Multiple epithelia are required to develop teeth deep inside the pharynx

Significance

Many vertebrates possess teeth deep in the pharynx. While teeth are known to derive from odontodes (skin denticles), it is unknown if an external epithelium is still required to produce a pharyngeal tooth, such as for odontode formation. Here, we show that the epithelial enamel organ of pharyngeal teeth in zebrafish is formed from endoderm, i.e. the internal germ layer. However, teeth develop 1) only when this endoderm becomes covered by a layer of cells with features of a periderm, i.e., the outer epithelial covering of the embryo; and 2) only when the endoderm has physically contacted the skin at the prospective gill slits. Thus, multiple epithelia are engaged in tooth formation, whether oral (mammals) or pharyngeal (teleosts).

Abstract

To explain the evolutionary origin of vertebrate teeth from odontodes, it has been proposed that competent epithelium spread into the oropharyngeal cavity via the mouth and other possible channels such as the gill slits [Huysseune et al., 2009, J. Anat. 214, 465–476]. Whether tooth formation deep inside the pharynx in extant vertebrates continues to require external epithelia has not been addressed so far. Using zebrafish we have previously demonstrated that cells derived from the periderm penetrate the oropharyngeal cavity via the mouth and via the endodermal pouches and connect to periderm-like cells that subsequently cover the entire endoderm-derived pharyngeal epithelium [Rosa et al., 2019, Sci. Rep. 9, 10082]. We now provide conclusive evidence that the epithelial component of pharyngeal teeth in zebrafish (the enamel organ) is derived from medial endoderm, as hitherto assumed based on position deep in the pharynx. Yet, dental morphogenesis starts only after the corresponding endodermal pouch (pouch 6) has made contact with the skin ectoderm, and only after periderm-like cells have covered the prospective tooth-forming endodermal epithelium. Manipulation of signaling pathways shown to adversely affect tooth development indicates they act downstream of these events. We demonstrate that pouch–ectoderm contact and the presence of a periderm-like layer are both required, but not sufficient, for tooth initiation in the pharynx. We conclude that the earliest interactions to generate pharyngeal teeth encompass those between different epithelial populations (skin ectoderm, endoderm, and periderm-like cells in zebrafish), in addition to the epithelial–mesenchymal interactions that govern the formation of all vertebrate teeth.

In chondrichthyans, basal sarcopterygians, amphibians, and actinopterygians not only the jaw margins but also the roof and floor of the pharynx can be tooth bearing, constituting a pharyngeal—next to an oral—dentition. Teeth—whether oral or pharyngeal—are evolutionarily derived from odontodes, also called skin denticles, dermal skeletal elements of ancient jawless vertebrates. The homology between odontodes and teeth is now well documented (1–6). Apart from being elements of the dermal skeleton, teeth and odontodes belong to the complex of ectodermal appendages whose development depends on reciprocal interactions between the surface epithelium (ectoderm) and the underlying (neural crest-derived) mesenchyme (7, 8). Accordingly, mutations of the ectodysplasin gene (EDA) or its receptor (EDAR) cause deficient development of hair, sweat glands, and teeth in humans, but also of pharyngeal teeth, scales, and dermal fin rays in zebrafish (9). As already noted by Charles Darwin, “Hairless dogs have imperfect teeth” (ref. 10, p. 30). However, different from the mammalian dentition that develops in an ectoderm-covered oral cavity (11), pharyngeal teeth in extant vertebrates develop in an endoderm-covered pharynx. How it was possible for dermal skeletal elements to develop deep inside the pharynx remains to be elucidated. It has been proposed that competent epithelium may have invaded the pharyngeal cavity via any channel that provides communication between the skin and the pharynx (12–14). In gnathostomes with open gill slits the pharynx can be covered with teeth, whereas in sarcopterygians (the lineage of tetrapods), pharyngeal teeth eventually disappear in the course of evolution together with the closure of the gill slits (15). Consequently, previous studies have stressed the importance of gill slits for pharyngeal tooth formation (12, 13).

Gill slits arise in areas where ectoderm meets endoderm. In vertebrates, the endodermal epithelium of the developing pharynx produces a series of bilateral outpocketings, called pharyngeal pouches, that eventually contact the skin ectoderm at corresponding clefts (16). In primary aquatic osteichthyans, most pouch–cleft contacts eventually break through to create openings, or gill slits (17–19). In teleost fishes, such as the zebrafish, six pharyngeal pouches form from anterior to posterior, separating the prospective pharyngeal arches (19–21). The first pouch (P1) is homologous to the spiraculum of chondrichthyans. It separates mandibular from the hyoid arch but does not usually open any longer in teleosts. The second pouch (P2) separates the hyoid arch from the third pharyngeal arch (also called first branchial or gill arch), pouch 3 (P3) separates pharyngeal arch 3 from 4, and so on. The pouches in vertebrates give rise to different organs crucial for immune responses and calcium homeostasis (16, 22). In zebrafish, teeth develop on the seventh (last) pharyngeal arch, i.e., posterior to pouch 6. Using various approaches, we have recently shown that periderm (the initial epithelial covering of the embryo) partially invades the pouches and that endogenous cells that resemble the periderm cells phenotypically, spread over the endoderm along the midline (23). Thus, at the time tooth formation is initiated the pharynx epithelium is composed of a double layer: a basal endoderm, that is overlain by a layer with periderm-like characteristics.

Using mutant and transgenic (Tg) zebrafish, as well as experimental manipulations (pharmaceutical inhibition experiments as well as mechanical perturbation of pouch development), this study now tests whether pharyngeal tooth initiation depends on the presence of pouch–cleft contacts and/or the presence of the periderm-like layer. We demonstrate that the enamel organ develops from the endodermal epithelium, as hitherto assumed based on the localization of the teeth deep in the pharynx. However, dental morphogenesis starts only after pouch 6 has made contact with the ectoderm, and only after a layer of periderm-like cells has covered the prospective tooth-forming endodermal epithelium. We conclude that the earliest interactions required to make a tooth deep in the pharynx encompass those between different epithelial populations, in addition to the epithelial–mesenchymal interactions that govern the formation of all vertebrate teeth.

