Loss of Stemness, EMT, and Supernumerary Tooth Formation in Cebpb−/−Runx2+/− Murine Incisors

Adult Cebpb KO mice incisors present amelogenin-positive epithelium pearls, enamel and dentin allopathic hyperplasia, fewer Sox2-positive cells in labial cervical loop epitheliums, and reduced Sox2 expression in enamel epithelial stem cells. Thus, Cebpb acts upstream of Sox2 to regulate stemness. In this study, Cebpb KO mice demonstrated cementum-like hard tissue in dental pulp, loss of polarity by ameloblasts, enamel matrix in ameloblastic layer, and increased expression of epithelial-mesenchymal transition (EMT) markers in a Cebpb knockdown mouse enamel epithelial stem cell line. Runx2 knockdown in the cell line presented a similar expression pattern. Therefore, the EMT enabled disengaged odontogenic epithelial stem cells to develop supernumerary teeth. Cebpb and Runx2 knockdown in the cell line revealed higher Biglycan and Decorin expression, and Decorin-positive staining in the periapical region, indicating their involvement in supernumerary tooth formation. Cebpb and Runx2 acted synergistically and played an important role in the formation of supernumerary teeth in adult incisors.


Results
Cebpb is associated with the maintenance of stemness in odontogenic epithelial stem cells and the labial cervical loop epithelium of adult maxillary incisors. To investigate the mechanisms underlying mineralized tissue formation around the labial cervical loop epithelium, such as those seen in supernumerary tooth structures, we first performed a detailed histological evaluation of adult Cebpb −/− mice (129 Sv background) ( Fig. 1A-C). We observed numerous amelogenin-positive epithelium pearls near the periapical tissue (Fig. 1D), enamel and dentin ectopic hyperplasia in the periapical tissue ( Fig. 1B-E,H), as well as miniaturization of the labial cervical loop epithelium that is specialized epithelial structure for stem cell niche (Fig. 1G,K-N). Based on these results, we concluded that odontogenic epithelial stem cells (OESCs) within the labial cervical loop epithelium lost their stemness in Cebpb −/− mice. Consequently, these OESCs differentiated into ameloblasts and secrete enamel matrix protein and enamel ectopically. Ectopic enamel formation was observed mainly in the maxilla, and was also observed in the mandible with low frequency (in a 129 Sv background) (Fig. 1H,I) 1 . We therefore investigated SRY (sex determining region Y)-box 2 (Sox2) expression in the labial cervical loop epithelium of maxillary incisors in Cebpb −/− mice. Sox2 localized to the inner and outer enamel epithelium, particularly the transit-amplifying zone, as well as the stellate reticulum (Fig. 1J), confirming a previous report 28 . There were significantly fewer Sox2-positive cells in the labial cervical loop epithelium of adult Cebpb −/− mouse incisors than in wild-type animals ( Fig. 1J-N). These findings suggest that Cebpb maintains Sox2-positive OESCs in the labial cervical loop epithelium during postnatal life.
Abrogation of Cebpb and Runx2 works synergistically in supernumerary tooth formation around the labial cervical loop epithelium in adult Cebpb −/− Runx2 +/− mice. We had previously observed mineralized tissue, akin to that of supernumerary tooth structures, around the labial cervical loop epithelium in adult Cebpb −/− mice 1 . Although Runx2 −/− mice die shortly after birth due to the absence of bone formation 22 , some Runx2 +/− mice, showed lingual bud formation in the embryonic stage. Therefore, to determine whether Cebpb and Runx2 acted synergistically and played an important role in supernumerary tooth formation, we performed a series