Identification of Nedd9 as a TGF-β-Smad2/3 Target Gene Involved in RANKL-Induced Osteoclastogenesis by Comprehensive Analysis

TGF-ß is a multifunctional cytokine that is involved in cell proliferation, differentiation and function. We previously reported an essential role of the TGF-ß -Smad2/3 pathways in RANKL-induced osteoclastogenesis. Using chromatin immunoprecipitation followed by sequencing, we comprehensively identified Smad2/3 target genes in bone marrow macrophages. These genes were enriched in the gene population upregulated by TGF-ß and downregulated by RANKL. Recent studies have revealed that histone modifications, such as trimethylation of histone H3 lysine 4 (H3K4me3) and lysine 27 (H3K27me3), critically regulate key developmental steps. We identified Nedd9 as a Smad2/3 target gene whose histone modification pattern was converted from H3K4me3(+)/H3K4me27(+) to H3K4me3(+)/H3K4me27(-) by TGF-ß. Nedd9 expression was increased by TGF-ß and suppressed by RANKL. Overexpression of Nedd9 partially rescued an inhibitory effect of a TGF-ß inhibitor, while gene silencing of Nedd9 suppressed RANKL-induced osteoclastogenesis. RANKL-induced osteoclastogenesis were reduced and stimulatory effects of TGF-ß on RANKL-induced osteoclastogenesis were partially abrogated in cells from Nedd9-deficient mice although knockout mice did not show abnormal skeletal phenotypes. These results suggest that Nedd9 is a Smad2/3 target gene implicated in RANKL-induced osteoclastogenesis.


Introduction
Skeletal homeostasis is strictly controlled by osteoclasts, which mediate bone resorption, and osteoblasts, which regulate bone formation. Osteoclasts are multinucleated cells derived from monocyte-macrophage lineage hematopoietic progenitor cells and specifically differentiated for bone resorption [1]. The differentiation of osteoclasts is regulated by two cytokines: receptor activator of nuclear factor kappa B ligand (RANKL) and macrophage colony-stimulating factor (M-CSF). In addition to these two essential cytokines, we recently reported a critical role for TGF-ß in osteoclastogenesis [2] [3]. TGF-ß is abundantly stored in bone matrix and has profound biological functions such as angiogenesis, cellular differentiation, apoptosis and bone homeostasis [4] [5]. The binding of TGF-ß to its type II receptors recruits and phosphorylates type I receptors, which in turn activate downstream signaling including Smad and non-Smad pathways [6]. Phosphorylated Smad2/3 forms a complex with Smad4, and the molecular complex translocates into the nucleus and regulates specific gene expression [7] [8] [9].
We previously reported that TGF-ß is required for osteoclast differentiation in response to RANKL and M-CSF by regulating the interaction of Smad2/3 with TRAF (tumor necrosis factor receptor-associated factor) 6, an adaptor molecule associated with RANK [2]. In addition, we identified Smad2/3-binding sites in open chromatin regions during osteoclastogenesis and found that Smad2/3 binding is necessary for the nuclear translocation of c-Fos, an essential transcription factor for osteoclastogenesis [3]. Moreover, it was reported that combined treatment of TGF-ß and TNF-α promotes maximal osteoclast formation compared to treatment with other cytokine combinations in the presence of RANKL based on a multiparameter cytokine assay [10]. However, direct target genes that regulate osteoclast differentiation downstream of TGF-ß-Smad2/3 pathways still remain elusive.
Multiple epigenetic modifications, such as DNA methylation and, histone acetylation and methylation, are involved in organization of chromatin structures at various levels and regulation of gene expression. The methylated sites in H3 or H4 are mainly located in the histone tail (H3K4, H3K9, H3K36 and H4K20) and the center of the nucleosome (H3K79) [11]. Among the five histones, which are designated as H1, H2A, H2B, H3 and H4 [12], Stahl et al. reported that the methylation of histone H3 at lysine 4 is highly conserved and correlated with transcriptionally active nuclei in Tetrahymena [13]. Bernstein et al. revealed that histone modifications such as trimethylation of histone H3 lysine 4 (H3K4me3) and lysine 27 (H3K27me3) play a critical role in gene expression, and in embryonic stem cells, key developmental genes tend to change histone modification patterns from the H3K4me3/H3K27me3 bivalent pattern to the H3K4me3 monovalent pattern [14]. Similar modifications of histone methylation have been observed in many other types of cells, and we previously reported that RANKL induced bivalent to monovalent changes in the nuclear factor of activated T-cells cytoplasmic 1 (NFATc1) gene during osteoclastogenesis [15]. Through chromatin immunoprecipitation with sequencing (ChIP-seq) analysis using anti-Smad2/3, anti-H3K4me3 and anti-H3K27me3 antibodies [16], here we investigated Smad2/3-regulating genes that are critically involved in the differentiation of osteoclasts and identified Nedd9 as a putative regulator of osteoclastogenesis downstream of TGF-ß-Smad2/3 pathways.

