BMP-2 induced Dspp transcription is mediated by Dlx3/Osx signaling pathway in odontoblasts

Dentin sialophosphoprotein (Dspp) as a differentiation marker of odontoblasts is regulated by BMP-2. However, the intimate mechanism is still unknown. Transcription factors Dlx3 and Osx are essential for odontoblasts differentiation. We hypothesized that BMP-2 regulation of Dspp transcription was mediated by Dlx3 and/or Osx in odontoblasts. In the present investigation, we found that BMP-2 stimulated expression and nuclear translocation of Dlx3 and Osx in odontoblasts both in vitro and in vivo. Osx was a downstream target of Dlx3 and both of them stimulated Dsp expression. Both Dlx3 and Osx were able to activate Dspp promoter from nucleotides (nt) −318 to +54 by transfections of luciferase reports containing different lengths of mouse Dspp promoters. The binding of Dlx3 and Osx with nt −318 to +54 of Dspp promoter was verified by chromatin immunoprecipitation in vivo. Two Dlx3 binding sites and one Osx binding site on Dspp promoter were found by EMSA. Furthermore, the exact biological function of these binding sites was confirmed by site-directed mutagenesis. At last, the protein-protein interaction between Dlx3 and Osx in odontoblasts was detected by co-immunoprecipitation. In conclusion, in this study we found a novel signaling pathway in which BMP-2 activates Dspp gene transcription via Dlx3/Osx pathway.

by BMP signaling, especially BMP-2, in vitro and in vivo 3,21,22 . Beads socked in human recombinant BMP-2 induce the mRNA expression of Dspp, after implantation onto dental papilla in organ culture 23 . Dspp expression was reduced in the Bmp-2 knock out mice 16 . Previously we also have found that BMP-2 up-regulates Dspp transcription through its regulatory region in mouse preodontoblast cells 3 . However, the intimate molecular mechanisms of BMP-2 regulating Dspp transcription in preodontoblasts or odontoblasts have not been completely understood.
Transcription factors Dlx3 and Osx (or Sp7, Osterix) are essential for osteoblast and odontoblast differentiation [24][25][26] . Both Dlx3 and Osx are suggested as downstream targets of BMP-2 signaling in osteogenic cells 27,28 . Although both odontoblasts and osteoblasts originate from mesenchymal cells and share many physical similarities, the molecular mechanisms regulating odontoblasts differentiation are different from osteoblasts. For example, BMPs are able to stimulate expression of Runx2 in osteoblastic cells but not in odontoblasts 2,29 . Whether Dlx3 and Osx mediate the BMP-2 induced Dspp expression in odontoblasts is still not known. During tooth morphogenesis, Dlx3 mRNA is initially expressed in the dental epithelium, and is later expressed in both the dental epithelium and the dental mesenchyme 25 . Mutations of DLX3 in human cause Tricho-Dento-Osseous (TDO) syndrome, which is an autosomal dominant disorder characterized by defects in ectodermal derivatives such as hair (kinky hair), teeth (enamel hypoplasia and taurodontism) and bone (increased bone density in cranium and long bones) [30][31][32] . Deletion of Dlx3 in neural crest, from which dental mesenchyme is derived, leads to severe dentin hypoplasia and dysplasia and remarkably down-regulated Dspp expression 33 . During tooth development, Osx expression was initiated at dental papilla, and remained highly expressed in differentiating odontoblasts at later stages 21 . Previously, we found overexpression of Dlx3 or Osx in odontogenic cells induced cell differentiation and Dspp expression 34,35 ,while knock-down of Osx caused the down-regulation of Dsp expression 36 .
Based on these biological functions of Dlx3 and Osx during odontoblasts differentiation and tooth development, we hypothesized that BMP-2 regulation of Dspp transcription was mediated by Dlx3 and/or Osx signaling pathway in odontoblasts.

BMP-2 stimulates expression and nuclear translocation of both Dlx3 and Osx in odontoblast cells.
We first detected the expressions of Dlx3 and Osx in odontoblasts of first molars in Bmp2-cKO°d mice.
