Revitalising the rudimentary replacement dentition in the mouse

Considerable effort has been made in the past decade using a rich variety of animal models to understand the molecular mechanisms underlying tooth replacement (Handrigan et al., 2010; Wu et al., 2013; Jussila et al., 2014; Yamanaka et al., 2007; Järvinen et al., 2008; Gaete and Tucker, 2013; Fraser et al., 2006). Across the animal kingdom, tooth development is a conserved and tightly regulated process, involving two-way molecular communication between an initially inductive epithelium and neural crest-derived mesenchyme. Together, these two tissues form the tooth germ, passing through well-described developmental stages (bud, cap, bell) before starting to produce mineralised tissue and finally erupt and become functional (Luckett, 1993). As teeth are worn down throughout life, new replacements are often required. Whereas certain animals possess the ability to replenish their lost dentitions continuously throughout their lifetime (Gaete and Tucker, 2013; Rasch et al., 2016; Handrigan and Richman, 2010b), in many others replacement is either limited to two generations or does not occur at all. This restriction to two generations, as observed in most mammals, is linked to the evolution of more complex tooth shapes adapted for certain diets, different tooth types and precise occlusion. Thus, there appears to be a trade-off between tooth replacement number and complexity (Järvinen et al., 2006; Tucker and Fraser, 2014).

In mammals, replacement teeth form from the dental lamina on the lingual side of the first tooth, known as the successional lamina (Järvinen et al., 2009; Juuri et al., 2013; Olley et al., 2014; Ooe, 1981; Berkovitz, 1972; van Nievelt and Smith, 2005; Wang et al., 2007). Later in embryonic development, and after the first generation of teeth have commenced development, the dental lamina begins to regress via a combination of processes, including cell death, disruption of the basement membrane and epithelial-to-mesenchymal transformation, preventing further tooth development (Buchtová et al., 2012). Sox2, a marker of stem/progenitor cells, localises to the dental lamina in diphyodont and polyphyodont animals (Juuri et al., 2013; Gaete and Tucker, 2013) and Sox2-positive cells in the dental placode give rise to all epithelial lineages of the incisor and molar teeth (Juuri et al., 2012, 2013). In the successional lamina, Sox2 expression is excluded from the very tip of the lamina, a region that expresses high levels of Wnt signalling, as shown by expression of the Wnt effector Lef1 (Juuri et al., 2013; Handrigan and Richman, 2010a; Gaete and Tucker, 2013) and localisation of nuclear β-catenin (Jussila et al., 2014). This complementary pattern of expression of Sox2 and activation of the Wnt pathway in the successional lamina is conserved in both diphyodont mammals and reptiles and therefore appears to be an essential component of tooth regeneration.

In taste buds and lung, Sox2 has been shown to antagonise Wnt signalling (Okubo et al., 2006; Hashimoto et al., 2012) with Sox2 capable of directly binding to the β-catenin/TCF/LEF1 transcriptional complex (Mansukhani et al., 2005). The mutually exclusive expression of Sox2 and canonical Wnt signalling in the successional lamina therefore suggest that Sox2 may also inhibit canonical Wnt signalling in this part of the tooth, confining Wnt activity to the cells at the tip of the lamina.

Although most mammals have two sets of teeth, a deciduous dentition replaced by a permanent dentition, the mouse possesses only one tooth generation. In order to cope with tooth wear, the incisors of the mouse are continuously growing (hypselodonty) with epithelial and mesenchymal stem cell niches housed within and around the labial cervical loop to provide a constant source of ameloblast and odontoblast progenitors (Li et al., 2012; Biehs et al., 2013; Kaukua et al., 2014; Juuri et al., 2012). The labial cervical loop expresses Sox2, with canonical Wnt signalling excluded from this region (Juuri et al., 2012; Sun et al., 2016). The mouse molars are not continuously growing and form from a single placode in an anterior-to-posterior direction. Molars of mammals are not replaced, this ability having been lost more than 200 million years ago (Kielan-Jaworowska et al., 2004). Despite this, a transient epithelial rudiment known as the rudimentary successional dental lamina (RSDL) is observed in many mammals during molar development (Dosedělová et al., 2015). The mouse RSDL protrudes from the lingual side of molar tooth germs, in a region equivalent to the lamina that forms successional teeth in diphyodont mammals, and houses Sox2-positive cells (Dosedělová et al., 2015). The RSDL shows high proliferation during early stages of formation [embryonic day (E) 16.5-E18.5]; however, after this point proliferation is reduced and the lamina regresses (Dosedělová et al., 2015).

Why this RSDL regresses and fails to make a second tooth germ is an interesting question, given the initial formation of a Sox2-positive structure. One possibility is that Sox2 inhibits Wnt activity in the tip of the forming successional lamina, preventing formation of a second tooth. Wnt activity has been shown to be absent, for example, in the regressing successional dental lamina of bearded dragon teeth, which are not replaced (Handrigan and Richman, 2010b).

