Tinkering with the inductive mesenchyme: Sostdc1 uncovers the role of dental mesenchyme in limiting tooth induction

According to Dollo's Law, evolution is not substantially reversible and a structure lost during evolution is unlikely to reappear. Whereas evolutionary reappearance of recently lost structures can still be assumed to be relatively common (Marshall et al.,1994), long-lost structures are exceedingly rare to reappear. Even long-lost features, however, may still be retained as vestigial structures before disappearing altogether. For example, the developing mouse dentition contains potential to make more teeth than are present in the adult dental pattern, which is often attributed to be due to the presence of vestigial teeth (Peterkova et al., 2002; Keränen et al., 1999). Vestigial tooth germs have been reported in the mouse toothless diastema region between the incisors and the first molar(Peterkova et al., 2002), and also in the mouse incisor region(Peterkova et al., 2002). These tooth rudiments, which would have been retained for over 45 million years, do not develop beyond early invagination of the epithelium, or the early bud stage, and they degenerate apoptotically during later development(Keränen et al., 1999; Peterkova et al., 2002). In addition to these missing tooth loci, rodents have lost the ability for tooth replacement from deciduous to permanent incisors. The loss of successional replacement of molars is evolutionarily basal for all living mammals.

Developmentally, vestigial mouse tooth rudiments disappear after the inductive potential of tooth formation has shifted from the epithelium to the mesenchyme (Vainio et al.,1993; Mina and Kollar,1987; Kollar and Baird,1969). The neural crest-derived dental mesenchyme can induce tooth formation when combined with non-dental epithelium, raising the question of what mechanism prevents the full development of rudimentary teeth? Several mutant and transgenic mouse strains have supernumerary molars, predominantly anterior to the first molar, where previous reports have identified vestigial tooth buds (Mustonen et al.,2003; Klein et al.,2006; Kassai et al.,2005). The analysis of the developmental origins of these supernumerary molars has indicated that they develop from the vestigial buds,and thus appear to `rescue' the teeth lost during evolution(Mustonen et al., 2003; Klein et al., 2006; Kassai et al., 2005). At the molecular level, these works suggest a role for multiple feedback loops activating and inhibiting tooth development and number, but it remains unknown how these interactions may modulate the inductive potential of the dental mesenchyme.

A signaling pathway affecting both incisor and diastema vestigial teeth in mice is the bone morphogenetic protein (BMP) pathway. BMP4 is required for the shift of the inductive potential from the presumptive dental epithelium to the dental mesenchyme after embryonic day 12.5 (E12.5)(Vainio et al., 1993; Chen et al., 1996). Epithelial BMP4 induces the expression of the homeobox gene Msx1, which is required for advancing tooth development and for the shift of Bmp4expression to mesenchyme (Vainio et al.,1993; Chen et al.,1996). In Msx1 mutants, development of all teeth is arrested at the early bud stage, and mesenchymal Bmp4 expression is lacking (Chen et al., 1996). This phenotype can be rescued by forced Bmp4 expression in the mesenchyme (Zhao et al., 2000; Bei et al., 2000). Tooth development also ceases at the bud stage in mice that lack the functional BMP receptor type 1A in the epithelium (Andl et al., 2004).

Whereas partial downregulation of BMP signaling decreases tooth number(Plikus et al., 2005),deletion of the BMP antagonist Sostdc1 rescues the vestigial molar tooth buds (Kassai et al.,2005). Sostdc1, which was discovered three times independently and has also been called ectodin, Usag1 and Wise (Laurikkala et al.,2003; Yanagita et al.,2004; Itasaki et al.,2003), is a secreted molecule identified as a BMP inhibitor binding to BMPs with high affinity(Laurikkala et al., 2003; Yanagita et al., 2004). BMPs themselves induce the expression of Sostdc1(Laurikkala et al., 2003). Sostdc1 has also been shown to regulate the Wnt pathway in a context-dependent manner. It is able to compete with Wnts for binding to the Wnt co-receptor Lrp6 (Itasaki et al.,2003).

