NELL-1, an Osteoinductive Factor, Is a Direct Transcriptional Target of Osterix

NELL-1 is a novel secreted protein associated with premature fusion of cranial sutures in craniosynostosis that has been found to promote osteoblast cell differentiation and mineralization. Our previous study showed that Runx2, the key transcription factor in osteoblast differentiation, transactivates the NELL-1 promoter. In this study, we evaluated the regulatory involvement and mechanisms of Osterix, an essential transcription factor of osteoblasts, in NELL-1 gene expression and function. Promoter analysis showed a cluster of potential Sp1 sites (Sp1/Osterix binding sites) within approximately 70 bp (from −71 to −142) of the 5′ flanking region of the human NELL-1 transcriptional start site. Luciferase activity in our NELL-1 promoter reporter systems was significantly decreased in Saos-2 cells when Osterix was overexpressed. Mutagenesis study demonstrated that this suppression is mediated by the Sp1 sites. The binding specificity of Osterix to these Sp1 sites was confirmed in Saos-2 cells and primary human osteoblasts by EMSA in vitro and ChIP assay in vivo. ChIP assay also showed that Osterix downregulated NELL-1 by affecting binding of RNA polymerase II to the NELL-1 promoter, but not by competing with Runx2 binding to the OSE2 sites. Moreover, NELL-1 mRNA levels were significantly decreased when Osterix was overexpressed in Saos-2, U2OS, Hela and Glioma cells. Correspondingly, knockdown of Osterix increased NELL-1 transcription and osteoblastic differentiation in both Saos-2 cells and primary human osteoblasts. These results suggest that Osterix is a direct transcriptional regulator with repressive effect on NELL-1 gene expression, contributing to a delicate balance of regulatory effects on NELL-1 transcription with Runx2, and may play a crucial role in osteoblast differentiation and mineralization. These findings also extend our understanding of the molecular mechanism of Runx2, Osterix, and NELL-1 and demonstrate their crosstalk during osteogenesis.


Introduction
Through promoter analyses, we recently established NELL-1, a Nel-like molecule-1 [1,2], as a novel direct transcriptional target of runt homology domain transcription factor-2 (Runx2) [3]. Sitedirected mutagenesis and chromatin immunoprecipitation (ChIP) assays revealed at least three functional consensus osteoblast specific binding elements 2 (OSE2) on the human NELL-1 promoter. Significantly, the overexpression of NELL-1 was originally found in pathologically fusing and fused sutures in nonsyndromic unilateral coronal synostosis (UCS) patients [4], and CMV-Nell-1 overexpression mice exhibited CS-like phenotypes that ranged from simple to compound synostoses [5]. These findings highly suggest that NELL-1 is a CS-associated factor with preferential osteogenic effects on cells of the osteochondral lineage. Furthermore, N-ethyl-N-nitrosourea (ENU)-induced Nell-1 defi-cient mice revealed major abnormalities in the skeletal system such as decreased calvarial bone mineralization and decreased vertebral disc volume, and perinatal death due to respiratory failure secondary to a deformed cartilaginous ribcage [6]. This Nell-1 deficient mouse model in addition to the overexpression transgenic mouse model further supports the critical role of Nell-1 in the Runx2 regulatory network of osteogenesis, however, the precise mechanism of action of Nell-1 remains unknown [7,8].
Osterix/Sp7 (Osx), a member of the Sp1 transcription factor family, is also essential for osteoblastogenesis [9,10,11]. Like Runx2-null mice, Osterix-null mice exhibit complete absence of bone matrix and osteoblasts, indicating an absolute requirement for Osterix in osteoblast formation [9]. However, Osterix-null mice exhibit normal cartilage hypertrophy while Runx2-null mice do not. In addition, Osterix-null mice exhibit normal Runx2 levels, while Osterix is not expressed in Runx2 null-mice suggesting that Osterix is downstream of and tightly regulated by Runx2. The Osterix promoter does contain at least one functional Runx2 binding site [12], however, Osterix can be induced by BMP2 in Runx2-null cells [13], possibly through upregulation of Dlx5 and its phosphorylation by p38. Thus, Osterix exhibits both Runx2 dependent and independent regulation. Previous studies have suggested that Osterix functionally segregates osteoblast and chondrocyte lineages whereby bipotential precursor cells initially express Runx2 and then express Osterix to suppress chondrogenic lineage and promote osteoblast differentiation [14]. Consistent with this, Kaback et al. demonstrated Osterix expression in perichondrium, immature chondrocytes, and osteoblasts, but not hypertrophic chondrocytes during development [15].
