Epithelial-mesenchymal interactions are required for msx 1 and msx 2 gene expression in the developing murine molar tooth

The murine first molar tooth develops through a series of morphologically distinct stages as a result of epithelial-mesenchymal interactions during days E9 to E18 of gestation (for review see: Thesleff et al., 1989). These events are characterized by, and are probably regulated by, changes in the extracellular matrix (Thesleff et al., 1979, 1987), production of peptide growth factors and their respective receptors (Partanen and Thesleff, 1989; Cam et al., 1990; Vaahtokari et al., 1991), and by DNA-binding transcription factors (MacKenzie et al., 1991a,b, 1992; Karavanova et al., 1992).

The epithelium, derived from the oral ectoderm, forms the enamel organ of the first molar which secretes the enamel matrix. The remainder of the tooth and its supporting tissues are derived from the mesenchyme of the first branchial arch (Lumsden, 1988). All of the mesenchyme of the first arch consists of neural crest cells which migrate from the first and second rhombomere region of the primitive hindbrain (Hunt et al., 1991). By recombination experiments, Lumsden (1988) showed that for tooth formation to occur the mesenchyme must be neural crest in origin, but that it need not be from the cranial region. Furthermore, tooth formation can only be initiated by the oral epithelium, which at around E10 possesses pre-patterned regions, beneath which mesenchymal condensation occurs (Mina and Kollar, 1987; Lumsden, 1988). During the condensation process, the mesenchyme becomes capable of inducing tooth formation when cultured in combination with non-oral epithelium (Mina and Kollar, 1987); this property of the mesenchyme persists until at least E16 (Kollar and Baird, 1969).

The odontogenic interactions only occur at specific zones of the epithelial-mesenchymal interface, separated by regions of non-odontogenic epithelium (Mina and Kollar, 1987). Each of these zones codes for a single, morphologically distinct tooth. Mutations resulting in the transposition of teeth are rare (Miles and Grigson, 1990). However, the patterning system is not flawless as missing, supplemental (extra teeth with indistinct morphology) supernumerary (additional teeth of specific morphology) and morphologically disrupted teeth are found in humans (Duterloo, 1991). There is only one reported case where molar-like teeth are found anteriorly in the mouth (Kantaputra and Gorlin, 1992).

Homeobox genes constitute a large, highly conserved, multigene family of developmentally regulated transcription factors. Combinatorial expression of the Antennapedia-like homeobox genes of mammals within mesoderm adjacent to the neural tube is responsible for establishing axial regional identity of the prevertebral segments (Kessel and Gruss, 1991), and a similar system within the neuroepithelium appears to establish the identity of the caudal branchial arches (Hunt and Krumlauf, 1991). The Antennapedia class of homeobox genes are not expressed in rhombomeres 1 and 2, which give rise to the neural crest population of the first branchial arch (Hunt and Krumlauf, 1991) and are thus unlikely to regulate specification of the jaws. However, msx 1 and msx 2 (formally Hox-7 and Hox-8 respectively) expression within the neural tube extends caudally from the forebrain and they are expressed in the branchial arches, including the region of the presumptive tooth germ (Hill et al., 1989; MacKenzie et al., 1991a,b, 1992). msx 1 and msx 2 are two members of a diverged homeobox family homologous to the Drosophila muscle-segment homeobox (msh) gene (Hill et al., 1989). In Drosophila, msh is mainly expressed in the central nervous system and in segmented striated muscles of the body wall. No mutations of msh are known at present (Robert et al., 1989). In the mouse there appear to be three distinct msh-like genes, which are found at different loci, and are not clustered. Duplication of the msh-like genes has been suggested to be correlated with the emergence of the vertebrate body plan (Thorogood and Hanken 1992; Holland, 1991).

MacKenzie et al. (1991a,b, 1992) proposed that the initiation of tooth formation and the subsequent ability of condensed neural crest mesenchyme to induce tooth formation are related to expression of the msx 1 and msx 2 genes. They are first expressed in the early maxillary and mandibular processes with an anteroposterior gradient within the mesenchyme (msx 1, MacKenzie et al., 1991a) and both epithelium and mesenchyme (msx 2, MacKenzie et al., 1992). During subsequent odontogenesis, msx 1 is expressed by all of the condensed dental mesenchyme. However, msx 2 is expressed by subpopulations of both epithelium and mesenchyme, suggesting that it may play a role in regulating epithelial-mesenchymal interactions.

