Zfhx4 regulates endochondral ossification as the transcriptional platform of Osterix in mice

Endochondral ossification is regulated by transcription factors that include SRY-box transcription factor 9, runt-related protein 2 (Runx2), and Osterix. However, the sequential and harmonious regulation of the multiple steps of endochondral ossification is unclear. This study identified zinc finger homeodomain 4 (Zfhx4) as a crucial transcriptional partner of Osterix. We found that Zfhx4 was highly expressed in cartilage and that Zfhx4 deficient mice had reduced expression of matrix metallopeptidase 13 and inhibited calcification of cartilage matrices. These phenotypes were very similar to impaired chondrogenesis in Osterix deficient mice. Coimmunoprecipitation and immunofluorescence indicated a physical interaction between Zfhx4 and Osterix. Notably, Zfhx4 and Osterix double mutant mice showed more severe phenotype than Zfhx4 deficient mice. Additionally, Zfhx4 interacted with Runx2 that functions upstream of Osterix. Our findings suggest that Zfhx4 coordinates the transcriptional network of Osterix and, consequently, endochondral ossification.

V ertebrate skeletal elements are formed through two distinct processes: intramembranous and endochondral ossifications 1,2 . Intramembranous ossification is the process by which mesenchymal cells differentiate into osteoblasts and is consequently responsible for the formation of the clavicle and most of the craniofacial skeleton 3 . Conversely, endochondral ossification involves numerous steps of chondrocyte differentiation followed by replacement of cartilage tissue with bone tissue, which gives rise to most bones that include vertebrae, ribs, and long bones 1,4 . The first step of endochondral ossification is mesenchymal condensation after which cells differentiate into resting and proliferating chondrocytes that produce components of the chondrogenic extracellular matrix including collagen type 2 alpha 1 chain (Col2a1), collagen type 11 alpha 2 chain (Col11a2), and aggrecan (Acan) 5,6 . The chondrocytes then differentiate into hypertrophic chondrocytes that produce collagen type 10 alpha 1 chain (Col10a1), Indian hedgehog protein (Ihh), and matrix metalloproteinase 13 (Mmp13) [7][8][9][10] . Subsequently, hypertrophic chondrocytes undergo apoptosis, cartilage matrices are calcified, vascular vessels invade the cartilage tissue and, finally, cartilage tissues are replaced by bone tissue 11 . These unique and complex steps are regulated by various critical transcription factors and coactivators 10 .
The transcription factor SRY-box transcription factor 9 (Sox9), is essential for the early stage of chondrogenesis, in which mesenchymal condensation and chondrocyte differentiation occur 12 . Mutations in the SOX9 gene in humans cause campomelic dysplasia (OMIM 114290) characterized by severe chondrodysplasia 13,14 . Similarly, mutations around the SOX9 gene cause Pierre Robin sequence (OMIM 261800) that involves a cleft palate 15 . The expression of transcriptional factors Sox5 and Sox6, both of which are crucial for early chondrogenesis, is induced by Sox9 [16][17][18] . A critical role of Sox9 in chondrogenesis has also been demonstrated by a study in which chondrogenesis was severely impaired in chondrocyte-specific Sox9 conditional knockout (KO) mice 19 . Moreover, Sox9 directly regulates the expression of early-stage chondrogenic genes that include Col2a1, Col11a2, and Acan 6,16,20 .
Conversely, runt-related protein (Runx) 2, Runx3, and Osterix are crucial for the late stages of chondrogenesis. Chondrocyte hypertrophy is severely impaired in Runx2 KO mice, whereas Runx2 and Runx3 double KO mice exhibit a complete lack of hypertrophic chondrocytes 21 . Furthermore, Runx2 plays a critical role in the regulation of Col10a1, Ihh, and vascular endothelial growth factor (Vegf) [21][22][23] . Osterix (also known as Sp7) functions as a downstream transcriptional partner of Runx2 during bone and cartilage development 10 . Chondrocyte-specific Osterix conditional KO mice show a lack of calcification in cartilage matrices and matrix vesicles as well as loss of Mmp13 expression 10 . Thus, the Sox9-Runx2-Osterix axis is responsible for spatiotemporal mediation of endochondral ossification.