Interestingly, the transduction of AdNell-1 inhibited Osterix mRNA expression without affecting Runx2 mRNA levels during osteoblastic differentiation of preosteoblastic MC3T3 cells [5,16], which may indicate a potential regulatory and functional relationship between Nell-1 and Osterix in addition to what has been discovered between Nell-1 and Runx2 in osteoblastic differentiation, leading us to pursue this current study. Here we demonstrated that overexpression of Osterix can suppress NELL-1 expression at the transcriptional level in multiple human osteoblast-like and non-osteoblastic cell lines, and verified that this inhibitory effect on NELL-1 expression with and without Runx2 induction involves Osterix direct binding of Sp1 sites in the NELL-1 promoter in a human osteosarcoma cell line, Saos2. We also verified that Nell-1 has inhibitory effects on Osterix expression during osteoblastic differentiation reciprocally. Taken together, we conclude that a delicate balance of regulatory effects on Nell-1 transcription by Osterix and Runx2 is crucial, and these novel findings provide new insights into the underlying mechanism of Nell-19s action during osteochondral differentiation.

Cell Culture
Saos-2 cell line was purchased from ATCC (Manassas, VA) and primary human osteoblast cells (HOb) were purchased from Cell Applications Inc (San Diego, CA). Cells were cultured in growth medium of Dulbecco's modified Eagles' medium (DMEM) with 10% fetal bovine serum (FBS), 100 units/ml penicillin, and 100 mg/ml streptomycin or osteoblastic differentiation medium containing additional 50 ug/ml ascorbic acid and 10 mM betaglycerophosphate in the growth medium. Cells were plated at a density of 2.5610 4 cells/cm 2 in DMEM for the experiments.

Reporter and Expression Constructs
The p2213WT-Luc and p325WT-Luc vectors containing the human NELL-1 promoter fragment were generated as described previously [3]. Mutant constructs of the p325WT-Luc promoter including p325mut all-Luc with mutation in all Sp1 sites of both cluster Site A and Site B, p325mutSiteA-Luc with mutation in cluster Site A, and p325mutSiteB with mutation in Site B were produced by GenScript USA Inc. (Piscataway, NJ). Human Osterix expression construct (pcDNA3.1-Osterix, pOsx), generated by RT-PCR from Saos-2 cell cDNA, and empty pcDNA3.1 control vector (pCtr) were utilized for all co-transfection experiments. The Runx2 expression plasmid (pRunx2) was a generous gift from Dr. Wenfang Wang of Harvard University. Renilla control plasmid was purchased from Promega (Madison, WI). All constructs were confirmed by reproducible sequencing (Laragen, Los Angeles, CA). Each experiment was repeated at least 3 times. The data is presented as the mean 6 SD. A t-test was performed, with p,0.05 considered statistically significant [3]. Cells in 6 well plates were transfected and grown to 80-90%  confluence with Lipofectamin 2000 (Invitrogen, Carlsbad, CA). For each reaction, the plasmid DNA (promoter-luciferase vectors p2213WT-Luc or p325WT-Luc plus Renilla plasmid, with pOsx or pCtr constructs) was transfected according to the manufacturer's instructions. Forty-eight hours after transfection, the cultures were harvested, and luciferase activity was assayed using Dual-Luciferase Assay System (Promega, Madison, WI), and then normalized by co-transfected Renilla activity. Each experiment was repeated at least 3 times. The data is presented as the mean 6 SD. A t-test was performed, and p,0.05 was considered statistically significant [3]. Total RNA of transfected cells was isolated by Trizol Reagent (Invitrogen, Carlsbad, CA), and the expression levels of NELL-1 and other genes were detected by realtime PCR.

