Mouse Serrate-1 (Jagged-1): expression in the developing tooth is regulated by epithelial-mesenchymal interactions and fibroblast growth factor-4

Cell-cell contacts are critical for embryonic development. A mechanism to organize short-range signaling would be to furnish one cell with a transmembrane receptor and the neighboring cell with a membrane-bound ligand. One such system is based on Notch-like receptors and ligands of the Delta/Serrate type (Artavanis-Tsakonas et al., 1991, 1995; Simpson et al., 1992). This signaling system appears to function in most multicellular organisms since Notch-like receptors have been found in Drosophila (Notch), Caenorhabditis elegans (lin-12 and glp-1) and vertebrates (Notch 1, 2 and 3). Genes encoding Delta/Serrate ligands are similarly conserved in both invertebrates and vertebrates. Thus, Drosophila harbors Delta and Serrate, C. elegans contains lag2 and apx-1 and there are two newly discovered Delta and Serrate homologues in vertebrates (for reviews see: Nye and Kopan, 1995; Simpson, 1995).

The Drosophila Delta and Serrate proteins are structurally related (Vässin et al., 1987; Kopczynski et al., 1988; Fleming et al., 1990; Thomas et al., 1991). Both are single transmembrane spanning proteins with 9 and 14 epidermal growth factor (EGF)-like repeats, respectively, at the extracellular side. The extracellular region of both proteins also contains a DSL (Delta/Serrate/LAG-2) domain, which is specific to ligands of this type (Nye and Kopan, 1995; Simpson, 1995). The principal structural difference between Delta and Serrate is the presence of a cysteine-rich domain between the EGF-repeats and the transmembrane region in Serrate. Delta and Serrate have different temporospatial expression patterns and distinct roles during Drosophila development. Delta is expressed early and its activity is required for many aspects of development, including formation of the fly nervous system (Muskavitch, 1994). Delta regulates neurogenesis through a mechanism referred to as lateral inhibition: future neuroblasts which are committed to a neural fate express Delta and, via activation of Notch receptors on surrounding cells, inhibit them from taking on the same developmental fate (Heitzler and Simpson, 1991). Serrate, in contrast, is expressed at later stages during Drosophila development and seems not to be involved in lateral inhibition. Serrate is required for proper development of anterior spiracles and mouth-hooks and for specifying cells at the dorsal-ventral wing margin (Speicher et al., 1994; Couso et al., 1995; Kim et al., 1995).

Recently, highly conserved homologues of Delta (Betten-hausen et al., 1995; Chitnis et al., 1995; Henrique et al., 1995) and Serrate (Lindsell et al., 1995; Myat et al., 1996) have been identified in vertebrates. Phenotypes obtained from expressing wild-type and mutated forms of Delta in Xenopus are in agreement with a role as Notch ligand and a function in lateral inhibition during neurogenesis (Chitnis et al., 1995). Jagged1, the rat Serrate-1 homologue, is similarly believed to participate in Notch signaling as it seems to activate Notch in myoblasts in culture, as reflected by an inhibition of their differentiation (Lindsell et al., 1995), a phenotype also observed after overactivation of Notch receptors (Kopan et al., 1994). Taken together these findings suggest that Serrate-1 (Ser-1) and Delta are ligands for the three mammalian Notch receptors (Reaume et al., 1992; Weinmaster et al., 1992; Lardelli et al., 1994).

Little is known about regulation of vertebrate Notch receptors and ligands. In this study, we observed that mouse Ser-1 expression often occurs at sites of epithelial/mesenchymal interaction. This prompted us to address the question of Ser-1 regulation during odontogenesis, since the developing tooth allows various experimental manipulations, including epithelial/mesenchymal explant cultures, in order to study gene regulation (Vainio et al., 1993; Mitsiadis et al., 1995a,b; Thesleff et al., 1995). Furthermore, it has previously been shown that the expression of the three mouse Notch genes is under control of inductive tissue interactions and specific signaling molecules in this organ (Mitsiadis et al., 1995a). The data presented here indicate that epithelial-mesenchymal interactions and growth factors are important also in regulation of Ser-1 expression during tooth development, but in a manner different from that of Notch genes.

