastogenesis from RAW264.7 cells, cells were seeded in 48-well plates at 1×10⁴ cells/well in α-MEM containing 10% FBS, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. Cells were cultured in the presence of 100 ng/ml RANKL without M-CSF addition. RAW264.7-derived pOCs were obtained after 2 days and mature osteoclasts were observed after ∼3.5-4 days.

In some experiments, RAW264.7 cells were cultured with 100 ng/ml RANKL for 2 or 3 days. Cells were transfected for 3 hours with the wild-type or dominant-negative JNK plasmid DNA using Lipofectamine 2000 reagent (Invitrogen) in α-MEM/10% FBS. At 16 hours after transfection, RANKL was added to 100 ng/ml and cells were further incubated for 24 hours before TRAP staining.

Experiments with human blood samples were approved by the Institutional Review Board for Medical Research Ethics. Peripheral blood of healthy donors was separated by centrifugation on Ficoll-Histopaque (25-072-CV, Cellgro), washed with PBS and then resuspended in α-MEM supplemented with 10% FBS. For osteoclast formation, PBMCs were cultured in α-MEM/10% FBS in the presence of 100 ng/ml M-CSF and 100 ng/ml RANKL as previously described (Chang et al., 2007). After 3 days of culture, pOCs were formed. Mature osteoclasts were generated after 7 days.

Recombinant retroviruses were prepared as previously described (Huang et al., 2000). pOCs derived from BMMs were infected with the prepared retroviruses for ∼12-24 hours.

Cultured cells (2×10⁵/ml) were incubated with FITC-zymosan for 45 minutes. After washing three times with 1% FBS/PBS to remove free FITC-zymosan, cells were fixed, permeabilized and stained with propidium iodide. The fluorescence of cells was detected by confocal microscopy (OLYMPUS-FV300) and FACS analyses (FACSCalibur, BD).

Cells (5×10⁵ cells/ml) were cultured in modified RPMI 1640 medium (BioWhittaker) in the presence of 10 ng/ml IL4 (Peprotech) and 10 ng/ml GM-CSF (Immunex/Amgen) for 3 days based on the method as described previously (Steptoe et al., 2002). To determine the phenotype of differentiated dendritic cells, cells were stained by FITC-labeled anti-F4/80 and PE-Cy7-labeled anti-CD11c, and were subjected to cytometric analyses (Alnaeeli et al., 2006).

Antibodies used for FACS analysis are described above. For surface-marker staining, cells were incubated with mAbs for 20 minutes on ice and washed three times with PBS. Flow-cytometric analysis was performed, using a FACSCalibur flow cytometer. CellQuest software (Becton Dickinson) was used for data acquisition and analysis.

For real-time PCR analysis, total RNA was isolated with the RNeasy total RNA isolation kit (Qiagen, Hilden, Germany) and reverse transcribed using RT Superscript II (Invitrogen, Carlsbad, CA) following the manufacturer's instruction. PCR reactions were performed using ABI Prism 7500 Sequence Detection System with SYBRGreen PCR master mix (Applied Biosystems, Warrington, UK) for 40 cycles of 15-second denaturation at 95°C and 60-second amplification at 60°C. Quantitative PCR primer sequences were as follows: 5′-AGATTTGTGGCTGTGGGCGA-3′ (TRAP forward); 5′-AAGTCAGCGCCCATCGTCTG-3′ (TRAP reverse); 5′-GCAACCGAACCTGGTCCAACTAT-3′ (CTR forward); 5′-AAGCAGCAATCGACAAGGAGTGA-3′ (CTR reverse); 5′-GACGGCACGCCTTGGTGTAG-3′ (MMP9 forward); 5′-GGCCCTCAAAGATGAACGGG-3′ (MMP9 reverse); 5′-TCAGCCTGCATCGCCTATATCC-3′ (CaMKIIα forward); 5′-AAGTGGACGATCTGCCATTTGC-3′ (CaMKIIα reverse); 5′-CTGCCTTTGAACCTGAAGCATTG-3′ (CaMKIIδ forward); 5′-TGAACGTGTGGGTTGAGGATGAT-3′ CaMKIIδ reverse); 5′-GGAGGAGACCTCCAGTATGGTGC-3′ (CaMKIV forward); 5′-CTCCTCAGTCATGGGGTCCATTT-3′ (CaMKIV reverse). Dissociation curves were analyzed with SDS software (Applied Biosystems) and data were presented as relative expression normalized by the expression of HPRT. The RT-PCR analyses were performed as described (Huang et al., 2006). The sequences of primers used are as follows: 5′-GGTGGAAGACGTACTTCCTAGCTG-3′ (NFATc1 forward); 5′-CTTCCAGGCTGGGCAGGT-3′ (NFATc1 reverse); 5′-CTGGCACACCTGGGTATCTT-3′ (CaMKIIδ forward); 5′-ATTCTGGTGACGGAAAATCG-3′ (CaMKIIδ reverse); 5′-AGCTGGTCACAGGAGGAGAA-3′ (CaMKIV forward); 5′-AATTTTGAGGGGTGCATCAG-3′ (CaMKIV reverse); 5′-ACCACAGTCCATGCCATCAC-3′ (GAPDH forward); and 5′-TCCACCACCCTGTTGCTGTA-3′ (GAPDH reverse).

Cells were disrupted in a lysis buffer containing 1% NP-40 and the cell lysates were subjected to western blotting as previously described (Ryu et al., 2006).

The siRNA target sense sequence for NFATc1 was 5′-AGACGTACTTCCTAGCTGCAA-3′ and the loop sequence was 5′-TTGATATCCG-3′. The siRNA-insert DNA was cloned into pSuper-retro vector (Oligoengine, Seattle, WA) using a BamHI and HindIII site. For knockdown of CaMK genes, RNA-duplex oligonucleotides (Invitrogen, Carlsbad, CA) were used. The sequences were 5′-UUCUGGAGAAAGAUACCCAGGUGUG-3′ for CamKII and 5′-UCCACAAUCCUGUCAAACAGUUCUC-3′ for CamKIV.

Human PBMCs and PBMC-derived pOCs treated with (SP+) or without (SP–) the JNK inhibitor SP600125 for 24 hours were used for the microarray analysis. Total RNA was extracted using the RNeasy total RNA isolation kit (Qiagen). RNA (10 μg) from each sample was transcribed to double-stranded cDNA using SuperScript II RT (Invitrogen) with oligo-dT primer containing the T7 RNA polymerase site on the 5′ end. The cDNA serves as the template for an in vitro transcription reaction in the presence of biotin-modified ribonucleotides to produce single-stranded RNA. The biotin-labeled RNA was fragmented and hybridized with the GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix, Santa Clara, CA) as per the manufacturer's instruction. The array chips were scanned with the GeneArray scanner (Affymetrix). Data analysis was performed using Microarray Suite (MAS) 5.0. (Affymetrix).

This work was supported by the Research Program for New Drug Target Discovery grant (M10748000257-07N4800-25710) and the 21C Frontier Functional Proteomics Project grant (M108KM010019-08K1301-01910) from the Korea Science & Engineering Foundation, Ministry of Science and Technology, Korea. E.J.C. was supported in part by the Korea Research Foundation grant funded by MOEHRD (KRF-2007-359-E00001).