With the knock-out of Bmp2 gene in odontoblasts (showed by in situ hybridization in Supplementary Fig. S1), the odontoblasts of first molars in Bmp2-cKO°d mice lost their differentiation and polarization, and accompanied with dramatically reduced expressions of Dlx3 and Osx compared with wild type mice, in which high expression of Dlx3 and Osx was seen in the nuclei of odontoblasts (Fig. 1A). Then we examined whether BMP-2 is able to induce expressions of Dlx3 and Osx in mouse pre-odontoblasts (MD10-F2). Our previous study indicated that BMP-2 (100 ng/ml) was sufficient to induce Dspp expression in MD10-F2 cells 3 . Therefore, a final concentration of 100 ng/ml of BMP-2 recombinant protein was added in the medium to stimulate MD10-F2 cells, and expressions of Dlx3, Osx, and Dsp were measured by Western blot. As illustrated in Fig. 1B, expression of Dlx3 was increased and reached the maximum level at 2-4 hours after BMP-2 treatment. Meanwhile, the expressions of Osx and Dsp were also up-regulated and reached the maximum level at 4-6 hours after BMP-2 treatment. All these in vivo and in vitro data indicated both Dlx3 and Osx are downstream targets of BMP-2 in odontoblasts. As transcription factors, Dlx3 and Osx, activated by BMP-2 should paly functions in nucleus. Then we detected the subcellular localization of Dlx3 and Osx after BMP-2 induction using immunofluorescence (Fig. 1C,D). In untreated cells, Dlx3 and Osx staining appeared mainly in the cytoplasm and weakly in nucleus. After induction with BMP-2, the staining of Dlx3 and Osx mainly localized into nucleus, compared to that in the untreated cells. These results suggested that BMP-2 not only stimulated expression of Dlx3, Osx and Dsp, but also induced nuclear translocation of Dlx3 and Osx in vitro and in vivo.
Osx is a downstream target of Dlx3 and both of them stimulate Dsp expression in MD10-F2 cells. It was reported that Osx is an explicit downstream target of Dlx5 in osteoblast-lineage cells 37 . Although Dlx5 and Dlx3 belong to the Distaless (Dlx) gene family, it is yet not known if Dlx3 is able to regulate Osx in odontoblasts. We therefore tested the relationship between Dlx3 and Osx in MD10-F2 cells. The results showed that overexpression of Dlx3 was able to induce Osx and Dsp expression ( Fig. 2A), whereas overexpression of Osx only induced Dsp expression but did not influence the expression of Dlx3 (Fig. 2B). To further confirm these results, we knocked down the endogenous Dlx3 expression using Dlx3-targeted siRNA (Fig. 2C). Transfection of Dlx3 siRNA for 48 hours reduced Dlx3 expression more than 50% in MD10-F2 cells. At the same time, Osx and Dsp expression was reduced to ~40% and ~90%, respectively, compared with their expression in cells transfection with control siRNA. Thus, Osx was a downstream target of Dlx3, and both of them were able to stimulate Dsp expression in MD10-F2 cells.

Dlx3 and Osx stimulate Dspp promoter activity. To confirm whether induction of Dsp expression by
Dlx3 and Osx is through stimulating Dspp promoter activity, luciferase reporters containing different lengths of mouse Dspp promoters, p1318, p591, and p318, were constructed (Fig. 3A) and transiently transfected into MD10-F2 cells. Transcription activity was measured when transfected with pcDNA-Dlx3, or pcDNA-Osx-Flag, or pcDNA3.1 empty vector (Fig. 3B). Both Dlx3 and Osx failed to change the transcription activity of p1318, p591, and pGL3-Basic, compared with transfection with pcDNA 3.1 empty vector. However, overexpression of Dlx3 or Osx significantly increased the transcription activity of p318, which indicated both Dlx3 and Osx were able to stimulate Dspp promoter activity from nucleotides (nt) −318 to +54.
