Snail2 controls mesodermal BMP/Wnt induction of neural crest

Neural crest is of interest for both evolutionary and medical reasons. Like the mesoderm, it is an induced tissue, arising at the boundary between the nascent neural plate and the embryonic epidermis (Alfandari et al., 2010; Basch and Bronner-Fraser, 2006; Klymkowsky et al., 2010; Minoux and Rijli, 2010). The neural crest represents an evolutionary innovation (Baker, 2008; Yu, 2010), responsible in part for the diverse cranial morphologies of vertebrates (Hanken and Gross, 2005). It provides a classic example of an epithelial-mesenchymal transition (EMT) and associated apoptotic suppression, and so serves as a model for cancer metastasis (Klymkowsky and Savagner, 2009). Neural crest is also the focus of tetratogenic birth defects in humans, such as those caused by thalidomide (McCredie, 2009) and ethanol (Webster and Ritchie, 1991).

Although there have been reports that mesoderm is not involved in the induction of neural crest in zebrafish (Ragland and Raible, 2004) and chick [(Basch et al., 2006), but see below], there is clear evidence for a role for early mesoderm in neural crest induction in amphibians in general [(Raven and Kloos, 1945), as cited by Baker and Bronner-Fraser (Baker and Bronner-Fraser, 1997)] and, more specifically, in Xenopus laevis (Bonstein et al., 1998; Hong et al., 2008; Marchant et al., 1998; Mayor et al., 1995; Monsoro-Burq et al., 2003; Steventon et al., 2009; Steventon et al., 2005). The situation is somewhat less clear in mouse and human, in part because of the challenges associated with experimental studies in the corresponding early embryonic stages (Aggarwal et al., 2010; Carver et al., 2001; Goh et al., 1997; O'Rourke and Tam, 2002; Xu et al., 2000).

The transcription factors Snail and Twist were first identified in Drosophila melanogaster, where they sit at the end of the Toll/Dorsal (NF-κB) signaling pathway (Stathopoulos and Levine, 2002; Valanne et al., 2011). Mutations in snail and twist lead to defects in mesoderm formation (Leptin, 1991; Thisse et al., 1987). snail-type genes encode C₂H₂-type zinc-finger transcription factors. Multiple members of this gene family have been characterized from placozoans through humans (Barrallo-Gimeno and Nieto, 2009). twist encodes a basic helix-loop-helix (bHLH)-type transcription factor; these proteins typically act as dimers (Barnes and Firulli, 2009). Both snail- and twist-like genes are found in the primitive chordate Ciona intestinalis (Shi et al., 2005) and in the jellyfish Podocoryne carnea (Spring et al., 2002; Spring et al., 2000). A BLAST analysis indicates that both the Ciona and Podocoryne Snail proteins more closely resemble mammalian Snail2 (Slug) than Snail1 proteins (our unpublished observation). Snail proteins appear to act primarily as transcriptional repressors, binding to DNA E-box (5′-CANNTG-3′) sequences. During Drosophila mesoderm specification and patterning, snail and twist expression are regulated by a molecular cascade involving dorsal, which encodes an ortholog of the NF-κB subunit protein RelA (Huguet et al., 1997; Ip et al., 1992). This network involves negative-feedback regulation of dorsal through the secreted factor WntD, the expression of which is regulated by Snail and Twist (Ganguly et al., 2005; Gordon et al., 2005). Genomic chromatin immunoprecipitation-microarray studies (Sandmann et al., 2007; Zeitlinger et al., 2007) suggest that Snail and Twist regulate a wide array of target genes: Twist targets almost 25% of all annotated Drosophila transcription factors (Sandmann et al., 2007). Interestingly, in the vertebrate X. laevis, snail1, snail2 and twist1 RNAs appear to be `immediate-early' targets of regulation by the NF-κB subunit protein RelA (Zhang et al., 2006).