Electrophoretic Mobility Shift Assays
Preparation of nuclear extracts from Saos-2 cells transfected with pCtr or pOsx were performed using Nuclear Extract Kit (Pierce, Rockford, IL) according to the manufacturer's protocol. Briefly, cells were harvested in ice-cold PBS supplemented with phosphatase and protease inhibitors, then lysed in hypotonic buffer and 1% Triton X-100. After collection of the cytoplasmic fraction, nuclei were lysed and the nuclear extract proteins were solubilized in lysis buffer supplemented with 10 mM DTT, and protease inhibitor cocktail. According to the human NELL-1 promoter analysis (Fig. 1), we designed two oligos, Site A containing three overlapping Sp1 sites and Site B containing a single Sp1 site. The mutated Site A oligo (Mut A ) has a mutation in every Sp1 site while the mutated Site B oligo (Mut B ) has the same mutation at its Sp1 site. Oligonucleotide probes labeled with and without biotin were generated from Invitrogen. Sequences used include: Site A: 59-GCGGTGGGGGGCGGGCTGGG-GGCGGGGCGGCCGGGGCTG-39, Mut A : 59-GCGGTGGG-GGAAGGGCTGGGGAAGGGAAGGCCGGGGCTG-39, Site B: 59-GCCGCAGCGGGGGGCCGGGCCCCGAGCCCC-39, and Mut B : 59-GCCGCAGCGGGGGGAAGGGCCCCGAGCC-CC-39. The EMSA protocol followed the LightShift Chemiluminescent EMSA Kit (Pierce, Rockford, IL) instructions. Briefly, 2 mg of nuclear extract protein (NE) plus biotin labeled oligonucleotide with and without anti-Osterix (Abcam, Cambridge, MA) in 20 ml binding system (containing 2.5% glycerol, 5 mM MgCl 2 , 50 ng/ml poly (dI?dC), 0.05% NP-40 and 1x binding buffer) was incubated at room temperature for 20 minutes. Five ml of 5x loading buffer was added to each 20 ml binding reaction, then 20 ml of each sample was loaded onto the 5% polyacrylamide gel in 0.5x TBE. Samples were electrophoresed until the bromophenol blue dye had migrated approximately 3/4 down the length of the gel. The gel and nylon membrane were sandwiched in a clean electrophoretic transfer unit, and transferred at 380 mA (,100V) for 30 minutes. Then DNA and membrane were crosslinked using a UV-light cross-linker instrument. Biotin-labeled DNA was detected by Chemiluminescence provided by the LightShift Chemiluminescent EMSA Kit (Pierce, Rockford, IL). The unlabeled oligos of Site A and Site B at 20x and 200x or 200x mutated Site A and Site B were added to compete specifically with labeled oligo binding in the competitive EMSA.

Chromatin Immunoprecipitation Assays
Saos-2 cells and primary human osteoblasts were used for chromatin immunoprecipitation (ChIP) assays with the One-day ChIP-qPCR kit (SABiosciences, Frederick, MD) according to the manufacturer's instructions. Briefly, cells were fixed with 1% formaldehyde for 10 minutes at 37uC. Cross-linked lysates were treated with glycine to stop the fixation reaction. Lysates were resuspended in shearing buffer and then sonicated to shear chromatin. Cross-linked lysates were immunoprecipitated with anti-Osterix (Abcam, Cambridge, MA), anti-Runx2 (Santa Cruz Biotechnology, Inc, Santa Cruz, CA), or non-specific IgG (SABiosciences, Frederick, MD) antibodies. Then DNA was recovered and used for qPCR with NELL-1 promoter primer pairs designed and provided by SABiosciences following the manufacturer's instructions. The human RNA Polymerase II ChIP-grade antibody and human GAPDH ChIP-qPCR control primer were provided in the ChampionChIP kit. The NELL-1 -1 kb primer set (NM_006157.2 201 kb) covers 1 kb proximal promoter region where all Sp1/Osterix binding sites and three OSE2 sites are located, and the NELL-1 22 kb primer set (NM_006157.2 202 kb) covers the NELL-1 promoter region from 21 kb to 22 kb where no Sp1 site and one OSE2 site exists ( Fig. 1). Each experiment was repeated at least 3 times. The data is presented as the mean 6 SD. A t-test was performed, with p,0.05 considered statistically significant [3].

Reverse Transcription-PCR and Real-Time PCR
One mg of DNase I treated total RNA were used for reverse transcription. The ABI Prism 7300 Real Time PCR System was utilized to quantify gene expression in the exponential phase of PCR reactions. TaqMan primer-probe sets for osteocalcin (Ocn), osteopontin (Opn), NELL-1, Osterix, and glyceraldehyde-3-phosphate dehydrogenase (Gapdh) were purchased (Applied Biosystems, Foster City, CA) and analyzed by real time PCR as previously described [5]. Relative gene expression profiles were calculated using the comparative quantification formula as 2 2DD Ct based on the evaluation of similar dynamic ranges for RT-PCR efficiency of both Gapdh and the target genes. Each experiment was repeated at least 3 times. The data is presented as the mean 6 SD. A t-test was performed, with p,0.05 considered statistically significant [3].

Transfection of Osterix siRNA into Saos-2 cells and primary human osteoblasts
Human Osterix siRNA oligos were designed and synthesized by Invitrogen (Silencer Select pre-designed siRNA s42458, s42459 and s42460 against Osterix mRNA). Saos-2 cells or human osteoblasts were seeded at 2.5610 4 /cm 2 into 6 well plates and allowed to reach 50% confluence the following day for siRNA transfection. Thirty pg/well of control siRNA or Osterix siRNA oligos mixture per well was added with the Lipofectamin 2000 reagent (Invitrogen, Carlsbad, CA) following the manufacturer's instructions. The blocking efficacy for the expression of human Osterix mRNA was measured at 2 days and 7 days post-transfection with real time PCR. Each experiment was repeated at least 3 times. The data is presented as the mean 6 SD. A t-test was performed, with p,0.05 considered statistically significant [3].

