Tsukushi controls ectodermal patterning and neural crest specification in Xenopus by direct regulation of BMP4 and X-delta-1 activity

Vertebrate neural crest cells originate at the border between the neural plate and the epidermis in the embryonic ectoderm. After their specification,neural crest cells migrate extensively along defined pathways, ultimately giving rise to diverse cell types that include the peripheral nervous system,melanocytes and the craniofacial skeleton(Hall, 1999; Le Douarin and Kalcheim,1999). Several factors secreted from the neural plate, non-neural ectoderm or mesoderm show neural crest inducing activity in ectodermal explants, when applied alone or in combination. Yet, how the neural crest is specified at the junction between the neural plate and the epidermis remains unclear.

It has been proposed that bone morphogenetic protein (BMP) signaling plays an important role in specification of cell fates at the neural plate/ectodermal border in mouse, chick and frog embryos(Kanzler et al., 2000; Liem et al., 1995; Marchant et al., 1998). During neural development in Xenopus, BMPs are first expressed in the whole ectoderm, but they are then gradually downregulated in the presumptive neural plate, while BMP antagonists such as Noggin, Chordin and Follistatin are secreted from the dorsal axial mesendoderm (reviewed by Sasai and De Robertis,1997). Several in vitro experiments suggest that a gradient of BMP activity in the ectoderm is established by interactions between BMPs expressed in the ectoderm and BMP inhibitors secreted from the axial mesendoderm,resulting in the specification of the neural crest at a specific threshold of BMP signaling levels (LaBonne and Bronner-Fraser, 1998; Marchant et al., 1998; Nguyen et al.,1998). However, neural crest formation is specifically restricted to the posterior border of the neural plate, excluding the anterior neural plate (Hopwood et al., 1989; Mayor et al., 1995). This cannot be explained by the BMP gradient alone, and posteriorizing factors such as Wnts, fibroblast growth factors (FGF) and retinoic acid (RA), are required for this anteroposterior localization(Villanueva et al., 2002). Activation of Wnt signaling in the Xenopus embryo or in neuralized animal cap induces neural crest specification, while Wnt inhibition by either Wnt8 knockdown or overexpression of a dominant-negative Tcf3abrogates neural crest specification in zebrafish(LaBonne and Bronner-Fraser,1998; Lewis et al.,2004; Mayor et al.,1995; Tan et al.,2001). These results led to the two-signal model, suggesting that the neural crest is specified in the posterior neural plate border by the intersection of canonical Wnt signaling with moderate levels of BMP signaling(LaBonne and Bronner-Fraser,1998; Marchant et al.,1998; Villanueva et al.,2002). Recently, the Notch signaling pathway has also been shown to play an important role in neural crest specification, by regulating the expression of BMP4 and the BMP target gene Msx1 at the lateral edge of the neural plate (Glavic et al., 2004).

Here, we show that the Xenopus Tsukushi gene (X-TSK) is strongly expressed at the border of the neural plate at the time of neural crest specification, and that its expression in the ectoderm is regulated by BMP signaling. By biochemical analysis and overexpression assays in Xenopus, we show that: (1) X-TSK works as a BMP antagonist by direct binding to BMPs; (2) X-TSK modulates activation of Notch signaling by directly binding to X-delta-1 extracellular region. Furthermore, X-TSK can regulate BMP4 transcription indirectly via modulation of the Notch signaling pathway. Both gain- and loss-of-function assays demonstrate that X-TSK plays a crucial role in patterning of the ectoderm and especially in neural crest specification. We argue that X-TSK functions in the crosstalk of BMP and Notch signaling pathways at the boundary between the neural and non-neural ectoderm to determine the correct specification of the neural crest progenitors.

A Xenopus cDNA library at stage 24 (high density cDNA filter 725,RZPD) was screened using C-TSK cDNA as a probe to isolate X-TSK cDNA (GenBank Accession Number, AB176536).

For RT-PCR, total RNA was isolated by RNeasy Kit (QIAGEN) and TRIZOL purification system (Invitrogen). The obtained total RNA was treated with DNase (Invitrogen) for removing genome DNA contamination, and first-strand cDNA was synthesized using oligo d(T)_12-18 primer and Superscript III reverse transcriptase (Invitrogen). PCR was performed with the Expand Long template PCR system 3 (Roche), with the following primers: Xslug-U 5′-CAATGCAAGAACTGTTCC-3′; Xslug-D 5′-TCTAGGCAAGAATTGCTC-3′; XAG-1(Sive and Bradley, 1996); XBF-1U, 5′-TCAACAGCCTAATGCCTGAAGC-3′; XBF-1D,5′-GCCGTCCACTTTCTTATCGTCG-3′; Xotx2(Blitz and Cho, 1995); Sox2 (De Robertis et al.,1997); Msx1(Tríbulo et al., 2003); BMP4 (Dale et al.,1992); XK81 (LaBonne and Bronner-Fraser, 1998); ODC(Agius et al., 2000); Cardiac actin (Stutz et al., 1986); Sox9 and Zic5(Monsoro-Burq et al., 2003); XESR-1 (Wittenberger et al.,1999); X-notch-1U, 5′-TCCTGATTTATATTGCTTATCCGAGT-3′;and X-notch-1D, 5′-TTACAGAAGTGTTAACAGCAACAACA-3′.

Xenopus embryos were obtained as previously described(Newport and Kirschner, 1982)and staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1967). For animal cap assays, mRNA was injected into the animal pole of two- or four-cell stage embryos. Animal caps (ACs) were dissected from stage 8-9 embryos in 1× MBS and cultured in 0.5× MBS until early neurula(stage 14), mid neurula (stage 17) or early tailbud (stage 21/22) stages,depending on the experiment. For dissections of ventral marginal zone (VMZ) or dorsal marginal zone (DMZ) explants, embryos were marginally injected into ventral or dorsal blastomeres at the four-cell stage with X-TSK, BMP4or BMP4 +X-TSK mRNAs. After culturing to the early gastrula stage (stage 10-10⁺), VMZ or DMZ explants (comprising about 60°of the VMZ or DMZ) were dissected in 1× MBS and cultured in 0.5×MBS until stage 24-26. Dorsal lateral marginal zone (DLMZ)-AC conjugates were prepared as described by Bonstein et al.(Bonstein et al., 1998).

