We cloned cichlid orthologues of genes required during oral epithelial organization and tooth germ initiation (bmp2, fgf8, pitx2, shh) 1138 as well as mesenchymal markers (bmp2, bmp4 36, pax9, runx2) involved in reciprocal signalling to the epithelium 11. Teleost tooth development has been well characterized in the zebrafish and thus our nomenclature for the early stages of tooth development will follow that model. Two stages of early odontogenesis are relevant: the thickened epithelium stage and the bell-shaped epithelium stage 394041. Developing teeth beyond this point will be referred to as tooth germs, spanning the progression of the tooth from a bell-shaped unit to various stages of functional maturity, characterized by cytodifferentiation.

The transcription factor pitx2, described as a putative odontogenic-commissioning gene 121315, has a broad expression pattern that encompasses both the developing tooth unit and the inter-tooth region, marking the extent of the dental-competent oral epithelium, including regions of future tooth rows (Figures 2a and 3A–C). pitx2 is one of the earliest dental epithelial markers (Figure 3C) with expression in the thickened dental epithelium (Figure 3C) and both the inner dental epithelium (IDE) and outer dental epithelium (ODE) of the maturing tooth (Figure 3A and 3B). Interestingly, the early pattern of pitx2 expression differs across the three species prior to and during morphogenesis of the first tooth, and reflects the future organization of these distinct dentitions (Figure 2a). L. fuelleborni shows the greatest region of dental competence (expression of pitx2), consistent with the later elaboration of teeth and tooth rows (Figures 1c and 2a). Future tooth rows also show expression of pitx2, labelling the lingual progression of the subsequent OB (Figure 3A and 3B). Similar to studies in other fishes 3841, we found that fgf8 expression is not associated with initiating tooth germs in Malawi cichlids (not shown).

pax9, one of the earliest mesenchymal markers of odontogenesis in the mouse, is either absent from or weakly expressed in the dentitions of zebrafish and Mexican tetra 38. By contrast, in Malawi cichlids pax9 is expressed initially in oral mesenchymal cells as a dental field along the mesiodistal jaw axis (Figure 2b, OB stage, and Figure 3D), then it is strongly up-regulated in the underlying mesenchyme at the epithelial thickening stage of the first tooth (Figure 2b, first-tooth stage, and Figure 3D). Expression of pax9 is then restricted to cells of the dental mesenchyme enveloping the tooth during morphogenesis and is absent from the cells of the dental papilla (Figure 2b, 3–4-teeth stage, and Figure 3E and 3F). The expression of runx2 essentially replicates that of pax9 for the stages examined (data not shown).

Expression of shh is up-regulated from the primary OB into the individual tooth germs (Figure 2c and 3G–I). shh continues to be expressed during tooth morphogenesis, marking the bell-shaped dental epithelium and later the IDE of the tooth during cytodifferentiation (Figure 3G–I). Notably, shh is never present in regions between or around teeth, called the zone of inhibition (ZOI) 151617. In Malawi cichlids shh expression continues to label OBs, marking subsequent initiation of more lingual tooth rows, one at a time (Figures 2c and 3G–I). C. afra does not develop a third OB (nor a third row), while M. zebra and L. fuelleborni initiate an OB for each future tooth row (see Figure 4). bmp2 is co-expressed in the competent epithelial OB with shh and pitx2. From the initial epithelial OB, bmp2 expression is up-regulated in the cells of the contorted bell-shaped epithelial germ and continues to be expressed during differentiation (Figure 3L) before becoming localized to cells of the mesenchymal dental papilla (Figure 3J–L). In addition, bmp2 is restricted to epithelial cells at the developing tooth tip (Figure 3K), which will differentiate to ameloblasts, partially responsible for the secretion of enameloid, the first mineralized tissue of the teleost tooth 42. These are an equivalent set of cells to the mammalian 'enamel knot'. bmp4 is expressed initially in the mesenchymal field along the mesiodistal axis prior to tooth germ initiation, much like pax9 and at the 3–4-teeth stage bmp4 is restricted to the dental papilla (data not shown 36).

