The transcription factor NKX2-3 mediates p21 expression and ectodysplasin-A signaling in the enamel knot for cusp formation in tooth development

Tooth morphogenesis is initiated by reciprocal interactions between the ectoderm and neural crest–derived mesenchyme. During tooth development, tooth cusps are regulated by precise control of proliferation of cell clusters, termed enamel knots, that are present among dental epithelial cells. The interaction of ectodysplasin-A (EDA) with its receptor, EDAR, plays a critical role in cusp formation by these enamel knots, and mutations of these genes is a cause of ectodermal dysplasia. It has also been reported that deficiency in Nkx2-3, encoding a member of the NK2 homeobox family of transcription factors, leads to cusp absence in affected teeth. However, the molecular role of NKX2-3 in tooth morphogenesis is not clearly understood. Using gene microarray analysis in mouse embryos, we found that Nkx2-3 is highly expressed during tooth development and increased during the tooth morphogenesis, especially during cusp formation. We also demonstrate that NKX2-3 is a target molecule of EDA and critical for expression of the cell cycle regulator p21 in the enamel knot. Moreover, NKX2-3 activated the bone morphogenetic protein (BMP) signaling pathway by up-regulating expression levels of Bmp2 and Bmpr2 in dental epithelium and decreased the expression of the dental epithelial stem cell marker SRY box 2 (SOX2). Together, our results indicate that EDA/NKX2-3 signaling is essential for enamel knot formation during tooth morphogenesis in mice.


Tooth morphogenesis is initiated by reciprocal interactions between the ectoderm and neural crest-derived mesenchyme.
During tooth development, tooth cusps are regulated by precise control of proliferation of cell clusters, termed enamel knots, that are present among dental epithelial cells. The interaction of ectodysplasin-A (EDA) with its receptor, EDAR, plays a critical role in cusp formation by these enamel knots, and mutations of these genes is a cause of ectodermal dysplasia. It has also been reported that deficiency in Nkx2-3, encoding a member of the NK2 homeobox family of transcription factors, leads to cusp absence in affected teeth. However, the molecular role of NKX2-3 in tooth morphogenesis is not clearly understood. Using gene microarray analysis in mouse embryos, we found that Nkx2-3 is highly expressed during tooth development and increased during the tooth morphogenesis, especially during cusp formation. We also demonstrate that NKX2-3 is a target molecule of EDA and critical for expression of the cell cycle regulator p21 in the enamel knot. Moreover, NKX2-3 activated the bone morphogenetic protein (BMP) signaling pathway by up-regulating expression levels of Bmp2 and Bmpr2 in dental epithelium and decreased the expression of the dental epithelial stem cell marker SRY box 2 (SOX2). Together, our results indicate that EDA/NKX2-3 signaling is essential for enamel knot formation during tooth morphogenesis in mice.
Tooth morphogenesis is initiated by reciprocal interactions between the ectoderm and neural crest-derived mesenchyme (1,2). A good model for understanding the mechanism of ectodermal organ development is tooth development, which features well-defined developmental stages and distinctive cell types. In mice, tooth morphogenesis is initiated by thickening of dental epithelium to form a dental placode, followed by invagination into the surrounding mesenchyme. Thereafter, tooth buds progress into the cap stage, and primary enamel knots are formed in dental epithelium, leading to tooth cusps. In the bell stage, a secondary enamel knot forms at the tip of the cusp-forming area to pattern the cusp and shape the tooth crown (3,4). In mice, only a single enamel knot is formed in incisors, whereas both primary and secondary enamel knots are formed in molars. An enamel knot is formed by a cluster of cells in dental epithelium and serves as a key signaling center that secretes sonic hedgehog (SHH), 4 bone morphogenetic protein (BMP), fibroblast growth factor (FGF), and members of the MMTV integration site (WNT) family, which are essential for forming the tooth crown shape and related cusps (4 -6). These signals are also essential for formation of the apical ectodermal ridge involved in limb development (7,8), indicating that organ formation is regulated by a common signaling pathway. During tooth growth, the mechanisms of differentiation during the later stages of dental epithelial cell development have been well defined (9 -12), whereas regulation and cell fate determination in the early stage of tooth morphogenesis are largely unknown.
Tabby and downless mutant mice have tooth, hair, and sweat gland defects (13). The tabby gene encodes ectodysplasin-A (EDA), a type II membrane protein of the tumor necrosis factor ligand family, whereas the downless gene encodes EDAR, a novel member of the tumor necrosis factor receptor family (14). Edar is expressed in the enamel knot, and mutations of Edar and its ligand the Eda gene lead to ectodermal dysplasia, which includes missing teeth and smaller teeth with reduced cusps, as shown in mutant mice (13,15). During this process, interactions between EDA and EDAR play a critical role in cusp for-mation. However, activity downstream of EDA signaling remains unclear.
NK2 homeobox families are tissue-specific transcription factors that feature a common DNA-binding structure and function as homeodomain genes, which critically regulate organ development of evolutionarily distant organisms ranging from the Drosophila genus to humans (16,17). NKX2-3 is expressed in the pharyngeal floor, as well as oral and branchial arch ectoderm tissues, whereas findings of Nkx2-3 knockout mice have revealed defects in maturation and cellular organization of the salivary glands, as well as largely absent molar cusps (18). Nevertheless, the role of NKX2-3 in tooth development remains largely unknown.
In the present study, we found a high level of Nkx2-3 expression in teeth obtained from embryonic day 14 (E14) mouse embryos using microarray analysis. Subsequently, the expression pattern of NKX2-3 was examined and found localized in the area of the enamel knot, which is closely related to cusp formation. Our findings also showed that Nkx2-3 is induced by EDA and regulates p21 transcription and expression by binding with its promoter region, thus altering cell proliferation. Finally, NKX2-3 induced enhancement of BMP signaling and inhibition of SOX2 expression. These results indicate important roles for NKX2-3 in cusp formation and tooth morphogenesis.