Whole-mount in situ hybridization was performed as previously described(Harland, 1991; Shain and Zuber, 1996). Pigmented embryos and explants were bleached after the color reaction(Mayor et al., 1995). The following probes were used: Xdlx3(Woda et al., 2003); ADAM13 (Alfandari et al.,1997); XAG-1 (Bradley et al., 1996); XK81(Jonas et al., 1985); Sox2 (Mizuseki et al.,1998); Xrx1 (Casarosa et al., 1997); N-tubulin(Richter et al., 1988);cardiac actin (Mohun et al.,1984); Xotx2 (Pannese et al.,1995); Xotx5b (Vignali et al.,2000); X-ESR-1(Wettstein et al., 1997). X-delta-1 was cloned independently(Kiyota et al., 2001). Xbmp4 cDNA was subcloned into pBSSK (-) from a BMP4expression vector. Xslug, Sox9, Zic5, Msx-1, Hairy2A and Trp-2 cDNAs were isolated by RT-PCR.

We raised polyclonal antibody against X-TSK in rabbits (QIAGEN). The specificity of this antibody was checked by SDS-PAGE and western blot. Myc-His tagged X-TSK and X-TSK protein in the lysate of embryo can be detected in blot(data not shown). For double or triple staining, the hybridized embryo was sectioned by cryostat, and detected by anti-X-TSK antibody with Cy3-conjugated secondary antibody (Jackson Immuno Research).

Capped mRNAs were synthesized from linearized plasmid templates with the mMessage Machine kit (Ambion). Embryos were injected with 500-4000 pg mRNA/embryo at the two- or four-cell stage in 0.1× MMR with 4% Ficoll and kept in it for a few hours. They were then transferred into 0.1× MMR for subsequent culturing. The following RNAs were made as previously described(Kuriyama and Kinoshita,2001): X-TSK (pCS2+-X-TSK), truncated BMP receptor (tBR) (Graff et al.,1994), Chordin (Sasai et al., 1994), BMP4(Fainsod et al., 1994), Xnr1(Jones et al., 1995), Notch Intracellular domain, X-delta-1 and dominant-negative form of Delta (Chitnis et al.,1995), Suppressor of hairless DNA binding mutant, and Su(H)/Notch ankyrin repeats (Ank) fusion construct(Wettstein et al., 1997).

We used the computer program provided on Dr Gregory Hannon's website(http://www.cshl.edu/gradschool/hannon.html)to predict the short-hairpin-siRNA sequence for gene silencing of X-TSK. Myc-His tagged X-TSK was transfected into COS-7 cells with either a control or X-TSK siRNA-expressing vector and downregulation of X-TSK protein synthesis was checked by western blotting. After selecting the effective sequences, we used a X-TSK siRNA designed for the following sequence: 5′-GATACCTCGTATCTGGATCTC-3′. C-TSK siRNA, which degrades C-TSK mRNA effectively, was used as control siRNA(Ohta et al., 2004). C-TSK mRNA was used for a rescue experiment of X-TSK siRNA. The applicable region of the C-TSK sequence was 5′-GATACCTCCTACTTGGATTTG-3′. In another loss-of-function experiment, we used a morpholino antisense oligo (MO)against X-TSK, designed to target the following sequence:5′-TCTAACAATGGCTCTTTCTTCTTGG-3′ (Gene Tools). To distinguish the effect of MO on lateral expression of X-TSK from MO injection to deplete the organizer expression (Ohta et al.,2004), we injected it into ventral border of animal dorsal blastomere at the eight-cell stage corresponding to future D121 and D122 in the cell fate map of 32-cell stage embryo(Moody, 2000).

Myc-His tagged X-TSK, Myc-His tagged C-TSK, a Flag tagged BMP4 plasmids or a Flag tagged X-delta-1 extracellular domain plasmids were transfected into COS-7 cells with LipofectAMINE 2000 (Invitrogen) and the supernatants were harvested after 96 hours culture in serum-free Opti-MEM (Invitrogen). The immunoprecipitation experiment was performed as previously described(Ohta et al., 2004). Myc-tagged and Flag-tagged proteins were detected after blotting using an anti-Myc antibody 9E10 or anti-Flag antibody M2 (Sigma), respectively.

mRNA electroporation was performed as described by Sasagawa et al.(Sasagawa et al., 2002). mRNA(200 nl) of X-TSK (2 μg/μl) and GFP (1 μg/μl), or only GFP (3μg/μl) were injected into the space under the vitelline membrane, and electroporated at stage 11.5. The position of electroporation was at the border between the lightly pigmented dorsal ectoderm and darkly pigmented lateral ectoderm. Electroporated embryos were selected under the fluorescent microscope for GFP expression at stage 14, fixed at the stage 15/16, then analyzed by whole-mount in situ hybridization.

The same procedure was used as previously described(Berry et al., 1998). Samples were fixed overnight in 4% paraformaldehyde/PBS, dehydrated in 95% ethanol and stained with Alcian Blue solution [95% ethanol:acetic acid:0.3% Alcian Blue in 70% ethanol (8 ml:2 ml:1 ml)]. After the staining, samples were left in 95%ethanol for 2 days and then stained with Alizarin Red solution [0.1% Alizarin Red in 95% ethanol:1% potassium hydroxide (0.25 ml:10 ml)] for 2 days. After 6 hours incubation in 1% potassium hydroxide, samples were transferred into 100%glycerol solution for image acquisition.

