Roles of Bmp4 during tooth morphogenesis and sequential tooth formation

Mammalian tooth development involves highly regulated reciprocal signaling interactions between the epithelium and mesenchyme. Classic tissue recombination experiments demonstrated that, although tooth inductive signals arise initially in the embryonic oral ectoderm, tooth inductive potential shifts from ectoderm to mesenchyme at the early tooth bud stage, with the developing tooth mesenchyme acquiring capability of inducing tooth organogenesis even when combined with non-dental epithelium (Kollar and Baird, 1970; Kollar and Fisher, 1980; Lumsden, 1988; Mina and Kollar, 1987; Ruch et al., 1973). As development proceeds, mesenchymal signals induce formation of an epithelial signaling center, termed the primary enamel knot (PEK), in the distal region of the tooth bud, which in turn drives tooth morphogenesis through the ‘cap’ and ‘bell’ stages by signaling into both the dental epithelium and mesenchyme. Many signaling molecules, including members of the Bmp, Fgf and Wnt families and Shh, are expressed during early tooth development (reviewed by Jernvall and Thesleff, 2000; Tucker and Sharpe, 2004; Zhang et al., 2005). Among these, Bmp4 exhibits an expression pattern coinciding with the shift of odontogenic potential from epithelium to mesenchyme during tooth bud formation (Vainio et al., 1993). Bmp signaling is required for activation of expression of the Msx1 transcription factor in the developing tooth mesenchyme (Tucker et al., 1998). In mice lacking Msx1 gene function, Bmp4 mRNA expression was downregulated in the tooth mesenchyme and tooth development arrested at the bud stage (Chen et al., 1996; Satokata and Maas, 1994). Addition of recombinant Bmp4 protein rescued Msx1^–/– mutant mandibular first molar tooth germs to late bell stage in explant cultures (Bei et al., 2000; Chen et al., 1996). Bmp4-releasing beads placed in contact with isolated dental epithelium induced localized expression of a PEK marker p21 (Cdkn1a) (Jernvall et al., 1998). Transgenic Bmp4 expression driven by an Msx1 gene promoter also partially rescued Msx1^–/– mutant first molar tooth germs to the cap stage with formation of a PEK (Zhao et al., 2000). In addition, mice homozygous for Pax9 null mutations exhibit tooth developmental arrest at the early bud stage accompanied by loss of Bmp4 and Msx1 expression in developing tooth mesenchyme (Peters et al., 1998; Zhou et al., 2011). In vitro biochemical assays showed that Msx1 and Pax9 act synergistically to activate the Bmp4 gene promoter (Ogawa et al., 2006). Moreover, tissue-specific inactivation of Bmpr1a in epithelial tissues resulted in tooth developmental arrest at the bud stage (Andl et al., 2004; Liu et al., 2005). These data led to the conclusion that Bmp4 is a key Msx1-dependent signal for induction of PEK formation to drive tooth morphogenesis beyond the bud stage (Bei et al., 2000; Miletich et al., 2011; O’Connell et al., 2012; Zhao et al., 2000). However, direct genetic analysis of the requirement for Bmp4 in early tooth morphogenesis has not been documented.

Teeth are iterative structures that form sequentially in an anterior-to-posterior direction but little is known about the molecular mechanisms regulating sequential tooth formation. In humans, heterozygous loss-of-function mutations in either MSX1 or PAX9 cause selective tooth agenesis, which often preferentially affects the posterior molars and second premolars (Stockton et al., 2000; Vastardis et al., 1996; Mostowska et al., 2012). Although mice heterozygous for either Msx1 or Pax9 do not exhibit tooth defects, many Msx1^+/–Pax9^+/– compound heterozygous mice lack third molars (Nakatomi et al., 2010). Interestingly, transgenic Bmp4 expression driven by the mouse Msx1 gene promoter rescued the tooth defects in Msx1^+/–Pax9^+/– compound heterozygous mice (Nakatomi et al., 2010). Moreover, whereas mice lacking the Osr2 transcription factor develop uniquely supernumerary teeth lingual to their molar teeth, supernumerary tooth formation in the Osr2^–/– mice is accompanied by lingual expansion of the domain of Bmp4 mRNA expression in the tooth mesenchyme (Zhang et al., 2009). Remarkably, in contrast to early tooth developmental arrest in Msx1^–/– mutant mice, mice lacking both Msx1 and Osr2 showed continued first molar morphogenesis to the late bell stage, accompanied by partially restored mesenchymal Bmp4 expression. However, the Msx1^–/–Osr2^–/– double mutant mice did not develop supernumerary or mandibular second molar teeth (Zhang et al., 2009). To directly investigate the roles of Bmp4 in tooth development and sequential tooth formation, we generated and analyzed tooth development in Bmp4^f/f;Wnt1Cre mice in which the Bmp4 gene is inactivated in neural crest-derived craniofacial mesenchyme, including the tooth mesenchyme. Although mandibular first molar development was arrested at the bud stage, the maxillary first and second molars as well as both upper and lower incisors developed to mineralized teeth in Bmp4^f/f;Wnt1Cre mutant mice. Further generation and analyses of compound mutant mice deficient in either Msx1 or Osr2 in the Bmp4^f/f;Wnt1Cre mutant background indicate that Bmp4 signaling suppresses Osr2 expression and synergizes with Msx1 to drive propagation of mesenchymal odontogenic activity during tooth morphogenesis and sequential tooth formation.