In the alligator, nuclear β-catenin expressed in the dental lamina during the initiation phase has been linked with an involvement of the Wnt pathway in regulating the quiescent and proliferative states of dental tissues during tooth regeneration (Wu et al., 2013), and in the leopard gecko dental epithelium Wnt signalling has been suggested to regulate stem cell fate by inducing slow-cycling stem cells to proliferate (Handrigan et al., 2010).

In mice, there have been a number of studies in which activation of canonical Wnt signalling using different transgenic lines in the oral epithelium has led to the production of a large number of supernumerary teeth (Järvinen et al., 2006; Xavier et al., 2015; Wang et al., 2009). In snakes, a similar general overexpression of Wnt signalling also leads to the formation of unregulated tooth production in explant culture (Gaete and Tucker, 2013). In the mouse, overexpression of Wnt/β-catenin in the mesenchyme has recently been shown to inhibit tooth number, with overexpression of Wnts in the mesenchyme counteracting the effects of overexpression in the epithelium (Järvinen et al., 2018). Precise regulation of Wnt signalling is therefore essential for the correct control of the number of replacement tooth generations.

Here, we have investigated the mechanisms preventing replacement tooth formation in the mouse and have attempted to re-awaken the tooth programme in the RSDL. In order to achieve this, we compared progenitor marker expression in the dental tissues of the minipig, a mammal with two sets of teeth, with those of the monophyodont mouse, to pinpoint similarities and differences that might be responsible for cessation of tooth replacement. Based on species-specific differences in progenitor cell localisation, we then tested whether activation of signalling in the rudimentary dental lamina could revitalise tooth replacement in the mouse. These data could lead to a better understanding of the molecular mechanisms required for tooth regeneration, with important practical implications for the field.

In diphyodonts, such as the minipig, the forming tooth germs are connected by a dental lamina, a continuous epithelial sheet spanning the length of the jaw. The dental lamina between the tooth germs is called the interdental lamina. The interdental lamina does not produce teeth, in contrast to the tooth-producing successional dental lamina located adjacent to the first generation teeth. Sox2 and Sox9 have recently been shown to have complementary patterns of expression in the growing molar tail, which will form the serially added molars at the back of the mouth (Gaete et al., 2015). We therefore analysed the expression of these two progenitor markers in the two parts of the minipig dental lamina, and compared their expression with that in the RSDL in the mouse.

In the minipig, Sox2 was localised on the lingual side of the successional lamina of the fourth premolar, and was markedly absent from the apical tip and the labial end of the structure, in agreement with expression shown in other diphyodonts (Juuri et al., 2013) (Fig. 1A). In contrast, in the interdental lamina, just anterior to the fourth premolar, Sox2 was present all the way to the apex (Fig. 1B). Sox9 was expressed in a complementary pattern to that of Sox2, with fewer positive cells in the lingual part of the deep dental lamina and abundant positive cells in the middle epithelium (Fig. 1E). Sox9, like Sox2, however, was excluded from the labial and lingual epithelium at the tip of the successional dental lamina (Fig. 1E). In the interdental lamina, Sox9 was expressed in the lingual epithelium all the way to the tip (Fig. 1F). The lingual and labial epithelium at the tip of the successional lamina housed mostly Wnt-active cells, shown by nuclear localisation of β-catenin in both the epithelium and surrounding mesenchyme when viewed at high power (Fig. 1I). The interdental lamina, in stark contrast, was completely devoid of cells with nuclear β-catenin, indicating absence of canonical Wnt signalling in this structure (Fig. 1J). This key difference could be the determining factor as to whether the dental lamina is poised to give rise to a further generation of teeth or not, with extended Sox2 and Sox9 expression potentially inhibiting Wnt activity in the interdental region to prevent further tooth formation.

In the mouse RSDL at E18.5, Sox2 was present in the RSDL and lingual oral epithelium but appeared absent from the tip. At this stage, therefore, the RSDL has some characteristics of the minipig successional lamina (Fig. 1C); however, unlike the minipig, the tip of the RSDL expressed high levels of Sox9 and no nuclear β-catenin (Fig. 1G,K). As the RSDL started to regress at postnatal day (P) 2, Sox2 and Sox9 became colocalised in the tip of the rudiment, similar to the situation observed in the minipig interdental lamina (Fig. 1D,H). Again, at this stage the mouse RSDL had no nuclear β-catenin (Fig. 1L). This data clearly suggests that it is lack of Wnt/β-catenin signalling in the RSDL that prevents the formation of a second generation of teeth, agreeing with previous analysis of rudimentary laminas in monophyodont reptiles (Richman and Handrigan, 2011).

