Shh signaling within the dental epithelium is necessary for cell proliferation, growth and polarization

The first visible indication of the initiation of odontogenesis is the appearance of a local thickening of the oral ectoderm, which subsequently grows into the underlying neural crestderived mesenchyme of the first branchial arch to form epithelial buds. Ectomesenchymal cells condense around the epithelial buds to form the dental mesenchyme. The primary enamel knot, a tooth signaling center (Vaahtokari et al.,1996; Jernvall et al.,1998), becomes prominent at the cap stage. At the bell stage, the tooth consists of an epithelial enamel organ (EEO) and a dental mesenchyme. The EEO components include proliferating preameloblasts and their progenitors,the cells of the inner dental epithelium (IDE), the secondary enamel knots at the tip of nascent cusps, the stellate reticulum (SR), the outer dental epithelium (ODE) and the stratum intermedium (SI). The latter consists of squamous cells adjacent to the IDE and preameloblasts. Later-arising secondary enamel knots, which form in the developing molars, are thought to control cuspal morphogenesis and terminal differentiation of odontoblasts(Thesleff et al., 2001). Mesenchymal cells of the dental papilla that lie adjacent to the IDE form the preodontoblast layer of proliferating cells; the rest of the dental papilla cells contribute to the later development of the dental pulp. Finally, dental mesenchyme that surrounds the tooth germ forms the dental sac, which gives rise to periodontal tissues.

During the cytodifferentiation stage, the terminal differentiation of odontoblasts is accomplished by their withdrawal from the cell cycle,elongation, polarization and secretion of a predentin matrix. This, in turn,triggers terminal differentiation of ameloblasts(Slavkin and Bringas, 1976;Frank and Nalbandian, 1989). Differentiation of the ameloblast into a highly-polarized complex secretory cell involves considerable growth, elongation of the cytoplasm, a change in nuclear polarity, a sequential development and change in polarity of organelles, and the appearance of a complex cytoskeleton(Slavkin, 1974). During this stage, there are other progressive changes within the enamel organ. Cells from the SI that are adjacent to polarizing post-mitotic ameloblasts become cuboidal in shape, except in the future enamel-free areas in rodent molars(Cohn, 1957;Gaunt, 1956;Hay, 1961). In addition, the SR is invaded by blood vessels and fibroblasts emanating from the dental sac(Lefkowitz et al., 1953;Hay, 1961).

The rodent incisor is unique in its tissue organization and consists of stem cells, differentiating cells and mature cells organized in defined regions along its anteroposterior and labial-lingual axes. The rodent incisor is asymmetrical, as the labial or amelogenic IDE gives rise to ameloblasts and enamel, whereas the IDE on the lingual side does not produce enamel. The posterior-most aspect of the incisor has been postulated to contain stem cells which give rise to the different dental cell populations(Smith and Warshawsky, 1975;Harada et al., 1999). A posteroanterior gradient of cytodifferentiation is thus present in the rodent incisor throughout life, with the less differentiated cells located posteriorly and the most mature cells anteriorly. Odontoblasts differentiate all along the epitheliomesenchymal interface of the incisor(Cohn, 1957; Warshavsky,1968).

Like many organs, morphogenesis and cytodifferentiation of the tooth is governed by sequential and reciprocal epithelialmesenchymal interactions mediated by several soluble bioactive proteins(Jernvall and Thesleff, 2000;Thesleff et al., 2001;Thesleff and Mikkola, 2002). In addition, cell-matrix interactions and cell-cell junctional complexes and cytoskeletal components have been implicated in the regulation of histomorphogenesis and proliferation(Thesleff et al., 1981;Lesot et al., 1982;Fausser et al., 1998).

Sonic hedgehog (Shh), a member of the vertebrate Hedgehog(Hh) family, encodes a secreted signaling peptide. Hedgehog signals are received within a target tissue by the general Hedgehog receptor Patched 1(Ptc 1 in mammals; Ptch — Mouse Genome Informatics) [for review of pathway see Ingham and McMahon (Ingham and McMahon, 2001)]. Transduction of the signal within a responding cell absolutely requires the activity of a second, multi-pass,membrane protein Smoothened (Smo). Smoothened activity leads to a conserved transcriptional response: up-regulation of Ptc1 and Gli1 in the target tissue. Gli1 encodes a member of the Ci/Gli family of transcriptional effectors of Hedgehog signaling.

The Hh signaling pathway is an evolutionarily well-conserved mechanism involved in a plethora of biological processes (for reviews, seeIngham and McMahon, 2001;McMahon et al., 2002).Shh is expressed exclusively in the epithelial component of the murine tooth from the dental lamina stage until cytodifferentiation(Bitgood and McMahon, 1995;Vaahtokari et al., 1996;Hardcastle et al., 1998;Dassule et al., 2000;Gritli-Linde et al., 2001). At the cap stage, Shh is confined to the primary enamel knot(Vaahtokari et al., 1996;Hardcastle et al., 1998). Expression spreads thereafter to the rest of the IDE laterally, the stratum intermedium (Dassule et al.,2000; Gritli-Linde et al.,2001) and the stellate reticulum(Gritli-Linde et al., 2001). At the cap stage, general transcriptional targets and effectors of Shh signaling, including Ptc1, Gli1 and Smoothened(Smo) are, however, expressed in both dental epithelium and mesenchyme, but are excluded from the enamel knot(Hardcastle et al., 1998). In contrast, Ptc2, while appearing to bind all mammalian Hedgehog proteins similarly to the related Hedgehog receptor Ptc1(Carpenter et al., 1998), is expressed in the enamel knot and IDE at the cap and early bell stages,respectively (Motoyama et al.,1998).