In vertebrates, there are two distinct twist-like genes: twist1 and twist2 (dermo1) (Li et al., 1995). twist1 has been implicated in mesoderm formation, as well as in a number of developmental events. twist1 haploinsufficiency leads to skeletal dysplasia (Miraoui and Marie, 2010). In the mouse, Twist1 is required for cranial neural crest migration as well as for the suppression of apoptosis (Chen and Behringer, 1995; Soo et al., 2002). In humans, mutations in TWIST have been implicated in mesenchymal stem cell differentiation and skeletal malformations (craniosynostosis) (Miraoui and Marie, 2010). There are two closely related snail-like genes in vertebrates, snail1 and snail2 (previously known as slug), as well as a number of more distantly related genes (Barrallo-Gimeno and Nieto, 2009; Manzanares et al., 2001; Nieto, 2002). In vertebrates, snail gene function was originally studied most intensely in the context of the neural crest (Aybar et al., 2003; Carl et al., 1999; LaBonne and Bronner-Fraser, 2000; Nieto et al., 1994; O'Rourke and Tam, 2002; Tribulo et al., 2004). In the chick, Snail2 is expressed in both mesoderm and premigratory crest, and appears to be involved in the formation and behavior of both tissues (Nieto et al., 1994). In the mouse, the domains of Snail2 and Snail1 expression are switched (Locascio et al., 2002; Sefton et al., 1998) and neither Snail1 nor Snail2 appears to be absolutely necessary for either mesodermal or neural crest formation (Carver et al., 2001; Jiang et al., 1998). That said, Snail1 null mice display a recessive embryonic lethal phenotype with clear gastrulation defects and morphologically abnormal mesoderm (Carver et al., 2001). Whether the roles of Snail1 and Snail2 in the early mouse embryo have been subsumed by other genes, such as Twist1, Zeb1 or Zeb2 (Sip1), which encodes an E-box-binding protein implicated in the regulation of EMT and tumor progression (Bracken et al., 2008; Liu et al., 2009; Schmalhofer et al., 2009; Wellner et al., 2009), remains unclear. That complex interactions may be involved is suggested by a report that Snail2 is required for Twist1-induced EMT in mice (Casas et al., 2011). In the mouse, Snail2 acts downstream of Sox9 in trunk neural crest specification (Cheung et al., 2005). Snail2^–/– mice display white forehead blaze, patchy depigmentation of the ventral body, tail and feet, macrocytic anemia, infertility and a deficiency in white adipose tissue mass (Perez-Mancera et al., 2007). The combination of a conditional Snail1 null mutation and Snail2^–/– leads to defects in left-right axis formation but not, apparently, to defects in neural crest formation (Murray and Gridley, 2006).

In X. laevis, snail1, snail2 and twist1 are expressed in the blastula stage embryo (Essex et al., 1993; Mayor et al., 1993; Mayor et al., 2000; Sargent and Bennett, 1990; Zhang and Klymkowsky, 2009). Previously, we presented evidence for a role for snail2 and twist1 in mesoderm and neural crest formation (Carl et al., 1999; Zhang et al., 2006; Zhang and Klymkowsky, 2009). The expression of snail2, snail1 and twist1 in both early mesoderm and neural crest raises a number of issues, illustrated in part by the work of Aybar et al. (Aybar et al., 2003). Based on the behavior of dominant-negative mutant forms of Snail2 and Snail1, they claimed that snail (snail1) precedes slug (snail2) in the genetic cascade involved in neural crest specification. Yet morpholino-based studies indicate that inhibition of snail2 expression disrupts mesoderm formation and leads to a decrease in snail1 RNA levels in the early embryo (Zhang et al., 2006; Zhang and Klymkowsky, 2009). To resolve these issues, we extended previous work using morpholinos to examine the role of snail1, snail2 and twist1 in the early Xenopus embryo. Blastomere injection and explant studies enabled us to discover distinct roles for snail2, as compared with snail1 and twist1, in mesoderm-based induction of neural crest.