of histological studies of maxillary incisor formation in Cebpb +/+ Runx2 +/+ , Cebpb −/− Runx2 +/+ , Cebpb +/+ Runx2 +/− , and Cebpb −/− Runx2 +/− mice ( Fig. 2A-D). We first investigated the synergistic action of Cebpb and Runx2 during lingual bud formation. We previously reported that Usag-1 and Runx2 acted in an antagonistic manner in lingual bud formation at embryonic day 15 (E15) 16 . Here, the Cebpb +/+ Runx2 +/− (F 2 background) maxillary incisor germ displayed lingual budding at E15 (Fig. 2C); however this genotype was smaller incidence rate of budding (Fig. 2E) and was significantly fewer cell number of budding than Cebpb −/− Runx2 +/− (Fig. 2F). These results support the hypothesis that Cebpb and Runx2 act synergistically in OESCs during embryonic development. In contrast, the lingual bud was highly enriched with Sox2-positive OESCs. At present, we could not find out how OESCs within the lingual bud lost their stemness during the embryonic stage. Next, we analyzed whether Cebpb and Runx2 acted synergistically during supernumerary tooth formation around the labial cervical loop epithelium in adult mice (Fig. 3). Five out of eight (62.5%) adult Cebpb −/− Runx2 +/+ (F 2 background) mice presented mineralized tissue around the labial cervical loop epithelium, akin to that of supernumerary tooth structures, roughly reflecting the value reported in Cebpb −/− (129 Sv background) mice (66.7%) 1 (Table 1). In severe cases, no labial cervical loop epithelium structures were seen in the aberrant incisors. These results indicate that there is no phenotypic difference between F 2 and 129 Sv mouse background. Four (33%) out of twelve Cebpb −/− Runx2 +/− 3-month-old (F 2 background) mice presented aberrant maxillary incisors, with developing or mature ectopic supernumerary teeth in the periapical tissue and dental pulp ( Table 1, Fig. 3A-D,H-K). The incidence of short incisors was higher in Cebpb −/− mice (129 Sv or F 2 background) than in Cebpb −/− Runx2 +/− mice (F 2 background), whereas the incidence of supernumerary teeth showed the opposite trend (Table 1). These data confirm the synergistic action of Cebpb and Runx2 in supernumerary tooth formation around the labial cervical loop epithelium during postnatal life. Moreover, this is the first report on supernumerary tooth formation in Cebpb −/− Runx2 +/− mice.  around the labial cervical loop epithelium. Compared to wild-type mice (Fig. 4A), differentiated ameloblasts in the maxillary incisor were deranged and lost their cell polarity in adult Cebpb +/+ Runx2 +/− animals (F 2 background) (Fig. 4B). Furthermore, loss of cell polarity and surface roughness of the enamel matrix were visible also in Cebpb −/− Runx2 +/+ and Cebpb −/− Runx2 +/− mice ( Fig. 4C-E). Concomitant to the EMT process is the disappearance of the apical-basal polarity of epithelial cells 29 . Therefore, we performed immunostaining for the EMT marker-N cadherin-in the maxillary incisors of 3-month-old Cebpb +/+ Runx2 +/− and Cebpb −/− Runx2 +/− mice. Non-polarized differentiated ameloblasts and the cell component surrounding ectopic enamel formation and cementum-like hard tissue within the dental pulp in the periapical region were positive for presence of N-cadherin ( Fig. 4F-H). Thus, based on these morphological changes, we suggest that Cebpb and Runx2 knockdown contributed to EMT in odontogenic epithelial cells of the maxillary incisor. Taken together, supernumerary tooth formation around the labial cervical loop epithelium of adult maxillary incisors is dependent on both Cebpb knockdown-induced loss of stemness in OESCs and EMT of odontogenic epithelial cells in Runx2 +/− and/or Cebpb −/− mice.

