Late-emigrating trunk neural crest cells in turtle embryos generate an osteogenic ectomesenchyme in the plastron



Background: The turtle plastron is composed of a keratinized epidermis overlying nine dermal bones. Its developmental origin has been controversial; recent evidence suggests that the plastral bones derive from trunk neural crest cells (NCCs). Results: This study extends the observations that there is a turtle-specific, second wave of trunk NCC delamination and migration, after the original NCCs have reached their destination and differentiated. This second wave was confirmed by immunohistochemistry in whole-mounts and serial sections, by injecting DiI (1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate) into the lumen of the neural tube and tracing labeled cells into the plastron, and by isolating neural tubes from older turtle embryos and observing delaminating NCCs. This later migration gives rise to a plastral ectomesenchyme that expresses NCC markers and can be induced to initiate bone formation. Conclusions: The NCCs of this second migration have properties similar to those of the earlier NCCs, but also express markers characteristic of cranial NCCs. The majority of the cells of the plastron mesenchyme express neural crest markers, and have osteogenic differentiation capabilities that are similar or identical to craniofacial ectomesenchyme. Our evidence supports the contention that turtle plastron bones are derived from a late emigrating population of cells derived from the trunk neural crest. Developmental Dynamics 242:1223–1235, 2013. © 2013 Wiley Periodicals, Inc.


The formation of turtle plastron bones remains controversial. The plastron consists of a plate of ventrally located dermal bones covered by keratinous ectodermal scutes (Zangerl, 1969). This structure is a synapomorphic character of the Order Testudines, although it may share common ancestry with the gastralia found in other reptile clades (Zangerl, 1939; Claessens, 2004). Each of the nine plastral bones of hard-shelled turtles forms by intramembranous ossification from a separate center, and the bony spicules radiating from these centers eventually interact with one another to form sutures (Rathke, 1848; Zangerl, 1939; Gilbert et al., 2001; Cebra-Thomas et al., 2007). The anteriormost plastron bones, the paired epiplastra and the central entoplastron between them, suture together shortly after their formation to make a bony anterior plate. The other plastral bones are paired laterally. The hyoplastron constitutes the plate that covers the mid-anterior portion of the trunk, as well as forming the anterior part of the bony bridge that unites the plastron with the carapace. In their posterior regions, the two hyoplastal bones remain unfused and form the rostral portion of the umbilical fontanel. The hypoplastron bones form posteriorly to them, creating the caudal portions of the bridge and the fontanel. Caudally, the paired xiphiplastron bones fuse together (Zangerl, 1969; Clark et al., 2001; Gilbert et al., 2001). The development of the plastron may be a key step in the evolution of the shell; Odontochelys, a fossil turtle from the late Triassic (220 mya), had a fully developed plastron, even though its carapace was only partially formed (Li et al., 2008).

Homologies of the plastron bones to any mammalian bone remain uncertain (see Hall, 2005; Vickaryous and Hall, 2006). It is generally thought that the anterior plastral bones, the epiplastra and entoplastron, are homologous, respectively, to the clavicles and interclavicle bones of other reptilian lineages (Parker, 1868; Gegenbaur, 1898; Zangerl, 1969; Rieppel, 1996; Cherepanov, 1997; Vickaryous and Hall, 2006, 2010), while the more posterior plastral bones are homologous to the gastralia (“floating ribs,” “abdominal ribs”) of other tetrapods (Zangerl, 1939; Claessens, 2004; Lyson et al., 2010).

Bones can form from paraxial mesoderm (axial skeleton), lateral plate mesoderm (limbs), and neural crest ectoderm (head; Hall, 2005). The cranial neural crest is responsible for forming the facial bones and parts of the calvareal skeleton (Noden, 1975; Gans and Northcutt, 1983; Lumsden et al., 1991; Serbedzija et al., 1992; Trainor et al., 2003; Yoshida et al., 2008; Lee and Saint-Jeannet, 2011; Hu and Marcucio, 2012; Koyabu et al., 2012). The prechordal cranium (the neurocranium rostral to the notochord) appears to be made exclusively from the cranial neural crest, rather than being generated by the head mesoderm or from a mixed population of crest and head mesoderm cells (Couly et al., 1993; McBratney-Owen et al., 2008; Wada et al., 2011). Intramembranous ossification and suture formation is characteristic of the bones forming the face and skull (see Noden, 1991; Hall, 2005; Rice and Rice, 2008).