Results
Identification of genes regulated by the TGF-β-Smad2/3 axis To identify genes regulated by the TGF-ß-Smad2/3 axis, we performed ChIP-seq analysis using anti-Smad2/3 antibody in M-CSF-dependent bone marrow macrophages (BMMs) treated with 2 ng/ml TGF-ß for 1.5 h. Total read number was 15,108,905, and 10,837,516 reads (71.7%) were mapped to the mouse genome. A total of 2,786 Smad2/3-binding regions (SBRs) were identified (peak signal ratio 8). Genes with peak positions of SBRs between 10 kb upstream from transcription start sites (TSSs) and first intron were defined as Smad2/3 target genes, and 903 genes were selected as Smad2/3 target genes.
TGF-β positively and RANKL negatively regulates Smad2/3 target genes Using 14,177 probes (8,839 genes) with expression values of more than 70 by MOE430 Gene-Chips at least one time point, we found that Smad2/3 target genes were significantly enriched in the genes whose expression was more than 2-fold upregulated, but not in those whose expression was less than 0.5-fold downregulated, by TGF-ß (Fig 3, upper and lower panels; P < 10 −5 by chi square test). Enrichment scores calculated by Gene Set Enrichment Analysis (GSEA) [17] exhibited statistically significant enrichment (P < 10 −6 ) (Fig 3, lower panel).
We then analyzed the change of the expression of Smad2/3 target genes by RANKL stimulation. We used 16,631 probes (10,004 genes) with expression values of more than 70 at least one time point for further analysis. Interestingly, Smad2/3 target genes were significantly enriched in the genes whose expression was less than 0.5-fold downregulated, but not in those whose expression was more than 2-fold upregulated, after 24, 48 and 72 h of RANKL stimulation (Fig 4, upper and lower panels; P < 10 −5 by chi square test). In addition, genes whose histone modification was changed from K4(+)K27(+) to K4(+)K27(-) patterns by TGF-ß treatment were enriched in downregulated genes by RANKL (Fig 4, upper panel). The statistical significance of the enrichment was analyzed by calculating GSEA enrichment scores (P < 10 −6 ) (
We then analyzed a potential function for Nedd9 in osteoclastogenesis. Retroviral overexpression of Nedd9 significantly increased the expression of Cathepsin K, a marker gene of osteoclasts ( Fig 6A). In addition, Nedd9 overexpression partially rescued the inhibitory effect of SB431542 on RANKL-induced osteoclastogenesis (P < 0.05; Fig 6B-6D). Conversely, Nedd9 knockdown by retrovirus carrying shNedd9 markedly suppressed RANKL-induced osteoclastogenesis (Fig 7A-7C). All of these findings strongly suggest a critical function for Nedd9 in RANKL-induced osteoclastogenesis.
Finally, we assessed the function of Nedd9 in osteoclastogenesis using Nedd9-/-mice. BMMs from Nedd9-/-mice exhibited reduced osteoclastogenesis similar to shNedd9-treated cells (Fig 8A and 8B). Stimulatory effects of TGF-ß on RANKL-induced osteoclastogenesis observed in wild-type BMMs was reduced in Nedd9-deficient BMMs (Fig 8B). Expression of Cathepsin K, as determined by Western blotting, was downregulated in Nedd9-deficient osteoclasts as compared to wild-type osteoclasts ( Fig 8C). However, no significant difference in the skeletal phenotypes, as assessed by soft X ray in the lower extremities, micro CT in lumbar vertebral bodies, and dual energy X ray absorptiometry was observed between Nedd9-knockout and wild-type mice (Fig 9A-9C).

