Nuclear Factor of Activated T-cells (NFAT) Rescues Osteoclastogenesis in Precursors Lacking c-Fos*

Osteoclasts are specialized macrophages that resorb bone. Mice lacking the AP-1 component c-Fos are osteopetrotic because of a lack of osteoclast differentiation and show an increased number of macrophages. The nature of the critical function of c-Fos in osteoclast differentiation is not known. Microarray analysis revealed that Nfatc1, another key regulator of osteoclastogenesis, was down-regulated in Fos-/- osteoclast precursors. Chromatin immunoprecipitation assay showed that c-Fos bound to the Nfatc1 and Acp5 promoters in osteoclasts. In vitro promoter analyses identified nuclear factor of activated T-cells (NFAT)/AP-1 sites in the osteoclast-specific Acp5 and Calcr promoters. Moreover, in Fos-/- precursors gene transfer of an active form of NFAT restored transcription of osteoclast-specific genes in the presence of receptor activator of the NF-κB ligand (RANKL), rescuing bone resorption. In the absence of RANKL, however, Fos-/- precursors were insensitive to NFAT-induced osteoclastogenesis unlike wild-type precursors. These data indicate that lack of Nfatc1 expression is the cause of the differentiation block in Fos-/- osteoclast precursors and that transcriptional induction of Nfatc1 is a major function of c-Fos in osteoclast differentiation.

Plasmids-Mouse Acp5-luciferase reporter plasmid was constructed in pGL3 vector (Promega) by transferring the promoter regions (Ϫ1453 down to the end of intron 1) from pKB5 (a gift from D. Roodman), and the NFAT site mutation was introduced by using the QuikChange kit (Stratagene). ⌬NFAT was constructed by assembling PCR-amplified fragments encoding amino acids 1-239 of enhanced GFP (Clontech) followed by Ser-Arg (an XbaI site) and amino acids 317-902 of human NFATc4 (20). The 2.5-kb ⌬NFAT fragment was cloned into both the cytomegalovirus-driven expression vector pRK5 and the retroviral vector pMX (21). pBJ5-human NFATc1 expression plasmid (pSH102) was a gift from G. R. Crabtree. The Ϫ797 and Ϫ94 Calcr-P3-pGL3basic constructs have been described (22). Additional Calcr-P3 5Ј-deletion constructs were generated by PCR using different forward Calcr primers with a BglII site and a common reverse Calcr primer with a HindIII site. The NFAT site mutations were incorporated into each deletion by sequential PCR reactions using mutant Calcr primers and vector primers. The c-Fos expression vector pMX-c-Fos-IRES-GFP was constructed by inserting the BamHI-SalI fragment of mouse c-Fos cDNA in pBabec-Fos (10). The sequence of each construct was confirmed. The small interfering RNA vectors were based on RVH1 and LTRH1 (23) (a gift from R. Medzhitov). The oligonucleotides encoding the mouse c-Fos small interfering RNA were: RNAi1, 5Ј-gatcccctgatgttctcgggtttcaattcaagagattgaaacccgagaacatcatttttggaac-3Ј and 5Ј-tcgagttccaaaaatgatgttctcgggtttcaatctcttgaattgaaacccgagaacatcaggg-3Ј; RNAi2, 5Ј-gatcccctccaagcggagacagatcattcaagagatgatctgtctccgcttggatttttggaac-3Ј and 5Ј-tcgagttccaaaaatccaagcggagacagatcatctcttgaatgatctgtctccgcttggaggg-3Ј.
Real-time Reverse Transcriptase-PCR-Calcr transcripts were quantitated on ABI PRISM 7000 (Applied Biosystems) using SYBR Green and were normalized to c-fms and Gapdh transcripts for co-cultures and osteoblast-free cultures, respectively.
Bone Resorption Assay-The surface of bone slices was visualized by backscattered electron imaging using a scanning electron microscope (S-2500CX, Hitachi). The extent of bone resorption was quantified with Metamorph (Universal Imaging).