To investigate the relationship of msx 1 and msx 2 expression to epithelial-mesenchymal interactions during odontogenesis, we performed a series of tissue recombination experiments and analyzed these by in situ hybridization. The recombination technique has been used successfully to examine the expression of matrix molecules and homeobox genes (Thesleff et al., 1990; Takahashi et al., 1991; Vainio and Thesleff, 1992).

Hybrid mice (CBA×C57BL) were time-mated and the day of locating the vaginal plug was designated as day 0. The regions of the first mandibular molar tooth germs were removed from embryos at days E12, E14, E16 and E18 of gestation, and the first mandibular arch from E10 embryos. For heterotypic recombinations; the second branchial arch was taken at E10, the palatal shelves or diastema tissue at days E12 and E14, and the distobuccal oral mucosa at E16 and E18. The tissues were initially dissected in phosphate-buffered saline (PBS) at 4°C, prior to final preparation in calcium and magnesium-supplemented Dulbecco’s buffered saline (DPBS).

For recombination experiments explants were incubated for 2 minutes in freshly defrosted 2.25% trypsin/0.75% pancreatin and then kept for at least 20 minutes at room temperature in minimum essential medium (MEM) supplemented with 10% fetal calf serum (FCS). The epithelium and mesenchyme were teased apart, and any follicular mesenchyme discarded. Isolated dental and non-dental tissue fragments were then recombined on a polycarbonate membrane (0.1 μm pore-size, Nuclepore Corp.). The recombinations were cultured for 48 hours in Trowell-type (Trowell, 1954) cultures in glutamine-supplemented MEM/10% FCS at 37°C. We have shown previously that these culture conditions are permissive for early tooth initiation at E10 (Vainio et al., 1991), and tooth-specific cell differentiation after E16 (Thesleff et al., 1989). Recombination experiments were performed using tissues from E10, E12, E14, E16 and E18 embryos. These times correspond to the stages of: induction of the mesenchyme by the epithelium, mesenchymal condensation, organization of the enamel organ, morphogenesis and tooth-specific cell differentiation, respectively. The numbers of recombinations performed are detailed in Table 1.

The tissues were fixed overnight in 4% paraformaldehyde at 4°C, prepared for wax histology and 7 μm serial sections were dried onto silanized slides.

In situ hybridization was done as previously described (Wilkin-son and Green, 1990). Sense and anti-sense [³⁵S]UTP-labelled single-stranded RNA probes were synthesized by in vitro transcription of linearized pSP72 vectors containing msx 1 (MacKenzie et al., 1991a) and msx 2 (Monaghan et al., 1991) cDNAs. All transcripts were shortened to around 100 bases by limited alkaline hydrolysis, purified by gel chromatography and precipitated with ethanol. Labelling was detected by autoradiography after 10 days exposure at +10°C.

In agreement with previous work (MacKenzie et al., 1991a, 1992), we found msx 1 expression only in the mesenchyme, and msx 2 expression in both epithelial and mesenchymal tissues. Hybridization of the control sense probes was minimal by comparison with the antisense (data not shown).

To investigate some aspects of the in vivo expression of msx 1 and msx 2 a limited survey of gene expression during molar tooth development was undertaken. At E10, the first sign of tooth development, an epithelium thickening is seen (Fig. 1A,B). Both msx 1 and msx 2 were expressed in the mesenchyme adjacent to, but anterior to, this thickening. There was no expression of msx 1 in the epithelium, but msx 2 was expressed at a low level in the epithelium from the tip of the process into the anterior edge of the dental epithelial thickening (Fig. 1B). There was negligible expression of either msx 1 or msx 2 at this stage in the anterior region of the second branchial arch (Fig. 1A,B). MacKenzie et al. (1992) have shown expression in all four arches; we attribute this difference to our mouse strain developing faster and, at E10, being equivalent to an E11.5 MFI strain mouse.

