Cadherins compose a large family of calcium-dependent cell surface molecules involved in the maintenance of tissular cohesion by mediating cell adhesion and cell recognition (Tepass et al., 2000; Yagi and Takeichi, 2000). Among them, classical cadherins are structurally related and share in common several basic functional features. Their extracellular domain is implicated in cis- and trans-binding to other cadherin molecules and to growth factor receptors, and their intracellular part is connected with the actin network by catenins. At the cellular level, classical cadherins are concentrated in adherens junctions where they provide a mechanical link between the cell membrane and the cytoskeleton and activate a variety of signaling cascades regulating cytoskeletal dynamics and intracellular trafficking (Braga, 2002; Perez-Moreno et al., 2003; Nelson and Nusse, 2004).
Classical cadherins are expressed by almost every cell type of the body, including epithelia, neurons, fibroblasts, and even in primordial germ cells, and they exert profound and varied effects on cell behavior and tissue organization both during embryonic development and in the adult (Halbleib and Nelson, 2006). In polarized epithelia, in the nervous system, and in muscles, they are able to support stable contacts between cells, thereby ensuring strong cohesion among them, and also contribute to the establishment of cell polarity and formation of barriers between tissular compartments. Conversely, in mesenchymal tissues or in undifferentiated cells, cadherins participate in dynamic morphogenetic events and mediate labile cell contacts allowing cell recognition, sorting, and migration. The differences in the ability of cells to establish either firm, stable, or labile contacts reside primarily in the type of cadherin expressed and in its prevalence on the cell's surface (Patel et al., 2003). Epithelial cells typically express E-cadherin whereas mesenchymal cells express various other cadherins such as N-cadherin, R-cadherin, cadherin-7, or cadherin-11. Consistent with this, biophysical studies demonstrated that E-cadherin mediates stronger bonds at the molecular level than any other cadherins (Dufour et al., 1999; Chu et al., 2006; Panorchan et al., 2006). Moreover, although E-cadherin and N-cadherin may transiently coexist in certain cells, they are generally mutually exclusive, and studies on epithelial cell lines or in mouse and Xenopus embryos in which E- and N-cadherins have been interchanged clearly indicate that they can only partially substitute each other (Luo et al., 2001; Kan et al., 2007; Wheelock et al., 2008; Nandadasa et al., 2009; Libusova et al., 2010).
Throughout life, cells do not maintain a constant repertoire of cadherins; rather they may execute successive changes in cadherin expression in relation to reorganizations of their interactions with their neighbors. In particular, the switch between E- and N-cadherin has been found to play a key role in early morphogenesis and in pathological situations, by promoting the process of epithelium-to-mesenchyme transition (EMT; Wheelock et al., 2008; Thiery et al., 2009). During gastrulation, for example, N-cadherin is up-regulated at the expense of E-cadherin on the surface of nascent mesodermal cells ingressing from the superficial ectoderm into the primitive streak (Ohta et al., 2007; Nakaya et al., 2008; Hardy et al., 2011). Similarly, dissemination of carcinoma cells from primary epithelial tumors is often associated with E-cadherin loss combined with N-cadherin expression, and inappropriate N-cadherin up-regulation in tumor epithelial cells has been shown to promote motility and invasion (Thiery et al., 2009).
There are examples, however, where the E- to N-cadherin switch does not result in deterioration of the epithelial structure and EMT. This is, in particular, the case of the early morphogenesis of the nervous system in vertebrates. In the chick embryo, the neural plate, from which the entire central nervous system is formed, is initially a flat epithelium of cuboidal cells that express E-cadherin and not N-cadherin. At completion of neurulation, when the neural tube is formed and fully separated from the superficial ectoderm, neural epithelial cells exhibit their typical elongated, radially-oriented shape and constitute a pseudo-stratified epithelium in which E-cadherin has been totally replaced by N-cadherin (Thiery et al., 1984; Hatta et al., 1987; Duband et al., 1988; Detrick et al., 1990; Radice et al., 1997). Yet, neural epithelial cells are unable to undergo spontaneous EMT; instead, they tend to remain coherent, even in two-dimensional culture where they retain their organization as a hollow tube (Duband et al., 2009). This E- to N-cadherin switch during neurulation has often been interpreted as an event necessary both for the acquisition of the morphological features of neural epithelial cells and for the movements of neurulation to occur. However, despite a variety of experiments in different animal models that uncovered the importance of N-cadherin in early neural development (Bronner-Fraser et al., 1992; Radice et al., 1997; Lele et al., 2002; Hong and Brewster, 2006; Nandadasa et al., 2009), these hypotheses have never been tested experimentally, and the biological significance of the change in cadherin expression during neurulation remains, therefore, unclear. In particular, it is not known why, contrary to epithelial cell lines established in vitro, E-cadherin loss in the neural epithelium does not provoke EMT. In other words, the question of the possible incidence of E-cadherin repression on the morphology of neural epithelial cells and on the elaboration of the neural epithelium has never been addressed directly. Moreover, the timing of the occurrence and the kinetics of the switch during the course of neurulation have not been established with precision, thereby rendering the interpretation of functional experiments uncertain. Finally, while many studies focused on the characterization of the transcriptional regulators involved in cadherin regulation during neural crest cell emigration (Duband, 2006), the identity of the transcription factors controlling cadherin expression during neurulation is not known. In the present study, we have performed a complete survey of the expression patterns of E- and N-cadherin proteins and messengers, in relation to those of their main transcriptional regulators, in the early nervous system of the chick embryo, from neural induction to completion of neurulation. Our observations indicate that deployment of the program of E- to N-cadherin switch during neurulation is progressive and orchestrated by a limited set of transcriptional regulators distinct from those recruited during gastrulation, segregation of the neural crest and formation of the notochord, possibly accounting for the transformation of the neural epithelium without deterioration of cell cohesion. In addition, we found that the timing and kinetics of cadherin switch vary considerably along the embryonic axis, in a manner that appears neither in synchrony with the movements of neurulation nor connected with the acquisition by neural cells of their morphological traits, but may rather reflect the spatial segregation of cells of the neurectoderm and their final commitment into neural tissues.
