Pitx2-Sox2-Lef1 interactions specify progenitor oral/dental epithelial cell signaling centers

Pituitary homeobox 2 (Pitx2), is a crucial transcription factor that regulates the asymmetrical development of organs (Logan et al., 1998; Piedra et al., 1998; Ryan et al., 1998; Yoshioka et al., 1998) and is crucial for tooth development. Mutations in the PITX2 gene have been found in individuals with Axenfeld-Rieger syndrome (ARS) who have developmental defects of the eyes, abdomen and teeth, indicating that it contributes to the development of these organs (Amendt et al., 1998; Semina et al., 1996). Mice with a general Pitx2 knockout are embryonic lethal with defects in the heart, lung, body wall and teeth, further demonstrating that Pitx2 plays an essential role in controlling organogenesis during development (Lin et al., 1999). In the case of teeth, the general knockout of Pitx2 leads to an arrest in development at the bud stage, involving downregulation of fibroblast growth factor 8 (Fgf8) expression in the dental epithelium and changes in the distribution of bone morphogenetic protein 4 (Bmp4) in the adjacent mesenchyme (Gage et al., 1999a; Lin et al., 1999; Lu et al., 1999). Since these mice die at early stages of embryonic development, the details of the cellular and molecular mechanisms by which Pitx2 controls early odontogenesis are unknown. However, using transgenic mice overexpressing a repressor of Pitx2, we have demonstrated that Pitx2 plays a role in dental epithelial cell differentiation into ameloblasts (Cao et al., 2013; Li et al., 2013). Although Pitx2 has long been considered as a master regulator of the transcriptional hierarchy in early tooth development, including stem cells (Pispa and Thesleff, 2003; Tucker and Sharpe, 2004), its specific role in cell differentiation and signaling has not been investigated due to embryonic lethality of global Pitx2 knockout mice. Therefore, we investigated the specific role of Pitx2 using a conditional knockout approach.

During embryonic development, stem-cell specification, and the proliferation and differentiation of transit-amplifying cells (TAs), which are a population in transition between stem cells and differentiated cells, are keys to organogenesis (Hsu et al., 2014). Organs such as teeth, hair follicles and mammary glands are derived from surface ectodermal cells via processes coordinated by interactions between the epithelium and mesenchyme (Jiménez-Rojo et al., 2012). As soon as the epithelial ectoderm receives signals from adjacent mesenchyme within a determined region, it forms a placode or a localized thickening of epithelial cells (Fig. 1B). Teeth develop from two epithelial cell populations within the dental placode: Sox2⁺ cells and Lef1⁺ cells (Sanz-Navarro et al., 2018; Sun et al., 2016) (Fig. 1B). The Sox2⁺ cells are the progenitors of various epithelial cell types that form bud- and cap-stage teeth, and the Lef1⁺ cells that are present during the placode, bud and cap stages produce growth factors that control tooth morphology. Ablation of Sox2 in the developing tooth leads to impaired proliferation of dental epithelial stem cells (DESCs) during the early stages of tooth development and prevents the renewal of DESCs in the adult (Sanz-Navarro et al., 2018; Sun et al., 2016). The ablation of Lef1 in the dental epithelium also leads to severe developmental defects of the teeth, causing an arrest during the transition from bud to cap stage (Sasaki et al., 2005; van Genderen et al., 1994). We have recently shown that conditional overexpression of Lef1 in the dental epithelium results in a new stem cell compartment and lack of dental epithelial cell differentiation (Sun et al., 2016). Furthermore, Pitx2 regulates Sox2 and Lef1 expression in the epithelium.

Tooth morphogenesis is stimulated by growth factors Fgf, Wnt, Shh and Bmp, which are produced by a cluster of cells known as the dental epithelial signaling center and by adjacent dental mesenchyme (Balic and Thesleff, 2015). Effective function of the epithelial signaling centers is crucial for tooth morphogenesis. There are two signaling centers, the initiation knot (IK), and enamel knot (EK), and the size of the IK signaling center correlates with that of the entire tooth at each stage (Ahtiainen et al., 2016) (Fig. 1B). Specifically, an epithelial signaling center at the cap stage (E13.5), the EK, is crucial for tooth morphogenesis during this stage. The EK is derived from the cells toward the posterior region of the tooth bud; it is not derived from the IK signaling center at bud stage (E12.5) (Ahtiainen et al., 2016; Du et al., 2017). Formation of the EK has been reported to be regulated by αE-cadherin-mediated restriction of YAP/TAZ activity in a cluster of non-proliferative cells (Li et al., 2016). We have also shown that Pitx2 regulates p21, which is localized to the EK during development (Cao et al., 2010a). However, further details of the mechanisms controlling dental epithelial signaling centers are not available.

Here, we report on our investigation of the mechanisms of Pitx2 function at the cellular and molecular levels during early stages of tooth development in mice. We found that conditional knockout of Pitx2 in the dental epithelium at approximately embryonic day 10 (E10.0) causes the disruption of tooth development. Pitx2 is necessary for the formation of the signaling centers, and the deletion of Pitx2 in the dental epithelium results in an arrest of tooth development. Pitx2 interacts with Sox2 and Lef1 to differentially regulate genes that form the early IK and later the EK. Pitx2 regulates the expression of Irx1- and Shh-expressing cells to initiate DESC differentiation and tooth growth at the bud stage.