We have shown previously that Shh protein produced by the enamel knot and IDE moves many cell diameters to reach the rest of the dental epithelium and the dental papilla, indicating that Shh has a long-range activity, consistent with the broad expression of Shh target genes(Gritli-Linde et al., 2001). Together, the above observations suggest that Shh signaling may be operative intra-epithelially as well as in mediating epithelial-mesenchymal interactions during tooth development. Finally, we have shown that genetic removal of Shh activity from the tooth leads to alterations in growth and morphogenesis and results in tissue disorganization, affecting both the dental epithelium and mesenchyme derivatives (Dassule et al.,2000). In this study, it was not possible to determine clearly whether the alterations in the dental mesenchyme and its derivatives were solely generated by lack of Shh signaling by the dental epithelium, or whether they were secondary to the lack of proper signaling — via other bioactive molecules — in the abnormal dental epithelium. Conversely,this approach left unanswered the question of whether abnormal development of the epithelial enamel organ in Shh mutant teeth was a result of a loss of intra-epithelial Shh signaling, or a secondary consequence of altered signaling by the underlying dysplastic dental mesenchyme. In order to distinguish between these alternatives, and further define the roles of Shh in regulating morphogenesis of the tooth, we abrogated Shh signal transduction by genetically removing the activity of Smo from the dental epithelium and its derivatives while maintaining Shh responsiveness in the dental mesenchyme.

The Smo conditional allele (Smo^c) and theSmo null allele (Smoⁿ) were generated as described (Zhang et al., 2001;Long et al., 2001). To remove Smo activity from the dental epithelium, we used mice expressing Cre recombinase under the regulation of the enhancer element of cytokeratin 14 (K14) (Dassule et al.,2000). Mice expressing the highest levels of Cre RNA in the skin were crossed to a Smo null allele(Zhang et al., 2001) to generate mice that were heterozygous for both the K14-Cre transgene and a Smoⁿ allele (K14-Cre;Smo^+/n). Mutant embryos were produced by crossing theK14-Cre; Smo^+/n mice toSmo^c/c mice. Embryos and new-borns were genotyped by PCR as described previously (Zhang et al.,2001; Long et al.,2001). The generation of mice lacking Shh activity in the dental epithelium was as described previously(Dassule et al., 2000).

The tips of the mandibles containing the lower incisors from control and mutant pups at 1 day post-partum (1 dpp) were processed for transmission electron microscopy (TEM) as described previously(Sun et al., 1998). Sections for histology were prepared according to routine procedures and stained with Hematoxylin and Eosin.

Sections from embryos and 1 dpp pups were prepared for in situ hybridization with ³⁵S-UTP-labeled riboprobes essentially as described previously (Wilkinson et al.,1987). The following probes were used: Bmp2, Bmp4 andBmp7 (Åberg et al.,1997); PDGFRα(Boström et al., 1996);amelin (Cerny et al.,1996; Fong et al.,1998); Shh, Ptc2, Gli1, Gli2, Gli3, Dentin Sialoprotein (DSP),Dentin Matrix Protein 1 (DMP1) and Msx2(Dassule et al., 2000);cyclin D1 (Long et al.,2001); Dlx3 and Dlx7(Zhao et al., 2000);Ptc1 (Gritli-Linde et al.,2001). A 1 kb Bmp5 pro-region fragment was also used to generate an antisense RNA probe.

Tissues were fixed in either 4% paraformaldehyde in 0.1 M phosphate buffer,pH 7.4, or in Sainte Marie's solution, embedded in paraffin and processed for immunohistochemistry as described previously(Gritli-Linde et al., 2001). The following antibodies were used: rabbit anti-amelin(Fong et al., 1998); rabbit anti-phospho-histone H3 (Ser 10) (Cell Signaling Technology); Ab80 anti-SHH(Bumcrot et al., 1995);Marti et al., 1995); rabbit anti-calbindin-D28K (Chemicon); rat anti-E-cadherin and rabbit anti-ZO1 (Zymed Laboratories); rabbit anti-collagen type IV (ICN Pharmaceuticals); mouse monoclonal anti-β-tubulin clone 5H1, IgM fraction (BD Pharmingen).

For staining of mineralized extracellular matrix, we used a modified method based on the von Kossa technique (Stevens et al., 1990).

We have previously described a K14-Cre transgene that is active specifically within the dental epithelium and its derivatives from shortly after the initiation of tooth development(Dassule et al., 2000) and aSmo conditional allele (Smo^c) which allows the Cre-dependent removal of Shh responsiveness in a target cell(Long et al., 2001). In the developing tooth, K14 is expressed in all components of the epithelial enamel organ (EEO) (Tabata et al., 1996; Dassule et al.,2000). We have previously shown(Dassule et al., 2000)efficient Cre activity in the skin and oral ectoderm, including the dental lamina at 11.75 days post-coitum (dpc), and throughout the dental epithelium by 14.5 dpc.