Considering the sequential and harmonious regulation of endochondral ossification achieved through this transcriptional network, it is possible that transcription factors that have not been identified may be involved in the multiple steps of cartilage development.
The present study aimed to identify novel transcription factors involved in the regulation of endochondral ossification and to elucidate the functional roles of these factors. We isolated zinc finger homeobox 4 (Zfhx4) from mouse limb buds and identified it as a highly expressed transcription factor in chondrogenic tissues. We demonstrated that Zfhx4 is required for endochondral ossification in conjunction with Osterix. We also found that Zfhx4 physically interacts with Runx2. Because Zfhx4 is a large transcription factor with four homeodomains and 22 zinc finger domains, Zfhx4 might function in the transcriptional platform and interact with several transcription factors and transcriptional regulators. Our findings provide insights into the role of Zfhx4 in the regulation of endochondral ossification.

Results
Identification of Zfhx4 as a transcription factor highly expressed in chondrocytes. Microarray analysis revealed high expression levels of p54 nrb , Arid5b, Bbf2h7, Cbfβ, ATF4, Sox9, Sox5, Sox6, Wwp2, Runx2, Dlx5, Dlx6, Msx2, and Osterix in mouse limb bud cells, all of which play important roles in endochondral ossification (Fig. 1a). Furthermore, Zfhx4 was highly expressed in limb bud cells (Fig. 1a). We focused on Zfhx4 for further analyses because of its putative role in 8q21.11 microdeletion syndrome (OMIM 614230) characterized by skeletal abnormalities 35 . Figure 1b shows the results of reverse transcriptase-quantitative polymerase chain reaction (RT-qPCR) analyses of newborn mouse tissues, which showed that Zfhx4 was highly expressed in several skeletal tissues that included ribs, limbs, and the calvaria. Consistent with the RT-qPCR data, expression of Zfhx4 was detected in limb bud cells by whole mount in situ and in situ hybridization analyses ( Supplementary  Fig. 1a, b). Additionally, in situ hybridization showed high expression of Zfhx4 in the growth plate of embryonic day (E) 16.5 mouse femurs (Fig. 1c).
Presence of skeletal abnormalities and cleft palate in Zfhx4deficient mice. To generate Zfhx4-deficient mice, we first generated Zfhx4 floxed mice ( Supplementary Fig. 1c) and confirmed germline transmission by Southern blot and PCR analyses ( Supplementary Fig. 1d). The generation of Zfhx4 heterozygousdeficient mice by further crossing was confirmed by PCR.