MacKenzie et al. (1992) reported that msx 2 expression in the dental papilla mesenchyme increased during the progression of the tooth from the bud to the cap stage. We observed the appearance of strong mesenchyme hybridization of msx 2 already at the bud stage, suggesting an induction at this stage (Fig. 1D). Expression of msx 2 by the dental mesenchyme was observed at all later stages. Mesenchymal msx 2 expression was highest during the late cap stage and was restricted to the mesenchyme condensed around or inside the enamel organ (Fig. 2C). By contrast, msx 1 expression in the mesenchyme was more widespread and always included the putative follicular mesenchyme (Fig. 2B). Within the epithelial enamel organ the enamel knot showed marked expression of msx 2, and this was often contiguous with a short length of the buccal external enamel epithelium and internal enamel epithelium (Fig. 2C). During the late cap and bell stages, we found expression of msx 1 within all of the dental mesenchyme, and msx 2 in the dental papilla mesenchyme and epithelial enamel organ, as previously described (MacKenzie et al., 1992).

The patterns of expression of msx 1 and msx 2 in these tissue recombinations are summarized in Fig. 6.

Homotypic recombinations of dental epithelium (DE) and mesenchyme (DM) were made to confirm that under these conditions the expression of msx 1 and msx 2 was equivalent to that observed in vivo. Fig. 3 shows that after 2 days of culture DE and DM from teeth at various stages of development reorganized and underwent epithelial-mesenchymal interactions. These interactions were accompanied by expression of both msx 1 and msx 2. msx 1 was seen in mesenchyme of all recombinations. In recombinants from E14 or earlier, it was usual to see a region of the DM which was msx 1 positive and msx 2 negative (Fig. 3E,F), but at E16 and E18 stages DM was positive for both msx 1 and msx 2. The DE was negative for msx 1 in all explants. There was no expression of msx 2 in the DE in recombinations of E10 tissues, slight expression adjacent to the mesenchyme at E12 (Fig. 3C) and widespread expression in the DE thereafter (Fig. 3F,I).

To ensure that tissues that did not express msx 1 or msx 2 in vivo behaved similarly in vitro, control homotypic recombinants of non-dental epithelium and mesenchyme were made. These showed no expression above background of either gene although an intimate epithelial-mesenchymal interface was formed (Figs 4A,B, 5A,B). When DM (i.e. jaw mesenchyme from the presumptive area of the tooth) from early embryos (E10 and E12) was cultured in isolation, it disaggregated, whereas in explants from older embryos the DM remained compact and often formed cartilage beads. msx 1 and msx 2 were not expressed in isolated mesenchyme except when explanted at E18. At this stage, the mesenchymal cells maintained expression of msx 1 (but lost expression of msx 2 ) during 2 days in vitro.

In recombinations of DM with non-dental epithelium (OE), msx 1 expression was maintained in the mesenchyme near to the epithelial-mesenchymal interface at stages E10 to E18 (Fig. 4D). This expression was usually restricted to several cell layers. In these heterotypic recombinations, msx 2 expression was never seen in the mesenchyme, and only occasionally in OE (at E12, data not shown).

Until the bell stage (i.e. E10 to E16), DE induced expression of msx 1 in the non-dental mesenchyme (Fig. 5D,G,I). This induction occurred through much of the mesenchyme and a gradient of expression could be discerned in large pieces of mesenchyme (Fig. 5D,I). From the density of the silver grains, it appeared, subjectively, that the inductive ability of the DE decreased with increasing age (compare Figs 5G and I). At E18 induction did not occur. It was only at E10 that DE (i.e. first arch epithelium) was able to induce expression of msx 2 in non-dental mesenchyme (Fig. 5E). msx 2 expression was not seen in the DE of any of these heterotypic recombinants.

Expression studies of msx 1 and msx 2 in vivo confirmed the previous results of MacKenzie et al. (1992) with the exception that we observed marked mesenchymal expression of msx 2 at the late bud and early cap stages. The transition of expression into the mesenchyme appears to occur around the time when the mesenchyme becomes competent to induce odontogenesis (Mina and Kollar, 1989). msx 1 was expressed in both the papillary and follicular mesenchyme in a pattern similar to that for the cell surface-proteoglycan syndecan (Thesleff et al., 1988; Vainio et al., 1991). The lack of msx 2 expression by the follicular cells clearly reflects the diverging differentiation pathways of the two cell populations. This pattern of expression is similar to that of int-2, where transcripts are restricted to the dental papilla (Thesleff et al., 1990 and unpublished data), but the reverse of IGFII where transcripts are restricted to the dental follicle and around the dental papilla (Ferguson et al., 1992). Follicular tissue differentiates into the progenitor cells of the cementum, periodontal ligament and alveolar bone after the late bell stage (Palmer and Lumsden, 1987).