The Bmp4^f/f, Bmp4^+/–, Msx1^+/–, Osr2^+/– and Wnt1Cre transgenic mice have been described previously (Danielian et al., 1998; Lan et al., 2004; Liu et al., 2005; Winnier et al., 1995). The Bmp4^+/–, Msx1^+/– and Osr2^+/– mice were maintained in the C57BL/6 background, Wnt1Cre mice were maintained in the CD1 background, and the Bmp4^f/f mice maintained by sibling intercrosses. The tooth phenotypes were indistinguishable between Bmp4^f/f;Wnt1Cre and Bmp4^f/–;Wnt1Cre mutant mice at various prenatal and postnatal stages examined, hence data were combined.

For histological analysis, embryos were collected from timed pregnant females, fixed in Bouin’s fixative or 4% paraformaldehyde (PFA), dehydrated through graded ethanol series, embedded in paraffin, and sectioned at 7 μm thickness. Sections were stained with Hematoxylin and Eosin, or using trichrome staining as described previously (Chen et al., 2009). Skeletons were prepared by using Dermestid beetles as described previously (Hefti et al., 1980).

For immunofluorescence staining and in situ hybridization, embryos were fixed in 4% PFA overnight at 4°C and processed for paraffin sections as described above. Indirect immunofluorescence staining was performed as described (Zhou et al., 2011). Anti-phospho-Smad1/5 antibody was purchased from Cell Signaling Technology. In situ hybridization of sections was performed as described previously (Zhang et al., 1999). The cRNA probe for Bmp4 was specific for the exon 4, which was flanked by loxP sites in the Bmp4^f allele.

Embryos were dissected on ice, and the heads were frozen immediately in Tissue-Tek OCT compound by immersion in liquid nitrogen. Fresh-frozen tissue was cryosectioned and collected on Arcturus PEN membrane glass slides (Applied Biosystems). The slides were immediately refrozen and stored at –80°C. The tooth mesenchyme tissues were isolated by laser capture microdissection (LCM) using Veritas Laser Microdissection instrument model 704. Maxillary and mandibular molar mesenchyme tissues were each pooled from three embryos of the same genotype and total RNAs extracted by using the RNeasy Micro Kit (Qiagen). cDNA templates were generated by using the NuGEN Ovation RNA-Seq v2 system. Sequencing libraries were generated by using Illumina Nextera DNA Sample Prep Kit and sequenced using Illumina Hiseq 2000. Sequenced reads were mapped to the reference mouse genome (mm9) using Bowtie version 0.12.7. Single-end reads were aligned using Tophat version 1.4.1 for Illumina (Brunskill and Potter, 2012).

The RNAseq raw data have been deposited into NCBI Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo) (accession number GSE39918). RNAseq data were analyzed using AvadisNGS software, with the reads per kilobase exon per million mapped sequences (RPKM) value calculated for each RefSeq gene for relative gene expression. For analyses of differential expression, the fold change cut-off was set at 1.5-fold or higher and P-values <0.01 from the Audic Claverie test were considered to be statistically significant, with Benjamini Hochberg FDR multiple testing correction (Brunskill and Potter, 2012).

Quantitative PCR amplifications were performed using the C1000 Touch Thermal Cycler (Bio-Rad) and the SsoAdvanced SYBR Green Supermix (Bio-Rad). PCR reaction was run in a program of 95°C for 30 seconds, followed by 40 cycles of 95°C for 5 seconds and 60°C for 30 seconds, with a melt curve generation cycle at the end. PCR reactions were carried out in duplicate and relative levels of mRNAs were normalized to Hprt using the standard curve method. Student’s t-test was used to analyze pair-wise differential expression and a P-value <0.05 was considered to be statistically significant.