Chick Tsukushi (C-TSK) consists of twelve leucine-rich repeats (LRRs) and two cysteine-rich clusters at both termini and belongs to the small leucine-rich proteoglycan (SLRP) family(Ohta et al., 2004). A Xenopus tailbud cDNA library was screened using C-TSK as a probe, obtaining a cDNA fragment encoding for a predicted protein of 338 amino acid residues, which corresponds to the Xenopus TSK gene(X-TSK). We have previously reported about other Tsukushi homologues in zebrafish, mouse and human (Ohta et al., 2004), all of which have similar characteristics of structure. At the amino acid level, X-TSK shares 59% identity with C-TSK, and 48-49% identity with zebrafish, mouse and human TSK.

Maternal X-TSK expression is observed in the animal hemisphere from unfertilized eggs to early gastrula embryos(Ohta et al., 2004). However,at the early neurula stage (stage 13), X-TSK expression is hardly detectable in the presumptive neural plate region, and restricted to the non-neural ectoderm (Fig. 1A),where its levels increase by stage 14, especially in the presumptive anterior neural fold (Fig. 1C). At these stages, the spatiotemporal expression pattern of X-TSK closely resembles that of Xbmp4 (Fig. 1D) (Fainsod et al.,1994; Hemmati-Brivanlou and Thomsen, 1995; Schmidt et al.,1995) and BMP target genes, such as Xdlx3(Fig. 1B)(Woda et al., 2003).

In addition, X-TSK is expressed in the prospective cranial neural crest, in a similar domain to that of Xslug(Fig. 1E,F)(Mayor et al., 1995). Sections of hybridized embryos at these stages revealed that X-TSK is expressed in a broad area overlapping with the neural fold, but it is hardly observed inside the neural plate or in the underlying mesoderm(Fig. 1I), while Xslugexpression is restricted to the neural crest region(Fig. 1J). We immunostained Xslug hybridized embryos with an anti-X-TSK antibody on sections(Fig. 1K,L). X-TSK protein is expressed in the superficial layer of the epidermis and the neural crest region, and also in the proximal edge of the neural crest region in the same layer of the Xslug expression domain(Fig. 1L, arrowhead). At the early tailbud stage (stage 23), X-TSK expression is observed in cranial neural crest cells, the dorsal retina and the lens placode(Fig. 1G). Comparison with Xbmp4 showed that X-TSK has a more distinct expression pattern at this stage (Fig. 1H). In the migrating cranial neural crest, X-TSKexpression is observed strongly in the mandibular crest segment, weakly in the distal tip of the hyoid crest segment, and in the anterior and posterior branchial crest segments (Fig. 1M). This expression is similar to that of Sox9(Fig. 1N)(Spokony et al., 2002) and ADAM13 (Fig. 1O)(Alfandari et al., 1997), while Xslug is downregulated in neural crest cells after they leave the neural tube (Fig. 1P).

As the early expression domains of X-TSK and Xbmp4 in the ectoderm partially overlap, we speculated that X-TSK expression might be under the control of BMP signaling. To test this hypothesis, we analyzed X-TSK expression in stage 14 animal caps from embryos injected at the two-cell stage with either 0.5 ng of the BMP antagonist Chordin mRNA or 1 ng of truncated, dominant-negative, BMP2/4 receptor(tBR) mRNA. In accordance with its expression in the non-neural ectoderm (Fig. 2I), X-TSK was clearly expressed in uninjected caps(Fig. 2D,G), which have epidermal character as shown by expression of the epidermal marker XK81 (Fig. 2A). By contrast, inhibition of BMP signaling by injection of Chordin or tBR mRNAs downregulated X-TSK expression, similar to XK81 (Fig. 2B,E,H). In caps co-injected with 0.5 ng Chordin and 1 ng of BMP4 mRNAs,both XK81 and X-TSK expression were restored(Fig. 2C,F). Similar results were obtained by semi-quantitative RT-PCR(Fig. 2J). In particular, Chordin-injected caps showed dose-dependent upregulation of X-TSK expression when increasing amounts of BMP4 mRNA were co-injected with Chordin mRNA, at both stages 14 and 17. These results indicate that BMP activity positively regulates X-TSKexpression in the ectoderm.

In a previous study, we showed that C-TSK has dorsalizing activity when overexpressed in Xenopus embryos, which is due to its ability to bind to BMPs directly and antagonize BMP signaling in the extracellular space(Ohta et al., 2004). In addition, C-TSK expression partially overlaps with BMP expression only at the posterior marginal zone and the posterior primitive streak during chick gastrulation. However, the expression of X-TSK largely overlaps with BMP4 in the ectoderm of Xenopus neurula embryos, and is controlled by BMP4. Thus, to confirm whether these dorsalizing and anti-BMP activities are conserved in X-TSK, we performed overexpression experiments of X-TSK mRNA in Xenopus embryos, and found that X-TSKcan dorsalize ventral mesoderm both in ventral marginal zone explants and in animal caps injected with low doses of Xnr1 mRNA, as detected by explant elongation and induction of cardiac actin expression (see Fig. S1 in the supplementary material). In addition, in co-injection experiments, X-TSK can antagonize the ventralizing effects of BMP4 overexpression in dorsal marginal zone explants (see Fig. S1 in the supplementary material). To examine whether X-TSK can bind to BMP4 directly, we carried out an immunoprecipitation assay between BMP4 and X-TSK or C-TSK proteins(Fig. 3A). When Myc-tagged X-TSK or C-TSK was reacted with FLAG-tagged BMP4, immunoprecipitation of both proteins with nickel chelating resins pulled down BMP4. These data indicate the direct binding of X-TSK to BMP4. Thus, the basic molecular characteristics of Tsukushi are also conserved in the Xenopus homologue.