A mouse Ser-1 cDNA fragment was amplified by PCR with the same degenerate oligonucleotide primers previously used to clone the chick Serrate-1 cDNA: CGI(T/C)TITGC(T/C)TIAA(G/A)(G/C)AITA-(C/T)CA and TCIAT(A/G)CAIGTICCICC(A/G)TT (Myat et al.,1996). First-strand random-primed cDNA, prepared from E9.5 mouse embryo RNA, was used to amplify a 1.8 kb cDNA fragment derived from the mouse Ser-1 gene. This fragment was cloned into the Blue-script KS I vector and sequenced, and the sequence analysed using the Wisconsin GCG set of programs. The nucleic acid sequence has been deposited with Genbank (X91012).

F1(CBA×C57BL or CBA×NMRI) mice were used at various embryonic stages. The age of the embryos was determined according to the vaginal plug (E0.5) and confirmed by morphological criteria. Animals were killed by cervical dislocation and E7.5-E18.5 embryos were surgically removed and fixed overnight in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS), pH 7.4. The embryos were then washed in PBS and stored in absolute methanol (MeOH) at −20°C until analysis by whole-mount in situ hybridization. For in situ hybridization on tissue sections, several of the embryos (E10.5-18.5) were dehydrated and embedded in paraffin wax. Serial sections (5 μm) were mounted on silanized slides, dried overnight and stored in air-tight boxes at 4°C.

For in situ hybridization studies, an 1,800 bp fragment of Ser-1 mouse cDNA, derived from the region encoding the extracellular domain of the Ser-1, was used. After linearization of the plasmid vector with EcoRI and HindIII (Promega), single-stranded antisense (pT3) and sense (pT7) Ser-1 [³⁵S]UTP- and digoxigenin-labeled cRNA probes were prepared by standard procedures, as previously described (Henrique et al., 1995; Mitsiadis et al., 1995a,b). The[³⁵S]UTP-labeled (= 1,000 Ci/nmol, Amersham) probes were used at 50,000 cts/minute/ml. The digoxigenin-labeled (Boehringer Mannheim) Ser-1, mouse Notch 1, Notch 2 and Notch 3 (Mitsiadis et al., 1995a) ribo-probes were diluted to 1 μg/ml in the hybridization mixture. The mouse Krox-20 riboprobe (a kind gift of Dr Robb Krumlauf) was labelled with fluorescein. Whole-mount in situ hybridization and in situ hybridization on sections was performed as described previously (Henrique et al., 1995; Mitsiadis et al., 1995a,b; Myat et al., 1996). Double label in situ hybridization using digoxigenin and biotinylated riboprobes was performed according to the same protocols. The Ser-1 signal was visualized with NBT/BCIP and the Krox-20 signal with INP/BCIP (Boehringer Mannheim).

Recombinant FGF-2 (Boehringer Mannheim, Germany), FGF-4 (British Biotechnology Products, UK), BMP-2 and BMP-4 (a kind gift of Dr Elisabeth Wang, Genetics Institute, Cambridge, Massachusetts) proteins were used. The proteins were stored at −70°C until use. Affigel blue agarose beads (100-200 mesh/75-150 μm diameter; Bio-Rad) and heparin acrylic beads (100-200 mesh/100-250 μm diameter; Sigma) were used as carriers of BMPs and FGFs, respectively, as previously described (Mitsiadis et al., 1995a,b). FGFs and BMPs proteins were diluted with 0.1% BSA in PBS, pH 7.4, to concentrations of 100-250 μg/ml per 5 μl per 50 beads. Beads were washed for 5 minutes in culture medium before their implantation in the explants. Control beads were incubated in 0.1% BSA in PBS and treated identically.

Molar tooth germs were carefully dissected in Dulbecco’s PBS from the rest of the developing maxillary and mandibular processess of E11.5-E16.5 mouse embryos, and incubated 3 to 5 minutes in 2.25% trypsin/ 0.75% pancreatin on ice. Epithelial and mesenchymal dental tissues were separated in Dulbecco’s Minimum Essential Medium (DMEM) supplemented with 10% fetal calf serum (FCS; Gibco), and then transferred with a mouth-controlled pipette on pieces of Nuclepore filters (pore size, 0.1 μm) supported by metal grids (Trowell-type). Isolated dental mesenchyme were cultured alone or as recombinants with isolated dental epithelia. Furthermore, isolated dental epithelia were recombined together (2-3 tissues), as previously described (Jernvall et al., 1994; Mitsiadis et al., 1995a), to avoid apoptosis, which occurs when individual epithelia are cultured. Beads preloaded with BMPs, FGFs and BSA were transferred with a mouth-controlled capillary pipette on top of E13.5-14.5 explants. The explants were cultured for 20 hours in DMEM supplemented with 10% FCS in a humidified atmosphere of 5% CO2 in air, at 37°C. After culture, explants were fixed in 4% PFA overnight at 4°C, washed in PBS and finally stored in MeOH at −20°C until analysis by whole-mount in situ hybridization. For in situ hybridization and immunohistochemistry on tissue sections, explants (E11.5-16.5) were embedded in paraffin wax and serially sectioned (5 μm). The tissue recombination and bead implantation experiments were repeated at least 20 times for each experimental setting.