As Sox2 can inhibit activation of canonical Wnt signalling in other systems, we wanted to bypass the potential inhibition of Wnt activity in the mouse RSDL using the cre-lox system to stabilise Wnt/β-catenin signalling in the RSDL. To do this, we constitutively activated Wnt/β-catenin signalling in Sox2-positive cells (Sox2ert2cre) and injected tamoxifen into pregnant mothers at E15.5 and E16.0. At this stage, Sox2 has a relatively restricted expression pattern in the tooth, with positive cells in the RDSL, lingual side of the dental stalk and in a few cells of the inner enamel epithelium, in addition to more widespread expression in the oral epithelium and neighbouring vestibular lamina (Dosedělová et al., 2015; Sun et al., 2016). Embryos were collected 3 days later at E18.5.

Activation of Wnt signalling in the RSDL from E16.0 resulted in the formation of an enlarged dental lamina on the lingual side of the first molar with heightened proliferation, as shown by bromodeoxyuridine (BrdU) incorporation, in the epithelium and surrounding dental mesenchyme (Fig. 2A,B). More proliferating cells were also observed on the lingual side of the second molar in the mutant compared with littermate controls (Fig. S1A,E). Laminin immunostaining was used to outline the forming RSDL, emphasising the larger size of the RSDL after activation of Wnt signalling (Fig. 2C,D). Laminin also highlighted the pattern of blood vessels in the mesenchyme, which formed concentric whirls around the developing Sox2^CreERT2/+;Ctnnb1^lox(ex3) RSDL (Fig. 2C,D).

A slightly earlier tamoxifen injection at E15.5 led to the formation of a more prominent structure forming from the free end of the dental lamina in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos (Fig. 3). These embryos exhibited a general thickening of the oral and palate epithelium, with multiple small epithelial protrusions throughout (Fig. 3A,E). The first molars, although largely normal, had a thickened dental lamina and slightly disorganised layer of preameloblasts (Fig. 3B,F,D,H). The odontoblast layer also appeared less organised, perhaps reflecting a defect in signalling from the ameloblasts to the odontoblasts (Thesleff et al., 2001). At this stage, the Sox2^CreERT2/+;Ctnnb1^lox(ex3) RSDL had formed a small cap-stage tooth germ, with a columnar inner enamel epithelium and condensed mesenchyme underneath it (Fig. 3C,G). The second molar showed a slightly more severe phenotype, with larger epithelial protrusions developing on its lingual side (Fig. S1D,H). This is consistent with the fact that the second molar is developmentally 2 days behind the first molar, and the expression of Sox2 is more widespread in the tooth at this earlier developmental stage.

The mice were mated to incorporate the floxed Tomato gene at the Rosa26 locus, which allowed the contribution of the cells with stabilised β-catenin to be followed. At E18.5, very few Tomato-positive cells could be observed in the developing tooth itself, as would be expected given the restricted dental expression of Sox2 at the time of tamoxifen injection (Fig. 3I). The red fluorescent cells were confined to the dental stalk and the RSDL, where a new tooth germ had developed from Sox2 lineage cells. As expected given the expression pattern of Sox2, nuclear localisation of β-catenin was observed in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos throughout the oral epithelium, vestibular lamina and in the putative tooth germ formed in the RSDL area (Fig. 3J, Fig. S1B,C,F,G). Interestingly, the cells with high levels of nuclear β-catenin aggregated into clusters surrounded by epithelial cells that did not have nuclear β-catenin. Positive clusters of cells were also found in the inner enamel epithelium of the second molar (Fig. S1F). Given the reciprocal relationship between Sox2/9 expression and Wnt signalling in the tip of the successional lamina in the minipig (Fig. 1), we investigated the expression of Sox9 after overexpression of Wnt signalling. In the controls, Sox9 was expressed at the tip of the RSDL (Fig. 1G, Fig. 3K), whereas in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos Sox9 was expressed in the forming successional tooth (Fig. 3L), in a similar pattern to that previously described for Sox9 at the cap stage (Kawasaki et al., 2015).