To further detect whether Dlx3 and Osx bind with the nt −318 to +54 of Dspp promoter, we performed ChIP assay. The immunoprecipitated and purified DNA was used as a template. The PCR amplifying bands with primers corresponding to the 5′-flanking region (from nt −318 to +54) of mouse Dspp gene were detected, and the band is more intense in cells transfected with pcDNA-Dlx3 or pcDNA-Osx-Flag plasmids (Fig. 3C, lanes 1-4, upper). Positive control was done using anti-TFIIB antibody to amplify GAPDH promoter (Fig. 3C, lane 5), and negative control was set using negative control IgG (Fig. 3C, lane 6). These results indicated that Dlx3 and Osx are able to directly bind to Dspp promoter region from nt −318 to +54 in vivo.

Identification of Dlx3 and Osx binding sites in the mouse Dspp promoter. It is well known Dlx3
binds to a conserved sequence of TAATT 38 , and Osx recognizes GC-rich sequence 39 . Analysis of the 5′-flanking region from nt −318 to +54 of the mouse Dspp gene showed two putative Dlx3 binding sites (Dlx3I and Dlx3II) and one putative Osx binding sites using computer software program and these sequences are highly conserved in several species, including mouse, human and rat (Fig. 4A). First, we examined whether Dlx3 and Osx could bind to their binding sites in the Dspp promoter using EMSA. The following oligonucleotide probes were pre-    Dlx3 and Osx depend on each other to increase Dspp promoter activity. To further confirm the functional effect of Dlx3 and Osx binding sites in the mouse Dspp promoter, we generated three p318 mutant constructs with deletion of Dlx3 or Osx binding sites ( Fig. 5A and B): p318(Dlx3I Del), p318(Dlx3II Del), and p318(Osx Del). Then p318 or p318 mutant constructs were transiently transfected into MD10F-2 cells, and their transcription activity was evaluated with the co-transfection of pcDNA-Dlx3 and/or pcDNA-Osx-Flag. The results showed that deletion of either Dlx3I or Osx binding site suppressed the p318 response to both Dlx3 and Osx, but deletion of Dlx3II binding site only suppressed the p318 response to Dlx3 but not to Osx (Fig. 5C). These results  (Table 1) were used to incubate with recombinant Dlx3 protein (B) or nuclear extracts from MD10-F2 cells (C). The black arrows indicate protein-DNA complexes and the white arrows indicate antibody-protein-DNA complexes (Supershift bands).

Name
Sequence (5′ to 3′) Dlx3 binds to Osx forming a protein complex. The result that Dlx3I element is necessary for Osx activating p318 transcription activity implied that Dlx3 may bind to Osx forming a complex to regulate Dspp transcription. Then, the protein-protein interaction between Dlx3 and Osx was detected with co-immunoprecipitation. As anticipated, Dlx3 was detected in the immunoprecipitation complex by Western blot with anti-Osx antibody as IP antibody (Fig. 5D, left penal). Meanwhile, Osx was also detected in the immunoprecipitation complex with anti-Dlx3 antibody as IP antibody (Fig. 5D, right penal), which indicated Dlx3 bound to Osx forming protein complex.