A rat monoclonal antibody against the human Ser-1 protein (a kind gift of Dr Artavanis-Tsakonas, Yale University, USA), and which cross-reacts with the mouse Ser-1, was used for immunohistochemical analysis (dilution 1:2000). Immunostaining was performed as previously described (Mitsiadis et al., 1995a,b).

Bromodeoxyuridine (BrdU) analysis in dental tissues

Cell proliferation in dental tissues was analyzed, both in vivo and in vitro, by using cell proliferation kits (Amersham, Boehringer Mannheim). For the detection of cell proliferation in vivo, several foster mothers were injected intraperitoneally with 5 mg/ml of BrdU (Sigma, St. Louis) in PBS (final concentration: 50 μg/g body weight) 1 to 2 hours before embryos were killed. BrdU-positive cells in teeth of E12.5-15.5 embryos were analysed on 5 μ m sections after immunostaining with anti-BrdU antibodies. For the detection of cell proliferation in vitro, the explants, after culture, were labeled for 1 hour with BrdU according to the manufacturer’s instructions and as described earlier (Mitsiadis et al., 1995b). Whole-mount immunohistochemistry with antibodies against BrdU was performed as earlier described (Mitsiadis et al., 1995b).

We have cloned a partial cDNA for the mouse Ser-1 gene, encoding a portion of the extracellular domain of the Ser-1 protein. The encoded protein is 85% identical to the chick Ser-1 protein (Myat et al., 1996) and 97% identical to the rat Jagged protein (Lindsell et al., 1995). Based on this homology and also on the conservation of the expression patterns (see below), we conclude that the cloned gene represents the mouse Ser-1 gene.

Expression of the mouse Ser-1 gene during postimplantation development was analysed by RNA in situ hybridization in whole-mount embryos and on sectioned material. Ser-1 mRNA is observed in embryonic mouse tissues from the earliest stage analysed (E7.5) and, at this stage, expression is confined to a small area around the primitive streak, presumably in embryonic endoderm (Fig. 1A). At E8.5, high levels of Ser-1 expression are found in two transverse stripes in the hindbrain region and lower levels are found in a third, more posterior, stripe (Fig. 1B). Double-labeling for Ser-1 and Krox-20 reveals that the second Ser-1 stripe corresponds to the presumptive rhombomere 3 and the third, posterior stripe to rhombomere 5 (Fig. 1D). The most anterior stripe corresponds to rhombomere 1 but the boundaries are not as sharp as for the stripe in rhombomere 3. Comparison between embryos at different developmental stages reveals that the stripe on rhombomere 5 is the last to appear (data not shown). Also, contrarily to Krox-20, the Ser-1 stripes are transient and disappear as the neural tube closes (not observed anymore at E9.5; Fig 1C and data not shown). These Ser-1 transverse stripes in alternating rhombomeres were not observed in the chick hindbrain (Myat et al., 1996) and raise the hypothesis that some of the molecular mechanisms underlying hindbrain segmentation are different between mice and chick embryos. At E9.5, the most intense hybridization signal is localized to a region between the first and second branchial arches, to the area just posterior of the second branchial arch, and to the otic vesicle (Fig. 1C). In Fig. 1E, an E9.5 embryo is shown after prolonged staining and cells expressing Ser-1 can be seen in the pronephros. Expression is also observed in a thin stripe of cells in the forming somite boundary, a feature not observed in chick embryos (Myat et al., 1996), but that can be functionally important, since targeted inactivation of the mouse Notch 1 and Su(H) genes affects somitogenesis (Conlon et al., 1995; Oka et al., 1995). At E11.5, Ser-1 expression is prominent in the facial processes, lens and in the growing limb buds (Fig. 1F) where Ser-1 mRNA is found in both the epithelium of the apical ectodermal ridge (AER) and in the most distal part of the mesenchyme (Fig. 1G). No hybridization signal was detected with sense probes at any developmental stages (data not shown).