Progressive commitment of the dental papilla mesenchyme is highlighted by the homotypic recombinants. In recombinants of E10 and E12 tissue, the DE was often small compared to the DM. This led to a bare region of DM which did not express the homeobox genes. At E18, even isolated mesenchyme continued to express msx 1 showing that by this stage the cells were independent of epithelial maintenance, although requiring the presence of DE to express msx 2. At E14, when isolated DM did not express either msx 1 or msx 2, the size of the DE was often adequate to cover the DM. All of the mesenchyme expressed msx 1, but msx 2 was only expressed in regions of the DM that were enclosed by an enamel organ-like structure. This implies that the inductive events for msx 1 and msx 2 are different, with msx 2 expression only occurring over part of the DE/DM interface. This effect is probably due to the interaction of relatively uncommitted dental mesenchyme with different aspects of the cap stage epithelial enamel organ. The internal enamel epithelium was shown by Ruch et al. (1982) to be able to stimulate the dental papilla cells to proliferate, whereas the external enamel epithelium could not. This effect was believed to be mediated by the specialized composition of their basement membranes. As the dental papilla of late cap and early bell staged teeth can form both papillary and follicular tissues (Palmer and Lumsden, 1987), we suggest that the DM adapts to the nature of epithelium adjacent to it in the recombination and expresses msx 2 accordingly.

As in vivo, the selected non-dental tissues did not clearly express either of the genes when recombined and cultured. We observed that both isolated DM and DE became disrupted in vitro and did not express the genes except at E18. Such dedifferentiation has been described in both oral (Smith and Hall, 1990) and limb (Coelho et al., 1991) systems. As non-dental branchial-arch epithelium was capable of maintaining the expression of msx 1 within the previously msx 1-expressing DM, we believe that its effect is permissive and that some epithelial signal is required prior to final dental commitment. The absence of msx 2 expression in these recombinations suggests that there are differences in the specificity of epithelial-mesenchymal interactions. Other experiments confirm the presence of non-specific epithelial-mesenchymal interactions. For example, it is clear that epithelium is important for the integrity of immature mesenchyme by maintaining the rate of tissue-specific proliferation (Minkoff, 1991). Additionally, in both in vivo and in vitro experiments the distribution of expression of syndecan around the developing tooth (Vainio et al., 1991) is very similar to that of msx 1 . Furthermore, like msx 1, its expression is induced in dental mesenchyme by non-dental epithelium (Vainio et al., 1989).

Although the dental mesenchyme has been reported to be able to induce tooth formation in heterotypic epithelium (Kollar and Baird, 1969; Mina and Kollar, 1987) we did not see formation of morphologically distinct dental structures, and only once found msx 2 expression in the non-dental epithelium (at E12, data not shown). The previous experiments in which such inductive events were described utilized reimplantation of the recombinants into host animals. It is possible that the improved nutritional conditions were permissive for reorganization and that such reorganization cannot occur during 2 or 3 days in vitro.

The signaling between the epithelium and mesenchyme is likely to be partly mediated by diffusible peptide growth factors. This has been speculated to be the case for the induction of syndecan expression in early dental mesenchyme by the presumptive dental epithelium (Vainio and Thesleff, 1992). Most probably there are many factors involved. Some of these may mediate non-tissue-specific interactions (resulting in expression of, for example, msx 1 and syndecan), whilst others signal tissue-specific fate, inducing the expression of molecules restricted to determined cell lineages (for example; msx 2 and Egr-1; Karavanova et al., 1992; and reviewed by Hall and Ekanayake, 1991).

The range over which non-dental epithelium could activate msx 1 in mesenchyme seemed to be less than the distance for DE in both homotypic DE/DM and in heterotypic DE/OM recombinations. This could represent an effect due to different combinations (and/or concentrations) of diffusing inducers interacting with the extracellular environment. Furthermore, msx 1 expression in the mesenchyme appeared to be greater when E12 DE was used than when E16 DE was used. A loss of inductive ability of the dental epithe limn was detailed by Mina and Kollar (1987). Although the DE retained an ability to induce msx 1 expression, the abil ity to induce msx 2 was only found at E10. As expression of msx 2 seems to be a more specific marker for dental tissues than does msx 1, this finding is in agreement with Mina and Kollar’s (1987) report that the DE loses its instructive potential at the bud stage.

