A nonneural epithelial domain of embryonic cranial neural folds gives rise to ectomesenchyme

Abstract

The neural crest is generally believed to be the embryonic source of skeletogenic mesenchyme (ectomesenchyme) in the vertebrate head and other derivatives, including pigment cells and neurons and glia of the peripheral nervous system. Although classical transplantation experiments leading to this conclusion assumed that embryonic neural folds were homogeneous epithelia, we reported that embryonic cranial neural folds contain spatially and phenotypically distinct domains, including a lateral nonneural domain with cells that coexpress E-cadherin and PDGFRα and a thickened mediodorsal neuroepithelial domain where these proteins are reduced or absent. We now show that Wnt1-Cre is expressed in the lateral nonneural epithelium of rostral neural folds and that cells coexpressing Cre-recombinase and PDGFRα delaminate precociously from some of this nonneural epithelium. We also show that ectomesenchymal cells exhibit β-galactosidase activity in embryos heterozygous for an Ecad-lacZ reporter knock- in allele. We conclude that a lateral nonneural domain of the neural fold epithelium, which we call “metablast,” is a source of ectomesenchyme distinct from the neural crest. We suggest that closer analysis of the origin of ectomesenchyme might help to understand (i) the molecular-genetic regulation of development of both neural crest and ectomesenchyme lineages; (ii) the early developmental origin of skeletogenic and connective tissue mesenchyme in the vertebrate head; and (iii) the presumed origin of head and branchial arch skeletal and connective tissue structures during vertebrate evolution.

In vertebrate embryos, dorsal neural epithelium of the neural tube (NT) undergoes an epithelium-to-mesenchyme transition (EMT) (1) to produce neural crest (NC) cells. These cells then disperse in embryonic interstitial spaces and eventually give rise to pigment cells and neurons and glia of the peripheral nervous systems (PNS). At cranial levels of amphibian, avian and mammalian embryos, the cells derived from neural fold (NF) epithelium also give rise to these derivatives. In addition, however, avian and mammalian cranial NFs, like the NFs at all axial levels of amphibian embryos, give rise to some connective tissue cells and skeletogenic mesenchyme [“ectomesenchyme” (EM)]. It is generally believed that the NC produces EM and that the crest at trunk axial levels of avian and mammalian embryos has lost the ability to do so (ref. 2; see also refs. 3 and 4).

However, Weston et al. (5) have challenged this idea, noting that mouse cranial NF epithelium is heterogeneous and that a sharp boundary exists between lateral E-cadherin⁺ (Ecad⁺) nonneural epithelium and the thickened E-cad⁻ neural epithelium (NE) in the dorsomedial ridge (see also refs. 6 and 7). Moreover, a subpopulation of E-cad⁺ cells in the lateral nonneural epithelium coexpresses PDGFRα, which is a well established marker of mesenchyme and connective tissue in somites and in the head and branchial arches (BA) (8–10). They suggested that EM arises from this PDGFRα⁺/E-cad⁺ NF epithelium in vivo. Our present results support the idea that the source of EM is not restricted to the dorsal NE and have allowed us to define the general location of an epithelial domain in the murine cranial NFs from which some EM originates.

Results

Cre-Recombinase Is Expressed in Lateral Neural Fold Epithelium Before EMT in Ht-PA-Cre Transgenic Embryos.

In transgenic mouse embryos expressing Cre-recombinase (Cre) under the control of the human tissue plasminogen activator promoter (Ht-PA-Cre/R26R), cells exhibiting β-galactosidase (β-gal) activity appear precociously in BA, frontonasal process, and periocular mesenchyme before any marked cells appear in cranial NF epithelium (Fig. 1). To reconcile this early appearance of labeled mesenchyme with its presumed NC origin, it was suggested (11) that some crest cells might disperse before the transgenic marker was activated. In early [embryonic days (E)8.0–E8.5] transgenic embryos, however, we observed Cre protein in cell nuclei both in the nonneural epithelium of the cranial NFs and in underlying mesenchyme (Fig. 2). Therefore, we conclude that Cre-expression precedes or at least is coincident with the delamination of labeled cells from the nonneural epithelium. These results are consistent with the prediction in ref. 5 that at least some EM originates precociously from a source other than NC.

Fig. 1.