We have previously shown that Sostdc1 affects the morphology of mouse molars by restricting the size and the placement of epithelial signaling centers: the enamel knots (Kassai et al.,2005). In addition, in Sostdc1 knockout mice, an extra tooth develops in the location of the vestigial premolar primordium(Kassai et al., 2005) and in the incisor region (Murashima-Suginami et al., 2007) (Fig. 1). In this study, we used Sostdc1 null mutant mice(Mus musculus Linnaeus) to explore the role of the mesenchyme in modulating tooth number. We focused on the incisor because previous reports have suggested the presence of multiple vestigial incisor buds in mice(Peterkova et al., 2002). Compared with living rodents, ancestors of rodents had both higher number of incisors and replacement of incisors from deciduous to permanent teeth. This raises the issue of whether the extra incisors in Sostdc1-deficient mice represent rescued teeth or rescued tooth replacement. Our results show that the extra incisors in Sostdc1-deficient mice are likely to be the latter: successional incisors. Furthermore, we show the amount of mesenchymal tissue alone may influence the outcome between induction and inhibition of extra teeth and that Sostdc1 is a central modulator of the inductive potential of dental mesenchyme.

The Sostdc1-deficient mouse line used in this study has been described previously (Kassai et al.,2005) and the heterozygous littermates were used as controls. Wild-type embryos were from the NMRI strain and from heterozygous Shh-GFP mice, which express a reporter gene [green fluorescence protein (GFP)] under a Shh-promoter(Harfe et al., 2004). Sostdc1-deficient mice were crossed with the Shh-GFP reporter line. The Catenb^Δex3k14/+ mice have been described previously (Järvinen et al.,2006). Two different reporter mouse lines were crossed with Sostdc1-deficient mice to study the localization of Wnt activity:TOPgal (Jackson laboratory) and BATgal(Maretto et al., 2003). The reporter constructs consist of multiple TCF/Lef-binding sites and a WNT target gene minimal promoter driving the expression of lacZ reporter gene(DasGupta and Fuchs, 1999; Maretto et al., 2003). The age of the embryos was estimated from the appearance of the vaginal plug (E0) and from their exterior features.

The teeth with some surrounding tissue were dissected from the lower jaw of E12 to E14 embryos and cultured using a Trowell type organ culture system(Sahlberg et al., 2002). The medium was changed every second day and contained DMEM and F12 (Ham's Nutrient Mix: Life Technologies) (1:1) supplemented with 10% fetal calf serum (PAA laboratories, Pasching, Austria), 150 mg/ml ascorbic acid, glutamine and penicillin-streptomycin. The explants were photographed every day using light(Olympus SZX9) or fluorescent (Leica MZFLIII) microscopy.

Recombinant BMP4 (0.5 ng/μl R&D), Noggin (1 ng/μl R&D), Dkk1(1 ng/μl R&D) and Sostdc1 (1 ng/μl, a kind gift from Dr N. Itoh,Kyoto, Japan) proteins were added into the medium. Canonical Wnt signaling was activated using a specific Gsk3beta inhibitor BIO (2 mM Calbiochem). In bead experiments, Affi-gel agarose beads (Biorad) were soaked in recombinant BMP4(100 ng/μl R&D) or BSA, and they were placed on top of the tooth explants.

Heads from wild-type and Sostdc1 knockout mouse embryos (E12-E17)were dissected and fixed with 4% paraformaldehyde (PFA) at +4°C overnight. They were dehydrated paraffin-embedded and serially sectioned at 7 μm. Sections were counterstained with Hematoxylin and Eosin. Jaws for whole-mount in situ hybridization were fixed similarly. The tissue culture samples were treated with 100% ice-cold methanol for 5 minutes before fixing with 4% PFA. Sample preparation was same for the TUNEL staining.

Digital pictures were taken of frontal serial sections of the incisor. For the shape 3D reconstruction, the pictures were imported into a stack with Scion Image software (version 4) and the individual slices were aligned with the register function using the midline of the jaw as reference. The epithelial shape was manually traced in Adobe Photoshop CS2 (PS) with the brush function. The distance between the sections (7 μm) was translated to a corresponding pixel distance based on a reference scale bar. The stack with the modified slices was re-sliced at a perpendicular angle in order to create a new stack with only one pixel distance between individual slices. The jagged epithelial shape in the resulting slices were smoothened in PS with a batch command: select white, expand five pixels, smooth 15 pixels, contract four pixels, fill white (shape), invert selection, fill black (background). The E16 Sostdc1 -/- stack was smoothened entirely manually using the brush function in PS. A projection was made of the stack and the sagittal view selected for presentation.