Results

Pharyngeal Teeth Develop from the Endoderm Lining Pouch 6.

The first pair of teeth (designated as 4V¹, ref. 24) appears at around 48 h postfertilization (hpf) as a placodal thickening of the pharyngeal epithelium on both sides of the midline at the level of pouch 6 (P6). More precisely, these teeth develop where the contact zone of pouch 6 with the ectoderm ends posteriorly (Fig. 1A). The development of tooth 4V¹ is soon followed by that of its successor, 4V², and the germs of teeth 3V¹ and 5V¹, medially and laterally of 4V¹, respectively (Fig. 1B). Using Tg(sox17:egfp) zebrafish, we could establish that the enamel organs are derived from sox17-positive epithelium (Fig. 1 C and D, and see SI Appendix, Fig. S1 A–C), clearly indicating that the epithelial component of the teeth is of endodermal origin. In particular, only the single layer adjoining the basal lamina folds into the prospective enamel organ (compare, e.g., figure 2J in ref. 25 with Fig. 1E). The requirement for endoderm in pharyngeal tooth formation was confirmed by the study of casanova mutants. Cas encodes a protein of the SoxF family that also includes Sox17 but it works upstream of Sox17 (26). casanova mutants do not develop endoderm (26). Serial sections of casanova mutants revealed the absence of pharyngeal pouches and of teeth (SI Appendix, Fig. S2), in line with the observation that enamel organs derive from sox17-positive endodermal epithelium.

Fig. 1.

Pharyngeal teeth develop from the endoderm lining pouch 6 after pouch 6 has made contact with the ectoderm. (A) Development of the placodes of teeth 4V¹ (thick arrows) from the endoderm at 48 hpf; on each body side, the contact point of pouch 6 (P6, thin arrows) with the ectoderm is indicated by asterisks. Posterior to this cross-sectional level, there is no longer contact of endoderm with ectoderm. No lumen is present in the pharynx yet. Arrowheads point to flattened midline cells (encircled by yellow dotted line) squeezed between endodermal layers (surrounded by black dotted line). (B) At 96 hpf, more tooth germs have developed. Pouch 6 has opened into a gill slit anterior to this cross-sectional level (visible on Fig. 1J) and the pharynx is now wide open. (C and D) Tooth germs (4V¹, thick arrows) at 56 hpf are derived from sox17-positive endoderm (outlined by a black and white dotted line, respectively). Periderm-like cells are outlined by a yellow dotted line. (E) Magnification of the tooth germ shown in D, indicating that folding into the prospective enamel organ occurs by the basal, sox17-positive, endodermal layer only. White and yellow dotted lines surround endoderm and periderm-like cells, respectively. (F–H) Successive stages in the formation of contact of pouch 6 (outlined by a dotted line) with the ectoderm. (F′–H′) Corresponding stages in Tg(sox17:egfp) embryos, immunostained for laminin. Depending on the angle of sectioning, pouch 5 can still be visible in the section along with pouch 6. Note breaking up of the basal lamina at 38 hpf (G′) and basal lamina of pouch 6 continuous with that underlying the ectoderm (H′). Arrowheads point to basal lamina of pouch 6 at its (prospective) contact place with skin ectoderm. (I) Cross-section of 72 hpf embryo at the level of P6 showing ongoing lumen formation (asterisks) (endoderm surrounded by dotted line). (J) At 96 hpf, pouch 6 has opened on either side into a gill slit. Cartoons show representative embryonic and postembryonic developmental stages and level of transverse sections shown in the different figures (boxed areas). Dark gray: cartilage; medium gray: endoderm; light gray: other tissues. Abbreviations: 3V¹, 4V¹, 5V¹: first-generation teeth in positions 3, 4, and 5 of the ventral tooth row; 4V²: second-generation tooth in position 4V (compare ref. 24); b: brain; CB5: ceratobranchial 5; e: eye; nt: notochord; ov: otic vesicle; P6: pouch 6; WT: wild type; y: yolk. (Scale bars in A–D and F–J, 50 µm.), (Scale bar in E, 20 µm.)

Thus, the endoderm layer is essential to produce the enamel organ of the pharyngeal teeth.

Teeth Develop Only When Pouch 6 Has Made Contact with the Ectoderm.

Earlier, we suggested that an ectodermal cellular contribution or signal may be essential for tooth formation (12, 13). Given that teeth always develop from midline endoderm immediately caudal to the contact area between the endodermal pouch 6 and the skin ectoderm, we investigated the possible contribution of ectoderm via this pouch. To do so, we first examined the timing of contact of pouch 6 with the ectoderm.

Observations on whole mount Tg(sox17:egfp) zebrafish of closely spaced developmental stages revealed that pouches 1 to 5 (P1–P5) are fully formed by 32 hpf, confirming earlier observations by Kopinke et al. (21) (i.e., P5 formed around 30 hpf) and Choe et al. (27) (i.e., P5 formed around 32 hpf). At 36 hpf, a bud-like thickening projects from the posterior end of pouch 5 and is recognizable as pouch 6 by 38 hpf. By 40 hpf, this pouch has thinned, while the pharynx continues to extend posteriorly to meet the alimentary canal (SI Appendix, Fig. S3 A–C and A′–C′). Sections of WT, as well as Tg(sox17:egfp) zebrafish stained for laminin, reveal details of the contact between pouch and ectoderm. At 36 hpf, the lateral part of the pouch has taken on a club shape and the lateralmost pouch cells extend filopodia toward the ectoderm. At this time point, the skin is thickened at the prospective contact point and is bilayered, with a single basal ectodermal layer covered by a periderm layer and sharply delimited from the underlying mesenchyme by a distinct basal lamina (Fig. 1 F and F′). At 38 hpf, the pouch endoderm has merged with the skin ectoderm, and the basal lamina between pouch 6 endoderm and ectoderm is now fragmented (Fig. 1 G and G′). At 40 hpf the pouch has thinned into a bilayer, covered by a basal lamina, which continues imperceptibly with the basal lamina underlying the ectoderm (Fig. 1 H and H′). These shape changes match those observed on whole mounts (SI Appendix, Fig. S3 A–C). At the cross-sectional level of pouch 6, several lumina first appear along the midline at around 56 hpf and soon become confluent (Fig. 1I). The pouch itself persists as a bilayered sox17-positive endodermal epithelium until the lumina merge into a single pharyngeal lumen and extend outwards, thus finally producing the open gill slit between the sixth and seventh arch at 72 hpf or beyond (Fig. 1J).