β-galactosidase⁺ mesenchymal cells appear precociously in the branchial arches. (A–C) X-Gal staining of E8-E8.5 Ht-PA-Cre/ROSA26 embryos at the 4-somite (A), 8-somite (B), and 10-somite (C) stages. (D) High magnification of outlined region in C, showing β-gal⁺ epithelial cells lateral to the dorsal ridge of the NF. Labeled cells are present in the BA (white arrowheads), in the frontonasal process (red arrowhead), and around the optic pit (black arrowheads) before any marked cells are present in the NE. The dorsal ridge of NE, comprising the mediodorsal domain of the NF (the NC), is indicated by dashed lines.

Fig. 2.

Ht-PA-Cre-recombinase is expressed in the nonneural cranial neural fold epithelium. (A–C and E–G) Double-staining of Cre and E-cad in cranial NF sections of E8 (4–5 som) Ht-PA-Cre transgenic embryos. Red arrowheads indicate cells of the nonneural epithelium that coexpress Cre and E-cad. White arrowheads mark instances of Cre-IR mesenchyme that also express weak cytoplasmic E-cad staining. (C and G) (Inset) Cells of the nonneural epithelium that coexpress Cre and E-cad cells shown at higher magnification. (F and G) Blue arrowheads show cells that appear to be down-regulating cell-surface Ecad during EMT, and intermingling with the neural cells. (D–H) Low-magnification DAPI-stained sections shown for orientation. Planes of section are indicated at right by lines (solid or dashed) on schematic of E8 embryos. Red segments on these lines indicate the location of the double-positive nonneural epithelial cells seen in both left and right NF domains, respectively, lateral to the dorsal ridge.

Wnt1-Cre Is Expressed in NonNeural Epithelium of Embryonic Cranial Neural Folds.

Previous cell labeling studies, using Wnt1-Cre transgenic embryos report that both BA mesenchyme and trunk NC derivatives express β-gal in E9.5 and older Wnt1-Cre/R26R transgenic embryos (11, 12). Because Wnt1 is generally assumed to be expressed only in dorsal neural tissue (13, 14), it is widely regarded as a definitive marker for NC-derived cells (e.g., ref. 15). When we examined Cre expression in the NFs of E8 (4–7 somite) transgenic embryos carrying Wnt1-Cre, we confirmed that Cre-immunoreactive (IR) cells are present in the dorsal neural epithelium of the NFs at and caudal to the Vagal axial level [data not shown; see supporting information (SI) Fig. S1]. However, in the fore- and midbrain NFs of these embryos, most Cre-IR was observed in the nonneural (E-cad-IR) epithelium lateral to the dorsal ridges and in underlying mesenchymal cells (Fig. 3). These results contradict the notion that Wnt1-Cre-IR cells arise solely from dorsal neural epithelium in these early embryos (see refs. 12, 15, and 16) and mitigate the inference that dorsal neuroepithelium-derived NC cells are the sole source of EM in vivo.

Fig. 3.

Wnt1-Cre-recombinase is expressed in the nonneural cranial neural fold epithelium. (A–C and E–G) Double-staining of Cre and E-cad in cranial NF sections of E8 (4–5 som) Wnt1-Cre transgenic embryos. Red arrowheads indicate instances of the nonneural epithelial cells (C and G) coexpressing Cre and E-cad. White arrowheads indicate Cre-IR mesenchyme with weak cytoplasmic E-cad staining. (C and G) (Inset) Cells of the nonneural epithelium that coexpress Cre and E-cad cells (indicated by asterisks on the arrowheads) shown at higher magnification. (D and H) Red segments indicate the location of the double-positive nonneural epithelial cells in the plane of section.

Nonneural Epithelial Cells and Underlying Ectomesenchyme Are Labeled in Embryos Where lacZ Is Under the Control of the E-cadherin Locus.

To confirm that EM cells are derived from E-cad-expressing epithelium, we examined E8 transgenic embryos in which lacZ is expressed under the control of the regulatory sequences of the E-cad locus, which are known to recapitulate endogenous E-cad expression patterns (17). These embryos reveal that β-gal is, or had recently been active in both nonneural epithelium of cranial NFs and in underlying EM cells (see Fig. 5 A–H and Fig. S2). We infer that these mesenchyme cells must have originated in the nonneural epithelium. In addition, however, we occasionally observed a few labeled cells within neural epithelium of E.8.5 Ecad-lacZ knockin embryos (data not shown, but see Discussion), suggesting that these cells also originate from the E-cad⁺ (nonneural) epithelium.