For the 3D reconstruction of expression patterns, first the shape 3D reconstruction was created according to the previous protocol. Subsequently,the blue X-Gal staining was isolated by applying the high contrast red filter with the `black&white' command in PS, the levels adjusted and inverted. The epithelial expression of interest was isolated and treated in the same sequence of steps as the shape and projected on the corresponding slice of the shape projection.

Whole-mount in situ hybridization was performed by using the InSituPro robot (Intavis AG, Germany) as described earlier(Laurikkala et al., 2003). BM Purple AP Substrate Precipitating Solution (Boehringer Mannheim Gmbh, Germany)was used to visualize the digoxigenin-labeled probes. Radioactive in situ hybridization for paraffin sections was carried out according to standard protocols using ³⁵S-UTP labeling (Amersham). The following probes were used: murine Shh (Vaahtokari et al., 1996), p21(Jernvall et al., 1998), Msx2 (Jowett et al.,1993), Sostdc1(Laurikkala et al., 2003; Kassai et al., 2005) and Bmp4 (Vainio et al.,1993).

Apoptotic cells were visualized by detecting DNA fragmentation in whole-mount samples and sections with the ApoTag Apoptosis Detection Kit.

Localization of Wnt activity through β-galactosidase in the embryonic tissue was revealed using X-Gal staining. E13-14 lower jaws were collected in Dulbecco's 1×PBS and fixed in 2% PFA, 0.2% glutaraldehyde for 30 minutes at 4°C. Tissues were incubated for 3×10 minutes in X-gal washing buffer (2 mM MgCl₂, 0.02% NP-40 in PBS) and stained with X-gal staining solution (1 mg/ml X-Gal, 5 mM K₃Fe(CN)₆, 5 mM K₄Fe(CN)₆, 2 mM MgCl₂ in X-gal washing buffer) in room temperature. Serial sections were counterstained with Nuclear Fast Red.

Microinjections were performed by injecting fluorescent DiI [1,1′-di-octadecyl-6,6-di(4-sulfophenyl)-3,3,3′,3′-tetramethylindocarbocyanine](Invitrogen) to the in vitro cultured incisor explants. The color (diluted in 2 μg/ml DMSO) was injected to the cells that will later form the extra incisor.

We first examined the expression pattern of Sostdc1 during the early stages of wild-type mouse incisor development(Fig. 1A-E). Whole-mount in situ analysis show that Sostdc1 expression surrounds the incisor placodes at E12 stage (Fig. 1B). At E13, Sostdc1 is expressed inside the incisor (and molar) tooth germs and around them, but now a Sostdc1-negative region has formed immediately next to the tooth germs(Fig. 1D). Histological sections show that at E12 and E13, Sostdc1 is expressed in the mesenchyme (Fig. 1A), and is upregulated in the tooth forming area around E13(Fig. 1C). In the dental epithelium, Sostdc1 is upregulated at E14 when the incisor has reached the cap stage (Fig. 1E). In the E14 mesenchyme Sostdc1 expression was localized in the lingual side of incisors(Fig. 1E). In the epithelium,expression was absent from the enamel knot and its underlying mesenchyme(Fig. 1E), agreeing with the earlier reports of developing molar teeth(Laurikkala et al., 2003; Kassai et al., 2005). As in molars, the incisor Sostdc1 expression correlates with Bmp4expression (Fig. 1F), except that in incisors, these genes are strongly expressed in the lingual side(Fig. 1E,F), whereas in the molars they are expressed in the buccal side(Laurikkala et al., 2003; Kassai et al., 2005). Furthermore, compared with Sostdc1, Bmp4 expression is located in closer proximity to the dental epithelium, including in the enamel knot cells themselves and throughout the underlying mesenchyme.