Interaction of Zfhx4 with Osterix during endochondral ossification. Because the skeletal phenotype of Zfhx4 −/− mice resembled that of Osterix KO mice with reduced Mmp13 expression and calcification of cartilage matrices 10 , we hypothesized that Zfhx4 interacted with Osterix during endochondral ossification. Coimmunoprecipitation and immunofluorescence analyses demonstrated that Zfhx4 physically associated with Osterix and that the proteins were colocalized in the nucleus (Fig. 3a, b). These results clearly demonstrated an association of Zfhx4 with Osterix. To understand the importance of this finding in terms of endochondral ossification, we generated double mutant mice of Zfhx4 and Osterix. Because Osterix KO mice exhibit a complete lack of Mmp13 expression and calcification of cartilage matrices 10 , we attempted to generate Zfhx4 −/− ; Osterix +/− mutant mice and compared the phenotype with Zfhx4 −/− mice. Double mutant Zfhx4 +/− ;Osterix +/− mice appeared to be similar to wild-type mice ( Fig. 3c-f, Supplementary Fig. 5a). Conversely, Zfhx4 −/− ;Osterix +/− mutant mice had less calcification of the humerus and femur at E16.5 than Zfhx4 −/− littermates (Fig. 3c). Histological analyses indicated that hypertrophy and calcification of the femur were more severely impaired in Zfhx4 −/− ;Osterix +/− mice at E16.5 than in Zfhx4 −/− littermates (Fig. 3d, e, Supplementary Fig. 6). Moreover, Mmp13 expression was diminished in Zfhx4 −/− ;Osterix +/− mice even at It has been shown that Mmp13 expression is regulated by a complex formed by Runx2 and Osterix 10 . Supplementary Figure 7a shows the results of coimmunoprecipitation of Zfhx4 and Runx2, which revealed that the two proteins interacted physically and were localized to the nucleus ( Supplementary  Fig. 7b). Consistent the notion that Runx2 is an upstream transcription factor of Osterix 10,36 , Osterix was not required for the interaction of Zfhx4 with Runx2 ( Supplementary Fig. 7c, d). We next examined the role of Zfhx4 in regulation of the Mmp13 gene promoter. A luciferase reporter assay indicated that Zfhx4 upregulated Mmp13 gene promoter activity and increased transcriptional activities of Osterix and Runx2 on their promoters ( Supplementary Fig. 8a). Consistently, chromatin immunoprecipitation analyses revealed binding of Zfhx4, Osterix, and Runx2 to Mmp13 gene promoter regions ( Supplementary Fig. 8b). Because Runx2 had a different binding ability to the Mmp13 gene promoter from Zfhx4 and Osterix, Zfhx4 may sequentially interact with Runx2 and Osterix.
Role of Zfhx4 in palate development. Because Zfhx4 −/− mice showed a complete cleft palate at 100% penetrance (Fig. 1d, Supplementary Fig. 9a), we examined expression of Zfhx4 in the palatal shelf. In situ hybridization analysis clearly indicated that Zfhx4 was expressed in the palatal shelf of E13.5 mice (Fig. 4a). Examination of palate development of wild-type and Zfhx4 −/− littermates at E13.5, E14.5, and E16.5 ( Fig. 4b and Supplementary  Fig. 9b-d) revealed that the growth and elongation of palatal shelves were normal in both Zfhx4 −/− and wild-type E13.5 mice. Palatal shelves were fused in E14.5 wild-type mice, whereas palatal shelves of Zfhx4 −/− mice were not elevated and had failed to fuse by E14.5 ( Fig. 4b and Supplementary Fig. 9c). Palatal shelves of E16.5 Zfhx4 −/− mice were degraded ( Fig. 4b and Supplementary Fig. 9d). Consistent with the results indicating that growth and elongation of palatal shelves were normal in Zfhx4 −/− mice, immunofluorescence with anti-Pcna and -Ki67 antibodies indicated normal cell proliferation in the palatal shelf of Zfhx4 −/− mice (Fig. 4c, d). Additionally, there was no detectable difference in the number of apoptotic cells in the palatal shelf between Zfhx4 −/− and wild-type littermate mice ( Supplementary Fig. 10), which suggested that apoptosis was not involved in the etiology of cleft palate in Zfhx4 −/− mice. To further examine the role of Zfhx4 in palatal development, we performed organ culture experiments using maxilla of E14.2 Zfhx4 −/− mice. Unexpectedly, palatal selves of Zfhx4 −/− mice were elevated and fused as well as wild-type littermates (Supplementary Fig. 11). These results suggested that Zfhx4 was unnecessary to elevate the palatal shelf itself.
