What the fossil record tells us
It is generally believed that modern birds (Neornithes) derive from theropod dinosaurs (e.g. Ostrom, 1976), although a close relationship between birds and crocodilians has also been proposed on the basis, amongst other characters, of similarities of tooth ornamentation (Whetstone & Martin, 1979; Martin et al. 1980). The last common ancestor of Avialae was equipped with clearly functional teeth. The oldest widely accepted Avialae is the well-known Archaeopteryx lithographica, which lived some 150 Ma and possessed teeth. More than 70 avialan genera are known from the Mesozoic (140–120 Ma; Chiappe & Dyke, 2002), which allows a phylogenetic tree to be drawn based on anatomical characters (Fig. 2). The most recent toothed Avialae in fossil records, the ornithurine birds Hesperornis regalis and Ichthyornis dispar, are known from the late Cretaceous (93–65 Ma; Marsh, 1872; Gregory, 1952). H. regalis was a swimming bird. It had a long beak with a rhamphotheca covering the pre-maxilla region only and was provided with small but effective conical teeth set firmly in the jaw (Gingerich, 1975). The teeth of the upper jaws were few in number and set in the back part, whereas those of the mandibles formed a complete series (Fig. 2). H. regalis therefore had half a beak and teeth. To date, I. dispar is the closest Avialae to the common ancestor of Neornithes (Clarke, 2004). The teeth of I. dispar are set in a groove as in H. regalis, and they are broad and flattened with highly expanded roots (Martin & Stewart, 1977). I. dispar was a powerful flighting bird and did not differ notably from the common flying birds of the present time (e.g. terns). Most molecular dates for the divergence of Neornithes imply that they existed 40 Ma (Van Tuinen & Dyke, 2004) prior to the oldest identified ornithurine fossils in the late Cretaceous (Campanian, 80 Ma; Fountaine et al. 2005). Therefore, this would place the origins of modern birds in the early Cretaceous (120 Ma; Smith & Peterson, 2002). This later date is still earlier than that estimated from the fossil records but establishing accurate calibration times for molecular phylogenies on the basis of fossil data is a difficult task. However, the recent discovery of a close relative of ducks (Anseriformes) in the Maastrichtian of Antarctica (70 Ma; Clarke et al. 2005) indicates that Neornithes originated long before the Cretaceous/Tertiary boundary, probably earlier than 80 Ma, even if they diversified later, during the early Cenozoic (65 Ma; Zhou, 2004).
We do not know whether the fossil taxa closer to the most recent common ancestor of Neornithes than I. dispar have teeth. Therefore, we can reasonably estimate that tooth loss in crown Aves arose maximally on the stem lineage between I. dispar and Neornithes, and minimally in the most recent common ancestor of Neornithes, the origin of modern birds, i.e. approximately 100 Ma.
What the developmental genetics tells us
It has been known for more than a century and a half that transient epithelium thickenings, homologous to the dental lamina stage in the mouse, appear in the chick oral cavity at embryonic day (E)5 (Geoffroy Saint-Hilaire, 1820; Blanchard, 1860; Gardiner, 1884; Röse, 1892; Carlsson, 1896). This could be interpreted as the ‘remnants’ of an ancestral toothed condition. In birds, early tooth development arrest could be the consequence of mutations that inactivated the genetic pathways leading to tooth formation. In the last few decades, taking advantage of recent technical advances, several scientists have attempted to stimulate the ‘dormant’ odontogenic pathway in chicken with the ultimate goal of resuscitating teeth. This dream seemed attainable through elegant experiments involving either mouse/chick tissue recombinations aiming to reinitialize epithelial/mesenchymal interactions or using beads impregnated with various signalling molecules to mimic as closely as possible such interactions. The results of these experiments are summarized below and schematically analysed in Fig. 3.