Next, we examined incisor formation in the Sostdc1-deficient mice. All the studied (n>25) Sostdc1-deficient embryos had extra incisors, which are also well visible after birth. With age, however,adult mice tend to lose the extra incisors, which, when retained, are substantially smaller than the main incisors and lack the yellow-brown pigmentation, typical for the rodent incisors, at their labial surface(Fig. 1G,H).

From E12, the expression of Shh, an epithelial placode and enamel knot marker in Sostdc1-deficient jaws is stronger and broader in incisors in comparison with the controls(Fig. 1I-L; see Fig. S1 in the supplementary material). The expanded expression of Shh in the epithelium correlates with the expanded and stronger expression of p21 expression (see Fig. S1 in the supplementary material), a differentiation marker of the enamel knots(Jernvall et al., 1998). Moreover, the expanded Shh expression in the Sostdc1-deficient epithelium corresponds to stages when Sostdc1 expression is limited to the mesenchyme in the wild-type incisors (Fig. 1C).

Because rudimentary tooth germs and enamel knots are known to disappear through apoptosis (Keränen et al.,1999; Peterkova et al.,2002), we used TUNEL-staining to study effects of Sostdc1on apoptosis (Fig. 2A,B). Almost no apoptosis was detected in the incisor buds at E12 and E13, and no difference was seen between wild-type and mutant incisors (not shown). However, at E14, the cap stage Sostdc1-deficient incisor epithelium shows a marked lack of apoptosis (n=6/6) in the enamel knot, and especially in the area closer to the oral epithelium(Fig. 2A,B). There was only limited apoptosis in the mesenchyme and it appeared similar in wild type and the Sostdc1-deficient mice.

In order to examine whether the area of apoptosis in the epithelium close to oral surface of E14 wild-type incisor is associated with the removal of a vestigial incisor rudiment, we localized enamel knot markers at an earlier stage - E13-E13.5 (Fig. 2C,D). Because the Wnt pathway is another potential Sostdc1 target, and enamel knots are sites of high Wnt signaling activity(Järvinen et al., 2006),we examined activity of Wnt pathway using TOPgal and BATgal reporter mice crossed with the Sostdc1-deficient mice. Epithelial Wnt activity was detected in the incisor enamel knot both in Sostdc1-deficient mice and controls at E13-E13.5 (Fig. 2C,D; data not shown). In addition, we detected a small domain of Wnt activity in the area of intense apoptosis in the E14 wild-type incisors(Fig. 2A).

We used 3D analysis of serial sections to determine the relationship between enamel knot dynamics and extra incisor formation. The results show that the small domain of Wnt activity located orally to the incisor enamel knot is larger and longer in duration in the Sostdc1-deficient incisors compared with the wild-type incisors(Fig. 3A). Furthermore, 3D analysis of later developmental stages(Fig. 3B) show that the extra incisor develops at the location previously showing Wnt activity and intense apoptosis in the wild-type incisors. The illustrated 3D renderings are projections from the medial side, and thus the location of the extra incisor closely matches the final position of this tooth in the jaw(Fig. 1H). Taken together, the extra incisor appears not to originate from a separate bud along the dental lamina, but rather to be developmentally part of the main incisor, much like replacement teeth in mammals with more complete tooth replacements.

Because we observed early expansion of epithelial enamel knot markers at the time when Sostdc1 expression is limited to the mesenchyme(E12-13), we tested whether we could phenocopy the effect of Sostdc1deficiency by minimizing the amount of dental mesenchyme(Fig. 4). We dissected most of the mesenchyme surrounding the incisor tooth germs and cultured the explants in vitro. Removal of mesenchyme resulted in a reduction of Sostdc1expression around the dental epithelium, as confirmed by whole-mount in situ hybridization (Fig. 4A,B). Compared with the untrimmed incisors (n=0/>25)(Fig. 4D), the trimmed explants showed development of an extra incisor (n=40/55)(Fig. 4C). The bud of the developing extra incisor was typically seen after 2 days culture at the lingual aspect of the explanted E13 incisor, and during the next 4 days it grew in size and acquired the characteristic shape of an incisor. The extra incisors also developed in these wild-type explants when the incisors were dissected at E12 (n=49/59). When the mesenchyme was trimmed in E14 tooth germs, no extra incisors developed during in vitro culture (not shown).