Discussion
We previously reported that TGF-ß is indispensable for RANKL and M-CSF-induced osteoclastogenesis. However, the effector genes acting downstream of TGF-ß-Smad2/3 pathways remain elusive. In the present study, we comprehensively analyzed Smad2/3-binding regions and identified Smad2/3 target genes in BMMs by ChIP-seq analysis. Smad2/3 target genes were enriched in TGF-ß upregulated genes as expected. Interestingly, these genes were enriched in the gene population whose expression was downregulated by RANKL treatment, indicating the possibility that the role of TGF-ß is to prepare an appropriate condition for BMMs to differentiate into mature osteoclasts upon RANKL stimulation.
Histone modifications play important roles in cell differentiation. H3K4me3 is enriched in the active and poised promoter regions [24,25], while H3K27me3 is involved in polycombmediated gene repression [26]. Recent studies have revealed that key developmental genes tend to change histone modification patterns from the H3K4me3/H3K27me3 bivalent pattern to the H3K4me3 monovalent pattern in various types of cells. In an attempt to narrow down the candidates of Smad2/3 target genes involving in osteoclast differentiation, we identified Nedd9 as a Smad2/3 target gene whose histone modification pattern changed from K4(+)K27(+) to of each SBR are shown by black bars. (B) Eight positive regions and two negative regions for Smad2/3 binding were selected from ChIP-seq data and validated by realtime PCR. ChIP using mouse IgG was used as control. Values are presented as n-fold enrichment over Hprt1. (C) Average Smad2/3 signal profile around transcriptional start site (TSS) in ChIP-seq analysis. Smad2/3 binding was enriched around TSS.
doi:10.1371/journal.pone.0157992.g001 Fig 2. (A) Genes with H3K4me3 peaks within +/-1 kb from TSS were defined as K4(+) genes, and genes with H3K27me3 peaks within +/-1 kb K4(+)K27(-) patterns in response to TGF-ß. Nedd9, also known as CasL and HEF1, was originally identified as a 105 kDa protein that is tyrosine phosphorylated by the ligation of ß1 integrins in peripheral T cells [18]. Nedd9 is induced by TGF-ß and directly interacts with Smads in various types of cells [27] [28]. We found that Smad2/3 binds to the promoter region of Nedd9 gene and TGF-ß upregulates Nedd9 expression in BMMs. Overexpression of Nedd9 promoted and knockdown or knockout of Nedd9 suppressed osteoclastogenesis, indicating a role of Nedd9 downstream of RANKL-RANK pathways. Previous studies reported that Nedd9 is The expression scores after TGF-ß stimulation of BMMs relative to time 0 are illustrated by a heat map with red or green color representing increased or decreased gene expression, respectively. The probes were sorted by the ratio at 24 h (ranked gene list). Horizontal bars indicate Smad2/3 target genes. GSEA enrichment scores are graphically shown in the right panel [17]. Smad2/3 target genes were significantly enriched in genes upregulated by TGF-ß. Lower: The expression scores after RANKL stimulation of BMMs relative to time 0 are illustrated by a heat map. Genes whose expression was upregulated more than 2-fold or downregulated less than 0.5-fold relative to RANKL treatment were counted. Enrichment of Smad2/3 target genes in up-or downregulated genes was evaluated by chi-square test (P < 10 −5 ).  involved in tumor differentiation, migration and metastasis [29], and therefore, it is also possible that Nedd9 supports osteoclast motility as well.
A previous study reported the association between Nedd9 and Smad6 or Smad7 [30], and our ChIP-seq analysis showed that Smad2/3 bound to the promoter region of Smad6 and Smad7. Although we did not address the association between inhibitory Smads and Nedd9 in BMMs, it is possible that Nedd9 regulates TGF-ß signaling in a negative feedback manner by interacting with inhibitory Smads.  The role of Nedd9 in skeletal homeostasis in vivo is not clear. Seo et al. generated Nedd9deficient mice and reported that lymphocyte trafficking was altered [22]. In addition, Katayose et al. recently reported that Nedd9-/-mice exhibited decreased onset of collagen-induced arthritis compared with wild-type mice and that joint destruction was reduced in the knockout mice [31]. We found that RANKL-induced osteoclastogenesis was impaired in BMMs from Nedd9-/-mice, and the stimulatory effect of TGF-ß on RANKL-induced osteoclastogenesis was partially abrogated. However, we were unable to find the abnormal bone phenotypes in RANKL-induced osteoclastogenesis. SB431542 suppressed osteoclastogenesis, which was partly recovered by Nedd9 overexpression. Cultures were stained by TRAP. Bars = 100 μm. (C) The expression of Nedd9 gene as determined by realtime PCR. (D) The number of TRAP positive osteoclasts was significantly increased by Nedd9 overexpression, and the suppression of osteoclastogenesis by SB431542 was partly suppressed by Nedd9 overexpression. *P < 0.05. Nedd9-/-mice. This may be due to the effect of Nedd9 on other types of cells such as osteoblasts or osteocytes, or the effect of Nedd9 deficiency is only observed in mice under pathological conditions such as ovariectomy and arthritis, or those treated by TGF-ß. Further studies are required to fully understand the role of Nedd9 in the skeletal milieu.