RESULTS
NFATc1 Is Down-regulated in Fos Ϫ/Ϫ Osteoclast Precursors-We set out to identify genome-wide novel c-Fos target genes in the osteoclast lineage. Three osteoclastogenic cultures derived from wild-type and Fos Ϫ/Ϫ splenocytes and wild-type bone marrow were prepared and co-cultured with stromal ST2-T cells. By day 6, the wild-type cultures produced abundant multinucleated osteoclasts, which were tartrate-resistant acid phosphatase (TRAP)-positive, whereas no such cells were generated in the Fos Ϫ/Ϫ culture (data not shown). The co-cultured cells were harvested in toto, and gene expression was analyzed by microarrays. The genes in which expression was detectable in wild-type cultures but absent or very low in the Fos Ϫ/Ϫ culture are summarized in Table I. The numbers for wild-type splenocytes (Spϩ/ϩ), Fos Ϫ/Ϫ splenocytes (SpϪ/Ϫ), and wild-type bone marrow cells (Bmϩ/ϩ) are GeneChip scores indicating RNA levels. In the Fos Ϫ/Ϫ culture, the expression of Nfatc1 was undetectable, and the expression of known osteoclast marker genes (25) was reduced. These include Acp5 (encoding TRAP), Ctsk (cathepsin K), Car2 (carbonic anhydrase 2), Mmp9 (matrix metalloproteinase 9), and Calcr (calcitonin receptor). However, Marco, a macrophage receptor, was not reduced in the Fos Ϫ/Ϫ culture (Table I).
To confirm the differential expression at the protein level, Western blot analysis was performed using total protein extracts prepared on day 6 from the osteoclastogenic co-cultures. Consistent with the RNA data, NFATc1 was not detectable in the Fos Ϫ/Ϫ culture (Fig. 1A). The size variation of NFATc1 in the wild-type bone marrow culture may be caused by either degradation products or by different isoforms of NFATc1 (26,27). Next we examined the subcellular localization of NFATc1 in mature osteoclasts freshly isolated from the femurs of wildtype mice. Immunofluorescence microscopy showed nuclear staining of NFATc1 but not NFATc2 in mature multinucleated osteoclasts generated in vivo (Fig. 1B). Western blot analysis of nuclear extracts from the macrophage-osteoclast precursor RAW264.7 cells demonstrated that nuclear NFATc1 was detectable only after RANKL stimulation (Fig. 1C). These data suggest that RANKL stimulates NFATc1 synthesis via c-Fos.
c-Fos Binds to the Nfatc1 Promoter-Putative c-Fos binding sites have been mapped in the promoter region of Nfatc1 ( Fig.  2A) (27,28). To examine whether the Nfatc1 promoter could be directly regulated by c-Fos in osteoclast precursors, we performed a chromatin immunoprecipitation assay using primary wild-type bone marrow cells and Fos Ϫ/Ϫ splenocytes treated with RANKL. The Nfatc1-P1 promoter fragment containing the distal block of homology between human and mouse sequences (27) was specifically precipitated with an anti-c-Fos antibody in samples prepared from wild-type cells treated with RANKL (Fig. 2B). The proximal block of homology could not be analyzed because of difficulty in PCR amplification. We also tested whether the Acp5 promoter ( Fig. 3A) was precipitated by antic-Fos antibody (Fig. 2B). In wild-type cells, c-Fos is present on the Acp5 promoter in the absence of RANKL, and c-Fos occupancy of the Acp5 promoter increases after RANKL treatment. These results suggest that c-Fos binds to the NFATc1 and Acp5 promoters during osteoclastogenesis. Acp5 and Calcr Promoters Contain Functional NFAT/AP-1 Sites-It is known that AP-1 composed of Fos/Jun dimers and NFAT transcription factors can cooperatively bind to promoter regions of various genes including the IL-2 gene (29,30). To study the molecular mechanisms by which NFAT is involved in the regulation of osteoclast-specific gene expression we examined promoter sequences of Acp5 and Calcr, two potential c-Fos target genes (Table I). First we searched for such composite binding sites for NFAT and AP-1 (NFAT/AP-1 sites) in mouse, human, and pig Acp5 promoter sequences. Two short stretches were highly conserved around the multiple transcription start sites, which we termed Acp5-160 and Acp5-120, respectively (Acp5 elements located at Ϫ160 and Ϫ120 relative to the 3Ј end of exon 1). Sequences of the conserved Acp5-120 were found to be similar to the prototypical 15-bp NFAT/AP-1 site in the IL-2 promoter (15) (Fig. 3A). Acp5-160 contains the binding site for the microphthalmia transcription factor (31). From EMSA results, the binding activity at Acp5-120 was indistinguishable from that observed with the IL-2 site (Fig. 3B). Binding of c-Fos to the Acp5 promoter in osteoclasts was demonstrated by chromatin immunoprecipitation assay (Fig. 2B). Then we mutated the putative NFAT site in Acp5-120 from GGAGAA to GGC-CCG in Acp5-luciferase reporter plasmids. Both human and mouse wild-type Acp5-luciferase constructs were most effi-ciently activated with NFATc4 compared with NFATc1, NFATc2, and NFATc3 in transient transfection assays (data not shown). Thus we constructed a constitutively active nuclear form of NFATc4, ⌬NFAT (20), fused to GFP. When the wildtype and mutant reporter plasmids were co-transfected with a ⌬NFAT expression plasmid into RAW264.7 cells, ⌬NFAT activated the wild-type Acp5 promoter more efficiently than the mutant promoter (Fig. 3C). Furthermore, the Acp5 promoter activity was enhanced by a co-transfected c-Fos expression vector and suppressed by small interfering RNA vectors for c-Fos in transient transfection assays (Fig. 3D). These data suggest that Acp5-120 is a functional NFAT/AP-1 binding site in the Acp5 promoter.
Next we searched for NFAT/AP-1 sites in the mouse osteoclast-specific Calcr-P3 promoter (22) and found eight putative NFAT/AP-1 sites in the Ϫ797 Calcr-P3 promoter (Fig. 4A). Co-transfection of both the Calcr-P3-luciferase reporter plasmids and the ⌬NFAT-expression vector into RAW264.7 cells resulted in a 30-fold increase in promoter activity above constitutive levels. Sequential 5Ј deletion of the Calcr-P3 promoter demonstrated that the Ϫ178 construct containing the putative NFAT/AP-1 sites 1-4 was sufficient for full activity. Site-specific mutagenesis of each NFAT site from GGAAAN to GGC-CCG revealed that site 2 at Ϫ93 was critical and that sites 1, 3, and 4 appear to cooperate with site 2 (Fig. 4A). In EMSA, site 2 was bound by NFATc1 using either RANKL-stimulated RAW264.7 nuclear extracts or in vitro translation products (Fig. 4B). Next we transfected RAW264.7 cells with Calcr-P3luciferase reporters and treated them with the calcium ionophore A23187. We observed that the Calcr promoter activities were enhanced only when site 2 was present presumably through activation of endogenous NFAT (Fig. 4C). Furthermore, the stimulatory effect of A23187 was blunted by pretreatment of the cells with cyclosporin A (Fig. 4D). These data suggest that site 2 is the critical NFAT site in the Calcr promoter.
NFAT Rescues Osteoclast Differentiation in Fos Ϫ/Ϫ Precursors-To test whether NFAT activity could rescue osteoclastogenesis in the absence of c-Fos, we introduced GFP or the GFP fusion ⌬NFAT into Fos Ϫ/Ϫ splenocytes by retroviral gene transfer. Infected cells were co-cultured with the ST2-T cells under osteoclastogenic conditions. At day 6, mRNA was harvested and microarray analysis was performed to compare gene expression between GFP and ⌬NFAT virus-infected Fos Ϫ/Ϫ cells. Strikingly, expression of ⌬NFAT activated about two-thirds of the genes that failed to be induced in Fos Ϫ/Ϫ cells including Acp5, Calcr, Ctsk, and endogenous Nfatc1 (Table I, ⌬NFAT/ GFP). This indicated that the differentiation block was to a large extent overcome by ⌬NFAT in the absence of c-Fos when RANKL from ST2-T cells was present. Indeed, whereas abundant GFP-positive cells were observed by day 6 with both viruses, TRAP-positive cells were generated only with ⌬NFAT virus (Fig. 5A). Next we tested whether the ⌬NFAT-expressing Fos Ϫ/Ϫ osteoclasts could resorb bone. Although resorption pits were not visible on bone slices in cultures of GFP virus-infected Fos Ϫ/Ϫ splenocytes, co-cultures containing ⌬NFAT virus-infected Fos Ϫ/Ϫ cells generated multiple resorption pits (bone surface resorbed, 2.3 Ϯ 0.5%) (Fig. 5B). Next we compared the rescue efficiency in two types of osteoclastogenic cultures, coculture using ST2-T cells and osteoblast-free cultures using only soluble M-CSF and RANKL. In co-cultures, the rescue with ⌬NFAT was comparable with that with c-Fos as judged by  Fig. 2B. B, EMSA using Acp5-120 (5), IL-2 distal NFAT/AP-1 site (I) and AP-1 consensus (A) or mutant (m) sites for probes and competitors (Comp.). Nuclear extracts containing NFAT binding activity were prepared from osteoclastogenic co-culture. C, transient transfection assay in RAW264.7 cells using wild-type (WT) and mutant (MUT) mouse Acp5 promoter-luciferase constructs. The activator plasmid expressing ⌬NFAT (⌬) was co-transfected. RLU, relative light units (normalized to co-transfected Renilla luciferase activity and relative to the unstimulated wild-type promoter). D, the Acp5 promoter-luciferase construct was activated by c-Fos expression vector (Fos) and suppressed by Fos small interfering RNA vectors (RNAi1, RNAi2) in transient co-transfection in RAW264.7 cells.