Our laboratory has accumulated evidence that the plastron bones, like the facial bones and portions of the skull, are derived from neural crest cells (NCCs). First, the plastron bones of hatchling turtles form by intramembranous ossification within the presumptive dermis (Gilbert et al., 2001), and the cells forming these bones express the HNK-1 epitope and platelet-derived growth factor receptor-alpha (PDGFRα), two markers, which in combination, mark skeletogenic NCCs (Clark et al., 2001). When turtle embryos were immunostained to look for the precursors of the plastron bones, a thick band of HNK-1+ mesenchymal cells was observed in the carapacial dermis of Greenbaum (2002) stages 17 and 18 (G17, G18; Cebra-Thomas et al., 2007). These stages are more than a week removed from those comparable to the embryonic stages where trunk neural crest emigration occurs in chicken embryos. By G17, dorsal root ganglia, sympathetic ganglia, and enteric neurons can be seen to have differentiated from the original wave of NCCs. This thick aggregation of HNK-1+ cells has not been reported in any other vertebrate embryo and can be interpreted as a “staging area” for NCCs. An antibody to the melanoblast-lineage marker C5 (Nishimura et al., 2005) stained cells adjacent to the ectoderm but not in this carapacial staging area, indicating that these HNK-1+ cells were not melanocyte precursors. Furthermore, HNK-1+ cells were found above the dorsal neural tube, dispersed in the dermis in the lateral regions of the embryo, and in aggregates in the ventral region, where they are seen surrounding the nascent plastron bones (Cebra-Thomas et al., 2007).

While the CD-57 epitope recognized by antibody HNK-1 is a commonly used marker for amniote NCCs, including those of the turtle (Tucker et al., 1984; Rickman et al., 1985; Bronner-Fraser, 1986; Erickson et al., 1989; Sadaghiani and Vielkind, 1990; Hou and Takeuchi, 1994; Jeffery et al., 2004; Hou, 1999), this epitope is also expressed on subsets of neurons and leukocytes (Tucker et al., 1984; Erickson et al., 1989; Chou et al., 2002). Similarly, PDGFRα is detected not only on skeletogenic NCCs, but also on rib precursors and the embryonic mesenchymal cells contributing to bone, hair, mammary gland, gut, and lung (Orr-Urtreger and Lonai, 1992; Betsholtz et al., 2001; Hoch and Soriano, 2003). To confirm that the dispersed HNK-1+ cells in the carapacial dermis are NCCs, we examined the expression of additional NCC cell markers. We found (Cebra-Thomas et al., 2007) that the cells within the carapacial staging area were also positively stained by antibodies against the low affinity neurotrophin receptor p75 (Rao and Anderson, 1997; Abzhanov et al., 2003; Takaki et al., 2006; Betters et al., 2010) and by antibodies against FoxD3, a transcription factor necessary for specifying NCCs and regulating their cell adhesion proteins (Labosky and Kaestner, 1998; Kos et al., 2001; Cheung et al., 2005; Barembaum and Bronner-Fraser, 2005; Tompers et al., 2005). The distribution of these FoxD3+ cells coincided with that of the HNK-1+ cells. In addition, FoxD3+ cells were located in the dorsal neural tube, suggesting that NCCs were still delaminating and emigrating. Moreover, because FoxD3 represses the melanoblast lineage specifying genes in chick NCCs (Kos et al., 2001; Thomas and Erickson, 2009), its expression further demonstrated that the HNK-1+ cells of the carapacial staging area were not premelanocytes. Therefore, these dorsal FoxD3+, p75+, HNK-1+, PDGFRα+ cells appear to represent a late-emerging skeletogenic population of trunk NCCs that are not found in chicks or mice (Cebra-Thomas et al., 2007).

However, trunk NCCs of model amniotes are not skeletogenic (Noden, 1975; Nakamura and Ayer-Le Lièvre, 1982; Hall, 2010), and the proposition that trunk NCCs from the relatively late turtle embryo form the plastral skeleton remains disputed (see Penissi, 2004; Kuratani et al., 2011). This study presents further evidence for the existence of a late-stage population of trunk NCCs in turtle embryos, and it demonstrates that these cells migrate to the ventral region of the turtle embryo and form a bone-generating ectomesenchyme. To show this turtle-specific wave of osteogenic NCC delamination from the trunk neural tube, we devised four separate lines of inquiry. First, we looked for evidence that neural crest migration was still occurring from turtle neural tubes long after the differentiation of sympathetic, sensory, and enteric neurons derived from the earlier set of NCC migrations. We used whole-mount and section immunohistochemistry to further document this NCC emigration from late-stage turtle neural tubes.

Second, if NCCs were still being produced in late-stage turtle embryos, then cells with neural crest characteristics should be seen migrating from isolated neural tubes at both early and later stages of turtle development. We used in vitro cell migration assays to demonstrate migratory cells with neural crest cell properties emigrating from isolated late-stage turtle neural tubes.

Third, if this NCC population were contributing to the formation of the plastron bones, then cells originating in the late neural tube should be traceable to the ventral side of the embryo. By injecting a lipophilic dye (DiI, 1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate) into the lumen of the embryonic neural tube such that it would only label neural tube cells, we could determine if a population of cells from the neural tube had migrated into the ventral regions of the embryo during the next days. Indeed, we found that labeled cells were later observed outside the neural tube, including in the mesenchyme of the ventrum, suggesting that these cells had once been part of the neural tube.

Last, if the plastron bones were formed by NCCs, we would expect to see cells having ectomesenchymal markers in the ventral trunk dermis, similar to those markers of the craniofacial ectomesenchyme. We therefore looked for molecular signatures of cranial NCCs in the ventral mesenchyme. Plastron mesenchyme cells expressed not only PDGFRα, but also noelin−1/2 (Moreno and Bronner-Fraser, 2002) and Id-2 (Martinson and Bronner-Fraser, 1998). In chick embryos, Id-2 is expressed in the cranial neural crest and its migrating derivatives, and it appears when trunk NCCs are experimentally induced to form skeletogenic cells (Abzhanov et al., 2003; Ido and Ito, 2006).