Reduction of
Reduced Cebpb mRNA expression leads to a decrease of Sox2, Runx2 knockdown inhibits adhesion molecules and EMT in OESCs. To confirm the above-mentioned function of Cebpb and Runx2 in odontogenic epithelial cells of the adult labial cervical loop epithelium, in vitro knockdown experiments were performed using mHAT9d cells. These cells are derived from the labial cervical loop epithelium of a mouse incisor and could undergo EMT 30 . Transfection efficiency was checked using a fluorescence-activated cell sorter (FACS) and was determined to be approximately 60% 24 h after addition of 1.5 µL Lipofectamine RNAiMAX and control stealth siRNA (Supplement Fig. 1). We used a semi-quantitative reverse transcription polymerase chain reaction (sqRT-PCR) to evaluate Cebpb and Runx2 type1 knockdown by stealth siRNA in mHAT9d cells 48 h after transfection (Fig. 5). The specific stealth siRNA effectively abolished Cebpb and Runx2 mRNA and protein levels (Fig. 5A,B). Furthermore, to confirm the specificity for the Runx2 type1 knockdown, we also investigated the expression of Runx1, other Runx2 isoforms, and Runx3 mRNA. We found that expression of other Runx2 isoforms and Runx3 declined, whereas that of Runx1 increased. Additionally, we showed that Cebpb acted upstream of Sox2 to maintain stemness of mHAT9d cells, as indicated by reduced Sox2 mRNA expression following Cebpb knockdown (Fig. 5A). However, expression of two ameloblast differentiation markers, amelogenin (Amelx) and ameloblastin (Ambn), could not be detected (Supplement Fig. 2). These results indicate that Cebpb alone is not sufficient for maintaining stemness and differential inhibition in mHAT9d cells in vitro. Next, we analyzed whether Runx2 type1 knockdown showed EMT in mHAT9d cells. SqRT-PCR analysis of Runx2 type1 knockdown by stealth siRNA demonstrated decreased mRNA expression of adhesion molecules (Fig. 5C). Expression of EMT markers snail family zinc finger 2 (Snai2) increased, as did that of the mesenchymal marker N-cadherin (Cdh2), whereas another mesenchymal marker, vimentin (Vim), and the epithelial marker E-cadherin declined. Similarly, Cebpb knockdown showed increased expression of Snai2, but a reduction of Vim and E-cadherin ( Fig. 5C). Biglycan (Bgn), decorin (Dcn), and bone morphogenetic protein 4/6/7 (Bmp4/6/7) mRNA expression was seen to increase in Runx2 type1 knockdown, and even more in Cebpb and Runx2 type1 knockdown (Fig. 5D). Additionally, bone morphogenetic protein (Bmp2/11) mRNA expression decreased in Runx2 type 1 knockdown and Cebpb and Runx2 type 1 knockdown (Fig. 5D). Because Dcn, encoding a small leucine-rich proteoglycan was particularly up-regulated in double knockdown, we immunostained for decorin in the incisors of 3-month-old F 2 -generation Cebpb −/− Runx2 +/− mice (Fig. 6). Positive staining was observed in the enamel epithelial pearl and extracellular matrix near the supernumerary tooth in the periapical region, as well as the extracellular matrix of the cell component surrounding cementum-like hard tissue in the dental pulp. These in vitro experiments indicated that knockdown of both Cebpb and Runx2 contributed to EMT in a synergistic manner, whereas Cebpb knockdown alone caused down-regulation of Sox2.

Discussion
The Bmp signaling network may provide a niche supporting transient Sox2-positive dental epithelial stem cells in murine incisors 31 . In undifferentiated mesenchymal stem cells (MSCs), CEBPB acts as a repressor, helping maintain stemness. Upon adipogenic, chondrogenic, or osteogenic stimulation, different signaling pathways activate Smad proteins, these bind to CEBPB and remove it from its target regulatory elements, thus converting the repressor into an activator 32,33 . Next, CEBPB binds to and activates its differentiation target genes and, together with other lineage-specific transcription factors, promotes mesenchymal differentiation into adipocytes, osteoblasts or chondrocytes. Here, part of the dental pulp caused ectopic cementogenesis in the maxillary incisor of Cebpb −/− mice (F 2 background) (Fig. 3C,H,J and Table 1, Figure Supplement 3A,B). The phenotype was thought to arise from transdifferentiation of OESCs, as these cells invaded the incisor dental pulp rather than its root analogue. Some cementum-producing cells (cementoblasts) may originate from dental epithelial cells 34 . Cebpb knockout promotes the differentiation of crown epithelia into the root lineage or even cementoblast-like cells rather than ameloblasts 35 (Figure Supplement 3C-J). Taken together, these findings suggest that Cebpb maintains OESCs in the labial cervical loop epithelium and dental pulp stem cells (DPSCs) in murine incisors (Fig. 7).