The mouse Ser-1 gene shows complex expression patterns during neurogenesis. In a sagittal section of an E10.5 mouse embryo, Ser-1 expression is observed in the ventricular region around the fourth ventricle, in the developing midbrain and telencephalon (Fig. 2A), and in dorsal root ganglia (see below). In addition, there are longitudinal stripes of Ser-1 expression in the spinal cord (Fig. 2A). In a transverse section of an E10.5 spinal cord four transected longitudinal stripes can be identified (Fig. 2B), identical to the pattern previously described for Ser-1 in the chick (Myat et al., 1996) and for Jagged in the rat (Lindsell et al., 1996). In the developing E11.5 midbrain and telencephalon, a subset of cells shows an intense Ser-1 hybridization signal (Fig. 2C). This ‘salt- and-pepper’ expression pattern extends across the neuroepithelium, which at this stage is not subdivided into discrete layers (Jacobson, 1991; Altman and Bayer, 1995), and persists until E13.5. At E15.5, mesencephalon is organized into a ventricular layer containing proliferating cells and outer layers with differentiated neurons. During this stage, most of the Ser-1-expressing cells are confined to the inner, ventricular region (Fig. 2D). This redistribution of Ser-1 expression from E11.5 to E15.5 in the developing brain is consistent with expression in cells destined to become neurons and suggests a possible role for Ser-1 during mammalian neurogenesis.

To learn whether Ser-1 expression correlates with expression of any particular Notch gene, we compared Ser-1 expression to that of Notch 1, Notch 2 and Notch 3 in serial sections of an E12.5 mouse embryo (Fig. 3). The Ser-1 gene expression largely overlaps with expression of the three Notch genes but does not perfectly correlate with any particular Notch gene (Fig. 3). Thus, we find that for example in dorsal root ganglia Ser-1 is predominantly corexpressed with Notch 1, and only to a minor extent with Notch 3 (Fig. 3). Similarly, Ser-1 and Notch 2, but not Notch 1 and 3, are expressed in specific regions of the maxillary process (Fig. 3). Ser-1 and Notch 3, but not Notch 1 and 2, are coexpressed in the aorta (Fig. 3) and in most other blood vessels (data not shown). This may be of importance, considering the recent implication of the human NOTCH 3 in familial stroke and dementia (CADASIL) (Joutel et al., 1996). Taken together, these data indicate that interactions between Ser-1 and all three Notch receptors may occur, and this notion is supported by the strong conservation in the extracellular, ligand-binding regions (epidermal growth factor-like repeats 11 and 12) in all three Notch receptors (Lardelli et al., 1994).

Ser-1 expression is frequently associated with sites where epithelial-mesenchymal interactions occur. During kidney development, Ser-1 expression is observed in the E9.5 pronephros (Fig. 1E), and transcripts are localized to the epithelial tubules at E10.5 (Fig. 4A). At E13.5 through E15.5, expression is observed in the glomeruli but not in the mesenchyme of the developing kidney (Fig. 4B-D). In Fig. 4E the mesenchyme surrounding the epithelial bud of the growing gallbladder is shown to express Ser-1 at E10.5. Similarly, Ser-1 transcripts are observed only in the parenchyma, but not in the epithelium, of the E12.5 lung (Fig. 4F). In the developing whisker follicles, Ser-1 transcripts are first found in the epithelium that invaginates the underlying condensing mesenchyme at E12.5 (data not shown) and thereafter, at E13.5-E14.5, the gene is expressed in both the epithelium and mesenchyme of the developing whisker follicles (Fig. 4G). In the developing palatal rugae (E13.5-E14.5), transcripts are observed in both the epithelial and mesenchymal components (Fig. 4H), while the Ser-1 gene is mainly expressed in epithelium of the E13.5-E14.5 pancreas (Fig. 4I). In the E14.5 testis, only the mesenchymal cells express Ser-1 (Fig. 4J).

The complex expression patterns at epithelial-mesenchymal boundaries suggests that Ser-1 expression is regulated by tissue interactions. This idea was tested in the developing tooth, an experimental system in which the influence on gene regulation exerted by epithelium, mesenchyme and specific signaling molecules can be analysed (Vainio et al., 1993; Mitsiadis et al., 1995a,b; Thesleff et al., 1995). The tooth develops as a result of sequential and reciprocal interactions between neural crestderived cells and the oral ectoderm. Tooth development is schematically depicted in Fig. 5A and starts as an invagination of the E11 stomodeal epithelium into the underlying jaw mesenchyme. By E13, the dental epithelium acquires the bud con-figuration while the surrounding mesenchyme condenses. Cells of the dental mesenchyme lying directly under the epithelium differentiate into odontoblasts while epithelial cells in close contact with the dental mesenchyme differentiate into ameloblasts.