To create a dental epithelium devoid of Smo activity, we crossedK14-Cre; Smo^+/n males to Smo^c/cfemales. As expected, approximatively 25% of newborns were the experimental genotype (K14-Cre; Smo^c/Smoⁿ), hereafter referred to as `Smo mutants' for simplicity. All mutant pups were of a normal size and displayed normal external features but died within one day of birth for unknown reasons. Throughout, we compare development of teeth fromK14-Cre;Smo<sup>c</sup>/Smo<sup>n</sup> to those of Smoheterozygotes (Smo<sup>c</sup>/Smo<sup>n</sup>), hereafter referred to as `controls'. Smo^c/Smoⁿ embryos and pups were phenotypically similar to wild-type animals.

In order to determine the outcome of Smo ablation from the dental epithelium, we analysed histological sections from maxillae and mandibles at 1 day post partum (dpp). As in wild-type pups, incisors and molars developed in both jaws of Smo mutants. Ablation of Shh activity within the tooth generates abnormally small and misshapen teeth(Dassule et al., 2000). In contrast, general growth of molars in Smo mutants was essentially similar to that of control littermates(Fig. 1A-D). Primary and secondary enamel knots developed normally in Smo mutant molars and expressed appropriate specific molecular markers, such as Bmp2, Bmp4, Shh,Ptc2 and Msx2 (see below and data not shown). Despite their normal growth, Smo mutant molars exhibited several morphological abberrations. Parasagittal sections at 1 dpp revealed that the first and second molars, in both the maxilla and mandible, were abnormally fused,forming a single gigantic anlage (Fig. 1A,B). The molars also fuse in Shh mutants; however, the single angle in those mice were much smaller than those of Smo mutant mice (data not shown). On parasagittal and frontal sections, the `first molars' of Smo mutant pups displayed shallow, broad, underdeveloped and misshapen cusps as compared to controls(Fig. 1A-D). As in Shhmutants (Dassule et al.,2000), Smo mutant molars developed close to the oral surface, reflecting the virtual absence of a dental cord(Fig. 1C,D). However, in contrast to Shh mutants (Dassule et al., 2000), at all developmental stages the mesenchyme of theSmo mutant molars appeared to have normal cellularity, temporospatial patterns of odontoblast terminal differentiation and secretion of a predentin matrix (Fig. 1C,D).

The principal cytological differences were observed in the dental epithelium derivatives. A disorganization within the EEO was evident at the late bell stage in the principal cusps of the `first molars'. In the less-developed cusps of the `first and second molars', the EEO cytological organization was normal, similar to that of controls(Fig. 1A,B). At the tip of the principal cusps of late bell stage first molars from control 1 dpp pups,elongated polarizing, post-mitotic ameloblasts lay adjacent to a predentin matrix. The ameloblast were overlaid by the SI, which had assumed a cuboidal shape. The ODE developed into a discontinuous layer, and the coronal aspect of the SR displayed the normal metaplastic changes secondary to the initiation of its invasion by early vascular loops and fibroblasts that emanate from the dental sac, through gaps in the ODE (Fig. 1C). In the most advanced cusps of the first molar region ofSmo mutant, ameloblasts were abnormally short and were overlaid by a scarce, squamous SI. The SR was also hypocellular and did not display the changes characteristic of the process of metaplasia. The absence of early vascular loops in the coronal aspect of the SR was verified by staining of blood vessel basement membrane with an anti-collagen type IV antibody (data not shown). Finally, the mutant ODE formed a continuous layer without the gaps observed in controls (Fig. 1D).

Similar cytological changes were observed, specifically on the labial side of the epithelium of Smo mutant incisors. They were, however, more dramatic, as the incisor is developmentally more advanced than the molar. At 1 dpp incisors from control pups contained, on their labial side, enamel matrix secreted by highly polarized secretory ameloblasts overlaid by cuboidal SI cells (Fig. 1E). At this stage,the ameloblasts of mutant incisors had failed to undergo polarized growth and formed a non-cohesive layer of non-polarized cuboidal cells with centrally located nuclei; the enamel matrix was absent(Fig. 1F). The SI overlying ameloblasts was sparse and exhibited a squamous morphology(Fig. 1F). In contrast to molars, the mutant incisors had a smaller diameter than the controls, which was first evident at the differentiation stage. This may reflect the differences between development of molars and incisors in rodents. In addition, the Smo mutant incisors exhibited abnormal foldings of the lateral and medial aspects of the IDE at the posterior and middle segments of the tooth, which at certain levels invaginated within the tooth (data not shown). These may also account for the slightly smaller size of the incisor(data not shown). In control incisors, during enamel secretion, a papillary layer consisting of merged SR and ODE was evident, but this structure was absent in the mutant incisors (data not shown).

In order to characterize the structural cytological alterations ofSmo mutant ameloblasts and SI at the subcellular level, we examined and compared their phenotype to that of controls by transmission electron microscopy (TEM) on ultrathin sections taken from incisors at their anterior segment. In control incisors, young secretory ameloblasts were highly columnar, elongated and polarized with oval-shaped nuclei elongated along the apical-basal axis (Fig. 1G). InSmo mutant incisors, the cuboidal ameloblasts were only 15% of the apical-basal height of the secretory ameloblasts of control incisors(excluding the Tomes' process) and contained centrally located round nuclei(Fig. 1H). Furthermore, several organelles, including mitochondria, RER and Golgi were both sparse and evenly distributed in the cytoplasm, whereas in controls these organelles were abundant and showed a polarized distribution within the cell. Tomes' processes and the terminal webs had not developed in mutant ameloblasts(Fig. 1H).