X. laevis embryos were staged, and explants and co-explants were generated following standard procedures (Klymkowsky and Hanken, 1991; Nieuwkoop and Faber, 1967; Sive et al., 2000; Zhang et al., 2004). Similar studies were carried out using X. tropicalis embryos using animals purchased from Xenopus I following methods analogous to those used in X. laevis, and modified as described in Khokha et al. (Khokha et al., 2002). Capped mRNAs were transcribed from linearized plasmid templates using mMessage mMachine kits (Ambion) following the manufacturer's instructions. For 2-cell stage studies, embryos were injected equatorially; for 32-cell stage studies, C2 and C3 blastomeres were co-injected with RNA encoding GFP, and examined at stage 10 to confirm the accuracy of injection. Animal caps were isolated from stage 8-9 blastula embryos in 0.5 MMR (Sive et al., 2000), transferred into wells of a 2% agarose plate (one animal cap per well) and combined with dorsolateral mesodermal zone isolated from stage 10.5 gastrula embryos according to Bonstein et al. (Bonstein et al., 1998). Explant recombinants were harvested when siblings reached stage 18. Images were captured using a Nikon CoolPix 995 camera on a Leica M400 photomicroscope or using a Nikon D5000 camera on a Wild stereomicroscope. Images were manipulated with Fireworks CS4 software (Adobe) using Auto Levels and Curves functions only.

Modified morpholino antisense oligonucleotides (MOs) were purchased from Gene Tools. The MO against snail1 (5′-TTTAGCAGCCGAGCACTGAGTTCCT-3′) was tested for its ability to block the translation of snail1 RNA using an RNA that contains its target site and encodes a Snail1-GFP chimera. Other MOs used were designed to block the translation of antipodean/vegT (5′-ACTTTACATCCGGCAGAGAATGCAT-3′) and xbra (5′-GCGCAGCTCTCGGTCGCACTCATTA-3′) RNAs. snail2 and twist1 MOs were as described previously (Zhang et al., 2006; Zhang and Klymkowsky, 2009). In addition, a new snail2 MO was designed (5-TCTTGACCAGAAAAGATCGTGGCAT-3′) that matches the single identified X. tropicalis snail2 gene perfectly (Vallin et al., 2001).

Plasmids encoding ΔNp63 and ΔNp63_R304W (Barton et al., 2009) were generously supplied by Ethan Lee. Plasmids encoding Sizzled (Salic et al., 1997), Cerberus (Bouwmeester et al., 1996), BMP4, Wnt8, Noggin (Smith and Harland, 1992) and Dickkopf (Dkk) (Glinka et al., 1998) were supplied by Eddy DeRobertis and Richard Harland. Snail2, Snail1 and Twist expression plasmids have been described previously (Zhang et al., 2006; Zhang and Klymkowsky, 2009); a GFP-tagged form of Snail1 was constructed for this project. RNA-injected embryos were selected based on GFP fluorescence.

For in situ hybridization studies, digoxigenin-UTP-labeled antisense probes were prepared following standard methods; specific probes for sox9, vegT/antipodean, chordin, endodermin, myoD, chd7 and xbra RNAs were used. In many cases, embryos were co-injected with RNA (50 pg/embryo) encoding β-galactosidase, the activity of which was visualized in fixed embryos using a brief Red-Gal (Research Organics) reaction in order to identify successfully injected embryos.

RNA isolation, RT-PCR and qPCR analyses were carried out as described previously (Zhang et al., 2003; Zhang et al., 2006). In brief, total RNA was isolated from embryos or dorsal lateral mesoderm regions; cDNA synthesis was performed from 1 μg purified RNA using a Verso cDNA kit (Thermo Scientific) according to the manufacturer's directions. Real-time PCR was carried out using a Mastercycler Epgradient Realplex device (Eppendorf). PCR reactions were set up using DyNAmo SYBR Green qPCR kits (Finnzymes). Each sample was normalized to the expression level of ornithine decarboxylase (ODC). The cycling conditions used were: 95°C for 5 minutes; then 40 cycles of 95°C for 15 seconds, 56°C for 15 seconds, 60°C for 30 seconds. The ΔΔCT method was used to calculate real-time PCR results. Primers for RT-PCR analyses were (5′ to 3′, forward and reverse):