Ser-1 is first expressed in the presumptive dental epithelium (E10.5-11) (Fig. 5B), where the inductive capacity for tooth formation resides (Lumsden, 1988). From E12-12.5 expression is downregulated in a layer of epithelial cells contacting mesenchyme concomitant with an upregulation in the dental mesenchyme. This pattern of expression is maintained at the bud (E13-13.5) (Fig. 5C) and cap stages (E14.5-15.5) (see below). From E16.5 (early bell stage) (Fig. 5D) to E18.5 (late bell stage) (Fig. 5E), Ser-1 mRNA is absent from dental papilla mesenchyme, whereas expression is observed in the epithelial cells of the stratum intermedium and stellate reticulum. No expression is observed in inner enamel epithelial cells/preameloblasts at these stages (Fig. 5D,E). Expression persists in the mesenchyme surrounding the tooth germ and which forms the dental follicle (Fig. 5D,E).

To learn whether the transient upregulation of Ser-1 expression in dental mesenchyme from E12.5 to E15.5 is a consequence of epithelial signaling, we generated recombinants of dental epithelium and mesenchyme from three different time points: E11-11.5, E12.5-14 and E16.5. In cocultures of E11-11.5 dental epithelium and mesenchyme, Ser-1 mRNA is absent from the mesenchyme after 20 hours of culture while the gene is expressed in all epithelial cells (Fig. 6A), a pattern that resembles the in vivo situation. In the E12.5-14 recombinants, Ser-1 transcripts are detected in the mesenchymal cells adjacent to the dental epithelium after 20 hours of culture (Fig. 6B; see also below Fig. 7A′,B‵). In contrast, epithelial cells adjacent to the Ser-1-expressing mesenchyme cells do not express the gene at detectable levels, whereas epithelial cells further away from the mesenchyme do (Fig. 6B). The distribution of Ser-1 protein follows that of the mRNA in E12.5-14 explants, i.e. a layer of epithelial cells adjacent to the mesenchyme is devoid of Ser-1 protein, while the protein is found in both the epithelium located further away and in the mesenchyme (Fig. 6C). In recombinants of E16.5 dental epithelium and mesenchyme, Ser-1 transcripts are observed only in epithelial cells located far from the mesenchyme and not in epithelium close to the mesenchyme nor in the mesenchyme itself (Fig. 6D).To analyse further the epithelial-mesenchymal signaling we next performed heterochronic recombination experiments. When E13.5 dental epithelium was recombined with E11.5 dental mesenchyme, Ser-1 expression is observed in the epithelium, except for cells located close to the mesenchyme (Fig. 6E). There are only very low expression levels in the mesenchyme (Fig. 6E). In the converse experiment, where E11.5 epithelium is recombined with E13.5 mesenchyme, high levels of expression are seen in the entire epithelium, including the cells juxtaposed to the mesenchyme, while expression is very low in the mesenchyme (Fig. 6F).

Fig. 7A′ shows how Ser-1 is regulated by E12.5-14 epithelial-mesenchymal interactions in whole-mount explants. As in the sections (compare Fig. 6B), there is a zone in which Ser-1-negative epithelium is juxtaposed to mesenchyme, in which Ser-1 expressed at least from E14 (Vaahtokari et al., 1996) and FGF-8 is already present in the E10 thickened early dental epithelium (Heikenheimo et al., 1994, for review see Thesleff and Nieminen, 1996). BMP-4 is expressed in dental epithelium from E10 (Vainio et al., 1993) and BMP-2 and BMP-7 from E13 (Vaahtokari et al., 1996; Kratochwil et al., 1996; for review see Thesleff and Nieminen, 1996).

To first test the role of FGF, beads preloaded with FGF-4 (100 μg/ml) were implanted in E13.5-E14.5 dental mesenchyme, which also was recombined with E13.5-14.5 epithelia. After 20 expression is upregulated (see arrow in Fig. 7A’). Interestingly, there is also a region of the epithelium that expresses Ser-1 and the mesenchyme flanking this part of the epithelium does not express Ser-1 (see arrowhead in Fig. 7A ‵).