This study proposes (1) that there is a late-emerging population of trunk NCCs in the turtle embryo, (2) that this late-emerging population of NCCs migrates ventrally to the plastron region, forming an osteogenic ectomesenchyme similar to that of the face, and (3) that these neural crest-derived cells form the plastron bones.


Confirmation of the Existence of a Second Period of NCC Emigration in the Turtle Embryo

Immunohistochemical staining of turtle embryos during sequential stages of development allowed us to see a biphasic pattern of HNK-1+ cells in the dermis adjacent to the trunk neural tube (Fig. 1). HNK-1+ cells emigrated from the dorsal region of the turtle trunk neural tube at G12–13, comparable to the developmental stages when NCCs emerge from the trunk neural tube in other amniote embryos, as previously shown (Hou and Takeuchi, 1994). As expected, these cells were found to contribute to traditional trunk neural crest-derived structures. By G15, mature derivatives such as the spinal dorsal root ganglia were clearly stained, but no single-cell staining was observed above the neural tube (Fig. 1A). In contrast, by G16 broad bands of staining could be observed along the dorsal midline (arrowheads, Fig. 1B,C). Similarly, when transverse sections were examined, neural crest derivatives such as the dorsal root ganglia were evident in HNK-1-stained sections at G15, but no stained single cells were observed below the dorsal ectoderm (Fig. 1D). By G16, however, HNK-1+ cells were seen in the newly thickened carapacial mesenchyme above the neural tube in sectioned embryos (Fig. 1E). By G17, large groups of HNK-1 cells were seen clustering in this carapacial staging area, forming a darkly staining band above the neural tube (Fig. 1F,K,L). To aid in comparison, all sections examined were from the anterior trunk at the level of the lungs. At G18, the HNK-1+ cells formed a broad band throughout the dorsal mesenchyme (Fig. 1G). At this stage, HNK-1+ cells were also seen in the lateral regions of the carapacial mesenchyme, and in the nascent plastron, forming clusters that presage the ossification centers (Cebra-Thomas et al., 2007). By G19, very few HNK-1+ cells were seen in the carapacial mesenchyme (Fig. 1H), and at G20 and 21, except for a few scattered HNK-1+ cells associated with the epidermal folds formed by the scute placodes, the carapacial mesenchyme cells were HNK-1 (Fig. 1I,J,M-O). Thus, the HNK-1+ cells of the carapacial staging area are a transient population, emerging at or near the midline around stage 16, spreading laterally, and declining by stage 20.

Figure 1.

HNK-1+ cells form a transient population in the carapacial staging area above the neural tube. A: Whole-mount antibody staining of a G15 (Greenbaum stage) embryo with HNK-1; mature ganglia in the trunk and limbs are stained. B,C: Whole-mount antibody staining of a G16 and G16+ embryo with HNK-1 showing labeled cells migrating from the midline (arrowheads) in addition to mature ganglia in the trunk and limbs. Labeling is most apparent at the anterior and posterior ends of the carapace, where the mesenchyme is less thick, suggesting that penetration of the antibody may be an issue. Subsequent paraffin sectioning confirmed the presence of labeled single cells in the carapacial mesenchyme (data not shown). D: Paraffin section of G15 T. scripta embryo stained with HNK-1, showing the neural tube and the dorsal root ganglia derived from early trunk neural crest cells (NCCs). No cells staining with HNK-1 are observed below the carapacial epidermis. E: Cells positive for HNK-1 expression are seen in the carapacial dermis above the neural tube at G16. F: Clustered cells in the carapacial staging area above the vertebra at G17 stain intensely with HNK-1. Staining is also apparent in the dorsal root ganglion. G: A broad band of HNK-1+ cells is observed in the carapacial dermis at G18. HNK-1+ cells begin to appear in the ventral mesenchyme (see Fig. 2). Staining is also apparent in the spinal ganglion. H: A few HNK-1+ cells remain in the carapacial dermis at G19. I,J: No HNK-1+ cells are observed above the neural tube in the carapacial dermis at G20 (I) and G21 (J). K: Low magnification of a section through the anterior trunk of a G17 embryo stained with HNK-1, but not counterstained (scale bar 300 μm). L: High magnification of the dorsal medial region of the section in (K) showing individual HNK-1+ cells above the neural tube (scale bar 40 μm). M: Low magnification through the anterior trunk of a G21 embryo stained with HNK-1, but not counterstained. N: High magnification of the dorsal medial region of the section in (M) showing the lack of HNK-1+ cells in the mesenchyme above the neural tube. O: A few scattered HNK-1+ cells remain in the lateral carapace, associated with the epidermal folds formed by the scute placodes N, notochord; nt, neural tube; drg, dorsal root ganglia; vrg, ventral root ganglia; V, vertebrae; lu, lung; liv, liver.

