Neural crest and the origin of ectomesenchyme: Neural fold heterogeneity suggests an alternative hypothesis

Abstract

The striking similarity between mesodermally derived fibroblasts and ectomesenchyme cells, which are thought to be derivatives of the neural crest, has long been a source of interest and controversy. In mice, the gene encoding the alpha subunit of the platelet-derived growth factor receptor (PDGFRα) is expressed both by mesodermally derived mesenchymal cells and by ectomesenchyme. Whole-mount immunostaining previously revealed that PDGFRα is present in the cephalic neural fold epithelium of early murine embryos (Takakura et al. [1997] J Histochem Cytochem 45:883–893). We now show that, within the neural fold, a sharp boundary exists between E-cadherin–expressing non-neural epithelium and the neural epithelium of the dorsal ridge. In addition, we found that cells coexpressing E-cadherin and PDGFRα are present in the non-neural epithelium of the neural folds. These observations raise the possibility that at least some PDGFRα⁺ ectomesenchyme originates from the lateral non-neural domain of neural fold epithelium. This inference is consistent with previous reports (Nichols [ 1981] J Embryol Exp Morphol 64:105–120; Nichols [ 1986] Am J Anat 176:221–231) that mesenchymal cells emerge precociously from an epithelial neural fold domain resembling the primitive streak in the early embryonic epiblast. Therefore, we propose the name “metablast” for this non-neural epithelial domain to indicate that it is the site of a delayed local delamination of mesenchyme similar to involution of mesoderm during gastrulation. We further propose the testable hypothesis that neural crest and ectomesenchyme are developmentally distinct progenitor populations and that at least some ectomesenchyme is metablast-derived rather than neural crest-derived tissue. Developmental Dynamics 229:118–130, 2004. © 2003 Wiley-Liss, Inc.

INTRODUCTION

The neural crest in amphibian, avian, and mammalian embryos is generally considered to be the population of cells located in the neural folds at the boundary between neural and epidermal ectoderm. During the process of neurulation, the dorsal ridges of the neural folds, which include neural crest progenitors, elevate and converge on the embryonic midline. At the completion of neurulation, the dorsal ridges of the neural folds fuse to form the neural tube overlain by ectodermal epithelium (Moury and Jacobson, 1989; Smith and Schoenwolf, 1997; Hall, 1999). Neural crest cells then emerge from the dorsal epithelium of the nascent neural tube into a migration staging area (Weston, 1991) and eventually disperse on distinct migration pathways in embryonic interstitial spaces. After dispersal, crest-derived cells differentiate in precise locations to form pigment cells of the skin and irides, glial and neuronal components of the peripheral and enteric nervous systems, and several neurosecretory cell types (LeDouarin and Kalcheim, 2000). In addition, the neural crest is generally considered to be the source of a population of cells, called “mesectoderm” or “ectomesenchyme” (see Hall and Hörstadius, 1988; Hall, 1999, for references and a more detailed discussion of the classic literature). Ectomesenchyme produces a variety of craniofacial skeletal and connective tissues, whose cellular phenotypes are remarkably distinct from neurogenic and melanogenic derivatives of the neural crest. These derivatives include components of the meningeal sheath covering the central nervous system (Raven, 1936, 1937; see Weston, 1970; Hall and Hörstadius, 1988; LeDouarin and Kalcheim, 2000, for classic citations), osteogenic cells of the jaw, palate, and parts of the skull (Chibon, 1969; Johnston, 1966; LeLievre and LeDouarin, 1975; LeLievre, 1978; Noden, 1978; Hall and Tremaine, 1979; Noden, 1988; Smith and Hall, 1990; Morriss-Kay et al., 1993), the corneal stroma (Hay, 1980), the odontoblasts of the dental papilla (Thesleff and Nieminen, 1996), and stromal cells of the thyroid, thymus, and probably other tissues whose differentiation involves interactions with pharyngeal endoderm (Bockman and Kirby, 1984; Weston, 1984). In terrestrial vertebrates, cells that form the vascular septa dividing the pulmonary and systemic outflow of the heart also arise from this mesenchymal population (Kirby et al., 1983; Morriss-Kay et al., 1993), and in aquatic vertebrates, ectomesenchyme produces connective tissue of the dorsal fin (DuShane, 1935; Hörstadius and Sellmann, 1946; see Hörstadius, 1950, for summaries). Compared with the emergence of crest cells from the dorsal neural tube in the embryonic trunk, at least some mesenchymal cells in the cranial domain of mammalian and avian embryos appear to emerge precociously, during neural fold elevation, well before the dorsal ridges fuse in the midline to form the neural tube (Vermeij-Keers and Poelmann, 1980; Nichols, 1981, 1986; Tan and Morriss-Kay, 1985).

Ectomesenchyme cells have been shown in various embryos to express the gene encoding the alpha subunit of the platelet-derived growth factor receptor (PDGFRα; Mercola et al., 1990; Morrison-Graham et al., 1992; Orr-Urtreger and Lonai, 1992; Schatteman et al., 1992; Endo et al., 2002; Bernard and Christine Thisse, personal communication). This gene is also expressed by mesodermally derived mesenchymal cells, which arise initially during gastrulation from epiblast epithelium, and subsequently from mesodermally derived epithelial somites. Thus, there appear to be two distinct sources of mesenchymal cells that express PDGFRα—the prechordal mesoderm and the ectomesenchyme; the latter is generally believed to originate from the cranial neural crest.

