Interplay between Notch signaling and the homeoprotein Xiro1 is required for neural crest induction in Xenopus embryos

The neural crest is a unique and highly specialized population of cells found in all vertebrate embryos. The neural crest is generated at the border of the neural plate, and following closure of the neural tube these cells delaminate from the dorsal neural tube to migrate along different pathways. On reaching their destinations in the embryo, they differentiate into a wide variety of different cell types (reviewed by LaBonne and Bronner-Fraser,1999; Mayor et al.,1999; Christiansen et al.,2000; Mayor and Aybar,2001; Aybar and Mayor,2002).

The generation of neural crest precursors is dependent on the interaction between the neural plate and the non-neural ectoderm(Moury and Jacobson, 1990; Selleck and Bronner-Fraser,1995; Mancilla and Mayor,1996; Mayor et al.,1997). From studies in chick, amphibian and zebrafish embryos,some of the signals involved in the induction of the neural crest have been identified, for example, BMPs, Wnts, FGF and retinoic acid(Liem et al., 1995; Selleck et al., 1998; Streit and Stern, 1999; Mayor et al., 1995; Mayor et al., 1997; LaBonne and Bronner-Fraser,1998; Deardorff et al.,2001; García-Castro et al., 2002; Saint-Jeannet et al., 1997; Villanueva et al.,2002). However, the molecular interactions that are involved in these induction processes seem to be different in the chick to those in Xenopus and zebrafish embryos.

In the chick, blocking BMP activity inhibits neural crest development, and augmenting BMP activity, or its ectopic application, expands the neural crest population (Liem et al., 1995; Selleck et al., 1998). However, in Xenopus and zebrafish it appears that the early induction of neural crest cells depends on a gradient of BMP activity (reviewed by Chitnis, 1999; Aybar and Mayor, 2002). As such, neural crest cells are specified at the border between the neural plate and the epidermis, where intermediate concentrations of BMPs are established,i.e. where the BMP4 concentration is lower than that required to induce epidermis formation and above that which induces neural tissue(Morgan and Sargent, 1997; Marchant et al., 1998; Wilson et al., 1997; LaBonne and Bronner-Fraser,1998; Villanueva et al.,2002; Nguyen et al.,1998).

The molecular mechanisms that underlie the differences in the way that BMP acts during neural crest induction in the chick and in Xenopus or zebrafish are not understood. Thus, in order to study the role of BMP signaling on neural crest induction in Xenopus, and to compare it with what it is known in the chick, we have analyzed two different molecules implicated in the control of BMP4 transcription. The Notch/Delta signaling pathway is thought to influence neural crest development in zebrafish and chick by controlling BMP transcription(Endo et al., 2002; Cornell and Eisen, 2000; Cornell and Eisen, 2002). Indeed, Notch/Delta signaling has already been shown to be involved in a wide variety of other developmental processes, including neurogenesis, gliogenesis,somitogenesis, compartment boundary formation and eye development (reviewed by Artavanis-Tsakonas et al.,1999; Chitnis et al.,1995; Cho and Choi,1998; Domínguez and de Celis, 1998; Kehl et al.,1998; Cavodeassi et al.,1999; Scheer et al.,2001). The Iro protein has been shown to control BMP transcription in the ectoderm and mesoderm of Xenopus embryos(Gómez-Skarmeta et al.,1998; Glavic et al.,2001; Glavic et al.,2002; Gómez-Skarmeta et al., 2001), and has been implicated in the development of the neural crest in zebrafish (Itho et al., 2002). The Iroquois genes participate in several developmental processes, including sensory organ development,compartment boundary formation in Drosophila, dorsal mesoderm formation, neural plate induction, dorsoventral patterning of the neural tube and midbrainhindbrain development(Bürglin, 1997; Cavodeassi et al., 2001; Gomez-Skarmeta and Modolell,2002; Leyns et al.,1996; Gomez-Skarmeta and Modolell, 1996; Papayannopoulos et al., 1998; Diez del Corral et al., 1999; Glavic et al., 2001; Kudoh and Dawid, 2001; Gomez-Skarmeta et al., 1998; Gomez-Skarmeta et al., 2001; Bellefroid et al., 1998; Bosse et al., 1997; Briscoe et al., 2000; Cohen et al., 2000; Glavic et al., 2002; Itoh et al., 2002).

Through conditional Notch/Delta and iro1 gain- and loss-of-function strategies, we demonstrate that Notch/Delta signaling and the iro1 protein in Xenopus play a direct role in neural crest induction by downregulating BMP4 transcription. Furthermore, a series of rescue experiments indicate that iro1 acts upstream of Notch/Delta in the cascade of neural crest induction. We also show that iro1positively regulates Delta1 transcription, in contrast to Snail, a gene that is specifically expressed in the neural crest and which negatively regulates Delta1. It should be mentioned that our experiments were performed using neural crest markers that are initially expressed only in the anterior neural crest. As a result, we discuss a model in which the interaction between iro1, Delta/Notch and Snailgenerates a pattern of gene expression in the anterior neural crest region that is required for the specification of these cells. Finally, our findings regarding the repression of BMP transcription through the activity of Notch/Delta signaling, and the ensuing induction of the neural crest, is in contrast to what has been observed in the chick, providing us with an explanation for the apparent differences between neural crest induction in chick and Xenopus embryos.