Animals
Nedd9-/-mice with C57BL/6J (B6) genetic background were generated as previously reported [22]. In short, the exon 2 encoding the N-terminal SH3 domain in the Cas-L protein was replaced with EGFP and a neomycin resistance gene. All animal procedures were approved by the Animal Care and Use Committee of the University of Tokyo.

Cell Culture
Murine bone marrow cells were collected from the femur and tibia of male ddY mice at 4-5 weeks of age. To prepare BMMs, cells were cultured in α-MEM/10% FBS with 100 ng/ml M-CSF for 5 days. BMMs were further cultured in the presence of 10 ng/ml M-CSF and 100 ng/ml RANKL for 4 days to generate osteoclasts. To examine the effect of the TGF-β-Smad pathway on osteoclastogenesis, 1 ng/ml TGF-β or 10 μM SB431542 was added with RANKL. Cells were stained with tartrate-resistant acid phosphatase (TRAP), and TRAP-positive cells containing more than three nuclei were counted as osteoclasts32.

Real-time PCR analysis
Total RNA was extracted with ISOGEN (Wako Pure Chemical Industries, Ltd.), and a 1-μg aliquot was reverse transcribed using a QuantiTect Reverse Transcription kit (QIAGEN) to produce singlestranded cDNA. PCR was performed on an ABI Prism 7000 Sequence Detection System (Applied Biosystems) using QuantiTect SYBR Green PCR Master Mix (QIAGEN) according to the manufacturer's instructions. All reactions were performed in triplicate. After data collection, the mRNA copy number of a specific gene was calculated with a standard curve generated with serially diluted plasmids containing PCR amplicon sequences and then normalized to rodent total RNA with mouse ß-actin serving as an internal control. Primer sequences were as follows: Nedd9 forward, 5 0 -CCACCCTCCTACC AGAATCA-3 0 ; Nedd9 reverse, 5 0 -ATACCCCTTGAGTGCTGTGG-3 0 ; Cathepsin K-forward, 5'-ACGGAGGCATTGACTCTGAAGATG-3'; Cathepsin K-reverse, 5'-GGAAGCACCAACG AGAGGAGAAAT-3'.

Expression constructs and gene transduction
For retrovirus construction, the full-length cDNAs were amplified by PCR using KODplus (Takara Bio Inc.), subcloned into Zero Blunt TOPO II vectors (Invitrogen), and inserted into pMX-puro vectors. A total of 2 × 10 6 BOSC23 packaging cells were transfected with 6 μg of vector using FuGENE 6 (Roche). After 24 h, the medium was replaced with fresh α-MEM/10% FBS, and cells were incubated for an additional 24 h. The supernatant was then collected as retroviral stock after centrifugation at 2,400 rpm for 3 min. A total of 5×10 6 BMMs were incubated with 8 ml of retroviral stock for 5 h in the presence of 6 μg/ml polybrene and 30 ng/ml recombinant mouse M-CSF. After 5 h of retroviral infection, the medium was changed to α-MEM/10% FBS and 100 ng/ml M-CSF, and cells were cultured for an additional 24 h. BMMs were recovered with trypsin, and puromycin-resistant cells were selected by incubation with α-MEM/10% FBS containing 2 μg/ml puromycin for 2 d and used for further experiments.

Microarray expression array analysis
Total RNA was extracted with TRIzol and subjected to GeneChip (Affymetrix; Santa Clara, CA, USA) expression analysis according to the technical manual. Briefly, biotin-labeled cRNA synthesized from total RNA was hybridized to a GeneChip Mouse Genome MOE430 2.0 oligonucleotide array (Affymetrix; Santa Clara, CA, USA). The arrays were stained with streptavidin-phycoerythrin and image data was collected with an Affymetrix scanner. Microarray Suite software 5.0 was used to calculate the average difference (AD) for each gene probe set, shown as the gene expression intensity value. The AD values were normalized for each array so that the average of all AD values was 100. One array datum was obtained for each sample. Obtained data were verified by qPCRs for various transcripts, and we had no conflicting results between the array data and qPCR data. Affymetrix probe IDs were converted to gene symbols.

Statistical analyses
The results are expressed as mean±SD. Statistical analyses were performed using a two-tailed unpaired Student's t test for continuous variables and chi-square tests for categorical variables. A p value of less than 0.05 was considered to be statistically significant.