Nkx2-3 highly expressed during tooth morphogenesis
Microarray analysis was performed to identify expression of tooth-specific genes in the tooth morphogenesis stage. Gene expressions in teeth were compared with those in the whole body of mice on E14. Scatter plot data revealed that a large number of genes were either up-or down-regulated in E14 teeth (Fig. 1A). Furthermore, the expression of Nkx2-3 was at a high level in teeth relative to the whole body, which may indicate its important roles during tooth development. To confirm the expression pattern of Nkx2-3 during tooth development, qRT-PCR was performed using total RNA obtained from tooth, heart, submandibular gland, skin, lung, eye, kidney, hair, brain, and stomach samples, as well as during each stage of tooth development (i.e. E11-E18, as well as postnatal day 0 (P0) to P7). Nkx2-3 expression was found to be elevated in teeth as compared with the other examined organs (Fig. 1B) and increased in accordance with tooth morphogenesis stage (E13-E16) (Fig. 1C), indicating that this gene has critical roles in tooth morphogenesis. To confirm whether Nkx2-3 is expressed in dental epithelium or mesenchyme, E14 dental epithelium was separated from mesenchyme under a microscope after treatment with dispase, and expression levels were determined by qRT-PCR. The dental epithelium-specific gene Krt14 was shown to be highly expressed in epithelium, whereas Vim, which encodes the mesenchyme-specific protein vimentin, was highly expressed in mesenchyme, and Nkx2-3 was mainly expressed in dental epithelium (Fig. 1D). Other members of the NK2 family showed different expression patterns as compared with the tooth specificity of Nkx2-3 ( Fig. 1B and Fig. S1), indicating that the NK2 family is composed of tissue-specific tran-scription factors, of which NKX2-3 may play an important role in tooth development. To examine the localization of NKX2-3 during tooth development, we performed immunohistochemistry using first molars obtained from embryo heads on E13, E14, and E16. NKX2-3 expression was shown to begin from the bud stage (E13) and continue in the cap stage (E14), mainly in the primary enamel knot area (Fig. 1E, enlarged right panel). In the bell stage (E16), its expression appeared in the second enamel knot, which is consistent with the enamel knot marker p21 (Fig. 1E, bottom panel). Furthermore, the basement membrane molecule perlecan was stained with NKX2-3 to reveal the border of epithelial and mesenchymal tissues (Fig. 1E). These results indicate that NKX2-3 is highly expressed in dental epithelial cells and localized in enamel knot areas.