BMP signaling has a crucial role in mediolateral patterning of the ectoderm(Knecht and Harland, 1997; Wilson et al., 1997). To further test the ability of X-TSK as a BMP antagonist, we performed animal cap assays and analyzed the induction of the dorsoanterior ectodermal markers. Using whole-mount in situ hybridization, we could detect strong induction of cement gland markers, but weak induction of neural markers, in X-TSK-injected caps (see Fig. S2 in the supplementary material). We then analyzed X-TSK-injected caps by semi-quantitative RT-PCR, which provides more sensitivity than in situ hybridization(Fig. 3B). In these experiments, X-TSK induced the cement gland marker XAG-1,the telencephalic marker XBF-1, the forebrain and cement gland marker Xotx2, and the pan-neural marker Sox2 in a dose-dependent manner (Fig. 3B, lanes 3-5). Interestingly, X-TSK overexpression had only a weak, if any, effect on the expressions of BMP4 and its target gene Msx1(Fig. 3B). Comparison with the effects of another BMP antagonist, Noggin, showed that, even at high doses of injected mRNA (4 ng), the effects of X-TSK mRNA on the induction of cement gland/neural markers and on the repression of BMP4 and Msx1 were clearly weaker than those of Noggin mRNA. Co-overexpression of X-TSK and Nogginalso slightly reinforced the effects of Noggin(Fig. 3B, lanes 1,2). These effects were direct and not due to the induction of dorsal mesoderm, as no activation of cardiac actin was detected in the injected caps. Altogether,these data indicate that, while BMP antagonists such as Noggin can strongly inhibit BMP activity and strongly neuralize ectodermal explants, X-TSK overexpression causes a moderate repression of BMP signaling. Thus, X-TSK may induce peripheral neural plate character in the ectoderm via direct induction of intermediate levels of BMP signaling.

Several studies in Xenopus have suggested that intermediate levels of BMP signaling are necessary for neural crest specification at the neural plate border (LaBonne and Bronner-Fraser,1998; Marchant et al.,1998). Thus, it was interesting to see, as described above, that X-TSK is a BMP antagonist strongly expressed in the presumptive cranial neural crest, which can promote intermediate levels of BMP signaling in ectodermal explants. To determine the role of X-TSK, we performed loss-of-function experiments using siRNA(Zhou et al., 2002). The inhibitory effect of X-TSK siRNA (X-TSK-si) sequence was confirmed by the inhibition of X-TSK-Myc protein production after co-transfection in COS cells(data not shown). X-TSK-si was also effective in reducing the levels of endogenous X-TSK mRNA in Xenopus embryos, as shown by RT-PCR(Fig. 4C). Although the endogenous expression of X-TSK increased from stage 13 in uninjected embryos, this increase was completely prevented in embryos injected with X-TSK-si. Injection of X-TSK-si into one animal blastomere at the four-cell stage (0.3-0.5 pmol/cell) caused abnormal neural fold formation in the neural plate and increased pigmentation at the site where the neural fold is normally formed (Fig. 4A,B).

We then analyzed the effects on X-TSK depletion on ectodermal patterning using molecular markers after co-injection of X-TSK-si with β-gal mRNA(see Table 1). The expression of the neural crest markers Sox9, Zic5 and Xslug was inhibited in the β-gal positive area(Fig. 4E,G,L). No effects were detected after injection of C-TSK siRNA (C-TSK-si)(Fig. 4D,F). In addition, while the expression of the neural plate markers Sox2 and Xrx1were not affected by X-TSK-si injection(Fig. 4H,J), the expression of the epidermal marker XK81 was enhanced(Fig. 4I) to the same extent of the increased pigmentation as shown in Fig. 4B. In these embryos, no changes were detectable in the expression of Xbmp4 (Fig. 4K). These effects were rescued by co-injection of both 500 pg C-TSK mRNA(Fig. 4L,M) or 250 pg X-TSK mRNA (Fig. 4N,O), indicating that these phenotypes are specifically caused by the depletion of the TSK protein (Table 2). Previously, we performed morpholino oligonucleotide (MO)injections into the prospective dorsal midline of the embryo to determine the effects of X-TSK depletion on the Spemann organizer and neural induction. In those experiments, we showed that X-TSK MO impairs anterior neural plate specification, as shown by a reduction in the expression domain of Sox2 and Xrx1 (Ohta et al., 2004). To confirm siRNA effects and to avoid indirect effects due to abnormal organizer formation, we performed MO injections into the prospective lateral ectoderm, as described in the Materials and methods section. In these conditions, both Sox9 and Xslug were diminished by X-TSK MO injection (Fig. 4P,Q). These effects were rescued by the injection of N-cad-X-TSK mRNA, where the N-terminal region of X-TSK, including the initiation codon and the signal peptides, were replaced by a myc-tagged N-cadherin signal peptide (Fig. 4R,S; see also Table 2). We also examined the effects of X-TSK depletion on the expression of Hairy2A and Msx1, which identify a pre-neural crest region before neural crest specification (Glavic et al.,2004; Tríbulo et al.,2003). Both Hairy2A and Msx1 were downregulated by X-TSK-si (Fig. 4U,W), but not C-TSK-si (Fig. 4T,V). To confirm the relationship between X-TSK protein reduction and the observed phenotypes, we sectioned the embryos hybridized with neural crest markers and immunostained them with an anti-TSK antibody(Fig. 4X). In the embryos showing downregulation of Sox9 and upregulation of XK81after X-TSK-si injection, X-TSK protein expression in the neural crest was reduced on the injected side, while X-TSK endogenous expression was detectable in the proximal edge of the neural crest region on the control side(Fig. 4X, arrowheads). Morphological and molecular analysis at later stages of development confirmed that, in agreement with the reduction of early neural crest markers at neurula stages, inhibition of X-TSK function strongly repressed formation of neural crest derivatives such as melanocytes and the branchial arch cartilages (see Fig. S3 in the supplementary material). Altogether, these results suggest that X-TSK is required at an early step of neural crest specification upstream of Hairy2A and Msx1, and that in the absence of TSK function in the ectoderm the presumptive neural crest region is at least partially specified as epidermis.