The transient upregulation of Ser-1 expression in mesenchyme at E12.5-15.5, both in vivo and in explants, suggests that Ser-1 expression is activated by epithelialderived signals present at these developmental stages. Factors that may mediate this signal include members of the FGF family and bone morphogenetic proteins (BMPs) (Thesleff and Nieminen, 1996). FGF-4 is hours of culture (Fig. 7B), the level of Ser-1 transcripts is increased in mesenchymal cells surrounding the FGF-4 beads (Fig. 7B ‵). At the epithelial/mesenchymal interface, the same pattern was observed as in Fig 7A ‵, i.e. a Ser-1-negative region is juxtaposed to Ser-1-positive mesenchyme (Fig. 7B ‵). An increase in Ser-1 expression is observed also when the FGF-4 bead was placed on mesenchyme cultured in isolation (Fig 7C ‵).

The appearance of a translucent area around the beads (Fig. 7B, C) confirms that biologically active protein is secreted from the beads (Vainio et al., 1993). When explants of E12.5-14 dental mesenchyme were cultured alone, Ser-1 expression is not detectable (data not shown).

To test if FGF-4 is involved in the control of Ser-1 expression also in dental epithelium, an FGF-4-releasing bead was added on top of E13.5-14.5 epithelial explants, which were cultured in cohorts because individual epithelia undergo apoptosis in culture (Fig. 7D). The expression of Ser-1 in the immediate vicinity of the bead appears not to be altered, suggesting that it is not influenced by FGF-4 (Fig. 7D ‵). It should be noted, however, that the individual epithelia show a composite expression pattern analogous to that seen in dental epithelium in vivo and after culture with dental mesenchyme, i.e. one part of the epithelium expresses Ser-1 and the other part does not (Fig. 7D’). This ‘patchy’ pattern in individual epithelia is stable for at least 20 hours in culture.

We next tested whether BMPs would regulate Ser-1 expression. Beads preloaded with 250 μ g/ml of either BMP-2 or/and BMP-4 were implanted in E13.5 dental mesenchyme. After 20 hours of culture the level of Ser-1 expression around a BMP bead is not increased above background (Fig. 8A). Regulation of Notch expression was then analysed in a similar manner. In contrast to Ser-1, Notch 3 transcription is significantly increased around the BMP-4-releasing bead in dental mesenchyme (Fig. 8B). Expression of Notch 1 and Notch 2 is also increased in response to BMP-4, although not as dramatically (data not shown). The E13.5 dental epithelium expresses moderate levels of Notch 3 mRNA (Mitsiadis et al., 1995a), but expression is upregulated around a BMP-4 bead (Fig. 8C). BMP-2 exerts the same effects on Notch expression as BMP-4 (data not shown). No increase in Notch expression is observed in dental mesenchyme after implantation of beads releasing either FGF-4 (Fig. 8D) or FGF-2 (data not shown).

In an attempt to learn how expression of Ser-1 correlates with cell proliferation, we first analysed the effects of BMP-4 and FGF-4 on BrdU incorporation in explants. BMP-4 or FGF-4 beads were placed on E13.5-14 dental mesenchyme and the mitotic activity was scored by labeling the explants with BrdU after 20 hours of culture. Beads releasing BMP-4 have no detectable effect on cell proliferation (Fig. 9A), while FGF-4 stimulates cell division close to the bead (Fig. 9B), indicating that the elevated Ser-1 expression correlates with an increase in cell proliferation in dental mesenchyme. As a control in Fig. 9A, a small piece of E12.5 dental epithelium was recombined with the E13.5 dental mesenchyme, and an increase in proliferation is observed in both the epithelium and in the mesenchyme contacting the epithelium. Recombination of an E11-12.5 epithelium and mesenchyme results in a strong increase in proliferation in the mesenchyme, but not in the epithelium (Fig. 9C). When more mature tissues are recombined (E12.5-14), proliferation is seen in both mesenchyme and epithelium contacting this mesenchyme (Fig. 9D). The increase in proliferation in dental epithelium coincides with the Ser-1 downregulation in the same region in E12.5-14 dental explants (compare Fig. 9D with Figs 6B and 7A‵,B‵).

