The novel Smad-interacting protein Smicl regulates Chordinexpression in the Xenopus embryo

Great progress has been made in coming to understand the series of inductive interactions that generates the body plan of the early amphibian embryo (Chan and Etkin, 2001; De Robertis and Kuroda, 2004; Heasman, 1997; Weaver and Kimelman, 2004). Briefly, fertilisation causes rotation of the cortical cytoplasm, which in turn brings about the Wnt11-dependent stabilisation of β-catenin on the dorsal side of the embryo at the blastula stage(Tao et al., 2005). This nuclear β-catenin is involved in the induction of two signalling centres in the blastula. The first, in dorsal ectoderm, is the recently identified BCNE (blastula Chordin- and Noggin-expressing) centre that is required for proper development of the nervous system. The second is the Nieuwkoop centre, which is formed in dorsal vegetal cells, where the domain of nuclear β-catenin overlaps the vegetally localised maternal mRNAs Vg1 and VegT (De Robertis and Kuroda,2004).

The Nieuwkoop centre induces the formation of Spemann's organiser in overlying equatorial cells and, through the action of Nodal-related proteins such as Xnr1, Xnr2, Xnr4, Xnr5 and Xnr6, activates the expression of genes such as Noggin and Chordin, which encode secreted inhibitors of BMP signalling (Agius et al.,2000; De Robertis and Kuroda,2004). The mechanism by which the Nodal-related proteins induce these genes is poorly understood, although it is known that Chordinis an indirect target of the Nodal-related proteins and of Activin, because its activation is inhibited by the protein synthesis inhibitor cycloheximide(Howell and Hill, 1997; Sasai et al., 1994).

In this paper, we provide new insight into the regulation of Chordin through our analysis of the novel Smad-interacting protein Smicl (Collart et al., 2005). Receptors of TGFβ family members such as the Nodal-related proteins and Activin signal by phosphorylating, and thereby activating Smad2 and Smad3(Miyazawa et al., 2002). Once activated, these Smad proteins bind Smad4 and translocate to the nucleus where they regulate gene expression. This is achieved through direct interaction with DNA or by interaction with other transcriptional regulators such as Fast1(Massague and Wotton, 2000). We show here that Smicl is expressed maternally in the Xenopus embryo and is required for the expression of Chordin, but not of Goosecoid or Xnr3, in Spemann's organiser. Significantly, the phenotype of embryos lacking Smicl resembles that of embryos in which Chordin is depleted. Smicl interacts specifically with Smad3 and is involved in the second step of an indirect pathway through which the Nodal-related proteins activate Chordin. In the first step,Smad3 activates the expression of Xlim. In the second a complex containing Smicl, Smad3 and the newly induced Xlim1 activates expression of Chordin in a direct manner. Our work defines the role of Smicl in the early Xenopus embryo and contributes new findings to the hitherto poorly understood regulation of Chordin.

Xenopus embryos were obtained by artificial fertilisation. X. laevis embryos were maintained in 10% normal amphibian medium (NAM)(Slack, 1984) and X. tropicalis embryos in 1% MMR. They were staged according to Nieuwkoop and Faber (Nieuwkoop and Faber,1975). Embryos at the one-cell stage were injected with RNA or with antisense morpholino oligonucleotides dissolved in water. For animal cap assays, embryos were dissected at stage 8, before the midblastula transition(MBT), and cultured in 75% NAM. Cycloheximide was dissolved in ethanol to a concentration of 10 mM and then diluted to a final concentration of 10 μM in 75% NAM containing 0.1% bovine serum albumin. Animal caps were frozen when sibling embryos reached stage 11.

The efficacy of antisense morpholino oligonucleotide XtMO1, directed against Xenopus tropicalis Smicl, was tested by injecting embryos with XtMO1 or the control oligonucleotide coMO followed by RNA encoding an HA-tagged form of XtSmicl. Embryos were allowed to develop to early gastrula stage 10 and were then subjected to SDS polyacrylamide gel electrophoresis and western blotting, using rat monoclonal anti-HA antibody 3F10 (Roche Diagnostics) to detect the HA epitope and a mouse monoclonal anti-GAPDH antibody (HyTest Ltd) as a loading control.