NELL-1 promoter contains multiple putative Sp1 sites (Sp1/Osterix binding sites)
In silico analysis of the human NELL-1 promoter identified a cluster of potential Sp1 sites (Sp1/Osterix binding sites) within approximately 70 bp of the 59 flanking region of the human NELL-1 gene (from 271 to 2142 bp upstream of the NELL-1 transcriptional start site) (www.gen matix.de). Within this cluster, Site A contains three overlapping potential Sp1 sites from 271 to 296 and Site B contains a single Sp1 site from 2133 to 2142 ( Fig. 1). In comparison to the human NELL-1 promoter, the mouse Nell-1 promoter also contains multiple putative Sp1 sites that reside at 2107, 2149, 2195, 2524, 2536 and 2971 bp upstream of the transcriptional start site. Similarly, a cluster of Sp1 sites was also identified by virtue of their proximity and overlapping consensus sequences between the fourth and fifth Sp1 sites (Fig. 1). To better illustrate the spatial distribution of Sp1 sites in both human and mouse Nell-1 promoters, the locations of previously identified Runx2 binding sites OSE2 are also displayed in Fig. 1, with three functional sites labeled A, B, C and a cryptic site H1 in the human NELL-1 promoter, and two sites labeled site 1 and 2 and a cryptic site m1 in the mouse promoter region. The presence of multiple Sp1 binding sites in the NELL-1 promoter across species of mouse and human is a good indication that NELL-1 transcription may also be directly regulated by Osterix in addition to Runx2.
Human NELL-1 promoter is responsive to Osterix by reporter assay To investigate whether the Sp1 sites contribute to NELL-1 transcription, we used the human NELL-1 promoter reporter systems 22213bp (p2213WT-Luc) and 2325bp (p325WT-Luc) as   [3]. These systems contain a 2213 bp or 325 bp length of genomic DNA upstream of the human NELL-1 transcription start site. We confirmed the presence of these putative Sp1 sites at predicated positions by full-length sequencing (Laragen, LA, CA) ( Fig. 1). Co-transfection experiments were performed to examine a dose-dependent response for both promoter constructs in Saos-2 cells. The Saos-2 cell line was chosen as a suitable system to study the human promoter and was previously used to study human NELL-1 promoter OSE2 sites [3]. Saos-2 cells have endogenous expression of both Osterix [17] and NELL-1, and have the necessary osteoblast-related factors for differentiation and formation of mineralized bone nodules [18,19]. Saos-2 cells transfected with the p2213WT-Luc and 0.25 mg to 1 mg Osterix expression vector (pOsx) had up to 80% decrease in luciferase activity in a dose-dependent manner compared to Saos-2 cells transfected with the p2213WT-Luc and control vector (pCtr) 2 days after transfection. Although the basal luciferase activity is reduced in the truncated p325WT-Luc construct group since it lacks 3 consensus OSE2 sites [3], the truncated construct with pOsx showed a similar percent decrease compared to control vector ( Fig. 2A), indicating that the short fragment containing these Sp1 sites in the p325WT-Luc construct are responsible for Osterix to down-regulation of luciferase activity.
Osterix forced expression decreases NELL-1 mRNA levels in human osteoblast-like and non-osteoblastic cells To investigate whether Osterix controls endogenous NELL-1 gene expression, Saos-2 cells were transiently transfected with either an empty vector or different amounts of Osterix expression constructs for 2 days and 7 days. Saos-2 cells were chosen for this analysis because we have previously shown that the high expression level of NELL-1 in these cells can induce several osteoblastic maker genes such as Opn and Ocn and promote their differentiation and mineralization [3]. The Osterix protein levels in Saos-2 cells transfected with different amounts of pOsx were detected by Western blot (Fig. 2B). In this study, the forced expression of Osterix in Saos-2 cells significantly decreased NELL-1 expression at 2 days post-transfection in a dose-dependent manner, and continued to reduce the levels for up to 7 days (Fig. 2B).
In addition to the Saos2 cells, the transcriptional repression of Nell-1 promoter by Osterix was also detected in other human cell lines including an immature osteoblast, U2OS cells and two nonosteoblastic cell lines, Hela and Glioma cells (Fig. 2C). Similar results in these cell lines compared to results in Saos2 cells further suggest that the transcriptional repression of NELL-1 promoter by Osterix exists in human cells irrespective of tissue origin or degree of osteoblastic maturity.