Reduction of Nkx2-3 interrupts cusp formation of molars in an organ culture system
To examine the role of NKX2-3 in enamel knots of developing teeth, E13 mandibular tooth germs were dissected and then transfected with control or Nkx2-3 siRNA ( Fig. 2A). Nkx2-3 was specifically down-regulated by 50% by Nkx2-3 siRNA in an ex vivo organ culture system (Fig. 2B). After 1 week of culture, tooth cusp sizes in both groups were measured (Fig. 2D), with no significant differences regarding width or total height between the groups (Fig. 2E). However, cusp height in the presence of Nkx2-3 siRNA was significantly reduced as compared with the control (Fig. 2, A and E). These results suggest that NKX2-3 is critical for molar cusp formation.
Cusp formation is believed to be regulated by an enamel knot expressing the cyclin-dependent kinase inhibitor p21. To clarify dental epithelial cell proliferation in cultured molars, we examined Ki67, a marker for cell proliferation, as well as p21 using immunohistochemistry. The number of NKX2-3positive cells was reduced in molars transfected with Nkx2-3 siRNA compared with the control, whereas Ki67-positive cells were increased in the presence of Nkx2-3 siRNA (Fig. 2, F and G). In contrast, p21 was down-regulated in the enamel knot area in Nkx2-3 siRNA-transfected molars (Fig. 2, F and G). In addition, qRT-PCR data revealed that the mRNA level of p21 (Cdkn1a) was also reduced (Fig. 2C), suggesting that NKX2-3 regulates p21 expression in dental epithelial cells. We considered that p21 disappearance in the enamel knot region may be a reason for the inhibition of cusp formation seen in Nkx2-3 siRNA-transfected molars.

Nkx2-3 inhibits cell proliferation in dental epithelial cells
We evaluated cell proliferation using the M3H1 dental epithelial cell line previously established by our group (19). Proliferation was decreased in the Nkx2-3 transfected group as compared with the control (Fig. 3A). Cell count and bromodeoxyuridine (BrdU) incorporation results also showed decreased cell proliferation in the Nkx2-3 transfectant group (Fig. 3B). Furthermore, the ratio of Ki67-positive cells was also reduced (Fig. 3C), which was consistent with our results shown by organ culture immunostaining (Fig. 2, F and G). Together, these findings suggest that NKX2-3 is necessary for regulation of epithelial cell proliferation. We also examined expressions of the cell cycle molecule cyclin D1 and cell proliferation-related

Nkx2-3 regulates tooth cusp formation and p21 expression
markers Ki67, p21, c-Myc, and c-Jun following transfection with Nkx2-3 by Western blot analysis. Ki67, c-Myc, and c-Jun were decreased by Nkx2-3 overexpression, whereas cyclin D1 showed no difference (Fig. 3D). Additionally, p21 was up-regulated in Nkx2-3-overexpressed cells, consistent with findings of the organ cultures. These results suggest that dental epithelial cell proliferation in vitro can be adjusted by NKX2-3.

Nkx2-3 regulates cell cycle regulator p21 and binds to its promoter region in dental epithelial cells
To confirm regulation of p21 expression by NKX2-3, we transfected an Nkx2-3-expressing vector into M3H1 cells and then evaluated the expression level of p21. Our results showed that p21-positive cells were increased at 48 h after transfection with Nkx2-3 (Fig. 3, E and F). qRT-PCR findings also showed that Cdkn1a mRNA expression was increased in the Nkx2-3transfected group (Fig. 3F), whereas Western blotting revealed that NKX2-3 induced p21 at the protein level (Fig. 3G). These results suggest that NKX2-3 induces p21 expression in mRNA and at the protein level in vitro. The expression level of p21 was also evaluated by qRT-PCR in Nkx2-3 siRNA-transfected M3H1 cells, which showed that Nkx2-3 expression was significantly inhibited and that of Cdkn1a (p21) was also decreased (Fig. 4A). Next, we examined whether NKX2-3 can potentially bind to the promoter region of Cdkn1a using the JASPAR database and found a motif sequence (Fig. 4B). This screening revealed a GTACTC motif at 328 bp upstream of the transcription start site (TSS) of the Cdkna1 gene (Fig. 4C). To identify binding of NKX2-3 in the Cdkn1a promoter region, Nkx2-3overexpressing M3H1 cells were fixed, and a ChIP assay was performed using antibodies against the V5 tag. DNA segments binding to NKX2-3 protein were detected by qRT-PCR using primer sequences at the proximal promoter region (site 1, bp Ϫ310 to Ϫ179) containing the consensus DNA-binding sequences as well as the distal region of the promoter (site 2, bp Ϫ2263 to Ϫ2130), used as a control (Fig. 4C). Expression of NKX2-3 associated with the site 1 region of the Cdkn1a promoter was ϳ4.2-fold higher as compared with the mock, whereas it was not significantly different in the control site ( Fig.  4E). To further confirm the target sequence of NKX2-3 in the Cdkn1a promoter, we constructed luciferase reporter vectors with insertion of the Cdkn1a promoter or a Cdkn1a promoter mutant by changing the nucleotide sequence GTACTC into GGGGTC (Fig. 4D). A Nkx2-3-expressing vector was transfected into M3H1 cells, and then promoter activities were evaluated using a luciferase reporter assay. Compared with the mock group, the Nkx2-3-overexpressed group had a 2-fold increase in transcriptional activity of the WT sequence Cdkn1a promoter, whereas the use of a triple-base mutation of the Cdkn1a promoter region resulted in a decrease of transcriptional activity as compared with the WT (Fig. 4F). These find-ings indicated that NKX2-3 directly binds to the promoter region of Cdkn1a and regulates its gene expression in dental epithelial cells.