The plastron is first evident as a thickened mesenchyme below the ventral ectoderm beginning at G17, and by G19 it forms a complete shield over the ventral region, with a central fontanel for the vitelline blood vessels (Fig. 2A). Although there were few HNK-1+ cells remaining in the dorsal carapace at G19 (Fig. 1H), HNK-1+ cells were visible in the lateral and ventral carapace, in the bridge region and forming the newly thickened plastron mesenchyme (Fig. 2B–E). The plastron ossification centers initiate at G19, and there were corresponding HNK-1+ condensations in the plastron (Fig. 2F). Some HNK-1+ cells were observed surrounding the rib, which is consistent with previous studies showing that cranial NCCs will associate with trunk skeletal elements (Le Douarin and Teillet, 1974; Le Douarin, 1982; McGonnell and Graham, 2002). Therefore, there appears to be an extended period of NCC production where cells migrate initially dorsally into the carapacial mesenchyme, and then laterally and ventrally to the plastron region.

Figure 2.

Plastron formation on the ventral side of T. scripta embryos is accompanied by an influx of HNK-1+ cells. A,B: By G19 (Greenbaum stage) the thickened plastron is visible on the ventral side of the embryo (A) and in cross section (B). C,D: A few HNK-1+ cells remained in the carapacial dermis (C), but most were visible apparently migrating around the periphery of the embryo (D). Some HNK-1+ cells were also observed surrounding the rib cartilage (C). This pattern was not observed in comparably stained chick embryos (Cebra-Thomas et al., 2007). E,F: HNK-1+ cells were located in the plastron mesenchyme (E) and in the condensations (F) representing ossification sites (arrow). NT, neural tube; FL, forelimb; Cara, carapace; Plas, plastron; * marks cartilage elements of the pectoral girdle.

These observations were confirmed in vitro by culturing neural tubes. Trunk neural tubes were isolated from G12 (early; Fig. 3A) and G17 (late; Fig. 3F) T. scripta embryos, and cultured on fibronectin and poly-D-lysine. Cells positive for neural crest markers HNK-1(Fig. 3B,C,G,H) and p75 (Fig. 3D,I) were observed migrating away from both the early and late turtle trunk neural tubes after 2 (for G12) or 3 (for G17) days in culture. These cells showed the filopodia characteristic of migratory NCCs (Teddy and Kulesa, 2004). More than 100 turtle neural tubes of both stages were analyzed, and in all cases, both the early stage and late stage turtle neural tube cultures had cells migrating away from the length of the neural tube. Critically, the early- and late-stage turtle NCCs differed in the expression of PDGFRα. The late migrating cells stained positively for this protein, while the early stage NCCs did not (Fig. 3E,J). PDGFRα is a marker for cranial NCCs and is thought to be involved in the ability to respond to osteogenic signals (Tallquist and Soriano, 2003; Eberhart et al., 2008). PDGFRα has been previously shown to be expressed by the cells that condense to form the developing plastron bones (Clark et al., 2001). Control cultures, stained without primary antibodies, were negative (Fig. 3Q).

Figure 3.

Late emergence of cells expressing neural crest cell (NCC) markers from isolated turtle neural tubes. A–J: Neural tubes were isolated from G12 (Greenbaum stage; A) and G17 (F) T. scripta embryos, and cultured on fibronectin and poly-D-lysine-coated plates. Neural crest-like cells can be observed migrating away from the neural tubes after 2 (G12) or 3 days (G17) in culture and staining with HNK-1 (B,C,G,H), anti-p75 (D,I) or anti–platelet-derived growth factor receptor-alpha (anti-PDGFRα; E,J). K–M: Neural tubes isolated from HH12 (Hamburger and Hamilton stage) chick embryos (K) and cultured as above for 2 days and stained with HNK-1 (L,M). N–P: Neural tubes isolated from HH26 chick embryos (N) and cultured as above for 3 days. No neural crest cells are observed migrating away from isolated HH26 chick neural tubes (O,P). Q: Negative control staining of migratory cells from G17 neural tubes; negative control staining from G12 neural tubes was similar (data not shown). Scale bar = 100 μm in L,O; 30 μm in D,E,I,J,M,P.

For comparison, chicken neural tubes at the crest-migrating stages (HH [Hamburger and Hamilton, 1951] stage 12; equivalent to T. scripta G12) and later stages (HH stage 26; equivalent to T. scripta G17) were isolated and cultured (Fig. 3K,N). As expected, NCCs migrated away from the early chick trunk neural tube (Fig. 3L,M). However, no neural crest emigration was seen from neural tubes isolated from the later (HH 26) chick embryos (Fig. 3O,P).