These data, in conjunction with data reported earlier for bmp4 36, suggest that early patterns of gene expression differentiate among Malawi cichlid dentition types. L. fuelleborni, M. zebra and C. afra exhibit marked variation in the field of odontogenic competence both in the epithelium (pitx2, shh) and mesenchyme (pax9), in the spacing (bmp2, bmp4, pax9, runx2, shh) and in the size (bmp4, pax9, runx2, shh) of the initial tooth germs. These differences are readily quantifiable (Table 1), arise before teeth acquire their functional adult shape, and correspond directly to the size and spacing observed in the adult dentitions 26. For example, C. afra embryos have the largest initial dental germs and surrounding ZOI (Table 1) while adults have the largest, fewest teeth (mean ± SD of 1.3 ± 0.20 per millimetre of jaw width) with the greatest inter-unit spacing compared with the other two species, M. zebra (3.3 ± 0.52) and L. fuelleborni (4.8 ± 0.74) 26. The measurements obtained from shh expression in Table 1 highlight an early developmental origin of dental diversity among these three species. Streelman and Albertson noted a similar pattern from bmp4 expression, with a comparable range of cichlid species 36. While some measurements shown here seem counterintuitive (for example, Table 1b, columns 3 and 4), it is because species also differ in more subtle aspects of tooth initiation. C. afra never initiates a tooth between the first three teeth to develop; M. zebra has an initiation order that places tooth 5 between teeth 1 and 3, whereas L. fuelleborni initiates tooth 5 between teeth 1 and 2 (Figure 2c). These early differences in tooth size, spacing and organization are perhaps surprising because cichlids (and all teleost fish) continuously replace their teeth; therefore, we might have expected inter-specific variety to develop gradually, over multiple rounds of tooth replacement, from a common dental 'ground' state, as is the case for tooth shape 26.

Given the set of molecules localized to the first tooth germs (notably shh, but also pitx2 and bmp2), we hypothesized that other factors expressed within the ZOI surrounding these germs might guide the size and spacing of early cichlid tooth units. We therefore analyzed the expression of genes involved in a putative spacing mechanism. We cloned three genes, eda, edar and wnt7b, with antagonistic effects on shh in the initiation of mouse teeth, mouse hair follicles and chick feathers 9304344454647484950. edar (data not shown) is expressed within the germs themselves in a pattern similar to shh (Figures 2c and 3G–I). We observed the expression of both wnt7b and eda surrounding the initial shh-positive tooth germs (Figure 5a–c) within the ZOI/inter-germ regions across the three species.

eda is expressed locally and strongly surrounding the first-tooth germ (expressing shh and edar) in all three species (first-tooth stage data not shown). By the 3–4-teeth stage, eda remains expressed in the mesenchyme locally and heavily at the lingual margin of the first tooth in C. afra and M. zebra (Figures 5a and 5b and 6A–C), but its lingual expression is broader and more diffuse in L. fuelleborni (Figure 5c). Notably, there appears to be a lateral bias in the expression of eda in both L. fuelleborni and M. zebra, which may reflect the influence of (or may influence) the initiation of tooth 5 between existing germs 1 and 2 or 1 and 3, respectively. An apparent eda-negative region exists in both L. fuelleborni and M. zebra where the fifth tooth in the series will appear, a different position in each species (Figure 5, middle column). C. afra lacks tooth initiation between these first three positions, an arrangement that continues into the adult dentition.wnt7b expression coincides with the ZOI surrounding the first teeth in all species (Figures 5 and 6D–F); this is best illustrated comparatively with image overlays with shh as depicted in Figure 5a–c.

Remarkably, these genes seem also to be employed in the initiation and spacing of future tooth rows, an iterated expression pattern similar to tooth germ organization within each row. shh labels each OB for subsequent tooth rows (Figures 2c and 4); eda and wnt7b are expressed between the first tooth row and the OB of the second (Figures 5 and 6). Specifically, eda expression partly overlaps that of shh in the lingual OB, while wnt7b is expressed either side of shh. Thus, eda from the enveloping mesenchyme (Figure 6A–C) may induce and maintain shh expression in tooth germs as well as in future tooth rows, and planar epithelial wnt7b (Figure 6D–F) may inhibit dental competence in these regions, similar to the role of these molecules in other systems 9304445464748.