Discussion
Odontoblast differentiation is regulated by many transcription and growth factors. BMP-2 signaling induces dental mesenchymal cell differentiation into odontoblasts 40 . Dspp has been characterized as a unique marker of odontoblasts differentiation 1, 2 . It has been reported that BMP-2 is able to induce Dspp expression, but up to now, the molecular mechanisms by which BMP-2 induces Dspp transcription have not been well elucidated. The present investigation reveals a new signaling pathway, which participates in the regulation of Dspp expression by BMP-2 in odontoblasts. Dlx3 contains a homeodomain, which is related to the distal-less domain of Drosophila and was detected even in structures involving epithelial-mesenchymal interaction, such as tooth germs and hair follicles. BMP-2 was able to induce Dlx3 gene expression in several tissue cells, including osteoblasts 27 , keratinocytes 41 , and dental follicle cells 42 . In addition to Dlx3, two other members of the distal-less family, Dlx2 and Dlx5, are also regulated by BMP-2 in osteogenic or chondrogenic cells 27,43 . Osx as a transcription factor was first identified in C2C12 myogenic cells treated with BMP-2 39 . Besides,it has also been reported that Osx was induced with BMP-2 in mouse progenitor cells and chondrocytes 39,44 , human and mouse mesenchymal stem cells 45,46 . Consistent with these previous investigations, we also demonstrated that Dlx3 and Osx were BMP-2-inducable in odontoblasts both in vitro and in vivo. Previously, an Osx-GFP fusing protein reporter system was constructed to track Osx translocation in osteogenic cells 47 . Osx remained mostly inactive in the cytosol of non-osteogenic cells and was activated and translocated into the nucleus with the osteogenic differentiation of osteogenic cells 47 . In the present The MD10-F2 cells were co-transfected with wild type or mutant p318 luciferase gene constructs with pcDNA-Dlx3 or/and pcDNA-Osx-Flag expression plasmids. Luciferase activity was determined, and the value obtained from the wild type p318 group was taken as 1-fold. *P < 0.05; **P < 0.01; compared with wild type p318 group. (D) Anti-Osx, anti-Dlx3 antibodies and negative control IgG were used to pull down by co-immunoprecipitation assay. Anti-Dlx3 or anti-Osx antibody was used for western blotting to confirm the presence of Dlx3 or Osx in the complex. Neg.: Negative control IgG; W.B.: Western blot. Uncropped images of blots were shown in Supplementary Fig. S5. investigation, as showed by immunofluorescence staining, we found both Dlx3 and Osx were activated and translocated into nucleus from cytoplasm with addition of BMP-2 in vitro and in vivo.
Although Osx mRNA expression is stimulated by BMP-2 treatment, pretreatment with cycloheximide, a protein synthesis inhibitor, blocks the BMP-2-induced expression of Osx mRNA 37 , which indicates that Osx is not the direct target of the BMP signaling cascade and the expression of Osx induced by BMP-2 requires the intermediation of newly synthesized proteins. In contrast to Osx expression induced by BMP-2, BMP-2-induced Dlx5 expression is unaffected by cycloheximide pretreatment 48,49 , which means Dlx5 is a direct target of BMP signaling. Moreover, inhibition of Dlx5 expression blocked Osx expression 50 , indicating Osx is down stream target of Dlx5. In the present investigation, we found the similar results: overexpression of Dlx3 was able to induce Osx expression, but overexpression of Osx did not influence the expression of Dlx3. Furthermore, down-regulated Osx expression in cell with transfection of Dlx3-targeted siRNA further confirmed Osx is a downstream target of Dlx3 in odontoblasts. The fact that the time of Dlx3 to reach the maximum level after BMP-2 treatment is earlier than that of Osx indicates that Osx is also not the direct target of the BMP-2 in odontoblasts.
Both positive and negative regulatory mechanisms are required for the spatial and temporal expression of Dspp gene. Many investigations characterized mouse and rat Dspp gene promoters and found inverted TATA and CAATT box sequences, Sp1, Nrf1, C/EBP and Runx2 binding sites, as well as several homeodomain (Dlx and Msx) motifs in the proximal regions 8, 51-53 . Previously, we found BMP-2 regulates Dspp expression through the activation of the heterotrimeric transcription factor NF-Y, and NF-Y binds to a BMP-2 response element in the mouse Dspp promoter, especially between nt −97 and −72 3 . In the present investigation, we found that in addition to NF-Y, BMP-2 also regulates Dspp expression also through the Dlx3 and Osx binding sites on Dspp promoter. Two Dlx3 binding sites (DLX3I, DLX3II) and one Osx binding site were identified in the