In situ hybridisation was carried out essentially as described previously(Harland, 1991), except that BMP purple was used as a substrate. A Chordin probe was as described(Sasai et al., 1994) and expression of XtSmicl was detected by transcription of a Smicl cDNA in the vector pCS107 derived from the Xenopus tropicalis EST database(http://www.sanger.ac.uk/Projects/X_tropicalis/;GenBank Accession Number AL675957). The plasmid was linearised with EcoRI and transcribed with T3 RNA polymerase.

Total RNA was prepared from five pooled X. laevis embryos, 30 X. tropicalis embryos or 10 X. laevis animal caps using the TriPure reagent (Roche), followed by DNAseI digestion, proteinase K treatment,phenol/chloroform extraction and ethanol precipitation. RNA was dissolved in water and used as a template for real-time RT-PCR.

Real-time RT-PCR with the LightCycler (Roche) was carried out using the manufacturer's RNA amplification kit. All determinations included a negative control and a serial dilution of embryo RNA was used to create a standard curve. Primers specific for Xbra, Goosecoid, Chordin and Ornithine decarboxylase (ODC) were as described previously(Piepenburg et al., 2004), Sox17 and Xlim1 were as described previously(Xanthos et al., 2001), Xnr3 was as described previously(Kofron et al., 1999) and Siamois was as described previously(Heasman et al., 2000). XtSmicl primers were 5′-AGCGCAGTCTGGCCATCATC-3′ and 5′-TCGGGAGACATAGACGTGGC-3′. All values were normalised to the level of ODC in each sample.

A mouse Smicl cDNA comprising the entire open reading frame except for the first six amino acids was cloned between the EcoRI and XbaI sites of pCS3, thereby introducing six N-terminal Myc tags. The construct was linearised with Asp718 and sense RNA was transcribed with SP6 RNA polymerase. An X. tropicalis cDNA comprising the entire XtSmicl open reading frame and 45 bp of 5′ UTR (GenBank Accession Number AY887083)was provided with a C-terminal HA tag by PCR and cloned between the EcoRI and XbaI sites of pCS2. Sense RNA was produced using SP6 RNA polymerase after linearisation of the plasmid with Asp718. A Myc-tagged Smad2 construct, cloned in pFTX5(Howell and Hill, 1997), was linearised with XbaI and sense RNA was transcribed with T7 RNA polymerase. The open reading frame of Chordin, cloned in pSP35T (the gift of E. De Robertis), was linearised with XbaI and transcribed with SP6 polymerase. The open reading frame of Xnr1 cloned in pCS2(Williams et al., 2004) was linearised with Asp718, and sense RNA was transcribed with SP6 RNA polymerase. A Smad3 construct in pCS2 was linearised with Asp718, and sense RNA was transcribed with SP6 RNA polymerase. Flag-Xlim1/3m, cloned in pCS2 (Yamamoto et al., 2003),was linearised with NotI and sense RNA was transcribed with SP6 RNA polymerase. A cDNA encoding Siamois, cloned in pBluescript RN3(Lemaire et al., 1995), was linearised with SfiI and sense RNA was transcribed with T3 RNA polymerase. A constitutively active β-catenin construct cloned in pSP64T(Domingos et al., 2001), was linearised with SfiI and sense RNA was synthesised using SP6 RNA polymerase. Myc-tagged Smad constructs were a gift from Dr K. Miyazono. Constitutively active forms of ALK6 and ALK4 were as described(Armes and Smith, 1997).

HEK293T cells were grown in Dulbecco's modified Eagle's medium containing 10% foetal bovine serum (FBS) supplemented with 4.5 g/l glucose. Cells were grown to 50% confluence in 9 cm dishes and transfected using Fugene (Roche Molecular Biochemicals) according to the manufacturer's protocol.