As previously described, dorsolateral marginal zone (DLMZ) explants dissected from early gastrula embryos at stage 10.25-10.5 induced expression of the neural crest marker Sox9 in conjugated animal caps grown to the late neurula stage (Fig. 5A; 78.9%, n=38)(Bonstein et al., 1998). Sox9 induction was much reduced in X-TSK-si-injected caps conjugates with DLMZ explants (Fig. 5B;15.9%, n=44), but not in uninjected caps conjugated with X-TSK-si-injected DLMZ (Fig. 5C; 75%, n=28). Expressions of the neural crest marker, Zic5 and Xslug were also decreased in a dose-dependent manner in X-TSK-si-injected caps after conjugation with DLMZ, as detected by RT-PCR analysis (Fig. 5G). By contrast, the expression of Sox9 was strongly increased in conjugates of DLMZ with animal caps overexpressing X-TSK mRNA(Fig. 5A,D; 100%, n=15), while it was decreased in uninjected caps conjugated with X-TSK-overexpressing DLMZ (Fig. 5A,E; 41.6%, n=24). These results indicate that X-TSK expression in the ectoderm enhances the response to neural crest inducing signals derived from DLMZ, while X-TSK overexpression in the dorsolateral mesoderm has the opposite effect. As X-TSK overexpression in the mesoderm has a dorsalizing effect (see Fig. S1 in the supplementary material), this opposite effect may be explained by dorsalization of the DLMZ.

In Xenopus embryos, neural crest specification requires the action of posteriorizing signals such as Wnts, FGFs and retinoic acid acting on cells with intermediate levels of BMP signaling at the border of the neural plate(Villanueva et al., 2002). These signals may be at least partially produced by the DLMZ(Bang et al., 1999; Monsoro-Burq et al., 2003; Monsoro-Burq et al., 2005; Wu et al., 2003). XWnt-8 overexpression induces ectopic neural crest markers in vivo(LaBonne and Bronner-Fraser,1998), and it can induce neural crest specification in animal caps in cooperation with BMP antagonists (Chordin or Noggin)(Christian et al., 1991; LaBonne and Bronner-Fraser,1998). Therefore, we tested whether X-TSK could induce neural crest specification in cooperation with XWnt-8. X-TSK-injected caps induced the cement gland marker XAG-1(Fig. 3B, Fig. 5I), but did not show any significant expression of the neural crest marker Xslug(Fig. 5L), while XWnt-8-injected caps showed weak expression of both XAG-1and Xslug (Fig. 5J,M). By contrast, when X-TSK and XWnt-8 were co-overexpressed, Xslug was strongly induced, while XAG-1 expression was reduced compared with X-TSK-injected caps(Fig. 5H,K). RT-PCR analysis confirmed that co-injection of X-TSK and XWnt-8significantly enhanced expression of the neural crest markers Sox9and Zic5, compared with single X-TSK or XWnt-8injection (Fig. 5N). Therefore,the BMP antagonistic activity of X-TSK is sufficient for neural crest specification in cooperation with XWnt-8.

In Xenopus, mRNA injections often show only earlier functions of the corresponding gene, preventing the analysis of later roles for the same gene. For example, while X-TSK or Chordin overexpression by mRNA injections into early embryos represses neural crest formation (see below), overexpression of a hormone-inducible version of the BMP antagonist Smad6 causes ectopic neural crest induction when its activity is induced after stage 9 (Wawersik et al.,2005). Thus, BMP antagonistic activity is required for neural crest formation after gastrulation. To study X-TSK function in vivo at these later stages, we performed mRNA electroporation in stage 11.5 Xenopus embryos (Fig. 5O-R). In these conditions, overexpression of X-TSK in the anterior neural fold extended the expression domain of Sox9(Fig. 5P, arrowhead), and induced ectopic expression in the epidermal region(Fig. 5Q, arrowhead), while GFP mRNA electroporation had no effect (Fig. 5O, arrowhead). Xslug expression was also enhanced in the electroporated side (Fig. 5R). The electroporation of X-TSK mRNA after early neurula stage 13 caused a slight enhancement of the cranial-facial neural crest at stage 21 (data not shown). These data strongly suggest that X-TSK upregulation in the lateral ectoderm after gastrulation during normal development(Fig. 1C,M) is involved in cranial neural crest specification.

In this study, gain-of-function and loss-of-function analyses clearly suggest the importance of X-TSK-mediated BMP antagonism in ectodermal patterning and neural crest specification. To shed light on X-TSK activity during ectodermal development, we performed additional overexpression experiments in Xenopus embryos. When X-TSK mRNA was injected into one dorsoanimal blastomere of four-cell stage embryos (1 ng/cell;co-injected with 200 pg of β-gal mRNA as a lineage tracer), the expression of the neural crest marker Xslug was suppressed on the injected side, while β-gal alone did not cause any specific phenotype(Fig. 6A,B). Sox9expression was also suppressed by X-TSK, and even more strongly by a membrane-bound form of X-TSK (X-TSK-CD2), obtained after fusion with the CD2 transmembrane domain(Chang et al., 2001)(Fig. 6C,D). The fact that the effects of both the secreted and membrane-bound forms of X-TSK were very similar, and that the secreted form of X-TSK affected only the area immediately around the β-gal-stained region, suggest that X-TSK acts as a short-range factor and does not disperse far away. Unilateral injection of X-TSK also caused the expansion of the neural plate marker Sox2 (Fig. 6E), the inhibition of the epidermal marker XK81(Fig. 6F), and a downregulation of Hairy2A and Msx-1 in the presumptive neural crest territory (Fig. 6K)(Table 3). β-Gal mRNA did not change the expression of Hairy2A and Msx-1 (data not shown) (Table 3). These effects were dose dependent. Similar to X-TSK overexpression, injection of Chordin mRNA, another BMP antagonist, also caused an expansion of the neural plate (Fig. 6G), and the inhibition of neural crest formation (Fig. 6H). These data confirm that X-TSK acts as a BMP antagonist in vivo, although less efficiently than Chordin. However, Chordin was not effective in restoring neural crest gene expression in X-TSK-depleted embryos (Fig. 6I,J), suggesting that the function of X-TSK in neural crest formation may not be simply explained by its anti-BMP activity.