To confirm the dental identity of these epithelial structures, we performed in situ hybridisation for genes known to be pivotal for the development of tooth germs, starting with enamel knot markers. The enamel knot is an important signalling centre, which first appears within the inner enamel epithelium of the late bud tooth germ. It simulates proliferation in surrounding cells and is important for the cuspal organisation of the tooth (Järvinen et al., 2006; Jernvall and Thesleff, 2012; Jernvall et al., 1994; Pispa et al., 2004; Thesleff et al., 2001; Tucker et al., 2000). Sonic hedgehog labels the cap-stage primary enamel knot and is subsequently expressed in the cells of the inner enamel epithelium and preodontoblasts (Vaahtokari et al., 1996). Shh was absent in the RSDL of the control tooth and expressed in the preodontoblast region of the first molar (M1) (Fig. 4A). In the Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples, Shh was expressed at the centre of the cap-like epithelial growth on the lingual side of M1, marking a putative enamel knot (Fig. 4A,F). Fgf4, another enamel knot marker (Kettunen et al., 1998), was also expressed in the same region (Fig. 4B,G). Fgf3 normally marks the enamel knot epithelium and underlying mesenchyme at the cap stage (Kettunen et al., 2000). Fgf3 was absent in the RSDL area in the control situation, whereas in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples it was present in the enamel knot-like structure and adjacent condensing mesenchyme (Fig. 4C,H). Bmp4 is normally expressed in the enamel knot and condensing mesenchyme at the cap stage (Åberg et al., 1997; Laurikkala et al., 2003). Bmp4 was expressed in the epithelium of the putative ectopic tooth germ, but, interestingly, was not expressed in the underlying mesenchyme (Fig. 4D,I). Ectodin (also known as Sostdc1) is a Wnt and Bmp inhibitor that is expressed around developing teeth but excluded from the enamel knot. In addition, in the ferret Ectodin has been shown to be expressed at the boundary between the dental lamina and the first generation tooth (Järvinen et al., 2009). Ectodin was dramatically upregulated in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) RSDL and in patches in the oral epithelium (Fig. 4E,J). These data suggest that the epithelial protrusions situated on the lingual side of the first molar dental lamina are cap-stage supernumerary tooth germs and recapitulate normal tooth development and associated gene expression. Given the odontoblast phenotype in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) M1, it was also evident that gene expression was disrupted in this area, with downregulation of Bmp4 and Fgf3 in the mesenchyme underlying the ameloblasts (Fig. 4C,D,H,I).

All of the aforementioned genes were also found to be expressed in the protrusions on the lingual side of the Sox2^CreERT2/+;Ctnnb1^lox(ex3) second molar (Fig. S2A-E,G-K). Pax9 normally marks the mesenchyme of the developing tooth germ and was strongly expressed in the mesenchyme surrounding the epithelial protrusions on the lingual side of the second molar (Fig. S2F,L). The vestibular lamina, which forms labial to the dental lamina and creates the lip furrow in later development, also produced epithelial cap-shaped protrusions that expressed enamel knot markers (Fig. S3), confirming the odontogenic potential of this tissue.

To gain an insight into the organisation of the dental tissues affected in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos, we generated 3D reconstructions from histological sections of our samples fixed at E18.5. Upon an initial inspection, the Sox2^CreERT2/+;Ctnnb1^lox(ex3) first and second molars appeared similar to controls, particularly when viewed from an oral or labial perspective. The phenotype was restricted to the lingual side of the dental lamina connecting the first molar to the oral epithelium, where cap-like structures could be visualised (Fig. 5A). In addition, the dental lamina connecting the first and the second molars together appeared thicker in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples compared with the controls (Fig. 5A).

To follow the development of these supernumerary teeth from the RSDL, we left the mice to litter down but unfortunately all pups were eaten by the mothers. We therefore dissected out the molar dental field before birth at E17.5 and either cultured the tissue or moved to a kidney capsule for longer development. TdTom Sox2^CreERT2/+;Ctnnb1^lox(ex3) were used so that the Sox2 cells and their progeny with activated β-catenin could be followed. In culture, the molar tooth germs were left to develop for up to 6 days, during which time the control tooth germs exhibited normal development with expansion of the first molar and development of the second molar (Fig. 5B,D), whereas in the Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples, ectopic tooth germs could be clearly seen developing on the lingual side of the first and second molars after 6 days in culture (Fig. 5C,E).

To assess whether the supernumerary tooth germs were able to mineralise, the dental field was dissected out from both control and Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos at E17.5 and placed into kidney capsules, where they were left to develop for 2 weeks. To avoid damaging the tooth germs, a larger piece of tissue around the molars was dissected that included the neighbouring vestibular lamina. MicroCT analysis of the kidneys revealed that the control tooth germs developed normally, forming a third molar posterior to the second and mineralising appropriately (Fig. 5F,G). In the mutant samples, multiple tooth germs formed surrounding the main first and second molars on the lingual and labial sides (Fig. 5H,I). The first and second molars themselves had relatively normal development (Fig. 5H,J). Based on the surrounding alveolar bone morphology, with larger and higher bony lamellae located on the labial side, the supernumerary teeth on the labial side (pink) of the first molar were small and conical in shape, whereas on the lingual side (green) the teeth were larger and more complex with development of cusps (Fig. 5H,J). This may reflect a difference in origin between the labially placed vestibular lamina and the lingually placed RSDL, respectively.