Transiently transfected HEK293T cells were frozen in liquid nitrogen,thawed on ice and solubilised in lysis buffer containing 1% NP40, 150 mM NaCl,20 mM Tris pH 7.5, 2 mM EDTA, 50 mM NaF, 1 mM sodium pyrophosphate,supplemented with protease inhibitors (Roche Molecular Biochemicals). Cell lysates were cleared by centrifugation, and precipitations were performed by overnight incubations with beads coupled to mouse monoclonal anti HA (Roche),mouse monoclonal M2Flag or mouse monoclonal 9E10 anti Myc (Santa Cruz). Unbound proteins were removed by washing four times with lysis buffer and once with phosphate-buffered saline at 4°C. Bound proteins were harvested by boiling in sample buffer, and they were resolved by SDS-polyacrylamide gel electrophoresis. Myc-tagged, Flag-tagged, HA-tagged proteins and endogenous Smad3 were visualised after western blotting using mouse monoclonal 9E10 anti-Myc (a gift from Innogenetics), anti-M2Flag (Santa Cruz), anti-HA (Roche)and rabbit polyclonal anti-Smad3 (Abcam) antibodies, in combination with horseradish peroxidase-conjugated anti-mouse and anti-rabbit secondary antibodies (Jackson), and the enhanced chemiluminescence kit (New England Nuclear).

DNA precipitations using biotinylated double-stranded oligonucleotides corresponding to base pairs -696 to -621 relative to the translation start site of X. tropicalis Chordin(http://genome.jgi-psf.org/Xentr3/Xentr3.home.html)were carried out as described (Hata et al., 2000). The sequence of the wild-type oligonucleotide was 5′CCATACTGATTATTCCCCAAATCTTGTCAAATTCTATGTAGCTTTCCCACATGCAATTATCTGCATGTCCCCCACT 3′. The sequence of a mutated oligonucleotide was 5′CCATACCTTTTATTCCCCAAATCTTGTCAAATTCTATGTAGCTTTCCCACATGACCAAGTCTGCATGTCCCCCACT 3′. Wild type and mutated Xlim1 binding sites(Mochizuki et al., 2000) are indicated in italics. DNA-bound proteins were collected with streptavidine-agarose beads (Sigma) and analyzed by western blotting.

Mouse Smicl (Smad-interacting CPSF-like protein) was identified as a candidate Smad-interacting protein in the yeast two-hybrid screen that also identified SIP1 (Collart et al.,2005; Verschueren et al.,1999). A cDNA encoding its putative Xenopus tropicalishomologue (XtSmicl) was identified by BLAST searching the Xenopus tropicalis EST database(http://www.sanger.ac.uk/Projects/X_tropicalis/),and a full-length version was cloned by reverse transcription PCR using cDNA derived from blastula stage embryos. The deduced amino acid sequence (GenBank Accession Number AY887083), comprising 827 residues, displays 43.1% identity and 44.7% similarity to the mouse sequence (GenBank Accession Number AJ516034), with a domain containing five C3H type zinc fingers being particularly highly conserved.

We first carried out co-immunoprecipitation experiments in HEK293T cells to determine whether XtSmicl, like mouse Smicl, is a Smad-interacting protein. XtSmicl was co-expressed with Smad1 or Smad5 (which act downstream of BMP family members), Smad2 or Smad3 (which act downstream of TGFβ, Activin and Nodal family members), or the common mediator Smad4, in the presence or absence of their cognate constitutively active (ca) receptors. Smad proteins were immunoprecipitated from cell extracts with anti-Myc antibody and the presence of HA-tagged XtSmicl in the immunoprecipitate was detected by western blotting. XtSmicl proved to interact weakly with Smad2(Fig. 1B, lanes 3 and 4, either co-transfected with caALK4 or not) and strongly with Smad3 and Smad4 when co-transfected with a constitutively active ALK4 receptor(Fig. 1B, lanes 6 and 11). A strong interaction between overexpressed HA-XtSmicl and endogenous Smad3 in presence of caALK4 could be detected after immunoprecipitation of HA-XtSmicl with anti-HA antibody and analysis of the immunoprecipitate after western blotting with anti-Smad3 antibody (Fig. 1C, lane 2).