RESULTS AND DISCUSSION
To investigate the timing and kinetics of E- to N-cadherin switch during early neural development, we analyzed the expression profile of their proteins and messages by immunofluorescence and in situ hybridization on serial sections of chick embryos as well as on whole embryos, between stages 3 and 19 of Hamburger and Hamilton (HH) (Hamburger and Hamilton, 1951), i.e., between the 12th and the 72nd hour of incubation. In parallel, we analyzed the distribution patterns of their main transcriptional regulators, i.e., Snail-1, Snail-2, Zeb-2 (also known as SIP-1), and Twist-1, acting as repressors of E-cadherin, and Sox-2, known as an activator of N-cadherin transcription (Nieto, 2002; Matsumata et al., 2005; Peinado et al., 2007). Sox-2 was also used as a marker for definitive neural tissues (Uchikawa et al., 2003, 2011; Linker and Stern, 2004).
Timing and Kinetics of E- to N-Cadherin Switch During Early Neural Induction
To determine first whether the switch from E- to N-cadherin during neural development occurs as a consequence of neural induction and is involved in the morphological transformation of neural epithelial cells (Takeichi, 1988; Halbleib and Nelson, 2006; Suzuki and Takeichi, 2008), we analyzed their expression patterns throughout the process of neural induction in the anterior region of the embryo. In the chick, neural induction starts during gastrulation at stage HH3 and becomes apparent at stage HH4 by expression of Sox-2 in the neural plate, in a horse-shoe-shaped area surrounding the Hensen's node (Uchikawa et al., 2003, 2011). At this stage (Fig. 1A, B), immunostaining for E-cadherin was strong on the whole surface of all superficial epithelial cells both in the nascent neural plate and in the non-induced epiblast. No E-cadherin was found in the mesoderm, but, as shown previously (Nakaya et al., 2008; Hardy et al., 2011), it was detectable in the primitive streak, in both ingressing and early migrating mesodermal cells. N-cadherin exhibited a reciprocal pattern to E-cadherin, with a strong expression in the mesoderm and a complete exclusion from the neural plate and the epiblast. In the primitive streak, it was codistributed with E-cadherin in the ingressing mesodermal cells. In situ hybridizations on whole embryos (Fig. 1C) revealed that E-cadherin messages were uniformly expressed in the superficial layer of the whole blastoderm, with a slight decrease in the primitive streak and neural plate. N-cadherin, in contrast, was conspicuous in the primitive streak and mesoderm, but neither in the neural plate nor the non-induced epiblast. Of interest, its expression pattern matched precisely that of Brachyury, a T-box transcription factor involved in mesoderm formation (Kispert et al., 1995). While they were not detected at stage HH3 (not shown, but see Sheng et al., 2003; Uchikawa et al., 2003, 2011), Sox-2 and Zeb-2 became both induced in conjunction at stage HH4 in the neural plate. Snail-2, in contrast, was absent from this region, being essentially confined to the primitive streak (see also Sefton et al., 1998; Acloque et al., 2011). Finally, among the other transcriptional repressors of E-cadherin tested, neither Twist-1 nor Snail-1 was expressed in the neural plate during early neural induction as well as at later stages during neurulation (Figs. 1C, 2B–D; see also Sefton et al., 1998; Tavares et al., 2001).