EDA/EDAR signaling regulates Nkx2-3 expression in dental epithelial cells
EDA/EDAR signaling is important for enamel knot formation and tooth morphogenesis. In mice, only a single enamel knot is formed in incisors, whereas both primary and secondary enamel knots are formed in molars. qRT-PCR results revealed that Nkx2-3 as well as Eda and Edar were expressed at higher levels in molars as compared with incisors (Fig. 5A), thus suggesting that NKX2-3 is a key regulator for enamel knot formation. To examine whether EDA/EDAR regulates Nkx2-3 expression, M3H1 cells were stimulated with Eda-a1 as well as BMPs, FGF9, Wnt3a, and SHH. Nkx2-3 was shown to be significantly induced by Eda-a1 (Fig. 5B). Furthermore, we added Eda-a1 to organ culture medium used for culturing teeth transfected with Nkx2-3 siRNA, which apparently abrogated the decrease in cusp height (Fig. 5C), suggesting that NKX2-3 is a target transcription factor regulated by the EDA/EDAR signaling pathway in the enamel knot (Fig. 5I). Other molecules, including BMPs, did not have effects on cusp formation in Nkx2-3 siRNA-treated organ cultures (data not shown).
BMP signaling is also an inducer of p21 and associated with epithelial-mesenchymal interaction in the enamel knot region (20). In an examination of the ability of BMPs to induce Nkx2-3, we found no change in expression level of Nkx2-3 in M3H1 cells after adding BMP2 or BMP4 (Fig. 5B). However, it was also noted that transfected Nkx2-3 induced Bmp2 and Bmpr2 mRNA expressions (Fig. 5D) as well as phosphorylation of Smad1/5/8 and Smad1 (Fig. 5E) in M3H1 cells. Thus, NKX2-3 may directly or indirectly regulate BMP signaling in the enamel knot (Fig. 5I).

Nkx2-3 decreases Sox2؉ cells in the buccal side of dental epithelial cells
Recently, a cell lineage tracing method was used to show that mouse molar anterior buccal secondary enamel knots are derived from primary enamel knots, which were originally Sox2ϩ cells (21). To elucidate whether NKX2-3 can drive stem cells into differentiation, we examined expression of the dental epithelial stem cell marker SOX2 in vitro. We cultured M3H1 cells transfected with Nkx2-3 or a mock vector for 48 h and noted that the number of Sox2ϩ cells among the Nkx2-3transfected cells was decreased (Fig. 5F). Western blot analysis also showed that SOX2 expression was decreased in Nkx2-3overexpressed M3H1 cells (Fig. 5G). We also used an ex vivo organ culture system with or without Nkx2-3 siRNA, and found that Sox2ϩ cells were increased on the buccal side of the tooth germ and in the enamel knot in the Nkx2-3 siRNA-transfected Figure 1. Expression of NKX2-3 and p21 during tooth morphogenesis. A, in samples obtained on E14, differentially expressed genes were identified by microarray analysis of teeth, and the results were compared with whole embryos. Highlighted plot, Nkx2-3. Red and blue plots, up-and down-regulated genes, respectively. B, qRT-PCR analysis of Nkx2-3 expression in tooth, heart, submandibular gland, skin, lung, eye, kidney, hair, brain, and stomach samples from E14 embryos after normalization to Gapdh mRNA expression. C, qRT-PCR analysis of Nkx2-3 expression in teeth obtained from E11 to P7 after normalization to Gapdh mRNA expression. D, qRT-PCR analysis of Nkx2-3, Krt14, and Vim expressions in tooth epithelium and mesenchyme. E, NKX2-3 (green) expression in E13, E14, and E16 mice, detected by immunocytochemistry. p21 (green) expression in E13, E14, and E16 mice, detected by immunocytochemistry. Perlecan (red) was used to detect the basement membrane. Nuclei were stained using DAPI (blue). Error bars, S.D. Scale bars, 50 m.