Taken together, tooth buds start to form from the endodermal layer ∼10 h after pouch 6 has contacted the ectoderm, yet well before the gill slit opens. Given this consistent developmental sequence of contact of pouch 6 with the ectoderm (below referred to as “pouch 6 contact”) and start of tooth formation, we next tested whether this contact is a prerequisite for teeth to form, using mutant zebrafish with anomalous pouches, and embryos having experienced mechanical perturbation of pouch formation (SI Appendix, Table S1).

Earlier, it was reported that van gogh (tbx1) mutants display defective pouch formation (28, 29), thus providing an excellent way to test for a link between pouch 6 contact and tooth formation. We examined 30 specimens ranging from 48 to 144 hpf, sectioned serially (not to miss any contact or tooth germ), and assessed each body side separately (n = 56; not all sides could be scored unequivocally for the presence of pouch 6 contact). We determined whether tooth germs were present, and whether pouch 6 endoderm contacted the ectoderm anterior to the position of the tooth germs (Fig. 2 A–F and SI Appendix, Table S1A). Teeth were present if pouch 6 had made contact with the ectoderm, possibly even opened into a gill slit (n = 20/56) (Fig. 2 B and C, arrow). Teeth were absent if there was no such contact (n = 18/56) (Fig. 2 A and B, arrowheads). Frequently, teeth were absent despite the presence of pouch 6 contact (n = 16/56). In only 2 cases out of 56, teeth were present in the apparent absence of pouch 6 contact (Fig. 2 D and E; compare to contralateral side, Fig. 2 E and F). However, at least in the case pictured, the epithelial buds are questionable as tooth germs (Fig. 2E). Interestingly, four embryos had teeth unilaterally. In two of those, absence of teeth corresponded with absence of pouch 6 contact at that body side, while the contralateral side presented with pouch 6 contact and teeth (Fig. 2 A–C). In the two others, one side presented with pouch 6 contact and teeth, the contralateral side pouch 6 contact but no teeth.

Fig. 2.

Teeth do not form in case of defective pouch 6 contact with ectoderm. (A–F) vgo^−/− mutant zebrafish at 5 dpf. (Middle) show midline with (arrow) or without (arrowhead) developing tooth germs; (Left and Right) show corresponding contact of pouch 6 with ectoderm and ensuing gill slit formation (Right) or absence thereof (Left), respectively. (G–I) Edema (thin arrow) causing extremely perturbed pouch formation in a vgo^−/− mutant zebrafish at 96 hpf (G), tooth absence (H, arrowheads), and corresponding lack of contact of pouch 6 with ectoderm (I, arrowhead). (J–L) Extreme unilateral edema (J, thin arrow) in a 7 dpf laf^−/− mutant zebrafish preventing pouch 6 contact with ectoderm, and corresponding tooth absence on that side (K, arrowhead), while pouch 6 at the contralateral side has opened into a gill slit (L) and a tooth is present (K and L, arrow). (M–O) Unilateral edema in a 96 hpf Tg(krt4:gfp) zebrafish preventing pouch 6 contact with ectoderm, and corresponding tooth absence on that side (M, arrowhead), while pouch 6 at the contralateral side has opened into a gill slit (O) and a tooth is present (N, arrow). Cartoons show the transverse sections from which the different details are taken (boxed areas). Dark gray: cartilage; medium gray: endoderm; light gray: other tissues. Abbreviations: asterisks: pharyngeal lumen; b: brain; ed: edema; nt: notochord; ov: otic vesicle; P6: pouch 6. (Scale bars, 50 µm.)

To further test the relationship between tooth formation and pouch 6 contact with ectoderm, we examined zebrafish in which pouch development and contact with ectoderm was perturbed mechanically. Pericardial or yolk sac edema provokes a severe disruption of contact between ectoderm and pouch endoderm or impedes this contact altogether. We examined serial sections of 14 embryos between 96 and 168 hpf that had developed pericardial edema, either naturally, or as the result of mutation or induction by ethanol. Given the (infrequent) asymmetric presence of teeth in the vgo mutants, and the occasional unilateral occurrence of edemas, we scored each body side separately (n = 28) (SI Appendix, Table S1B). In all four vgo^−/− mutants that had developed an edema, teeth were absent on both sides (n = 8) and no pouch 6 contact was observed (Fig. 2 G–I). In nine other specimens with edema, teeth were present on both sides (n = 18) along with pouch 6 contact. Finally, in one (168 hpf old) laf^−/− mutant (Fig. 2 J–L) and in one (96 hpf old) Tg(krt4:gfp) specimen (Fig. 2 M–O), teeth were absent unilaterally, coinciding with extremely disturbed pouches on the edentulous side.

In vertebrates, pouch–cleft contacts have been reported to act as signaling areas as judged by the localized expression of certain genes, such as Eya1 (30) or wnt4a (27). These genes are important for normal pouch development, respectively in mammals (31) and zebrafish (27). Rather surprisingly, the six dog-eared (eya1, dog^tm90b) mutant specimens (12 sides) did not display a tooth phenotype (SI Appendix, Fig. S4 A and B and Table S1C). Mutants for wnt4a also displayed a normal dentition (27).