The expression pattern of XtSmicl during early development was assayed by quantitative RT-PCR and in situ hybridisation. Transcripts are abundant in the animal hemisphere of the fertilised egg(Fig. 1D, part a, Fig. 1E) and decline thereafter, although XtSmicl RNA is still detectable at early and mid gastrula stages in involuting dorsal mesoderm(Fig. 1D, parts c,d). By the neurula stage, expression is most abundant in neurectoderm and migrating neural crest cells (Fig. 1D,parts e-h,j), and a section through an embryo at stage 23 shows expression in the notochord, somites and neural tube(Fig. 1D, part i).

To analyse the function of Smicl during early Xenopus tropicalisdevelopment, we designed two antisense morpholino oligonucleotides: one directed against the start codon of XtSmicl (XtMO1) and the other against a region within the 5′ untranslated region of the mRNA (XtMO2)(Fig. 2A). We also designed a control oligonucleotide (coMO), which differs by eight bases from the sequence of XtMO2 (Fig. 2A). A morpholino oligonucleotide targeted against the start codon of the X. laevis homologue of Smicl (XlMO) was based on the sequence of an X. laevis Smicl EST. Ten bases of XlMO differ from the equivalent X. tropicalis sequence (Fig. 2A). The specific morpholino oligonucleotide XtMO1 inhibited translation in a dose-dependent manner of an HA-tagged form of XtSmicl after injection of XtMO1 into Xenopus embryos followed by injection of HA-XtSmicl mRNA (Fig. 2B).

Injection of the three Smicl antisense morpholino oligonucleotides causes similar phenotypes in X. tropicalis and X. laevis. The first observed effect is a delay in the onset of gastrulation(Fig. 2C, parts b,d) and by neurula stages this delay manifests itself as a failure of the blastopore to close (Fig. 2C, parts f,h). At tadpole stages dorsoanterior structures are reduced and ventroposterior structures somewhat expanded; the anteroposterior axis is shortened; and embryos are microcephalic (Fig. 2C, parts j,l). Injection of coMO causes no detectable defects in development.

We note that injection of XtMO1, but not of coMO, causes the upregulation of Smicl mRNA at stage 10.5 (Fig. 2E). This elevated transcription may reflect an attempt by the embryo to regulate levels of Smicl protein following inhibition of translation by the antisense morpholino oligonucleotide, or it may be due to stabilisation of the RNA.

The observation that three different morpholino oligonucleotides yield similar phenotypes in two species of Xenopus argues that the effects of these reagents are specific. To confirm this impression, we carried out rescue experiments in Xenopus laevis using a mouse Smicl construct that lacks the first six amino acids and contains a Myc tag (see Materials and methods), so that its translation is not inhibited by the antisense oligonucleotide. This mRNA caused significant rescue of the phenotype caused by XlMO (Fig. 2D, part d), and indeed XlMO was able to rescue the spina bifida phenotype that is caused by mis-expression of the mouse Smicl construct(Fig. 2D, part c). Together,these experiments indicate that our Smicl antisense morpholino oligonucleotides function in a specific manner and that the mouse and Xenopus proteins are functional homologues.