Junctional complexes and the cytoskeleton are important in maintaining epithelial cell polarity and cell-cell interactions (Farquar and Palade, 1963;Gundersen and Cook, 1999). To assess whether defects in the EEO in Smo mutant teeth were secondary to alterations in the development of these structures, we analyzed by immunohistochemistry the distribution of some markers of these cellular components.

First, we studied the distribution of Zonula Occludens-1 (ZO-1), a member of the membrane-associated guanylate kinase homologues (MAGUKs) and a scaffolding protein associated with tight junctions and other cell-cell contact sites (Tsukita et al.,1999; Vasioukhin and Fuchs,2001). In control embryos at the early bell stage, ZO-1 staining was abundant in the ODE and less so in the IDE (data not shown). By the late bell stage, in control molars, ZO-1 protein gradually increased at the apicolateral and basolateral domains of polarizing ameloblasts adjacent to the first layer of predentin. At a similar stage, the ameloblasts of mutants showed barely detectable levels of ZO-1(Fig. 2A-D). The ODE of control molars exhibited reduced ZO-1 immunostaining and formed a discontinuous layer interrupted by gaps penetrated by connective tissue invading the coronal part of the SR (Fig. 2C). In contrast, the ODE of Smo mutants remained as a continuous layer of cells exhibiting strong ZO-1 staining (Fig. 2D). At more advanced developmental stages, as demonstrated in incisors, Smo mutant ameloblasts totally lacked ZO-1 staining,whereas the ODE and SI displayed a strong cytoplasmic accumulation of the protein, in striking contrast to control incisors(Fig. 2E-H).

E-cadherin, the epithelial prototype of the transmembrane core of adherens junctions, which is associated with homotypic cell-cell adhesion, also showed a dynamic distribution in polarizing ameloblasts and ODE of control molars that was similar to that of ZO-1 (Fig. 2I,K). In Smo mutant molars, some ameloblasts facing predentin had an apicolateral accumulation of E-cadherin similar to that of control ameloblasts (Fig. 2J,L), but in the less-developed cusps, immunostaining was much weaker in the Smo mutant preameloblasts than controls(Fig. 2I,J). Similarly,E-cadherin accumulated at the apicolateral and basolateral membrane domains of polarizing ameloblasts but was weak and disorganized in mutants(Fig. 2M,N). At a more anterior level, E-cadherin staining became strong in cells of the papillary layer(Fig. 2O). In Smomutant incisors, the shrunken ameloblasts showed a weak cytoplasmic staining and the papillary layer was unrecognizable(Fig. 2P).

The polarization of both the ameloblast and odontoblast populations is characterized by the accumulation of microtubules. In control teeth, whereasβ-tubulin was present at high levels in functional odontoblasts, in polarizing ameloblasts (Fig. 2Q) and in secretory ameloblasts (data not shown), no such accumulation of β-tubulin protein was detected in Smo mutants(Fig. 2R). In contrast,odontoblasts in the mutants displayed normal levels of β-tubulin staining.

Thus, Shh signaling is also necessary for normal development of both the cytoskeletal organization and cell-cell junctional complexes, which are likely to play a role in establishing and maintaining ameloblast polarity and function.

Interestingly, analysis of the expression of Ptc2 andGli1 transcripts indicated that their subcellular distribution was polarized in the developing ameloblasts. Ptc2 and Gli1transcripts were enriched at the basal and perinuclear compartments of polarizing (see Fig. 5K below),presecretory and secretory ameloblasts closest to the Shh-expressing SI (Fig. 3A-C and see below). In contrast, transcripts of Dlx7, Bmp7, Bmp5 (data not shown),Gli2 (see below), Msx2, Ptc1 and Dlx3(Fig. 3D-F) were either distributed uniformly throughout the ameloblast cytoplasm or enriched at its apical pole. Thus a polarized response to the SI, mediated by Shh signaling could play a role in the Shh-dependent ameloblast polarization that is deficient in Smo mutants.

In order to gain insight into the molecular basis of the cytological alterations within the epithelial enamel organ following removal of Smo activity, we first analysed in detail the expression of Shh and its targets and effectors at different developmental stages of tooth development.Shh expression and responsiveness in control and mutant molars at the cap stage are represented in the schematics(Fig. 4A,B). At the cap stage,Shh was detected in the primary enamel knot in both controls andSmo mutants (Fig. 4C,D). Ptc1 and Gli1, primary targets of Shh signaling, were expressed in both the EEO (excluding the enamel knot) and mesenchyme in control teeth (Fig. 4E,G). In contrast, in Smo mutants, the EEO exhibited a dramatic loss of Ptc1 and Gli1 expressions, whereas the mesenchyme maintained normal expression of both of these genes(Fig. 4F,H). Expression ofGli2 and Gli3 (data not shown), Ptc2 in the enamel knot and Hip1 (Fig. 4I-L) was similar to controls. These data indicate that, as expected, Shh responsiveness was absent from the EEO but maintained in the dental mesenchyme at the cap stage.