• snail2, GATGCACATCAGGACACACAC and CTGCGAATGCTCTGTTGCAGT;

• snail1, AAGCACAATGGACTCCTT and CCAATAGTGATACACACC;

• twist1, AGTCCGATCTCAGTGAAGGGC and TGTGTGTGGCCTGAGCTGTAG;

• ODC, CAGCTAGCTGTGGTGTGG and CAACATGGAAACTCACAC;

• sox9, TGCAATTTTCAAGGCCCTAC and GCTGCCTACCATTCTCTTGC;

• sizzled, CATGTCCGGAGTCTTCCTGC and GGATGAACGTGTCCAGGCAG;

• cerberus, CCTTGCCCTTCACTCAG and TGGCAGACAGTCCTTT;

• wnt8, TGATGCCTTCACTTCTGTGG and TCCTGCAGCTTCTTCTCTCC;

• bmp4, TGGTGGATTAGTCTCGTGTCC and TCAACCTCAGCAGCATTCC;

• dkk, ACGGAAGATGATGACTGTGC and CTCTTGATCTTGCTCCACAGG;

• noggin, AGTTGCAGATGTGGCTCT and AGTCCAAGAGTCT CAGCA;

• frzb1, TGGACTCATTCCTGCTACTGG and AATTGCCAGGATAGCATTGG;

• sfrp2, TCTGTGTGAGCAGGTGAAGG and GTCATTGTCATCCTCGTTGC;

• crescent, GGCTCTCTGCTATGACATTGG and CACACAGACTGCGACATGG;

• chordin, CCTCCAATCCAAGACTCCAGCAG and GGAGGAGGAGGAGCTTTGGGACAAG;

• fgf8, TGGTGACCGACCAACTAAGC and ACGATTAACTTGGCGTGTGG;

• eomes, TCCTCAGGCTACCAGTACAGC and GCCGGTTACAGAGGAAGACC;

• gata4a, CTGGACTGCTTCTCCATTCG and ACATTGCTCCACAGTTGACG; and

• tbx6, CAGCCAATCAGGAACAAGG and GTTCTGTGCTGCATCTGTGG.

It is clear that snail1 RNA is present in X. laevis embryos following the onset of zygotic transcription (Mayor et al., 2000; Sargent and Bennett, 1990; Zhang and Klymkowsky, 2009). In past studies, it was reported that injection of snail1 RNA could partially rescue the effects of snail2 antisense RNA and snail2 and twist1 morphant phenotypes (Carl et al., 1999; Zhang et al., 2006; Zhang and Klymkowsky, 2009), as well as the effects of dominant-negative versions of Snail2 (LaBonne and Bronner-Fraser, 2000). Levels of snail1 RNA are reduced in snail2 and twist1 morphant embryos (Zhang and Klymkowsky, 2009). To determine whether snail1 expression is involved in early mesoderm formation and/or maintenance, which does not appear to have been examined previously, we designed a morpholino (MO) directed against the 5′ UTR of the single identified snail1 mRNA in X. laevis (see Fig. S1 in the supplementary material). The previously employed snail2 MO (Zhang et al., 2006) has 12 out of 25 mismatches to the analogous region of the snail1 mRNA, whereas the snail1 MO has 17 and 18 out of 25 mismatches to the snail2a and snail2b mRNAs, respectively, as identified by Vallin et al. (Vallin et al., 2001). As we have yet to identify a functional antibody for Xenopus Snail1 (or Snail2 or Twist), we examined the effects of snail1 MO on the translation of an RNA, UTR-Snail1-GFP, that contains the MO 5′ UTR target sequence. When UTR-Snail1-GFP RNA and snail1 MO were co-injected into fertilized eggs, there was a reduction in the level of Snail1-GFP protein accumulation (Fig. 1A) and a reduction in the level of Snail1-GFP fluorescence (Fig. 1B,C). RT-PCR (Fig. 1D) and quantitative RT-PCR (qPCR) (Fig. 1E) analyses indicated that injection of snail1 MO into both blastomeres of a 2-cell embryo led to a reduction in the amounts of both twist1 and snail2 RNA.