As might be expected, because both ectomesenchyme and mesodermally derived mesenchyme in the embryonic trunk express abundant PDGFRα message in early embryos, deletion or loss of function of the gene encoding PDGFRα (PDGFRa^Patch) causes multiple defects in connective tissue derivatives, leading to embryonic lethality. In early mouse homozygotes, for example, maxillary and mandibular processes fail to fuse and the palate does not form. Homozygous embryos also exhibit corneal defects and absence of septation of the cardiac outflow tract (Stephenson et al., 1991; Morrison-Graham et al., 1992; Schatteman et al., 1992). It is important to note, however, that these mutations do not cause cell-autonomous defects in the early development of neurogenic and melanogenic derivatives of the neural crest (Morrison-Graham et al., 1992; Schatteman et al., 1992; Wehrle-Haller et al., 1996; Soriano, 1997).

PDGFRα expression and function distinguish ectomesenchymal cells from neural crest-derived neurogenic and melanogenic cells, suggesting that ectomesenchyme and neural crest represent developmentally distinct progenitor cell populations, at least by the time receptor tyrosine kinase function is required for normal development. It is essential, therefore, to learn when these two developmentally distinct cell populations segregate from each other. In this regard, the report (Takakura et al., 1997) that PDGFRα immunoreactivity first appears in the epithelial neural folds lateral to the dorsal ridge at cranial and caudal axial levels of embryonic day (E) 8 murine embryos, raises at least two important questions: first, is the lateral epithelial domain phenotypically different from the dorsal ridge of the neural fold that produces neural crest precursors in the dorsal neural tube? And second, do PDGFR-expressing ectomesenchyme cells arise from this spatially distinct lateral, non-neural domain of the neural folds instead of, or in addition to, the dorsal neural tube epithelium?

To address these questions, we exploited previous reports (Thiery et al., 1984; Duband et al., 1988; Duband et al., 1995; Newgreen et al., 1997) that embryonic neural and non-neural (epidermal) epithelia express different cell adhesion molecules. Specifically, because it has been shown that L-CAM (E-cadherin) is expressed by non-neural but not by neural epithelia, we carefully examined the boundaries of E-cadherin immunoreactivity in the cranial neural folds of murine embryos. Next, to assess the possibility that E-cadherin–expressing non-neural epithelia might generate PDGFRα-expressing mesenchyme, we asked whether any E-cadherin–immunoreactive epithelial cells coexpress PDGFRα.

Our results suggest the directly testable hypothesis that ectomesenchyme cells arise from non-neural epithelium in the embryonic neural fold and that ectomesenchyme and “authentic” neural crest cells, which emerge later from the dorsal neural tube epithelium, represent spatially, temporally, and developmentally distinct populations. Both intrinsic marking methods to identify cells undergoing epithelium-to-mesenchyme transition (EMT; see, for example, Pietri et al., 2003a), and extrinsic marking protocols to follow the fates of specific cell lineages (see, for example, Schilling and Kimmel, 1994) will allow the predictions of this hypothesis to be tested.

RESULTS

E-Cadherin Immunoreactivity Reveals a Sharp Boundary Between Neural and Non-neural Epithelium in Murine Embryonic Neural Folds

Immunostaining with monoclonal antibody ECCD2, which recognizes E-cadherin, defines distinct embryonic domains that include surface and gut epithelia but not neural epithelium. Nascent, epithelial somites were also E-cadherin⁺ (not shown; H. Yoshida and T. Era, unpublished observations). As expected, intense E-cadherin immunoreactivity is localized at the junctions between epithelial cells. The immunoreactive cells in the neural folds form a precise boundary with neural epithelia as the dorsal ridges of the folds elevate and converge toward the embryonic midline. As the neural folds elevate, this precisely defined E-cadherin⁺ epithelium comes to lie atop, but clearly distinct from, the dorsal ridge of the neural epithelial fold (Fig. 1b,f, and c,h). As previously described (see Thiery et al., 1984; Duband et al., 1988; Martins-Green, 1988; Smith and Schoenwolf, 1997), when the dorsal ridges meet in the embryonic midline to form the neural tube, the neural and the E-cadherin⁺ non-neural epithelia remain distinct, with the latter forming the surface epithelium overlying the neural tube.

Figure 1. E-Cadherin and platelet-derived growth factor receptor (PDGFRα) expression in embryonic day 8.5 murine embryos. a: Whole-mount embryo subjected to diaminobenzidine staining after exposure to APA5 (rat anti-PDGFRα monoclonal antibody). Notice immunostaining in the somites as well as in the lateral cranial and caudal neural folds. Lines indicate planes of section in similar embryos illustrated in b–e. b,c: Sections of embryos stained with ECCD (rat anti–E-cadherin monoclonal antibody). Note intense staining surrounding non-neural epithelial cells and distinct boundary between surface epithelium and the thickened neural epithelium. Boxes indicate regions enlarged in f and g, and h and i, respectively. f,h: Note reduced cytoplasmic expression of E-cadherin in cells in and below the lateral neural plate (arrowheads). d,e: Same sections as b and c, showing double staining with ECCD and APA5 (rat anti-PDGFRα monoclonal antibody). Note colocalization of staining (yellow) in epithelia lateral to (d) and overlying (e) the lateral neural plate epithelium (arrows). g,i: Also, note in the enlargements, the apparent colocalization of PDGFRα staining in cells, in and below the neural plate epithelium, that also show weak cytoplasmic E-cadherin immunoreactivity (arrowheads).