Xenopus embryos were obtained as described previously(Gómez-Skarmeta et al.,1998) and staged according to Niewkoop and Faber (Niewkoop and Faber, 1967). Dissections were performed as described by Mancilla and Mayor(Mancilla and Mayor, 1996) and dexamethasone was employed as described by Kolm and Sive(Kolm and Sive, 1995). Dexamethasone was included in the culture medium at stage 2, 10 or 12 and maintained until the embryos were fixed.

Inducible DNA constructs of Xmsx1 were prepared by fusing the entire coding region of Xmsx1 (amino acid residues 1-294) to the ligand-binding domain of the human glucocorticoid receptor (GR; amino acid residues 512-777). A dominant-negative DNA construct (dnXmsx1) was prepared by fusing the homeodomain region of Xmsx1 (amino acid residues 156-294) to the GR domain. Coding sequences were amplified by PCR,using a high fidelity polymerase (Roche Molecular Biochemicals, Mannheim,Germany) and the following primers:

• Xmsx1, 5′-ATGGGGGATTCGTTGTATGGATCGC-3′ and 5′-GAGCTCCGGACAGATGGTACATGCTGTATCC-3′; and

• dnXmsx1, 5′-GAATTCATGAGCCCACCCGCCTG-3′ and 5′-GAGCTCCGGACAGATGGTACATGCTGTATCC-3′.

The PCR products were purified and cloned into pGEM-T Easy vector(Promega), digested with EcoRI/SacI, and ligated with a SacI/XhoI-digested GR fragment into a pCS2+ vector digested with EcoRI/XhoI. Both fusion constructs were automatically sequenced on both strands at the junctions (BRC, Cornell University, Ithaca,NY, USA).

The Xiro1, Notch, Delta, Su(H), SnailGR and Snail dominant-negative (SnailNGR) constructs have all been described previously (Gomez-Skarmeta et al.,2001; McLauglin et al., 2000; Aybar et al., 2003). All cDNAs were linearized and transcribed as described by Harland and Weintraub(Harland and Weintraub, 1985),using a GTP cap analog (New England Biolabs), and SP6, T3 or T7 RNA polymerases. After DNAse treatment, RNA was extracted with phenol-chloroform and precipitated with ethanol. GFP mRNA was used as a control for injections. For injection, mRNA was resuspended in DEPC-water and injected into two-cell stage embryos using 8-12 nl needles.

Dejellied embryos were placed in 75% NAM containing 5% Ficoll. One blastomere of two-cell stage embryos was injected with different amounts of capped mRNA in a solution containing 1-3 μg/μl of lysine fixable fluorescein dextran, as previously described(Aybar et al., 2003)

Total RNA was isolated from embryonic tissue by the guanidinethiocyanate phenol-chloroform method (Chomczynski and Sacchi, 1987), and cDNA was synthesized using AMV reverse transcriptase (Roche Biochemicals) and an oligo(dT) primer. For PCR analysis,the primers for H4 used were those described previously(Aybar et al., 2003). The primers used to analyze Xenopus Delta1 expression amplify a 331 bp product corresponding to the 3′UTR region:5′-GTCCTGGAGAGCAATATGCTCCAG-3′ and 5′-CCATTGTACTGTGAACACAGCATGC-3′.

PCR amplification with these primers was performed over 30 cycles and the PCR products were analyzed on 1.5% agarose gels. PCR was performed simultaneously with RNA that had not undergone reverse transcription to control for genomic DNA contamination. Quantification of PCR bands was performed using ImageJ software (NIH, USA) on 8-bit grayscale JPG files. The values were normalized to the levels of H4 from the same sample and expressed as relative intensities for comparison (sample/H4×10).

Antisense RNA probes for Xiro1(Gómez-Skarmeta et al.,1998), Xslug (Mayor et al., 1995), Foxd3(Sasai et al., 2001), Hairy2A (Wettstein et al.,1997), Bmp4(Hemmati-Brivanlou and Thomsen,1995), Xmsx1 (Suzuki et al., 1997), Serrate(Kiyota et al., 2001) and Notch (Coffman et al.,1990) were synthesized from cDNAs incorporating digoxigenin or fluorescein (Boehringer Mannheim) tags. Embryo specimens were prepared,hybridized and stained according to the method of Harland(Harland, 1991). The alkaline phosphatase substrates used were NBT/BCIP, or BCIP alone.