Injection of snail2 antisense RNA, or snail2 or twist1 MOs, into one cell of 2-cell embryos leads to the loss of the mesodermal markers xbra, antipodean/vegT and myoD, an increase in the expression domain of the endodermal marker endodermin (edd), and the loss of expression of the neural crest marker sox9 (Carl et al., 1999; Zhang et al., 2006; Zhang and Klymkowsky, 2009). Injection of the snail1 MO produced similar effects: loss of xbra (Fig. 2A,A′) and antipodean/vegT (data not shown) expression, an increase in edd expression (Fig. 2B,B′) and expansion of the edd expression domain into the underlying mesoderm of stage 11/12 embryos (Fig. 2C,C′), and the loss of expression of sox9 (Fig. 2D,E) and myoD (Fig. 2F,F′) in later stage embryos. Levels of other neural crest markers, specifically snail2, twist and chd7 (Bajpai et al., 2010), were reduced in snail1 morphant embryos (see Fig. S2 in the supplementary material). A quantitative analysis of the effects of snail1 MO are presented in Table 1. As noted in past studies, a number of morphant embryos failed to gastrulate, suggesting that later phenotypes are hypomorphic rather than null (amorphic) (see Zhang et al., 2006; Zhang and Klymkowsky, 2009). The effects of the snail1 MO were rescued by the injection of an RNA encoding Snail1-GFP (Fig. 2G-J), an RNA that does not contain the target sequence of the snail1 MO. As in the case of snail2 and twist1 morphant embryos, the snail1 MO phenotype was rescued, albeit not as efficiently, by snail2 and twist1 RNAs (Fig. 2K). The extent to which snail2, snail1 and twist1 RNAs rescue the expression of `late' genes, such as sox9 and myoD, was lower than their ability to rescue the expression of the `earlier' genes xbra and edd, perhaps because later `differentiation' genes are more redundantly regulated.

To confirm that loss of mesoderm per se leads to the failure of neural crest induction, as described previously (Bonstein et al., 1998), we disrupted mesoderm formation in two complementary ways. xbra (Smith et al., 1991) and vegT/antipodean (Stennard et al., 1996; Stennard et al., 1999) encode T-box-type transcription factors involved in mesoderm formation in X. laevis. Injection of an xbra MO (10 ng/embryo) into one cell of a 2-cell embryo had little effect on the expression of the myotomal marker myoD, whereas injection of an antipodean/vegT MO led to the loss of myoD expression, as described previously (Fukuda et al., 2010) (see Fig. S3 in the supplementary material); neither had any apparent effect on sox9 expression (data not shown). By contrast, when injected together (5 ng each/embryo), the xbra and antipodean/vegT MOs caused the loss of sox9 expression in ∼80% of embryos (Fig. 3A-C; Table 2).

That the antipodean/vegT MO alone had no effect on sox9 expression suggested possible compensatory processes. Mesoderm formation involves multiple pathways that act in independent, interdependent and cooperative ways (Koide et al., 2005; Loose and Patient, 2004). A second mesoderm pathway active in X. laevis involves the transcription factor p53 and its modulation of TGFβ signaling through interactions with SMAD proteins (Cordenonsi et al., 2003; Cordenonsi et al., 2007; Sasai et al., 2008; Takebayashi-Suzuki et al., 2003). Barton et al. (Barton et al., 2009) described a splice variant of the p53-related protein p63, ΔNp63, that inhibits the p53-SMAD interaction and blocks mesoderm formation in X. laevis. Injection of ΔNp63 RNA (∼600 pg/embryo) into one cell of a 2-cell embryo led to loss of the neural crest marker sox9 and of the mesoderm/muscle marker myoD in ∼50% of embryos (Fig. 3D-G; Table 2). Injection of ΔNp63 RNA into both cells of a 2-cell embryo led to a reduction in xbra and vegT RNAs, as determined by qPCR, as well as to the reduction of another mesoderm marker, myf5 (Fig. 3J) (Hopwood et al., 1991). Injection of RNA encoding ΔNp63_R304W, a mutated form of ΔNp63 that no longer binds to DNA (Barton et al., 2009), had no effect on sox9 expression (Fig. 3H,I). When injected together, xbra and antipodean/vegT MOs and ΔNp63 RNA led to the near complete loss of sox9 expression (Fig. 3C; Table 2), suggesting that the two pathways (T-box and p53/SMAD) cooperate in terms of mesodermal induction of neural crest (Table 2).