In late bell stage control molars, Shh was strongly expressed in the IDE, SR, SI, preameloblasts and differentiating ameloblasts(Fig. 5A). However, in mutant molars, preameloblasts adjacent to either polarized odontoblasts or to predentin, SI and SR showed barely detectable levels of Shh (arrowFig. 5B). At this developmental stage, Ptc1 and Gli1 were expressed in both the EEO and mesenchyme, whereas Ptc2 expression was confined to the EEO in control molars (Fig. 5C,E,I). In Smo mutants, Ptc1, Gli1 and Ptc2 expressions were severely decreased in the EEO, whereas the mesenchyme showed normal Shh responsiveness (Fig. 5D,F,J). In Smo mutant molars, the expression of Hip1 in the mesenchyme was similar to that of controls(Fig. 5G,H).

Immunohistochemistry for Shh protein showed staining in the EEO and mesenchyme as well as in the basement membrane and predentin matrix in molars from control pups (Fig. 5K). The mutant molars (Fig. 5L)exhibited normal Shh immunostaining in the dental mesenchyme and SR; however,immunostaining was dramatically reduced in preameloblasts adjacent either to polarized odontoblasts or to a thin layer of predentin, as well as in the SI overlying these cells.

As in molars, alterations of Shh expression and responsiveness were also found in the mutant incisors. In the middle segment of control incisors at 1 dpp, Shh was strongly expressed in the IDE, SI,preameloblasts and differentiating ameloblasts on the labial side of the tooth(Fig. 5M). In contrast, inSmo mutants, Shh expression was severely decreased in preameloblasts adjacent either to polarized odontoblasts or to predentin matrix (Fig. 5N). The SI continued to express Shh for only a short time before expression was decreased in this cell layer too (Fig. 5N). At this stage, Ptc1 (data not shown) andGli1 (Fig. 5O) were expressed in both the mesenchyme and labial epithelium of control incisors. In contrast, expression of both genes (Fig. 5P and data not shown) was severely decreased in the epithelium but not in the mesenchyme of Smo mutant incisors. In the anterior segment of control incisors, Shh expression declined in polarizing ameloblasts (arrowhead in Fig. 5Q), dramatically decreased in presecretory ameloblasts (arrow inFig. 5Q), and was totally abrogated in secretory ameloblasts (data not shown). The SI continued to express Shh (Fig. 5Q),even at a more advanced secretory stage (data not shown). In mutant incisors,Shh expression was dramatically decreased(Fig. 5R). In control incisors,polarizing, presecretory and young secretory ameloblasts and the SI, expressedGli3 (data not shown), Ptc1, Ptc2, Gli1 and Gli2;however, with the exception of Gli3 (data not shown), expression of all these genes was decreased to background levels in these cell populations in Smo mutants (Fig. 5S-Z).

These data demonstrate that removal of Smo activity from the dental epithelium affects Shh responsiveness within all components of the EEO without altering Shh signaling in the dental mesenchyme. Furthermore, the results provide evidence that in the developing tooth, with the exception of the enamel knot, Ptc2 transcription is Shh dependent, suggesting that Ptc2, like Ptc1, is a target of Shh signaling, and implicating Ptc2 in the regulation of Shh signaling activity within the EEO.

The Smo mutant molars were small, misshapen and had shallow cusps and the incisors were also abnormally small. The enamel knots have been suggested to control cusp morphogenesis in molars. However, primary and secondary enamel knot development in Smo mutant molars appeared normal at both the morphological and molecular level. Cell proliferation profiles have been suggested to govern tooth morphogenesis and cytodifferentiation (Ruch,1990). To address whether cell proliferation, altered bySmo loss-of-function, might contribute to the morphological alterations we observed, we first examined the distribution of phosphorylated histone H3 (PH-H3), a marker of mitosis. In control molars and incisors,preameloblasts and the overlying SI facing polarized odontoblasts showed PH-H3 staining. PH-H3 staining was lost thereafter in ameloblasts adjacent to predentin matrix (Fig. 6A and data not shown). In contrast, in Smo mutant molars and incisors,PH-H3 staining was lost at an earlier stage, in preameloblasts and SI facing polarized odontoblasts, before predentin secretion(Fig. 6B and data not shown).

Cyclin D1 is a G1 cyclin, which has recently been shown to be transcriptionally induced by Shh and Indian hedgehog(Kenney and Rowitch, 2000;Long et al., 2001). We found that expression of cyclin D1 was similar in control and Smomutant molars from the cap stage until the early bell stage (data not shown). In the EEO of control molars at 1 dpp, cyclin D1 was expressed in the IDE, SI and preameloblasts adjacent to polarized odontoblasts. Expression was thereafter abrogated in post-mitotic ameloblasts adjacent to predentin matrix,whereas the SI continued to express cyclin D1(Fig. 6C). In Smomutant molars, however, cyclin D1 transcripts were lost from preameloblasts and the adjacent SI in young cusps that contained polarized odontoblasts but before predentin secretion by these cells(Fig. 6D). As expected, no alterations in PH-H3 staining and cyclin D1 expression were found in the dental mesenchyme (Fig. 6A-D). Thus, it is likely that mutant preameloblasts and cells of the SI exit the cell cycle prematurely, and that Shh may promote proliferation within the dental epithelium at least in part through the transcriptional regulation of cyclin D1.