By stage HH7 (Fig. 2A, B), during formation of the head fold, while E-cadherin remained conspicuous in the entire neural plate, expression of N-cadherin transcripts was initiated in its anterior-most portion. However, no N-cadherin protein could be detected in neural epithelial cells by immunofluorescence at this stage. At stage HH8 (Fig. 2C), N-cadherin expression became significantly more pronounced in the anterior neural plate up to the first somites, and this was accompanied by gradual accumulation of the protein on the cell surface (see Fig. 4A–C). In contrast, E-cadherin persisted in the anterior neural plate until stage HH10 (Fig. 2C, D), and its complete repression was observed only at stage HH11 (see Figs. 4, 5). Between stages HH4 and HH8, Sox-2 and Zeb-2 expressions were markedly increased in the entire neural plate ahead of the regressing node, with the notable exception of the notochord, while Snail-2 remained limited to the primitive streak as earlier.
Therefore, our results show that, at the onset of neural induction at stage HH3, cells of the prospective neural plate were undistinguishable from the rest of the epiblast in their shape and cadherin content, and expressed none of the known transcriptional regulators of E- and N-cadherins. During progression of neural induction at stage HH4, neural epithelial cells started to acquire their morphological traits, in particular their elongated, spindle-like shape. However, this was not accompanied by an immediate and significant modification in cadherin expression. Changes in cadherin expression occurred only gradually over time, first with increase in N-cadherin messages well after neural induction and formation of the neural plate, then with repression of E-cadherin even later (Fig. 3A). These observations fully support previous genetic studies in the mouse and Zebrafish showing that, in the absence of N-cadherin, onset of formation of the neural epithelium develops normally. The first signs of neural tube malformations are detected relatively late, in the mouse, by the appearance of undulations along the neural tube after closure (Radice et al., 1997) and, in the fish, by the absence of cavitation due to alterations in movements of convergence and intercalation (Hong and Brewster, 2006). Therefore, rather than a prerequisite to the morphological changes, cadherin switch appears merely as a secondary event during neurulation.
Of the transcriptional regulators of cadherin examined here, only Sox-2 and Zeb-2 appeared to exhibit a spatial expression pattern consistent with the cadherin switch in neural epithelial cells, as they were uniformly expressed throughout the entire neural plate and absent from the ectoderm. Previous genetic studies in the mouse showed that embryos lacking Zeb-2 exhibit persistent expression of E-cadherin in the neural tube, thereby revealing that Zeb-2 is necessary for E-cadherin repression in the neural epithelium (Van de Putte et al., 2003). On the other hand, misexpression of Sox-2 in the cranial ectoderm of early chick embryos has been found to cause ectopic N-cadherin expression, indicating that Sox-2 is sufficient to drive expression of N-cadherin in neurectodermal tissues (Matsumata et al., 2005). In addition, in agreement with the fact that both factors have been ascribed critical roles in neural induction in chick and Xenopus (Eisaki et al., 2000; Sheng et al., 2003; Uchikawa et al., 2003, 2011; Takemoto et al., 2006), we also found that Sox-2 and Zeb-2 are induced almost coincidently in neural plate cells at an early step of neural induction. These observations therefore suggest that Sox-2 and Zeb-2, in conjunction, are key players in the spatiotemporal control of the E- to N-cadherin switch in the nervous system. However, rather unexpectedly, we observed that induction of N-cadherin and complete extinction of E-cadherin occurred long after onset of expression of Sox-2 and Zeb-2 in the neural plate (Fig. 3A). While the 6-hr delay between Sox-2 and N-cadherin expression in the neural plate is consistent with the 4-hr delay reported previously for the sensory placodes (Matsumata et al., 2005), repression of cadherin expression more than 12 hr after induction of Zeb-2 suggests that, albeit necessary, this factor may not be sufficient to modify transcription of the E-cadherin gene in neural epithelial cells and that additional co-factors required for complete repression may be induced only at late stages during neural induction. Additionally, it cannot be excluded that post-transcriptional regulatory events may account for the delay between the timing of expression of Sox-2 and Zeb-2 and the changes in surface expression of E- and N-cadherin in neural epithelial cells. Indeed, both E- and N-cadherins may be subjected to post-transcriptional regulations, such as proteolytic cleavage or cytoplasmic trafficking (Halbleib and Nelson, 2006), that may locally modulate their activity. However, the overall consistent expression of E- and N-cadherin messages and proteins throughout neurulation suggests that cadherin activity in neural epithelial cells is primarily regulated at the transcriptional level. Moreover, we found no major difference in immunostainings using antibodies directed against the cytoplasmic tail or the external part of E- and N-cadherin (not shown).