E-Cadherin Is Down-regulated in Epithelial Cells Undergoing EMT

At early stages of elevation of the neural folds, a population of cells with low-level E-cadherin immunoreactivity in their cytoplasm lies just beneath the E-cadherin⁺ epithelium at the lateral boundary of the E-cadherin–negative neural plate epithelium (Fig. 1b,f). These cells are present in a loose, multilayered mass, distinct from both the neural and the non-neural epithelia. The pattern of immunoreactivity in the cytoplasm of this cell population suggests that, as they undergo EMT, E-cadherin is removed from the surfaces of these cells. It is noteworthy that some cells with low-level cytoplasmic E-cadherin immunoreactivity also appear to intermingle with cells of the thickened neural epithelium (Fig. 1c,h).

E-Cadherin and PDGFRα Are Colocalized in the Lateral Non-neural Epithelium and on the Dorsal Ridges of the Neural Folds

Whole-mount alkaline phosphatase immunostaining confirmed that PDGFRα immunoreactivity appears in the lateral neural fold epithelium, at both cranial and caudal axial levels (Fig. 1a) as previously reported (Takakura et al., 1997). To determine the precise location of the PDGFRα-immunoreactive cells seen in whole-mount preparations, we double-stained histologic sections of similarly staged embryos with both anti-PDGFRα and anti–E-cadherin monoclonal antibodies (see Experimental Procedures, Shindler and Roth, 1996). Although subject to somewhat increased background fluorescence, the double-staining method revealed that PDGFRα is colocalized with E-cadherin in a region of cranial neural fold epithelium, immediately lateral to the thickened neural plate epithelium (Fig. 1d, arrows; Fig 1g). In addition, PDGFRα immunoreactivity is observed in a population of cells with reduced cytoplasmic E-cadherin expression lying beneath this non-neural epithelium (region marked by arrowhead in Fig. 1g). As expected, fluorescently stained PDGFRα-immunoreactive cells were also observed throughout the mesenchyme of the cephalic neural folds and branchial arches, as well as in the lateral body wall mesoderm.

At axial levels where neural folds were more elevated, the E-cadherin⁺/PDGFRα⁺ non-neural epithelium lies atop the dorsal ridge of the neural epithelium (Fig. 1e, arrows; Fig 1i). At this stage, some PDGFRα⁺ cells are also transiently present in the lateral domain of the thickened neural epithelium. This PDGFRα immunoreactivity appears to colocalize with cells that also express reduced cytoplasmic E-cadherin immunoreactivity (see above, and Fig. 1i, arrowhead).

E-Cadherin/PDGFRα Coexpressing Cells Are Present in Cell Populations Derived From Murine Neural Folds

Although the immunostaining results above indicate that E-cadherin⁺ and PDGFRα⁺ cells are colocalized in histologic sections, it cannot be directly concluded that cells coexpress these two antigens. Therefore, to confirm that some E-cadherin⁺ cells coexpress PDGFRα, we subjected dissociated cell populations from cranial neural folds (Fig. 2a) and from caudal embryonic trunk to fluorescence-activated cell sorting (FACS) based on E-cadherin and PDGFRα expression (see Experimental Procedures section). As expected, populations from both sources contained abundant subpopulations that uniquely expressed either E-cadherin or PDGFRα (Fig. 2b,c). In addition, however, cells dissociated from both the cranial neural folds and trunk segments containing nascent somites also contained limited subpopulations that expressed both E-cadherin and PDGFRα immunoreactivity (Fig. 2b,c). This result confirms the in situ immunostaining results and shows that colocalization of immunostaining in situ is the consequence of coexpression of PDGFRα by cells in the non-neural epithelium in the lateral neural folds.

Figure 2. Flow cytometry of cells dissociated from embryonic day 8.5 murine embryonic head and trunk fragments. a: Collection of neural folds and associated tissues dissected from embryos before dissociation (see Experimental Procedures section). b: Plot of cells dissociated from embryonic heads (a), and subjected to fluorescence-activated cell sorting (FACS) with respect to expression of platelet-derived growth factor receptor (PDGFRα; vertical axis) and E-cadherin (horizontal axis). c: Plot of cells dissociated from isolated trunk tissue and subjected to FACS as described. In b and c, note points in scatter plot (to the right and above the threshold lines) that represent cells coexpressing both PDGFRα and E-cadherin. FITC, fluorescein isothiocyanate.

DISCUSSION

The embryonic neural crest has been the focus of interest and speculation since its initial description and characterization (de Beer, 1947; see Hall and Hörstadius, 1988, for a review of the classic literature). The assertion that the neural crest contributes to skeletal and connective tissue derivatives originally caused considerable controversy, because it appeared to violate the notion that mesoderm, and not ectoderm, was the source of such tissues (see Hall, 1999, for additional discussion of this controversy). By now, however, it has become “conventional wisdom” that the neural crest is the source of cranial skeletal and connective tissues. Indeed, the heuristic and widely accepted idea that the neural crest was an “evolutionary invention” that gave rise to the craniofacial structures distinguishing vertebrates from other chordates (Gans and Northcutt, 1983; Hall, 1999) has now lent the cranial neural crest a special cachet and increased the credulous enthusiasm with which it is regarded. In the recent literature, not only are branchial arch-derived skeletal and connective tissues generally considered to be neural crest derivatives, but as discussed below, phenotypic markers expressed by such derivatives are routinely, but somewhat uncritically, used to indicate “cranial neural crest” identity of the expressing cells—largely ignoring expression of these markers by non–neural crest-derived cell types, and conversely, lack of expression of some of these markers by neural crest-derived subpopulations.