In support of the above interpretation, we found that Chordininhibited Xbmp4 expression in the lateral ectoderm(Fig. 6N), while X-TSKoverexpression enhanced it (Fig. 6M), and that BMP4 overexpression also caused a repression of neural crest formation (Fig. 6O). It has been shown that Notch signaling and its target gene Hairy2A are involved in neural crest specification in Xenopus, by regulating the expression of BMP4 and Msx-1 at the lateral border of the neural plate(Glavic et al., 2004). In fact, similar to the results described by Glavic et al.(Glavic et al., 2004),overexpression of Notch-ICD mRNA into early embryos repressed Xslug expression (Fig. 6P), while overexpression of X-delta-1^Stu, a dominant-negative ligand of Notch, caused upregulation of Xbmp4expression (Fig. 6R). As these effects are similar to those of X-TSK overexpression, we then examined in more detail whether the effects of TSK overexpression are mediated by modulation of the Notch signaling pathway. We first checked the effects of TSK overexpression on primary neurogenesis, which is known to be controlled by Notch signaling(Chitnis et al., 1995). Analogous with what has previously been shown for Notch overexpression(Chitnis et al., 1995; Chitnis and Kintner, 1996), TSK overexpression repressed the expression of N-tubulin, a marker of differentiated neurons, and that of X-delta-1(Fig. 6S,T). Moreover, X-TSK overexpression upregulated the Notch-target gene, X-ESR-1 (Fig. 6V). Finally, we found that the repression of neural crest specification by TSK overexpression could be overcome by co-injection with Su(H)^DBM mRNA, which acts as an antagonist of Notch signaling (Wettstein et al.,1997) (Fig. 6W)(Table 3), while Su(H)^DBM mRNA alone very slightly enhanced Sox9expression (data not shown) (Table 3). By contrast, co-overexpression of X-TSK with Su(H)/Ank, which works as an agonist of Notch signaling as described by Wettstein et al. (Wettstein et al.,1997), strongly inhibited Xslug and Sox9(Fig. 6Z, data not shown)(Table 3). Su(H)/Ankalone also inhibited the expression of neural crest markers, though more weakly than Notch ICD (Fig. 6Y) (Table 3). However, Xbmp4 upregulation by X-TSK overexpression in the lateral ectoderm could not be rescued by co-injection of Su(H)^DBM mRNA (Fig. 6X) (Table 3),consistent with the fact that the same effect is obtained after overexpression of X-delta-1^Stu (Fig. 6R). By contrast, the upregulation of Xbmp4 expression by X-TSK overexpression could be partially prevented by co-injection of Su(H)/Ank mRNA, though some patches of ectopic Xbmp4expression in the lateral area of β-gal-positive cells were still detectable (Fig. 6BB, arrows). Su(H)/Ank mRNA alone inhibited Xbmp4 expression very weakly(Fig. 6AA, arrowhead). Altogether, these results suggest that TSK overexpression has complex, bimodal effects on the regulation of Notch signaling. Specifically,the repression of neural crest markers and primary neurogenesis appear to be consistent with enhanced Notch signaling, while the activation of Xbmp4 transcription seems to be related to diminished Notch signaling.

Our results suggest that X-TSK overexpression can modulate Notch signaling in the neural crest region. In order to determine whether there is a direct interaction between X-TSK and the Notch signaling pathway, we stained Notch-transfected cells and Delta-transfected cells with TSK-myc tagged protein and anti-myc antibody. Binding of TSK protein to Delta-transfected cells, but not to Notch-transfected cells, was detectable (data not shown). To confirm this binding, we performed immunoprecipitation experiments, and found that the X-TSK protein precipitates the X-delta-1 extracellular domain(Fig. 7A). This result suggests that X-TSK modulates Notch signaling by binding to Delta in the extracellular space.

To clarify whether TSK works as an activator or an inhibitor of Notch signaling, we performed animal cap experiments by using the Notch-target gene XESR-1 expression as a readout for the activation of the Notch signaling pathway (Fig. 7B). As the expression of X-delta-1 in early gastrula stages is localized to the marginal zone, uninjected animal caps express Notch, but not X-delta-1 (Wittenberger et al.,1999). Thus, Notch activation does not normally occur in isolated animal caps (Kiyota and Kinoshita,2002). We then activated Notch signaling in animal caps by injection of X-delta-1 mRNA, and assessed the effects of X-TSK gain-or loss-of-function on this activation. Synthetic mRNAs and/or X-TSK MO were injected into four-cell stage embryos, animal caps were excised from stage 8.5 embryos, and analyzed by RT-PCR at stage 9.5. As expected, both Notch ICD mRNA and X-delta-1 mRNA caused induction of XESR-1(Fig. 7B, lanes 3, 4). By contrast, co-injection of X-TSK with X-delta-1 caused a much lower induction of XESR-1 compared with X-delta-1 alone,while X-TSK alone could not induce X-ESR1(Fig. 7B, lanes 4-6). Remarkably, co-injection of X-TSK-MO together with X-delta-1 also strongly blocked X-ESR1 activation by X-delta-1(Fig. 7B, lanes 7, 8). Taken together, these data suggest that the endogenous levels of X-TSK present in uninjected animal caps (Fig. 2D,J) are required for activation of Notch signaling by X-delta-1. Conversely, higher levels of TSK, as those resulting from X-TSK overexpression, can cause inhibition of X-delta-1 activity, suggesting that X-TSK may differentially modulate X-delta-1 activity in a dose-dependent manner in vivo.