The slow replacement of E-cadherin by N-cadherin in the neural plate contrasts strikingly with the situation found in the primitive streak, where ingressing mesodermal cells express Snail-2 and Brachyury and immediately execute a complete EMT program with breakdown of the basement membrane, loss of cell polarity, and a rapid E- to N-cadherin switch (Sefton et al., 1998; Nakaya et al., 2008; Acloque et al., 2011). Beside Snail-1, Snail-2 is considered as the prototype of EMT inducers. It is often aberrantly overexpressed in metastatic tumor cells and is commonly recruited in tissues normally undergoing rapid and massive EMT (Thiery and Sleeman, 2006; Peinado et al., 2007; Thiery et al., 2009). Moreover, functional studies in the chick have demonstrated that it is sufficient to induce ectopic cell delamination in the epiblast during gastrulation (Acloque et al., 2011). Concerning Brachyury, although it has been regarded essentially as an inducer of mesodermal identity, it is also necessary for mesodermal cell movements during gastrulation as shown by clonal analyses in the mouse embryo (Wilson and Beddington, 1997), and recent findings revealed that it can trigger EMT in epithelial cell lines both directly and indirectly through Snail-1/2 (Fernando et al., 2010). Thus, in the early chick gastrula, Snail-2 in possible conjunction with Brachyury may promote a rapid switch between E- and N-cadherin in primitive streak cells, associated with complete EMT. On the other hand, although Zeb-2 has been found to perform the same roles as Snail-1/2 in epithelial cells maintained in vitro (Thiery and Sleeman, 2006; Peinado et al., 2007; Vandewalle et al., 2009), our data show that, during neural induction, its expression does not correlate with events of EMT and previous functional studies revealed that, rather than inducing disintegration of the epithelial structure, its ectopic expression in the epiblast at the gastrula stage prevents EMT and cell movements through the primitive streak (Sheng et al., 2003). Moreover, Zeb-2 is also known to be a potent repressor of Brachyury (Papin et al., 2002; Sheng et al., 2003). Intriguingly enough, ectopic expression of Snail-2 in the neural tube at a time when it also expresses Zeb-2 is not sufficient to provoke EMT of neural epithelial cells (del Barrio and Nieto, 2002; Cheung et al., 2005; Théveneau et al., 2007). All together, these observations suggest possible antagonistic roles for Snail-2 and Zeb-2 in the control of EMT in neurectodermal cells, irrespective of their function as cadherin regulators. Thus, it can be hypothesized that Snail-2 and Brachyury, on the one side, and Zeb-2 and Sox-2, on the other side, may define two distinct territories in the chick blastoderm with divergent fates, but both characterized by a switch from E- to N-cadherin (Fig. 3B). Cells expressing Snail-2 and Brachyury may become fated to become mesodermal cells and undergo a rapid E- to N-cadherin switch followed by EMT and ingression through the primitive streak whereas cells expressing Sox-2 and Zeb-2 would acquire a neural identity and execute a slow transition between E- and N-cadherin while being prevented from losing their epithelial character and ingressing in the underlying space.
Timing and Kinetics of E- to N-Cadherin Switch During Primary Neurulation
To determine whether the switch from E- to N-cadherin in the neural epithelium correlates with the movements of neurulation and neural tube closure (Takeichi, 1988; Bronner-Fraser et al., 1992; Redies, 1995; Halbleib and Nelson, 2006), we studied their expression patterns during the whole process of neurulation at various defined positions along the neural axis, from the midbrain up to the posterior trunk. In the chick, the neural tube is formed by two independent mechanisms involving different morphogenetitc movements and termed primary and secondary neurulations (Colas and Schoenwolf, 2001; Lowery and Sive, 2004). Primary neurulation occurs in the head and upper trunk by rolling of the neural plate and formation of neural folds that then fuse dorsally, resulting in the typical hollow shape of the neural tube.
Analysis of the expression patterns of cadherins at different levels of the neural tube formed by primary neurulation revealed essentially two regions that differ in the timing and kinetics of the E- to N-cadherin switch: the first one encompasses the midbrain and anterior hindbrain and the second one extends from the posterior hindbrain to the anterior trunk. At the midbrain and anterior hindbrain levels, movements of neurulation start early at stage HH7. At this stage (Fig. 4A), both the neural plate and the ectoderm expressed uniformly high amounts of E-cadherin but no N-cadherin. During elevation of the neural folds at stage HH8 (Fig. 4B), while E-cadherin remained conspicuous over the entire cell surface of all neural epithelial cells, a faint, punctate staining for N-cadherin could be detected at the apical side of cells situated close to the midline (arrowhead in Fig. 4B). At stages HH8+ and HH9, during apposition of the neural folds (Fig. 4C, D), staining for N-cadherin became gradually stronger, but still concentrated at the apical side of cells, in the entire neural tube, with the notable exception of the prospective neural crest cells in the folds. Coincidently, staining for E-cadherin declined slowly and became markedly less intense than in the ectoderm. In contrast to N-cadherin staining, which propagated gradually from the midline up to the lateral (then dorsal) sides of the neural tube, E-cadherin staining decreased uniformly throughout the neural tube, with no obvious converse gradient from lateral to medial. After fusion of the folds and at the time of crest cell emigration from the neural tube at stage HH10 (Fig. 4E), N-cadherin staining became prominent over the entire surface of all neural epithelial cells. However, although it was completely closed and separated from the ectoderm, the neural tube still expressed detectable amounts of E-cadherin. Strikingly, neural crest cells expressed no N-cadherin but retained E-cadherin expression throughout delamination, as well as at the initial steps of migration. At stage HH11 (Fig. 4F), after completion of neural tube closure and neural crest cell emigration, E-cadherin staining has totally disappeared from the neural tube, which instead expressed uniformly high levels of N-cadherin.