Fig. 5.

E-cadherin promoter activity is revealed in both nonneural epithelium of the cranial neural folds and the underlying mesenchyme. (A and B) E8 (4–5 som) Ecad-lacZ embryos exhibit β-gal activity in domains of the cranial NFs (white arrowheads) that are distinct from the NE. (C and E) In sections of such embryos, β-gal activity (blue cytoplasm or inclusions) is present in nonneural epithelial cells of the head (red arrowheads) and in underlying mesenchyme (white arrowheads and Insets). (G and I) PDGFRα staining (brown) performed on X-Gal-stained Ecad-lacZ embryos reveals that β-gal⁺ cells in the nonneural epithelium and underlying mesenchyme coexpress PDGFRα. Instances of such double-positive mesenchymal cells are indicated by white arrowheads. (Insets) Magnified cells in insets denoted by asterisks. (D, F, H, and J) Low magnifications and planes of the sections for C, E, G, and I, respectively. Red segments show the location of β-gal⁺ or β-gal⁺/PDGFRα⁺ cells observed in the head mesenchyme.

Cells Derived from Nonneural Epithelium also Coexpress PDGFRα.

BA mesenchyme persistently expresses PDGFRα⁺, and requires its function for normal development of EM derivatives (8–10). Therefore, if EM originated from nonneural epithelium (see refs. 5 and 10), we would expect that (i) some cells in the NF epithelium of Wnt1⁻ and Ht-PA-Cre transgenic embryos would express both PDGFRα and Cre, (ii) some of these PDGFRα⁺ cells would also coexpress E-cad, and (iii) β-gal-expressing cells in Ecad-lacZ knockin embryos would express PDGFRα. These predictions were verified: Cre/PDGFRα double-positive cells were present in both the nonneural epithelium of NFs and in underlying mesenchyme of E8 Ht-PA-Cre (Fig. 4 A–D) and Wnt1-Cre (Fig. 4 E–H) embryos. Likewise, labeled nonneural epithelium and mesenchyme in Ecad-lacZ knockin embryos coexpressed PDGFRα (Fig. 5 G and I). Thus, it seems likely that skeletogenic (PDGFRα⁺) mesenchymal cells delaminate from this nonneural domain of the NF.

Fig. 4.

Nonneural epithelial cells in the cranial neural fold express PDGFRα. Double-staining of Cre (A, C, E, and G) and PDGFRα (B, C, F, and G) in Ht-PA-Cre (A–D) and Wnt1-Cre (E–H) transgenic embryos. PDGFRα-IR is present as sparse punctate staining at cell surfaces. Red and white arrowheads indicate cells of the nonneural epithelium or underlying mesenchyme, respectively, expressing both Cre and PDGFRα. (C and G) Dashed lines indicate the boundary of E-cad⁻ NE. (D and H) Red segments on the lines show the plane of section and the location of the double-positive nonneural epithelial cells.

Discussion

A Lateral Nonneural Domain of Cranial Neural Folds Produces Ectomesenchyme.

Our results have allowed us to map the general location of a lateral nonneural epithelial domain in the rostral NF of early murine embryos (Fig. 6A, green shading) that produces ectomesenchyme. This so-called metablast domain (5) is spatially separate from the locations of neurogenic placodes (Fig. 6A, blue shading; see also ref. 18). The metablast epithelium is also phenotypically and developmentally distinct from the mediodorsal NE, which is believed to give rise to NC-derived melanogenic and neurogenic cells at all axial levels. We suggest that the metablast epithelium produces a population of PDGFRα⁺mesenchymal cells, presumably skeletogenic EM, that was previously thought to originate from the NC.

Fig. 6.