To determine the cellular and molecular mechanisms underlying Pitx2 specification of the early stages of tooth development, we conditionally knocked out Pitx2 (Pitx2^cKO) in the oral and dental epithelia by crossing Pitx2^flox/flox mice to the keratin14-Cre recombinase (Krt14^Cre) mouse line. The Pitx2^flox mice were previously generated, with loxP sites flanking the fourth exon of the Pitx2 gene (Gage et al., 1999b). We tested the recombinase activity of Krt14^Cre line by crossing these mice with Rosa26^GFP mice, in which active Cre recombinase can initiate the expression of green fluorescent protein (GFP). Immunostaining of K14^CreRosa26^GFP mice for GFP at E11.5 and E14.5 showed GFP expression in the oral and dental epithelia starting at E11.5 (Fig. S1A). The Pitx2^cKO mice survived to weaning stage (postnatal day 21) and their overall body size was significantly smaller than that of their Pitx2^flox/flox (WT) counterparts (Fig. S1B,C). In addition, the Pitx2^cKO mice completely lacked teeth (Fig. 1A, Fig. S1D).

The development of teeth (lower incisor) begins with a localized thickening of the oral epithelia at E11.5 with the neural crest-derived mesenchyme condensing around the invaginating epithelia (Fig. 1B). The placode stage (E10.5-11.5) is denoted by juxtaposed expression of Sox2 (posterior region) and Lef1 (anterior region) (Sun et al., 2016). At E12.5 (bud stage) the first signaling center or initiation knot (IK) is formed (red region, Fig. 1B). At E13.5, the EK has formed and expresses Lef1, Shh and p21 (Cdkn1a). Furthermore, the Sox²⁺ cells, which are located first in the posterior region, have transitioned to a group of Sox²⁺ cells toward the anterior region (black speckled region, Fig. 1B) in the expanding tooth bud and adjacent to a transient structure termed the vestibular lamina (VL); thus, the enamel knot is more posterior. At E14.5 (the cap stage), the DESC niche is forming and condensed to a smaller region; at E15.5, both centers are separated and the EK will eventually be lost as the incisor develops. However, because the rodent lower incisor continuously grows, the DESCs in the labial cervical loop will provide the progenitor cells needed for growth of the incisor as it wears down due to gnawing and eating by the mouse and to the asymmetric nature of enamel deposition.

To identify the tissue-specific defects that result in toothless Pitx2^cKO mice, we examined Pitx2^cKO embryos at various stages of development and found early dental developmental defects (Fig. 1C, Fig. S1E). Specifically, Pitx2^cKO embryos had smaller bud stage teeth and undetectable cap and bell stage teeth compared with wild type (Fig. 1C). In addition, invagination of both the dental epithelium and vestibular lamina were delayed in the lower incisors of Pitx2^cKO embryos (Fig. 1C). We noticed similar defects in the developing molars of Pitx2^cKO embryos. Although the initiation of molar development was not affected, morphogenesis was arrested at a very early stage of tooth development (Fig. S1E).

The disruption of odontogenesis in Pitx2^cKO embryos could be due to either increased cell death or decreased cell proliferation. Analyses of cell death using cleaved caspase 3 as a marker, in E12.5 and E14.5 wild-type and Pitx2^cKO embryos, did not reveal a significant increase in apoptosis (Fig. S2). We next performed a 5-bromo-2′-deoxyuridine (BrdU) incorporation assay to determine whether the rate of cell proliferation was reduced in Pitx2^cKO versus wild-type tooth buds. BrdU was administered to pregnant mice at E11.5 and E12.5, and the embryos were harvested 2 h after BrdU injection. Immunostaining for BrdU revealed that cell proliferation was significantly reduced in the lower incisors of Pitx2^cKO embryos at E12.5 but not at E11.5 (Fig. 2A,C). The IK is shown as non-proliferating cells in the E12.5 wild-type incisor (Fig. 2A, IK, white arrow). Staining of E14.5 wild-type and Pitx2^cKO lower incisors for Ki-67 showed that, whereas all dental epithelial cells outside the EK were positive in wild-type mice, fewer cells were positive for Ki-67 in the Pitx2^cKO lower incisors (Fig. 2B,D). BrdU incorporation was used to identify DNA replication in cells and staining for the Ki-67 protein marks cell proliferation. We used both assays to validate cell proliferation. Thus, the absence of an EK structure in the E14.5 Pitx2^cKO lower incisors (Fig. 2B) resulted in lack of necessary signals for these cells to proliferate and invaginate.

Label-retaining assays were performed, delivering 5-chloro-2-deoxyuridine (CIdU) and 5-iodo-2-deoxyuridine (IdU) to wild-type and Pitx2^cKO embryos at different time points (Fig. 2E). CIdU was injected into pregnant females at E12.5, and IdU was injected into pregnant females at E14.5 and E16.5, 2 h before the embryos were harvested. In wild-type mice, the LaCL region of the lower incisor tooth germ contained cells positive for IdU (green fluorescence) at both E14.5 and E16.5; however, the lower incisors contained less CIdU (red fluorescence staining) at the later versus earlier stage (Fig. 2E, top panels). The CldU-retaining cells have moved out of the LaCL to populate the ameloblast layer and are fully differentiated cells. In the Pitx2^cKO group, the lower incisors contained fewer IdU cells and more cells that retained the red fluorescent label (Fig. 2E, bottom panels).

Generation of a 3-dimensional model of bud-stage lower incisor tooth germs of E12.5 wild-type and Pitx2^cKO serial sagittal sections revealed that, in comparison with wild-type counterparts, the tooth germs in the Pitx2^cKO group were shorter along the labio-lingual axis and thinner at the widest point of the mediolateral axis (Fig. 2F). Based on these findings, we concluded that, in the Pitx2^cKO lower incisor tooth germs, the epithelial progenitors were not actively proliferating, resulting in a smaller tooth germ structure.