In Runx2 heterozygous and null mice, budding was observed in maxillary incisors at E15 (Fig. 2C,D). Both Runx2 mutants displayed lingual buds in front of the maxillary molars, which are in line with Runx2 preventing the formation of buds for successional teeth 24,25 . There is a difference between the phenotypes of OESCs in Cebpb −/− Runx2 +/+ and Cebpb −/− Runx2 +/− incisors (F 2 background) between E15 and adult animals. No buddings were observed at E15 in Cebpb −/− mice (F 2 background) (Fig. 2B), however they were seen at E15 in Cebpb +/+ Runx2 +/− and Cebpb −/− Runx2 +/− mice (Fig. 2C,D) and disappeared on postnatal day 7 (Figure supplement 4). Meanwhile, some adult Cebpb −/− mice (129 Sv background) possessed an unusual incisor that presented ectopic hyperplasia of enamel and dentin in the periapical tissue. Moreover, 33% of 3-month-old Cebpb −/− Runx2 +/− mice (F 2 background) had aberrant incisors, characterized by developing or mature ectopic supernumerary teeth in the periapical tissue and dental pulp ( Table 1, Figs 1B-D and 3A-D,H-K). Indeed, in humans, supernumerary teeth are less common in deciduous dentition (first generation of teeth) than in permanent dentition (second generation of teeth) 36 . In mice, the difference may be linked to stem cell aging in the incisor. Common contributing factors of aging in different organisms, but particularly in mammals, are genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, and altered intercellular communication 37 . As another example of epithelial-mesenchymal interactions, hair graying is the most obvious sign of aging in mammals. Irreparable DNA damage, as that caused by ionizing radiation, abolishes renewal of MSCs in mice and results in hair graying inasmuch as it also triggers MSC differentiation into mature melanocytes in the niche 38 . The hallmarks of OESCs can change according to aging.
Differentiated ameloblasts lost their cellular polarity and derangement of ameloblasts was observed in Cebpb +/+ Runx2 +/− , Cebpb −/− Runx2 +/+ , and Cebpb −/− Runx2 +/− mice (F 2 background) ( Fig. 4B-E). Furthermore, we observed the positive staining of N-cadherin during EMT within the dental epithelium in the periapical region (Fig. 4F,G). SqRT-PCR analysis of Runx2 knockdown by siRNA in mHAT9d cells resulted in decreased expression of the epithelial marker E-cadherin (Fig. 5C). The same cells exhibited also increased expression of EMT markers Cdh2 (N-cadherin) and Snai2, but not the other mesenchymal marker Vim. It should be noted that Cebpb knockdown confirmed increased expression of Snai2, as well as a decline in Vim (Fig. 5C). This in vitro knockdown experiment may be unable to reproduce in vivo conditions sufficiently and may only reflect the initial state of EMT, carried out by epithelial cell rests of Malassez (ERM). Nevertheless, these results indicate the additive effect of Cebpb and Runx2 type1 knockdown on EMT. Rodent incisors are a continuation of Hertwig's epithelial root sheath (HERS), which consists of two epithelial layers and plays a role in inducing odontogenesis during root development, becoming fragmented after that. The fragmentation of HERS cells is associated with EMT [39][40][41] . Growth inhibition of the inner enamel epithelium and the active proliferation of the outer enamel epithelium and/or stellate reticulum result in HERS formation 42 . ERM are capable of undergoing EMT under osteogenic conditions, when Vim is down-regulated 43 , as reported also in this study. Strong Bmp signaling in crown/labial  Bilateral 0(0%) 0(0%) 4(80%) 0(0%)

Characteristics of aberrant incisors
Short incisor 0(0%) 0(0%) 3(60%) 0(0%) With supernumerary teeth 0(0%) 0(0%) 2(40%) 4(100%) phenotypic change, as indicated by the onset of EMT in BMPR-1A-depleted tissues during the differentiation stage 35 . Cebpb might also be regulated by Bmp signaling 45 . Knockout of Cebpb mimics a situation characterized by strong Wnt signaling and no Bmp signaling, thereby leading to EMT and the cementum-like structure of ameloblasts. Therefore, we believe that EMT enabled disengaged OESCs bud and develop supernumerary teeth following induction of DPSCs, MSCs originating from OESCs, or those present in connective tissue near the periapical region. These MSCs may have interacted with free ameloblasts, stimulating their differentiation and the formation of amelogenin-positive enamel epithelial pearls and free blobs. SqRT-PCR analysis of Runx2 knockdown by siRNA in mHAT9d cells showed that Sox2 increased during EMT, allowing cells to gain stemness (Fig. 5A). Cebpb knockdown and Runx2 type1 knockdown act on stemness in an opposing manner (Fig. 5A). This behavior has been reported to indicate a direct link between EMT and acquisition of epithelial stem cell properties including self proliferation 46,47 . SqRT-PCR analysis demonstrated increased expression of Bmp4/6 in Runx2 type1 knockdown cells and even more so in Cebpb and Runx2 type1 knockdown cells (Fig. 5D). Bmp signaling plays a role in tumor generation and progression by affecting tumor EMT and stemness, thus increased Bmp4/6 expression could induce