Antibody staining after in situ hybridization of the embryos was performed according to the method described by Turner and Weintraub(Turner and Weintraub, 1994),using a mouse anti-Myc monoclonal antibody from BabCo. The 12/101 polyclonal antiserum from the Developmental Studies Hybridoma Bank was used to label somites (Griffin et al.,1987).

In order to examine the possible role of Notch signaling and of the homeoprotein gene Xiro1 in the induction of the neural crest, we first analyzed the expression of Xiro1, Delta1, Serrate, Hairy2A and Notch in the presumptive crest territory, comparing their distribution with that of the neural crest marker Xslug. This analysis was performed using double whole-mount in situ hybridization and care was taken to follow individual embryos for the staining of both genes. At the late gastrula stage (stage 12-13), Xiro1 expression was readily detected in the region of the neural plate, although weak expression could also be observed outside of the neural plate in the anterior region of the embryo (Fig. 1A,B; star). When the distribution of Xslug, characteristically expressed in the anterior neural crest cells, was visualized in the same embryos(Fig. 1C,E; arrowhead), it became evident that Xiro1 is expressed in the neural plate, neural crest and tissue adjacent to the neural crest territory(Fig. 1C-E). The Delta1 and Serrate genes have a very dynamic pattern of expression, although both are expressed in a similar manner. At the late gastrula stage (stage 12.5), Delta1 is expressed along the neural anteroposterior axis, but there is a characteristic gap in its expression at the anterior neural plate border (Fig. 1F; arrowhead). In this tissue devoid of Delta1, Xslug is expressed (Fig. 1G; arrowhead). Double in situ hybridization for the Delta1 and Xslug genes confirmed the complementary expression of these genes, the cells expressing Xslug are clearly surrounded by cells expressing Delta1(Fig. 1H,I). This expression pattern was more readily apparent in sections of the stained embryos(Fig. 1J). The same pattern was observed for Serrate expression, Serrate-positive cells surrounded those expressing Xslug(Fig. 1K). The expression of Notch is strong in the neural territory and, in contrast to Delta1 and Serrate, it overlaps with the neural crest marker Xslug (Fig. 1L,M;arrowhead) (Coffman et al.,1993). Finally, from early in development Hairy2A, a downstream target of the Notch signaling pathway(Dawson et al., 1995; Wettstein et al., 1997), is expressed at the neural plate border, coinciding with the territory of Xslug expression (Fig. 1N,O; bracket and arrowhead). However, like Delta1 and Serrate, Hairy2A expression extends into the posterior neural crest at stages when no Slug transcripts can be detected in these cells(Fig. 1N,O; arrow). At the late neurula stage, the expression of Hairy2A can be seen in the prospective forebrain region, whereas Xslug is expressed in the migrating neural crest (Fig. 1P).

In summary (Fig. 1Q,R), Notch, like Xiro1, is present in the neural plate and crest territory, where it could interact with Delta1 and Serrate,which are present at the border of the prospective neural crest territory. The potential interaction of Notch with one of its ligands is compatible with the expression of the target gene Hairy2A in the crest cells.

Based on the pattern of Notch expression and its ligands, we set out to determine whether Notch signaling might be involved in the induction of the neural crest. It has become clear that an interaction between the neural plate and the epidermis, and signals from the paraxial mesoderm, are involved in the induction of the neural crest (Selleck and Bronner-Fraser, 1995; Mancilla and Mayor, 1996; Bonstein et al., 1998; Marchant et al.,1998; Monsoro-Burq, 2003). It has also been established that Notch signaling is involved in the development of the neural plate and mesoderm(Coffman et al., 1993). Thus,we took care not to interfere with the development of the mesoderm and the neural plate when studying the role of Notch signaling in the induction and development of the neural crest. It is known that the mesoderm is specified earlier than the neural tissues, and it has been reported that the neural plate is specified earlier than the neural crest(Smith and Slack, 1983; Servetnick and Grainger, 1991; Mancilla and Mayor, 1996; Woda et al., 2003). Therefore,in order to specifically study neural crest development, Notch signaling was interfered after the mesoderm and the neural plate had already been specified. For this reason, inducible constructs that activated or inhibited Notch signaling were used to control the timing of intervention.