As Hörstadius (1950) and subsequently others (e.g., Tan and Morriss-Kay, 1985; Moury and Jacobson, 1990) have correctly pointed out, however, there has always been some ambiguity about precisely what portion of the embryonic neural folds gives rise to neural crest cells (see also Raven, 1931; Baker and Graves, 1939). This ambiguity is based on fundamental confusion in the literature about where in the neural folds neural epithelium ends and epidermal epithelium begins. As discussed below, this confusion has probably resulted in misinterpretation of the results of marking and grafting experiments indicating that the neural crest is the source of ectomesenchyme-derived skeletal and connective tissue of the head. To reconcile various discrepancies and ambiguities in the literature, therefore, we propose the following testable alternative hypothesis: Although ectomesenchyme and dispersing neural crest cells share interstitial spaces in the embryonic head, just as do sclerotomal mesenchyme and neural crest cells in the embryonic trunk, ectomesenchyme emerges from epithelia that are spatially and developmentally distinct from those that generate neural crest. The arguments justifying this proposal are discussed below (see also, Verwoerd and van Oostrom, 1979).

Non-neural Epithelium in the Neural Fold Appears to Be a Source of Mesenchyme Cells

Improved histologic resolution, made possible with tissue-specific cadherin immunostaining, has clarified the location of the boundary between neural and non-neural epithelium in the neural fold. Previous reports (Duband et al., 1988, 1995) of the pattern of A-CAM (N-cadherin) immunoreactivity in avian embryos suggest that N-cadherin expression complements the E-cadherin expression reported above and that both function to preserve the developmental distinctions between neural and non-neural epithelia. Although it is important to emphasize that cadherin expression is not an unequivocal indicator of cell fate, the presence of such cell adhesion molecules at the surface of specific cell populations does ensure that epithelial integrity is maintained, which in turn, is believed to preserve distinct developmental compartments (Matsumami and Tabeichi, 1995). Thus, it seems likely that the continued tissue-specific expression of E-cadherin maintains the sharp boundary between neural and non-neural epithelium and preserves developmental distinctions between these neighboring cell populations. Consistent with these inferences, our results clearly show that even as it comes to lie atop the dorsal ridge of the neural epithelium in elevated murine neural folds, E-cadherin⁺ non-neural epithelium remains sharply separate and distinct from the neural plate epithelium—presumably the result of cell type-specific adhesion molecules expressed in these epithelia. Eventually, as the folds fuse in the midline to form the neural tube, the E-cadherin⁺ epidermal epithelium separates completely from the cells of the dorsal neural tube (see Tan and Morriss-Kay, 1985; Martins-Green, 1988; Smith and Schoenwolf, 1997; Lawson and England, 1998; Colas and Schoenwolf, 2001).

We have also seen that a unique domain of E-cadherin–expressing cells, abutting the neural plate, coexpresses PDGFRα, which is also expressed by branchial arch and other mesenchymal cells in the head, as well as by mesodermally derived mesenchyme in the embryonic trunk (Morrison-Graham et al., 1992; see also Orr-Urtreger and Lonai, 1992; Schatteman et al., 1992). Of interest, both E-cadherin and PDGFRα are also coexpressed in the primitive streak and nascent somites (H. Yoshida and T. Era, unpublished observations). At later stages of neurulation, we have observed some PDGFRα-immunoreactive cells beneath the strongly E-cadherin⁺ non-neural epithelium. This PDGFRα⁺ mesenchymal population exhibits low-level, cytoplasmically localized E-cadherin immunoreactivity. The reduced E-cadherin immunoreactivity in the cytoplasm of these PDGFRα⁺ cells is consistent with the reports of Snail/Slug-induced repression of E-cadherin synthesis (Cano et al., 2000; Locascio et al., 2002), coupled with endocytosis-mediated turnover of cadherin from the cell surface (see Miller and McClay, 1997). Importantly, the pattern of dual immunoreactivity by such subepithelial cells clearly indicates that they originate from the overlying E-cadherin⁺ non-neural epithelium and that they lose epithelial morphology as a consequence of E-cadherin down-regulation from their cell surfaces. The doubly immunoreactive cells reside at the boundary between neural and non-neural epithelia, often beneath an epithelial groove (see also, Waterman, 1976; Nichols, 1981). This groove, sometimes designated the neurosomatic junction (see O'Rahilly and Muller, 1999), bears a remarkable resemblance to the primitive streak that transiently exists at early epiblast stages of avian and mammalian development as the site of involution of mesodermal mesenchyme during gastrulation. It is noteworthy that the PDGFRα⁺ mesenchymal cells that appear to delaminate from the non-neural epithelium of the neural fold probably correspond to the cells previously reported to emerge precociously from elevating neural folds (Nichols, 1981, 1986; for specific examples, see Figs. 3 and 4 in Nichols, 1986).