Because Bmp4-null mouse embryos die prior to embryonic day (E) 9.5 (Winnier et al., 1995), before tooth morphogenesis occurs, we investigated the roles of Bmp4 in tooth development using Cre/loxP-mediated conditional gene inactivation of the Bmp4^f allele containing loxP sites flanking exon-4 (Liu et al., 2005). The Wnt1Cre transgenic mice express Cre recombinase transiently in premigratory neural crest cells and have been shown to efficiently delete loxP-flanked DNA sequences from neural crest-derived craniofacial mesenchyme, including the developing tooth mesenchyme (Chai et al., 2000; Danielian et al., 1998; Han et al., 2003). Indeed, in contrast to strong Bmp4 mRNA expression in the developing tooth mesenchyme at E13.5 in control embryos (Fig. 1A), Bmp4 mRNA was undetectable in the Bmp4^f/f;Wnt1Cre mutant tooth mesenchyme by in situ hybridization analyses using cRNA probes specific for the exon-4 sequences (Fig. 1B). To quantify the efficiency of inactivation of the Bmp4 gene in the tooth mesenchyme, we isolated maxillary and mandibular molar tooth mesenchyme from E13.5 wild-type and Bmp4^f/f;Wnt1Cre mutant littermates using LCM (supplementary material Fig. S1) and compared the levels of exon-4-containing Bmp4 mRNAs by real-time RT-PCR. The Bmp4^f/f;Wnt1Cre mutant maxillary and mandibular molar tooth mesenchyme each contained <1% of the amount of exon-4-containing Bmp4 mRNAs present in the wild-type counterparts (Fig. 1C). Consistent with loss of Bmp4 function in the tooth mesenchyme, little phosphorylated Smad1/5 (pSmad1/5) was detected in the developing tooth mesenchyme in the mutant embryos whereas high levels of pSmad1/5 were present in both the dental epithelium and mesenchyme in the control littermates at E13.5 (Fig. 1D,E). Furthermore, in contrast to strong expression of Dlx2 mRNA in the buccal side of the developing tooth epithelium in both the maxillary and mandibular molar tooth germs in E13.5 control mouse embryos (Fig. 1F), Dlx2 mRNA expression was undetectable in the developing upper and lower molar tooth epithelium in the Bmp4^f/f;Wnt1Cre mutant littermates (Fig. 1G). These data indicate that Bmp4 function was effectively inactivated in the neural crest-derived tooth mesenchyme in the Bmp4^f/f;Wnt1Cre mutant mice.

We examined newborn Bmp4^f/f;Wnt1Cre mutant mice for tooth developmental defects by histological analysis. Whereas both maxillary and mandibular first molar tooth germs progressed to the cytodifferentiation stage in the Bmp4^f/+;Wnt1Cre control littermates (supplementary material Fig. S2A,C), Bmp4^f/f;Wnt1Cre mutants showed retarded maxillary first molar tooth germs at the late bell stage and absence of the mandibular first molar tooth germ (supplementary material Fig. S2B,D). We also detected substantially smaller maxillary second molar tooth germs and absence of mandibular second molar tooth germs in the Bmp4^f/f;Wnt1Cre mutants in comparison with the well-developed late-bell stage second molar tooth germs in the control littermates (supplementary material Fig. S2E-H). Although previous studies suggested that Bmp4 is an essential signal downstream of Msx1 in the developing secondary palate (Zhang et al., 2002), we found that only ∼16% of the Bmp4^f/f;Wnt1Cre mutant mice examined at late gestation or postnatal day (P) 0 exhibited cleft palate. Indeed, most Bmp4^f/f;Wnt1Cre mutant mice survived postnatally but exhibited postnatal growth retardation by three weeks of age. Skeletal preparations showed that, whereas the control littermates had three well-mineralized molars and one incisor in each quadrant (Fig. 2A,C), the Bmp4^f/f;Wnt1Cre mice had mineralized upper and lower incisors as well as maxillary first and second molars, but completely lacked mandibular molars (Fig. 2B,D). The mutant maxillary molars were smaller with shallower cusps than their counterparts in the control mice. In addition, the mutant mandibular incisors exhibited a round shape, in contrast to the sharp mandibular incisors of the control littermates (Fig. 2C,D). Histological analyses of P5 mouse incisor tooth germs indicate that both the odontoblast and ameloblast layers of the lower incisor tooth germs were malformed, resulting in significantly reduced enamel deposition on the labile surface of the lower incisors in the mutant mice (supplementary material Fig. S3).

We analyzed embryos from E12.5 to E16.5 and found that mandibular first molar development was arrested at the bud stage in the Bmp4^f/f;Wnt1Cre mutant embryos (Fig. 3A-F). By contrast, the maxillary first molar tooth germs in Bmp4^f/f;Wnt1Cre mutant embryos progressed to the cap stage by E14.5, similar to the control littermates (Fig. 3G,H). By E15.5, the mutant maxillary first molar tooth germs were obviously smaller than those in the control littermates (Fig. 3I,J). At E16.5, the maxillary first molar tooth germs continued to expand in size in both control and mutant embryos, but the size and morphology of the mutant maxillary first molar tooth germs appeared similar to the E15.5 control maxillary first molar tooth germs (Fig. 3I-L).