We first analyzed the effect of activating Notch signaling at different developmental times on the formation of the mesoderm, neural plate and neural crest. Ligand activation of Notch results in the proteolytic cleavage of its transmembrane domain and the release of the cytoplasmic region (NICD)(Struhl and Adachi, 2000). NICD can then translocate to the nucleus, where it interacts with the transcriptional repressor Suppressor of Hairless (Su(H)), forming a transcriptional activator complex(Artavanis-Tsakonas et al.,1999). Here, we have used an inducible form of NICD(NICDGR) in order to control the time of its activation. We injected mRNA encoding NICDGR into one blastomere of a two-cell stage embryo,and induced its expression, by exposure to dexamethasone, immediately after the injection (stage 2), at the blastula stage (stage 6-8) or at the gastrula stage (stage 12). The development of the mesoderm was assessed by analyzing the expression of the somite antigen 12/101; development of the neural plate and neural crest induction were assessed by analyzing Sox2 and Xslug expression, respectively. As for non-inducible forms of activated Notch (Coffman et al.,1993), early activation of NICDGR provoked both the expansion of the somites and neural plate on the injected side(Fig. 2A,C), as well as the inhibition of the anterior neural crest(Fig. 2E). Similar results were obtained when NICDGR was activated prior to stage 8. By contrast,when induced at stage 12, NICDGR had no effect on somite or neural plate development (Fig. 2B,D),but rather a clear expansion of the neural crest markers was observed(Fig. 2F). These results indicated that to study the specific effects of Notch signaling on neural crest development, and to avoid any influence on the mesoderm or neural plate,all the Notch signaling constructs should be activated at stage 12. Indeed,using inducible constructs of Dlx proteins, an early effect was observed on neural plate and neural crest development, whereas a later induction produced alterations specific to the neural crest(Woda et al., 2003). Thus, in all the following experiments inducible constructs were activated at stage 12.

Several molecular tools have been developed to modify the activity of the Notch signaling pathway at different levels(Coffman et al., 1993; Chitnis et al., 1995; McLaughlin et al., 2000). Thus we were able to analyze the effects of both gain- and loss-of-function on neural crest development. Activation, at stage 12, of a NICD(NICDGR), or of an inducible ankyrin activator fusion of Su(H) [Su(H)ankGR], provoked an expansion of the Xslug and Foxd3 domains of expression(Fig. 3A,B,E,F). By contrast,the injection of mRNA encoding the dominant-negative Delta^Stu (Dl^Stu) or Su(H)DBMGR into one blastomere of a two-cell embryo, and induction at the late gastrula stage (stage 12), inhibited the expression of the neural crest markers Xslug and Foxd3(Fig. 3C,D,G,H).

It has been shown that inhibition of BMP activity in Xenopus and zebrafish embryos leads to an expansion of the neural crest territory and an increase in Xmsx1 expression(Marchant et al., 1998; Nguyen et al., 1998; Tríbulo et al., 2003). Thus, we analyzed the effect of activating or inhibiting Notch signaling on both BMP4 and Xmsx1 transcription. In contrast to the chick(Endo et al., 2002),activating Notch signaling, by inducing NICDGR and Su(H)andkGR expression, provoked the inhibition of BMP4expression (Fig. 3I,J) and the upregulation of Xmsx1 transcription(Fig. 3M,N). In addition,inhibition of Notch signaling, by Dl^stu and Su(H)DBMGR, promoted the expansion of the BMP4 expression domain (Fig. 3K,L), while inhibiting Xmsx1 expression (Fig. 3O,P).

Finally, to confirm that these constructs were indeed acting on the Notch signaling pathway, we analyzed their effects on the expression of Hairy2A, a known target gene of Notch(Dawson et al., 1995). Each of the constructs that augmented Notch signaling provoked an expansion of the Hairy2A expression domain (Fig. 3Q,R). By contrast, those that inhibited Notch signaling diminished the expression of Hairy2A(Fig. 3S,T). Thus, we concluded that the activation of Notch signaling enlarges the neural crest territory and the domain of Xmsx1 expression, while inhibiting BMP4transcription. Conversely, inhibition of Notch signaling produces exactly the opposite effect.

Hairy2A is a vertebrate target of Notch signaling that belongs to the Enhancer of Split complex. This bHLH transcription factor can act as a transcriptional repressor and has been implicated in the repression of neuronal differentiation (Dawson et al.,1995; Wettstein et al.,1997). We analyzed whether overexpression of Hairy2A also influenced the expression of neural crest markers. Overexpression of Hairy2A repressed N-tubulin expression, a control for the activity of Hairy2A mRNA, at the sites where primary neurons form(Fig. 4A). As we had previously shown that an early activation of Notch signaling leads to an expansion of the somites and, in turn, to an indirect effect on neural crest induction, we took care of injecting the Hairy2A mRNA specifically into the blastomeres fated to become ectoderm. We performed the injection of Hairy2A mRNA into two animal blastomeres of an eight-cell stage embryo. In order to show that there was no effect on mesodermal development, the somite antigen 12/101 was analyzed. No effect on 12/101 was observed in the injected side(Fig. 4B). Interestingly, the same group of embryos that exhibited normal somite development showed an increase in Xslug expression (Fig. 4C). In addition, the expression of Bmp4 was also decreased in these embryos, although the expression of Xmsx1augmented (Fig. 4D-F). These results suggest that the expansion of the neural crest population upon the activation of Notch signaling may be a consequence of the increase in Hairy2A expression provoked in these embryos.