As reported above, PDGFRα-immunoreactive cells also appear to be present in the lateral neural plate epithelium (e.g., Fig. 1e). These results appear to suggest that PDGFRα⁺ cells also arise in the neuroepithelium of the neural folds, and, therefore, that the PDGFRα⁺ ectomesenchymal cell population must, at least, have a dual origin from both neural and non-neural epithelia. This possibility, which is consistent with the observation that crest can be induced from non-neural epithelium by juxtaposed neural ectoderm (Moury and Jacobson, 1989; Selleck and Bronner-Fraser, 1995), cannot presently be excluded. However, the immunostaining pattern in this region suggests an alternative hypothesis. Thus, although N-cadherin is reported to be present in the thickened neural plate (Duband et al., 1995), it is eventually down-regulated in cells at the lateral edges of the neural plate, and in the dorsal portion of the neural tube before emergence of crest cells from neural epithelium (Nakagawa and Takeichi, 1998). As with E-cadherin (see above), this down-regulation is probably the consequence of Snail/Slug-mediated repression of cadherin synthesis followed by endocytosis (see above; Cano et al., 2000). Thus, the low-level E-cadherin immunoreactivity that we observe in the cytoplasm of PDGFRα⁺ cells within the thickened lateral neural epithelium of the neural folds strongly suggests that they originate from E-cadherin⁺ epithelium, just like the PDGFRα⁺ cells that lie beneath the E-cadherin⁺ non-neural epithelium. Therefore, as an alternative to dual origin of these PDGFRα⁺ cells, we suggest that they originate in the non-neural epithelium and transiently intermingle with neural epithelial cells as both cell populations down-regulate their tissue-specific cell surface cadherins. Such intermingling has been reported to occur in model systems when cadherin levels change in the surfaces of confronting cell populations in vitro (Duguay et al., 2003, and M.S. Steinberg, personal communication). Accordingly, although we cannot presently eliminate the possibility that ectomesenchyme has a dual origin in both neural and non-neural epithelia, we suggest that at least some of the PDGFRα⁺ cells within the thickened neural epithelium might actually originate from overlying non-neural epithelium. Although these cells have down-regulated cell surface E-cadherin, they might still represent a developmentally distinct compartment within the dorsal ridge of the neural folds and, therefore, might also be distinct from “authentic” neural crest cells, whose precursors emerge later from the dorsal neural tube epithelium at both cranial and trunk axial levels.

Phenotypes of Ectomesenchyme and Neural Crest Cells Are Distinct

The possibility that precociously delaminating ectomesenchyme is developmentally distinct from the neural crest finds support in several observations and inferences in the literature. Thus, in addition to expressing PDGFRα, cells of the lateral non-neural (epidermal) epithelium also transiently express β3 integrin (Pietri et al., 2003b) and a proteoglycan link protein characteristic of cartilage cells (Colas and Schoenwolf, 2001). Trunk neural crest cells derived from the neural epithelium of the dorsal neural tube do not express these gene products. Moreover, ectomesenchyme cells, like mesodermally derived fibroblasts but unlike “authentic” neural crest cells, also synthesize and secrete fibronectin (Newgreen and Thiery, 1980; Duband and Thiery, 1982) and the hyaluronan receptor CD44 (Corbel et al., 2000). Consistent with these molecular traits, ectomesenchyme cells in vitro exhibit a large fibroblastic morphology similar to somite fibroblasts, whereas on identical culture substrata, trunk neural crest cells, as well as cells that emerge belatedly from cranial neural fold explants, exhibit a manifestly different, small stellate morphology (see Newgreen and Thiery, 1980; Newgreen and Minichiello, 1995).

Other gene products have also been used to indicate neural crest identity. These include FoxD3 (Kos et al., 2001; Sasai et al., 2001), Zic5 (Nakata et al., 2000), Twist (Hopwood et al., 1989; Soo et al., 2002), and AP2 (Luo et al., 2003b). It is noteworthy, however, that expression of these putative crest markers was first recognized in mesodermally derived tissues of vertebrate embryos (see Furchtbauer, 1995; Stoetzel et al., 1995; Gitelman, 1997; Soo et al., 2002). It is also noteworthy that, in addition to PDGFRα mutations, there are several mutations affecting branchial arch and cranial skeleton that show no cell-autonomous pleiotropic effects on pigmentation or on the development of the peripheral nervous system (Sasaki and Hogan, 1993; Satokata and Maas, 1994; Foerst-Potts and Sadler, 1997; Srivastava et al., 1997; Thomas et al., 1998; see also Henion et al., 1995; Kelsh et al., 1996). It seems likely, therefore, that like mutations in the PDGFRα gene, these mutations act in a cell population that is already developmentally distinct from the rest of the neural crest. It is not yet known if this distinction arises during segregation of cell lineages from progenitors within the dorsal ridge of the neural folds (i.e., from precursors common to both neural crest and ectomesenchyme), or if the mutations affect genes that function only in non-neural epithelium that is distinct from neural crest precursors (see below and Tonegawa et al., 1997).

In apparent contradiction to the idea, discussed above, that neural crest-derived cells and ectomesenchyme express distinct phenotypes, many recent contributors have relied extensively on expression of the Snail gene family as markers for both neural crest-derivatives and ectomesenchyme cells (e.g., Carl et al., 1999; Mayor et al., 1999; LaBonne and Bronner-Fraser, 2000; Aybar et al., 2003). It is important to recognize, however, that Snail/Slug expression is developmentally dynamic and appears at many different times and in many cell types. In Xenopus laevis embryos, for example, Snail is expressed just before EMT in the lateral non-neural epithelium of the neural folds and appears in the trunk of slightly older (stg. 22–25) Xenopus laevis embryos in the epithelium above the dorsal ridge of the closing neural folds, which is believed to contribute to fin mesenchyme (Mayor et al., 1995). Snail/Slug is also expressed in the epiblast of avian and mammalian embryos, where mesodermal cells undergo involution through the primitive streak (see Cano et al., 2000), and by epithelial somite cells before formation of sclerotomal and dermal mesenchyme (Locascio et al., 2002). Later, Snail/Slug is expressed by the precursors of “authentic” neural crest derivatives (PNS and pigment cells) before they undergo EMT in the dorsal neural tube epithelium of the head and trunk of zebrafish (Thisse et al., 1995), Xenopus (LaBonne and Bronner-Fraser, 2000), and avian (Endo et al., 2002) embryos. As has been emphasized recently, however, Snail/Slug expression precedes, and probably controls, E-cadherin down-regulation before EMT in all these epithelia (Cano et al., 2000). Accordingly, Snail/Slug expression indicates execution of a morphogenetic program and should not be considered a cell type-specific marker either for ectomesenchyme, or subsequently, for neural crest precursors in the dorsal neural tube.