Given that tooth morphogenesis was disrupted in Pitx2^cKO mice, we performed double immunostaining for Sox2 and Lef1 expression in sagittal sections of bud- (E12.5) and cap- (E14.5) stage lower incisors to determine the specific cell populations affected by Pitx2 ablation (Fig. 3A). In the bud-stage (E12.5) of lower incisor germs of the wild-type embryos, Sox2 was posteriorly expressed, whereas Lef1 was anteriorly expressed, marking a cell population that appears to be the IK and the adjacent mesenchyme. In E12.5 Pitx2^cKO sections, the Sox2⁺ cell population was smaller and the position of the IK marked by Lef1 expression was shifted to the bottom of the bud (Fig. 3A, white arrow). In cap-stage (E14.5) lower incisors of wild-type embryos, Sox2⁺ cells were concentrated in the labial cervical loops (LaCL) and Lef1⁺ cells in the enamel knot (EK) (white arrow) (Jernvall and Thesleff, 2000). However, in the Pitx2^cKO incisors at this stage, the Sox2⁺ cells did not segregate to the LaCL and Lef1⁺ cells were present in the posterior region (Fig. 3A, white arrow). We conclude that tooth development was not just delayed, but also impaired, in E14.5 Pitx2^cKO embryos, based on the changes in Sox2 and Lef1 expression, and the organization of the IK and EK signaling centers as well as a lack of a well-defined LaCL.

A lack of differentiation in Pitx2^cKO embryos was further supported by reduced Sox9 expression. In wild-type incisors of E12.5 embryos, Sox9 was expressed in the region adjacent to the oral cavity that lacked Sox2 expression (Fig. 3B, top panel, white arrow), and at E14.5, progenitors of the stellate reticulum (SR) and stratum intermedium (SI) showed Sox9 expression (Fig. 3B, top panel). In the Pitx2^cKO lower incisors, by contrast, at E12.5 the Sox9-positive signal (Sox9⁺) was reduced (Fig. 3B, bottom panel), and at E14.5 only a few Sox9⁺ cells were present in either the oral or dental epithelium (Fig. 3B, bottom panel). Statistical analysis of the number of Sox9⁺ cells revealed that they were significantly reduced in the lower incisor tooth germs of Pitx2^cKO mice at both E12.5 and E14.5 (Fig. 3C). Because Sox9 is associated with dental epithelial differentiation, these data demonstrate a reduction in differentiation of progenitor cells regulated by Pitx2.

Genetic studies have established that the Hippo pathway plays a crucial role in organ size, controlling cell number by modulating cell proliferation and apoptosis (Buttitta and Edgar, 2007; Hamaratoglu et al., 2006; Huang et al., 2005; Lai et al., 2005; Wu et al., 2003; Zhao et al., 2008a). The unphosphorylated (i.e. active) forms of YAP and TAZ associate with transcription factors (TFs) of the TEAD/TEF family in the nucleus, activating the expression of target genes and thereby promoting cell proliferation and inhibiting apoptosis (Cao et al., 2008; Zhao et al., 2008a,b). We have shown recently that Pitx2 interacts with Yap to regulate gene expression, and both Yap and Pitx2 share DNA-binding sites on gene promoters, indicating that they regulate multiple genes together in the Hippo signaling pathway (Tao et al., 2016). We asked whether unphosphorylated Yap and/or phosphorylated Yap (P-Yap) were affected in the Pitx2^cKO incisors. Immunostaining for Yap demonstrated that the Pitx2^cKO embryos had decreased Yap at E12.5 and E14.5 compared with the Pitx2^Flox controls (Fig. 4). There was little difference in the expression of P-Yap at E12.5 between the Pitx2^cKO and control embryos (Fig. 4B). However, at E14.5, P-Yap was not detected in the Pitx2^cKO lower incisor (Fig. 4D). Interestingly, the EK did not show expression of Yap in the control embryos (Fig. 4C). This is consistent with an earlier report showing that the EK contains low levels of nuclear Yap (Li et al., 2016). Thus, in addition to decreased cell proliferation in the epithelial compartment of the Pitx2^cKO lower incisor, the embryos also showed decreased Yap, consistent with the role of Pitx2 in regulating Yap interactions and a Hippo gene regulatory network. The disruption of Hippo signaling also contributes to the lack of cell proliferation and cell signaling.

We next investigated whether Pitx2 was essential for growth of the lower incisors in the adult and during homeostasis, using an inducible Pitx2 knockout mouse model. We crossed Pitx2^flox/flox mice with Rosa26^CreERT to generate Rosa26^CreERTPitx2^flox/flox (Pitx2^icKO) mice. We allowed the Rosa26^CreERTPitx2^flox/+ (Pitx2^iHet) and Pitx2^icKO progeny to mature and then injected them with tamoxifen (TAM) to ablate Pitx2 expression (Fig. S3A). Eight days after TAM injection, the left lower incisor of each mouse was clipped and the difference in length of the left and right lower incisors was measured and recorded (Fig. S3B); 3.5 days later, the difference in length between the lower incisors of each mouse was recorded. In these experiments, inducible deletion of Pitx2 did not affect growth of the adult lower incisor (Fig. S3B,C), although levels of Pitx2 transcripts in tissues surrounding the cervical loop region were significantly lower in the TAM-treated Pitx2^icKO versus Pitx2^iHet mice (Fig. S3D). We next analyzed the enamel layer, under a scanning electron microscope, for dental enamel defects and found disruption of enamel rod organization in the Pitx2^icKO but not Pitx2^iHet mice (Fig. S3E). Thus, in adult mice lower incisors, after the signaling centers have been patterned and LaCL (DESCs) formed, Pitx2 is not required for the proliferation of adult DESCs but is required for dental epithelial cell differentiation and amelogenesis during homeostasis (Cao et al., 2013, 2010b; Li et al., 2013).