In a variety of systems Sox2 has been shown to inhibit Wnt signalling (He et al., 2017; Hashimoto et al., 2012; Mansukhani et al., 2005). Reduction of Wnt signalling has also been shown to lead to upregulation of Sox2 in the lung suggesting the presence of a feedback loop (Ostrin et al., 2018). In order to identify whether altering Wnt signalling has an effect on Sox2 expression tooth, we assessed the expression of Sox2 in the tissues where Wnt/β-catenin signalling had been activated. In control embryos, Sox2 could be seen localised in the nuclei of the RSDL and dental stalk, as expected (Fig. 6A). Immunostaining in the E18.5 Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos, however, revealed a clear loss of Sox2 in the dental field as well as surrounding oral epithelium (Fig. 6B). This contrasts with a presence of Sox2-positive cells in other epithelia of the same Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples, indicating that the effect of Wnt/β-catenin signalling on Sox2 is specific for the dental area (Fig. 6C).

To confirm that Wnt/β-catenin signalling has a downregulatory effect on Sox2, we dissected out molar placodes at E14.5 and cultured them in media with either 20 μM 6-bromoindirubin-3′-oxime (BIO) or DMSO for controls. BIO acts to stimulate the Wnt/β-catenin pathway by inhibiting GSK3β, the kinase which phosphorylates β-catenin to target it for degradation in the destruction complex (MacDonald et al., 2009). After culturing the placodes for 2 days, there was a marked downregulation of Sox2 compared with the DMSO controls (Fig. 6D,E). These data indicate that Wnt/β-catenin signalling negatively regulates Sox2 expression in the molar.

The experiments described above highlight the potential of the RSDL to form a tooth if given the correct stimulus. This leads to the question of what inhibits Wnt expression in the RSDL in normal development. One possibility is that the first generation tooth inhibits the formation of a successive tooth. For example, in the shrew, in which the deciduous teeth are rudimentary, it has been suggested that early activation of the permanent replacement teeth might repress the development of the first teeth (Järvinen et al., 2008). To test this possibility, we dissected slices of the first molar at E16.5, when the RSDL could be clearly observed as a bump on the lingual side of the tooth (Fig. 7A). The RSDL was then cut from the rest of the tooth and cultured in isolation (Fig. 7D,E), or the whole tooth slice (molar and RSDL) was cultured intact (Fig. 7A,B). To allow Wnt signalling to be monitored, we used Axin2-lacZ reporter mice. In the intact cultures, the RDSL was still visible after 4 days of culture as a small protrusion on the side of the first molar (Fig. 7B,C). Very low levels of canonical Wnt activity were evident in the RDSL as indicated by X-gal staining (blue) after fixation (Fig. 6B). In contrast, the isolated RSDL pieces enlarged and formed cap stage-like tooth germs with upregulated Wnt signalling in the epithelium and mesenchyme (Fig. 7F,G). Under the confocal microscope, a clear cap-shape morphology was evident, with X-gal staining in the enamel knot and underlying mesenchyme in 6/10 slices (Fig. 7H). Using phase contrast, apoptotic bodies were identified in the enamel knot epithelium, confirming the identity of the structure as a cap-stage tooth germ (Fig. 7I). We repeated the experiment a day later at E17.5. At this stage of isolation, a cap-like structure was only observed in 1/10 RSDL pieces, suggesting that the potential of the RSDL to form a tooth has reduced by this time point (data not shown).

A long-standing issue in the field of tooth replacement has been what governs whether a tooth is replaced or not. This covers why monophyodonts only have one generation of teeth, but also why most mammals are restricted to two sets.

Key to this is the molecular signals that direct whether the dental lamina forms a new tooth or starts to regress. Our results indicate a molecular identity for the parts of the dental lamina that will go on to make a second generation of teeth in diphyodont mammals. In the minipig, we have shown that it is possible to distinguish molecularly between the successional dental lamina and connected interdental lamina, by investigating the expression patterns of Sox2, Sox9 and nuclear β-catenin localisation. The successional lamina, which goes on to form a tooth, does not express Sox2 or Sox9 at its tip, which has high levels of nuclear β-catenin. This restriction of Sox2 from the tip of the odontogenic dental lamina appears to be conserved across amniotes, the exclusion of Sox2 from this region perhaps being required to maintain Wnt activity.

The RSDL in the mouse appears more similar molecularly to the interdental lamina of the minipig, and does not normally have odontogenic potential. The absence of canonical Wnt activity in the RSDL is particularly evident. Here, we show that activating Wnt signalling in the RSDL in a transgenic mouse leads to proliferation of the RSDL and the formation of a cap-stage tooth germ, which can form a mineralised tooth in culture. Similar studies have previously been published, using different modalities of activating the Wnt pathway in oral epithelia. Stabilising β-catenin in a non-inducible K14Cre line (targeting all oral and dental epithelium from E11.5) led to the formation of multiple epithelial invaginations in both jaws (Järvinen et al., 2006). When transplanted into kidney capsules, these structures formed multiple small teeth, which were able to mineralise. No indication of normal molar development was observed in these mice and this phenotype appears to arise as a consequence of repetitive formation of ectopic tooth germs within the molar placode. In contrast to this, our inducible transgenic mice driven by Sox2Cre produced a subtler phenotype with limited disruption to the first molar, and targeting of the RSDL.