Sequence specific binding of Osterix to Sp1 sites within the NELL-1 promoter Since all these Sp1 sites lie within the 325bp promoter of the proximal NELL-1 transcriptional start site, to determine the functional relevance of all Sp1 sites in Osterix-mediated decrease of NELL-1 promoter activity, we generated a mutant promoter construct (p325mut all-Luc, Genscript Co, Piscataway, NJ) containing an alteration of the Sp1 sites by point mutation known to disrupt Osterix binding [9] (Fig. 3A). This mutant construct, p325mut all-Luc with mutations in all Sp1 sites of both cluster Site A and Site B, was transfected into Saos-2 cells and the downstream reporter gene luciferase activity was analyzed with and without forced Osterix expression. The Osterix-induced suppression of luciferase activity was statistically significant in the wild type construct p325WT-Luc (p,0.05) (Fig 3B). Furthermore, the complete suppression of Osx inhibitory effect was observed in the p325mut all-Luc construct as compared to p325WT-Luc in the setting of Osterix overexpression (p,0.05). This result strongly indicates that these Sp1 sites of the Nell-1 promoter are required for Osx binding in regulating NELL-1's transcription.
To determine which Sp1 site is more important to induce the suppression, we made two additional mutant reporter constructs, p325mutSiteA-Luc with mutations in cluster Site A, and p325mutSiteB with mutation in Site B. Notably, the suppression of luciferase activity by expression of Osterix was still observed when either p325mutSiteA or p325mutSiteB constructs were used. However, the levels of Osterix overexpression-mediated suppression in these constructs were significantly diminished in comparison to p325WT construct (p,0.05) (Fig. 3B). These results indicate that both Site A and Site B have functional roles in the suppression of NELL-1 when bound by Osterix, and both are absolutely necessary and responsible for the complete suppression of Osx inhibitory effect on NELL-1's transcription when they are mutated simultaneously (p,0.05).
To further examine DNA-protein interactions at these Sp1 sites, EMSAs using Saos-2 nuclear extracts and the respective oligonucleotide probes containing the Sp1 sites were performed. Since these Sp1 sites were within 70 bp in the NELL-1 proximal promoter, we divided them into two respective oligonucleotide probes in our experiment-Site A (from 2101 to 262 bp upstream of NELL-1) containing three overlapping potential Sp1 sites and Site B (from 2153 to 2124 bp upstream of NELL-1 transcriptional start site) containing a single Sp1 site.
Since Tomohiro reported that Sp1/Sp3 can also occupy the Sp1 sites in Saos-2 cells [17], we force expressed Osterix in Saos-2 cells to increase its amount in the nuclear extract (NE) for gel shift assay. We were then able to find the main protein-DNA (Osterix-Site A and Osterix-Site B) complex bands in the two groups with and without Osterix overexpression (Fig. 4A). These protein-DNA complex bands were confirmed by super shift assay when Osterix antibody was added (Fig. 4A).
Furthermore, specificity of binding to the Sp1 site was confirmed by competition analysis with WT probes and by mutant probes containing 2 bp mutations known to disrupt Osterix binding described in mutant reporter systems. Incubation of NE from Saos-2 cells transfected with pOsx led to the formation of protein-DNA complexes with WT oligonucleotides. In contrast, the oligonucleotides harboring the 2 bp mutations had no protein-DNA complexes detected (Fig. 4B). Binding of the promoter sites by labeled WT probe was diminished with the addition of 20x and 200x unlabeled WT oligonucleotide, but no effect was seen with the addition of 200x unlabeled mutated oligonucleotides (Fig. 4B). These results demonstrate that Osterix is indeed part of the DNAprotein complex in the NELL-1 promoter and indicate that NE from Saos-2 cells has Osterix that can specifically bind to the Sp1 sites of the human NELL-1 promoter.

Osterix binding to the endogenous NELL-1 promoter in vivo
Since our reporter assay and EMSA studies described above indicate that Osterix protein recognized these Sp1 sites in the human NELL-1 promoter, we next examined whether Osterix is able to interact directly with the NELL-1 promoter in vivo. To study this we used ChIP-qPCR assay (ChampionChIP One-Day kit, SABiosciences, Frederick, MD) to examine the binding of Osterix to the NELL-1 promoter region in Saos-2 cells. The qPCR products between anti-Osterix and control IgG showed a significant difference in the NELL-1 promoter 21 kb group, but the difference was not seen in the NELL-1 promoter 22 kb group (Fig. 4C), suggesting that Osterix can directly interact with NELL-1 promoter in vivo and specifically bind to the chromatin fragment containing these Sp1 sites region covered by specific NELL-1 promoter 21 kb qPCR primer. To further confirm this finding we also performed the ChIP-qPCR assay in primary human osteoblast cells. The results indicated the existence of similar recessive regulatory relationship of Osterix and Nell-1 through direct binding of Osterix to Sp1 sites of Nell-1 promoter (Fig. 4D). These results further verify that endogenous Osterix both in Saos-2 cells and primary human osteoblast specifically associate with these Sp1 sites in the endogenous human NELL-1 promoter, and that Osterix is directly involved in the regulation of NELL-1 gene activity through its physical interaction with the NELL-1 promoter.