We next set out to identify the molecular mechanism by which Pitx2 regulates tooth morphogenesis during early stages of tooth development by investigating the localization of β-catenin, and the expression of Fgf8 and Bmp4 transcript levels. The immunostaining for β-catenin in bud-stage lower incisor tooth germs showed more membrane-associated β-catenin in wild-type compared with Pitx2^cKO lower incisors (Fig. S4). Interestingly, the epithelial cells in the Pitx2^cKO at E12.5 appeared less polarized compared with wild type. To determine whether β-catenin nuclear localization was impaired in the E12.5 Pitx2^cKO embryos, sections were stained using the PY489-β-catenin antibody and co-stained for E-cadherin (Fig. S4B). Nuclear β-catenin was observed in the WT invaginating tooth bud, but was decreased in the Pitx2^cKO tooth bud (Fig. S4B). Furthermore, E-cadherin expression revealed that cells were less polarized in the Pitx2^cKO tooth germ compared with wild type (Fig. S4B). These data suggest that a decrease in polarized cells and nuclear β-catenin affect epithelial invagination.

The in situ hybridization for Fgf8 and Bmp4 transcripts revealed that Fgf8 levels were decreased in Pitx2^cKO incisors compared with wild-type embryos, and Bmp4 transcript levels were slightly decreased (Fig. S5). In addition, assessment of the levels of Bmp2, Shh and Wnt10b transcripts (in situ hybridization), and the p21 protein (immunostaining) in the signaling centers of E12.5 incisors showed that only Shh was reduced in the Pitx2^cKO (Fig. 5A). There is a consensus Pitx2-binding site 3505 bp upstream of the Shh transcription start site, allowing Pitx2 to activate Shh expression. The expression domains of Shh and p21 mark the IK at E12.5 (Ahtiainen et al., 2016). The expression of Fgf4, Shh and Wnt10b transcripts, and immunostaining of p21, in E14.5 (cap-stage) lower incisors revealed positive signals for Fgf4, Shh and Wnt10b in wild-type mice but not in their Pitx2^cKO counterparts (Fig. 5B). Moreover, there were very few p21-positive cells in the posterior region of the Pitx2^cKO lower incisor (Fig. 5B). Taken together with the immunostaining results for the early EK marker Lef1 (Fig. 3A) and p21 (Jernvall et al., 1998; Jernvall and Thesleff, 2000), and the in situ hybridization results for the mature EK markers Fgf4, Shh and Wnt10b (Fig. 5B) in E14.5 lower incisors, these results led us to conclude that formation of the EK was disrupted in the Pitx2^cKO lower incisors.

Given that Shh levels were reduced in the dental epithelial signaling centers of Pitx2^cKO lower incisors, we tested the hypothesis that downstream targets of the Shh signaling pathway were downregulated. We show that Gli1 expression was decreased in E12.5 and E14.5 Pitx2^cKO lower incisors compared with wild type (Fig. 5C). Cyclin D2, another downstream target of the Shh signaling pathway, was also decreased in E12.5 Pitx2^cKO lower incisors (Fig. 5C); interestingly, cyclin D2 marks the wild-type IK (Fig. 5C, yellow arrow). Notably, the cyclin D2 reduction was also apparent in the Sox2⁺ cell population and appeared to be decreased compared with those in the wild-type group in cells of the IK (Fig. 5C). These findings support the conclusion that Pitx2 regulates cells expressing Shh, p21 and Ccnd2 in the IK dental epithelial signaling center at E12.5.

Recently, we reported that Irx1 regulates lung progenitor cell differentiation and marks alveolar type 2 cells; in dental development, it also marks cells of the stratum intermedium (SI) and outer enamel epithelium (OEE) (Yu et al., 2017). In a previous RNA-seq analysis, we found that Irx1 levels were higher in craniofacial tissues of Pitx2c-overexpressing transgenic mice compared with wild-type embryos. E14.5 tooth germs were checked for co-expression of Pitx2 and Irx1. While Irx1 expression is highest in the OEE and dental epithelium, Irx1 is also expressed at low levels in the EK, and both Pitx2 and Irx1 are co-expressed in the OEE and dental epithelium (Fig. S7). However, Irx1 was not expressed in the Pitx2^cKO incisor. In addition, a Pitx2-binding site was identified 1.6 kb upstream of the Irx1 transcription start site (Fig. 6A). We therefore investigated, both in vitro and in vivo, the possibility that Pitx2 regulates Irx1 transcription. Chromatin immunoprecipitation (ChIP) analysis revealed that endogenous Pitx2 protein binds the Irx1 promoter in LS-8 cells (Fig. 6A,B). Pitx2 and Irx1 directly interact with each other, as shown by co-immunoprecipitation assays (Fig. 6C). LS-8 epithelial cells endogenously express Pitx2 and overexpress Irx1; the Pitx2 antibody was used to pull down the Pitx2-Irx1 protein complex. We did not perform an IP experiment using the Irx1 antibody to pull down Pitx2 as LS-8 cells do not endogenously express high levels of Irx1. In a gel-shift assay, Pitx2 protein bound to the Irx1 promoter probe with the Pitx2-binding motif (Fig. 6D), and a luciferase assay showed that Pitx2 activated an Irx1 promoter construct in LS-8 cells (Fig. 6E). Finally, immunostaining for Irx1 in lower incisors revealed that protein levels were lower in E12.5 and E14.5 Pitx2^cKO embryos than in their wild-type counterparts (Fig. 6F). Taken together, these data show that Pitx2 directly regulates Irx1 transcription during tooth development.