Palatal development involves multiple steps, which include palatal shelf outgrowth, elevation, adhesion, and fusion, and palatal bone formation. Odd-skipped-related (Osr) 1 37,38 , Osr2 37,38 , fibroblast growth factor 10 (Fgf10) 39 , msh homeobox 1 (Msx1) 40,41 , and Paired box 9 (Pax9) 42,43 play essential roles in palatal development. Cranial neural crest-specific Sox9 conditional KO mice 44 and Runx2 KO mice 45 exhibit cleft palates. In the present study, we found no significant difference between wild-type and Zfhx4 −/− littermates in terms of the expression of Osr1, Osr2, Fgf10, Msx1, or Pax9 in palatal shelves (Fig. 4e, f and Supplementary Fig. 12). The expression of Runx2 and Osterix also remained unchanged in Zfhx4 −/− compared with wild-type littermates (Fig. 4e). Therefore, these results suggested that expression of these genes was not regulated and unknown Zfhx4 target genes might be involved in palatal development. Identification of target genes of Zfhx4 in the palatal shelf may contribute to understanding the molecular basis of palatal development, which involves Zfhx4.

Discussion
Cell differentiation, proliferation, and death are strictly regulated by transcriptional machinery complexes that comprise specific transcriptional factors and transcriptional coregulators. Although transcriptional regulation during endochondral ossification has been extensively investigated 1 , the details of the sequential and harmonious mechanism have not yet been elucidated. To address this, we attempted to identify transcription factors that regulate the transcriptional network system in endochondral ossification. In this study, we identified Zfhx4 as a transcriptional factor that is highly expressed in cartilage and limb buds. Notably, we found that Zfhx4 functioned as a transcriptional partner of Osterix. Coimmunoprecipitation and immunofluorescence analyses indicated a physical interaction between Zfhx4 and Osterix. Furthermore, genetic experiments revealed that Zfhx4 and Osterix interacted functionally in the late stage of endochondral ossification. Our data also indicated that Mmp13 expression was regulated by Zfhx4. Additionally, luciferase reporter and chromatin immunoprecipitation analyses indicated regulation of the Mmp13 gene promoter by Zfhx4. Because Osterix is critical for the regulation of Mmp13 expression in chondrocytes 10 , our results support the importance of the association between Zfhx4 and Osterix in endochondral ossification.
Col10a1 expression was diminished in Zfhx4 −/− mice as determined by in situ hybridization experiments (Fig. 2c). Conversely, we observed similar expression level of Col10a1 and Col10 between wild-type and Zfhx4 −/− littermates (Fig. 2d and  e), although the delayed expression pattern of Col10 was consistent with the in situ hybridization experiments. We assume that these difference resulted from slightly different time points of these experiments. Therefore, we believe that Col10a1 is not a critical transcriptional target for Zfhx4.
Because of the structure of Zfhx4, it is possible that Zfhx4 associates with numerous transcriptional regulators that function in the platform of the transcriptional network during endochondral ossification. In addition to Osterix, we found that Zfhx4 physically interacted with Runx2, a critical transcription factor in endochondral ossification 46 . Furthermore, Runx2 functions upstream of Osterix by controlling Mmp13 expression through its interaction with Osterix 10 . Thus, our study suggests that Zfhx4 facilitates the Runx2-Osterix axis, which results in coordination of the sequential steps of the transcriptional network that underlies endochondral ossification.