In this paper, we describe the Xenopus homolog of the Tsukushi gene. The present results suggest that X-TSK is required for proper ectodermal patterning and neural crest specification,acting both as a BMP antagonist and as a modulator of Notchsignaling.

X-TSK encodes for a secreted protein belonging to the SLRP family. Although some members of this family, such as decorin, have previously been shown to bind TGF-β and modulate its activity(Yamaguchi et al., 1990),their role during embryonic development was not clear. We recently described the chick TSK homolog and showed that it works as a BMP antagonist during chick gastrulation (Ohta et al.,2004), while another SLRP member, Biglycan, has also been shown to modulate BMP activity during Xenopus early development(Moreno et al., 2005). Similar to its chick counterpart, X-TSK works as a BMP antagonist, as indicated by overexpression experiments in Xenopus embryos and in vitro assays. In fact, X-TSK overexpression can dorsalize ventral mesoderm (see Fig. S1 in the supplementary material), and it can induce cement gland and neural tissue, but not the dorsal mesodermal marker cardiac actin in animal caps(Fig. 3, see Fig. S2 in the supplementary material), thus mimicking the effects of other known BMP inhibitors, such as Chordin, Noggin or a dominant-negative BMP receptor (Sasai et al., 1995; Sive and Bradley, 1996). In whole embryos, X-TSK overexpression can enhance dorsoanterior fates and repress ventrolateral fates, and these effects are similar to those produced by Chordin overexpression(Fig. 6G). The opposite effect,namely a reduction of dorsal and an expansion of ventral structures, is observed after X-TSK depletion (Fig. 4) (Ohta et al.,2004). Finally, X-TSK can antagonize the ventralizing activity of BMP4 in mesodermal explants, and it can bind BMP proteins in vitro. Altogether, these data indicate that X-TSK is endowed with a clear BMP antagonistic activity.

However, not all the effects of TSK gain or loss of function can be simply ascribed to its anti-BMP function. In particular, although chick TSK could efficiently rescue the reduction in neural crest formation caused by X-TSK depletion, Chordin could not(Fig. 4M-O,R,S, Fig. 6I,J). Moreover, although Chordin overexpression repressed BMP4 expression in the ectoderm, X-TSK overexpression had the opposite effect(Fig. 3B, Fig. 6M,N). These data suggest that X-TSK has some signaling activities that are different from BMP antagonism. Thus, it is interesting to see (Figs 6 and 7) that modulation of the Notch signaling pathway is also involved in X-TSK function, as evidenced by the direct binding of X-TSK to X-delta-1 in vitro(Fig. 7A). Moreover, both Notch ICD and X-TSK had similar effects on neurogenesis(Chitinis et al., 1995), neural crest formation(Glavic et al., 2004) and the expression of the Notch-target gene X-ESR-1(Wettstein et al., 1997). Finally, suppression of neural crest formation by X-TSKoverexpression was efficiently rescued by a dominant-negative Su(H),which inhibits Notch signaling, but not by an active form of Su(H)(Fig. 6). Together, these data clearly suggest that X-TSK may also work, at least in part, via activation of Notch signaling. However, the effect of X-TSK on BMP4 expression in the anterior peripheral region of the neural plate is similar to that of a dominant-negative Notch ligand, X-delta-1^Stu(Glavic et al., 2004), and this X-TSK effect can be prevented, at least in part, by an active form of Su(H), but not by a dominant-negative Su(H)(Fig. 6). This suggests that X-TSK overexpression may also cause inhibition of Notch signaling at the neural/epidermal border. To gain more insight into the interaction between X-TSK and X-delta-1, we performed gain- and loss-of-function experiments of X-TSK in isolated animal caps where Notch signaling was simultaneously activated by X-delta-1 overexpression(Fig. 7B). Surprisingly, we found that while the endogenous levels of X-TSK in isolated animal caps are necessary for X-delta-1 activity, higher X-TSK levels resulting from X-TSK overexpression inhibits X-delta-1 function. Although further work will be needed to dissect the biochemical mechanism by which X-TSK exerts these dual effects on X-delta-1 activity, these results identify X-TSK as a new modulator of the Notch signaling pathway, which works by directly binding X-delta-1 in the extracellular space.

The work described in this paper uncovers a novel function for TSK family members, i.e. the control of neural crest specification. During gastrulation and neurulation, X-TSK expression is downregulated in the middle of the neural plate, while it remains in the non-neural ectoderm and it accumulates to strong levels at the neural plate border(Fig. 1). This is the region where the neural crest is specified, and therefore it was reasonable to see that both X-TSK gain and loss of function affected neural crest formation. In particular, neural crest specification was inhibited in X-TSK-depleted embryos, which also showed an expansion of the epidermal ectoderm and, depending on the approach, a reduction of the neural plate (Fig. 4)(Ohta et al., 2004). In these experiments, the observation that the neural plate was reduced in embryos injected with a X-TSK-targeted morpholino(Ohta et al., 2004), but not siRNA, may be explained with the fact that siRNA did not apparently affect X-TSK levels during gastrulation, when X-TSK is expressed in the dorsal ectoderm and mesoderm, while the morpholino may be already effective at these stages (Ohta et al.,2004). By contrast, X-TSK overexpression after gastrulation caused an expansion of the neural crest(Fig. 6). In addition, though X-TSK on its own was not able to induce neural crest markers in animal caps, it could strongly cooperate in this process with the dorsolateral mesoderm or XWnt-8, both of which provide a posteriorizing signal required for neural crest specification(Fig. 5)(Villanueva et al., 2002). These experiments also showed that, consistent with it expression pattern at the neural plate border (Fig. 1), X-TSK function is required in the ectoderm for neural crest specification.