DiI Labeling of Turtle Neural Tubes

DiI labeling also confirmed (1) that NCCs were still emerging from the neural tube of late-stage turtle embryos, and (2) that these cells migrated ventrally to the plastron region. Comparably staged chick and turtle embryos (HH stage 26 chicks; G stage 17–18 turtles) were prepared for dorsal trunk explant culture as previously described (Cebra-Thomas et al., 2005). DiI was injected into the lumen of the trunk neural tubes, which were photographed under epifluorescence immediately after injection (day 0), as well as after two or three days of incubation. Any visible “spreading” of the fluorescent dye away from the neural tube would represent migration of cells that were originally in contact with the neural tube lumen (del Barrio and Nieto, 2002). As predicted, no movement of labeled cells away from the later stage (HH26) chick neural tube laterally into the dermis was observed (Fig. 4A). However, in later-stage (G 17) turtle embryos, migration of labeled cells away from the trunk neural tube after two days of incubation was readily observed (Fig. 4B). These cells were found in thick bands above and on either side of the neural tube, far from the injection sites, and also more laterally in the mesenchyme between the ribs of the turtle. Migration of cells away from the neural tube was observed in all of the injected turtle embryos.

Figure 4.

Late emergence of DiI (1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate) -labeled cells from the turtle neural tube. A,B: Trunk explants of day 5 (HH26, Hamburger and Hamilton stage) chick embryo (A) and day 18 (G17, Greenbaum stage) T. scripta embryo (B) were injected with DiI into the lumen of the neural tube and incubated for 2 days to allow neural crest cell migration. Labeled cells are readily observed spreading laterally in the turtle (but not the chick) embryo. C: Intact G16 turtle embryo injected with DiI into the lumen of the neural tube and cultured for 2 days. D,G: DiI fluorescence of cryosections through the anterior trunk region of the embryo in (C) as indicated (bar = 300 μm). The majority of the DiI is confined to the lumen of the neural tube. E,F,H,I: Higher magnification of the ventral side of the sections in (D and G; indicated by boxed regions). DiI-labeled cells were observed in the mesenchyme between the epidermis and the thoracic cavity (scale bar = 50 μm). FL, forelimb; HL, hindlimb; nt, neural tube; H, heart; lu, lung.

While this explant culture technique enabled us to observe cells leaving the late turtle neural tube, the lack of ventral tissue in the cultured dorsa prevented the assessment of whether the DiI+ cells could reach the plastron region. To address this question, intact stage G16 turtle embryos were injected with DiI into the neural tube at two sites: at the level of the hindbrain and posterior to the hindlimbs. The embryos were cultured in suspension for 3 days, and viability was affirmed by continued trunk muscle contractions. Embryos where the lumen of the trunk neural tube filled with DiI were selected for cryosectioning (Fig. 4C). Sections were taken from regions between the two injection sites to insure that labeled cells had only been exposed to dye from within the lumen. In all cases, epifluorescence at day 3 confirmed that most of the dye was in the neural tube or in cells touching the neural tube (Fig. 4D,G). But in these same embryos, DiI+ cells were also located on the opposite side of the embryo, in the ventral mesenchyme (Fig. 4E,F,H,I). Therefore, cells leaving the trunk neural tube of late stage (G16) turtle embryos can migrate to the plastron.

Characterization of Plastron Mesenchyme

If these late emigrating NCCs reach the ventral region of the embryo and contribute to the plastron bones, then the plastral mesenchyme should be, at least in part, neural crest-derived. Plastron mesenchyme cells were taken from G19 turtle embryos (i.e., after the migration of the HNK-1+ cells from the carapacial staging area and their appearance in the ventrum. As fibronectin has been shown to provide an essential substrate for migratory NCCs (Rovasio et al., 1983), the isolated plastron mesenchyme cells were allowed to attach to fibronectin, and the adherent cells were stained for NCC markers. The fibronectin-adherent cells exhibited the morphology of migratory cells (Fig. 5A), and nearly 100% of them stained positively for neural crest markers HNK-1 and p75 NGF-receptor (Fig. 5B,C), and for fibroblast growth factor-2bek (FGFR2bek; Fig. E) which binds FGF2, an essential growth factor required for cranial NCC survival (Abzhanov et al., 2003). The plastral mesenchyme thus contains a significant population of neural crest-derived cells. In addition, most of the cells stained positively for PDGFRα (Fig. 5D), and some cells appeared to express low levels of procollagen Type I (Fig. 5F), which is associated with osteogenesis (Åberg et al., 2005). At the molecular level, G18 and G19 plastron mesenchyme cells expressed Id-2 and noelin-1 (Fig. 5G), which are characteristic of cranial, rather than trunk, NCCs. The levels of Id-2 were comparable to the levels in craniofacial mesenchyme, while noelin-1 was present at lower, but detectable levels.

Figure 5.

Plastron mesenchyme cells from G19 (Greenbaum stage) turtle embryos express neural crest cell markers. Plastrons were isolated, and the condensation centers and ectoderm were removed by manual dissection. The mesenchyme was dissociated by enzymatic digestion, and the resulting cells were plated overnight on fibronectin and poly-D-lysine-coated dishes. A–F: The attached cells (A,F) were stained as indicated for markers of neural crest cells (HNK-1 and p75 NGF-receptor; B,C), cranial skeletogenic neural crest (platelet-derived growth factor receptor-alpha [PDGFRα] and fibroblast growth factor-2 [FGFR2]; D,E), and osteoblast differentiation (procollagen I; F). Scale bar = 30 μm. G: Polymerase chain reaction analysis of craniofacial (CF), G18 and G19 plastron mesenchyme using gapdh, Id-2, and noelin-1 primers; (−), no cDNA control. Noelin-1 (*) was expressed in plastron mesenchyme at lower levels than in CF mesenchyme; Id-2 was expressed at comparable levels.