Figure 3. A shift in the positioning of the odontogenic epithelium relative to the dental competent mesenchyme could explain the loss of the ability to form teeth in the modern bird ancestor. Schematic drawings summarizing the chick tooth experiments. (A) Mouse molar developmental stages, from bud [embryonic day (E)12.5] to cap ( E14.5) to bell (E16.5). The condensing mesenchyme around the bud stage tooth germ expresses Bmp4 and Msx1 and induces development of the enamel knot at the cap stage, which expresses signalling molecules such as Shh. The inner enamel epithelium forms the ameloblasts that form enamel, whereas the adjacent mesenchyme forms the odontoblasts that form dentine (see Caton & Tucker, 2009). (B) Chick development. At Hamburger & Hamilton (HH) stage 28 a bud-like thickening of the oral epithelium is observed. Expression of Bmp4 and Msx1 is not, however, associated with this region. No further tooth development is observed at later stages and the thickening regresses. Note that, at an earlier stage (stage 24), Bmp4 expression is epithelial and shifts into the mesenchyme at stage 28 (Francis-West et al. 1994). (C) When a bead impregnated with Bmp4 and Fgf4 is implanted into the chick epithelium, the expression of Bmp4 and Msx1 in the mesenchyme extends around the developing tooth bud. This leads to the extension and folding of the bud epithelium, and induction of Shh. No further progression of the tooth germs is observed, however (Chen et al. 2000). (D) When mouse mesenchyme is combined with chick epithelium (either by recombination of mandible tissue or by earlier neural crest grafts of mouse neural crest into a chick embryo), the chick epithelium induces Msx1 and Bmp4 in the mouse mesenchyme. The tooth germ progresses to the cap stage and forms an enamel knot-like structure expressing Shh. The mouse tissue differentiates into odontoblasts and forms a bell stage tooth germ. Tooth differentiation does not proceed beyond this stage and enamel is not deposited (Wang et al. 1998; Mitsiadis et al. 2003). (E) In the chick mutant talpid2 a shift in the positioning of the epithelium and mesenchyme has been described (indicated by dashed lines and arrows). The chick epithelium is able to induce expression of Bmp4 in the underlying mesenchyme and expresses Shh. The tooth germ develops by evagination, similar to that observed in alligator embryos. At later stages differentiated odontoblasts are identified by histology but no further differentiation occurs (Harris et al. 2006).
Download figure to PowerPoint
Kollar & Fisher (1980) performed a simple experiment in which they recombined dental epithelium of E5 chick embryos with molar mesenchyme of E16–E18 mouse embryos. These recombinant tissues were cultivated within the anterior chamber of a mouse eye. Several days later they obtained teeth with a dentine cone and an enamel cover, the famous ‘hen's teeth’. This meant that the cells of the E5 chick dental epithelium not only had retained the genetic potential to respond to the induction from the mouse mesenchymal cells for more than 100 Ma (tooth loss in a Neornithe ancestor) but they were also able to develop until the last developmental step (enamel matrix deposition). However, a possible contamination of mouse mesenchyme by mouse epithelium makes such interpretation uncertain. Indeed, it is difficult to completely eliminate contamination of mouse mesenchyme with residual epithelium and such a contamination would allow tooth formation (Arechaga et al. 1983). Eighteen years later, in another series of recombination experiments, Kollar's group showed that chick epithelium was able to induce cell proliferation and the expression of key developmental genes (Msx1, Msx2 and Bmp4) in the mouse mesenchyme, leading to odontoblast differentiation and formation of tooth germs (Wang et al. 1998) (Fig. 3D).
Another important finding was obtained by Chen et al. (2000) who showed that the early odontogenic pathway remained inducible in chick mandibles. During mandible development, they analysed the expression of crucial genes known to regulate mouse tooth morphogenesis. They discovered that the main markers for dental lamina formation in the mouse are expressed in the developing chick mandible (Fgf8, Pitx2, Pax9, Barx1, Msx1 and Msx2). However, the expression of three key genes (Bmp4, Msx1 and Msx2) was missing in the distal mandibular mesenchyme facing the presumptive chick dental lamina (Fig. 3B). When beads impregnated with Bmp4 and Fgf4 were implanted into the epithelium, facing cells in the distal mandibular mesenchyme expressed Msx1 and Msx2 (Fig. 3C). These experiments suggest not only that the odontogenic signalling pathway is conserved in the chick and can be reactivated but also that a defect in Bmp4 signalling could be responsible for the lack of Msx1 and Msx2 expression. A defect in Bmp4 signalling may therefore have occurred in a Neornithe ancestor, leading to a premature arrest of tooth development. However, exogenous addition of Bmp4, even in the presence of Fgfs, did not allow tooth development in the chick to proceed beyond the cap stage, i.e. to the stages of tooth germ differentiation. Structures that could be defined as teeth were not formed (Chen et al. 2000).
These findings obtained in vitro were confirmed in vivo by Mitsiadis et al. (2003) who performed transplantations of mouse (E8) neural crest cells in 1-day-old chick embryos. They showed that the avian dental epithelium was still able to induce a non-avian developmental programme in mouse neural crest-derived mesenchyme, resulting in tooth germ formation. Here again they did not obtain functional teeth but tooth germs developed until an advanced stage of differentiation in which some dentine-like matrix was deposited by the ectomesenchymal cells (odontoblast-like cells) and started to mineralize (Fig. 3D). No enamel-like structure was observed, however. This in-vivo study also indicates that endogenous factors coming from the mouse ectomesenchyme are more efficient at supporting tooth formation beyond the cap stage than exogenous factors used in the previous study (Chen et al. 2000). Unfortunately, the duration of these experiments was too short to determine whether or not tooth differentiation would have eventually reached a more advanced stage and the problem remains as to what extent tooth programmes are maintained in the chicken.