Nkx2-3 regulates tooth cusp formation and p21 expression
group (Fig. 5H). Together, these results suggest that NKX2-3 contributes to cell fate determination of the enamel knot for cusp formation (Fig. 5I).

Discussion
Hypodontia, congenital absence of teeth, can occur without or with association with certain genetic syndromes. Several genes are known to be involved with isolated or nonsyndromic hypodontia, such as PAX9 (22), MSX1 (23), AXIN2 (24), and EDA (25), whereas other reports have described isolated tooth agenesis or microdontia in families with mutations of EDA (26 -28) and EDAR (29 -31), suggesting that EDA/EDAR signaling is a critical component of tooth development. Furthermore, these molecules associated with tooth anomalies also regulate enamel knot formation and function. In the present study, we demonstrated that NKX2-3 is a potential target gene of EDA/EDAR signaling, as it regulates cell proliferation through p21 transcription in the molar enamel knot.
The NK2 homeobox family members are known to function as tissue-specific homeobox transcription factors that regulate various fundamental cellular processes, including head patterning, cardiac and lung development, and neural cell specification (32,33). Nkx2-1 encodes thyroid transcription factor-1 (TTF-1), which is a protein that regulates transcription of genes specific for the thyroid, lungs, and diencephalon. Nkx2-1 mutations lead to surfactant protein dysregulation, which causes interstitial lung disease in patients with brain-lung-thyroid syndrome (34,35). Our results showed that Nkx2-1 was highly expressed in lung and brain tissues obtained from E14 mouse embryos (see Fig. S1), suggesting that the specific expression pattern of the NK2 homeobox family is critical for tissue-specific growth and development. NKX2-3 controls development of small intestine splenic morphology as well as shaping of lymphocyte dynamics and vasculature (36 -41) and also salivary gland and tooth development (18). Absence of Nkx2-3 in mouse molars was reported to result in no cusp formation, whereas there were no phenotypical changes in the incisors (18), suggesting that NKX2-3 is a critical regulator for molar cusp formation. However, the molecular mechanism of NKX2-3 in regard to tooth development remains unclear.
An important event during tooth development is formation of the enamel knot and its expression of several signaling molecules. During knot formation, a group of proliferating cells exit the cell cycle and are regulated by the cyclin-dependent kinase inhibitor p21 (20). In the present study, we found NKX2-3 to be highly expressed in the enamel knot region of molars and involved in regulation of p21 expression in both primary and secondary enamel knots as a homeobox transcription factor. BMP2 and BMP4, as well as BMP7 are expressed in early dental epithelium and regulate tooth morphogenesis (42)(43)(44)(45)(46). The expression of Bmp4 shifts to the mesenchyme at the bud stage (47), indicating that BMPs are critical regulators for epithelial-mesenchymal interaction during early tooth development. BMP signaling is an inducer of p21 and associated with apoptosis in the enamel knot (20). Together, these findings suggest that BMP2 in the enamel knot may induce BMP4 in mesenchyme through an epithelial-mesenchymal interaction at the bud stage, and then BMP signaling is accelerated by stimulation of BMP2 and BMP4 in the enamel knot area in an autocrine or paracrine manner (Fig. 5I). BMP signaling may also be critically involved in cusp formation and tooth germ development by regulating cell proliferation and differentiation in the enamel knot. The present findings revealed that NKX2-3 regulates Bmp2 and Bmpr2 expression, indicating NKX2-3 as a regulator of the BMP signaling pathway in the enamel knot. Furthermore, we found that BMP2 had no effects on cusp formation in an Nkx2-3 siRNA-treated organ culture (data not shown), which may be related to down-regulation of Bmp2 and Bmpr2 expression in Nkx2-3-silencing cells.
SOX2 is a member of the family of SRY-related HMG transcription factors that are important for cell fate and differentiation in developmental processes and are also essential for embryonic development (48 -50). During tooth development, Sox2ϩ cells contribute to all epithelial cell lineages and function as dental epithelial stem cells in incisor and molar development (51,52). Using a lineage tracing method, a recent study revealed the cell lineage of enamel knots (21), which demonstrated that the mouse incisor enamel knot is derived from Sox2ϩ cells, some of which differentiate into ameloblasts but not into stem cells in the cervical loop. On the other hand, in molars, primary enamel knot cells give rise to a buccal secondary enamel knot. Thus, cusp formation is controlled by construction of signaling centers that were originally SOX2-expressing stem cells. The present findings showed that NKX2-3 regulates SOX2 expression in M3H1 cells and ex vivo in the tooth germ, suggesting a crucial role for NKX2-3 in enamel knot formation and cell fate determination. Furthermore, differences in tooth shape between incisors and molars are due to the level of expression of NKX2-3 during odontogenesis.
We also identified NKX2-3 bound to the promoter region of Cdkn1a (p21) in ChIP assay findings. An NKX2-3 binding motif exists near the TSS of Cdkn1a, indicating that NKX2-3 directly regulates Cdkn1a expression by binding at the promoter region. NKX2-3 has been shown to bind with a variety of different genes, suggesting that other genes may be targeted for control of cell proliferation during tooth development.
In summary, we found that NKX2-3 is highly expressed in teeth, especially during the early developmental stage. Nkx2-3 is induced by EDA as a target molecule of the EDA/EDAR pathway in dental epithelial cells. NKX2-3 regulates cell proliferation by controlling p21 gene expression by binding directly to its promoter or through the BMP signaling pathway. We con-