Taken together, our data provide a strong indication that pouch 6 needs to contact the overlying ectoderm before tooth formation can occur, yet the molecular player(s) remains elusive.

Teeth Develop Only once a Periderm-Like Layer Has Covered the Endoderm.

Using Tg(krt4:gfp) zebrafish, we observed that the placode of the first tooth forms only after a layer of krt4-positive cells has appeared along the midline, squeezed between the two layers of endodermal epithelium, and with cells more flattened than in the endoderm (Fig. 3A). Earlier, we traced the origin of these krt4-positive, periderm-like cells back to a source of cells more anteriorly, at the level of pouch 2, expanding posteriorly in the pharynx along the midline. These cells express sox17 from early onwards (SI Appendix, Fig. S1D). They reach the level of pouch 6 at around 48 hpf (Fig. 1A, and see ref. 23), followed by the appearance of small lumina in between them (Fig. 3B and SI Appendix, Fig. S5). The separate lumina coalesce into a larger lumen that eventually separates the future floor from the roof of the pharynx (Fig. 3C). The future pharynx floor is then constituted of a deep layer of endodermal cells resting on the basal lamina, covered by a superficial layer of more flattened periderm-like cells (further referred to as “midline cells”), facing the lumen (Fig. 3C). The same two tiers, in mirror image, form the future roof of the pharyngeal cavity. While the endoderm loses its sox17 expression, the midline cells continue to strongly express sox17 (Fig. 3C), while maintaining krt4 expression (Fig. 3A and SI Appendix, Fig. S1E). Pouch 6 itself persists as a sox17-positive endodermal bilayer until at least 56 hpf (Figs. 1D and 3C).

Fig. 3.

Teeth develop only once a periderm-like layer has covered the endoderm. (A) From 48 hpf onwards krt4-positive cells (white arrowheads) cover the (unlabeled) endodermal lining of the pharynx (white arrows) at the posterior margin of pouch 6 (P6). The outline of the endoderm is indicated by a dotted line. (B) Same area as in A, shown in TEM. The originally two layers of endoderm, composed of high cylindrical cells, now become separated from each other by flattened to cuboidal periderm-like cells (white arrowheads, outlined by a dotted line). Between the latter, the first lumina have appeared (red arrow). These will separate dorsal from ventral pharyngeal epithelium, each constituted of a basal endodermal and a superficial, periderm-like layer. (C) The superficial layer (so-called midline cells, outlined by a white dotted line) starts to express sox17, even stronger than the basal endodermal layer (white arrow). (D–F) Control Tg(krt4:gfp) embryos at 72 hpf, treated with DMSO (dimethylsulfoxide) only, showing opening of pouch 2 (P2) into a gill slit (D), yet a still closed pouch 6 (E) and tooth germs clearly present (F, thick arrows). Note the presence of a krt4-positive layer (E and F white arrowheads) covering the unlabeled endoderm (indicated by a dotted outline in E and F). (G–I) Corresponding images of a DAPT-treated 72 hpf Tg(krt4:gfp) embryo at levels corresponding to those shown in D–F. Note opened pouch 2 (G), closed pouch 6 (H), presence of a krt4-positive layer (H and I, white arrowheads) covering the endoderm (outlined by a dotted line in H and I), and clear presence of tooth germs (thick arrows in I), similar to control animal. (J–L) WT embryos at 72 hpf showing an open pouch 2 (J), open pouch 6 (K), and tooth germs clearly present (L, thick arrows). (M–O) mib^−/− mutant embryo at levels corresponding to those shown in J–L. Note open pouch 2 (M), closed pouch 6 (N), presence of midline cells covering the endoderm (N, white arrowheads), and clear presence of tooth germs (O, thick arrows). Endoderm is outlined by a dotted line in N and O. Cartoons show representative postembryonic developmental stages and level of transverse sections shown in the different figures (boxed areas). Dark gray: cartilage; medium gray: endoderm; light gray: other tissues. Abbreviations: asterisk: pharyngeal lumen; b: brain; e: eye; nt: notochord; ov: otic vesicle; P2: pouch 2; P6: pouch 6; >P6: level posterior to pouch 6; y: yolk. (Scale bars, 50 µm in A, C–O, and 10 µm in B.)

Because of the striking temporal coincidence between the appearance of these krt4-positive midline cells and the start of tooth formation, we scored all 50 embryos—previously examined for pouch 6 contact—for the presence of these cells in the midline at the level of pouch 6. Their presence was established either through a GFP signal, or by microscopic observation at high magnification (SI Appendix, Table S1 and Fig. S6). In all but the two youngest (48 hpf) embryos such cells were present, either as individual cells, or up to a complete additional layer. Tooth germs were present in only half of the embryos (n = 50/100 sides), but present, without exception, only when midline cells were present as well.

Because we previously established that the midline cells, observed to cover the enamel organ-forming endoderm, do not derive from skin periderm (23), we expected that mutants with perturbed skin periderm would not yield a tooth phenotype. This was confirmed by studying goosepimple mutants, which carry a mutation in myosin Vb, involved in peridermal plasma membrane homeostasis (32) (SI Appendix, Fig. S4 C and D).

In conclusion, tooth germs form from the endodermal layer but only after periderm-like cells have covered this endoderm.

Signaling Pathways Known so Far to Affect Tooth Development Do Not Interfere with Pouch 6 Contact or Presence of Periderm-Like Cells.