Recent observations made in a mutant chick [talpid2 (ta2) affected gene unknown], in which the development of several organ systems is affected, have brought additional light to the investigation on the mechanisms underlying tooth loss in birds. Indeed, the ta2 mutant was shown to develop rudimentary teeth under the rhamphotheca (Harris et al. 2006). These outgrowths from the distal mandible were conical and caniniform-shaped, morphological features that were similar to the so-called rudimentary teeth that develop in crocodile embryos. In the latter, these teeth do not develop further than the stage of dentine deposition and degenerate without any enamel covering being deposited. In contrast to the ta2 mutant, in crocodilians and lepidosaurians (lizards and snakes) when the rudimentary teeth are degenerating, first-generation teeth start to form, which are functional at hatching (and covered with enamel). Unfortunately, the oldest ta2 mutant embryos died at E16, several days prior to hatching, and so further developmental stages beyond the early dentine-like deposition were not available. It is therefore not possible to determine whether or not these rudimentary teeth would have degenerated and been replaced, as in crocodile embryos, by functional teeth. In ta2 mutants, genes necessary for tooth formation in the mouse are expressed in the mandible, including Bmp4 (Fig. 3E). The formation of advanced tooth germ in ta2 mutants is associated with defects in the specification of the oral/aboral boundary. This leads to a developmental repositioning of the presumptive dental epithelium to overlie mesenchyme competent to form teeth. Therefore, the authors proposed that changes in the relative position of a lateral signalling centre over competent odontogenic mesenchyme led to the loss of the ability to form teeth in an ancestral Neornithe.
In summary, these elegant experiments provide strong support for the mechanism of tooth loss in birds being a consequence of a developmental shift in the oral epithelium. This resulted in signalling molecules from the epithelium no longer reaching their targets in the mesenchyme and leading to a failure in the induction of key molecules, such as Msx1 and Bmp4, and an early arrest of tooth development. These experiments also indicate that, under appropriate conditions, the odontogenic capacity of the chicken dental epithelium can be reactivated. However, if the reactivation of such odontogenic pathways is a prere-quisite to initiate tooth development and to proceed further until tooth differentiation, it appears insufficient to form true functional teeth, i.e. a dentine cone covered with enamel. At the end of the developmental pathway, genes encoding for structural proteins might be activated but it appears that they were not activated in any of the experiments. In fact, they could not. Indeed, all of the genes encoding dental-specific proteins (dentine sialophosphoprotein for dentine; amelogenin, ameloblastin and enamelin for enamel) have disappeared from the chicken genome after chromosomal rearrangement or are pseudogenes (Sire et al. 2008). The revival of hen's teeth will therefore remain elusive.
Adaptations to tooth loss in birds
The fossil record tells us that the ability to form teeth was lost several times in non-avialan theropodan and avialan lineages (Sander, 1997). Three unrelated, extinct avialans lack teeth: Confuciusornis sanctus (early Cretaceous of China; approximately 120 Ma), Gobipteryx minuta and Apsaravis ukhaana (Campanian; approximately 80 Ma) (Chiappe et al. 1999, 2001; Clarke & Norell, 2002) (Fig. 2). However, the closest Avialae to modern birds is I. dispar, a toothed bird. Therefore, the loss of teeth occurred independently at least four times during Avialae evolution, including the lineage leading to modern birds. It is difficult to believe that there were four different mechanisms that led to tooth loss in Avialae and a shift in the mutual position of the competent tissues involved in epithelial/mesenchymal interactions may have been responsible in each case. However, how is it possible to survive the loss of such important tools? Indeed, we can suppose that tooth loss would have been lethal if these taxa were not already equipped with an alternative tool, the beak.
It is noteworthy that tooth loss in Avialae coincides with the presence of a beak. Indeed, although horny beaks are not preserved in fossils there are indications, for instance in the subjacent bone, of the probable presence of a rhamphotheca. Theropod and early toothed Avialae had no keratinized beak and had teeth on both the maxilla and pre-maxilla. H. regalis had a keratinized beak covering the pre-maxilla only, and no teeth in this region (Gingerich, 1975). This is a good example of a morphological intermediate structure between toothed birds, which lack a horny beak, and beaked, toothless birds. C. sanctus is the earliest bird known in the fossil record to have a toothless, horny beak, like modern birds (Fig. 2). It was hypothesized that C. sanctus fed on plant materials due to its toothless beak (Zhou & Zhang, 2003) but recent findings indicate that its diet comprised fish, like modern sea birds (Dalsätt et al. 2006).
In Avialae the horny beak represented a dramatic innovation resulting from the transformation of the reptilian snout into a beak. The rhamphotheca is assumed to have evolved from reptilian keratinized scales (e.g. Zweers et al. 1997). Interestingly, the BMP pathway was shown to be involved in the fine-tuning of beak morphoregulation in birds (Wu et al. 2004, 2006). It is speculated that this innovation was retained in Avialae because the beak compensated for the dedication of forelimbs to flight, which meant that the limbs were no longer efficient tools for finding (and manipulating) food. With beaks selected as a new tool for food uptake, the strong selective constraints on teeth were relaxed and their loss could have occurred with no drastic consequence for bird survival.