Based on the considerations above, we have predicted that neural crest and ectomesenchyme precursors are developmentally distinct at early developmental stages. It is of interest, therefore, that, where fate restrictions have already been examined by lineage analysis, ectomesenchymal cells appear to be developmentally distinct from “authentic” crest cells. Thus, in zebrafish, mesenchymal cells arising from tissue lateral to cranial neural epithelium appear to be fate-restricted to form connective tissue derivatives of the branchial arch skeleton as soon as they can be recognized and labeled for lineage analysis (Schilling and Kimmel, 1994). Coincidentally, these cells, which the authors called the “premigratory mass” of neural crest cells, also appear to express markers such as PDGFRα (Bernard and Christine Thisse, personal communication) and Snail (Thisse et al., 1995).

Dissimilarities in Experimental Protocols Can Account for the Inferred Differences Between Cranial and Trunk Crest Fates

The suggestion that non-neural epithelium gives rise to ectomesenchyme finds support from an unlikely source. Thus, the elegant quail–chicken grafting studies, which were used to define the fates of cranial crest cells in avian embryos (LeLievre and LeDouarin, 1975), indicated that donor-derived meningeal cells and connective tissue mesenchyme (including vascular smooth muscle and corneal stroma) were present in host embryos that received orthotopic grafts in the cranial region but not in the embryonic trunk. These results led to the inference that in avian embryos, trunk crest cells lack the potential to give rise to ectomesenchymal derivatives (LeLievre and LeDouarin, 1975; LeLievre, 1978). These conclusions differed from those derived from the results of the original vital staining and grafting experiments in amphibian embryos (Raven, 1931; Hörstadius and Sellman, 1946; see also Chibon, 1964; and the summary discussions in Hall and Hörstadius, 1988). In these earlier studies, labeled or graft-derived meningeal cells and connective tissue mesenchyme (such as dorsal fin mesenchyme) were present in both the embryonic head and trunk.

It should be emphasized that the inferred lack of ectomesenchyme potential in avian trunk neural crest in vivo is based on the implicit assumption that the tissues used in all the grafting experiments were comparable. However, this is not the case. In fact, all the classic experimental studies on cephalic and trunk crest in amphibian embryos involved marking, ablation, or transplantation of neural folds—tissues that contain both neural and non-neural (epidermal) epithelium (see, for examples, Hall and Hörstadius, 1988; and Sadaghiani and Thiebaud, 1987). Likewise, the experiments with avian embryos also used neural fold tissue as grafts in rostral axial levels. In contrast, however, neural tubes, which in most grafting protocols, lack non-neural epithelial cells, were transplanted at more posterior axial levels (posterior to somite 6 of embryos with 7 or more somites; LeLievre and LeDouarin, 1975). In these experiments, donor-derived ectomesenchymal tissues (branchial arch skeletal and connective tissue derivatives, corneal stroma, meninges, etc.) were obtained in avian embryos only when cephalic neural folds were used as grafts (LeLievre and LeDouarin, 1975; Noden, 1975, 1978). Such derivatives were not observed after orthotopic transplantation of neural tubes. Consequently, the different results, which led to the inferred differences between the developmental potential of avian cranial and trunk crest populations, can most straightforwardly be explained simply by the presence or absence of non-neural epithelium in the grafted tissue, rather than by intrinsic properties of the grafted crest cells from different axial levels.

Do Ectomesenchyme and Crest Derivatives Share a Common Precursor?

Despite all of our circumstantial arguments supporting the idea that neural crest and ectomesenchyme originate from distinct embryonic sources, the well-known results of clonal cell culture (Baroffio et al., 1988, 1991; Ito and Sieber-Blum, 1991) appear decisively to refute this idea. From the results of such cloning studies, it has been inferred that the cephalic neural crest contains common precursors for neural and mesectodermal (ectomesenchyme) derivatives. As is often the case with cell culture experiments, however, the results might be vulnerable to misinterpretation about events in vivo. Indeed, current understanding of the technical problems inherent in such long-term clonal cultures (see Henion and Weston, 1997) suggests that the data do not support such inferences. In the first place, clonal efficiency, after 2 weeks of culture, was probably significantly underestimated, because small clones (e.g., clones containing fewer than 20 cells) were not analyzed, and we now know that many small clones are probably derived from fate restricted progenitor cells (Henion and Weston, 1997; Luo et al., 2003a). Then, in one carefully presented and widely accepted analysis (Baroffio et al., 1991), only 10 clones of the hundreds that were analyzed contained cartilage nodules that operationally represented the sought ectomesenchyme phenotype. Of these, just two clones also expressed derivatives (e.g., neurons, pigment, or other markers) characteristic of “authentic” crest. Because, for the reasons mentioned above, plating efficiency was probably underestimated in these experiments, these two clonal progenitors are likely to represent no more than 0.2% of the viable cells obtained from dissociated neural folds. More importantly, these two clones were highly exceptional because they alone contained 20,000–50,000 cells—two to three orders of magnitude more cells than any other cartilage-containing clone. Furthermore, the crest cell marker HNK1 was expressed by some cells in these populations but not the same cells that expressed cartilage traits. Therefore, two alternative interpretations of the clonal data seem likely: (1) that, despite the care taken by the authors, these exceptionally large colonies were not clonally derived, or (2) that the highly proliferative clonal progenitors that produced them were actually rare embryonic stem cells whose development was affected by their prolonged exposure to a rich culture environment. In either case, we suggest that the results of these clonal analyses do not unequivocally support the authors' conclusion that ectomesenchyme derivatives, melanocytes, and components of the peripheral nervous system share common neural crest cell precursors in vivo.