In the posterior hindbrain and anterior trunk, movements of neurulation commence between stage HH8 and HH9 when the paraxial mesoderm is still unsegmented <zaq 4>(Fig. 5A, B). In this region, although the neural plate was initially organized as a flat epithelial layer in continuity with the ectoderm as in the midbrain, it clearly differed from the more anterior levels by the spatial distribution of cadherins over its surface. E-cadherin showed a latero-medial gradient in the neural plate with strong and uniform expression in the ectoderm and neural folds, weaker and apically-polarized expression in the intermediate neural tube, and a very faint, barely detectable expression in the median part. N-cadherin exhibited an almost exactly complementary pattern in the neural plate, being present medially, though at low levels, and absent laterally. Of note, unlike more anteriorly, codistribution of E-and N-cadherins was limited both spatially and temporally to cells of the intermediate portion of the neural tube. During elevation and apposition of the neural folds at stage HH10 (Fig. 5C, D), expression of E-cadherin disappeared almost entirely from the neural tube to become restricted to a few cells in the dorsal midline. Conversely, N-cadherin was expressed by all neural tube cells with the notable exception of the prospective neural crest population. By stage HH11 (Fig. 5E), after fusion of the neural folds and healing of the ectoderm, E-cadherin has entirely disappeared from all neural tube cells. Most remarkably, neural crest cells that have not yet started delaminating remained negative for N-cadherin. At stage HH12 (Fig. 5F), neural crest cell migration has just began and neither E-cadherin nor N-cadherin was expressed on their surface during delamination and early migration. As at more anterior levels, cessation of neural crest cell delamination was accompanied by a gradual increase in N-cadherin expression in the dorsal midline (not shown).
In situ hybridization analyses revealed that, at both head and upper-trunk levels, expression of E- and N-cadherin transcripts matched approximately the spatio-temporal profile of their corresponding proteins (Figs. 2, 6, 7). In addition, Sox-2 exhibited the same expression pattern as N-cadherin in cells derived from the neurectoderm: it was abundant in neural tube cells, but absent from the ectoderm and progenitors of the neural crest in the folds (see also Wakamatsu et al., 2004). As amply described previously (for a review, see Duband, 2006), Snail-2 was restricted to prospective neural crest cells until they became segregated entirely from the neural tube. Immunostaining using antibodies to Snail-2 gave essentially the same pattern (not shown). In contrast to Snail-2 and Sox-2, which marked distinct cell populations in the neural tube, Zeb-2 was equally expressed in neural tube and neural crest cells during neural tube closure and it was excluded from the ectoderm. Interestingly, at posterior hindbrain and truncal levels, while Zeb-2 was uniform throughout the neural tube prior to neural crest cell emigration, it became markedly enhanced in crest cells as soon as they commenced their migration (Fig. 7D).
These observations, therefore, reveal that, in the rostral half of the embryo, from the midbrain up to the anterior trunk, neural epithelial cells underwent a complete switch from E- to N-cadherin during neurulation. However, both the timing and the kinetics of replacement of E-cadherin by N-cadherin varied significantly along the embryonic axis (Fig. 8). At the cranial level, the E- to N-cadherin switch occurred only gradually with a relatively long time period where both molecules were coexpressed in cells, and the complete disappearance of E-cadherin from the neural epithelium was complete only after neural tube closure and neural crest cell emigration. In the anterior trunk, the switch also occurred gradually, but much earlier during the process of neural tube closure than in the midbrain. It started before bending of the neural plate and was complete at apposition of the neural folds. In addition, in contrast to the midbrain level where the domain of overlap between E- and N-cadherins extended over the entire neural tube, in the trunk, this domain was restricted to the intermediate zone of the neural tube.