Interestingly, we found that the Sox2 protein can inhibit Pitx2 protein binding to the Irx1 promoter in a dose-dependent manner (Fig. 6D) and negatively regulates the transcriptional activity of Pitx2 in LS-8 cells (Fig. 6E). We also demonstrated this regulation in mice by immunostaining for Irx1 and Sox2 protein in the lower incisors. Sox2 and Irx1 were colocalized in the posterior part of the dental epithelium in E12.5 wild-type lower incisors and then were expressed in distinct cell populations (Sox2 in LaCL and Irx1 in OEE & SR) in E14.5 wild-type lower incisors (Fig. 6F). Irx1 was relatively higher in E14.5 lower incisor tooth germs compared with E12.5 lower incisor tooth germs, indicating that the lack of Sox2 expression in OEE cells at E14.5 allowed Pitx2 to activate Irx1 expression. Sox2 also inhibits the synergistic activation of the Lef1 promoter by Pitx2 and Lef1 in a dose-responsive mechanism (Fig. S6). These new molecular mechanisms demonstrate that Sox2 inhibits Pitx2 transcriptional activity and Lef1 expression to maintain the Sox2⁺ domain cells such as the LaCL stem cell niche (Sun et al., 2016). However, Sox2 is not expressed in the EK signaling center, which expresses Lef1. Pitx2 expression is required to pattern the cells for each signaling center; Lef1 expression is later restricted to the EK and Sox2 expression is restricted to the LaCL stem cell niche.

Pitx2 is the earliest transcriptional marker for tooth development, and is required for the differentiation of stem cells and progenitors in several organs (Gage et al., 1999a; Lin et al., 1999; Lu et al., 1999; Mucchielli et al., 1997). In this paper, we show that Pitx2 regulates development of the signaling centers within the dental epithelium during tooth development. Furthermore, Pitx2 regulates the cells expressing Shh and promotes the formation of the enamel knot (EK) during the cap stage of tooth development. Pitx2 acts as a checkpoint for formation of the signaling centers and later for DESC differentiation to positively affect the accumulation of rapidly proliferating transit-amplifying cells during the early stages of tooth development. We have previously shown that Pitx2 is expressed in the DESCs of the murine lower incisor LaCL and the transit amplifying cells (Cao et al., 2013). In addition to controlling signals from the dental epithelial signaling centers (IK and EK), we show a new mechanism whereby Sox2 interacts with Pitx2 to negatively regulate Irx1 transcription in Sox2⁺ cell lineages, thereby contributing to the early stages of tooth morphogenesis. These types of protein interactions directly regulate and control gene expression in specific cells of the developing tooth organ.

Interestingly, cell proliferation in Pitx2^cKO lower incisors was not affected at E11.5; however, it was significantly decreased at later embryonic stages, indicating that Pitx2 influences cell proliferation during embryonic tooth development. If Pitx2 regulated cell cycle progression directly, cell proliferation should have been downregulated in the Pitx2^cKO lower incisor tooth germs at all stages of development. It is possible that the E11.5 and E12.5 dental epithelium of these mice retained sufficient functional Pitx2 protein to maintain cell proliferation, despite the fact Pitx2 transcripts were ablated at E11.5 by Krt14^Cre (demonstrated by GFP immunostaining in Rosa26^GPF mice). However, it is more likely that Pitx2 regulates early progenitor cell proliferation in the developing tooth, and that the reduced cell proliferation in these mice at bud and cap stages is a consequence of an arrest of stem-cell proliferation and differentiation. In the tooth, stem cells give rise to TAs, which rapidly proliferate. In the lower incisors of E14.5 (cap stage) Pitx2^cKO embryos, Sox2⁺ stem cells failed to differentiate into such cells. In addition, the IK and EK did not form in Pitx2^cKO lower incisors, and given that the dental epithelial signaling centers provide Shh, Fgf4, Bmp2 and Wnt10b for the proliferation and differentiation of dental epithelial cells, it is reasonable to propose that cell proliferation is significantly downregulated in E14.5 Pitx2^cKO lower incisors due to a lack of an EK dental epithelial signaling center.

The fact that we did not observe a difference in growth of the lower incisor in the inducible Pitx2 knockout indicates that Pitx2 is not required for this process in adult mice after development of the signaling centers has taken place. Interestingly, in individuals with Axenfeld-Rieger syndrome (ARS) with mutations in PITX2, their surviving teeth are brittle and fall out (Amendt et al., 2000). Our new data support a role for Pitx2 in late stage enamel formation, supporting a mechanism for brittle teeth in individuals with ARS. Furthermore, our demonstration that Pitx2 directly regulates the expression of Irx1, a gene that regulates the differentiation of Sox2⁺ stem cells into outer enamel epithelial cells (Yu et al., 2017), leads us to conclude that Pitx2 controls the differentiation of dental epithelial progenitor cells during tooth morphogenesis at later stages and not proliferation. We speculate that, in the continuously growing murine lower incisor during homeostasis, Sox2 expression and other factors, such as Shh, Bmi1, Lrig and Tbx1, regulate the stem cell and transient-amplifying cell proliferation (Gao et al., 2015), rather than Pitx2. However, during incisor homeostasis, a recent report clearly demonstrated that, upon injury, repair of the incisor relied upon cells from the SI that repopulated the ameloblast layer to facilitate rapid tooth repair (Sharir et al., 2019). With these new data and the fact that Pitx2 and Irx1 expression is shifted to the ameloblasts, OEE and SI during homeostasis, Pitx2 and Irx1 may have a unique role in this repair process.

The IK is a transient structure that does not express Sox2 but is derived from Sox2⁺ cells located in the posterior region of the placode (Du et al., 2017). The IK does express Shh, Bmp2, Wnt10b and p21 (Ahtiainen et al., 2016), also shown in this report, but also cyclin D2 and Lef1. In the Pitx2^cKO E12.5 embryos, the IK appears to be positioned in the middle of the tooth bud with reduced expression of the IK markers. The attenuation of Wnt signaling may be required for the clearance of the IK (Ahtiainen et al., 2016) and Lef1 expression is absent from the IK region after E12.5 and reappears in a de novo activation to mark the EK at E13.5.