Schematic representation of cranial neural fold complexity during the morphogenetic events leading to neural tube formation and EMT. (A) Based on combined staining of all transgenic embryos, the green shaded area on the diagram suggests the general location of the inferred metablast domain within the NF epithelium. Note that this domain is distinct, at this developmental stage, from the locations of cranial placodes (blue shaded areas) that represent later sources of dispersing cells in the head. n, nasal placode; t, trigeminal placode; o, otic placode; e, epibranchial placodes. (B) Schematic summary of EMT events in metablast and neural crest epithelium. As inferred from immunostaining results, the NF epithelium has distinct domains with sharply defined boundaries: The E-cad⁺, nonneural cells of NF epithelium are represented by red circles. PDGFRα-IR cells within, and delaminating from, the E-cad⁺ NF epithelium are represented by green circles. NC cells, which arise in and begin to delaminate from the mediodorsal NE domain, are represented by blue circles. In these diagrams, PDGFRα⁺ cells in the nonneural epithelium overlying and lateral to the NE (a) delaminate precociously into underlying cell-free interstitial spaces (b). This nascent EM, like somite-derived sclerotomal mesenchyme, then establishes the interstitial environment into which neural crest cells delaminate from the dorsal NE (c). As NFs fuse in the midline to form the NT (d), nonneural (E-Cad⁺) epithelium (red) separates from the dorsal NE of the nascent NT (blue), EM (green) occupies the interstitial spaces between neural and nonneural epithelia, and NC cells (blue) intermingle with and disperse ventrolaterally among the EM cells, similar to trunk crest cell dispersal among sclerotomal mesenchyme.

We have also observed both Ecad-lacZ activity and PDGFRα-IR in a few cells within the thickened NE of the NF (see above). Likewise, weak cytoplasmic E-cad-IR is visible in some of the Cre-IR cells in marginal NE and in underlying EM (see Figs. 2 B and F and 3 B and F). Based on these observations, we suggest that such E-cad⁺ cells within the NE also originate from nonneural epithelium, some of which overlaps the NE margin, and that these cells transiently intermingle with cells in the NE after down-regulating E-cad from their surfaces during EMT (see ref. 5).

Is the Metablast a Subdomain of the Neural Crest?

In much of the current literature, the cranial NF epithelium and the NC are assumed to be equivalent. Accordingly, what we have called metablast might be considered a subdomain of the NC and not distinct from it. The fact that cells of this NF epithelium share the expression of conventional “neural crest markers” (e.g., Snail gene-family members, Foxd3, and Msx1) is certainly consistent with this idea. However, because many of these genes are actually widely expressed in a variety of cell types before specific morphogenetic events like EMT (see ref. 5), their use as specific markers of NC-derived cells could be misleading.

Despite these reservations, localized gene expression patterns clearly do distinguish the two epithelial domains in mouse cranial NFs: In the nonneural epithelial domain, E-cad and PDGFRα, a marker characteristic of mesodermally derived connective tissues (9, 10), are coexpressed at cell surfaces, whereas NE and eventually neuro/gliogenic lineages of the PNS exclusively express Sox1 (19). The metablast and NC domains of the cranial NF epithelia also differ in a number of other ways. Thus, part of the lateral nonneural epithelial NF has been reported to express a proteoglycan link protein (20) and transiently, β₃-integrin (21), both of which are characteristic of skeletogenic cells, and neither of which is expressed by dorsal NE or trunk NC-derived cells in mouse embryos. Therefore, because cells in the metablast domain express phenotypic traits characteristic of cells that will produce skeletal/connective tissue, and, because some of these cells delaminate precociously to produce a population of PDGFRα⁺ EM cells, we consider it likely that some are the precursors of craniofacial skeletogenic and connective tissues. Although we acknowledge that the ultimate fates of the metablast-derived cells remains to be determined, it is important to emphasize that different genetic regulatory pathways appear to operate in the cell populations derived from the two epithelial domains and therefore that it would probably be appropriate to analyze the mesenchymal cells that emerge from them as precursors of distinct lineages.

Can Cells Delaminating from Metablast Epithelium Account for All of the Ectomesenchyme That Appears in Developing Branchial Arches?

It is impossible to estimate the number of BA mesenchyme cells that delaminate from the entire metablast domain by sampling individual histological sections of embryos at early developmental stages. It is likely, however, that these mesenchyme cells are continuously recruited from the metablast epithelium as the neural folds elevate and fuse during cranial development (see ref. 22). Consequently, the cells illustrated in the accompanying figures would represent only a very small fraction of the total number that delaminate during this morphogenetic process. The dramatic growth of the BAs (22) clearly must also involve remarkable rates of cell proliferation, both in the nonneural epithelium and within the mesenchyme itself (ref. 23; see also refs. 24 and 25).

Ectomesenchyme Provides the Conditions Necessary for Neural Crest Cell Dispersal.