EMT and increased Sox2 expression in vitro 48 . Taken together, Runx2 inhibits EMT, differentiation, and branching of undifferentiated enamel epithelial cells, whereas Cebpb and Runx2 knockdown contribute to EMT additively, with EMT promoting Sox2 mRNA expression and acquisition of stemness (Fig. 7). Decorin and biglycan are small leucine-rich proteoglycans distributed intracellularly, on cell surfaces, and within extracellular matrices. Biglycan and decorin regulate early stages of tooth formation and mediate mineralization of dental hard tissues. Both Bgn and Dcn null mice exhibited hypomineralized dentin, with the effect being more conspicuous in the latter. Enamel formation increased strikingly in newborn Bgn null mice and was retarded in the case of Dcn deletion 49 . The expression of Bgn in odontoblasts is highly differentiation  stage-dependent. Dcn mRNA is expressed by odontoblasts abutting the premineralized matrix and predentin, and is synchronously related to expression of type I collagen. Bgn mRNA is expressed by pre-secretory and secretory ameloblasts 50 . Bgn may be associated with odontoblast and ameloblast differentiation, enamel formation, and restraining of collagen fibrillogenesis in predentin. Meanwhile, Dcn may be committed to dentin formation as a regulator of collagen mineralization, possibly through its specific association with type I collagen fibrils [49][50][51] . SqRT-PCR analysis demonstrated increased expression of Bgn, Dcn, and Bmp4/6 in Runx2 type1 knockdown cells and even more so in Cebpb and Runx2 type1 knockdown cells (Fig. 5D). From the above results, it appears that increased Bgn and Dcn expression in OESCs is likely involved in promoting the development of supernumerary teeth in Runx2 +/− Cebpb −/− mice. BGN is essential for Bmp4-induced osteoblast differentiation in neonatal murine calvarial cells 52 . BGN may drive tooth formation by cooperating with Bmp4. DCN inhibits multiple growth factors, such as Tgfb1, and directly antagonizes several members of the receptor tyrosine kinase (RTK) family, including the epidermal growth factor receptor, the hepatocyte growth factor receptor, and the insulin-like growth factor receptor I 53 . Soluble decorin has been characterized as a pan-RTK inhibitor, with a simultaneous powerful downstream signaling inhibitory effect 54 . Higher Dcn expression enabled the collagen fibers to be markedly thinner and caused no mineralized matrices in vitro 55 . These functions of decorin may prevent proliferation of dental epithelial cells and mineralization of mesenchymal cells in the dental pulp, and help the development of a nearly normal supernumerary tooth conformation in the periapical tissue of Cebpb −/− Runx2 +/− mice (F 2 background) (Figs 3C,D and 6, and Table 1).
Molecular targeted therapy uses small molecules, monoclonal antibodies, or other substances to identify and attack specific types of cells by interfering with cellular targets. It could be used for de novo whole tooth regeneration by stimulating arrested tooth germs in animal models with congenital tooth agenesis 16 . Tooth development is under genetic control and is the result of reciprocal and reiterative signaling between oral ectoderm-derived dental epithelium and cranial neural crest cell-derived dental mesenchyme 56 . Two different types of cell components (epithelial and mesenchymal cells) are necessary and essential for whole tooth regeneration. Our results demonstrate that OESCs had the potential to differentiate odontoblasts by EMT and regenerate de novo whole teeth (supernumerary teeth) by synergistic abrogation of Cebpb and repression of Runx2. In recent years, there have been many studies on stem cells in relation to the development of regenerative therapies for various organs. This has led to the discovery that adult zebrafish regenerate nephrons (the functional units of the kidney) throughout their lives. The zebrafish kidney contains renal stem cells, which differentiate into epithelial or mesenchymal cells of the nephrons 57 . The EMT in Sox2-expressing stem/progenitor cells contributes to the regeneration of the pituitary gland 58 . It has been confirmed that Sox2, which is one of the molecular marker of OESCs, localizes to the dental lamina of developing human primary molars 59 . OESCs are suitable for molecular targeted therapy using Cebpb and/or Runx2 (Fig. 7). However, from our analysis, it remains unclear whether OESCs differentiate into odontogenic epithelia and mesenchyme directly or through other cells. Future studies tracking individual cells are needed to demonstrate cell dynamics during de novo whole tooth regeneration. Cebpb +/− and Runx2 +/− mice were gifts from Shizuo Akira (Osaka University) and Toshihisa Komori (Nagasaki University), respectively 22,60 . Cebpb +/− mice were maintained in a 129 Sv background. Runx2 +/− mice were maintained in a C57BL/6 background. We interbred heterozygous Cebpb and Runx2 mice and analyzed the F 2 generation. Runx2 null mice were embryonic lethal. Day E0 was taken to be midnight prior to finding a vaginal plug. Wild-type mice were used as control.