The above findings hint at the potential importance of endoderm–ectoderm contacts and presence of periderm-like, midline cells in tooth formation. Thus, we next tested whether manipulating signaling pathways, some of which are known to perturb zebrafish tooth development, also affect either pouch 6 contact, the presence of midline cells, or any combination. Because we previously established that these midline cells blend with periderm invading via pouch 2, we included observations on pouch 2 in our analysis, again using serial sections and scoring each body side separately not to miss potential asymmetries. We assigned pouch number only as far as it could be identified in case of highly deformed pouches. For this analysis, we used mutants targeting particular signaling pathways, and/or pharmaceutically inhibited the pathway (SI Appendix, Table S2).

Because the close cellular contact between midline cells and tooth-forming endodermal epithelium suggested potential juxtacrine signaling, we started by investigating the role of Delta-Notch signaling. DAPT (N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester), a known and carefully characterized γ-secretase inhibitor (33), severely interferes with Notch signaling in zebrafish embryos. DAPT was applied at 40 or 44 hpf, i.e., when midline cells have not yet expanded far enough posteriorly as to cover the endoderm of the future tooth-forming region. Controls and treated specimens were killed at 72 hpf. In controls of this age, pouch 2 is open (Fig. 3D), pouch 6 makes contact with the ectoderm (Fig. 3E), but is not open yet, and a layer of periderm-like, krt4-positive, midline cells is present, as are tooth germs (Fig. 3F). Continuous application of DAPT until 72 hpf did not prevent expansion of the periderm-like cells, pouch 6 contact, or tooth formation (n = 6/6) (Fig. 3 G–I). mind bomb (mib) is a ubiquitin ligase that is essential for efficient activation of Notch signaling by Delta (34). In mib mutants, like in controls, pouch 2 was observed to be open, pouch 6 made contact with the ectoderm, teeth were present, as was a layer of midline cells (n = 6/6) (Fig. 3 J–O) (SI Appendix, Table S2A).

Hedgehog signaling is important at multiple stages of tooth development in zebrafish (35). Thus, we investigated the effect of impeding hedgehog signaling on spreading of periderm-like cells over the midline endoderm, and on tooth formation. Inhibiting hedgehog signaling through cyclopamine A (CyA) from 40 hpf onwards until killing at 96 hpf did not prevent spreading of midline cells until the posterior pharynx. Yet, despite the presence of midline cells in the tooth-forming region, and despite pouch 6 contact, teeth did not develop, consistent with findings reported in ref. 35 (Fig. 4 A–F) (SI Appendix, Table S2B).

Fig. 4.

Defects in signaling and absence of teeth coincide with absence of a periderm-like layer. (A–C) WT embryo at 96 hpf. Pouch 2 is open (A), pouch 6 is opening (B), midline cells cover the endoderm and tooth germs are present (C, thick arrows). (D–F) WT embryo at 96 hpf treated with cyclopamine A to inhibit the Shh pathway. Note open pouch 2 (D), pouch 6 contact (E), presence of midline cells covering the endoderm but clear absence of tooth germs (F, thick arrows). (G–I) WT embryo at 80 hpf, showing an open pouch 2 (G), yet a still closed pouch 6 (H), presence of midline cells (I, black arrowhead) covering the endoderm (outlined by a dotted line in H), and tooth germs clearly present (I, thick arrows). (J–L) WT embryo at 80 hpf, treated with 25 µM SU5402 between 40 and 80 hpf, imaged at levels corresponding to those shown in G–I. Note open pouch 2 (J), closed pouch 6 (K), presence of midline cells (K, black arrowhead) covering the endoderm (outlined by a dotted line), and clear presence of tooth germs (L, thick arrows), similar to control animal. (M–O) WT embryo at 72 hpf, showing an open pouch 2 (M), yet a still closed pouch 6 (N), presence of midline cells (N, black arrowhead) covering the endoderm (outlined by a dotted line in N) and tooth germs clearly present (O, thick arrows). (P–R) WT embryo at 72 hpf, treated with 25 µM SU5402 between 40 and 72 hpf, imaged at levels corresponding to those shown in G–I. Note open pouch 2 (P), closed pouch 6 (Q), presence of midline cells (Q, black arrowhead) covering the endoderm (outlined by a dotted line in Q), and absence of tooth germs (R, thick arrows). Cartoons show representative postembryonic developmental stages and level of transverse sections shown in the different figures (boxed areas). Dark gray: cartilage; medium gray: endoderm; light gray: other tissues. Abbreviations: asterisk: pharyngeal lumen; b: brain; CB5: ceratobranchial 5, e: eye; nt: notochord; ov: otic vesicle; P2: pouch 2; P6: pouch 6; >P6: level posterior to pouch 6; y: yolk. (Scale bars, 50 µm.)

Next to Shh signaling, Fgf signaling has been shown to be essential for tooth formation (25). We performed different experiments whereby SU5402 was applied starting either from 30 or from 40 hpf onwards (i.e., before, respectively after fully established contact between pouch 6 and ectoderm), until killing at 56, 72, or 80 hpf. Again, we scored each side separately. Whereas controls displayed normal tooth development associated with the presence of at least two epithelial tiers, pouch 6 contact and an open pouch 2 (25 embryos, n = 50/50) (Fig. 4 G–I), the treated 30 embryos showed a variable, yet symmetric outcome. In about half of the embryos treated from 30 hpf onwards (n = 16/30), and about half of the embryos treated from 40 hpf onwards (n = 14/30), tooth germs were present (albeit delayed), as was a layer of midline cells (Fig. 4 J–L). In the other half of the embryos treated from 30 or 40 hpf, tooth germs were missing, with the layer of midline cells either absent (n = 10/30), or frequently also present (n = 20/30) (Fig. 4 M–R) (SI Appendix, Table S2C). In all 30 embryos, pouch 2 was normally developed and pouch 6 contact was present on both sides.

Recently, FoxF genes have been shown to have an essential role in tooth formation, as triple FoxF mutants (foxf1^−/−;foxf2a^−/−;foxf2b^−/−) are edentulous (36). Close examination nevertheless revealed an open pouch 2, pouch 6 contact, and presence of midline cells (SI Appendix, Fig. S4 G and H and Table S2D).