Although previous reports have shown significant loss of Bmp4 mRNA expression in the molar tooth mesenchyme by E13.5 in Msx1^–/– mutant embryos as detected by in situ hybridization (Chen et al., 1996; Zhang et al., 2009), a recent microarray-based gene expression profiling study of LCM-isolated tooth tissues from wild-type and Msx1^–/– mutant embryos revealed that the levels of Bmp4 mRNA in the E13.5 Msx1^–/– mutant mandibular molar tooth mesenchyme was reduced by only ∼26% in comparison with the control littermates (O’Connell et al., 2012) (see also ToothCODE database http://compbio.med.harvard.edu/ToothCODE). To directly compare the amount of loss of Bmp4 mRNA in the Msx1^–/– versus Bmp4^f/f;Wnt1Cre mutant tooth mesenchyme, we isolated the maxillary and mandibular molar tooth mesenchyme from E13.5 and E14.0 Msx1^–/– mutant and Msx1^+/– control littermates using LCM and carried out quantitative real-time RT-PCR analyses of exon-4-containing Bmp4 mRNAs. We found that the E13.5 Msx1^–/– mutant maxillary and mandibular molar mesenchyme contained 87% and 65%, respectively, and the E14.0 Msx1^–/– mutant maxillary and mandibular molar mesenchyme contained 88% and 64%, respectively, of the amount of Bmp4 mRNAs present in the molar mesenchyme of littermate controls (Fig. 4). Thus, although the Bmp4^f/f;Wnt1Cre mutant embryos showed much more dramatic loss of Bmp4 mRNAs in both maxillary and mandibular molar mesenchyme than did the Msx1^–/– mutant embryos, the maxillary first molars in the Bmp4^f/f;Wnt1Cre mutant continued to develop to mineralized teeth but the corresponding maxillary first molars in the Msx1^–/– mutant embryos were developmentally arrested at the bud stage. We verified further that the Msx1^–/– mutant mouse tooth phenotype has not changed from previous studies: all nine late term Msx1^–/– mutant mouse embryos examined showed bud stage arrest of both maxillary and mandibular molar tooth germs. These data indicate that the developmental arrest of the maxillary first molar tooth germs in the Msx1^–/– mutant mice could not be due only to the decrease in mesenchymal Bmp4 expression and suggest that normal bud-to-cap transition of molar tooth development requires at least one other Msx1-dependent odontogenic factor.

Previous studies suggested that Bmp4 and Msx1 function in a positive-feedback loop to regulate the expression of each other in the developing tooth mesenchyme (Chen et al., 1996; Tucker et al., 1998; Vainio et al., 1993). We found that the levels of Msx1 mRNA expression were consistently lower from E13.5 to E14.5 in both the maxillary and mandibular molar tooth mesenchyme in the Bmp4^f/f;Wnt1Cre mutant embryos in comparison with control littermates by using in situ hybridization analyses (Fig. 5A,B,D,E). Quantitative real-time RT-PCR analyses of LCM-isolated tooth mesenchyme samples showed that the levels of Msx1 mRNAs were decreased by >50% in the mandibular molar mesenchyme in the Bmp4^f/f;Wnt1Cre mutant embryos by E14.5 in comparison with control littermates (Fig. 5C,F). By comparison, the reduction in Msx1 mRNA expression in the maxillary molar tooth mesenchyme was not as dramatic as in the mandibular molar mesenchyme in the Bmp4^f/f;Wnt1Cre mutant embryos (Fig. 5C,F). To test whether the differences in Msx1 expression levels could account for the distinct outcome of the maxillary and mandibular molar teeth, we examined tooth development in Msx1^+/–Bmp4^f/–;Wnt1Cre mutant mice. Whereas the Bmp4^f/–;Wnt1Cre mutant mice formed maxillary first and second molars that are reduced in size in comparison with their counterparts in the control littermates (Fig. 6A,B), the Msx1^+/–Bmp4^f/–;Wnt1Cre mutant mice formed only a much smaller maxillary first molar (Fig. 6C). Histological analyses of E18.5 embryos showed that the maxillary first molar tooth germs in the Msx1^+/–Bmp4^f/–;Wnt1Cre mutant embryos were dramatically smaller than the counterparts in the control and Bmp4^f/–;Wnt1Cre mutant embryos but still developed to the bell stage (Fig. 6D-F). By contrast, whereas the maxillary second molar tooth germs in the Bmp4^f/–;Wnt1Cre mutants were smaller than those in the control embryos (Fig. 6G,H), the second molar tooth germs failed to develop in the Msx1^+/–Bmp4^f/–;Wnt1Cre mutant embryos (Fig. 6I). These data indicate that the distinct maxillary and mandibular molar tooth phenotypes of the Bmp4^f/f;Wnt1Cre mutant mice could not be explained simply by the differences in the levels of Msx1 expression in the tooth mesenchyme but that the level of Msx1 expression is important for sequential molar tooth development.