proximal mouse Dspp promoter that mediates BMP-2-stimulated Dspp expression. EMSA and ChIP analyses verified that Dlx3 and Osx bind to their target sites in the mouse Dspp promoter. To further evaluated the function of these binding sequences, direct deletion in the Dspp promoter was used, and showed that deletion of either Dlx3 or Osx binding site decreased Dspp promoter activity in MD10-F2 cells, which means each Dlx3 or Osx binding site in the Dspp promoter was functional. Duverger et al. 33 demonstrated the direct binding of Dlx3 to Dspp promoter, which is in consistent with our findings. In distal-less family, Dlx5 also mediates Dspp expression regulated by BMP signal 22 . Although both Dlx3 and Dlx5 can bind to TAAT box, but the two Dlx3 binding sites found in the present research only respond to Dlx3 but not to Dlx5 22 , which indicates Dlx3 and Dlx5, as distal-less family members, show different mechanism in regulation of Dspp transcription. Both Dlx3 and Osx can increase the promoter activity of p318, but not p591 and p1318. One possibility is that there are Dlx3 and Osx binding sites at upstream of nt −318 on Dspp promoter, but Dlx3 and Osx act as suppressors when binding to these binding sites. Another possibility is that there are strong negative regulatory elements for the binding of other transcription factors at upstream of nt −318 on Dspp promoter, and Dlx3 and Osx as co-factors could increase the suppressing activity of these elements. Cao et al. 22 found that deletion of the region from −791 to −427 strongly increased basal Dspp promoter activity, and inferred there might be a strong negative regulatory element between nt −794 and −427. Narayanan et al. 8 also identified a repression domain between nt −700 and −400 of Dspp promoter. Further analysis of this region identified two binding elements, which have 90% homology to the DNA binding sites for two bZIP transcription factors, Nrf1 and the CCAAT enhancer-binding protein (C/EBP) β. A possible response element for Yy1, a negative transcription factor, was also found (at position of nt −622) in this repression region of Dspp promoter, and the basal promoter activity of Dspp increased after the Yy1 binding sites was deleted 22 .
By means of co-immunoprecipitation assay we have clearly demonstrated that Dlx3 associates and functionally cooperates with Osx during regulation of Dspp transcription. Mutational studies suggest that the Dlx3 and Osx depend on each other to increase Dspp promoter activity. Dlx3 could not activate the promoter activity by binding itself to the Dspp promoter containing the mutated Osx binding site. All these results indicate the bindings of Dlx3 and Osx to their binding sites on Dspp promoter are necessary for each of these two transcription factors to activate Dspp promoter.
In conclusion, in this study we provide the evidence that BMP-2 activates Dspp gene transcription via Dlx3/ Osx signaling pathway: BMP-2 interacting with its receptor on cell membrane induces Dlx3 expression, then Dlx3 induces Osx expression; subsequently, both Dlx3 and Osx are activated, enter nucleus and bind to their target sites in Dspp promoter; then Dlx3 interacts with Osx and synergistically stimulate Dspp transcription.

Methods
Animals and Tissue Preparation. All experimental procedures involving the use of animals were reviewed and approved by the Institutional Animal Care at the University of Texas Health Science Center at San Antonio. All experiments were performed in accordance with the relevant guidelines and regulations. The generations of Bmp2 f/f and 3.6Col1a1-Cre mice have been described previously 54,55 . To conditionally knock out Bmp2 gene in odontoblasts, 3.6Col1a1-Cre;Bmp2 f/f (referred to as Bmp2-cKO od ) mice were generated by crossing 3.6Col1a1-Cre;Bmp2 f/+ allele with Bmp2 f/f allele. Mice at postnatal day 3 (P3) were put to death. Mandibles of the mice were immediately dissected and fixed with 4% paraformaldehyde (PFA) for 24 hours at 4 °C. After washes with PBS, samples were demineralized in 8% EDTA for 2 weeks. Then, the tissues were processed for paraffin embedding and sectioned at 5 µm.
Cell Culture and Treatment. A mouse immortalized preodontoblast cell line, MD10-F2 was used 56 .
Cells were grown at 33 °C under 5% CO 2 in alpha minimum essential medium (α-MEM) supplemented with 10% fetal bovine serum, 100 units/ml penicillin/streptomycin, 50 µg/ml ascorbic acid and 10 mM sodium β-glycerophosphate. For BMP-2 induction, cells were starved overnight in FBS-free medium, then cells were induced with 100 ng/ml of recombinant human BMP-2 protein (R&D) for the indicated times. Then the cells were subjected to Western blot or immunofluorescence staining.