Zebrafish defective in eda/edar signaling display perturbed tooth development (9). We examined serial sections of 10 finless (edar) mutants; in seven embryos older than 48 hpf, scored for each side separately (n = 14), tooth germs were present, along with pouch 6 contact and presence of midline cells. Teeth were not necessarily in full numbers, their presence indicating nevertheless that tooth initiation is not fully impeded (SI Appendix, Fig. S4 E and F and Table S2E).

Finally, we reexamined masterblind (mbl) mutants for the status of the pouches and the presence or absence of midline cells. mbl encodes for axin1, part of the beta-catenin destruction complex, and thus these mutants mimic overactivation of Wnt signaling. These mutants were previously reported to have a normal tooth phenotype (37) (SI Appendix, Table S2F). In all 11 specimens studied and scored for each side (n = 22), pouches 2 and 6 presented a normal contact with ectoderm and midline cells were present.

Together, these data indicate that signaling pathways known so far to affect tooth development do not interfere with pouch 6 contact or presence of periderm-like midline cells.

Pouch 6 Contact with Ectoderm and the Presence of Periderm-Like Midline Cells Are Required but Not Sufficient for Tooth Formation.

To assess the role of pouch contact with ectoderm and the presence of periderm-like cells in tooth initiation, we pooled all of the data on mutant, transgenic, and experimental animals (SI Appendix, Fig. S7 and Table S3). We did not consider embryos younger than 48 hpf because tooth formation in WT becomes morphologically discernible only from 48 hpf onwards. Taking all data together, it can be inferred that teeth are sometimes present in the absence of pouch 2, while its contact with ectoderm, or even opening, is not sufficient to allow tooth formation, suggesting that pouch 2 has no causal role in tooth initiation (SI Appendix, Fig. S7 and Table S3C). There are two questionable exceptions (2/30) (discussed further) to the observation that pouch 6 must contact the ectoderm in order for teeth to be present. Yet, as for pouch 2, the contact is not sufficient for tooth initiation, as around 30% of the body sides examined display pouch 6 contact without teeth. The most stringent requirement appears to be the presence of periderm-like cells; in not one single case were tooth germs observed in the absence of such cells. Finally, we scored the absence or presence of teeth if both pouch 6 contact and the presence of midline cells are considered. With the same two exceptions, it can be inferred that both periderm-like cells and pouch 6 contact must be present if teeth are to be found, but that they are not sufficient for tooth initiation.

Discussion

Our findings show that 1) the epithelial component of pharyngeal teeth in zebrafish (the enamel organ) is derived from medial endoderm just posterior to pouch 6, yet, 2) dental morphogenesis starts only after pouch 6 has made contact with skin ectoderm, and 3) only after a layer of periderm-like cells has covered the prospective odontogenic epithelium (SI Appendix, Fig. S8), and finally, 4) interfering with signaling pathways known to affect tooth development supports the conclusion that pouch 6 contact and the presence of midline cells are required, but not alone sufficient, for tooth initiation.

The Enamel Organ of Pharyngeal Teeth Derives from Endoderm.

That the epithelial component of pharyngeal teeth in teleosts derives from endoderm has long been assumed based on circumstantial evidence, such as their position on the posteriormost pharyngeal arches, but conclusive evidence was missing so far. A reconsideration of the outside in theory for the evolutionary origin of teeth had furthermore fueled the idea that ectoderm may invade the pouches and occupy the tooth-forming region (12, 13). This had cast doubt on the endodermal nature of the enamel organ. Here we show that only the layer adjoining the basal membrane, i.e., the sox17-positive endodermal layer, undergoes morphogenesis to produce the early enamel organ of pharyngeal teeth. The absence, in zebrafish, of teeth on the mandibular arch precludes any conclusion on the germ layer contributing to oral teeth in teleosts. In urodele amphibians, the enamel organ can be formed in endoderm, ectoderm, or both (reviewed in refs. 38, 39). In contrast, in mammals it has now been convincingly shown that all teeth derive from ectodermal epithelium (11).

Pouch–Ectoderm Contact Is Required for Pharyngeal Tooth Formation.

Teeth start to develop morphologically only after pouch 6 has made contact with the skin ectoderm (two questionable exceptions out of 216 body sides examined), yet before the pouch opens into a gill slit. Likewise, in medaka (Oryzias latipes), pharyngeal teeth start to form first on pharyngobranchial 4 at stage 29 (40, 41), which is after pouch 5–ectoderm contact is established (42). Malformation of endodermal pouches in zebrafish prdm1a mutants is accompanied by the loss of pharyngeal teeth (43). Together these observations suggest that some sort of signaling may occur at the pouch–cleft contact to allow tooth initiation from midline endoderm. Signaling between endoderm and ectoderm is required for the development of several organs. For example, defective development of the thymus in nude mice is because third pouch endoderm no longer contacts the ectoderm, thus depriving it from its normal inducing agent (44). Both in the chick (45) and in zebrafish (46), neurogenesis in epibranchial placodes requires a signal from pouch endoderm. In zebrafish vgo mutants studied here, teeth were frequently absent despite the presence of an endodermal–ectodermal contact (n = 16/56), but in those cases, the contact zone could be established as being narrower than 16 µm, i.e., less than two cell diameters (i.e., the width of a pouch). Interestingly, Holzschuh et al. (46) likewise found a close correlation between enlarged or reduced contacts made by pouches with the ectoderm and epibranchial defects in vgo mutants.

One could raise the argument that the contact between pouch 6 and the ectoderm on the one hand, and the start of tooth formation on the other hand, is purely coincidental, and merely a reflection of overall sufficient maturation of endoderm. However, observations on the vgo^−/− mutants argue against this interpretation, as some specimens have teeth on just one body side, i.e., the side where pouch 6 makes contact with ectoderm. Asymmetric pouch development is also observed in ace (fgf8) mutants, yet teeth develop symmetrically, albeit that they are malformed, smaller, and less mature (20, 25, 47). However, it is not clear from these studies whether defective pouches actually fail to contact the ectoderm.