Moreover, the plastron mesenchyme also appears to have strong skeletogenic potential. When cultured in media containing ascorbic acid, β-glycerophosphate and dexamethasone, which had previously been shown to support bone formation in cultured cranial and trunk NCCs (McGonnell and Graham, 2002), the plastron mesenchyme cells divided rapidly, reached confluence, and aggregated into structures expressing high levels of procollagen type I while maintaining expression of HNK-1, p75, PDGFRα and FGFR2 (Fig. 6). These properties indicate osteoblast differentation. Many of the plastron mesenchymal cells stained positively with Fast Red (Fig. 6G), indicating alkaline phosphatase activity, which is also consistent with their becoming osteoblasts. There was only very weak expression of collagen II in the most heavily clustered regions (Fig. 6F). Thus, like the craniofacial ectomesenchyme, the plastral mesenchyme appears to be capable of readily differentiating along skeletogenic pathways.

Figure 6.

Plastral ectomesenchymal cells differentiate in culture along skeletogenic pathways. A–G: Fibronectin-adherent plastral mesenchyme cells (see Fig. 6) were cultured in osteoblast media for 1 week, and stained as indicated for markers of neural crest cells (HNK-1 and p75; A,B); cranial skeletogenic neural crest (PDGFRα and FGFR2; C,D); osteoblast differentiation (procollagen I and Fast Red reactivity; E,G) and chondrocyte differentiation (collagen II; F). H: Negative control staining.


We have previously shown that each of the intramembranously generated plastron bones (but none of the costal carapace bones) stained positively for HNK-1, supporting a neural crest origin for the plastron (Clark et al., 2001). Moreover, we identified a population of cells, bearing neural crest markers HNK-1, FoxD3, and p75, whose location is consistent with the hypothesis that they emerge from the dorsal neural tube of the turtle trunk, collect in the carapacial dermis, migrate ventrally, and aggregate to form the plastron bones (Cebra-Thomas et al., 2007). The nascent plastron bones are positive for these markers. Because there is no marker solely specific for NCCs (see Betters et al., 2010), and because transgenic lineage tracers are not yet available for turtle embryos, the “triangulation” of these three markers, coupled with the DiI labeling of the HNK-1+ cells of the carapacial staging area, has provided the major evidence for the trunk NCC origin of the plastron bones.

The present study finds further evidence that the plastron bones are generated from a late-migrating population of trunk NCCs. First, we show that the previously reported population of HNK-1+ cells in the dorsal mesenchyme are a transient population, and thus not a subset of the carapacial dermis (Fig. 1). Their disappearance from the dorsal mesenchyme correlates with their appearance in the ventrum (Fig. 2). Next, we demonstrate that the late-stage neural tube is still generating a migrating cell population. In one set of experiments, DiI injected into the neural tube lumen of intact stage 16 turtle embryos was later shown to be present on ventral mesenchymal cells in the plastron (Fig. 4). For any cell to be labeled, it had to have had resided in the neural tube at the time of DiI injection. Thus, cells that had been in the neural tube at the time of injection were found in the plastron regions. In a second set of experiments, neural tubes were isolated from turtle embryos at later stages of development, when the earlier set of NCCs had already differentiated. HNK-1+, p75+ cells of neural crest-like morphology were seen emigrating onto the culture dishes (Fig. 3). Thus, the trunk neural tubes of late-stage turtle embryos were still generating migratory NCC-like cells, but the expression of PDGFRα (Fig. 3J) suggests that they are similar to cranial NCCs.

We have also shown in this study that the fibronectin-adherent mesenchyme of the plastron expresses several markers of cranial NCCs (Fig. 5) and that it responds like facial ectomesenchyme when placed in medium that promotes osteogenesis (Fig. 6). This provides further and stronger evidence for the proposition (Cebra-Thomas et al., 2007; Gilbert et al., 2007) that the turtle embryo forms its plastron bones by elaborating a late population of NCCs that migrates ventrally to form a skeletogenic ectomesenchyme.

The hypothesis that the plastron bones form from trunk NCCs has been controversial (see Pennisi, 2004; Vicaryous and Sire, 2009; Kuratani et al., 2011), because trunk NCCs do not form skeletal elements in other amniotes, and cranial NCCs (which are skeletogenic) have not been shown to migrate more posteriorly than the collarbone and shoulder (Noden, 1975; Nakamura and Ayer-Le Lièvre, 1982; Le Douarin, 1982; Hall, 2005; Matsuoka et al., 2005). The only region of the neural crest capable of forming bone is the Facial Skeletogenic Neural Crest (FSNC; Le Douarin and Dupin, 2012) that extends from the diencephalon through rhombomere-2. This region forms the entirety of the facial skeleton, and most of the skull (see Hall, 2010; Lee and Saint-Jeannet, 2011).