To investigate the molecular basis of the dramatic difference in maxillary and mandibular molar tooth phenotypes in the Bmp4^f/f;Wnt1Cre mutant mice, we isolated maxillary and mandibular molar tooth mesenchyme from E13.5 Bmp4^f/f;Wnt1Cre mutant and wild-type littermates using LCM and carried out RNAseq analyses. Consistent with the real-time RT-PCR results (Fig. 1C), we found that the Bmp4^f/f;Wnt1Cre mutant maxillary and mandibular molar tooth mesenchyme contained <1% of wild-type levels of exon-4-containing Bmp4 mRNAs and that the Bmp4 mRNA levels were not significantly different between the maxillary and mandibular molar tooth mesenchyme in either the wild-type or the Bmp4^f/f;Wnt1Cre embryos (supplementary material Table S1).

We investigated whether other members of the Bmp family of ligands were differentially expressed in the developing maxillary and mandibular tooth mesenchyme. Of all Bmp family members other than Bmp4, only Bmp3 and Bmp6 mRNAs were detected at >5 reads per kilobase of transcript per million of mapped reads (RPKM) in maxillary and/or mandibular molar mesenchyme in the E13.5 wild-type embryos (supplementary material Table S2). These data are in agreement with previous in situ hybridization results showing expression of Bmp3, Bmp4 and Bmp6 mRNAs in the developing tooth mesenchyme whereas expression of Bmp2 and Bmp7 mRNAs was restricted to the tooth bud epithelium at E13.5 (Aberg et al., 1997). The RNAseq data show no significant differences in expression of other Bmp ligands that could account for the differences in maxillary and mandibular molar tooth defects in the Bmp4^f/f;Wnt1Cre mutant mice.

Analyses of the RNAseq data revealed that >170 genes are differentially expressed by more than twofold (P<0.01) in the wild-type maxillary versus mandibular molar tooth mesenchyme at E13.5. Among these, Dkk2 transcripts were 3.5 times as abundant in the mandibular as in the maxillary molar mesenchyme (supplementary material Table S3). Transcripts for another Wnt antagonist, Wif1, were 2.4-fold as abundantly expressed in the mandibular as in the maxillary molar mesenchyme. Whereas Wif1 expression levels were similar in the Bmp4^f/f;Wnt1Cre mutant and control molar tooth mesenchyme, Dkk2 mRNA expression was significantly upregulated in both the maxillary and mandibular molar mesenchyme in the Bmp4^f/f;Wnt1Cre mutant embryos in comparison with the wild-type embryos (supplementary material Table S3). These differences in Dkk2 and Wif1 mRNA levels between the maxillary and mandibular molar mesenchyme in the control and mutant embryos were verified further by using real-time RT-PCR analyses (Fig. 7). As Wnt signaling is crucial for the bud-to-cap transition during tooth development (Liu et al., 2008; Chen et al., 2009), the substantially higher levels of Dkk2 and Wif1 in the mandibular than maxillary molar mesenchyme could contribute to the more severe mandibular tooth defects in the Bmp4^f/f;Wnt1Cre mutant mice.