The nature of the putative inductive signal mediating endoderm–ectoderm interactions remains elusive. Genes important in zebrafish pouch outgrowth and cellular rearrangement into a bilayer are tbx1, fgf8a, and wnt4a, yet none of the mutants for these genes lacks teeth (this study and ref. 25, 27). Different bmps (Bmp7 in the chick, bmp2b and bmp5 in zebrafish), as well as fgf3, have been identified as crucial components of the endodermal signals that induce epibranchial neurogenesis (45, 46, 48). Knockdown in zebrafish of bmp2b and bmp4 does not cause tooth absence (49). Likewise, mutants for other genes expressed at pouch–cleft contacts and candidates for mediating ectodermal–endodermal signaling, such as eya1 (50) or pax9 (51), have not yielded a tooth phenotype so far. The absence in zebrafish of a tooth phenotype both in eya1 mutants and pax9 morphants is remarkable also because of the high conservation of the Pax-Six-Eya regulatory network in vertebrate pharyngeal pouches (16).

One, largely unexplored, possibility is that tissue mechanics, together with molecular effectors, coordinate morphogenesis (52). Thus, mechanical cues issuing from pouch–cleft contact, rather than a molecular prepattern, may promote morphogenesis of the basal endodermal layer to form the enamel organ. Our data on failure of tooth initiation in the case of mechanically perturbed pouch–cleft contacts support this idea. Thus, while Fgf and Shh signaling pathways are crucial for tooth initiation, they could act downstream of pouch 6 contact (as well as of appearance of midline cells, see below).

Pharyngeal Tooth Initiation Requires a Periderm-Like Cell Layer Covering the Presumptive Enamel Organ.

At the time tooth morphogenesis starts (48 hpf), pouch 6 contact is still closed. The tight bilayer nature of the pouch prevents any precocious invasion of ectoderm or periderm into the pouch, similar to conclusions for medaka (42). However, recently we have demonstrated that a group of krt4-positive, periderm-like cells appears at the level of pouch 2 at around 26 hpf, and that this population expands throughout the pharynx along the midline, thereby covering the endoderm (23). These cells, here called midline cells, connect to the superficial skin layer, the periderm, first via pouch 2, later via more posterior pouches. The remarkable correlation between the presence of midline cells and the presence of teeth (no exceptions out of 216 sides), strongly suggests that these midline cells are required for initiation of tooth formation. Again, one could argue that the presence of a periderm-like layer and the start of tooth formation is coincidental but not causally linked. While we have no other argument than correlative evidence, it is useful to point out that we know of no instances where vertebrate teeth are initiated from a nonstratified epithelium. It may well be that enamel organs simply cannot form from a monolayer and that the periderm-like cells may play a mechanical role similar to that of the suprabasal canopy in mammalian tooth development (53). In mice, Fgf8-expressing cells migrate toward the molar placode’s Shh-expressing cells to initiate tooth development (54). It is interesting to note that in zebrafish, two Fgf receptors, fgfr1a and fgfr2, are expressed in pharyngeal endoderm (55). Knockout of three Fgf receptors, fgfr1a, fgfr1b, and fgfr2, apparently leads to tooth absence (figure 6C in ref. 56), consistent with results reported in ref. 25 and here on the use of SU5402. That endoderm morphogenesis is essential for normal craniofacial development has been firmly established in numerous papers (57); yet, our data warrant an investigation into whether the midline cells, in addition to the endoderm, play a crucial role. The intimate contact between the apical periderm-like layer and the basal endodermal layer, as well as the long cell processes issuing from the former and contacting the basal lamina (23), provide a further incentive to do so. In the mammalian oral cavity, a periderm is present, superficial to the tooth germs (58). Different from zebrafish, this periderm is derived from ectoderm. Its role in mammalian tooth initiation has not properly been investigated, although its reported timing of appearance (59) clearly warrants an investigation into this question.

Pharyngeal Teeth from an Evolutionary Developmental Biology Perspective.

The primary incentive for the present study was to test whether a contribution from the ectoderm (cellular or via signaling) is still required for pharyngeal teeth to develop, as proposed in the modified outside-in hypothesis (12–14). Atukorala et al. (42) tested this hypothesis on medaka. They saw no evidence of any ectodermal contribution to pharyngeal tooth formation and concluded for an intrinsic odontogenic competence of the rostral endoderm. However, we have demonstrated, both in ref. 23 and in the present paper, that cells with periderm-like characteristics, different from endoderm, come to overlie the endoderm and appear to be required to start the process of tooth formation. Importantly, in medaka the endoderm is covered by hatching gland cells, itself overlain by flattened cells reminiscent of the midline cells observed in zebrafish (60). Other teleost species also possess a layer of flattened periderm-like cells lining the pharynx (23). The decisive experiment to demonstrate a key role for the periderm-like layer will be a selective ablation of these cells.

The developmental origin of the midline cells still needs to be elucidated but they have been termed periderm-like because of shared morphology and markers, and compatible adhesion with periderm (23). The periderm itself is the first epithelial covering of the embryo. In teleosts the periderm derives from the enveloping layer (EVL) that surrounds the embryo during gastrulation. The EVL segregates from the blastoderm and becomes restricted to a peridermal fate by 4 h, at late blastula stage (61). Still, homology of the EVL across taxa is not clearly established (62). However, periderm (upper layer) and ectoderm (lower layer) together constitute the bilayered epidermis of the skin of teleost embryos and posthatching stages, and the periderm persists well into juvenile life (63, 64). Given that early embryonic development is to a large extent determined by the amount of yolk (65), it will be interesting to study the periderm in representatives of basal, nonteleost actinopterygian lineages, instead of the telolecithal eggs of zebrafish.