One way in which inhibition of Smicl function might disrupt gastrulation is by interfering with gene expression in the organiser, and indeed the Smicl phenotype resembles quite closely that of the organiser-specific gene Chordin, obtained by targeting both Xenopus laevis Chordin pseudo-alleles with antisense morpholino oligonucleotides(Oelgeschlager et al., 2003). To address this point, we studied expression levels of the pan mesodermal marker Xbra (Fig. 3A),the endodermal marker Sox17 (Fig. 3B) and the organiser-specific genes Xnr3(Fig. 3C), Goosecoid(Fig. 3D) and Chordin(Fig. 3E-G). The only one of these genes to be affected by inhibition of Smicl function, in X. laevis and in X. tropicalis, was Chordin. This was confirmed by in situ hybridisation, which showed that inhibition of Smicl function both reduces the expression level of Chordin and decreases the size of its expression domain (Fig. 3H), while the expression pattern of the other organiser markers,also analyzed by in situ hybridisation, is normal (data not shown). As an additional control, we observed that the X. tropicalisoligonucleotide XtMO1, which differs by ten bases from XlMO, did not decrease expression of Chordin in X. laevis (data not shown), and the downregulation of Chordin caused by XlMO was rescued by co-injection of mRNA encoding mouse Smicl (Fig. 3H).

To ask whether the Smicl loss-of-function phenotype is caused in part by the downregulation of Chordin, we attempted to rescue the effects of the Xenopus laevis Smicl antisense oligonucleotide by co-injection of RNA encoding Chordin. This mRNA brought about partial rescue of the anterior structures of the embryos (Fig. 3J, part d), which are significantly reduced in embryos injected with XlMO (Fig. 3J, part c).

Together, these experiments indicate that Smicl is required for expression of Chordin in the Xenopus organiser, and that the phenotype caused by loss of Smicl function is due in part to the downregulation of Chordin. We therefore went on to investigate the role of this Smad-interacting protein in the regulation of Chordin in more detail.

Previous work indicates that the expression of Chordin in the organiser of Xenopus is initiated by β-catenin signalling and that its maintenance depends on high levels of Nodal related proteins such as Xnr1 derived from the Nieuwkoop centre(Wessely et al., 2001). Consistent with this idea, activation of Chordin in isolated animal pole regions by members of the TGFβ family is inhibited by cycloheximide(Howell and Hill, 1997; Sasai et al., 1994),suggesting that induction requires the synthesis of intermediate proteins and is therefore indirect.

To investigate the activation of Chordin in more detail, we first asked whether its activation by β-catenin is direct or indirect. RNA encoding Xnr1 or β-catenin was injected into Xenopus embryos at the one-cell stage, and animal pole regions were dissected before the mid-blastula transition (that is, before the onset of zygotic transcription)and incubated in the presence or absence of cycloheximide until the equivalent of the early gastrula stage. Xnr1 and β-catenin both activate expression of Chordin in animal caps, but induction by β-catenin, like induction by Xnr1, is inhibited by cycloheximide and is therefore indirect(Fig. 4A,B).

It is possible that the indirect induction of Chordin byβ-catenin occurs through Siamois(Lemaire et al., 1995), a transcription factor that is expressed in the organiser in response toβ-catenin and that can activate transcription of Chordin(Wessely et al., 2004). Further experiments demonstrated that Siamois activates Chordin in a direct manner (Fig. 5A). To examine the possibility that Smicl is involved in this process, we asked whether Smicl antisense morpholino oligonucleotides prevent induction of Chordin by Siamois in animal caps. This proved not to be the case(Fig. 5C). Moreover, inhibition of Smicl function does not inhibit Siamois expression in intact embryos (Fig. 5B). We conclude that Smicl is not involved in the induction of Chordin through theβ-catenin/Siamois pathway, although we cannot exclude the possibility that β-catenin induces Chordin via other genes.

The inductive effects of Nodal-related signalling are mediated by Smad2 and Smad3, which, on receptor activation, associate with a co-Smad and accumulate in the nucleus where they are recruited to particular promoters by specific transcription factors (Hill,2001). Preliminary experiments revealed that both Smad2 and Smad3 are able to induce expression of Chordin in isolated Xenopus laevis animal caps (Fig. 6A). Smad3 proved to be a more powerful inducer of Chordin than did Smad2, and further experiments using Smad3 revealed that induction of Chordin by both Xnr1 and by Smad3 requires Smicl(Fig. 6B).