Additionally, we tested whether the removal of the surrounding mesenchyme from wild-type molar buds would also rescue the development of the vestigial bud anterior to the first molar, which develops to an extra tooth in Sostdc1 null mutants (Kassai et al., 2005). We dissected the mandibular molar tooth germs at E12 stage and removed most of the surrounding mesenchyme. In a small number of cases (n=4/20) supernumerary teeth formed anterior to the first molars and their development resembled the phenotype of Sostdc1-deficient mice (see Fig. S2 in the supplementary material).

When the surrounding mesenchyme was trimmed from the Sostdc1-deficient incisor explants, the extra incisor that forms in vivo developed also in vitro (Fig. 5). However, the formation of extra incisors appears to continue,and additional de novo incisors appeared after a few days of culture from the lingual cervical loop area (Fig. 5A-D) (n=48/109). Because the de novo incisors develop in immediate proximity to the other incisors, we examined the origins of these teeth by microinjecting DiI-label in Sostdc1-deficient and control explants expressing GFP in the Shh locus(Fig. 6A-C). DiI was injected after 4 days of culture to the lingual epithelium of E13 incisor at the site where the `normal' extra incisor form. The DiI-label spreads during subsequent culture and labels the extra incisor as it grows in both Sostdc1-deficient and control explants(Fig. 6A-C; see Fig. S3 in the supplementary material). However, the de novo incisor in the trimmed Sostdc1-deficient explant appears as a new DiI-negative outgrowth from, or close to, the cervical loop of the main incisor(Fig. 6A-C).

Next, we tested, using the in vitro culture system, the role of BMP signaling in the development of the extra and de novo incisors in Sostdc1-deficient mice (Fig. 7). In tooth explants from Shh-GFP expressing Sostdc1-deficient mice, we observed changes in the enamel knot size after treatment with recombinant BMP4 protein. The results show that overnight treatment with BMP4 causes increased Shh expression in Sostdc1-deficient incisors, as well as in wild-type incisors where the mesenchyme has been trimmed (Fig. 7A,B).

We introduced exogenous BMP4 to dissected E13 control and Sostdc1-deficient incisor explants at the site where Bmp4 is normally co-expressed with Sostdc1 in the mesenchyme. BMP4-soaked beads were placed on the lingual side of incisor buds (owing to the flattening of in vitro cultured teeth, lingual position is an approximation) and the day when extra incisors appeared was tabulated(Fig. 7C). In control samples,extra incisors did not develop either in BMP4-treated or in the control explants (Fig. 7F). By contrast, in homozygous Sostdc1-deficient explants, BMP4 beads accelerated the formation of extra teeth(Fig. 7F). They had already started to form at the lingual aspect of the incisors during the second culture day, whereas in the absence of BMP bead, the formation of extra incisors was first seen after 4 days of culture(Fig. 7C).

When the timing of the initiation of extra incisor development was compared between the different explants, the Sostdc1-deficient teeth cultured with BMP4 beads showed the fastest initiation of extra incisors(Fig. 7C). However, the wild-type incisors with trimmed mesenchyme showed faster initiation of extra incisors than did Sostdc1-deficient incisors cultured with mesenchyme, suggesting the presence of other mesenchymal inhibitory factors in addition to Sostdc1.

We tested whether known inhibitors of BMP and Wnt signaling could mimic the function of Sostdc1 and prevent the formation of extra incisors in Sostdc1-deficient embryos. Dissected E13 incisor explants were cultured in the presence of Noggin and Dkk1, antagonists of BMP and Wnt signaling, respectively. In addition, recombinant Sostdc1 was used as a control to rescue Sostdc1 deficiency. All three inhibitors prevented the formation of the extra incisors in Sostdc1-deficient and control explants (see Fig. S4 in the supplementary material).