We have shown that by influencing Bmp4 transcription, Notch signaling is involved in specifying the neural crest. Another factor that is known to affect the early transcription of BMP4 is Xiro1(Glavic et al., 2001; Gómez-Skarmeta et al.,2001). Given that Xiro1 is co-expressed with the neural crest marker Xslug, and that the zebrafish Iroquois genes are involved in neural crest formation, we analyzed whether Xiro1 might also influence Xenopus neural crest development. In order to overcome the early effects of Xiro1 in mesoderm and neural plate development,inducible fusion constructs were used as described previously(Glavic et al., 2001; Gomez-Skarmeta et al., 2001; Glavic et al., 2002).

It has been shown that Xiro1 acts as a transcriptional repressor(Glavic et al., 2001; Gomez-Skarmeta et al., 2001). However, when mRNA encoding both Xiro1 (not shown) and its inducible repressor fusion (HDGREnR) was injected and then activated at stage 12, Xslug expression was augmented(Fig. 5A). Conversely,activation, at stage 12, of both the inducible dominant-negative fusion(HDGR) and the inducible activator fusion (HDGRE1A)inhibited Xslug expression (Fig. 5B,C). By contrast, transcription of Bmp4 at the neural plate border was repressed in embryos injected with HDGREnR(Fig. 5D) but increased in embryos overexpressing HDGRE1A and HDGR(Fig. 5E,F). It should be noted that Bmp4 has a complex and dynamic pattern of expression in the neural folds, and that the inhibition of Xiro1 not only affects the levels of Bmp4 expression but also its distribution. The expression of Xmsx1 was augmented and expanded when Xiro1 and HDGREnR was injected into embryos(Fig. 5G), whereas the levels of transcripts diminished and its expression pattern was disrupted in embryos injected with the mRNAs encoding for the activator and dominant-negative constructs (Fig. 5H,I). Finally, overexpression of HDGREnR de-repressed Hairy2Aexpression in the neural fold (Fig. 5J), whereas injecting HDGRE1A and HDGRdecreased Hairy2A expression (Fig. 5K,L). Thus, Xiro1, in addition to being involved in the expression of neural crest markers, also influences Bmp4 and Hairy2A expression in the neural crest precursor domain.

Having established that both Xiro1 and Notch signaling are involved in the specification of the neural crest, we set out to investigate the relationship between these elements by performing rescue experiments. Activation of injected Xiro1 dominant-negative mRNA (HDGR)at stage 12 clearly inhibited Xslug expression(Fig. 6A). By contrast, this effect was prevented, and in some cases Xslug expression was enhanced, if HDGR was co-injected with Hairy2A mRNA or with an activator fusion of Notch signaling (e.g. Su(H)ankGR; Fig. 6B,C). However, the inhibition of Xslug expression induced by blocking Notch signaling could not be rescued by activating the Xiro1 gene (not shown). Taken together, these results suggest that Notch signaling and Hairy2A are likely to be downstream of Xiro1 activity in specifying the neural crest. The inhibition of Notch signaling produced by Su(H)DBMGRrepressed Xslug expression (Fig. 6D), an effect that was reversed by the co-injection of Hairy2A or XmsxGR mRNA(Fig. 6E,F). This suggests that the effect of suppressing Notch activity on neural crest specification depends mainly on Hairy2A and, in addition, that this Notch activity is likely to be upstream of Xmsx1. Finally, the enlargement of the Xslug expression domain produced by NICDGR(Fig. 6G) was reversed by blocking Xmsx1 activity with an inducible dominant-negative construct of Xsmx1, dnXmsxGR (Fig. 6H). This observation provides further evidence that Notch signaling depends on Xmsx1 activity to influence neural crest specification. In all rescue experiments, an unrelated mRNA such as GFP was co-injected, and no effects of GFP on rescue activity were observed (an example on the effect of NICDGR is shown; Fig. 6I).

We have shown that Xiro1 is likely to be upstream of Notch signaling and that the expression of Xiro1 overlaps with that of Delta1. Therefore, we tested whether Xiro1 could regulate the transcription of Delta1. When HDGREnR or HDGRmRNA was activated at stage 12, and cultured until stage 17, the activation of the Xiro1 gene produced a moderate upregulation of Delta1expression in the neural crest region (Fig. 7A; arrowhead). By contrast, however, inhibition of Xiro1by HDGR expression produced a complete inhibition of Delta1expression, even at the border of the neural crest territory(Fig. 7B). Thus, we further examined the regulation of Delta1 by Xiro by injecting one-cell embryos with Xiro1 or Xiro3 mRNA, dissecting out the animal caps from these embryos at stage 9, and then culturing these to the equivalent of stage 18, when the expression of Delta1 was analyzed. Although no expression of Delta1 was observed in control animal caps(Fig. 7C), Delta1transcripts were detected by in situ hybridization in animal caps injected with Xiro1 or Xiro3 mRNA(Fig. 7D,E). When analyzed by RT-PCR, low levels of Delta1 mRNA could be detected in the control animal caps (Fig. 7F,G),probably due to the expression of Delta1 in the ciliary cells of the epidermis. However, after injection of Xiro3 mRNA, a significant upregulation of Delta1 mRNA expression was observed. Taken together,these results strongly suggest that Xiro1 (and Xiro3) is able to activate Delta1 transcription. However, it is likely that in the embryo other signals are present that repress Delta1transcription, which might explain why Delta1 is only expressed in a subdomain of Xiro1 expressing cells.