Scientific RepoRts | 7: 10775 | DOI:10.1038/s41598-017-10908-8 Immunohistochemistry and Immunofluorescence Staining. The expressions of Dlx3 and Osx in mandibular first molars of both wild type and Bmp2-cKO od mice were analyzed by immunohistochemistry as described previously 57 . MD10-F2 cells cultured on glass slides were fixed with cold methanol/acetone (1:1) for 10 minutes and permeabilized with 0.5% Triton-X for 15 minutes. To block the non-specific binding of antibodies, slides were incubated with 10% goat serum for 30 minutes, followed by primary antibodies, anti-Dlx3 (Abcam) and anti-Osx (Santa Cruz), overnight at 4 o C. After washing with PBS, slides were incubated with secondary IgG antibodies conjugated to Alexa-Fluor 568 (Invitrogen) were added and incubated for 1 hour. Hoechst (Pierce) was used to stain nucleus. Western Blot. Western blot was performed with whole cell lysates from MD10-F2 cells. Cells were washed with cold PBS and lysed with RIPA buffer (1× PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 10 mg/ml phenylmethylsulfonyl fluoride, 30 μl/ml aprotinin, 100 mM sodium orthovanadate). Proteins (40 μg/ well) were resolved by 10% SDS-PAGE and transferred to a Trans-blot membrane (Bio-Rad). Western blot was performed as described earlier 2 . Anti-Dlx3, anti-Osx, and anti-Dsp (Santa Cruz) was used as primary antibodies. β-actin antibody (Santa Cruz) was used as a loading control.

Construction of Reporter Gene Constructs, Expression Plasmids, and Mutagenesis. All reporter
gene constructs containing 5′ deletions of the Dspp promoter were generated using standard cloning procedures as described previously 58 . Briefly, the 1372-bp BglII-HindIII fragment of the mouse Dspp gene from nucleotides (nt) −1318 to +54 was cloned into the BglII and HindIII sites of pGL3-Basic luciferase vector (Promega) and designated p1318. The 646-bp EcoRI-HindIII fragment of the mouse Dspp gene from nt −591 to +54 was cloned into the EcoRI and HindIII sites of the pGL3-Basic and called p591. The 372-bp XhoI-HindIII fragment of the mouse Dspp gene from nt −318 to +54 was cloned into the XhoI and HindIII sites of pGL3-Basic vector and called p318. All these various plasmids contained part of the exon 1 noncoding region of Dspp gene.
For the generation of expression constructs, the full length of Dlx3 cDNA or Osx cDNA containing a Flag tag was subcloned into pcDNA 3.1 vector (pcDNA-Dlx3, pcDNA-Osx-Flag). All constructs were confirmed by DNA sequencing.
Site Luciferase reporter assay. Cell were seeded in 48-well plates and cultured overnight. Then wild-type or mutant reporter plasmids or empty pGL3-Basic plasmid and pRL-TK Renilla luciferase reporter (Promega), as well as pcDNA-Dlx3 or/and pcDNA-Osx-Flag were co-transfected into cells using the Lipofectamine 2000 for 48 hours. After that, the cells were collected and lysed in passive lysis buffer (Promega). The luciferase assay was performed using the Dual Luciferase Reporter Assay System (Promega) according to the manufacture's protocols. Firefly and Renilla luciferase activities were qualified using the Glomax Luminometer (Promega). Firefly luciferase activities was normalized against Renilla luciferase activities. All luciferase assays were performed in triplicate at least three times.