qPCR studies of xbra and antipodean/vegT morphant (Fig. 3K) and ΔNp63 RNA-injected (Fig. 3L) embryos (both cells of 2-cell embryos injected) revealed decreased levels of snail1, snail2 and twist1 RNAs. Given that loss of snail1, snail2 or twist1 expression inhibits mesoderm formation and neural crest induction (see above), we examined whether xbra and antipodean/vegT MO-induced loss of sox9 expression could be rescued by the injection of snail2, snail1 or twist1 RNAs. Such studies produced a modest rescue of sox9 expression (Fig. 3C). The loss of sox9 expression in the xbra and antipodean/vegT morphant and ΔNp63 RNA-injected embryos supports an active role of the mesoderm in neural crest induction in Xenopus (Bonstein et al., 1998).

A limitation of whole- and half-embryo studies is that the injected reagents often influence a range of developing tissues. These studies are also complicated by the fact that the rates of diffusion of different reagents (e.g. antibodies, MOs and RNAs) within the embryo differ (Zhang et al., 2004) (our unpublished observations). To circumvent this effect, we used two complementary methods: targeted blastomere injection and combinatorial (sandwich) explant studies. Although it is certainly the case that there is no blastomere within the 32-cell embryo that contributes to a single tissue type or lineage, owing in part to the slow intermixing of cells (Dale and Slack, 1987; Moody, 1987; Nakamura et al., 1978; Wetts and Fraser, 1989), it is possible to favor certain tissues. For example, in the 32-cell embryo, the C2 and C3 blastomeres give rise primarily to paraxial mesoderm and make only moderate contributions to the neural crest. In our studies, we injected the C2/C3 blastomeres (Fig. 4A) with MOs against snail1, snail2 or twist1 RNAs together with RNA encoding GFP as a lineage tracer. Embryos were sorted at stage 10/10.5 (Fig. 4B) to select those accurately injected. When fixed at stage 18 and stained in situ, we found that sox9 expression was dramatically reduced in snail2 morphant embryos, but not in snail1 or twist1 morphant embryos (Fig. 4C-F). Subsequent studies indicated that whereas all three MOs led to a reduction in the levels of the mesodermal markers xbra (stage 11) and myoD (stage 25), only the snail2 MO produced a decrease in the neural border/early neural crest marker c-myc (stage 16) (Bellmeyer et al., 2003) and in the later neural crest marker sox9 (stage 18) (Lee et al., 2004; Spokony et al., 2002) (Fig. 4F). Preliminary studies in X. tropicalis indicate that the same pattern holds true there: C2/C3 snail2, but not snail1 or twist1, morphant embryos show a loss of sox9 expression (see Fig. S4 in the supplementary material).

The effects of C2/C3 snail2 MO injection on sox9 expression could be partially rescued by injection of high levels (600 pg/embryo) of snail2, snail1 or twist1 RNAs (Fig. 4G-I). RNA titration studies indicated that snail2 RNA was more effective at every dose (150 to 600 pg/embryo) than snail1 or twist1 RNAs at rescuing the snail2 morphant phenotype (Fig. 4J-M).