Delayed Delamination, Similar to the Gastrulation Process in Embryonic Epiblast, Can Give Rise to Ectomesenchyme From Non-neural Epithelium in the Neural Folds

Although our suggestion might mitigate an otherwise insurmountable difficulty for our hypothesis, it raises another portentious question—namely, if not from neural crest precursors, where do progenitors of ectomesenchyme originate?

Before considering the possible origin of ectomesenchyme progenitors, it is important to re-emphasize the observations that E-cadherin and PDGFRα are coexpressed both in the neural fold epithelium and in the primitive streak (H. Yoshida and T. Era, unpublished observations), and that at least some ectomesenchyme appears to originate from a region of lateral neural fold epithelium with striking morphologic similarity to the primitive streak in the epiblast (Waterman, 1976; Nichols, 1981, 1986). Based on these considerations, we propose to refer to the non-neural epithelial source(s) of ectomesenchyme in neural fold as “metablast,” to emphasize the special properties of such localized domains within the neural folds, and the marked similarity of ectomesenchymal cells to mesodermal fibroblasts that arise during gastrulation. The prefix of our proposed name, “metablast,” means “occurring later than, or in succession to [epiblast]” and suggests that the delamination of mesenchyme from the non-neural epithelium of the neural fold represents a delayed morphogenetic event similar to the gastrulation process that occurs earlier in the embryonic epiblast.

Next, it is appropriate to recall that both mesodermal (dermal and sclerotomal) mesenchyme in the trunk and ectomesenchyme in the head occupy embryonic interstitial spaces early and subsequently share some of these spaces with dispersing neural crest-derived cells. Like somitic mesenchyme in the trunk (see Erickson, 1997; Erickson and Reedy, 1998), ectomesenchyme also appears to be required for the normal morphogenetic behavior of later-migrating crest cells in the head (see Kanzler et al., 2000; Soo et al., 2002). For all the reasons summarized above, it seems useful to maintain a formal and operational distinction between “metablast-derived” ectomesenchyme and “authentic” neural crest cells that emerge from dorsal neural tube epithelium. We posit that this is not “merely” a semantic distinction to be gratuitously ignored or dismissed. Indeed, to do so seems likely to lead to erroneous inferences about cell type-specific developmental regulatory mechanisms, because as we will assert below, one cannot easily elucidate the regulation of development of one tissue or cell type by exploring regulatory mechanisms in a different tissue or cell type.

Where, then, does the putative “metablast” originate? The answer to this question is, of course, presently unknown. However, blastoderm mapping studies (Hatada and Stern, 1994; Foley et al., 2000) provide intriguing hints about possible origins of these cells. These workers identified an area of the early avian blastoderm containing cells that move anteriorly and converge toward the midline well before primitive streak formation. However, these cells do not appear to undergo involution through the primitive streak. Therefore, in light of the remarkable similarity between delamination of mesoderm in the epiblast, and delamination of ectomesenchyme from putative metablast epithelium, as well as the many similar phenotypes among ectomesenchyme-derived and nascent mesoderm-derived cells, it is tempting to speculate that cells fated to contribute to the metablast originate as a consequence of the signaling events that lead to the initial segregation of mesodermal precursors in the epiblast epithelium. Such prospective metablast cells would bide their time in the anterior epiblast during primitive streak regression, but somehow evade incorporation into the primitive streak domain. Finally, in this scenario, these precociously specified cells would undergo delayed involution/delamination in the neural folds to produce ectomesenchyme that is phenotypically and functionally similar to sclerotome mesenchyme. This, then, suggests the crucial distinction between putative metablast and neural crest: Cells of the former would be specified in the early epiblast by intercellular signals that induce mesoderm, whereas the latter result from signals between neural epithelium and adjacent non-neural ectoderm (see Moury and Jacobson, 1989, 1990; Selleck and Bronner-Fraser, 1995). Consistent with this idea, it is noteworthy that, in the case of the interactions between neural plate and non-neural ectoderm in vitro, cellular phenotypes characteristic of “authentic” neural crest, but not cells expressing ectomesenchymal (mesectodermal) phenotypes, were detected in the cultures.

Of course, the definitive tests of the hypothesis that ectomesenchymal cells arise from the non-neural epithelium of the neural fold will require that such non-neural epithelial cells be selectively identified—either by extrinsic marking for cell lineage tracing (e.g., Schilling and Kimmel, 1994), or by intrinsically marking specific cell populations undergoing EMT, using the ROSA26 transgenic mouse strain mated to transgenic embryos expressing Cre-recombinase driven by the promoters of genes differentially expressed during this morphogenetic process (e.g., Pietri et al., 2003a)—to show that they give rise to skeletogenic mesenchyme in the embryonic head. Such lineage tracing studies will require high-resolution fate maps of the neural fold to be constructed to distinguish the locations of putative metablast from placodal epithelia (see, for example, D'Amico-Martel and Noden, 1983).