How does X-TSK control neural crest specification? Clearly, its BMP antagonistic activity is likely to play a role. In fact, other BMP inhibitors,such as Chordin or Noggin, can mimic X-TSK ability to induce neural crest markers in cooperation with XWnt-8(LaBonne and Bronner-Fraser,1998; Mayor et al.,1995). In addition, both gain- and loss-of-function analysis suggest that X-TSK has a more general role in the mediolateral patterning of the ectoderm, which is best explained with its anti-BMP function. A specific gradient of BMP signaling is well-known to be required to specify the neural plate, neural crest and epidermal domains in the Xenopus ectoderm(Marchant et al., 1998). Therefore, one possibility is that the BMP-inhibitory activity of X-TSK,localized at the neural plate border and in the non-neural ectoderm, is essential for the proper shaping of the BMP gradient required for ectodermal patterning, and that the action of BMP antagonists secreted from the dorsal midline, such as Chordin and Noggin, is not sufficient in this respect. This would be in agreement with previous observations that removal of the dorsal marginal zone does not prevent neural crest specification, though it strongly affects neural plate induction (Marchant et al., 1998).

The fact that another BMP antagonist, namely Chordin, could not efficiently rescue the neural crest reduction in X-TSK-depleted embryos(Fig. 6I,J) suggested the interesting possibility that X-TSK may also control neural crest specification via alternative mechanisms distinct from direct BMP antagonism through protein-protein interaction. First, different from other BMP antagonists,X-TSK can upregulate BMP4 expression in the non-neural ectoderm,while BMP signaling can positively regulate X-TSK expression. In addition, as described above, we found clear indications that X-TSK may also work in neural crest specification via modulation of the Notch signaling pathway. Remarkably, it has been already shown that Notch signaling plays a role in neural crest specification and that it can regulate expression of BMP4 and genes associated with BMP signaling in the presumptive neural crest region (Glavic et al.,2004). Moreover, during Xenopus development, it has been described that Notch and its target gene Hairy2A are expressed in the neural crest territory, while the Notch ligands, Delta and Serrate, are expressed in the cells surrounding the prospective crest cells(Glavic et al., 2004). An attractive hypothesis is that, by directly interacting with both BMP4 and X-delta-1 in the extracellular space at the neural/epidermal border, and by indirectly regulating BMP4 transcription in this region, X-TSK may work as a crucial molecular intersection between BMP and Notch signaling in the territory where the neural crest is specified.

Our results also suggest that X-TSK may differentially modulate X-delta-1 activity in a dose-dependent manner, by activating it at lower X-TSK levels and repressing it at higher X-TSK levels. On this respect, it is interesting to note that the distribution of TSK expression levels shows some correlation with the localization of the neural crest. For example, the neural crest is specified at the lateral neural border, where X-TSK expression is lower, and excluded from the anterior neural border, where X-TSK expression is higher(Fig. 1A,C). Moreover, the proximal edge of the neural crest in the lateral neural border overlaps with a region of higher X-TSK expression, while X-TSK expression inside the neural crest domain is lower (Fig. 1K,L; Fig. 4X). As Notch is expressed in the whole neural plate, while X-delta-1 is expressed both at the anterior neural boundary and at both the proximal and the distal edges of the neural crest in the lateral neural boundary(Glavic et al., 2004), the expression pattern of Notch and Delta can only partially account for the specific localization of the neural crest region. Furthermore, although caudalizing signals such as Wnt, FGF and RA are thought to be responsible for the localized induction of the neural crest at the lateral, but not the anterior, neural border, some FGFs and RA-producing enzymes, such as FGF8 and Raldh2, are expressed in the anterior neural border from early neurula stages(Lupo et al., 2005). Therefore, it is tempting to speculate that the presence of higher levels of X-TSK in the anterior neural border and in the proximal edge of the neural crest in the lateral neural border, compared with the neural crest itself,might be an additional mechanism to restrict the boundaries of the neural crest domain

Previously, two-signal models of neural crest specification were proposed(LaBonne and Bronner-Fraser,1998; Villanueva et al.,2002), which suggested the requirement of intermediate levels of BMP signaling and posteriorization factors. Based on the results presented in this paper, we propose a model of consecutive steps in neural crest specification that contemplates the dynamic expression pattern of X-TSK and sequential molecular interaction during ectodermal development.

First, during gastrulation, X-TSK is expressed in the whole animal ectoderm and in the dorsal mesoderm (Ohta et al., 2004). Dorsal midline signals, such as Chordin,Noggin and X-TSK, bind to BMPs and directly inhibit their activity in the dorsal ectoderm (Sasai et al., 1994) (Fig. 8A, phase I), leading to the specification of the neural plate at these stages (reviewed by Sasai and De Robertis, 1997).

Subsequently, X-TSK expression is downregulated in the presumptive neural plate, while it is maintained and enhanced in the ectoderm flanking the neural plate (phase II). At these stages, in the presumptive neural crest region, X-TSK controls BMP activity through direct binding to BMPs and it also modulates BMP transcription, possibly via modulation of the Notch signaling pathway (Fig. 8B). Acting via simultaneous modulation of BMP and Notch signaling, TSK regulates the establishment of a pre-neural crest region expressing Hairy2A and Msx1. Finally, posteriorizing signals such as XWnt-8converge on this molecular network downstream of Msx1, to control the activation of Pax3 expression and the specification of the neural crest within the Msx1-positive region(Monsoro-Burq et al., 2005)(Fig. 8B).

We thank Dr Takeshima (Nagoya University) for kindly suggesting the method of the electroporation, and thank all members of our laboratories for help. This work was supported in part by PRESTO of the Japan Science and Technology Corporation (K.O.), by a Grant-in-Aid from the Ministry of Education, Science,Sports, and Culture of Japan to H.T. and K.O, and by a Grant-in-Aid for 21st Century COE Research from Ministry of Education, Culture, Sports, Science and Technology `Cell Fate Regulation Research and Education Unit' to S.K. and H.T. G.L. was supported by an EMBO long-term post-doctoral fellowship. S.O. was supported by the Cancer Research UK and the Royal Society. W.H. was supported by the Wellcome Trust and the EC.