Osterix down regulates NELL-1 expression without disruption of Runx2 binding to NELL-1 promoter OSE2 sites Because our previous study showed that Runx2 positively regulates NELL-1 expression through three consensus and one cryptic OSE2 site (H1) in the NELL-1 promoter, here we cotransfected Osterix and Runx2 with p2213-Luc or p325-Luc constructs to detect Osterix effects on Runx2 induced NELL-1 promoter activity. The reporter systems were co-transfected with pRunx2 and reporter constructs plus pOsx or pCtr. The luciferase activity after treatment with pOsx was significantly reduced compared to treatment with pCtr in both p2213WT-Luc and p325WT-Luc groups (Fig. 5A). This result suggests that high level Osterix can suppress Runx2 induced NELL-1 expression. In order to determine whether this suppression is mediated through these Sp1 sites, we used the construct (p325mut all-Luc) with all mutated Sp1 sites disrupting Osterix binding (Fig. 3A). As expected, in the p325mut all-Luc group, there was no difference in luciferase activity between pOsx and pCtr (Fig. 5A) indicating that the suppression of Runx2 induced NELL-1 expression by Osterix requires functional Sp1 sites.
Our previous NELL-1 promoter analysis also showed that these Sp1 sites (271,2142 bp) are located proximal to the Runx2 OSE2 binding site (H1, 2247 bp). It is possible that Osterix down regulation of NELL-1 promoter activity is mediated by suppression of Runx2 binding to the H1 site. Therefore, ChIP-qPCR assay  . Osterix specifically binds to Sp1 sites in the NELL-1 promoter in vitro and in vivo. (A) EMSA of Saos-2 nuclear proteins transfected with pCtr or pOsx binding to SiteA (containing three proximal Sp1 sites) and SiteB probes. Supershifts with specific Osterix antibody indicate the specific Osterix-DNA complexes. (B) EMSA depicting primary human osteoblast cell (hOB) nuclear proteins transfected with pOsx binding to the SiteA probes, with competition by 20x and 200x unlabeled SiteA and 200x unlabeled mutated SiteA (MutA) oligonucleotides. Note that the MutA probes failed to bind nuclear proteins. The same pattern is also seen by competition of SiteB element and MutB. (C-D) Osterix binds to endogenous Sp1 sites of the human NELL-1 promoter in Saos-2 (C) and hOB (D) cells. The NELL-1 -1 kb primer set covers 1 kb proximal promoter region containing all Sp1/ Osterix binding sites and three OSE2 sites. The NELL-1 -2 kb primer set covers NELL-1 promoter region from 21 kb to 22 kb where no Sp1/Osterix binding site but one OSE2 site exists. The qPCR products depict DNA amplified from Chromatin Immunoprecipitation with cells utilizing Control IgG and Osterix antibody. Input DNA represents positive genomic DNA control. doi:10.1371/journal.pone.0024638.g004 was used to detect binding between Runx2 and NELL-1 promoter with and without Osterix forced expression. The same amount of chromatin was used for ChIP assay plus control IgG, Osterix antibody, Runx2 antibody and general transcriptional factor RNA polymerase II antibody. ChIP-qPCR products were normalized by endogenous GAPDH amounts between Osterix transfection and control vector groups. The results showed that Osterix binding to NELL-1 promoter was significantly increased in the Osterix forced expression group compared to control vector group. There was no obvious difference seen in Runx2 binding to NELL-1 promoter with and without Osterix forced expression (Fig. 5B). Interestingly, the general transcription factor RNA polymerase II binding to NELL-1 promoter was significantly decreased in the Osterix overexpression group (Fig. 5B), indicating one possible mechanism for Osterix negative regulation of NELL-1 promoter activity.

Osterix siRNA increases NELL-1 mRNA levels in Saos-2 cells concomitantly with increased mineralization
The data showed that Osterix forced expression decreases NELL-1 mRNA levels in Saos-2 cells (Fig. 2B). To further demonstrate the effect of suppression, we also analyzed other osteoblastic marker mRNA levels after Osterix overexpression in Saos-2 cells and primary human osteoblasts. Interestingly, some markers such as Ocn and Opn expression levels also decreased following the decrease of NELL-1 expression at 2 days posttransfection (Fig. 6A). However, by 7 days post-transfection, Ocn and Opn expression levels showed no significant difference between the pCtr and pOsx groups in Saos-2 cells. Moreover, Ocn expression level also decreased in a similar fashion as Nell-1 at 2 days post-transfection in primary human osteoblasts (Fig. 6B), but Opn expression patterns were different between Saos-2 osteosarcoma cells and normal primary human osteoblast cells, which may indicate that overexpression of Osterix plays a transient and more complicated role with variable effects on bone marker gene levels at different stages of maturation of human osteoblasts.