The first cells to emerge in outgrowths of explanted cranial NFs are reported to produce fibronectin, whereas cells that emerge later from the same explants, and NC cells cultured from trunk NT explants do not do so (26). Because NC cells apparently require fibronectin-containing extracellular matrix as a substratum for migration, the precocious emergence of EM in the cranial region could establish the conditions necessary for subsequent dispersal of NC-derived cells in the same region. These ideas are summarized in a schematic diagram (Fig. 6B), showing our inferences that EM emerges precociously by delamination from the lateral nonneural domain of cranial NF, and that NC cells subsequently enter the interstitial space where they comingle with EM and disperse in the matrix it produces. In this sense, EM in the head would serve the function of somite-derived sclerotome cells in the trunk of amniote embryos to provide appropriate substrata for NC cell migration.

Assumptions Underlying Classical Grafting Experiments Probably Led to Their Misinterpretation.

Pioneering experiments in amphibian embryos suggested that the entire range of “NC-derived” phenotypes, including EM, arises from grafted tissues (see ref. 3). It is important to note, however, that in all these experiments, the entire NF was marked and transplanted. Because NFs contain phenotypically distinct epithelial domains, the results of these grafting experiments cannot exclude the possibility that EM originates from a domain of the NF that is distinct from the NC epithelium.

Although not widely acknowledged in the literature, but suggested by E-cad staining patterns (e.g., see Fig. 2), it is impossible to dissect the neural from nonneural epithelia in early embryonic NFs. Consequently, the cranial NF grafts in avian embryos would also contain both mediodorsal (neural) and lateral (nonneural) tissues. Accordingly, we suggest an alternative explanation for the widely accepted inference that skeletogenic ability is lost by trunk neural crest cells in avian embryos (2). These authors reported that when cranial NFs were grafted, EM derivatives of donor origin were present in avian host embryos. In contrast, when only NT was transplanted in the trunk, hosts contained donor-derived melanocytes and components of the PNS, but lacked donor-derived EM. Based on our present results, we suggest that the apparent difference in developmental potential between cranial and trunk grafts is most parsimoniously explained by differences in experimental protocol: NF and NT were grafted at head and trunk axial levels, respectively, and only the NF, which contains a nonneural epithelial domain, would include the precursors of donor-derived EM.

Reports that skeletogenic subpopulations are present in long-term mouse trunk neural crest cell cultures (27) might appear to be inconsistent with our conclusions about metablast-derived mesenchyme in vivo. In this work, however, NFs from early (E8, 4–7 somite) embryos appear to have been explanted before neural and nonneural epithelial domains had separated during midline fusion of the folds. As mentioned above, we have also occasionally observed E-cad⁺ cells within NE of younger Ecad-lacZ knockin embryos. It is possible, therefore, that skeletogenic cells and reported mesenchyme stem cells within cultured trunk crest cell populations (19) might also originate from nonneural epithelium of the NF.

It Will Be Productive to Distinguish the Two Neural Fold Domains in Future Experimental Analyses.

There are several additional reasons to consider NC and metablast to be developmentally distinct epithelial domains within the NFs: First, if metablast and NC represent spatially and developmentally distinct embryonic tissues, it would be unnecessary to postulate regulatory mechanisms at this stage of development to account for the loss of skeletogenic potential by trunk neural crest cell precursors (see ref. 2 and 4), nor would it be necessary to assume the existence of, and therefore to strive to identify, gene regulatory pathways that operate in a common NC “stem cell” population to produce both neurogenic and skeletogenic cells in vivo. Although it has been argued (25, 28) that in vitro cloning studies support the idea that skeletogenic and neurogenic lineages share a common neural crest-cell precursor, we must reiterate that most of these cloning experiments suffer significant technical limitations (29), and that the published evidence from clonal analyses for such common precursors (e.g., ref. 24) is, at least, susceptible to alternative interpretations (see ref. 5). Consequently, the assumptions underlying these clonal analyses, and their relevance to events in vivo, might beneficially be revisited.

Second, if it were acknowledged that two distinct epithelial domains exist in NFs, the need to define NC markers more precisely would be compelling. As previously mentioned, for example, the expression of genes like Wnt1 (and also Snail family members; e.g., ref. 30–32) is unlikely to indicate NC origin because their gene products are present in both mediodorsal and lateral NF epithelia. Thus, these expression patterns appear to be spatially and temporally too general to be used to infer embryonic origins.