The expression of Delta1 and Serrate is restricted to the border of the neural crest region (Fig. 1). This observation suggests that a repressor of Delta1might be present in neural crest cells. Many transcription factors that act as transcriptional repressors have been identified (reviewed by Mayor and Aybar, 2001). One such factor is Xsnail, which also seems to be upstream of the genetic cascade of transcription factors that act in the neural crest territory(Aybar et al., 2003). Thus we tested whether Xsnail could repress Delta1 transcription in the neural crest territory. Animal caps taken from embryos co-injected with Xiro3 and Xsnail mRNA were cultured until the equivalent of stage 18, and their mRNA analyzed by RT-PCR. Strong inhibition of Delta1 expression was observed in these animal caps when compared with controls or those injected with Xiro3 mRNA alone(Fig. 7F,G). We have recently developed two specific dominant-negative constructs of Snail, one that contains the Snail zinc finger (ZnfSnailGR) and another that includes the N-terminal (SnailNGR) domain(Aybar et al., 2003). The mRNAs that encode these dominant-negative constructs were injected into one cell of a two-cell embryo, and the expression of Delta1 was analyzed by in situ hybridization after their activation. The expression of Delta1was clearly upregulated in the injected side of the embryo injected with both ZnfSnailGR (Fig. 6H-J)or SnailNGR (not shown). We also examined the effect of inducing the expression of SnailGR at stage 12 and, in these embryos, a moderate but consistent inhibition of Delta1 expression was observed in the ectodermal regions (not shown). Taken together, these results support the idea that Snail could repress Delta1 transcription in the neural crest territory.

We have analyzed the role that Notch signaling and Xiro1 play in neural crest specification. The activation of these elements at the late gastrula stage using inducible constructs has enabled us to examine their specific effects on crest induction without producing any detectable effect on mesoderm or neural plate development. As a result, we have produced a schematic model of the molecular interactions involved in the generation of the neural crest in Xenopus embryos

In Xenopus embryos, the expression patterns of Notch, the Notch ligand Delta1 and the Notch downstream gene Hairy2Asuggest that these molecules might be implicated in the formation of the neural crest. Interestingly, in contrast to the homogenous expression described previously (Kiyota et al.,2001), we observed that another Notch ligand, Serrate, is expressed in a complex pattern very similar to that of Delta1. Thus,both ligands are expressed in cells that surround those expressing Xslug and hence they could activate Notch and, thus, Hairy2Ain the neural folds. The restricted pattern of Hairy2A expression overlaps that of Xslug, suggesting that other elements either repress Hairy2A transcription in the adjacent epidermis and neural plate, or permit the expression of this gene in the neural fold region. One of these elements could be Notch itself.

In Xenopus, Notch is detected in neural tissue and is excluded from the non-neural ectoderm, thereby accounting for the absence of Hairy2A expression in the epidermis(Coffman et al., 1990) (this work). Our analysis of Notch signaling demonstrates that increasing Notch activity at the early gastrula stage produces an expansion of the neural crest territory. Interestingly, the increase in Xslug and Foxd3expression produced by Notch activation is in contrast to the repression of Slug upon changes in Notch activity previously described in the chick(Endo et al., 2002). In addition, inhibition of Notch signaling by Delta^Stu, or by a dominant-negative form of Suppressor of Hairless, produces a reduction in the number of Xslug- and Foxd3-positive cells. Furthermore, direct overexpression of the Notch target gene Hairy2Aleads to the induction of neural crest cells. Thus, our results provide evidence of a role for Notch and its downstream elements in the specification of Xenopus neural crest.

The molecular mechanism by which Notch signaling controls the induction of the neural crest in the chick appears to involve the upregulation of BMP4 expression, necessary for neural crest induction(Liem et al., 1995; Endo et al., 2002). However,in Xenopus, the activity of BMP is opposite to that of the chick, and a decrease in BMP activity relative to that seen in the non-neural ectoderm induces neural crest cells. Therefore, the observed increase of Xslugand Foxd3 expression is most likely due to the repression of Bmp4 transcription. Indeed, here we show that the activation of Notch represses Bmp4 expression in Xenopus embryos. In addition,inhibition of Notch signaling by Delta^Stu, or by a dominant-negative form of Suppressor of Hairless, produces an increase in Bmp4 transcription. Our analysis of the influence of Notch signaling on the BMP pathway further showed that the precise pattern of Xmsx1 expression, a BMP target gene, is finely regulated in the neural crest precursor domain.