We recently reported that the Osr2 transcription factor antagonizes Msx1-mediated propagation of mesenchymal odontogenic activity along the buccolingual axis to restrict molar tooth development in a single row (Zhang et al., 2009). Comparison of RNAseq data from the control and the Bmp4^f/f;Wnt1Cre mutant molar mesenchyme revealed that expression of Osr2 mRNAs was increased in both the maxillary and mandibular molar tooth mesenchyme in the Bmp4^f/f;Wnt1Cre mutant embryos in comparison with the control samples (supplementary material Table S3). Both RNAseq and quantitative real-time RT-PCR analyses showed that Osr2 mRNA expression was more significantly upregulated in the mandibular than in the maxillary molar mesenchyme such that Osr2 mRNA levels were significantly more abundant in the mandibular than the maxillary molar mesenchyme in Bmp4^f/f;Wnt1Cre mutant embryos (Fig. 7; supplementary material Table S3). To investigate whether the increased Osr2 expression contributed to the mandibular molar tooth developmental arrest in the Bmp4^f/f;Wnt1Cre mutant mice, we introduced the Osr2 null mutation into the Bmp4^f/f;Wnt1Cre mice. We found that deleting one allele of Osr2 partially rescued each mandibular first molar to a small mineralized tooth in the Bmp4^f/f;Wnt1Cre mutant mice (Fig. 8A-C). Examination of Bmp4 mRNA expression in the developing molar tooth germs at E14.5 by in situ hybridization analyses showed that both Bmp4^f/f;Wnt1Cre and Osr2^+/–Bmp4^f/f;Wnt1Cre mutant embryos lack Bmp4 expression in the tooth mesenchyme, but both maxillary and mandibular first molar tooth germs progressed to the cap stage and strongly expressed Bmp4 mRNA in the PEK in the Osr2^+/–Bmp4^f/f;Wnt1Cre mutant embryos, in contrast to the bud stage arrest of the mandibular first molars in the Bmp4^f/f;Wnt1Cre mutant embryos (Fig. 8D-F). Real-time RT-PCR analyses of LCM-isolated maxillary and mandibular molar tooth mesenchyme confirmed that there was no restoration of Bmp4 expression in the tooth mesenchyme in Osr2^+/–Bmp4^f/f;Wnt1Cre mutant embryos (supplementary material Fig. S4). Moreover, pSmad1/5 activation was similarly deficient in the developing molar tooth germs in Bmp4^f/f;Wnt1Cre and Osr2^+/–Bmp4^f/f;Wnt1Cre mutant embryos (Fig. 8G-I). However, both maxillary and mandibular molar tooth germs in the Osr2^+/–Bmp4^f/-;Wnt1Cre mutant embryos exhibited strong expression of p21, Lef1 and Fgf3 in the PEK at E14.5, similar to the Bmp4^f/+;Wnt1Cre control embryos and in contrast to the lack of PEK formation in the mandibular molar tooth germs in the Bmp4^f/-;Wnt1Cre mutant littermates (Fig. 9A-I). We further analyzed Dkk2 expression in Osr2^+/–Bmp4^f/f;Wnt1Cre mutant embryos and found that Dkk2 mRNA expression remained differentially expressed in the maxillary and mandibular mesenchyme and still significantly upregulated in both the maxillary and mandibular mesenchyme in comparison with control embryos (supplementary material Fig. S4). These data suggest that Bmp4 signaling plays an important role in the activation of mesenchymal odontogenic activity by suppressing the expression of tooth developmental antagonists, including Dkk2 and Osr2, in the developing tooth mesenchyme.

We previously showed that Osr2^–/– mutant mice develop supernumerary teeth from oral epithelium lingual to the molar teeth owing to expansion of the mesenchymal odontogenic field and that initiation of supernumerary tooth formation in the Osr2^–/– mice requires Msx1 function (Zhang et al., 2009). To investigate the role of mesenchymal Bmp4 in sequential tooth formation, we generated and analyzed molar tooth development in Osr2^–/–Bmp4^f/f;Wnt1Cre mutant mice. Histological analyses of newborn Osr2^–/–Bmp4^f/f;Wnt1Cre mutant mice did not detect supernumerary tooth germs lingual to their molars, although all Osr2^–/– mutant littermates showed supernumerary teeth (Fig. 10A,B). We then analyzed the mutant embryos at E15.5 to investigate whether the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant embryos failed to initiate supernumerary tooth formation. An area of aberrantly thickened oral epithelium expressing the tooth epithelial marker Pitx2 was clearly observed lingual to each of the mandibular first molar tooth germs in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant embryos, similar to observations in the Osr2^–/– mutant littermates (Fig. 10C-F), indicating that supernumerary tooth development was initiated in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant embryos. However, in contrast to the expansion of Msx1 mRNA expression to the mesenchyme underlying the supernumerary tooth placode in the Osr2^–/– mutant embryos, Msx1 mRNA expression was significantly reduced in the mandibular fist molar tooth mesenchyme and did not expand to the mesenchyme underlying the supernumerary tooth placode in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant embryos (Fig. 10G,H). The lack of expansion of Msx1 expression to the mesenchyme underlying the supernumerary tooth placode accounts for the subsequent failure of supernumerary tooth morphogenesis in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant mice and indicates that mesenchymal Bmp4 signaling is required to propagate Msx1 expression during sequential tooth formation.

In addition to lack of supernumerary tooth initiation, the Msx1^–/–Osr2^–/– double mutant mice did not develop second molars although their first molars developed beyond the bell stage and there was clearly detected Bmp4 expression in the developing molar tooth mesenchyme (Zhang et al., 2009). By contrast, we consistently detected a small second molar tooth germ that had progressed to the cap stage in addition to a late bell stage first molar tooth germ in each half of the mandible in P0 Osr2^–/–Bmp4^f/f;Wnt1Cre mutant mice (Fig. 11; supplementary material Fig. S5). As the Msx1^–/–Osr2^–/– mutant embryos showed clearly detectable Bmp4 mRNA expression in the molar tooth mesenchyme (Zhang et al., 2009), the formation of supernumerary and second molar tooth germs in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant but not in Msx1^–/–Osr2^–/– mutant embryos indicate that other Msx1-dependent odontogenic factors in addition to Bmp4 are required for the induction of sequential molar tooth formation.