To further confirm Osterix suppression of NELL-1 expression, we inhibited Osterix mRNA level using siRNA in Saos-2 cells and primary human osteoblasts. Data showed that NELL-1 mRNA levels increased almost 3 fold 2 days after Osterix siRNA transfection at which time Osterix mRNA expression levels were decreased by 80% in Saos-2 cells (Fig. 6A). Ocn and Opn expression also increased slightly 2 days after transfection. At posttransfection day 7, when Osterix mRNA levels were still less than 30%, NELL-1 mRNA levels continued to be elevated. NELL-1 and Ocn mRNA levels also increased in a similar pattern at 7 days posttransfection (Fig. 6A). To further confirm Osterix regulation of NELL-1 in mature osteoblast cells, these experiments were performed in human primary osteoblasts. Although the inhibition efficiency of Osx-siRNAs (around 50%) in this cell line is less than that in Saos-2 cells at Day 2, NELL-1 mRNA levels showed significant increase along with significant changes in other bone markers (Ocn and Opn) after 7 days post Osterix siRNA transfection (Fig. 6B). Alizarin Red staining was also used to detect mineralization during osteoblast differentiation. Osterix siRNA transfection increased the mineralization of Saos-2 cells at 9 days posttransfection (data not shown), consistent with bone marker gene mRNA level increase in Osterix siRNA assay.

Discussion
NELL-1 is a novel osteoinductive factor under direct transcriptional regulation of Runx2 [3,5], the master transcription factor of osteogenesis [20]. Osterix is another essential transcription factor for osteoblast differentiation and bone formation directly down-stream of Runx2 [12]. In this study we sought to determine the regulatory and functional relationship between these two downstream targets of Runx2, in particular to validate the functional characteristics of potential Osterix binding sites in the human NELL-1 promoter revealed by in silico analysis. Our data showed that Osterix exhibits repressive instead of assumed inductive effect on NELL-1 expression at the transcriptional level by binding directly to Sp1 sites in the NELL-1 promoter region; a surprising finding given the fact that NELL-1 and Osterix are both considered pro-osteogenic factors [5,9,10,16,21]. This adds NELL-1 as a member of Osterix regulated molecules that include Col 1a [9], Col 11a2 [17], DKK1 [22] and IL-1a [23]. Like IL1-a, NELL-1 is also negatively regulated by Osterix. In addition, we also found that the Sp1 binding elements in the human NELL-1 promoter, identified as two clusters, Site A and B, have similar capacity to be fully occupied by Osterix to mediate repression. The release of this repression can occur only when Site A and B are mutated simultaneously.
The definitive mechanisms underlying the activating or inhibitory effects of Osterix on target promoters of these molecules remain unclear. Interestingly, basic transcription element B1 (BTEB1), a Sp1-like protein, has been found to activate transcription on promoters containing multiple GC boxes but act as a repressor on promoters containing only a single GC box [24]. This differential effect on multiple versus single GC box in gene promoters also applies to Osterix direct targets including activation of Col 11a2 and DKK1 which both have multiple binding sites, and repression of IL-1a which has a single binding site. However, this rule does not apply to all targets of Osterix, as Col 1a which has a single binding site is activated, not repressed, by Osterix, while Nell-1 with multiple sites is repressed. Col 1a regulation is more complex, as its regulation has been reported to also involve NFATc1 as a co-factor that forms a complex with Osterix to bind the consensus Sp1 binding site [10]. It is possible that NFATc1 may modulate Osterix-mediated transactivation by recruitment of other transcriptional co-activators [10]. Most recently, another co-factor of Osterix, NO66, a Jumonji family histone demethylase, has been reported to impair transcriptional activation of Osterix through interaction with the Osterix activation domain. In particular, the interaction between Osterix and NO66 is believed to regulate Osterix target genes in osteoblasts through modulating histone methylation [25]. Osterix transcriptional repression of Nell-1, a gene expressed preferentially in osteoblasts, may therefore also involve a co-factor leading to the negative effect on NELL-1 promoter activity.