Our data suggest that the ZOI has an important role in patterning the size (and spacing) of the shh-positive tooth germs, especially the first unit to initiate. We speculated that the first tooth might possess unique regulatory properties as a source for continued induction and patterning of the dental program. We tested the role of the first tooth as a source of communicative signal for the organization of the dentition using targeted chemical inhibition of the hedgehog pathway at the first epithelial thickening stage. C. afra embryos at the first-tooth stage (five days post-fertilization (dpf); Figure 2c) were treated for 24 hours in 50 μM cyclopamine in 1% DMSO (based on protocols in 52), thus spanning the time from the initiation of the first tooth to the three-teeth stage (6 dpf), by which time the second row OB had established territory. A subset of treated embryos further developed for an additional 24 hours under standard conditions (7 dpf; Figure 7a and 7b); the remaining embryos from the same brood were allowed to develop for an additional six days (12 dpf; Figure 7c and 7d) to span the period of both first row eruption and development of the lingual tooth row (Figures 4, 7 and 8). Treated C. afra fixed at 7 dpf showed varying low levels of shh expression localized to a reduced number (one or two) of tooth germs on each side of the dentary and the OB for the second row (Figure 7b). shh expression appears within the region allocated for the ZOI (seen as shh-negative inter-germ regions in all controls; Figures 2c and 7a) normally expressing eda and wnt7b (Figures 5 and 6), suggesting a breakdown in both the initiation and spacing mechanisms (Figure 7b). Expression of eda in treated C. afra was absent in regions of tooth development (data not shown), implying that eda may respond to signals from, or downstream of, the hedgehog pathway. 1% DMSO control C. afra showed patterns of shh expression identical to standard controls (untreated).

In all treated individuals that were allowed to develop for a further six days (fixed at 12 dpf), we found that the first tooth continues partial development and shows signs of mineralization, although it does not complete development or attachment (Figure 7d). With the exception of a mineralized remnant of the first tooth, all other teeth, adjacent to the first and in subsequent rows, failed to develop (Figure 7d). Knockdown of the hedgehog pathway at the 3–4-teeth stage resulted in a functional, patterned and replacing dental system (data not shown). These observations demonstrate that when perturbed (via the hedgehog pathway) at the first-tooth stage, the dental programme cannot recover, despite continued cycles of periodic patterning past this stage in untreated individuals.

The comparison of gene expression across Malawi cichlid species with divergent dentitions suggests a simple model implicating pitx2, eda and wnt7b, and their interaction with shh and edar, as primary features of a periodic pattern generator for diversity in Lake Malawi cichlid dentitions (Figure 8). The model accounts for two aspects of dental patterning: how to put tooth rows in jaws, and how to put teeth in tooth rows. Our data suggest that the combination of pitx2 and shh is required for a competent field of tooth initiation (the OB, Figures 2, 3, 4 and 8). M. zebra and L. fuelleborni exhibit expanded expression of pitx2 lingually on the embryonic lower jaw; C. afra does not (Figure 2). pitx2 and shh are also co-expressed in each subsequent OB for M. zebra and L. fuelleborni (Figure 4); C. afra does not initiate a third OB. Therefore, the lack of lingual/oral co-expression of pitx2 and shh in C. afra (Figures 4, 8F and 8I) may account for the reduction in row number compared with the other species (Figures 4, 8G and 8J, H and 8K, respectively). The lack of combinatorial expression of shh and pitx2 in the oral region of zebrafish may partially explain the lack of teeth 38. Here we show that this mechanism likely accounts for variation in tooth row number among Malawi cichlids. Thus, molecular mechanisms used to pattern the first row of teeth (the only row of teeth in mammals and most vertebrates) are redeployed as 'triggers' of dental competence and initiation in each subsequent row. We suggest that the initiation of new tooth rows follows a 'copy and paste' mechanism wherein the dental expression network is redeployed for each new tooth row. Therefore, our model posits that preceding tooth rows are required as a source of signal to initiate the next lingual row during sequential addition.