The altered expression of Shh, Bmp2, Fgf4 and Wnt10b suggests that Pitx2 regulates the development of the EK at the cap stage in Pitx2^cKO embryos. Furthermore, p21 and Lef1 expression, early markers of dental epithelial signaling centers in the teeth (Jernvall and Thesleff, 2000; Keränen et al., 1999), revealed a small cell population (IK) that is positioned toward the anterior end of the developing tooth bud. This cell population (Sox2⁻ cells) does not represent progenitors of the EK, as the incisor EK cells are not derived from the IK (Ahtiainen et al., 2016; Du et al., 2017). The EK has been reported to be induced from de novo progenitor cells in the tooth bud (Ahtiainen et al., 2016; Du et al., 2017). Thus, Pitx2 may be involved in initiating the development of the EK, as we have shown that Pitx2 activates Lef1 and p21 expression (Amen et al., 2007; Cao et al., 2010a). Furthermore, the EK structure is not formed in the Pitx2^cKO embryos. Given that formation of the EK is the key event in the transition from bud to cap stage, the unique function of Pitx2 makes this transcription factor an extremely important determinant of tooth development.

The Shh signaling pathway plays crucial roles during the early stages of tooth development. Shh is expressed in the EK of both molars and incisors, and the BMP-Shh signaling network has been reported to control the fate of Sox2⁺ epithelial stem cells in molars (Li et al., 2015). In the tooth, Shh knockout leads to severe disruption of morphogenesis, leading to smaller teeth and disorganization of both the ameloblast and odontoblast layers (Dassule et al., 2000). Notably, overexpression of Shh in the dental epithelium also led to an arrest of tooth development at the bud stage (Cobourne et al., 2009), indicating that tooth development is very sensitive to Shh levels. The demonstration that Shh expression in the lower incisors of E12.5 Pitx2^cKO mice was decreased, causing reduced invagination of the dental epithelium, delays in the formation of bud stage teeth and an arrest in the differentiation of Sox2⁺ epithelial stem cells is consistent with the known role for Shh in tooth development. Moreover, Gli1 is transcriptionally activated by Shh in mice (Hynes et al., 1997), and Gli1 was downregulated in Pitx2^cKO lower incisor tooth germs, which is consistent with the other effects on Shh signaling. In addition, tooth development is arrested at the bud stage when both Gli2 and Gli3 are knocked out in mice (Hardcastle et al., 1998), which is a phenotype similar to the Pitx2^cKO mice.

Pitx2 operates in a gene regulatory network involving Sox2⁺ and Sox2^– cells. Sox2 protein in turn interacts with Pitx2 and represses Pitx2 transcriptional activity, providing a unique cellular and molecular mechanism of Pitx2-mediated regulation. Here, we show that this mechanism allows the formation of the signaling centers and DESCs, as well as their differentiation during tooth development.

Interestingly, the expression domains of Sox2, Lef1 and Irx1 overlap with Pitx2 during early tooth development. This raises the question of how Sox2, Lef1 and Irx1 are differentially expressed and regulated by Pitx2? We have previously shown that Pitx2 regulates Sox2 and Lef1 expression (Amen et al., 2007; Sun et al., 2016; Vadlamudi et al., 2005). Furthermore, we have demonstrated that Hmgn2, a small high-mobility group protein, can directly interact with Pitx2 to remove it from DNA and inhibit Pitx2 transcriptional activity (Amen et al., 2008). Sox2 contains a Hmg domain that represses Pitx2 transcriptional activation of Sox2, Lef1 and Pitx2 auto-regulation (Sun et al., 2016). We show, in this report, that Sox2 can inhibit Pitx2 from binding the Irx1 promoter as a direct mechanism to control Pitx2 regulation of Irx1 in a Sox2 expression domain during development. We have shown previously that Pitx2 directly interacts with Lef1 to synergistically activate the Lef1 promoter (Vadlamudi et al., 2005). As a mechanism to regulate Lef1 in the signaling centers, we show a dose response relationship between Sox2-mediated inhibition of the synergistic activation of the Lef1 promoter by Pitx2-Lef1. Thus, Sox2 controls the transcriptional activity of Pitx2 and represses Pitx2 activation of Lef1, Pitx2, Irx1 and other factors regulated by Pitx2. Therefore, in Sox2 expression domains, it acts to inhibit Pitx2-mediated transcriptional activation, but outside the Sox2 domains (the IK and EK), Pitx2 then activates a gene regulatory network required for IK and EK formation, and progenitor cell differentiation. A potential mechanism for the clearance of the IK after E12.5 resides in the growing and changing cell population expressing Sox2. As the Sox2⁺ cells transition from the posterior region of the placode and early tooth bud to the LaCL at E14.5, Sox2 expression would attenuate Pitx2 transcriptional activity, which would result in decreased Lef1, p21 and Shh expression. This would clear the IK structure and favor the re-establishment of the EK away from the Sox2⁺ cell domain. Sox2 expression domains are also regulated and maintained by other mechanisms, such as signaling factors and microRNAs. Because Lef1 and Sox2 expression domains appear to be nonoverlapping, Lef1 and Shh with other factors may also be controlling Sox2 expression (or lack of) in the IK and EK.

In summary, we have identified a novel molecular mechanism whereby Pitx2 regulates the expression of Irx1 and cells expressing Shh to control the early stages of tooth morphogenesis. Pitx2 is expressed throughout the placode, bud and cap stage during tooth development (Cao et al., 2013; Hjalt et al., 2000). Specifically, our data show that Pitx2 regulates the differentiation of progenitors of the signaling centers. At the E12.5 bud stage, Pitx2 promotes Lef1, p21 and Shh expression within the IK signaling center of the lower incisor, and these cells do not express Sox2. At the same time, Pitx2 directly promotes Irx1 transcription and at later stages differentiation to Sox9⁺ cells and other cell types. At the cap stage, Pitx2 directly regulates Irx1 transcription in progenitors of the outer enamel epithelium, stratum intermedium and the stellate reticulum, promoting their differentiation. At this stage, Pitx2 is essential for the formation of the EK: the signaling center within the cap-stage lower incisors (Fig. 7). Sox2 directly interacts with Pitx2 to attenuate Pitx2 transcriptional activity and activation of Sox2, Irx1, Shh and Lef-1, when these two proteins share domains in the developing incisor. We speculate that this is a method to control Sox2⁺ cells and separate those cells from the other types of cells expressing Pitx2. Our findings further demonstrate that Pitx2 is crucial for the communication between the DESCs and the epithelial signaling centers, suggesting that it is a master regulator of tooth development.