Contrary to our expectations, activation of Notch often produced an increase in Xmsx1 expression, even though Bmp4 transcription was inhibited. Accordingly, treatments that blocked Notch signaling, and that therefore activated Bmp4 expression, produced embryos where Xmsx1 expression was impaired. These results support the conclusion that Xmsx1 expression is induced at a specific level of BMP activity(Tríbulo et al., 2003). We also observed that, when overexpressed in embryos, Hairy2Aproduced similar effects on Xslug, Bmp4 and Xmsx1expression, and that it is able to rescue the effect of Su(H)DBMGR in blocking Notch signaling.

In conclusion, Notch signaling activates the expression of Hairy2Ain the region of the neural folds, and thereby represses Bmp4transcription. This effect of Notch signaling is dependent on Xmsx1activity, as the inhibition of Notch by Su(H)DBMGR can be reversed by Xmsx1, and the effects produced by activating Notch can be blocked by a dominant-negative Xmsx1 construct. Our results also provide a possible explanation for the apparent discrepancy in the role played by BMP in chick and Xenopus or zebrafish neural crest induction. At the time of neural crest induction, the levels of BMP at the neural plate border are high in both Xenopus and zebrafish, and low in the chick. If we assume that an intermediate level is required to induce neural crest in all these vertebrates, then an increase in BMP levels in the chick would establish similar levels to those generated by a decrease in Xenopus and zebrafish. Thus, because of the initial differences in the levels of BMP in these two groups of organisms, the molecular machinery that induces neural crest formation (e.g. Notch/Delta, Xiro1) must adjust the specific levels of BMP by producing opposing effects on BMP expression. Thus,Notch/Delta signaling induces the neural crest by increasing BMP expression in the chick (Endo et al., 2002),and decreasing it in Xenopus.

Genes of the Iroquois family have been implicated in a variety of developmental processes, including dorsal mesoderm formation, neural induction, compartment specification in the eye imaginal disc of Drosophila and midbrain-hindbrain boundary formation(Glavic et al., 2001; Kudoh and Dawid, 2001; Papayannopoulos et al., 1998; Diez del Corral et al., 1999; Gomez-Skarmeta et al., 1998; Bellefroid et al., 1998; Bosse et al., 1997; Briscoe et al., 2000; Glavic et al., 2002; Itoh et al., 2002). Our results extend the role of Xiro1 during development to that of neural crest specification. Indeed, it has already been demonstrated that Xiro1 can bind to the Bmp4 promoter, and, by acting as a repressor, it can inhibit Bmp4 transcription in both the Spemanns'organizer and the neural plate(Gomez-Skarmeta et al., 2001; Glavic et al., 2001).

Our observations show that Xiro1 is expressed in the neural crest territory and that its activation produces an enlargement of this territory. By contrast, inhibition of Xiro1 leads to a reduction in the expression of neural crest markers. Like Notch signaling, Xiro1 also represses Bmp4 transcription and activates Hairy2Aexpression in the neural folds, as well as expanding the domain of Xmsx1 expression. The effects of inhibiting Xiro1 on neural crest specification can be reversed by activating Notch signaling, or by co-injecting the Notch target gene Hairy2A. Taken together, these results indicate that Xiro1 activity is upstream of Notch signaling.

Although the regulation of Notch activity by Xiro1 could operate at different levels, we have presented evidence that Xiro1 can upregulate Delta1 transcription. Activation of Xiro1 in animal caps or whole embryos, led to an upregulation of Delta1,whereas impairing Xiro1 produced an inhibition of Delta1expression in the neural crest territory. Thus, Xiro1 seems to positively regulate Delta1 expression. However, as the expression of Delta1 and Xiro1 do not completely overlap, additional factors must be required either to activate Delta1 where Xiro1 is not expressed, or to inhibit its expression in those cells expressing Xiro1 but not Delta1.