It is possible that the late-migrating trunk neural crest cells become re-specified as cranial neural crest cells. A re-specification of trunk NCCs to FSNC “cranial” NCCs might actually represent a reacquisition of skeletogenic potency. Smith and Hall (1993) and Smith and colleagues (1994) have argued that skeletogenesis is a primitive property of the neural crest and that the trunk neural crest has generally lost this ability during vertebrate evolution. Moreover, trunk neural crest can be induced to form bone experimentally in vivo (Lumsden, 1987; Graveson et al., 1997) and in vitro (McGonnell and Graham, 2002; Abzhanov et al., 2003; Ido and Ito, 2006). The avian trunk neural crest appears to contain a population of multipotent NC progenitors, capable of producing skeletogenic derivatives in addition to neural cell types, but these are at a much lower frequency than in cranial neural crest (Calloni et al., 2007). Several experiments (Couly et al., 1998; Creuzet et al., 2002; Abzhanov et al., 2003) have suggested that Hox gene products may be critical in suppressing the skeletogenic ability of trunk NCCs. The newly acquired turtle transcriptome (Kaplinsky et al., 2013) should allow us to test hypotheses for neural crest cell specification as well as for the factors regulating the extent of neural crest cell formation, the direction of cell migration and the location of foci of osteogenesis.


Eggs, Antibody Staining and Alkaline Phosphatase Activity

Embryonated Trachemys scripta (red eared slider) eggs at various days of incubation were purchased from the Kliebert Turtle and Alligator Farm (Hammond, LA). The eggs were washed in 10% bleach, rinsed in distilled water, and washed in 70% ethanol. Chick eggs were cleaned in 70% ethanol. Embryos were isolated by dissection and staged according to Greenbaum (2002), fixed overnight with cold 4% paraformaldehyde (Sigma, St. Louis, MO) in phosphate buffered saline (PBS), washed, and embedded in Paraplast+ paraffin wax (Fisher, Morris Plains, NJ). Eight-micrometer transverse sections were cut with a Hacker-Bright rotary microtome and adhered to SuperFrost+ slides (Fisher). Antibody staining of paraffin sections and cultured, adherent cells was performed as described (Cebra-Thomas et al., 2007) with anti-CD57 (HNK-1, BD Pharminogen, San Diego, CA; 1:100–200), anti-p75 (Clontech, Mountain View, CA; 1:100), anti-PDGFRα (Santa Cruz Biotechnology, Santa Cruz, CA; 1:200–250), anti-FGF receptor 2 (bek, Santa Cruz; 1:50), anti-procollagen I (SP1.D8-s, Developmental Studies Hybridoma Bank [DSHB], Iowa City, IA; 1:4 dilution of tissue culture supernatant), and anti-collagen II (cIIc1-s, DSHB, 1:4). Extracellular alkaline phosphatase activity was assessed in unpermeabilized cells by incubation in 5-bromo-4-chloro-3-indolyl-phosphate and nitro blue tetrazolium (Fisher) or 0.5% Fast Red (Sigma). HNK-1 whole-mount immunohistochemical staining was performed based on the procedure of Metscher and Müller (2011), using anti-CD57 (1:1,000) and peroxidase conjugated goat anti-mouse IgG (Invitrogen, Carlsbad, CA; 1:1,000). Detection of the peroxidase-conjugated secondary antibody by enzyme metallography was performed using an EnzMet for General Research Applications kit according to the manufacturer's instructions (Nanoprobes Inc., Yaphank, NY).

DiI Injection

DiI (Molecular Probes, Eugene, OR) was injected into the lumen of the neural tube of G17 and G18 T. scripta turtle embryos and HH 26 chick embryos (Hamburger and Hamilton, 1951). Turtle G17 and chick HH 26 are comparable based on developmental landmarks. The embryos were prepared for injection by removing the head and ventrum so that the embryos would lie flat on Transwell membranes (Corning, Tewksbury, MA) above Dulbecco's Modified Eagles Medium (DMEM; Sigma) containing 50 μg/ml gentamicin (Invitrogen) and 2% fetal calf serum (FCS; Hyclone, Logan, UT). DiI (9.2 nl of a 0.5 mg/ml solution in 0.3 M sucrose) was injected into the posterior lumen of the neural tube using a Nanojector II (Drummond Instruments, Broomall, PA). The injected embryos were photographed immediately after injection using a Leica Fluo11 fluorescence stereomicroscope and KL1500 LCD camera to ensure that the DiI label had traveled along the lumen of the neural tube, and then incubated at 37°C and 5% CO2 for chick embryos, and 33°C and 5% CO2 for turtle embryos. The embryos were again photographed and fixed in 4% paraformaldehyde (PFA, Sigma) in PBS after 2–3 days of incubation. Approximately 60 stage 16–18 turtle embryos and 36 late-stage chick embryos were injected for this experiment.