Nkx2-3 regulates tooth cusp formation and p21 expression
cluded that NKX2-3 is a critical regulator of cusp formation and has an important role in regulating the fate of enamel knot cells.

Microarray analysis
All animal experiments were approved by the ethics committee of the Kyushu University Animal Experiment Center (pro-tocol A26-208-0). Total RNA samples from lower first molars and whole embryos from E14 mice were isolated using TRIzol reagent (Life Technologies, Inc.) and purified with an RNeasy Mini kit (Qiagen, Valencia, CA) according to the manufacturer's protocol. RNA quality was verified using an Experion

Nkx2-3 regulates tooth cusp formation and p21 expression
RNA and 9.9 for whole-embryo RNA. Labeling and array hybridization were performed using standard protocols at the Research Support Center of the Research Center for Human Disease Modeling of Kyushu University. Gene expression profiles were analyzed using a chip-based gene array (MouseWG-6 version 2, Illumina, Santa Clara, CA). Data were normalized using Genome Studio (Illumina). Gene expression analysis was performed using Subio Platform version 1.18 (Subio, Amami, Japan).

Tissue preparation
Pregnant mice were euthanized by anesthesia, and mouse embryos were immediately dissected. Embryo heads were fixed with 4% paraformaldehyde in PBS for 5 h at 4°C, and samples were embedded in O.C.T. compound (Sakura Finetek, Tokyo, Japan). Frozen sections of mandibular molars and incisors were obtained from mouse heads at each stage (E13, E14, and E16) and cut into 10-m-thick sections.

Organ cultures
Tooth germs from mandibular molars dissected from E13 mice embryos were seeded into cell culture inserts (BD Falcon, BD Biosciences) and grown using an air-liquid interface culture technique in Dulbecco's modified Eagle's medium (DMEM)/F-12, supplemented with 20% fetal bovine serum (Gibco/Life Technologies), 180 g/ml ascorbic acid, 2 mM L-glutamine, and 50 units/ml penicillin/streptomycin at 37°C in a humidified atmosphere of 5% CO 2 for 7 days, as described previously (19,53). For siRNA-mediated knockdown, tooth germs were transfected with siRNA for Nkx2-3 (ON-TARGET Plus L-057189-01-0002, Dharmacon) or control siRNA (ON-TARGET Plus Nontargeting Control Pool D-001810-05, Dharmacon) at a concentration of 500 nM using Lipofectamine 3000 reagent, according to the manufacturer's protocol. EDA-A1 (200 ng/ml; R&D Systems) was added to organ culture medium. Organs were cultured for 2 days and embedded into frozen blocks, and then the sections were immunostained. For evaluation of tooth size, E13 tooth germs were cultured for 7 days, and tooth size was determined using ImageJ software (Wayne Rasband, National Institutes of Health).