Chromatin Immunoprecipitation (ChIP). ChIP were performed according to the instructions provided by ChIP-IT TM kit (Active Motif). Briefly, MD10-F2 cells were co-transfected with pcDNA-Dlx3, pcDNA-Osx-Flag expression plasmids and p318 plasmid. After 72 hours of transfection, cells were washed in PBS and incubated for 10 minutes with 1% formaldehyde. After quenching the reaction with 0.1 M glycine, the cross-linked material was sonicated into chromatin fragment of an average length of 200-800 bp. After pre-cleaning, 1% of each sample was saved as input fraction. The chromatin solution was precleared by adding protein G beads with salmon sperm DNA for 2 hours at 4 °C. Then immunoprecipitations were performed with protein G beads and 5 μg of anti-Dlx3 or Osx antibodies overnight at 4 °C. Anti-TFIIB antibody and negative control IgG (all from Santa Cruz Biotechnology) were used as positive and negative controls. Immunoprecipitated materials were washed, and cross-links were reversed by incubating samples for 5 hours at 65 °C in 200 mM NaCl and 10 μg of Rnase A to eliminate RNA. Recovered material was treated with proteinase K, and the DNA fragment was purified by Qiagen columns (Qiagen). The purified DNA was analyzed by PCR amplifying promoter regions (−318/+54) of the Dspp gene using the following primer pairs: forward −318 5′-GAAATGCAGGGTGACAGAGTCTAAGTGGCT-3′ −289 , and reverse +54 5′-CGAGGGG ACTTTGAAAATCCAGATT-3′ +30 . The sequence of positive control PCR primers was provided by the kit, which flank the GAPDH promoter. The sequence of negative control PCR primers was also provided by the kit, which flank a region of genomic DNA between the GAPDH gene and the chromosome condensation-related SMC-associated protein (CNAP1) gene.
Scientific RepoRts | 7: 10775 | DOI:10.1038/s41598-017-10908-8 Electrophoretic Mobility Shift Assay (EMSA). Recombinant Dlx3 protein was purchased from Novus Biologicals. Nuclear extracts from MD 10-F2 cells were prepared using the method of Dignam et al. 59 . Protein concentration was determined by using the Bradford assay 60 . Oligonucleotides used in EMSA were synthesized as listed in Table 1. Dlx3 (Con) Oligo and Osx (Con) Oligo were set as positive control, which were verified containing two copies of the Dlx3 binding sites and one GC-rich Osx binding site, respectively 39,61 . Dlx3I Oligo , Dlx3II Oligo , and Osx Oligo consist of the putative Dlx3 or Osx binding region of Dspp promoter. These oligonucleotides were labeled with [γ-32 P]ATP. Double-stranded probes were generated by annealing [γ-32 P]ATP-labeled complementary oligonucleotides. EMSA was performed as described previously 58 . For the competition binding reactions, the unlabeled competitor in 100-fold molar excesses of the labeled probe was included in the reaction. Antibody super-shift experiments were performed with anti-Dlx3 and anti-Osx antibodies. The antibodies were added to recombinant Dlx3 protein or the nuclear extracts 10 min prior to the addition of the radiolabeled prob. Free and protein-bound DNA complexes were loaded onto a 5% native polyacrylamide gel in 1 × Tris/boric acid/EDTA (TBE) buffer, electrophoresed, dried, and exposed to X-ray film.
Co-immunoprecipitation (Co-IP) Analysis. MD10-F2 cells cultured in 10-cm dish were transfected with 10 μg of pcDNA-Dlx3 and pcDNA-Osx-Flag. After 72 hours of transfection, cells were lysed with RIPA lysis buffer, and centrifuged at 12,000 rpm at 4 °C for 20 minutes. The supernatant was collected and precleared with 40 μl of protein A/G plus agarose beads (Santa Cruz) at 4 °C for 30 minutes. Immunoprecipitation was performed with anti-Dlx3, anti-Osx, anti-Flag (Sigma) antibodies or negative control IgG (3 μg each) and immunocomplexes were precipitated with 50 μl of protein A/G plus agarose beads. Beads were washed three times with wash buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10% glycerol, 1% NP-40, and 2 mM EDTA). Protein was eluted by boiling in 50 μl of sample buffer for 10 minutes and analyzed by Western blot.

Statistical Analysis
Quantitative data were presented as means ± SD from three independent experiments and analyzed with Analysis of Variance (ANOVA) or Student's t test with SPSS. For Western blot, densitometry of immunoblot bands were analyzed with ImageJ software.