At stage 11, the dorsolateral mesodermal zone (DLMZ) of C2/C3 injected morphant embryos was dissected and analyzed by quantitative RT-PCR (Fig. 4N). The levels of RNA encoding the mesodermal markers eomes, myf5 and tbx6 (Hug et al., 1997; Li et al., 2006; Lou et al., 2006; Ryan et al., 1996) were preferentially reduced in snail2 morphant DLMZ. Interestingly, fgf8 RNA levels (Fletcher et al., 2006; Fletcher and Harland, 2008; Monsoro-Burq et al., 2003) appeared to be selectively regulated by Snail1. To resolve this and other issues, we have begun RNA SEQ analysis of morphant X. tropicalis embryos.

Because C2/C3 blastomeres give rise not only to mesoderm but also to other tissues, including neural crest, we examined inductive interactions between ectoderm (animal cap) and DLMZ explants. On its own, the wild-type animal cap remains as atypical ectoderm, expressing only low levels of the neural crest marker sox9 compared with whole embryos analyzed at the same stage (Fig. 5A,F). As demonstrated previously, DLMZ induces a number of markers of neural crest, including snail2, foxD3, zic5 and sox9 (Bonstein et al., 1998; Monsoro-Burq et al., 2003). Fig. 5B,F illustrates the effect of wild-type DLMZ on sox9 expression in explants. snail2 morphant DLMZ induced a much reduced level of sox9 expression (Fig. 5C,F), whereas twist1 morphant DLMZ behaved very much like wild-type DLMZ (Fig. 5E,F). snail1 morphant DLMZ explants produced a reduced, but readily detectable, level of sox9 expression (Fig. 5D,F).

Previous studies have implicated FGF, Wnt and BMP signaling in mesodermal induction of the neural crest in X. laevis (Monsoro-Burq et al., 2003; Monsoro-Burq et al., 2005; Steventon et al., 2009). We examined this situation in the context of C2/C3 snail2 morphant embryos. Injection of RNAs (25 pg/embryo) encoding either BMP4 or Wnt8 were able to substantially rescue sox9 expression (Fig. 6A-D,F). When wnt8 and bmp4 RNAs were injected together, there was an improved (although not dramatic) rescue response (Fig. 6E,F). Injection of fgf8 RNA (25 pg/embryo) alone had little effect on the sox9 expression phenotype (Fig. 6F), and was not studied further. At lower RNA levels (10 pg/embryo), neither bmp4 nor wnt8 RNAs rescued the snail2 C2/C3 morphant sox9 phenotype, but together they produced a strong rescue (Fig. 6G). Levels of chordin RNA, which encodes a secreted BMP signaling inhibitor (Sasai et al., 1995; Sasai et al., 1994), were upregulated in snail2 C2/C3 morphant embryos (Fig. 6H,I). qPCR analyses of C2/C3 injected, and subsequently dissected, DLMZ indicate that levels of wnt8, bmp4 and frzb1 RNAs were decreased, whereas levels of cerberus, sizzled and chordin RNAs were increased in snail2 morphant tissue; a distinctly different pattern of changes in RNA levels was observed in snail1 and twist1 morphant tissues (Fig. 6J).

The role of mesoderm in neural crest induction in X. laevis is well established (Bonstein et al., 1998; Green et al., 1997; Mayor et al., 2000; Monsoro-Burq et al., 2003; Steventon et al., 2009; Zhang et al., 2006; Zhang and Klymkowsky, 2009). Besides demonstrating that snail1 is involved in mesoderm formation/maintenance and neural crest induction, here we report that Snail2 plays a distinct role in the DLMZ of the late blastula/early gastrula stage embryo. Loss of Snail2 in this embryonic region led to the loss of expression of the early neural border/neural crest marker c-myc and of the late neural crest marker sox9. That the effect on neural crest is mediated by inductive signals is supported by `sandwich'-type explant studies between wild-type ectoderm and morphant DLMZ. Rescue studies suggest that both BMP4 and Wnt8 are essential components of this inductive signal; although this appears to partially contradict the conclusions of Steventon et al. (Steventon et al., 2009), it is likely that the details of the different experimental scenarios influence the behavior of the embryo (for example, differences in the concentrations and potency of the RNAs used, which can be influenced by the efficiency of the in vitro capping reaction.) Our conclusion is bolstered by the observation that the loss of snail2 activity led to a decrease in the levels of bmp4 and wnt8 RNAs, as well as to a reduction in the RNA encoding the selective Wnt inhibitor Frzb1 (Fig. 6J) (Leyns et al., 1997; Wang et al., 1997a; Wang et al., 1997b) and to increases in the levels of RNAs encoding the BMP4 antagonist Chordin, the Wnt and BMP antagonist Sizzled, and the Wnt, BMP and Nodal antagonist Cerberus. cerberus is an apparent `immediate-early' target of Snail2 regulation (Zhang and Klymkowsky, 2009). Together, these observations suggest that it is a balance between Wnt and BMP signaling that is critical for mesodermal induction of neural crest.