To assess whether the premature withdrawal from the cell cycle of mutant ameloblasts was associated with their differentiation, we examined the expression of amelin and its protein product as well as calbindin D28K immunostaining. These markers are expressed in differentiating and in mature ameloblasts (Elms and Taylor,1987; Fong et al.,1998). In control molars and incisors, amelin mRNA and protein were present in the dental papilla, preodontoblasts and polarizing odontoblasts (Fig. 6E,I and data not shown). Thereafter, amelin mRNA and protein appeared in differentiating ameloblasts and was strongly increased and maintained in ameloblasts (Fig. 6K and data not shown). In contrast, in mutant molars and incisors, amelintranscripts and protein were already detected in preameloblasts that were adjacent to either preodontoblasts or polarizing odontoblasts, prior to predentin secretion (Fig. 6F,Jand data not shown), and remained strong at later stages(Fig. 6L and data not shown). Immunohistochemistry with anti-calbindin D28K antibody showed staining in preameloblasts that are either adjacent to polarized odontoblasts or a thin layer of predentin (Fig. 6G),whereas in Smo mutants molars calbindin staining was present at a developmentally earlier stage, in preameloblasts adjacent either to preodontoblasts or polarized odontoblasts(Fig. 6H). These data suggest that mutant ameloblasts `mature' precociously and express genetic markers of post-mitotic ameloblasts, without undergoing the cytological changes characteristic of the ameloblast proper.

As Shh signaling appears to be intact in the dental mesenchyme and its derivatives, we expected that these would not be affected in Smomutants. However, the altered differentiation and responsiveness of ameloblasts in Smo mutants might have indirectly affected mesenchymal components. To address this issue, we examined the expression of a number of genetic markers, that are known to be initiated or upregulated upon ameloblast or odontoblast differentiation, including PDGFRα, DSP,DMP1, Dlx3, Dlx7, Msx2, Bmp2, Bmp4, Bmp5, and Bmp7(Fig. 7 and data not shown).

Odontoblasts differentiated normally on schedule and secreted normal predentin/dentin matrix in Smo mutant teeth. This is supported by the presence of normally polarized odontoblasts and normal mineralization of dentin matrix, as shown by von Kossa staining(Fig. 7A,B) and odontoblast expression of PDGFRα, DMP1, Bmp2, Dlx7 and DSP(Fig. 7C-P and data not shown). In contrast, cells of the ameloblastic lineage showed temporospatial alterations of the expressions of Dlx7, DSP, Bmp4, Bmp5 andBmp7. In Smo mutant incisors and molars, expression ofDlx7 in differentiating ameloblasts was considerably less than in controls (Fig. 7G-J). These data suggest a possible regulation of Dlx7 by Shh during ameloblast differentiation. It is noteworthy that in the dental epithelium of incisors,Dlx7 expression is confined to the amelogenic labial epithelium(Fig. 7G). As shown in control incisors and molars, in addition to its expression in odontoblasts,DSP is known to be transiently expressed in preameloblasts adjacent to the first layer of predentin (Fig. 7K,M) and then lost in secretory ameloblasts(Fig. 7O). However, inSmo mutant incisors and molars(Fig. 7L,N) DSPtranscripts were lost precociously from ameloblasts(Fig. 7L,N). This was not secondary to cell death in mutant ameloblasts, as these cells continued to express other late genetic markers, such as Msx2 and amelin(see Fig. 6P and data not shown) at very high levels in the anterior portion of the incisor. These data provide further evidence for the premature differentiation of Smomutant ameloblasts. Finally, while in control incisors, Bmp4, Bmp5and Bmp7 (Fig. 7Q and data not shown) were expressed in secretory ameloblasts, expressions of these genes were severely down-regulated in mutant ameloblasts(Fig. 7R and data not shown). However, at an earlier stage in both incisors and molars from Smomutants, preameloblasts and ameloblasts expressed Bmp4 andBmp5 as in controls (Fig. 7S,T and data not shown).

We have shown previously (Dassule et al., 2000) that conditional removal of Shh activity from the dental epithelium, the sole tooth compartment responsible for Shh production,alters growth, morphogenesis and cytological organization in both the dental epithelium and mesenchyme. By the conditional ablation of Smo within the dental epithelium, we now provide evidence that Shh signaling within the EEO itself is crucial for the normal control of cell proliferation, cell differentiation, cell growth and cell polarity, and consequently for appropriate morphogenesis of the tooth (summarized schematically inFig. 8).

Loss of Smo activity in the dental epithelium led to the expected abrogation of Shh responsiveness within the EEO, whereas Shh signaling was maintained in the dental mesenchyme, which developed normally at all stages up to 1 dpp. Smo mutant molars were fused and displayed alterations in cusp morphogenesis. In Smo mutant teeth, preameloblasts and cells of the SI withdrew from the cell cycle prior to predentin secretion, as evidenced by loss of cyclin D1 mRNA and PH-H3 protein, and exhibited molecular features of more mature cells. These data suggest a direct requirement for Shh activity in regulating proliferation within the dental epithelium. Rather than exiting the cell-cycle coincident with secretion of the first layer of predentin matrix (Ruch, 1987;Ruch, 1990), in the absence of Shh signal transduction, preameloblast and cells of the SI withdraw before matrix accumulation. Thus, Shh may exert a direct mitogenic effect on these cells, at least in part, by promoting cyclin D1 transcription to control the G₁/S transition. After exiting the cell cycle,preameloblasts underwent premature differentiation. Thus, maintaining these cells in a proliferative state may be one mechanism by which the pace and position of ameloblast differentiation is controlled in the IEE. Alternatively, Shh may regulate cell proliferation and cell differentiation independently. In this regard it is noteworthy that a second member of this family, Ihh, coordinates proliferation and differentiation of chondrocytes by distinct mechanisms during morphogenesis of the endochondral skeleton(reviewed by McMahon et al.,2002).