Because the bud stage tooth developmental arrest in Msx1^–/– mutant mice was associated with a reduction in Bmp4 mRNA expression in the tooth mesenchyme and addition of exogenous Bmp4 protein or transgenic overexpression of Bmp4 was able to partially rescue Msx1^–/– mutant mandibular molar tooth germs through the bud-to-cap transition, it has been concluded that mesenchymal Bmp4 is responsible for and required for driving molar tooth morphogenesis through the bud-to-cap transition (Chen et al., 1996; Bei et al., 2000; Zhao et al., 2000). This conclusion has been widely accepted and applied as an established principle for interpreting tooth developmental phenotypes, although the role of endogenous mesenchymal Bmp4 in tooth development has not been previously directly investigated. For example, a 24-hour stalling of the molar tooth buds from E13.5 to E14.5 in Barx1^–/– mutant embryos was suggested to be caused by a reduction of mesenchymal Bmp4 mRNA expression in the tooth mesenchyme at E13.5 as detected by in situ hybridization (Miletich et al., 2011). Miletich et al. further hypothesized that the subsequent resumption of molar tooth morphogenesis in the Barx1^–/– mice, following the bud stage stalling, was due to mesenchymal Bmp accumulating and reaching an optimal level by E14.5 (Miletich et al., 2011). In this study, we have directly investigated the role of endogenous mesenchymal Bmp4 in tooth development. We found that maxillary first molar tooth germs in the Bmp4^f/f;Wnt1Cre mutant mice developed through the bud-to-cap transition in pace with those in the control littermates although the Bmp4 gene was irreversibly inactivated in the cranial neural crest cells and the residual level of Bmp4 mRNA in Bmp4^f/f;Wnt1Cre maxillary molar tooth mesenchyme was much lower than that in Msx1^–/– maxillary molar tooth mesenchyme. These data indicate that at least the maxillary molar tooth bud arrest in Msx1^–/– mutant mice was not solely due to the reduction in mesenchymal Bmp4 expression. Moreover, as there was clearly detected Bmp4 mRNA expression in the molar tooth bud mesenchyme in Barx1^–/– mutant embryos at E13.5 (Miletich et al., 2011), in contrast to undetectable Bmp4 mRNA expression in the Bmp4^f/f;Wnt1Cre maxillary molar mesenchyme by in situ hybridization, the molecular mechanism underlying bud stage stalling of molar tooth germs in Barx1^–/– mutant mice is likely to involve other Barx1-regulated factors in addition to Bmp4, which warrants further investigation.

If the molar tooth developmental arrest in Msx1^–/– mutant mice was not solely due to decreased Bmp4 expression, why was exogenous Bmp4 protein or transgenic Bmp4 overexpression able to rescue morphogenesis of the Msx1^–/– mutant molar tooth germs (Bei et al., 2000; Zhao et al., 2000)? We previously demonstrated that Msx1^–/–Osr2^–/– double homozygous mutant mice exhibited first molar morphogenesis to the bell stage (Zhang et al., 2009), indicating that Msx1 and Osr2 act antagonistically to regulate mesenchymal odontogenic activity but at least one other endogenous factor could activate mesenchymal odontogenic activity in the absence of both Msx1 and Osr2. In this study, we found that expression of Osr2 was significantly upregulated in the Bmp4^f/f;Wnt1Cre mutant molar mesenchyme and that reduction of Osr2 gene dosage partially rescued Bmp4^f/f;Wnt1Cre mutant mandibular molar morphogenesis without restoring mesenchymal Bmp4 expression. Thus, it is possible that the exogenous or overexpressed Bmp4 was able to restore odontogenic activity of the Msx1^–/– mutant molar mesenchyme through downregulation of Osr2 and/or activation of other odontogenic activators. In addition, other factors in the fetal calf serum or chick embryo extract used in the explant culture medium might have acted synergistically with exogenous Bmp4 protein as a high percentage of the Msx1^–/– tooth germ explants supplemented with exogenous Bmp4 progressed to the stage of dentin formation whereas the Bmp4 transgene driven by the Msx1 promoter only rescued a low percentage of Msx1^–/– mutant molar tooth germs to early cap stage without further morphogenesis (Bei et al., 2000; Zhao et al., 2000). Thus, although our findings indicate that endogenous mesenchymal Bmp4 plays crucial roles in tooth development, this study reveals an important gap in the understanding of Msx1-mediated regulation of mesenchymal odontogenic activity. Moreover, our data indicate an essential role for Bmp4 signaling in the regulation of mesenchymal odontogenic activity through suppression of Osr2 expression and in modulation of Wnt signaling during tooth development through suppression of Dkk2 expression.