Mouse maintenance and mouse-related procedures were performed using protocols approved by the Institutional Animal Care and Use Committee of the University of Iowa. The Pitx2^flox/flox mouse strain has been previously described (Gage et al., 1999b) and was a generous gift from the laboratory of James F. Martin (Baylor College of Medicine, Houston, TX, USA). The K14^Cre mouse strain has also been described previously (Cao et al., 2010b; Dassule et al., 2000). The Rosa26^CreERT [B6.129-Gt(ROSA)26Sor^{tm1(cre/ERT2)Tyj}/J] (stock number 008463) and Rosa26^GPF [B6.129(Cg)-Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo}/J] (stock number 007676) mouse strains originated from the Jackson Laboratory, and were generous gifts from the laboratory of John Engelhardt (The University of Iowa, Iowa, USA). Mice were genotyped using the primers listed in Table S1 following the standard genotyping protocol. Embryos were staged by checking for vaginal plugs in the crossed females, with the day of vaginal plug formation defined as E0.5.

The following primary antibodies were used for immunostaining: anti-IRX1 (Sigma, HPA043160, 1:100-1:300), anti-Lef-1 (Cell Signaling, 2230, 1:150), anti-Sox2 (R&D Systems, AF2018, 1:150), anti-Sox9 (Millipore, AB5535, 1:400), anti-GFP (Abcam, ab290, 1:1000), anti-Ki-67 (Abcam, ab15580, 1:500), anti-cleaved caspase-3 (Cell Signaling, 9661, 1:500), anti-BrdU (Abcam, ab6326, 1:1000), anti-CIdU (Accurate Chemical, OBT0030, 1:250), anti-IdU (Roche, 11170376001, 1:250), anti-p21 (BD Pharmingen, 556430, 1:100), anti-PY489-β-catenin (DSHB, University of Iowa), anti-β-catenin (Upstate, 06-734, 1:250), anti-Pitx2 (R&D Systems, AF7388, 1:500), anti-Myc (Cell Signaling, 9B11, 1:500) and anti-cyclin D2 (Santa Cruz, sc-593, 1:400). The following secondary antibodies were used for immunostaining: Alexa Fluor488 donkey anti-rabbit IgG (Invitrogen, A21206, 1:500), Alexa Fluor488 donkey anti-mouse IgG (Invitrogen, A21202, 1:500) and Alexa Fluor594 donkey anti-mouse IgG (Invitrogen, A21203, 1:500). Other staining reagents used were: DAPI (Invitrogen, D1306, 1 μg/ml), BrdU (Invitrogen, 00-0103), CIdU (Sigma, C6891) and IdU (Sigma, 17125).

A 2200 bp Irx1 promoter containing the Pitx2-binding site was cloned and inserted into the pTK-luc vector. The LEF-1 promoter has been previously described (Amen et al., 2007). Standard transient transfection by electroporation and luciferase assay were carried out in the ameloblast-derived LS-8 line (Chen, 1992) according to the methods previously described (Cao et al., 2013). LS-8 cultures (Chen, 1992) were seeded in flasks or plates in Dulbecco's Modified Eagle's Medium (DMEM) with 10% fetal bovine serum, and fed at least 24 h prior to the experiments.

Mouse embryos were harvested and washed with ice-cold 1× PBS, then fixed with 4% PFA and washed with 1× PBS three times (10 min each). Embryos were then taken through a standard dehydration and paraffin-embedding process. Sections of 5-8 µm were cut and then subjected to staining with either standard Hematoxylin and Eosin (H&E) or antibodies. For immunostaining, sections were incubated in citrate buffer (pH=6.0) in a 100°C water bath for 20 min, blocked with 10% donkey serum and incubated with primary antibodies overnight in 4°C. They were washed with 1× PBS, incubated with Alexa Fluor488 and 594 secondary antibodies, and stained with DAPI. Images were captured using a Nikon Eclipse microscope or a Zeiss 700 confocal microscope.

BrdU, CIdU and IdU were administered to pregnant females by intraperitoneal injection at different time points. Following the standard tissue processing steps described above, sections were treated with 2 M HCl for 40 min and neutralized with 1× PBS. They were then treated with citrate buffer and subjected to standard immunostaining as described above. Details of the methods used for IdU and CIdU double staining have been described previously (Sun et al., 2016).

PFA-fixed paraffin-embedded tissue sections were used for in situ hybridization. Embryo sections were cut into segments of 8 µm, and subsequently prepared as described previously (Gregorieff and Clevers, 2015). Briefly, sections were processed by a standard dewaxing and rehydration protocol, and rinsed twice in DEPC-treated water. Sections were next treated with 0.2 N HCl for 15 min at room temperature and incubated with proteinase K in 1× PBS buffer. Sections were then washed with 1× PBS quickly, and post-fixed for 5 min with 4% PFA. They were then treated with fresh acetic anhydride solution twice, and rinsed with 5× SSC buffer (pH 7.5). Sections were incubated with hybridization solution at 65°C for prehybridization, and then with hybridization solution containing 500 ng/ml of digoxigenin-labeled probes at 62-68°C for 36 h. Sections were washed with 50% formamide/2× SSC buffer (pH 4.5) at 65°C, blocked with blocking buffer and incubated with sheep anti-digoxigenin antibody (Roche, 1:2000) overnight at 4°C. Last, sections were incubated with NBT/BCIP solution, and images were taken using a bright-field microscope. The digoxigenin-labeled probes were generated using a DIG RNA Labeling Kit (Roche, 11175025910). Primers used to generate probes are listed in Table S2.