A key role of hedgehog members as mitogens has been shown in several developmental settings (for a review, seeMcMahon et al., 2002). A direct hedgehog input controlling cell proliferation has been best characterized in the cerebellum (Dahmane and Ruiz i Altaba, 1999;Wallace, 1999;Wechsler-Reya and Scott, 1999;Kenney and Rowitch, 2000) and cartilage (Long et al., 2001),as well as in the optic lamina and eye in Drosophila(Huang and Kunes, 1996;Huang and Kunes, 1998;Duman-Scheel et al., 2002). Recent evidence indicates that hedgehog proteins are likely to exert their mitogenic activity through regulation of G₁ and G₁/S cyclins. In the cerebellum, Shh stimulates granule cell progenitor (GCP)proliferation by promoting D-type cyclins(Kenney and Rowitch, 2000). In the developing cartilage, Ihh's mitogenic activity also promotes cyclin D1 transcription (Long et al.,2001). Finally, recent studies in Drosophila established a link between hedgehog and D-type and E-type cyclins in controlling cell proliferation and growth (Duman-Scheel et al., 2002). Together, these data indicate that transcriptional induction of D-type cyclins is a likely common and conserved mechanism by which hedgehog proteins promote cell proliferation. In addition, studies in transfected cells have shown that Shh ligand disrupts an interaction between Ptc1 and cyclin B1 at the G₂/M transition, thereby allowing the activation of M-phase-promoting factor and cell cycle progression(Barnes et al., 2001). It would be of interest to determine whether these events also occur in normal cells that are responsive to Shh.

During the late bell stage, both the proliferating preameloblasts and the cells of the SI strongly express Shh, and both respond to Shh signaling as judged by expression of key targets such as Ptc1 andGli1. Whether Shh operates in an autocrine or paracrine manner within and between these cell layers, respectively, is unclear at present. One possible way to examine this issue would be by removal of Shh activity specifically from the SI.

What makes preameloblasts exit the cell cycle? Our analyses show that both the proliferating preameloblasts and the newly differentiated ameloblasts both express high levels of Shh and are Shh responsive. This raises the question of what normally triggers exit from the cell cycle in ameloblasts. The same situation is found in the developing cerebellum, where GCPs migrate towards Purkinje neurons, the source of Shh, and thus become exposed to higher concentrations of the protein. However, GCPs stop dividing by the time they reach the internal part of the external germinal layer, although they remain highly responsive to Shh. It is unlikely that termination of preameloblast proliferation is due to reduced exposure or responsiveness to Shh. This may be the result of conversion of a proliferative response into a differentiative one or to the activation of antagonizing signals.

Smo mutant first and second molars were fused and developed into a single anlage. It is noteworthy that molar fusions can occur in both animals and humans (Sofaer and Shaw,1971; Turell and Zmener,1999), but their etiology is unknown. In addition, Smomutant molars had abnormally shaped, small and shallow cusps. Signaling from the enamel knots has been shown to pattern cuspidogenesis in molars(Thesleff et al., 2001). InSmo mutants molars, the primary and secondary enamels knots developed normally and expressed several genetic markers of these structures. We therefore propose that cusp dysmorphogenesis in Smo mutant molars may be secondary to the alterations in cell proliferation profiles within the EEO.

Smo mutant ameloblasts failed to assume the morphological features of differentiating cells, and the SI remained squamous in nature. At this developmental stage in Smo mutant teeth, odontoblasts developed and differentiated normally, and predentin/dentin matrices were secreted on schedule. These data indicate that Shh activity within the dental epithelium is necessary for proper cytodifferentiation of preameloblasts and cells of the SI.

In the mouse incisor, Shh and Dlx7 were expressed exclusively in the labial amelogenic epithelium, and Dlx7 expression was down-regulated in Smo mutant teeth. Transcripts for Ptc2and Gli1 showed a polarized localization at the basal and perinuclear pole of polarizing ameloblasts, which is adjacent to the SI.

The events leading to ameloblast cytodifferentiation are not clear. Early tissue recombination studies using dental and nondental tissues have shown that ameloblast cytodifferentiation requires functional odontoblasts(Kollar and Baird, 1970;Ruch et al., 1973), and that acellular dental matrices can promote ameloblast cytodifferentiation as well(Karcher-Djuricic et al.,1985). In those studies (and J. V. Ruch, personal communication),the development of an EEO (and not only a monolayer of IDE) from non-dental and dental epithelia was shown to be a prerequisite for induction of ameloblast cytodifferentiation by the mesenchymal components. Thus, there is additional need for an epithelial endogenous control, possibly involving interactions between the SI and preameloblasts. However, the different developmental fates of the labial amelogenic and lingual non-amelogenic IDE of the incisor have been shown to be governed by an endogenous epithelial regulation, independent of odontoblasts/predentin/dentin(Amar et al., 1986;Amar et al., 1989;Ruch, 1990). The role of the SI is unknown. This cell layer has been suggested to play a role in ameloblast differentiation based on morphological differences in the relationships between ameloblasts and the SI in the amelogenic zones as compared to the non-amelogenic zones of rodent teeth(Wakita and Hinrichsen, 1980;Nakamura et al., 1991). The SI has also been suggested to give rise to cells of the SR(Hunt and Paynter, 1963). The polarized Shh responsiveness in ameloblasts suggests that Shh emanating from the SI may play a role in ameloblast cytodifferentiation. However, Shh is expressed in both the SI and the differentiating ameloblasts.