We found that endochondral ossification was impaired in the proximal limbs of Zfhx4-deficient mice, which included the scapula, humerus, and femur. The anterior-posterior patterning of the proximal limb is regulated by homeobox (Hox) 9 and 10 paralogous groups 47-50 as well as short stature homeobox 2 (Shox2) 51 . These genes are expressed in a spatiotemporally restricted pattern in domains along the axes of limb buds and provide anterior-posterior positional information for proximal limb patterning. Hoxa9/Hoxd9 double mutant mice have shortened humeri and loss of deltoid crests 50 , and triple mutant mice (Hoxa10/Hoxc10/Hoxd10) show short humeri, loss of deltoid processes, and remarkably shortened femurs 49 . Mutation of Shox2 causes marked shortening of the humerus and femur, as well as loss of the deltoid crest and ossification of the femur 52 . Double mutation of HoxD and Shox2 (HoxD −/− ;Shox2 +/− ) causes the phenotype of shortened humeri and lack of deltoid crests, whereas HoxD +/+ ;Shox2 +/− mutants show intact proximal limbs 51 . Considering the morphological phenotypes of Hox9, Hox10, and Shox2 mutant mice, Zfhx4 may be involved in the patterning processes of proximal limbs, possibly in association with Hox9, Hox10, and Shox2. Clinically, shortening of limbs is classified into four types: rhizomelic (dysplasia of proximal parts of the limbs), mesomelic (dysplasia of middle parts of the limbs), acromelic (dysplasia of distal parts of the limbs), and micromelic (generalized shortening of entire limbs) 53 . On the basis of the skeletal phenotype of Zfhx4-deficient mice, this mutation induces rhizomelic type skeletal dysplasia. Rhizomelic type skeletal dysplasia, achondroplasia, rhizomelic chondrodysplasia punctata, and spondyloepiphyseal dysplasia congenita are well characterized and several genes have been identified to be responsible for these diseases 53 . Achondroplasia (OMIM100800) is caused by heterozygous mutations in the gene that encodes fibroblast growth factor receptor 3 54 , rhizomelic chondrodysplasia punctata type1 (RCDP1) (OMIM 215100) is caused by homozygous or compound heterozygous mutations in the gene that encodes peroxisomal biogenesis factor 7 (PEX7) 55 , and spondyloepiphyseal dysplasia congenital (OMIM183900) is caused by heterozygous mutations in COL2A1 56 . However, the causative genes and pathological mechanisms of some rhizomelic type skeletal dysplasias, which include cleidorhizomelic syndrome, Patterson-Lowry rhizomelic dysplasia, and multiple epiphyseal dysplasia with Ribbing type, are yet to be elucidated. The Zfhx4 mutation described in the present study may be a good model to evaluate the pathogenesis of rhizomelic type skeletal dysplasia and such assessment might provide insights into the pathogenesis of rhizomelic dysplasia.
Our finding that Zfhx4 is highly expressed in calvariae suggests that Zfhx4 plays roles in both membranous and endochondral ossifications. However, Zfhx4-deficient mice showed no remarkable phenotype in terms of membranous ossification. We therefore assume that Zfhx4 has a partially redundant role and that its function might be compensated by other genes. A potential candidate gene is zinc finger homeobox 3 (Zfhx3; also known as ATBF1) that belongs to the Zfhx family and has the highest homology with Zfhx4 of all family members. The Zfhx3 protein is 406 kDa and contains four homeodomains and 23 zinc finger domains 57 . Supporting our theory, P10 Zfhx3 heterozygous mice are much smaller in terms of body size and weight than wild-type littermates 58 . This is presumably due to impaired skeletal development. Although dissection of Zfhx3/Zfhx4 double mutant mice may provide insights into the potential roles of Zfhx family proteins in skeletal development, this was beyond the scope of the present study. We intend to carry this out as the next step of this study.
We unexpectedly identified cleft palate at the elevation stage in Zfhx4-deficient mice. Notably, Zfhx4 was specifically expressed in palatal shelves, but not in the tongue. However, organ culture experiments showed that elevation and fusion abilities of palatal shelves from Zfhx4 −/− mice and wild-type littermates were intact. An explanation for this discrepancy between in vivo and ex vivo experiments is that Zfhx4 is required to remove the tongue from the middle of developing palatal shelved to create the space for palatal shelves to elevate and fuse. Considering specific expression of Zfhx4 in palatal shelves, it is likely that Zfhx4 regulates the expression of genes involved in palatal development. However, the expression levels of well-known palatal marker genes appeared normal in Zfhx4-deficient mice. Further investigation is required to identify Zfhx4 target genes in palatal shelves. The Zfhx4 gene has been proposed to be a candidate gene responsible for 8q21.11 microdeletion syndrome 35 and a report has shown this disease manifesting as a cleft palate in eight unrelated individuals. Thus, Zfhx4-deficient mice might be a suitable animal model to investigate the molecular mechanisms of palatal development and the pathogenesis of 8q21.11 microdeletion syndrome.