Stimulation of Wnt signaling in dental epithelium by forced activation ofβ-catenin causes dramatic stimulation of both molars and incisors in mice(Järvinen et al., 2006). Although the formation of extra molars in these mice was analyzed in detail and it was demonstrated to occur continuously from previously formed teeth,the details of the generation of the extra incisors were not studied. In order to examine whether there were similarities with the formation of the extra teeth in Sostdc1 knockout mice, we observed the development of Catenb^Δex3k14/+ incisors in organ culture. E13-E15 incisor tooth germs were dissected from the lower jaws of Catenb^Δex3k14/+ embryos and cultured for 1 week (see Fig. S5 in the supplementary material). The formation of the supernumerary teeth was progressive, and they formed successively, especially from the lingual incisor epithelium of previously formed incisors (see Fig. S5A in the supplementary material). These supernumerary incisors started to form earlier in vitro than in vivo, which was similar to our observations in Sostdc1-deficient explants. The de novo formation of incisors was morphologically very similar in the two mouse mutants and indicates that the lingual dental epithelium has the capacity for tooth formation. Additionally,we administered BIO, a specific inhibitor of GSK3b, into the culture medium,thus activating the canonical Wnt signaling pathways throughout the explants. No extra incisor formed in these explants (not shown), but as BIO also had inhibitory effects on the main incisor of the mutant, as well as on control teeth, this may indicate inhibitory effects of enhanced mesenchymal Wnt signaling on tooth development.

Classically, the inductive potential of tooth formation is considered to shift from the epithelium to the mesenchyme after E11(Mina and Kollar, 1987). After this, the neural crest-derived dental mesenchyme can induce tooth formation when combined with non-dental epithelium. Furthermore, several mutants are known that interfere with the mesenchymal induction of the epithelial signaling centers, enamel knots, causing the tooth development to arrest at E13 bud stage (Kratochwil et al.,1996; Bei and Maas,1998; Peters and Balling,1999; Åberg et al.,2004). Whereas the inductive role of mesenchyme in tooth development is well established (Vainio et al., 1993; Mina and Kollar,1987), our results show that induction of additional teeth can be achieved by reducing the amount of dental mesenchyme, perhaps uncovering a role for the mesenchyme in limiting excess tooth induction.

In the Sostdc1-deficient embryos, markers of tooth formation have broader expression domains, encompassing the areas giving rise to the extra teeth (Fig. 1). In themselves,these kind of broader expression domains are to be expected because similar changes have been reported in the molar region of genetically altered mouse strains with extra molars (Kangas et al.,2004; Mustonen et al.,2003; Klein et al.,2006; Zhang et al.,2003). Our results also agree with previous reports on enlarged expression domains of enamel knot markers, such as p21, in Sostdc1-deficient molars and incisors(Kassai et al., 2005; Murashima-Suginami et al.,2007; Murashima-Suginami et al., 2008). These patterns are intriguing because, until E14, Sostdc1 expression is predominantly limited to the mesenchyme(Fig. 1), and the E13 broadened expression patterns in the epithelium are thus suggestive of a role for Sostdc1 in interfering with epithelio-mesenchymal induction.

The in vitro culture system allowed us to test how reducing the amount of surrounding mesenchyme, and consequently decreasing Sostdc1expression, would affect the incisor formation in the wild-type teeth(Fig. 4). Our result, i.e. the in vitro formation of the extra incisor, essentially phenocopies the Sostdc1-deficient phenotype. The peak effect of mesenchymal trimming was reached at E13, well within the stage of development when the inductive potential for tooth formation resides in the mesenchyme(Mina and Kollar, 1987). Our work thus appears to have uncovered a complex role for the mesenchyme in the regulation of tooth induction. Initially, the mesenchyme is obviously required for normal tooth morphogenesis and the induction of epithelial enamel knots(Kollar and Baird, 1970a; Kollar and Baird, 1970b; Jernvall et al., 1998), but it might acquire inhibitory roles as the development proceeds and the successional incisor starts to develop. One possibility for the observed activation of the successional tooth after mesenchymal tissue reduction could be slightly different expression domains of genes that induce and inhibit tooth formation. For example, Sostdc1 appears to have broad expression domains that expand laterally around developing teeth(Fig. 1). Furthermore, peak Sostdc1 expression intensity in E14 cap stage is further away from the center of the tooth than Bmp4 expression(Fig. 1), which has been implicated in the induction of enamel knots and the regulation of Sostdc1 itself (Laurikkala et al., 2003; Kassai et al.,2005; Bei et al.,2000). Interestingly, the bioengineering of tooth germs from dissociated dental cells in vitro has been shown to lead to the formation of multiple incisors (Nakao et al.,2007); it is plausible that these in vitro procedures disrupt the inhibitory mechanisms of in vivo mesenchyme.