By comparing the time of complete E-cadherin repression in neural epithelial cells with that of completion of neural tube closure at different axial levels from the midbrain to the anterior trunk, we found that loss of E-cadherin occurred approximately at a single time point at all levels (between stages HH10 and HH11), whereas closure of the neural tube was effective in these regions in a sequential manner from stage HH10 to HH12 (Fig. 8). This indicates that the timing and kinetics of E- to N-cadherin switch are not synchronized with the movements of neurulation. Analysis of the expression patterns of E- and N-cadherin transcripts in whole embryos at stages HH7–11 (Fig. 2B–D) revealed large domains along the anteroposterior axis in which cadherin expression evolves in a coordinated manner within each domain and differently from the other domains. Thus, three regions could be identified, encompassing the whole brain up to the otic placode for the first one, the anterior trunk covering approximately the first 10–15 somites for the second one, and the rest of the trunk for the third one. Surprisingly, Sox-2 and Zeb-2 displayed uniform expression domains showing no obvious correspondence with these domains, suggesting that additional, yet undefined co-regulators may contribute to the fine tuning of cadherin expression in the neural tube. It should be stressed, however, that although Sox-2 expression in the early embryonic nervous system appears uniform, it is subdivided into several domains by multiple enhancers with distinct spatio-temporal specificities (Uchikawa et al., 2003). Given that the regulatory sequences of the N-cadherin gene itself contains several region-specific enhancers that all rely on Sox-2 for their activity (Matsumata et al., 2005), it is conceivable that the spatial control of N-cadherin expression in the nervous system also depends on Sox-2 activation.
At the onset of neurulation, the neurectoderm comprises exclusively E-cadherin-expressing cells, while at its completion, at least three populations with distinct cadherin repertoires are segregated spatially: laterally, the ectoderm in which E-cadherin expression persists; medially, the neural tube at the origin of the central nervous system in which E-cadherin is substituted for N-cadherin; and at the connection between both compartments, the neural crest progenitors in which E-cadherin is lost but not replaced by N-cadherin. These domains are also characterized by different repertoires of cadherin regulators: none in the ectoderm, Sox-2 and Zeb-2 in the neural tube, and Zeb-2 and Snail-2 in the neural crest. Besides cell adhesion, one of the major roles of cadherins is to mediate cell recognition and segregation. Using transfected cell lines (Duguay et al., 2003) showed that sorting depends on both the type of cadherin expressed by the cells and its relative amount on the cell surface. Therefore, it is tempting to propose that the timing and kinetics of E- to N-cadherin switch during neurulation reflects the spatial segregation of the different lineages arising from the neurectoderm along the medio-lateral axis.
Expression Pattern of Claudin-1 During Primary Neurulation
Loss of occludin, a component of tight junctions, from the surface of neural epithelial cells has been proposed to occur at the time of increase of N-cadherin expression during neural tube closure (Aaku-Saraste et al., 1996). To define whether the disappearance of tight junctions correlates with E- to N-cadherin switch in neural epithelial cells, we analyzed the expression pattern of Claudin-1, another component of tight junctions expressed early during avian development (Simard et al., 2005). As shown in Figures 6 and 7, Claudin-1 expression matched precisely that of E-cadherin in the ectoderm and neural epithelium, indicating that they are co-regulated during reorganization of junctional complexes occurring during neurulation. Of note, transcriptional regulators of E-cadherin, and notably Zeb-2, have also been shown to repress Claudin-1 expression in epithelial cells (Vandewalle et al., 2005).
Timing and Kinetics of E- to N-Cadherin Switch During Secondary Neurulation in the Caudal Trunk
Secondary neurulation starts at closure of the posterior blastopore at stage HH13, and concerns a large portion of the trunk extending beyond the 28th somite and corresponding to the future lumbosacral region. In this region, the neural tube derives from the tail bud, a mesenchymal population of cells that coalesce into a rod and transform into an epithelium. The lumen is formed secondarily by cavitation to produce a hollow tube (Colas and Schoenwolf, 2001).
In the caudal extremity of the embryo at stages HH14–18, undifferentiated tail bud cells formed a uniform solid cord in which early neural progenitors expressing low levels of Sox-2 were morphologically undistinguishable from the paraxial mesoderm. At this level, while E-cadherin was confined to the overlying superficial ectoderm, N-cadherin was uniformly expressed in all tail bud cells (Fig. 9A). During epithelialization of the neural blastema and cavitation, N-cadherin became preferentially accumulated in the apical side of the Sox-2-positive neural epithelial cells, while it remained uniform in the mesenchymal paraxial mesoderm (Fig. 9B–E). In situ hybridizations studies (Fig. 10) confirmed that the expression profiles of E- and N-cadherins showed no region of overlap. E-cadherin expression was restricted to the ectoderm and to the caudal tip of the tail bud, corresponding to the remnants of the primitive streak (Ohta et al., 2007), abutting precisely the region where N-cadherin is expressed. The expression domain of N-cadherin appeared significantly larger than that of Sox-2 in the tail bud, suggesting that, contrary to more anterior levels, Sox-2 plays no role in the initiation of N-cadherin expression in the caudal neural tube. Rather than Sox-2, Tbx-6L, a close relative to Brachyury, exhibited an expression profile matching precisely with that of N-cadherin in the tail bud. More anteriorly, however, in the region where the neural tube is being formed, Tbx-6L was replaced by Sox-2 in the N-cadherin-expressing cells of the neural tube (see also Takemoto et al., 2011). Finally, reflecting the absence of E-cadherin in neural progenitors throughout the course of secondary neurulation, Zeb-2 was not found in the caudal neural tube.