The combination of comparative gene expression data and perturbation of the hedgehog pathway suggests that the correct initiation and maintenance of the first-tooth germ, via activation of shh, is necessary for the periodically patterned dental programme in Malawi cichlids (Figure 8). Comparison of the cyclopamine phenotype at the first-tooth to the 3–4-teeth stages shows that disturbing the development of the first-tooth germ has an effect on the entire dentition, whereas disrupting the dentition at later stages results in a mildly reduced phenotype with additional teeth forming and completing development. We do not yet understand the molecular mechanisms (for example, decreased epithelial proliferation and/or increased cell death) of this severe dental phenotype at the first-tooth stage.

Our data imply that eda and wnt7b, expressed in the ZOI, regulate initial tooth germ size and position within rows, through interactions with shh; wnt7b inhibits the germ through planar epithelial signals (Figures 5 and 6D–F) and eda maintains the tooth germ (shh and edar) from within the surrounding mesenchyme (Figures 5 and 6A–C). The ZOI may not lie solely within the layers of the epithelium and we suggest that inhibitor/activator controls signal from within the underlying mesenchyme that envelops the thickened dental epithelium 34. Once the periodic pattern is established, other molecules may act as inhibitors from within the developing tooth unit, for example bmp2, which is present both in the early epithelial thickening and within the dental papilla (mesenchyme) during maturation (Figure 3J–L), and bmp4, which is restricted to the dental papilla (data not shown; 36).

The expression of eda in the mesenchyme surrounding the developing dental germs of cichlids (Figures 5 and 6A–C) is more similar to that deployed during the patterning of feather placodes and salivary primordia 9 than that observed in mammalian dentitions, where it is restricted to epithelium 4446. In our model, a large initial tooth germ in C. afra results from sustained local and intense eda expression on a comparatively similar inhibitory background of wnt7b (Figure 5a). The size of this tooth germ is reduced in M. zebra (Figure 5b) and L. fuelleborni (Figure 5c) because the eda expression broadens earlier (especially for L. fuelleborni), a heterochronic imbalance setting the stage for more, closely packed shh-positive tooth germs (Table 1 and Figure 8). Consistent with our results, transgenic mice (K14-eda) with increased levels of Ectodysplasin expression exhibit larger tooth germs 4852. Furthermore, Eda null mutant mice have reduced tooth germs 48535455. However, in the mouse, effects of Eda on tooth size correlate positively with effects on tooth number; for example, higher levels of Eda lead to a single extra molar 4952. Our data and model point to an important distinction between overall levels of eda and its spatial expression over time. An earlier dispersion of eda expression after initiation of the first tooth (as in L. fuelleborni), rather than continued localized expression around that first-tooth germ (as in C. afra), may in fact lead to the production of more, smaller tooth germs (Figure 8).

The position of subsequent tooth rows is also specified in part by the expression of wnt7b and eda in our model. Mesenchymal eda plays a permissive role in the positioning of the lingual OB (Figures 5 and 6A). In C. afra, its expression is strongest medial to the first tooth, while in M. zebra and L. fuelleborni it appears more as a band along the mesiodistal axis (Figures 5 and 6A–C; also see the second row tooth positions in Figure 1). wnt7b also appears to demarcate the location of the second row, as its expression is either side of the shh-positive second OB (Figures 5 and 6D–F) and, in a similar iterative manner to the patterning of individual tooth units, wnt7b is restricted to the inter-row space (Figure 5a–c). Once the initiation of the primary dental pattern for each row is established, the essential nature of shh and genes that occupy the ZOI is lost; although they likely continue to be expressed during further morphogenesis (Figures 3 and 6), these molecules are probably no longer required for initiation of the secondary, replacement dentition 13.

Embryos and fry of three species of Lake Malawi cichlids (C. afra, M. zebra and L. fuelleborni) were raised to the required stage in a recirculating aquarium system at 28°C. Embryo ages (in dpf) were set after the identification of mouth brooding females (day 0). Embryos were then removed from the mouths of brooding females and, if required, were maintained for further development in separate culture tanks at 28°C.