Both Bmp4 and Sostdc1 are intensely expressed in the mesenchyme at the lingual side of incisor, corresponding to the location of extra incisors in the Sostdc1-deficient mouse. This contrasts with the buccal expression bias of Sostdc1 and Bmp4 in molars(Jernvall et al., 1998; Laurikkala et al., 2003; Kassai et al., 2005). However,in molars of Sostdc1-deficient mice, the cuspal morphology is substantially altered in the buccal side. We interpret these results as further supporting the role of Sostdc1 as a feedback inhibitor of BMP4 in tooth development, and the role of BMP4 in regulating the lateral bias in tooth patterning (Kassai et al.,2005).

In the toothless mouse diastema, both the epithelium(Cobourne et al., 2004) and the mesenchyme (Yamamoto et al.,2005; Yuan et al.,2008) have been proposed to inhibit diastema tooth development. Recently Yuan et al. (Yuan et al.,2008) have shown that, whereas post E12 diastema mesenchyme retains the capacity to form teeth, it cannot initiate tooth formation by inducing epithelium. Because Sostdc1 expression domains partially cover the diastemal regions, we propose that Sostdc1 may be one of the genes suppressing diastemal tooth formation.

In our in vitro cultures of wild-type teeth, the experimental phenocopying of the Sostdc1-deficient extra incisors was possible until the incisors reached E14 cap-stage. By this developmental stage, Sostdc1expression is upregulated also in the epithelium(Fig. 1). This alone suggests that trimming the E14 dental mesenchyme in vitro may not reduce Sostdc1 expression enough to allow the development of extra incisors. Another possibility for the decreased tooth making potential of E14 explants may be apoptotic removal of the epithelial area that has potential to form extra incisors. Increased apoptosis has been reported in multiple rudimentary teeth (Tureckova et al., 1996; Peterkova et al., 2002; Keränen et al., 1999) and we detected a distinct population of apoptotic cells in the dental epithelium of the wild-type incisors at the site where the extra incisors of Sostdc1 mutants develop. There was a dramatic lack of apoptotic bodies in the dental epithelium of Sostdc1-deficient teeth at E14,both in the enamel knot and especially in the region of the epithelium giving rise to the extra incisors (Fig. 2). Unlike Murashima-Suginami(Murashima-Suginami, 2007; Murashima-Suginami, 2008), we did not detect marked changes in mesenchymal apoptosis in the Sostdc1-deficient teeth. Taken together, because earlier removal of mesenchyme expressing Sostdc1 causes the development of the extra incisors in the wild-type explants and because this ability is lost with the upregulation of Sostdc1 expression and apoptosis in the epithelium,we suggest that apoptosis may be a relatively downstream event in the in vivo suppression of extra tooth formation.

Because the trimming of the mesenchyme in the wild-type explants consistently produced the extra incisor found in Sostdc1-deficent mice in vivo and in vitro, we tested whether the trimming of mesenchyme in the mutants had any effect. The results showed that, in addition to the one extra incisor, additional teeth were initiated later from the lingual cervical loop epithelium of the previously initiated incisors (Figs 5 and 6). This de novo incisor formation in the trimmed Sostdc1 null explants suggests that there are additional genes expressed in the mesenchyme that suppress extra teeth. The phenotype of these de novo incisors resembles the supernumerary teeth that form when epithelial β-catenin is constitutively stabilized in transgenic mice (Catenb^Δex3k14/+) (see Fig. S5 in the supplementary material). Normally, the stabilization of epithelialβ-catenin requires canonical Wnt signaling and the additional mesenchymal factors may thus involve modulators of Wnt signaling pathway. As Sostdc1 has been implicated in the inhibition of both BMP and Wnt signaling (Laurikkala et al.,2003; Itasaki et al.,2003; Yanagita et al.,2004), there may be partly redundant signaling pathways that restrict extra tooth formation.