In the hindbrain, expression of the Antennapedia-like homeobox genes first occurs in the neuroepithelium and is maintained in the neural crest cells as they migrate into the caudal branchial arches. The neural crest cells are then able to induce expression of the same homeobox genes in the overlying ectoderm (Hunt et al., 1991). As msx 1 and msx 2 are expressed in the neural folds and adjacent crest cells, it is possible that a similar mechanism is responsible for the expression of msh-like genes in the mesenchyme and ectoderm of the first branchial arch and facial processes, creating the pattern of expression observed by MacKenzie et al. (1991a, 1992) at E9.

Expression of the msh-like genes is found in the vertebrate eye (Monaghan et al., 1991), heart and limb (Robert et al., 1991; Suzuki et al., 1991). Within these tissues, which develop by epithelial-mesenchymal interactions, msx 1 and msx 2 often show complementary expression in the tissues, suggesting that they may be involved in regulating interactions. The concept of two phases of msh-like homeobox gene expression is most clear in the chicken limb. In the early limb bud (H&H stage 17), the chicken homologue of msx 1 (GHox-7) is expressed throughout the mesoderm and GHox-8 expressed most intensely anteriorly (Coelho et al., 1991). This expression is independent of epithelial-mes-enchymal interactions. However, later in development (H&H stage 20 onwards), expression of both genes requires tissue interactions involving the specialized apical ectodermal ridge epithelium (Coelho et al., 1991; Robert et al., 1991). This repeating expression pattern during epithelial-mesenchymal organogenesis has also been described by Suzuki et al. (1991). Indeed, in Aves, it now appears clear that the second stage of expression of GHox-7, in chicken limb mesoderm (Coelho et al., 1991), and Quox-7, in quail branchial arch mesenchyme, are dependent upon the presence of competent epithelium. Coelho et al. (1991) further speculate that expression of GHox-8 by the epithelium is essential for normal mesoderm homeobox gene activation. From our data it also appears that localized expression of the msx 1 and msx 2 genes in the developing tooth is regulated by local epithelial-mesenchymal interactions.

Interestingly, members of the recently identified Dlx homeobox gene family (homologous to Drosophila distalless gene) show a similar pattern of distribution to the msh-like genes in both facial and limb tissues (Dollé et al., 1992). Like the msh-like genes, the Dlx homeobox genes are present only as single copies in insects, but amplified in vertebrates. Gaunt (1991) suggests that duplication of Hox families may be related to the evolution of new subsets of tissues and organs.

Expression of msx 2 by the dental mesenchyme appears to be indicative of a dental phenotype. This is suggested by its pattern of in vivo localization and its appearance in tissue recombinants. Early dental epithelium is capable of inducing msx 2 in mesenchyme, which then gains some inductive potential. Expression of msx 1 by the dental mesenchyme also requires epithelium. Until the late cap stage, the dental epithelium can induce msx 1, but not msx 2, expression in non-dental mesenchyme. We conclude that expression of msx 1 in mesenchyme can be induced by a relatively non-specific epithelial signal, whereas msx 2 expression in both epithelium and mesenchyme requires reciprocal interactions between specialized dental cell populations.

The similarity of expression patterns of msx 1 and msx 2 in murine dental tissues (our data), the eye (Monaghan et al., 1991) and in chick limb (Coelho et al., 1991; Robert et al., 1991) tissues, and the absence of clustering (cf. Antp-like genes; Gaunt, 1991), leads us to suggest that msx 1 and msx 2 are involved in epithelial-mesenchymal interaction-controlled organogenesis, rather than positional specification, and that msx 2 is expressed by those cells possessing the organizing potential.

We thank Dr R. Hill of the MRC Human Cytogenetics Unit, Edinburgh for the msx 2 probe. The skillful technical assistance of Ms Merja Mäkinen and Ms Riikka Santalahti is gratefully acknowledged. This work was supported by grants from The Finnish Academy, the Sigrid Juselius Foundation and by NIH grant DE09399. A. K. J. was a Royal Society European Exchange Fellow funded by The Finnish Academy.