To ask whether a Smicl/Smad3 complex might activate Chordindirectly, animal pole regions were dissected from embryos expressing exogenous Smad3 and in which endogenous Smicl is also present. The animal caps were incubated in the presence or absence of cycloheximide, and assayed for expression of Chordin at the early gastrula stage. Cycloheximide proved to inhibit activation of Chordin(Fig. 6A), indicating that Smad3 acts indirectly, presumably through the induction of another gene `X'. This gene is unlikely to be Smicl, because neither Xnr1 nor Smad3 increases expression of Smicl in animal caps or in intact embryos(data not shown).

The experiments described above indicate that Smicl regulates the expression of Chordin by functioning in concert with a factor X that is produced in response to signalling by Xnr1(Fig. 6C). One candidate for X is Xlim1, which is involved in the induction of organiser-specific genes such as Goosecoid and Cerberus(Mochizuki et al., 2000; Yamamoto et al., 2003), as well as Chordin (Taira et al.,1994). Moreover, expression of Xlim1 is induced in isolated animal pole regions by members of the TGFβ family such as Activin (Taira et al.,1992).

To investigate whether Xlim1 is involved in the Xnr1/Smad3 signalling cascade that leads to induction of Chordin, we first tested the abilities of Smad2 and Smad3 to activate Xlim1 in isolated animal pole regions. Smad3 proved to induce strong expression of Xlim1 in a direct manner; induction by Smad2 was weaker and indirect(Fig. 7A).

We next asked whether Xlim1, as would be expected of factor X, can induce expression of Chordin in isolated animal pole regions. These experiments made use of Xlim1/3m, a constitutively active variant of Xlim1 in which two inhibitory Lim domains are inactivated(Taira et al., 1994). As previously reported (Taira et al.,1994), expression of Xlim1/3m does activate Chordin in isolated animal pole regions, and this induction proved to be direct(Fig. 7B). Depletion of Xlim in Xenopus embryos does not cause downregulation of Chordin or other organiser-specific genes at very early gastrula stages, but it remains possible that Xlim1 plays a role in the maintenance of their expression (Hukriede et al., 2003).

Together, these experiments indicate that Xlim1 can be induced directly by Smad3 and that Xlim1 in turn can activate Chordin in a direct fashion. Smicl is not involved in the first of these steps, because inhibition of Smicl function by injection of XlMO does not affect expression levels of Xlim1 in Xenopus laevis(Fig. 7C). However, use of the same antisense morpholino oligonucleotide shows that Smicl is required for activation of Chordin by Xlim1/3m(Fig. 7D). Together, these experiments indicate that the factor X that is required downstream of Xnr1 and Smad3 is Xlim1, and that induction of Chordin by Xlim1 requires the Smad-interacting protein Smicl.

The requirement for Smicl in the induction of Chordin by Xlim1 suggests that the two proteins might physically interact. This possibility was tested by co-immunoprecipitation experiments showing that XtSmicl associates with Xlim1/3m following expression of the two proteins in HEK293T cells(Fig. 7E, lane 5). Xlim1 does not interact directly with Smad3 (Fig. 7E, lane 4), but the Smad-interacting protein Smicl can recruit Smad3 to create a complex containing these two proteins and Xlim1(Fig. 7E, lane 1).

Our results predict that Xlim1 should bind to the Chordinpromoter. Previous work has demonstrated that Xlim1 can bind, in either orientation, the sequence TAATXY, where XY is TA, TG, CA or GG(Mochizuki et al., 2000). Inspection of the Xenopus tropicalis Chordin promoter region revealed two putative Xlim1-binding sites positioned at nucleotides -638 to -643 and-685 to -690 relative to the translation start site (see Materials and methods). To confirm that these sites can indeed bind Xlim1, we carried out a DNA-mediated pull-down assay (Hata et al.,2000). Flag-tagged Xlim1/3m derived from transfected HEK293T cells was efficiently precipitated by biotinylated Chordin promoter oligonucleotides containing the two Xlim1-binding sites(Fig. 7F, lane 3) and further experiments showed that HA-tagged Smicl could be co-immunoprecipitated with Xlim1 (Fig. 7F, lane 5). Mutation of the Xlim1 binding sites abolished binding(Fig. 7F, lane 8).