Together, these data suggest that although the dental mesenchyme is necessary for ameloblast cytodifferentiation, it is not sufficient, and autocrine and/or paracrine epithelial-epithelial interactions within the EEO play an important role. Thus, our Smo conditional allele provides a molecular basis for understanding the early experimental studies, and suggests that Shh may be an endogenous epithelial factor regulating ameloblast cytodifferentiation. The asymmetrical expression of Shh andDlx7 in the incisor and the downregulation of Dlx7expression in mutant ameloblasts raise the question of whether Shh is, indeed,involved in the determination of the different developmental fates between the labial and lingual dental epithelia through the regulation of Dlx7expression.

The preameloblast is a unique epithelial cell, as upon differentiation into a secretory ameloblast it reverses its polarity: the pole of the cell that was originally basal (towards the basement membrane) becomes structurally and functionally apical (Frank and Nalbandian,1967). In addition, differentiation of the ameloblast is accompanied by a significant increase in size and by an extensive development of cytoplamic organelles (Frank and Nalbandian, 1967). At this time, cells of the SI also increase in size (Wakita and Hinrichsen,1980).

At the differentiation stage, despite the presence of predentin/dentin matrices, Smo mutant ameloblasts and cells of the SI failed to grow in size, remained unpolarized, and were characterized by a paucity of organelles. In spite of this, mutant ameloblasts expressed several genetic markers of differentiated cells. As in polarizing ameloblasts, Ptc2and Gli1 transcripts were enriched in the basal and perinuclear compartment of control presecretory and secretory ameloblasts. By this time,Shh production was barely detectable in these cells, whereas the overlying SI continued to produce Shh. These data suggest that ameloblast growth and polarization and the development of a cuboidal SI are Shh dependent. Hence,Shh may have anabolic activities regulating organelle and membrane biosynthesis. In addition, the phenotype of Smo mutant ameloblasts suggests that control of differentiation may be uncoupled from regulation of growth and polarization.

Smo mutant ameloblasts failed to show the typical polarized and localized distribution of E-cadherin and ZO-1 and lacked β-tubulin accumulation. Polarized epithelial cells interact with each other through specialized junctional complexes (Farquhar and Palade, 1963). Junctional complexes and microtubules are involved not only in maintaining epithelial cell polarity and partitioning the plasma membrane into apical and basolateral domains, but also in integrating mechanical and signaling pathways(Kirkpatrick and Peifer, 1995;Barth et al., 1997;Gundersen and Cook, 1999;Tsukita et al., 1999;Vasioukhin and Fuchs, 2001;Jamora and Fuchs, 2002). Previous TEM and immunological studies have shown that during the cytodifferentiation stage, numerous junctional complexes are established within the ameloblast and SI layers as well as between these two cell layers(Wakita and Hinrichsen, 1980;Fausser et al., 1998). Together, these data reveal roles for Shh signaling in controlling epithelial cell size and polarity.

The intracellular localization of Ptc2 and Gli1transcripts to the basal and perinuclear compartments of presecretory and secretory ameloblasts is an intriguing finding. This suggests that signaling from the SI to ameloblasts may be linked to the polarized distribution of these RNAs, potentially providing an efficient mechanism for the targeting of their protein products to the appropriate subcellular compartments. Epithelial polarity is established and maintained through specific subcellular trafficking of proteins to different subcellular compartments or membranes(Ikonen and Simons, 1998;Mostov et al., 2000). Protein trafficking and cell polarity involve not only the directed movement of specialized vesicles after protein synthesis(Ikonen and Simons, 1998), but also occur via the subcellular localization of transcripts (RNA sorting) prior to translation (Palacios and St. Johnston,2001). Why Ptc2 and Gli1 transcripts show this distribution in the ameloblast cytoplasm, but Ptc1 and Gli2transcripts do not, is not clear.

In conclusion, the expression patterns of Shh signaling components and the tooth phenotype of Smo mutant mice provide evidence for the presence of subtle and yet exquisite Shh signaling within the dental epithelium. Tooth development is orchestrated by Shh-dependent epithelial-mesenchymal and epithelial-epithelial interactions.

We are grateful to Drs C. Betsholtz, H. Edlund, A. George, D. Kingsley, U. Lendahl, M. MacDougall, E. Robertson, P. Sharpe, I. Thesleff, K. Weiss, J. Wozney and T. Wurtz for probes and antisera. We also thank Drs I. Thesleff and J. V. Ruch for advice and kindness. We wish to express our gratitude to T. Chadashvili and J. McMahon for mouse genotyping, H. Dassule for initially setting up mouse crosses, P. Cassidy for efficiency and kindness and to K. Hallberg for generously linearizing some of the plasmids. This work was supported by the Swedish Research Council (Grants 2789 and 14100) and the County Council of Västra Götaland to A. G. L. and A. L.; and by grants from the NIH to M. B. (K22DE14230), A. P. M. (NS 33642) and R. M. (DE 11697).