We finally wanted to establish whether Ser-1 expression and cellular proliferation correlate also in vivo. Pregnant mice were injected with BrdU and molars of E12.5 and E15.5 embryos were analysed in parallel for cellular proliferation and Ser-1 expression. The regions of Ser-1 expression and cell proliferation show considerable overlap in dental mesenchyme, both at E12.5 (Fig. 9E,F) and at E15.5 (Fig. 9G,H). In contrast, in the dental epithelium, we find no obvious correlation between the distribution of Ser-1-expressing and BrdU-positive cells: at E12.5 (Fig. 9E,F), there are proliferating cells in Ser-1-negative areas closest to the mesenchyme and, at E15.5 (Fig. 9G,H), the absence of Ser-1 expression in inner enamel epithelium is accompanied by an increase in proliferation.

The mouse Ser-1 gene shows a complex and dynamic expression pattern during postimplantation development. Ser-1 expression largely overlaps with expression of the three Notch genes, but does not perfectly correlate with any particular gene, which suggests that interactions between Ser-1 and all three receptors are likely to occur. Mouse Ser-1 expression is often associated with sites of epithelial-mesenchymal interactions and a link between Notch signaling and these types of interactions has been suggested based on Notch expression patterns in hair follicles (Kopan and Weintraub, 1993) and after experimental tissue recombination in the developing tooth (Mitsiadis et al., 1995a). We find that, in some organs, such as kidney and pancreas, Ser-1 expression is predominantly epithelial, while in the developing gallbladder, lung and testis, Ser-1 mRNA is largely confined to the mesenchyme. In the whisker follicle and palatal rugae, both the epithelium and the mesenchyme express Ser-1. These data suggest a complex interplay of signals from both epithelium and mesenchyme to regulate the Ser-1 gene.

Ser-1 expression appears to be directly regulated by signals from mesenchyme and epithelium during odontogenesis. Ser-1 is first expressed in the dental placode where the inductive capacity for tooth formation resides (Lumsden, 1988). Thereafter, expression is upregulated in mesenchyme at a time corresponding to the shift of the odontogenic potential from epithelium to mesenchyme (E12.5; Lumsden, 1988) and persists until the early bell stage (E16). The in vivo expression pattern indicates that epithelial inducers are responsible for the transient upregulation of Ser-1 expression in dental mesenchyme. This upregulation was reproduced in tissue explants cultured in vitro: only dental epithelia derived from the E12.5 to E15.5 stage generate an upregulation of Ser-1 expression in isochronic recombined mesenchymal explants. Interestingly, no upregulation of Ser-1 expression in mesenchyme is seen in the heterochronic recombinations, i.e. neither the E13.5 epithelium/E11.5 mesenchyme nor the E11.5 epithelium/E13.5 mesenchyme explants generated elevated levels of Ser-1 mRNA in the mesenchyme. This suggests that the E11.5 mesenchyme is not competent to respond to epithelial signals and that the E11.5 epithelium does not possess the adequate repertoire of signaling molecules (e.g. FGF-4) involved in Ser-1 upregulation in mesenchyme.

Conversely, another signal appears to cause a downregulation of expression in epithelial cells, but only in the cells juxtaposed to dental mesenchyme, i.e. in the future preameloblasts. Downregulation is restricted in time: it is observed from E12.5 and onwards, but not at E11.5, either in vivo or in the isochronic explant cultures. The data from the heterochronic experiments show that E11.5 mesenchyme can downregulate Ser-1 expression in E13.5 epithelial cells contacting the mesenchyme, but not in epithelium located further away. Since no downregulation was observed in epithelium in the isochronic E11.5 explants, this indicates that the E11.5 epithelium is not competent to the mesenchymal downregulating signal(s). This notion is further supported by the heterochronic E11.5 epithelium/E13.5 mesenchyme experiment, in which no downregulation is seen in the E11.5 epithelium.

The source of the downregulating signal is not yet known, but could involve growth factors secreted from the mesenchyme and acting over a short distance. Diffusion of secreted factors is limited by binding to components of the extracellular matrix and of the basement membrane (Jessel and Melton, 1992), which is rapidly restored between epithelium and mesenchyme cultured in recombination (Ruch, 1987). The differential expression of Ser-1 and Notch genes (Mitsiadis et al., 1995a) in the future preameloblasts and other dental epithelial cells suggests that Notch signaling plays a role for the acquisition of dental epithelial cell fates by suppressing the ameloblast fate in cells that express both Ser-1 and Notch.