Another study making use of the K14Cre transgenic line with a mosaic pattern of recombination activity showed that the development of supernumeraries is also possible when ablating a restrictive component of the canonical Wnt signalling pathway, adenomatous polyposis coli (APC; Wang et al., 2009). When APC was ablated, β-catenin was able to accumulate in the nucleus and Wnt signalling was therefore constitutively active. This mosaic pattern of Wnt activation led to a phenotype more similar to our mutant embryos, with normally developed main molars, and supernumerary teeth arising from the labial oral epithelium (between the dental and vestibular laminae) as well as from the vestibular lamina itself. Supernumerary teeth can also form around the continuously growing murine incisors throughout adulthood (Wang et al., 2009; Xavier et al., 2015).

Revitalisation of the RSDL appears to involve enhancement of proliferation in the epithelium in this region and was accompanied by condensation of the surrounding mesenchyme and recruitment of blood vessels around the RSDL as a tooth bud started to form. Our in situ hybridisation assays showed that the lingual supernumerary tooth germs recapitulate normal gene expression. Tooth germs were also evident in the vestibular lamina. The Sox2⁺ vestibular lamina develops with intimate connection to the molars (Peterkova et al., 2014) and in a normal situation lacks expression of genes related to tooth development. However, like the tooth germs on the lingual side of the tooth, these labial tooth germs formed after activating Wnt signalling were able to produce mineralised teeth. The ability of the vestibular lamina to form teeth may be linked to its developmental origin, as lineage-tracing experiments have shown that the two laminas (dental and vestibular) form from a common Shh-expressing placode (Hovorakova et al., 2016). In patients, odontomas, tooth-like tumours, have been observed to form in the lip furrow region, again suggesting that this tissue has odontogenic potential that can be re-awakened in pathological situations (Hovorakova et al., 2016). These results suggest a discreet balance in the Sox2-Wnt/β-catenin network is required for normal development and identity of oral ectodermal organs. Tipping this balance in favour of Wnt/β-catenin can lead to a change in fate of the vestibular lamina to a dental fate.

An interesting observation was the loss of Sox2 protein in the dental field and oral epithelium in the mutant embryos, but not in nasal and soft palate epithelia. This suggests dynamic and tissue-dependent relationships between Sox2 and Wnt signalling. In taste buds and lung, Sox2 negatively regulates Wnt expression (Okubo et al., 2006; Hashimoto et al., 2012). In the airway submucosal glands, the situation has been shown to be more complex with Sox2 having both inductive and repressive effects on Lef1 depending on the presence of other factors (Xie et al., 2014). In dental tissue, Sox2 expression and Wnt activity are complementary, as shown here in the dental lamina, but also in the mouse incisor (Juuri et al., 2012, 2013), suggesting a repressive effect. However, when Sox2 is downregulated in tooth culture using siRNA, Wnt targets are downregulated (Lee et al., 2016), and overexpression of Lef1 can partially rescue the dental defects in conditional Sox2 knockout mice, suggesting that Sox2 can positively regulate Wnt activity (Sun et al., 2016). In addition to the effect of Sox2 on Wnt activity, Wnt activity can impact on Sox2 expression. In the lung, reduction of Wnt signalling at E11 led to ectopic Sox2 expression in the epithelium (Ostrin et al., 2018). This effect, however, was highly stage dependent with later loss having no impact on Sox2, even though Sox9 levels were downregulated at both time points. Wnt activity therefore appears to negatively regulate Sox2, but only in some contexts. This agrees with our findings that activating Wnt signalling in culture or in vivo leads to a loss of Sox2 in the dental epithelium, but not other associated epithelia, suggesting a negative-feedback loop in action specifically in the dental lamina. This highlights the importance of carefully controlling the balance between stemness and Wnt-directed differentiation for limited or unlimited tooth formation. It is possible that the Wnt-driven loss of Sox2 in the lamina was enough to stimulate tooth development. This could be investigated further by specifically downregulating Sox2 in the RSDL without altering Wnt signalling.

During mammalian evolution, reduction of Wnt signalling at the tip of the successional lamina may have led to an expansion of Sox2 expression into this region. Presence of Sox2 would have led to further inhibition of Wnt signalling at the tip of the lamina, reinforcing the negative-feedback loop and leading to a complete loss of Wnt activity and subsequent loss of odontogenic potential. This would then have been followed by physical loss of the lamina, using a combination of apoptosis and epithelial-mesenchymal transition, preventing the possibility of any further tooth replacement (Buchtová et al., 2012).