Construction of expression vectors
Nkx2-3 expression vectors were constructed using a Gateway cloning system (Life Technologies), according to the manufacturer's protocol. Briefly, the coding sequence of mouse Nkx2-3 cDNA without a stop codon was cloned into a pENTR/D-TOPO vector, and then the following forward and reverse primers were used: 5Ј-cacccaattaagtggccctgatgatgttaccaagcccggtcacc-3Ј and 5Ј-caaaggatccaccgccggt-3Ј. Expression vectors were constructed using an LR recombination reaction (pcDNA-DEST40; Life Technologies) tagged with V5-His.
For transfection, cells were cultured in 12-well plates at a density of 2 ϫ 10 5 cells/well in low-Ca 2ϩ DMEM and then transfected with an expression vector using Lipofectamine 3000 with Plus reagent (Life Technologies), according to the manufacturer's protocol.

Cell proliferation and BrdU incorporation
M3H1 cells were seeded into 96-well plates and then transfected with an Nkx2-3 expression vector. Proliferation was determined after culturing for 1, 2, 3, 5, and 7 days using a Cell Counting Kit (CCK)-8 (Dojindo Laboratories), according to the manufacturer's protocol. BrdU incorporation was assayed

Nkx2-3 regulates tooth cusp formation and p21 expression
using a BrdU labeling kit (Roche Diagnostics). Cells transfected with Nkx2-3 were cultured for 48 h, and then BrdU was applied to the plates for 30 min, and incorporated BrdU was detected after washing with PBS, according to the manufacturer's protocol. BrdU-positive cells were counted under a fluorescent microscope.

ChIP assay
ChIP assays were performed using M3H1 cells transfected with an Nkx2-3 vector or mock vector with a ChIP-IT express kit (Active Motif), according to the manufacturer's protocol. Cross-linked chromatin lysates were sonicated and incubated with protein G-magnet beads with a V5 antibody (M215-11, MBL) for 6 h at 4°C. Cross-linking was reversed using a reverse cross-linking buffer, and then DNA fragments were analyzed by qRT-PCR using the following primer pairs: Cdkn1a site 1, 5Ј-tccttttctgggaagtggtg-3Ј and 5Ј-tcctctggggctgatagaaa-3Ј; Cdkn1a site 2, 5Ј-cctagaaagcaagcctgtgg-3Ј and 5Ј-agcagcccataaacatccat-3Ј.

Luciferase assay
A Cdkn1a reporter plasmid was constructed by inserting a Ϫ435 to ϩ81 promoter sequence into a pGL4.15 vector (Promega). The following forward and reverse primers were used: 5Ј-tctatcctgaccctcgtgct-3Ј and 5Ј-tggggtctctgtctccattc-3Ј. A Cdkn1a mutant reporter plasmid was constructed using a QuikChange II XL site-directed mutagenesis kit (Agilent Technologies), and the following forward and reverse primers were used: 5Ј-ctgacagtccctctttgggggtcccctgtccttttctgg-3Ј and 5Ј-gactgtcagggagaaacccccaggggacaggaaaagacc-3Ј. These two reporter vectors were transfected into M3H1 cells either with a mock vector or the Nkx2-3 expression vector. A pRL-TK vector that encodes Renilla luciferase was used as an internal control to co-transfect M3H1 cells. Activity was determined 48 h later using a Dual-Luciferase reporter assay system (Promega) with a luminometer (LB942, Berthold). Firefly luciferase activity was normalized for Renilla luciferase activity as an internal control.

Statistical analysis
All experiments were repeated at least three times to confirm reproducibility. Statistical significance was determined using a two-tailed unpaired Student's t test with Prism version 6 (GraphPad Software, La Jolla, CA). One-way analysis of variance and Tukey's multiple-comparison test were utilized for quantification between multiple groups. Differences with a p value Ͻ 0.05 were considered to be statistically significant.
Author contributions-X. H., K. Y., and S. F. designed the study, performed experiments, analyzed data, and wrote the paper. K. M., C. A., K. F., T. Y., Y. C., K. S., T. I., A. Y., and T. T. participated in performing the experiments and analysis of data. S. F. and I. T. edited the paper. All authors have read and approved the final version of the manuscript.