Since the periderm in zebrafish partially invades the mouth cavity (23), the absence of oral teeth cannot be attributed to a lack of such cells. This is in line with conclusions reached in ref. 66 that changes in transacting regulators of Dlx genes were responsible for cypriniform tooth loss. On the other hand, the periderm-like layer expands into the esophagus, but not further caudally (23). Based on our current findings, we predict that teeth can be present in the digestive system but not beyond the esophagus. Interestingly, several fish species indeed possess esophageal teeth (67), while we know of no teleosts with teeth in the stomach or beyond.

Conclusion

Summarizing the data from mutant and transgenic zebrafish and from our experimental approaches, we conclude that there is a double requirement before pharyngeal teeth are initiated. These are contact of the endodermal pouch with skin ectoderm and the presence of a periderm-like layer covering the endoderm. Both are necessary, but not sufficient, for tooth formation. We find that pouch–cleft contacts are necessary, although not to allow an influx of ectodermal (or peridermal) cells. Pouch–cleft contacts may act as signaling centers, as demonstrated for other vertebrate pharyngeal derivatives. Alternatively, they may deliver mechanical cues to coordinate endoderm morphogenesis. Likewise, the need for a layer of cells covering the endoderm, with periderm-like features, is strongly underscored by observations on more than a hundred embryos, deriving from various mutant and transgenic lines and experiments. The nature of the signal issuing from the midline cells and allowing initiation of tooth placode formation in the endoderm remains elusive. Our results are nevertheless in line with the findings for mouse molar teeth, where two spatially distinct cell populations, one expressing Shh, the other Fgf8, are involved in tooth initiation (54). Whether the spatial expansion of periderm-like cells can be likened to the intraepithelial migration found in the mouse, still needs to be assessed. Nevertheless, it is clear that both in zebrafish and in the mouse, the earliest interactions necessary to produce a tooth include those between different epithelial populations, in addition to the known epithelial–mesenchymal interactions that govern all vertebrate teeth (8). This suggests a high level of conservation of tooth development, whether oral or pharyngeal.

Materials and Methods

Transgenic Zebrafish Lines.

Tg(sox17:egfp) zebrafish (68), in which the endoderm is labeled, were obtained from the laboratory of R. Opitz, Vrije Universiteit Brussel (VUB), Brussels, Belgium. Tg(krt4:gfp) (69) and Tg(krt4:tomatoCAAX) zebrafish, in which the outer skin layer, the periderm, is GFP- respectively tomato-positive, were a gift from M. Hammerschmidt, University of Köln, Köln, Germany. Tg(fli1:gfp) were obtained from A. Willaert, Ghent University Hospital, Ghent, Belgium. Adult fish were maintained and spawned according to ref. 70. Embryos were raised in egg water at 28.5 °C and staged according to ref. 18.

Mutant Zebrafish Embryos.

We selected a slate of mutant zebrafish, chosen to cover a broad range of defects in the pharyngeal epithelium (SI Appendix, Supplementary Material and Methods). All mutants, along with their wild type (WT) siblings, were obtained as embryos, fixed either in a mixture of 1.5% paraformaldehyde (PFA) and 1.5% glutaraldehyde in 0.1 M cacodylate buffer (a mixture abbreviated as PG), or in 4% PFA, and processed for embedding in epon or glycol methacrylate (GMA), as described below.

Pharmaceutical Inhibition.

Inhibition experiments were conducted for three different signaling pathways, as described in SI Appendix, Supplementary Material and Methods. The role of Notch signaling was investigated using DAPT (33). The role of Shh signaling was examined through the use of CyA, an established hedgehog pathway antagonist, known to affect pharyngeal tooth development (35). Fgf dependency was tested by inhibition with the generalized Fgf inhibitor SU5402 (25).

Mechanical Perturbation of Pharynx Development.

Zebrafish embryos that had developed pericardial or yolk sac edema, either provoked by mutation (laf^−/−; vgo^−/−), or as an occasional phenotype in transgenic lines [Tg(fli1:gfp) and Tg(krt4:gfp)], were considered separate from nonaffected siblings. Edemas were also artificially induced in Tg(sox17: egfp) embryos by ethanol treatment (for details, see SI Appendix, Supplementary Material and Methods).

Immunohistochemistry.

Immunohistochemistry for laminin on whole mount embryos was performed as described in ref. 71 using a polyclonal rabbit anti-laminin primary antibody (Sigma-Aldrich, L9393) and a goat anti-rabbit secondary antibody (DyLight 488 nm, Abcam). After immunohistochemistry, embryos were processed for embedding in GMA.

Histology and Transmission Electron Microscopy.

High-resolution histology (light and transmission electron microscopy [TEM]) was achieved by fixation in PG and embedding in epon as described previously (72) (for details, see SI Appendix, Supplementary Material and Methods). To visualize GFP on sections, embryos were fixed in 4% PFA. The GMA embedding protocol followed (73), a technique that preserves GFP signals and requires no extra staining (SI Appendix, Supplementary Material and Methods).

Observations and Microphotography.

Observations of epon or GMA sections were done on a Zeiss Axio Imager Z1 (https://www.zeiss.com/corporate/int/home.html). Photomicrographs were taken with a Zeiss Axiocam 503 camera and processed using ZEN software (Zeiss, https://www.zeiss.com/corporate/int/home.html). All photomicrographs were manually aligned so as to be able to count and identify individual pouches. Computer-generated images were processed for color balance, contrast, and brightness only, and applied to all parts of the figures equally.

Ethical Statement.

Animal care, experimentation, and killing complied with European Directive 2010/63/EU of September 22, 2010. All animal procedures used in this study were carried out under the laboratory permit number LA 1400452.

Data Availability.

All sections used for this study are kept in the slide collection of the Research Group “Evolutionary Developmental Biology” at the Biology Department of Ghent University, and are available for inspection upon request.