It has previously been suggested that replacing and successional teeth might inhibit the formation of each other. For example, the successionally developing molars of the mouse inhibit each other, with the second molar forming precociously when isolated from the first molar in culture (Kavanagh et al., 2007). Similarly, the early development of the shrew permanent dentition has been suggested to inhibit the development of the deciduous tooth germs (Järvinen et al., 2008). Here, we show that removal of the first generation tooth in the mouse dramatically frees the RSDL so that it can now form a tooth bud. Given the need for Wnt signalling to initiate tooth development, it is tempting to speculate that a Wnt inhibitor is expressed by the main tooth, inhibiting Wnt activity in the RSDL. Interestingly, the soluble Wnt inhibitors dickkopf 2 and 3 have been shown to be strongly expressed in the mesenchyme around the developing RSDL at E16 and E18 and may act to inhibit Wnt activity in this region (Fjeld et al., 2005). Restriction of the mouse to one tooth generation could therefore be reliant on the relative positioning and timing of development of the first tooth and the successional lamina.

Although most mammals have two sets of teeth, it may be possible to produce a third set by controlled stimulation of the Wnt pathway. A rudimentary successional lamina from a permanent tooth has been described in both human and bat embryos, suggesting the potential for a third set of teeth (Ooe, 1981; Popa et al., 2016). Forming a third set, however, would only be feasible by targeting a small window of time after formation of the replacement dentition but before loss of the dental lamina, and would not be possible later in life.

Strain LiM minipig embryos were collected at the Libechov animal facility in the Czech Republic (ethical approval 020/2010). Insemination day was regarded as day 0 of gestation. Embryos were collected at E67 and fixed in 4% paraformaldehyde (PFA) overnight. Samples were washed in PBS and dehydrated through a series of increasing methanol concentrations, followed by isopropanol. Heads were embedded in paraffin wax after clearing in 1,2,3,4-tetrahydronaphtalene. Samples were sectioned frontally at a thickness of 5 µm and collected for fluorescence immunohistochemical staining. Sections showing the successional lamina were taken at the level of the fourth premolar, whereas the interdental lamina sections were anterior to the fourth premolar.

Sox2^CreERT2 mice were generated and described by Andoniadou et al. (2013). The Ctnnb1-lox(ex3) mice have been previously described by Harada et al. (1999). The two strains were mated to generate Sox2^CreERT2/+;Ctnnb1^lox(ex3) embryos. Mice were also mated to Rosa-tdTomato (Madisen et al., 2010) heterozygous mice to generate Sox2^CreERT2/+;Ctnnb1^lox(ex3)/+; R26^TOM/+ mice in order to allow overexpressing cells to be followed. Axin2-lacZ mice were used as previously described as a readout of canonical Wnt activity (Lohi et al., 2010).

Mating of C57Bl6 adult mice was set up in the evening, with vaginal plug checks performed in the morning. E0.5 was marked at midday of the day a vaginal plug was observed. Adult pregnant females were injected intraperitoneally with 20 mg/ml tamoxifen (Sigma-Aldrich) and 10 mg/progesterone (Sigma-Aldrich) dissolved in 10% ethanol in corn oil at E15.5 and E16.0 and culled by exposure to CO₂ gas at E17.5 (for kidney capsule and explant culture experiments) or E18.5 (for wax embedding). Controls were injected with corn oil only. A single BrdU (Sigma-Aldrich) injection was given to pregnant mice 30 min before culling.

Embryos were dissected out, genotyped and processed for wax embedding. For histology studies, 8-μm-thick paraffin sections were stained with Picro Sirius Red, Haematoxylin and Alcian Blue (trichrome stain).

For this study, n=8 Sox2^CreERT2/+;Ctnnb1^lox(ex3) and n=8 control [tamoxifen-injected Sox2^CreERT2/−;Ctnnb1^lox(ex3) or corn oil-injected Sox2^CreERT2/+;Ctnnb1^lox(ex3)] embryos from three separate litters were used. For the 6-day culture experiment, n=3 control and n=5 Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples were used. For the kidney capsule experiments, n=2 control and n=3 Sox2^CreERT2/+;Ctnnb1^lox(ex3) samples were used. For the BIO culture experiments, n=3 control and n=4 treated samples were assessed for whole-mount immunohistochemistry.

All culling followed Schedule One methods as approved by the UK Home Office and was performed by trained individuals. All procedures were performed under approved personal and project licences in accordance with the Animal (Scientific Procedures) Act of 1986, United Kingdom.