Cloned sequences used to generate digoxigenin-labelled antisense riboprobes from Malawi cichlid species have been deposited in GenBank (accession numbers: EU867210 – EU867219). Many of the genes were identified through partial genome assemblies of L. fuelleborni and M. zebra 60 and cloned from M. zebra and L. fuelleborni cDNA libraries. Sequences of cDNA used to generate the probes are identical across the three species. Overall, these species exhibit almost no sequence divergence; the average nucleotide diversity for comparisons across the Malawi assemblage is 0.2%, less than among laboratory strains of the zebrafish 60.

To ensure the embryos of the three species were of equivalent stages (especially during gene expression comparisons), specimens were stage-matched based on external features, including pectoral and caudal fin development and eye development and maturity. Specimens for in situ hybridization were anaesthetized in tricaine methanesulfonate (MS222, Argent) and fixed overnight in 4% paraformaldehyde (PFA) in 0.1% phosphate-buffered saline (PBS) at 4°C. Whole-mount in situ hybridization experiments were based on protocols from 12 and modified as follows: embryos were transferred to methanol for dehydration and stored at -20°C. Specimens were rehydrated through to PBS with Tween-20 and digested with 4–10 μg/ml proteinase K (PK); the final concentration was based on the specific stage of embryo/fry (for example, embryos at approximately 5 dpf were digested with 5 μg/ml PK). Following hybridization, embryos were washed in TST (10 mM NaCl, 10 mM Tris-HCl, Tween-20 in depc-H₂O). During the colour reaction stage of the protocol, all embryos were allowed to fully develop the colour. Thus, embryos were continuously transferred into fresh NBT/BCIP solution (Roche) in NTMT until full staining had ensued; this was determined after multiple regions of known expression became positive. Specimens were stage-matched based on external features, including pectoral and caudal fin development and eye development and maturity. All in situ hybridization experiments were performed with multiple specimens (multiple individuals were fixed at regular intervals, within single broods, then repeated at least twice with alternative broods) to fully characterize the expression patterns within and across the three species. After colour reaction (NBT/BCIP, Roche) embryos were washed in PBS and fixed again in 4% PFA, before whole-mount imaging using a Leica Microsystems stereomicroscope (MZ16). Embryos were embedded in gelatin and chick albumin with 2.5% gluteraldehyde. The gelatin-albumin blocks were post-fixed in 4% PFA before sectioning. Thin sections were cut at 15–25 μm using a Leica Microsystems VT1000 vibratome.

From a single brood of 24 individuals, 14 C. afra embryos were treated with cyclopamine (LC Laboratories) compound (50 μM) from a stock (5 mM cyclopamine in DMSO) to make up a final 1% DMSO solution in fish water. Five C. afra individuals were used as a 1% DMSO control, under the same incubation conditions as the treated embryos (Figure 7a and 7c). A further five individuals were kept as standard controls (wild-type), developing in the Georgia Institute of Technology aquarium. Treatment and control experiments were performed in ventilated Petri dishes spinning at 28°C in an oscillating platform culture incubator (Barnstead Lab-Line Max 4000). Following the treatment experiments and for the controls with DMSO, fishes were washed 10 times in fresh fish water to remove any remnant of cyclopamine compound or DMSO before transferring to culture vessels containing at least 300 ml of fish water, changed daily until ready for analysis. Although initial experiments with 50 μM cyclopamine using 1% (of 95%) ethanol as the solvent (suggested by the manufacturer, LC Laboratories and previous reports 5161) showed differential expression patterns of shh to the 1% ethanol control experiments, alizarin red preparation of embryos raised to 12 dpf showed gross phenotypic effects on the ethanol-administered controls. Therefore, we substituted 1% DMSO for ethanol solvent, after which controls could not be distinguished from standard controls (untreated). While DMSO is not the best solvent for cyclopamine because of limited solubility above concentrations of 4 mg/ml, at the low concentrations used for enhanced viability of treated embryos, DMSO proved to be a better solvent than ethanol because of lower solvation temperatures and faster solvation times from -20°C storage temperatures.