Although the absence of expression in dental epithelial cells contacting the mesenchyme appears to be a consequence of a signal emanating from dental mesenchyme for both Notch (Mitsiadis et al., 1995a) and Ser-1, there seems to be a difference in how this downregulation is accomplished. In vivo, downregulation of Notch in epithelium occurs from the earliest stage of tooth induction (E11). In contrast, Ser-1 downregulation in epithelium occurs when the odontogenic potential has shifted to the mesenchyme (E12.5) (Lumsden, 1988). In culture, isolated epithelia rapidly reexpress the Notch genes in all cells of the explant (Mitsiadis et al., 1995a), while the epithelia maintain a Ser-1 expressing and a non-expressing zone after 20 hours in culture. This ‘patchy’ pattern (Fig. 7D‵) is very similar both to the pattern observed in vivo (Fig. 5C) and when epithelia are recombined with mesenchyme (Fig. 7A‵, B‵). These findings indicate that putative negative mesenchymal signals are required for both initiation and maintenance of Notch downregulation in epithelium, whereas the negative signal acting on Ser-1 expression in epithelia is only needed to initiate downregulation of expression.

It is noteworthy that, in recombinants, the Ser-1-expressing and non-expressing regions of the epithelia may influence mesenchyme differently: Ser-1-expressing epithelium was consistently juxtaposed with Ser-1-negative mesenchyme, while epithelium not expressing Ser-1 was flanked by Ser-1-positive mesenchyme (see Fig. 7A‵). These Ser-1-positive and -negative regions of the epithelium may thus be endowed with different capacities to interact with juxtaposed mesenchyme. It is therefore possible that dental epithelium contains different signaling centers that participate in successive steps of tooth induction and patterning. An epithelial structure, called the enamel knot, has recently been reported to form such a center affecting tooth development (Vaahtokari et al., 1996), but other organizing centers may also exist.

Growth factors may act either as survival factors, stimulating the proliferation of certain predisposed cell populations, or as morphogens, influencing cell fate decisions (Gurdon, 1992). FGFs and BMPs can mimic different aspects of dental mesenchyme induction (Thesleff and Nieminen, 1996). BMPs have previously been shown to play important roles in the initiation and maintenance of tooth development, inducing the expression of several transcription factors (Vainio et al., 1993), and we show here that Notch, but not Ser-1, expression is induced by BMP in dental mesenchyme. While BMP-2 and BMP-4 increase the Notch expression in the bead assay and are present in dental epithelium at the relevant stages, we can not exclude the possibility that the effects on Notch expression in vivo are mediated by other BMPs or BMP-related molecules. FGFs lead to the opposite response in Notch and Ser-1 regulation: Ser-1 expression is upregulated in dental mesenchyme, while Notch expression is not. We do not know if the in vivo effects on Ser-1 expression are mediated only by FGF-4 or also by other members of the FGF family. FGF-4 is expressed from E14 in dental epithelium (Vaahtokari et al., 1994), whereas FGF-8 is expressed already at E10 in thickened early dental epithelium (Heikenheimo et al., 1994).

Exposure to BMP did not increase cellular proliferation, suggesting that BMPs act instructively to influence the fate of dental cells. In contrast, FGFs stimulate cell proliferation, which argues that FGFs act rather selectively to support survival of lineage-committed progenitors. We observe a close correlation between Ser-1 expression and BrdU-incorporation in dental mesenchyme both in vivo, after induction from epithelium in the explant cultures and after FGF stimulation. This is in keeping with data from both flies and vertebrates that Notch signaling is often observed in proliferative, uncommitted cells (Artavanis-Tsakonas et al., 1995). The situation is however more complex in dental epithelium, where we find no apparent correlation between Ser-1 expression and proliferation.

In conclusion, the data presented here provide evidence that signals from both dental epithelium and mesenchyme influence the expression of the Ser-1 gene in the developing tooth. It also appears that two major induction pathways, i.e. the FGFs and BMPs, are involved in regulating the expression of members of the Notch signaling pathway in the developing tooth, but affect Notch ligands and receptors differently. The developing tooth organ may therefore utilize growth factors in different ways to generate cellular diversity.

We are grateful to Dr Spyros Artavanis-Tsakonas for the gift of the monoclonal antibody against the human Ser-1, Dr Robb Krumlauf for the gift of the Krox-20 probe, Dr David Ish-Horowicz for valuable comments on the manuscript and Erik Nilsson for excellent technical assistance. This work was supported by grants from the Swedish Cancer Society (T. M. and U. L.), the Swedish Medical Research Council and Åke Wibergs Stiftelse (U. L.), the Imperial Cancer Research Fund (D. H.), the Finnish Academy (I. T.), and by short term fellowships from EMBO and the European Science Foundation (T. M).