FIG. 4. Identification of NFAT/AP-1 sites in the Calcr-P3 promoter.
A, a series of Calcr-P3 constructs was tested in RAW264.7 cells. The numbering is relative to the transcriptional start. Open ovals are sites 1-8 (22), and X is a site-specific mutant. Asterisks indicate nucleotides that are identical between Calcr-P3 site 2 and the IL-2 site. Relative light units (RLU) are normalized to micrograms of protein and are relative to unstimulated pGL3, a promoterless luciferase vector. B, EMSA using Calcr-P3 site 2. The DNA-binding protein source was either nuclear extracts prepared from RAW264.7 cells cultured in the absence (Ϫ) or presence of RANKL or in vitro translated NFATc1 (c1). Arrows and arrowheads indicate specific NFAT binding activity and supershifts, respectively. The oligonucleotide competitors were wildtype site 2 (2), site 2 with the putative AP-1 site mutated (2mA), site 2 with the NFAT site mutated (2mN), and wild-type AP-1 (AP-1). Supershifts were done with antibodies to NFATc1 (␣c1) and NFATc2 (␣c2). C, induction of Calcr-P3 by A23187 (1 M), which was added to transfected RAW264.7 cells 2 h before harvest. WT, wild type. D, transient transfection assay in RAW264.7 cells using the Ϫ319 Calcr-P3-luciferase construct. Cyclosporin A (1 g/ml) was added 1 h before transfection, and A23187 (1 M) was added for 4 h before harvest.
Calcr expression, TRAP-positive cell numbers, and resorption (Fig. 5C, st). In contrast, in the absence of stromal cells the rescue with ⌬NFAT was lower than that of c-Fos based on all three parameters (Fig. 5C, MϩR). Apart from the differences in efficiency in both cultures, ⌬NFAT substituted at least in part for the osteoclastogenic function of c-Fos to the extent that Fos Ϫ/Ϫ splenocytes formed bone resorption pits in the presence of RANKL. In addition, gene transfer of the human full-length NFATc1 also rescued Fos Ϫ/Ϫ osteoclastogenesis in vitro (data not shown). These data collectively indicate that the lack of NFATc1 is a major reason for the differentiation block in Fos Ϫ/Ϫ osteoclast precursors.