Although often considered in isolation, there is growing evidence that the BMP and Wnt signaling pathways interact (Itasaki and Hoppler, 2010; Katoh, 2010; Row and Kimelman, 2009). For example, it appears that the secreted Wnt inhibitors Dkk1 and Sost are regulated by BMP signaling (Kamiya et al., 2010) and that BMP signaling requires functional Wnt signaling (Tang et al., 2009). The Wnt signaling modulator Lpr4 interacts with Wise, a secreted and endoplasmic reticulum-localized protein that can either enhance or inhibit Wnt signaling (Guidato and Itasaki, 2007; Guidato et al., 1996; Itasaki et al., 2003). The Lrp4-Wise complex can, in turn, modulate BMP activity (Lintern et al., 2009; Ohazama et al., 2010). It is worth noting that the DLMZ of the late blastula/early gastrula stage X. laevis embryo is characterized by the expression of a number of other secreted signaling agonists and antagonists, in addition to chordin, sizzled, cerberus, frzb1, bmp4 and wnt8; these include other canonical and non-canonical acting Wnts (Kuhl, 2002), the TGFβ inhibitor Follistatin (Fainsod et al., 1997; Iemura et al., 1998; Tashiro et al., 1991), the BMP inhibitor Noggin (Smith and Harland, 1992) and the Wnt inhibitors Dkk, SFRP2, Sizzled2 and Crescent (Bradley et al., 2000; Glinka et al., 1998; Semenov et al., 2001; Shibata et al., 2000), as well as membrane-bound signaling regulators such as Kremen (Nakamura and Matsumoto, 2008). qPCR analyses support the view that the differences in the effects of snail1, snail2 and twist1 MOs in the C2/C3 lineage are due to differences in their regulatory targets (Fig. 6J).

It is very likely that these signaling factors interact. For example, Steventon et al. (Steventon et al., 2009) reported that snail2 RNA levels are reduced in chordin morphant embryos. Assuming a simple (and certainly incorrect) linear model for the BMP-Chordin interaction, loss of chordin expression would increase BMP signaling activity, thereby mimicking the ventral region of the embryo, where snail2 (as well as snail1 and twist1) RNA levels are low compared with the dorsal region (Zhang and Klymkowsky, 2009). Dissecting the interactions involved in the mesodermal induction of the neural crest network demands a much more global analysis – an analysis that we have begun by exploiting the recently released genomic data for X. tropicalis (Hellsten et al., 2010). Together with RNA and ChIP SEQ data, it should be possible to identify and study, in detail, the differences in the regulatory targets of Snail2, Snail1 and Twist1 in the DLMZ. In this light, it is worth noting that preliminary studies indicate that loss of snail2 function in the C2/C3 lineage in X. tropicalis (see Fig. S3 in the supplementary material) causes a loss of sox9 expression similar to that observed in X. laevis, suggesting that a similar regulatory network is involved in the two species.

We thank the Xenopus community, particularly Ethan Lee, Kristin Kroll, Hazel Sive, Eddy DeRobertis, J. Wysocka and Richard Harland for reagents; Mustafa Khokha and Richard Harland for advice on X. tropicalis; and the NIH (GM84133) for financial support. Deposited in PMC for release after 12 months.