Paraffin sections were dewaxed in xylene and rehydrated through 100%, 90% and 70% methanol, followed by a wash in deionised water. Heat-mediated antigen retrieval was performed in a solution of Tris-EDTA, pH 9.0 (DAKO) for 15 min, followed by cooling and incubation with blocking solution (10% goat serum, 0.1% Triton X-100 in PBS). Sections were incubated at 4°C overnight with primary antibodies (rabbit anti-Sox2, 2748s, Cell Signaling Technology; rabbit anti-Sox9, AB5535, Sigma-Aldrich; rabbit anti-β-catenin, C2206, Sigma-Aldrich; rabbit anti-laminin, L9393, Sigma-Aldrich; rat anti-BrdU, AB6326 (Abcam) at a concentration of 1:200 in blocking solution. Incubation with the secondary antibody [Alexa 568 goat anti-rabbit (A11034) or goat anti-rat (A11077), Invitrogen] was carried out the next day over a period of 2 hours at room temperature. All antibodies used showed expression that agreed with published expression patterns in other species, or in other tissues in the head. Imaging was carried out using a Leica SP5 confocal microscope. Images were cropped and adjusted for levels using Adobe Photoshop.

Mouse probes for Shh (EcoR1, T7), Fgf3 (HindIII, T7), Fgf4 (XbaI), Bmp4 (EcoR1, Sp6), Pax9 (BamHI, T3) and ectodin (NotI, T7) were linearised and transcribed. The Bmp4 probe was a gift from Prof. Karen Lyons (University of California at Los Angeles, USA; Lyons et al., 1990); the ectodin probe (Wise) was a gift from Dr Atsushi Ohazama (King’s College London, UK; Ohazama et al., 2008); Fgf3 and Fgf4 probes were gifts from Prof. Ivor Mason (King’s College London, UK; Mahmood et al., 1996); the Pax9 probe was a gift from Dr Heiko Peters (Newcastle University, UK; Peters et al., 1998); and the Shh probe was a gift from Prof. Andy McMahon (University of Southern California, Los Angeles, USA; Bitgood and McMahon, 1995). In situ hybridisation of craniofacial sections was carried out as previously described by Gaete et al. (2015).

Serial histological sections containing the upper molar tooth set of E18.5 control and Sox2^CreERT2/+;Ctnnb1^lox(ex3)/+ embryos were imaged in an anterior-to-posterior sequence. The images were imported in Fiji (Fiji is Just ImageJ 1.47v), grouped as a stack and aligned using a combination of automatic and manual alignments. The dental tissues were painted manually over each image and from this a 3D object was generated, which was then photographed from labial, lingual and oral perspectives.

Sox2CreGOFTOM pregnant females were injected with tamoxifen and progesterone at 15.5 days of gestation and embryos were collected at E17.5. Control and Sox2^CreERT2/+;Ctnnb1^lox(ex3) tooth germs were carefully dissected out in DMEM/F12 supplemented with penicillin and streptomycin. Three 7-week-old CD1 males were anaesthetised using 80 mg/kg ketamine and 16 mg/kg xylazine (Sigma-Aldrich) by intraperitoneal injection. The area of incision on the dorsal right side of the mouse was cleaned with 70% ethanol, then a 1-cm-long incision was made above the right kidney area. The kidney was gently placed outside of the abdominal cavity and two dental organ explants were inserted under the membrane covering the kidney. The kidneys were then placed back inside the abdominal cavity, and the incision was sutured. Two weeks after the surgery, the mice were sacrificed in a CO₂ chamber and the kidneys were removed, fixed in 4% PFA and scanned using a GE locus SP microCT scanner. Images were reconstructed using microview software. After scanning, samples were decalcified and prepared for histological analysis.

Lower jaw molar regions were dissected at E14.5 and incubated with Wnt activator BIO for 2 days using a modified Trowell culture method. The BIO was dissolved in DMSO and added to Advanced Dulbecco's Modified Eagle Medium F12 (DMEM F12) culture medium (plus Glutamax and Pen/Strep) at a final concentration of 20 μM. Control cultures were incubated with DMSO. After 2 days, samples were processed for whole-mount immunofluorescence.

Lower jaws were dissected at E16.5 and E17.5 from wild-type and Axin2-lacZ/+ embryos. The molar placodes were isolated and the bone removed using fine tungsten needles. Placodes were then sliced into 250-μm-thick frontal sections using a McIlwean tissue chopper. First molar slices with clear RSDL on the lingual side were selected. Slices in which the RSDL was not clear were excluded. In some cases, the slices were cultured for several hours before cutting to allow more accurate visualisation of the RSDL. The RSDL was then separated from the rest of the developing tooth using a scalpel blade in half the slices (n=10 for each stage). Slices were photographed before and after isolation using a MZFLIII Leica microscope and cultured on filters for 4 days (as described by Alfaqeeh and Tucker, 2013). After culture, the slices were fixed for 15 min in 4% PFA. Axin2-lacZ/+ positive slices were stained with X-gal for 6 h at 37°C. Slices were then permeabilised in PBT (phosphate-buffered saline plus 0.1% Tween) for several hours, and then placed in DAPI to stain the nuclei. Slices were mounted and photographed on a Leica SP5 confocal microscope.

Thanks to Marcia Gaete for help culturing and photographing the tooth germ slices and to Juan Fons Romero for confocal imaging of the slices.