In addition, 26 intact G16–17 turtle embryos were immobilized in 4% low melting point agarose (Sigma) and injected with 0.5 mg/ml DiI in 0.3M sucrose into the neural tube of the hindbrain and posterior to the hindlimbs. The DiI traveled along the lumen of the neural tube in the trunk region, and the embryos were cultured in suspension in DMEM supplemented with 10% FCS, 50 μg/ml gentamicin (Invitrogen), 2.5 μg/ml fungizone (Invitrogen) and 67 U/ml Nystatin (Sigma) for 2–3 days. The turtle embryos remained viable, as assessed by continued muscle contractions. The embryos were fixed in 4% PFA and photographed under epifluorescence using an Olympus SZX20 stereomicroscope and DP-20 camera. Attempts to culture chick embryos in a similar manner were unsuccessful.

In some cases, the cultured injected explants and intact embryos were embedded in 1:1 30% sucrose: Tissue Tek O.C.T. (Fisher) and 14 μm sections were cut on a Leica CM30505 cryostat. The sections were post-fixed in 4% PFA, counterstained with 0.2 μg/ml Hoescht 33342 (Sigma) and photographed under epifluorescence using an Olympus SZX20 stereomicroscope and a Nikon Eclipse E400 compound microscope, and Olympus DP-20 camera.

Neural Tube Dissection and Migration Assays

The neural tubes were isolated manually by microdissection in both chicks and turtles. The entire trunk neural tube, beginning in the cervical region anterior to the forelimb bud and ending posterior to the hind limb bud, was isolated from turtle embryos at G12 and G17 and chicken embryos at comparative stages (HH 12–14 and HH 26). If necessary, the dorsal root ganglia were removed. The isolation of the neural tube in the earlier embryos was facilitated by a brief incubation in 0.3 mg/ml dispase (Worthington Biochemicals, Lakewood, NJ). The wells of 12-well tissue culture plates (Corning) were precoated with αMEM (Sigma) containing fibronectin (25 μg/ml; Sigma), and poly-D-lysine hydrobromide (25 μg/ml; Sigma) for 1 hr. Small round glass cover slips (Fisher) were placed into the wells before coating in some experiments. After the neural tubes were cleaned of surrounding cells, they were placed in the prepared wells with 2–3 drops of neural crest culture media (αMEM, 10% FCS, 10 ng/ml FGF2 [Sigma], 50 μg/ml gentamicin, 2.5 μg/ml fungizone, 67 U/ml Nystatin). The neural tubes were allowed to adhere, and an additional 1 ml of culture media was added. Those plates containing turtle neural tubes were cultured at 33°C, 5% CO2; those containing chick neural tubes were cultured at 37°C, 5% CO2. Neural tubes were incubated for approximately 2 days for younger stages and 3 days for older stages, to allow cell emigration. The G17 neural tubes are large, and had a smoother surface. Often they did not adhere to the bottom of the wells at first. The longer culture period was used to consistently obtain large numbers of emigrating cells. The culture medium and in some cases, the neural tube, was then removed and the wells were rinsed with 1× PBS and fixed in 4% PFA.

Isolation and Culture of Plastron Mesenchyme Cells

Plastra from G19 turtle embryos were isolated by manual dissection; the ectoderm and bone primordia were removed with fine forceps; and the mesenchyme was dissociated by enzymatic digestion using 0.1% trypsin (Life Technologies, Grand Island, NY) and 0.1% collagenase 4 (Worthington) in calcium- and magnesium-free Tyrodes solution for one hour at 33°C. Single cells were separated from debris and spun through fetal calf serum. The resulting cells were cultured overnight on fibronectin and poly-D-lysine coated dishes at 0.5–1 × 105 cells/ml in NC culture medium containing 10 ng/ml FGF2 as above. The cells were fixed for antibody staining with 4% paraformaldehyde, or cultured for 7–10 days in αMEM, 10% FCS, 50 μg/ml gentamicin, 2.5 μg/ml fungizone, 67 U/ml Nystatin, supplemented with 50 μg/ml ascorbic acid (Sigma), 10 mM β-glycerophosphate (Sigma), 1 × 10−7 dexamethasone (Sigma; McGonnell and Graham, 2002) and then fixed. Similar results were obtained with mesenchyme cells from G18 embryos.

Polymerase Chain Reaction Analysis of Gene Expression

Total RNA was isolated from plastron mesenchyme microdissected from G18 and G19 embryos as described above, and the craniofacial mesenchyme of G17 embryos, using TRIZOL (Sigma), and cDNA synthesized using oligo-dT and AMV reverse transcriptase (Macherey-Nagel, Bethlehem, PA). Polymerase chain reactions for noelin-1 and Id-2 were conducted essentially as described (Abzanhov et al., 2003), but with lower annealing temperatures (noelin-1, 45°C; Id-2, 50°C). Primers for the housekeeping gene gapdh (George-Weinstein et al., 1996) were used as described to control for cDNA quality.


The authors thank A. Vihera for technical assistance. Funding for L.G., Y.H., and E.B. was provided by the Howard Hughes Medical Institute grant to Swarthmore College. J.C.-T., A.T., S.S., G.M., and K.B. were also supported by Millersville University Faculty Research Grants and by the Pennsylvania State System of Higher Education, and S.G. was also supported by Swarthmore College Faculty Research Awards and the Academy of Finland. The cIIc1-s (Holmdahl et al., 1986) and the SP1.D8-s (Foellmer et al., 1983) monoclonal antibodies were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242.