Delta1 is excluded from the center of the prospective neural crest region, and its transcripts can only be seen at the border of the crest region. This pattern of Delta1 expression suggests that a repressor is acting in the crest region. Many transcriptional repressors are expressed in the neural crest, including Snail(Aybar et al., 2003), Slug (LaBonne and Bronner-Fraser,1999; Mayor et al.,2000), Foxd3 (Sasai et al., 2001) and Zic5(Nakata et al., 2000). Moreover, Snail appears to be upstream in this genetic cascade(Aybar et al., 2003). We show here that Snail can repress Delta1 expression in animal caps and in whole embryos, and that the inhibition of Snail activity provokes an upregulation of Delta1 expression in the neural crest territory. Our results strongly suggest that the expression of Delta1in the neural crest could be patterned by the activity of Snail. It is worth mentioning that the effect of Snail on Delta1expression was not only seen in the ectoderm but also in the somites, where Snail is also expressed (Essex et al., 1993). Thus, it seems feasible that Delta1expression, which plays an important role in somite formation(Jen et al., 1997), could also be under the control of Snail. Indeed in Drosophila, Snailhas been shown to represses Delta expression during the dorsoventral patterning of the embryo (Cowden and Levine, 2002; Ip and Gridley,2002). It is also interesting to note that Snail is weakly expressed in the anterior neural fold at the early gastrula stage, but at the end of gastrulation, when Delta1 is strongly expressed in the anterior neural fold, Snail expression is downregulated in that region (Aybar et al., 2003). This complementary pattern of expression between Snail and Delta1 also supports the idea that Snail is indeed a repressor of Delta1 transcription. Finally, Snail may not only serve to repress Delta1 in the neural crest, overexpression of Snail induces the appearance of neural crest markers in animal caps and in whole embryos (Aybar et al.,2003). Indeed, it is likely that the influence of Snailon neural crest markers is independent of its repression of Delta1. It is important to mention that Slug or Foxd3 are never expressed in the anterior neural fold, being also putative inhibitors of Delta1 in the crest region.

The role of the Iroquois genes in establishing embryonic boundaries seems to be extended across this gene family. As mentioned before, Iroquois genes participate in the development of the imaginal disc compartment in Drosophila (Papayannopoulos et al., 1998; Diez del Corral et al., 1999; Cavodeassi et al.,1999), and, in Xenopus, Xiro1 is involved in the formation of the midbrain-hindbrain boundary(Glavic et al., 2002). It is noteworthy that Notch signaling is also involved in both these processes(Papayannopoulos et al., 1998; Domínguez and de Celis,1998). In Drosophila, the Iroquois genes influence Notch signaling through the expression of Fringe, thereby defining the dorsal and ventral compartments (Cavodeassi et al., 1999). In Xenopus, the Notch target genes Hes1 and Hes3 (Hirata et al., 2001), and the Hes-related 1 gene (Xhr1)(Shinga et al., 2001), have been implicated in establishing the midbrainhindbrain border, and in particular in midbrain development. Recently, Xiro1 has been shown to be involved in the establishment of this region by controlling Gbx2and Otx2 expression (Glavic et al., 2002). It is thus tempting to speculate that Xiro1might regulate Hes1, Hes3 and/or Xhr1 expression at the midbrain-hindbrain boundary. Here, we present evidence that Xiro1 is also involved in the establishment of the boundary between the neural plate and the epidermis, i.e. the region in which the neural crest cells are generated.

The data generated over the past years, together with our present observations, lead us to propose the following model for neural crest induction (Fig. 8). It should be noted that this model is predominantly based on data from the analysis of neural crest markers that are initially expressed only in the anterior neural crest. Therefore, additional studies using specific posterior neural crest markers should be carried out to determine whether our model is also valid for posterior neural crest cells.

At the early gastrula stage, the coordinate action of Xiro1, as a positive regulator, and Snail, as a repressor, restricts the homogenous expression of Delta1 to a ring of cells at the border of the neural crest territory (Fig. 8A). At the late gastrula stage, Xiro1 continues to induce the expression of Delta1 at the border of the neural crest territory, where Delta1 interacts with Notch to activate Hairy2A in the neural fold region(Fig. 8B). Later in development, Hairy2A acts as a repressor of Bmp4transcription, ensuring that the optimal level of Bmp4 to specify the neural plate border in this region is reached(Fig. 8C). This intermediate level of Bmp4 in turn activates msx1 expression, which is also required for the specification of the neural plate border(Tríbulo et al., 2003). Finally, the action of additional signals (WNTs, FGFs, retinoic acid) in this newly defined domain induces the production of neural crest cells(Mayor et al., 1995; Mayor et al., 1997; LaBonne and Bronner-Fraser,1998; Deardorff et al.,2001; Villanueva et al.,2002; García-Castro et al., 2002).

We would like to thank: C. Kintner for the Serrate(Ser1.2), and D. Turner for the Hairy2A cDNAs; K. McLauglin for the different constructs for the Notch/Delta pathway; N. Ueno for the Xmsx1 clone; and Florencio Espinoza for technical support. Work was supported by an International Research Scholar Award from the Howard Hughes Medical Institute to R.M., and by grants from Fondecyt (#1020688), the Millennium Program (P99-137F) and the MRC to R.M. M.J.A. was supported by postdoctoral grants from Fondecyt (#3010061) and Fundación Antorchas(#14169-3).