Examination of transfection efficiency using fluorescent oligos. For each transfection reaction on a 24-well plate, fluorescent oligo-Lipofectamine ® RNAiMAX (Thermo Fisher Scientific) complexes were prepared as follows.
BLOCK-iT TM Alexa Fluor Red Fluorescent Control (Thermo Fisher Scientific) (6 pmol) was diluted in 100 μL of Opti-MEM ® I Reduced Serum Medium (Thermo Fisher Scientific). Lipofectamine ® RNAiMAX (0.5, 1.0, and 1.5 μL) was added to the mixture and incubated at room temperature for 15 min. The cell suspension (500 μL) was diluted with the above-mentioned medium and was added to a well containing the fluorescent oligo and Lipofectamine ® RNAiMAX mixture. This gave a final fluorescent oligo concentration of 10 nM, a final volume of 600 μL, and 2 × 10 5 cells. The cells were incubated at 37 °C in a CO 2 incubator until they were ready for harvesting and assayed for the gene of interest. Transfection efficiency and thus target gene knockdown was measured using a FACS calibur (BD Biosciences) 24, 48, and 72 h after transfection. SqRT-PCR. RT-PCR was performed as follows. Total RNA from 70% confluent mHAT9d cells was extracted using TRIzol reagent (Thermo Fisher Scientific), 0.2 mL chloroform, and 0.2 mL phenol chloroform per 1 mL of TRIZOL. The RNA was precipitated from the aqueous phase by mixing with isopropyl alcohol and 16.7 mg/mL glycogen or with NucleoSpin ® RNA Clean-up XS columns (U0903, TaKaRa Bio). Total RNA (3 μg) was reverse transcribed using the SuperScript ® IV First-Strand Synthesis System (Thermo Fisher Scientific).
Each siRNA was the same in quantity (10 nM each) when mixed. 1/10, 1/30, and 1/90 indicates dilution ratio. 925 and 1623 indicate Cebpb and Runx2 type1 siRNA, respectively. RT-PCR analysis of Cebpb and Runx2 type1 expression after knockdown with negative control stealth siRNA or the corresponding siRNA (final concentration 20 nM). All experiments were performed at least thrice, and representive data are shown. GAPDH was used as an internal control for RT-PCR. The bar graphs indicate relative mRNA level of genes, which were calculated by image j software. Values of them were normalized to GAPDH and mRNA expression data on negative control stealth siRNA were used as denominator of ratios.
Statistical analysis. Data are presented as mean ± standard deviation. Statistical significance was assessed by analysis of variance in R. Gaussian distribution was determined with an Anderson-Darling normality test.
To analyze the labial cervical loop epithelium area, the number of Sox2-positive cells and Sox2-positive area/Labial cervical loop epithelium area, which was measured using the imageJ software, in the maximal area of all sections in Cebpb genetically modified mice (129 Sv), statistical significance was determined using a Mann-Whitney U test. Sox2-positive area (diaminobenzidine coloring) was extracted with Photoshop elements 12 (Adobe Systems, San Jose, CA, USA). To analyze the area and the cell number of lingual buds in the maximal area of all sections in Cebpb +/+ Runx2 +/− and Cebpb −/− Runx2 +/− F 2 (129 Sv/C57BL/6) mice, statistical significance was determined using a Mann-Whitney U test.