Third, it is now widely accepted that the “invention” of NC played an important role in the evolution of the vertebrate head (see ref. 33). However, as we suggest, connective tissue and skeletogenic mesenchyme in the vertebrate head could have a distinct developmental origin from that of peripheral neurons, glia, and pigment cells. If this were acknowledged, even tentatively, it would provide an incentive to consider heuristic alternatives to the proposed role of the NC in vertebrate evolution. The discovery of “NC-like” migratory cells in ascidians (34), recently understood to be the vertebrate sister group (35), adds support to our argument. Some of these ascidian cells produce pigment cells, but not other derivatives usually attributed to the neural crest, notably skeletogenic cells. As these authors point out, other migratory cell types may have joined with NC-like cells to account for the appearance of such new vertebrate derivatives.

Finally, in light of recent reports of the remarkably expanded and convergent dermal and chondral skeletogenic abilities for both mesodermal and putative NC-derived mesenchyme (15), it now seems appropriate to explore the possibility that they have a common developmental origin, distinct from the NC, and to consider where and when the so-called metablast epithelium might originate. Based on earlier intriguing reports (36, 37), we suggest a previously unsuspected contribution to cranial mesenchyme from ectopic mesoderm-like cells: Results reported by these authors indicate that such cells could (i) originate in the early epiblast before the formation of the primitive streak, (ii) initially evade incorporation into the nascent primitive streak and subsequently remain as a distinct population within ectodermal epithelium during gastrulation, and (iii) eventually undergo delayed involution in NF epithelia through transient NF structures whose morphology resembles the primitive streak (see ref. 38). Likewise, as noted in ref. 5, EM derivatives were absent when NC was experimentally induced in nonneural epithelium by interaction with NE (39, 40). This also provides support for the suggestion that metablast and NC epithelia have different developmental origins.

Ultimately, it will be important to follow the fate of epithelial cells in the putative metablast domain of the rostral NF. Precise marking of individual cells in the metablast epithelium would definitively test the prediction that they are the source of skeletogenic and connective tissue lineages in the head and face of vertebrate embryos. Such experiments are feasible in avian, amphibian, and zebrafish embryos.

Methods

Transgenic Embryos.

Mice carrying Cre-recombinase under the control of a promoter of human tissue plasminogen activator (Ht-PA-Cre) (11) were mated with the β-gal reporter strain ROSA26 (R26R) (41). E8–E9 transgenic embryos used for whole-mount X-gal staining (n = 15) were removed in cold PBS plus 5% FCS and fixed briefly in cold 1% formaldehyde, 0.2% glutaraldehyde, and 0.02% Nonidet-P40. For immunostaining, E8 Ht-PA-Cre embryos (11) (n = 7) and embryos from mice carrying Wnt1-Cre (14) [founders kindly provided by A. McMahon (Harvard University, Cambridge, MA)] (n = 7), were removed in cold PBS, and fixed in cold 4% PFA for 1.5 h. After washing two times in PBS, fixed embryos were incubated successively in 12% (1.5 h), 15% (2 h) and 18% (overnight) sucrose solutions, then embedded in Tissue-Tek matrix, frozen, and sectioned at 7 μm on a Leica cryostat microtome before staining. Heterozygous Ecad-lacZ mice (Ecad-In2flox) (17) were crossed to C57BL/6 females to obtain E8 (4–5 somite) embryos (n = 24). Embryos were fixed in 1% PFA and, after whole-mount staining, dehydrated in an alcohol series, embedded in wax, and sectioned at 7 μm. Sections of embryos stained for X-gal only (n = 13) were counterstained with FastRed (Vector).

Staining Procedures.

β-galactosidase activity was revealed in whole-mount preparations as described in refs. 17 and 42. Antibodies used for immunostaining included rat anti-mouse PDGFRα (APA5) (5 μg/ml) (10), rat anti-mouse E-cad (ECCD-2) (10 μg/ml) (43) and rabbit anti-Cre recombinase (1/1,000) (PRB-106C; Covance; lot no. 135028002). Sections were incubated for 2 h in blocking solution consisting of 10% FCS, 0.1% Triton X-100, and 0.5× Blocking Reagent (Roche) in PBS. The sections were then incubated overnight at 4°C with primary antibodies in blocking solution, rinsed several times in PBS, and incubated with secondary antibodies (for APA5 and ECCD-2, goat anti-rat Alexa Fluor 594; for anti-Cre, goat anti-rabbit Alexa Fluor 488) for 2 h at room temperature in the dark. Whole-mount APA5 staining was performed as described in ref. 10.