These observations, therefore, indicate that, during secondary neurulation, elaboration of the neural tube does not involve an E- to N-cadherin switch consecutive to transition from a neuroectodermal progenitor to a neural fate. Instead, the neural tube forms from a pool of cells at the origin of the paraxial mesoderm and caudal neural tube (Wilson et al., 2009). These cell populations express N-cadherin from the very beginning on and become gradually segregated probably as a result of selective adhesion. However, we observed no striking difference in the relative cadherin content over time between the neural and mesodermal precursors, indicating that differential level of expression of N-cadherin does not constitute the triggering signal for cell segregation. Yet, expression of cadherin on the surface of the neural cells gradually evolved from uniform to polarized, most likely accompanying the transition from a mesenchymal organization to an epithelial structure.
Cadherin Switch During Node Regression and Formation of the Notochord
Neurulation is accompanied by the formation of the notochord, which is laid down during regression of the Hensen's node. To investigate whether cadherins are involved in notochord segregation, we analyzed the expression patterns of E- and N-cadherins in the node at different stages of its regression during primary neurulation. At stages HH6 to HH8, node regression generates the notochord corresponding to the head and anterior trunk levels, while at stage HH10 it lays down the notochord of the mid-trunk. At stages HH6–8, node cells were organized as a compact mass of Sox-2-negative cells, distributed over the midline and splitting the neural plate into two lateral parts (Fig. 11A, B). These cells could be easily visualized by their high N-cadherin and low E-cadherin content, contrasting with both neural plate cells that expressed exclusively E-cadherin and primitive streak cells that co-expressed both E- and N-cadherins. At stage HH10, the neural plate is no longer organized as a flat, uniform epithelium along the plan of the embryo; rather, it is composed of two epithelial sheets facing each other and oriented along the dorso-ventral axis. At this level, the node is located deeply at the junction between the two neural plate sheets (Fig. 11C). It is often referred to as chordo-neural hinge. As in more proximal regions, node cells could be discerned from the neighboring neural cells by the lack of Sox-2 expression and their strong and uniform staining for N-cadherin. These observations, therefore, suggest that notochord progenitors may segregate from the rest of the neural epithelium owing to their high N-cadherin content. Interestingly, of the various transcriptional regulators of cadherins studied, neither Sox-2, Zeb-2, Snail-1, Snail-2, nor Twist-1 exhibited a spatiotemporal pattern consistent with the selective increase of N-cadherin in node cells. Brachyury, in contrast, was conspicuous at all stages in the node and in notochord (Figs. 1C, 2B–D). The recent demonstration that Brachyury may control cadherin expression and EMT (Fernando et al., 2010) raises the intriguing possibility that it may up-regulate N-cadherin in nodal cells, allowing their segregation from the rest of the neural plate.
Several features emerge from our detailed survey of the expression patterns of E- and N-cadherins during neurogenesis in the chick embryo. First, there is no strict correlation between the E- to N-cadherin switch and the major morphological events accompanying neural induction and neural tube formation. Thus, in the rostral half of the embryo, expression of N-cadherin on the surface of neural epithelial cells follows rather than precedes changes in their cell shape from cuboidal to elongated and pseudo-stratified while the complete disappearance of E-cadherin is not synchronized with closure of the neural tube and separation from the ectoderm. In the caudal half of the embryo, in contrast, N-cadherin pre-exists the process of neurogenesis and there is no switch between E- and N-cadherin. Second, in regions where it occurs, the timing and kinetics of the E- to N-cadherin switch is in general progressive both in time and space and orchestrated by a set of transcriptional regulators distinct from those deployed during EMT, possibly accounting for the transformation of the neural epithelium without substantial modification of its organization. Third, cadherin switch accompanies the gradual partition and final commitment of the neurectoderm into three distinct populations, the ectoderm, neural crest, and neural tube, characterized by different cadherin repertoires under the control of specific sets of transcriptional regulators.
Fertilized chicken eggs (Gallus gallus) from a commercial source (Morizeau, Dangers, France) were incubated at 38°C in a humidified incubator. Embryos were staged according to somite numbers and to the Hamburger and Hamilton staging chart.