Whatever its early origin, the foregoing discussion suggests that it would be prudent to consider the ectomesenchyme population to be just as distinct from later-dispersing cephalic neural crest cells (see Erickson and Reedy, 1998), as the somitic (sclerotomal and dermal) mesenchyme is from trunk neural crest cells that disperse amongst them. If, in the fullness of time, the predictions of our hypothesis were validated, the notion that neural crest and ectomesenchyme represent spatially and temporally distinct lineages in the early embryo would have at least two significant consequences. First, it would eliminate the conceptual difficulty of explaining how unique skeletal and connective tissue phenotypes (ectomesenchyme derivatives) differentiate at cranial axial levels from the same progenitors as pigment cells and components of the peripheral nervous system. Second, it would mitigate a practical impediment to identifying specific genetic mechanisms regulating cell specialization in these two embryonic tissues. Thus, if we acknowledge that ectomesenchyme and neural crest are developmentally and not merely semantically distinct, we will be more likely to know precisely where and when to look to identify unique regulatory mechanisms in the two progenitor populations.

EXPERIMENTAL PROCEDURES

Flow Cytometry

Preparation of single cell suspensions from embryonic cranial and trunk domains.

Pregnant ICR strain mice were purchased from Japan SLC, Ltd. (Shizuoka, Japan). Noon of the day a vaginal plug was observed was considered to be 0.5 days post coitum (E0.5). At E8.5, embryos were dissected and those with three to eight somites were selected (ca. 100 embryos). Older embryos that had already undergone cranial neural fold fusion were excluded. The cranial region (the region of neural folds and branchial arches rostral to the heart; see Fig. 2a), and a segment of trunk containing posterior somites and the caudal neural folds were dissected and collected separately. These tissues were dissociated by using a combination of Dispase (Gibco) and cell dissociation buffer (Gibco) as previously described for embryonic endothelial cell dissociation (Nishikawa et al., 1998). To allow surface antigens to regenerate, dissociated cells were incubated for 30 min at 37°C in Alpha Minimal Essential Medium (αMEM), containing 10% fetal calf serum (Gibco). Cells were then washed and resuspended in Hanks' buffered salt solution (HBSS) containing bovine serum albumin.

Antibodies.

Dissociated cells were exposed to rat anti-mouse PDGFRα monoclonal antibody (APA5, Takakura et al., 1997) conjugated to allophycyanin (APC), and rat anti-mouse E-cadherin monoclonal antibody (ECCD2, Takakura et al., 1996) conjugated to fluorescein isothiocyanate (FITC). Alternatively, cells were exposed to biotinylated APA5 and subsequently fluorescently labeled with streptavidin-conjugated APC (SA-APC; Molecular Probes).

Flow cytometry.

Cell suspensions from cranial and trunk tissue domains of E8.5 embryos were incubated with APC-labeled APA5, and ECCD2-FITC, for 20 min at 4°C. Labeled cells were then washed and re-suspended in HBSS containing propidium iodide to identify and exclude dead cells. Fluorescent cell sorting and analysis was performed using the Vantage FACS system (Becton Dickinson, San Jose, CA).

Immunostaining

Whole-mounts.

Embryos were fixed for 10 min in 4% paraformaldehyde at 4°C, washed in several changes of phosphate buffered saline (PBS), and dehydrated and stored in 100% methanol until use. Before whole-mount staining, embryos were rehydrated and immunostained by using previously published methods (Wakamatsu and Weston, 1997; Wakamatsu et al., 1998; detailed methods available upon request).

Histologic sections.

Immunostaining of mouse embryos fixed as above was performed with APA5 and ECCD2 (see above). Because both of these antibodies are produced by rat hybridomas, the following protocol, adapted from Shindler and Roth (1996), used tyramide signal amplification (TSA) for double-staining: dehydrated embryos were treated with 0.3% hydrogen peroxide in absolute methanol, then rehydrated, rinsed in PBS, and blocked with TSA kit blocking reagent in PBS for 1 hr at room temperature (as described in the TSA kit; Dupont-NEN, Boston, MA). Embryos were incubated overnight in highly diluted APA5 (1:27,000) antibody in PBS+TSA blocking reagent. Embryos were washed by gentle rocking 3× for 60 min each in PBST (PBS + 0.05% Tween-20). Embryos blocked again for 1 hr at room temperature and then incubated with secondary antibody (goat anti-rat HRP, 1:500) in PBS + TSA blocking reagent overnight at 4°C. After washing as above, Tyramide working solution (prepared according to TSA kit instructions) was applied to the samples for no more than 3 min. After washing 3× for 10 min each in PBST, the embryos were again blocked and then incubated overnight with ECCD2 (1:500 in PBS+ TSA blocking reagent), followed by extensive washing, reblocking, and incubation overnight with second secondary antibody (goat anti-rat Alexa488, 1:200 in PBS + TSA blocking reagent). After extensive washing as above, embryos were stored in PBST for photographing as whole-mounts and subsequent sectioning.

Acknowledgements

The original work reported here was initiated while J.A.W. was a Visiting Professor in the Graduate School of Medicine, Kyoto University Medical School. J.A.W. thanks the Department of Molecular Genetics for their generous hospitality and friendship. He also thanks the students and colleagues, too numerous to mention, for their indulgence and help during the gestation of the ideas contained in this study. He is especially indebted to Judith Eisen, Carol Erickson, Charles Kimmel, Andrew Lumsden, Angela Nieto, and Jean Paul Thiery, who offered their stimulating insights, constructive criticisms, and always useful, if sometimes ignored, skepticism, at various stages in the elaboration of these ideas. They have all made the process truly joyful. This study is dedicated to the memory of J. P. Trinkaus, who pointed the way (see Trinkaus, 2003)