NFAT Rescue of Fos Ϫ/Ϫ Precursors Is RANKL-dependent-To examine the role of RANKL in NFAT-induced osteoclast formation, we introduced the GFP fusion ⌬NFAT into RAW264.7 cells by transient transfection. This resulted in the induction of the endogenous Acp5 and Calcr genes as early as 1 day after transfection even in the absence of RANKL (Fig.  6A). This is consistent with the reported osteoclastogenic activity of NFATc1 in the absence of RANKL (5). Next, we tested whether ⌬NFAT could rescue Fos Ϫ/Ϫ precursors in the absence of RANKL. ⌬NFAT produced bone-resorbing TRAP-positive cells from wild-type precursors (Bmϩ/ϩ, Spϩ/ϩ) in the absence or presence of RANKL (Fig. 6B). However, Fos Ϫ/Ϫ precursors (SpϪ/Ϫ) hardly produced any TRAP-positive cells upon introduction of ⌬NFAT in the absence of RANKL (Fig.  6B), and no bone resorption pits were observed (data not shown). These data suggest that the rescue of Fos Ϫ/Ϫ cells with NFAT activity requires receptor activator of NF-B (RANK) signaling. DISCUSSION It has been established that the lack of c-Fos expression results in a differentiation block in the osteoclast lineage (7)(8)(9). However, whether this is a cumulative effect of numerous deregulated c-Fos target genes or an effect of one critical c-Fos target gene is unclear. Our results show that the absence of Nfatc1 expression in Fos Ϫ/Ϫ precursors is the major cause of the differentiation block because an active form of NFAT alone rescued osteoclast-specific gene expression and bone resorptive function.
We have identified NFAT/AP-1 sites in the Acp5 and Calcr promoters. EMSA showed that NFAT and AP-1 cooperatively bind to the Acp5 NFAT/AP-1 site, and the chromatin immunoprecipitation assay indicated that c-Fos binds to the Acp5 promoter in osteoclasts. These observations are consistent with the idea that NFATc1 and c-Fos synergize to activate the Acp5 promoter (5). On the other hand, the rescue of osteoclastogenesis by ⌬NFAT alone in the absence of c-Fos clearly demonstrates that c-Fos is not essential for activation of the Acp5 promoter. To activate these promoters in Fos Ϫ/Ϫ precursors, ⌬NFAT may interact with Jun-Jun homodimers (32) or may act alone in the absence of cooperative partners (33). EMSA using Fos Ϫ/Ϫ cell extract in combination with in vitro translated NFATc1 will help to address this issue. Although binding of Transcripts were analyzed by reverse transcriptase-PCR with primers for Acp5, Calcr (all isoforms), and glyceraldehyde-3-phosphate dehydrogenase (Gapdh). The reverse transcripts-PCR products for Calcr were also detected by hybridization with an internal probe (hyb). B, rescue of TRAP-positive Fos Ϫ/Ϫ cells by ⌬NFAT requires RANKL. M-CSF-dependent macrophages derived from wild-type and Fos Ϫ/Ϫ splenocytes (Spϩ/ϩ and SpϪ/Ϫ) and wild-type bone marrow cells (Bmϩ/ϩ) were infected with the GFP ⌬NFAT virus and then were cultured for 3 more days in the presence or absence of RANKL. The bottom row shows expression of GFP ⌬NFAT fusion protein.
NFAT to the Calcr promoter was unambiguously demonstrated by EMSA, binding of AP-1 to the Calcr promoter needs to be rigorously tested in the future.
Importantly, osteoclast-specific gene expression is not entirely rescued with ⌬NFAT. Those genes for which expression is not rescued, for example Mmp9, may be more strictly dependent on c-Fos or additional c-Fos-dependent transcription factors. Curiously, the rescue activity of ⌬NFAT was similar to that of c-Fos when Fos Ϫ/Ϫ precursors were co-cultured with ST-2 but was lower than that of c-Fos when soluble M-CSF and RANKL were used. Therefore, c-Fos dependence appears to increase as stromal factors decrease. One of the stromal factors involved might be the ligand of osteoclast-associated receptor (OSCAR) (34). Whereas in wild-type precursors ⌬NFAT or NFATc1 induce osteoclast differentiation even in the absence of RANKL (5), ⌬NFAT expression in Fos Ϫ/Ϫ precursors failed to rescue osteoclast differentiation in the absence of RANKL. This suggests that in the absence of RANKL, NFAT requires c-Fos, presumably as a binding partner or possibly indirectly to exert its osteoclastogenic function. It also suggests that RANKL may induce an alternative partner for NFAT that can substitute for c-Fos function in Fos Ϫ/Ϫ cells.
Taken together, these results demonstrate that a major function of c-Fos during osteoclast formation is to trigger a transcriptional regulatory cascade by producing and cooperating with NFATc1, thereby activating a number of target genes involved in osteoclast differentiation and function. These yet to be identified novel target genes together with Nfatc1 may provide additional drug targets for bone diseases including osteoporosis and rheumatoid arthritis.