In this paper, we study the function of Smicl, a novel Smad-interacting zinc-finger protein that is expressed maternally in the Xenopusembryo. Inhibition of Smicl function by injection of antisense morpholino oligonucleotides causes the loss or reduction of anterior and dorsal structures, and the shortening of the anteroposterior axis. This phenotype is preceded, and is partially caused by, a reduction in Chordinexpression at the early gastrula stage: we note that the phenotype of embryos in which Smicl function is inhibited resembles that of embryos injected with antisense morpholino oligonucleotides directed against Chordin(Oelgeschlager et al., 2003),and that injection of Chordin mRNA rescues, in part, the effects of loss of Smicl function (Fig. 3J).

Our results are of interest for three reasons. First, they define the function of a novel Xenopus Smad-interacting protein, and in doing so they reveal a surprising degree of specificity in this protein; although we have not examined a large panel of markers, inhibition of Smicl function does not downregulate all genes expressed in the organiser, just Chordin. In this regard, we note that expression of Chordin in the embryo is not completely inhibited by antisense morpholino oligonucleotides directed against Smicl. There are four possible explanations for this observation. First, the concentration of the Smicl antisense oligonucleotides used in these experiments may be too low to elicit the most extreme phenotype. Second,maternal protein may persist long enough to provide some `rescue' of the effects of inhibiting de novo translation. Third, Smicl antisense morpholino oligonucleotides may not inhibit the initial activation of Chordinexpression, but may prevent its maintenance. And finally, other signalling pathways may be involved in Chordin regulation, although it is not clear, at present, whether FGF signalling plays a role(Delaune et al., 2005; Mitchell and Sheets,2001).

A second point of interest concerns the regulation of Chordin,which has long been recognised as an indirect target of TGFβ signalling,to the extent that it is sometimes used as a control for the efficacy of cycloheximide treatment (Howell and Hill,1997). Our work defines the steps involved in this indirect activation. And finally, our results are of note because they define differential activities for Smad2 and Smad3 in the early Xenopusembryo. This point and the other issues mentioned above are discussed below.

Smicl was identified in a yeast two-hybrid screen designed to identify Smad-interacting proteins (Collart et al.,2005; Verschueren et al.,1999). The C terminus of the protein contains five CCCH-type zinc fingers that display homology to a domain in CPSF30, the 30 kDa subunit of cleavage and polyadenylation specificity factor (CPSF). CPSF is required for the cleavage and polyadenylation of pre-mRNA and, like CPSF, the zinc-finger domain of Smicl can bind single-stranded DNA as well as cleave RNA in vitro(Collart et al., 2005).

To investigate the role of Smicl during early Xenopus development,we injected specific antisense morpholino oligonucleotides into embryos of Xenopus laevis and Xenopus tropicalis. Such embryos develop with small heads, reduced dorsal tissues, increased ventral and posterior structures, and shortened trunks. Interestingly, this phenotype resembles that of Xenopus and zebrafish embryos in which Chordin function is inhibited or absent (Leung et al.,2005; Oelgeschlager et al.,2003; Schulte-Merker et al.,1997), and indeed of the genes we investigated only Chordin proved to be affected by the inhibition of Smicl function(Fig. 3). Consistent with this observation, the phenotype of embryos lacking Smicl can be rescued quite significantly by injection of RNA encoding Chordin(Fig. 3J), although the fact that rescue is not complete suggests that there are other Smicl target genes yet to be identified. Some such genes have been identified in a preliminary microarray analysis, but none of these has yet proved to be organiser-specific(C.C., J. Ramis and J.C.S., unpublished).