Previous studies of Inhba^–/– and Dlx1/2^–/– mutant mice have suggested that development of the maxillary and mandibular molar teeth might involve distinct molecular pathways (Ferguson et al., 1998; Thomas et al., 1997). In particular, maxillary molar tooth development, uniquely, does not require activin signaling although expression of genes downstream of activin signaling, including follistatin and Irx1, were downregulated in both the maxillary and mandibular molar tooth germs in the Inhba^–/– mutant embryos (Ferguson et al., 1998; Ferguson et al., 2000). Similarly, we found that the Bmp4^f/f;Wnt1Cre mutant embryos exhibited loss of pSmad1/5 as well as dental epithelial expression of Dlx2 in both maxillary and mandibular molar tooth germs, although only the mandibular molar tooth germs were arrested at the bud stage. Our RNAseq analysis revealed that the wild-type maxillary and mandibular molar mesenchyme exhibit significantly different gene expression profiles, which is consistent with the hypothesis that maxillary and mandibular molar development employ distinct molecular pathways. However, we found that two secreted Wnt antagonists, Dkk2 and Wif1, are expressed at much higher levels in the mandibular than maxillary molar mesenchyme. Wnt signaling is required in both the epithelium and mesenchyme for either maxillary or mandibular tooth development to progress beyond the bud stage (Chen et al., 2009; Liu et al., 2008). Thus, it is likely that mandibular molar tooth development requires higher levels of mesenchymal odontogenic activity than the maxillary molar germs to overcome the significantly higher levels of Wnt antagonists. In the Bmp4^f/f;Wnt1Cre mutant embryos, expression of Dkk2 and Osr2 are further increased in the mandibular molar mesenchyme. Genetic reduction of Osr2 gene dosage by 50% in Bmp4^f/f;Wnt1Cre mutant embryos was able to rescue the mandibular first molar tooth development through the bud-to-cap transition. As Osr2 is known to antagonize Msx1-mediated activation of mesenchymal odontogenic activity (Zhang et al., 2009), and mesenchymal Bmp4 expression was not restored in the Osr2^+/– Bmp4^f/f;Wnt1Cre mutant embryos, a logical explanation is that reducing Osr2 expression allowed sufficient activation of other odontogenic signals by Msx1 to drive the mandibular first molar tooth germs through the bud-to-cap transition in these embryos. Thus, although maxillary and mandibular molar tooth germs exhibit distinct gene expression profiles, the distinct phenotypes of maxillary versus mandibular molar tooth development in Bmp4^f/f;Wnt1Cre mutant embryos appear to be due to the mandibular molar mesenchyme expressing higher levels of antagonists of the common odontogenic pathways than the maxillary molar mesenchyme. Further analysis of the molecular mechanisms underlying the differences in maxillary and mandibular molar tooth phenotypes in the Inhba^–/– mutant mice is necessary to clarify this issue.

Using tooth germ explant cultures, Kavanagh et al. (Kavanagh et al., 2007) showed that mouse mandibular first molar tooth germ inhibited second molar tooth development and they proposed an inhibitory cascade model, in which initiation of the posterior molar depended on a balance between intermolar inhibition and mesenchymal activation. Similar activator-inhibitor mechanisms have been proposed for periodic dentition patterning in other vertebrates (Smith, 2003; Streelman et al., 2003). However, the molecular mechanisms controlling sequential tooth initiation have not been elucidated. In this study, we found that the Bmp4^f/f;Wnt1Cre mutant mice developed mineralized maxillary first and second molars but lacked the third molar. Reducing the Msx1 gene dosage by 50% in the Bmp4^f/f;Wnt1Cre mutant background blocked maxillary second molar development. By contrast, reducing the Osr2 gene dosage by 50% partially rescued mandibular first molar development and complete knockout of Osr2 also partially rescued mandibular second molar development in the Bmp4^f/f;Wnt1Cre mutant mice (Fig. 11). Whereas the downregulation of Msx1 expression in the first molar mesenchyme and lack of expansion of Msx1 expression into the supernumerary tooth mesenchyme in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant embryos support a crucial role of the Bmp4-Msx1 positive feedback loop for propagation of Msx1 and Bmp4 expression during sequential tooth development, the findings that both supernumerary tooth germ and second molar development were initiated in the Osr2^–/–Bmp4^f/f;Wnt1Cre mutant embryos but not in Msx1^–/–Osr2^–/– mutant embryos, which had clearly detectable Bmp4 expression in the molar mesenchyme (Zhang et al., 2009), indicate that other Msx1-dependent mesenchymal odontogenic activity, in addition to Bmp4, is required for sequential tooth development. Taken together, these data indicate that Bmp4 and Msx1 positively regulate each other’s expression in the developing tooth mesenchyme and that Msx1 activates other mesenchymal odontogenic factors in addition to Bmp4 to drive tooth morphogenesis through the bud-to-cap transition as well as to induce sequential tooth formation.

We thank Steven Potter and Eric Brunskill for technical advice on laser capture microdissection and RNAseq data analyses. We thank the Gene Expression Microarray Core and the Genetic Variation and Gene Discovery Core facilities at Cincinnati Children’s Hospital Medical Center for RNA quality analysis, cDNA amplification, and next-generation sequencing services.