Runx2 is known as the master regulator of osteochondrogenesis, promoting commitment, clonal expansion, and early osteoblastic differentiation [26,27], and is a direct upstream regulator of NELL-1 gene expression [3]. Our previous studies have demonstrated that Runx2 directly activates NELL-1 transcription by physically binding to OSE2 sites on its promoter region [3]. In this current study, reporter system assays confirmed that Osterix directly represses Runx2-induced NELL-1 expression through binding of multiple Sp1 sites on its promoter. Mechanistically, by using CHIP-qPCR assay, we were able to demonstrate that there was no difference in Runx2 binding of NELL-1 promoter OSE2 sites with and without Osterix forced expression. This demonstrates that Osterix-mediated down-regulation of NELL-1 expression does not involve disruption of Runx2 binding of the NELL-1 promoter OSE2 sites. Instead, we found that general transcription factor RNA polymerase II binding to the NELL-1 promoter is significantly decreased when Osterix is overexpressed, which may interfere with initiation of NELL-1 gene transcription [28]. However, the exact role Osterix plays, along with RNA polymerase II, in the negative regulation of NELL-1 with and without Runx2 induction remains unclear and warrants further study. Notably, there has been no evidence to date that Osterix and Runx2 interact with each other directly to alter their DNA binding and promoter transactivating activities [26].
To determine how Osterix repressive transcriptional regulation of NELL-1 affects its osteogenic activity, we performed in vitro osteoblastic differentiation studies with either overexpression or specific siRNA knockdown of Osterix in Saos2 as well as in normal primary human osteoblast cells. Expectedly, the mRNA expression of NELL-1 was severely inhibited by overexpression of Osterix. Notably, NELL-1 repression was associated with the early transient decrease of Ocn and Opn mRNA indicating some level of impairment of NELL-1 osteoinductive capacity. In line with these findings, over two fold upregulation of NELL-1 mRNA along with increase of Opn at the early phase, and increase of Ocn and mineralization at the late stage of osteoblastic differentiation were observed after Osterix knock down by specific siRNA. Interestingly, the different pattern of Opn expression between Saos-2 osteosarcoma cells and normal primary human osteoblast cells suggests a more complicated role for Osterix in osteoblastic differentiation at different maturation stages of human osteoblasts. Taken together, these data definitively demonstrate the functional impact and significance of Osterix repression of NELL-1. Furthermore, the forced expression of NELL-1 remarkably reduced Osterix mRNA levels in Saos-2 cells (data not shown), demonstrating reciprocal repression of Osterix by NELL-1. This further confirmed our previous study on MC3T3 cells that showed transduction of AdNELL-1 inhibited Osterix mRNA expression without affecting Runx2 mRNA levels [5,16].
The repressive regulation of NELL-1 by Osterix may seem paradoxical given that both are known to be pro-osteoblastic, with many reports having shown that Osterix and NELL-1 can positively regulate osteoblast differentiation [5][6][7][9][10][11]. However, in reality, this is not uncommon. For instance, Osterix, a proosteogenic regulator, negatively regulates the Wnt signaling pathway which is known to play a crucial role in the control of bone mass [22]. Osterix inhibits the Wnt signaling pathway through several mechanisms, including binding to and activating the Wnt antagonist DKK1 promoter, or interrupting TCF binding to its DNA elements and then suppressing downstream b-catenin activity [22].
Studies on the inter-relationship among various factors involved in the transcriptional regulatory network of osteogenesis are few in number and provide only limited answers likely owing to the high complexity of this area of study. What is known is that NELL-1 is a critical component in regulating osteoblastic differentiation, and that both Runx2 and Osterix are involved in its transcriptional regulation and osteogenic function. Runx2, a positive regulator of NELL-1, is highly expressed during transition from mesenchymal cells to preosteoblasts and immature osteoblasts [26]. NELL-1 may be an effector of a large portion of Runx29s role, as it is a key downstream functional mediator in this process [7]. Osterix negative regulation of NELL-1, which is also tightly regulated by Runx2, may result from a delicate balancing of various driving forces in this regulatory network, modulating NELL-1 expression levels as needed at different developmental time points. Moreover, the overexpression of NELL-1 also affects Runx2 expression levels or bioactivity reciprocally, adding to the complexity of the regulatory network (Fig. 6C).
We expect that the regulatory relationship between NELL-1 and Osterix presented here from our in vitro studies is likely also true in vivo. Our preliminary studies revealed a higher level of Osterix expression with less mineralization in neonatal calvarial bones in ENU-induced Nell-1 deficient mice compared to that of wild type mice suggesting existence of a similar regulatory relationship between Nell-1 and Osterix in vivo (data not shown). More extensive investigation of mouse skeletal development is needed for conclusive results, and is an area of future study for us.
Collectively, our findings through investigation of Osterix transcriptional regulation of and functional impact on NELL-1 represent further understanding of the complex regulatory network that governs osteochondrogenesis.