The phenotype of embryos lacking Smicl function differs from that of zebrafish embryos lacking no arches, the zebrafish homologue of CPSF30(Gaiano et al., 1996). Such embryos, as their name implies, lack pharyngeal arches and eyes. Preliminary experiments using an antisense morpholino directed against Xenopus tropicalis CPSF30 reveal a more severe phenotype in which epidermal cells no longer adhere to the underlying mesodermal tissue (A.R., C.C. and J.C.S.,unpublished). We do not yet know why the two phenotypes should differ; perhaps there is another CPSF30 in the zebrafish genome.

Chordin is expressed in the organiser of the Xenopusembryo. It encodes a secreted factor that binds to, and inhibits the function of, BMP family members such as BMP4, and thereby functions as an important mediator of the inducing and patterning activities of the organiser(Sasai et al., 1995; Sasai et al., 1994). Previous work has demonstrated that Chordin is an indirect target of TGFβsignalling and more recent experiments suggest that its expression is initiated by β-catenin-mediated transcriptional activation via Siamois and maintained by Nodal-related signalling pathways(Wessely et al., 2001).

Our work reveals that Smicl is involved in the Xnr-mediated maintenance of Chordin expression, and that it acts in the second component of a two-step pathway. In the first, TGFβ signalling activates the expression of Xlim1 directly via Smad3. In the second, our results suggest that Xlim1 cooperates with Smicl, perhaps in a ternary complex with Smad3, to activate Chordin directly (Figs 7, 8). Consistent with this suggestion, we note that Chordin(Sasai et al., 1994), Smicl (Fig. 1C), Smad3 (Howell et al.,2001) and Xlim1(Taira et al., 1992) are all expressed in dorsal mesoderm at early gastrulation stages.

The activity of Xlim1 is regulated by Ldb1(Breen et al., 1998; Jurata et al., 1998), which is believed to counteract the effects of an inhibitory protein and thereby cause Xlim1 to shift to an activated state in which it can bind cell specific transcriptional co-activators (Hiratani et al., 2001). Our data suggest that Smicl is such a co-activator. Indeed, the ability of Xnr1 to induce expression of Goosecoid in Xenopus animal caps is abolished by inhibition of Smicl function(data not shown), although the fact that loss of Smicl activity does not inhibit expression of Goosecoid in the dorsal mesoderm of intact embryos (Fig. 3C) suggests that another factor can substitute for Smicl in this region of the embryo.

Our observations suggest that the closely related proteins Smad2 and Smad3 play different roles in the early Xenopus embryo and that these roles differ not only because Smad2 is expressed at much higher levels than Smad3(Howell et al., 2001). In particular, we note that Smad3 activates expression of Xlim1 directly, while Smad2 induces Xlim1 in an indirect fashion, in the sense that activation is inhibited by cycloheximide. Previous work has demonstrated that induction of Xlim1 by Activin and Nodal-related signalling requires the Smad-interacting protein Fast1. This transcription factor, together with receptor activated Smad proteins, acts as a direct transcriptional inducer of Xlim1 through a cluster of Fast1/Smad4 sites located in the first intron of the gene (Watanabe et al.,2002). Fast1 contains two Smad-binding domains, the Smad interaction motif (SIM) and the Fast/FoxH1 motif (FM)(Randall et al., 2004). While the SIM can bind both Smad2 and Smad3, the FM binding site is highly specific for Smad2. This observation, together with our own data, suggests that the Smad interaction motif of Fast1 but not the Fast/FoxH1 motif is required for direct transcriptional activation of Xlim1. The existence of these distinct Smad-binding motifs in Fast1 might provide the molecular basis for the differential activities of Smad2 and Smad3 in the induction of Xlim1 transcription.

This work was funded by a Wellcome Trust Programme Grant awarded to J.C.S.,by grants from the Fund of Scientific Research-Flanders (G.0105.02) and the Research Council of the University of Leuven (OT/00/41), and by funding through the Interuniversity Attraction Pole Networks (IUAP5/35) to D.H.