We have previously linked Sostdc1 to BMP signaling in tooth development (Laurikkala et al.,2003; Kassai et al.,2005). BMP4 can induce the expression of Sostdc1, which in turn antagonizes the induction of enamel knot marker p21 by BMPs in tissue cultures (Laurikkala et al.,2003; Kassai et al.,2005). The role of BMP4 as mesenchymally expressed enamel knot inducer was strengthened in the absence of Sostdc1, as shown by our experiments where we placed BMP4-releasing beads on the lingual side of the in vitro mutant explants (Fig. 7). BMP4 accelerated extra incisor formation even further than did the trimming of the mesenchyme. However, we observed acceleration of extra incisor formation in only the Sostdc1-deficient explants, perhaps further supporting the role of Sostdc1 as a feedback inhibitor of BMP signaling(Kassai et al., 2005). Additionally, we have previously shown a relatively subtle acceleration of normal molar initiation by BMP4 (Kavanagh et al., 2007), and it is conceivable that the `rescue' of extra teeth may require additional factors.

In addition to BMP signaling, Sostdc1 has been linked to Wnt signaling (Itasaki et al.,2003). This link is indirectly supported by the resemblance of the de novo supernumerary tooth formation in Sostdc1-deficient mice with the phenotype of Catenb^Δex3k14/+ mice (see Fig. S5 in the supplementary material). In addition, Dkk1, a potent Wnt signal inhibitor inhibited extra incisor formation in vitro. The BMP antagonist Noggin had a similar inhibitory effect, as also shown recently by Murashima-Suginami et al.(Murashima-Suginami et al.,2008) (see Fig. S4 in the supplementary material). These inhibitory effects agree with previous works implicating both Wnt and BMP as activators of tooth initiation(Järvinen et al., 2006; Bei et al., 2000; Wise and Stock, 2006). It is noteworthy that the de novo incisors formed only in the absence of epithelial Sostdc1 expression. Whereas mesenchymal Sostdc1 expression may predominantly antagonize mesenchymally expressed Bmp4, the epithelially expressed Sostdc1 may also interfere with epithelial Wnt signaling crucial for tooth induction, a possibility supported by the increased Wnt activity in the Sostdc1-deficient incisors(Fig. 3). In developing hair placode epithelia, Bmp expression is upregulated as a result ofβ-catenin stabilization (Närhi et al., 2008), and it is conceivable that Wnt and BMP signaling may function both up- and downstream from each other at different stages of development.

Our results are indicative of a role for mesenchyme, through Sostdc1 and at least one additional factor, in inhibiting extra tooth formation. The exact developmental origin of the extra incisor(Fig. 3) is reminiscent of the location from which normal replacement teeth are initiated in many mammals(e.g. Luckett, 1985; Järvinen et al., 2009). Therefore, we propose that the most parsimonious interpretation of the extra incisor in Sostdc1-deficient mice is to consider it as a replacement tooth. This fits with the views that, evolutionarily, rodent incisors are interpreted to be second, and sometimes first, deciduous incisors(Luckett 1985; Meng et al., 2003; Asher et al., 2005). The permanent replacement incisor was lost early in the evolutionary history of rodents, together with the reduction of the number of incisors to the one pair found in all rodents. This scenario for the role of Sostdc1 may be further supported by the de novo supernumerary incisors formed in the Sostdc1 mutants with trimmed mesenchyme. These teeth resemble the continuous tooth budding found in the Catenb^Δex3k14/+ mice. Hence, a central role of Sostdc1 in normal tooth development may be to modulate BMP and Wnt signaling in limiting tooth replacement. In addition to tooth replacement, the Sostdc1-Wnt-BMP signaling may be part of the developmental program limiting the induction of teeth spatially. Interestingly, the region-specific absence of dentition in teleost fish seems to correlate with the lack of Bmpexpression (Wise and Stock,2006), and it remains to be tested whether Sostdc1-like inhibition may also be involved in spatial delineation of tooth-forming areas in other vertebrates. Considering the role of mesenchyme in tooth induction and the design of tissue engineering protocols, our work may have uncovered how delicate control of tissue quantities alone may influence the outcome between induction and inhibition.