Primary Antibodies and In Situ mRNA Probes
Several antibodies to E- and N-cadherins were selected for their complete lack of cross-reaction. These are the mouse monoclonal antibodies (mAb) to chick N-cadherin (FA5, GC4, and ID723 clones) purchased from Sigma (St. Louis, MO). FA5 and GC4 clones both recognize epitopes in the molecule situated close to the N-terminus and involved in cadherin binding while ID723 clone recognizes a site proximal to the membrane. All three mAbs gave identical stainings by immunofluorescence, irrespective of the developmental stages of the embryos. The mAb to mouse E-cadherin (clone 36/E-cadherin) directed against its cytoplasmic tail and cross-reacting with chick E-cadherin (also known as L-CAM) was from BD Transduction Laboratories (San Diego, CA). This mAb produced the same staining as rabbit polyclonal antibodies to the external part of the L-CAM molecule (Thiery et al., 1984). The rabbit polyclonal antibody to chick fibronectin was described previously (Rovasio et al., 1983). The rabbit polyclonal antibody to human Sox-2 was from Abcam (Cambridge, MA) and the mouse mAb anti-chick Snail-2 (clone 62.1E6) from the Developmental Study Hybridoma Bank at the University of Iowa (Iowa City, IA). Plasmids for mRNA probe synthesis for chick N-Cadherin was obtained from M. Takeichi, for chick E-Cadherin from W. Gallin, for chick Brachyury from K. Storey, for chick Claudin-1 from A. Ryan, for chick Sox-2 from P. Scotting, for chick Snail-1 and Snail-2 from A. Nieto, and for chick Tbx-6L from M.-A. Teillet. The probes for chick Zeb-2 and Twist-1 were produced from EST clones (ChEST0090B17 and ChEST0313G23, respectively) developed by the BBSRC-UMIST ChickEST Library and purchased from ARK Genomics (Midlothian, Scotland, UK). Linearized DNA was used to synthesize digoxigenin-UTP (Roche, Indianapolis, IN) labeled antisense probes with RNA polymerases from Promega (Madison, WI) and RNA probes were purified with Illustra ProbeQuant G-50 microcolumns (GE Healthcare, Piscataway, NJ).
Histological Sections and Immunolabelings
For immunolabelings of sections, embryos were fixed in most cases in a 1% paraformaldehyde solution in PBS supplemented with 4% sucrose and 0.1 mM CaCl2 (PBS-sucrose) for 2 hr at room temperature for the preservation of the antigenicity of Sox-2 and Snail-2. In some cases, embryos were fixed overnight at 4°C in 4% paraformaldehyde in PBS-sucrose. After rinsing in PBS-sucrose, embryos were embedded first in a 15% sucrose solution, then in a 15% sucrose-7.5% gelatin solution, and frozen in chilled isopentane. Sections were cut at 10–12 μm on a cryostat and collected on Superfrost/Plus slides (CML, Besancon, France). Sections were permeabilized with 0.1% Triton-X100 in PBS, subjected to immunofluorescence labeling using appropriate secondary antibodies conjugated to Alexa-fluor 488 or 555 (Invitrogen, Carlsbad, CA), and processed, before mounting, for DAPI staining to visualize cells' nuclei. Preparations were observed with a Nikon microscope equipped for epifluorescence and data were collected using the QCapture Pro software (QImaging, Surrey, Canada) and processed using Adobe Photoshop software. Data were acquired using equal exposure times.
In Situ Hybridizations to Whole Mount Embryos and to Tissue Sections
For in situ hybridizations on tissue sections, embryos were fixed and processed for sectioning as for immunolabelings, except that sections were cut at 20 μm. After extensive washes in PBS-sucrose, sections were hybridized for 15 hr at 65°C with the digoxygenin-UTP-labeled RNA probes in 50% formamide, 10% dextran sulfate, and Denhart's buffer (10 μl probe at 0.5–1 mg/ml for 1 ml hybridization buffer and 500 μl buffer/slide). Sections were washed twice during 30 min in 50% formamide, 1× SSC, and 0.1% Tween 20 at 65°C, then 4 times during 20 min at room temperature in 100 mM maleic acid, 150 mM NaCl, pH 7.5, and 0.1% Tween 20 (MABT buffer). After a pre-incubation of 1 hr in MABT buffer containing 10% blocking reagent (Roche) and 10% heat-inactivated lamb serum, sections were incubated overnight at room temperature with the anti-digoxygenin antibody (Roche). After several rinsings with MABT, the sections were rinsed in 100 mM NaCl, 50 mM MgCl2, 1% Tween 20, and 25 mM Tris-HCl, pH 9.5, and stained with NBT-BCIP (Roche) following the manufacturer's guidelines. Preparations were observed and data collected and processed as for immunostainings. Essentially the same procedure was used for the in situ hybridizations on whole mount embryos.
We are indebted to Maria Jussila, who was supported by the Erasmus European Exchange Program, for her excellent work at the early stages of the study. We thank A. Nieto, A. Ryan, P. Scotting, K. Storey, and M. Takeichi for providing cDNA probes, and Sophie Gournet for advice on the illustrations. Alwyn Dady is a recipient of doctoral fellowships of the Ministère de l'Enseignement Supérieur et de la Recherche and the Association pour la Recherche contre le Cancer.