Pitx2^iHet and Pitx2^icKO mice at ∼40 days of age received daily peritoneal injections of tamoxifen (2 mg/10 g body weight) for 10 days. After 8 days of injection, the left lower incisor was clipped and the difference in length with the contralateral incisor was measured using a caliper. Tamoxifen was injected daily for an additional 2 days, and at day 3.5 post-clipping, the mice were sacrificed and the teeth were measured again.

For analysis of the enamel, incisors were isolated by hemimandibular section. All incisors were fixed in 4% paraformaldehyde overnight, after which they were rinsed three times. One of the incisors from each mouse was immediately subjected to gradient alcohol dehydration, whereas the other was prepared for analysis of a cross-section. Specifically, the incisors were fractured at the level of the gingival margin, with the line of fracture parallel to the enamel surface, after which hydrochloric acid was applied to the enamel for 30 s and samples were rinsed three times. In both sample sets, following alcohol dehydration, critical-point drying was performed by submerging the sample in hexamethyldisilazane (HMDS) for 1 h and drying overnight. Samples were mounted on stubs and sputter-coated with gold/palladium alloy. Gross morphology of the complete incisor was assessed using a Hitachi S-4800 microscope in secondary electron mode, at 3 kV and an emission current of 7400 nA. Enamel cross-sections from fractured teeth were imaged in secondary electron mode, at 5 Kv and an emission current of 10,600 nA. All images were taken under high vacuum. The working distance ranged from 8 mm to 10.8 mm.

The procedure for the ChIP assay has been described previously (Sun et al., 2016), and the ChIP Assay Kit (Zymo Research, Zymo-Spin ChIP Kit, D5210) was used for the analysis. LS-8 cells were washed twice with ice-cold 1× PBS solution and crosslinked (1% formaldehyde, room temperature, 7 min). They were then subjected to three rounds of sonication (6 s duration, 25% of maximum amplitude), which caused lysis and the shearing of genomic DNA into ∼200-1000 bp fragments. DNA and protein complexes were then immunoprecipitated using 5 µg Pitx2 antibody (Capra Sciences, PA-1023). In the negative control, the same amount of normal rabbit IgG replaced the Pitx2 antibody. Irx1 primers used for PCR were as follows: primer 1 forward, 5′-CAAGAAGAGGTCACAATTAGGAGCT-3′; primer 1 reverse, 5′-CTTGG-AATGTCAGGACCTGCTTT-3′; primer 2 forward, 5′-TCGTGGGAGACTCAAAGACAGG-3′; primer 2 reverse, 5′-AGACGCGGAGAGTCAACACGA-3′. All PCR products were visualized on a 1.5% agarose gel, and their identities were confirmed by sequencing. LS8 cells were transfected with an Irx1 overexpression plasmid in pCDNA3.1 or empty pcDNA3.1 using PEI. Transfected cells were harvested, suspended in IP lysis buffer, sonicated and incubated with 1 μg of Pitx2 or normal rabbit IgG in the presence of Protein A Magnabeads (Fisher Scientific) for 2 h. The samples were then washed, resuspended in 4× Laemmli buffer and probed on a western blot. Myc-tag antibody listed above (1:200) was used for detection and an IgG-binding-depleted anti mouse-HRP was used (Abcam, ab131368, 1:4000).

Fluorescent dye-labeled probe and purified mouse proteins were used for EMSA according to the protocol from the Odyssey Infrared EMSA Kit (LI-COR, P/N: 829-07910). Oligonucleotides from a region of the Irx1-promoter including a Pitx2-binding motif were generated by the Iowa Institute of Human Genetics of The University of Iowa. The sequences were: 5′-/IRdye700/TACCCTCCAGACTAAATCCAAGCAGGAAAGCAG-3′ and 5′-TACCCTCCAGACTAAATCCAAGCAGGAAAGCAG-3′. Complementary oligonucleotides were annealed and 50 nM of probe was added into the binding reaction. The purified mouse Pitx2c and Sox2 proteins were prepared as previously described (Tao et al., 2016), and 5 µg of Pitx2c and 0-10 µg of Sox2 were added to the binding reaction. After 20-30 min of incubation at room temperature, 2 µl of 10× Orange Loading Dye was added to the 20 µl EMSA reaction. The entire binding reaction was then loaded onto a 6% native polyacrylamide gel and run until the dye reached the bottom of the gel. Finally, the gel was scanned using an Odyssey scanner.

Serial sagittal sections from E12.5 wild-type and Pitx2^cKO lower incisors were stained with DAPI, and serial images were generated using a Zeiss 700 confocal microscope. Then images were automatically aligned using the StackReg plug-in for ImageJ, and the final 3-dimensional reconstructions were rendered using the Imaris software (Bitplane).

For each condition, a minimum of three experiments were carried out and data are presented as mean±s.e.m. An independent two-tailed t-test was used to determine the significance of differences between wild-type and Pitx2^cKO groups.

We thank Tom Moninger for the help with the 3D reconstruction, Dr James F. Martin for the Pitx2^flox mice, Dr Bernd Fritzsch for sharing the protocol of designing in situ hybridization probes, and Dr John Engelhardt for the Rosa26^CreERT and Rosa26^GFP mice. We thank Dr Christine M. Blaumueller for scientific editing and advice on the manuscript; and Drs Hank Qi, Martine Dunnwald, Robert Cornell, Fenghuang Zhan and Fang Lin, and all of the members of the Amendt laboratory for comments and ideas. Some data for the publication was the result of a PhD thesis submitted by Wenjie Yu in 2018 at the University of Iowa.