In conclusion, our results demonstrate that Zfhx4 is a transcriptional partner of Osterix and regulates the late stage of endochondral ossification. Our findings contribute to improving our understanding of the sequential and harmonious regulation of endochondral ossification.

Methods
Cell culture. Anterior and posterior limb buds were dissected from E12.5 Institute of Cancer Research (ICR) mouse embryos (SLC, Shizuoka, Japan) and digested by incubation in Dulbecco's modified Eagle's medium (DMEM) (Sigma-Aldrich, St. Louis, MO, USA) with 0.1% collagenase type II (Sigma-Aldrich) and 0.1% trypsin (Sigma-Aldrich) at 37°C 29 . Cells were dissociated by pipetting and then centrifuged at 300 × g for 5 min. The supernatant was removed and the cell pellet was resuspended with α-minimum essential medium (Sigma-Aldrich) with 10% fetal bovine serum (FBS; JRH Bioscience, Lenexa, KA, USA) to a final density of 2 × 10 7 cells/ml. Cells were incubated at 37°C in a humidified atmosphere with 5% CO 2 . 293FT and SW1353 cell lines were purchased from Thermo Fisher Scientific (Waltham, MA, USA) and the American Type Culture Collection (Manassas, VA, USA), respectively. Cells were cultured in DMEM with 10% FBS at 37°C in a humidified atmosphere with 5% CO 2 .
Animal experiments. All animal experiments were approved by the Osaka University Graduate School of Dentistry animal care committee and the Institutional Animal Care and Use Committee (IACUC) of RIKEN Kobe Branch. Animal experiments with E12, E12.5, E13.5, E14.5, E15.5, and E16.5 embryos, and newborn littermates were performed by following protocols approved by the Osaka University Graduate School of Dentistry animal care committee. CAG-Cre transgenic mice (RBRC01828) were provided by the RIKEN BioResource Research Center (Ibaraki, Japan).
Zfhx4 mutant mice were designed with floxed exons 7 and 8 ( Supplementary  Fig. S1c). Deletion of exons 7 and 8 caused a nonsense frame-shift mutation in most functional domains of the Zfhx4 protein, including all homeobox domains, 13 zinc finger domains and a region coding nuclear localization signal. To generate Zfhx4 floxed mice (accession number CDB0954K, http://www2.clst.riken.jp/arg/ micelist.html), TT2 embryonic stem (ES) cells were transfected with a targeting vector that contained a neomycin resistance gene (Supplementary Fig. S1c) by electroporation 59 . Homologous recombination was confirmed by Southern blot analysis. The ES cells were injected into eight-cell-stage ICR embryos and Zfhx4 floxed chimera mice were obtained. The Zfhx4 floxed chimeras were mated with C57BL/6 mice and then germline transmission of the mutant allele was achieved as confirmed by Southern blotting and genomic PCR analyses. The 590 bp probe (Supplementary information) located in exon 10 of the Zfhx4 gene ( Supplementary  Fig. 1c) was used for Southern blotting analyses of EcoRI-digested genomic DNA from mouse tails. The sequences of primers used for genomic PCR were as follows: 5′-GCCAAAGGCTGACTCAAAAC-3′ and 5′-GGGTCCCCACTGTGATTTCT-3′. To generate Zfhx4 global knockout mice, heterozygous Zfhx4 floxed mice were mated with CAG-Cre transgenic mice. Deletion of the floxed allele was confirmed by genomic PCR. Deletion of the Cre transgene was achieved by mating Zfhx4 heterozygous-deficient mice with C57BL/6 mice as confirmed by genomic PCR. Zfhx4-deficient mice were produced by interbreeding of Zfhx4 heterozygous mice. To determine mouse genotypes, the following primers were used in PCR: 5′-TCAC TGTGCATGAGGCAAAAC-3′ and 5′-GGGTCCCCACTGTGATTTCTA -3′. Osterix KO mice were used as described previously 10 .
Microarray analysis. We analyzed microarray data deposited in the Gene Expression Omnibus (Accession No: GSE126945). Briefly, total RNA was isolated from limb buds of E12.5 ICR mice using an Ambion® WT Expression Kit (Applied Biosystems, Branchburg, NJ, USA). Single-stranded DNA was synthesized from total RNA and subsequently hybridized to a GeneChip®Mouse Gene 1.0 ST Array (Affymetrix, Santa Clara, CA, USA). After 16 h of hybridization, the array was washed, scanned, and finally quantified by an Affymetrix Expression Console (Affymetrix) 10 .
Reverse transcriptase-quantitative polymerase chain reaction (RT-qPCR). Total RNA was isolated from cells using a NucleoSpin RNA Plus Kit (Macherey-Nagel, Duren, Germany). Total RNA was denatured by incubation at 65°C for 5 min, after which cDNA was synthesized using ReverTra Ace Quantitative PCR RT Master Mix with gDNA Remover (Toyobo, Osaka, Japan) 27 . Real-time RT-qPCR amplification was performed using the TaqMan PCR protocol and Step-One plus real-time PCR system (Applied Biosystems, Carlsbad, CA, USA). TaqMan probes used for amplification are listed in Table 1. The expression level of mRNA was normalized to β-actin mRNA expression 60 . Table 1 List of sequences of Taqman probe sets for real-time RT-PCR experiments.
von Kossa calcium staining. Sections were deparaffinized, incubated in a 5% silver nitrate solution for 30 min with exposure to sunlight, and then rinsed with two changes of distilled water. Sections were incubated in a 5% sodium thiosulfate solution for 2 min, rinsed with two changes of distilled water, and then incubated in a nuclear fast red solution for 5 min 10 .
Construction of expression vectors. Myc-tagged Osterix and Venus-tagged-Osterix expression vectors were used as described previously 10 . The Myc-Runx2 expression vector was generated by subcloning PCR-amplified, full-length Runx2 cDNA (variant 3, 1791 bp) into the EcoRI and XbaI sites of a pcDNA3 expression vector with six tandem repeats of the Myc tag at the N-terminal. Venus-tagged Runx2 was generated by subcloning the corresponding cDNA into the Venustagged expression vector. Flag-tagged Zfhx4 was generated using the Gateway Cloning System (Thermo Fisher Scientific) 65 . The PCR amplification products of Flag-tagged Zfhx4 cDNA were subcloned into the Kpnl and Xhol sites of the Gateway entry clone and the resulting clone was transferred into the Gateway pcDNA Zeo destination vector using LR Clonase enzyme mix, thereby constructing the Flag-tagged Zfhx4 expression vector 66 .
Transfection. Transfection of expression vectors was carried out using a X-treamGene 9 (Roche) in accordance with the manufacturer's protocol.
Coimmunoprecipitation analysis. Cells transfected with Flag-Zfhx4, Myc-Osterix, or Myc-Runx2 were washed twice with ice-cold PBS and lysed in lysis buffer.
Lysates were centrifuged at 12,000 rpm for 20 min at 4°C and incubated with an anti-Flag antibody for 16 h at 4°C, followed by immunoprecipitation with Dyna-beads™ Protein G for Immunoprecipitation (Thermo Fisher Scientific). Immunoprecipitates were washed five times with ice-cold PBS and then boiled in SDS sample buffer with 0.5 M β-mercaptoethanol at 95°C for 5 min. Supernatants were subsequently subjected to western blot analysis 62 .
Statistics and reproducibility. Randomization and blinding were not performed in the animal studies. Data were statistically analyzed using a two-tailed and unpaired Student's t-test for intergroup comparisons. Values of P < 0.05 were considered statistically significant. All results were performed three times independently and reproduced with similar results.