δEF1 and SIP1 are differentially expressed and have overlapping activities during Xenopus embryogenesis

Authors

  • Leo A. van Grunsven,

    1. Department of Developmental Biology (VIB7), Flanders Interuniversity Institute for Biotechnology (VIB) and Laboratory of Molecular Biology (Celgen), University of Leuven, Campus Gasthuisberg, Leuven, Belgium
    Search for more papers by this author
    • Drs. van Grunsven and Taelman contributed equally to this work.

  • Vincent Taelman,

    1. Laboratoire d′Embryologie Moléculaire, Université Libre de Bruxelles, Institut de Biologie et de Médecine Moléculaires (IBMM), Gosselies, Belgium
    Search for more papers by this author
    • Drs. van Grunsven and Taelman contributed equally to this work.

  • Christine Michiels,

    1. Department of Developmental Biology (VIB7), Flanders Interuniversity Institute for Biotechnology (VIB) and Laboratory of Molecular Biology (Celgen), University of Leuven, Campus Gasthuisberg, Leuven, Belgium
    Search for more papers by this author
  • Karin Opdecamp,

    1. Laboratoire d′Embryologie Moléculaire, Université Libre de Bruxelles, Institut de Biologie et de Médecine Moléculaires (IBMM), Gosselies, Belgium
    Search for more papers by this author
  • Danny Huylebroeck,

    Corresponding author
    1. Department of Developmental Biology (VIB7), Flanders Interuniversity Institute for Biotechnology (VIB) and Laboratory of Molecular Biology (Celgen), University of Leuven, Campus Gasthuisberg, Leuven, Belgium
    • Department of Developmental Biology (VIB7), University of Leuven, Campus Gasthuisberg (Bldg. Ond&Nav, box 812), Herestraat 49, B-3000 Leuven, Belgium
    Search for more papers by this author
  • Eric J. Bellefroid

    1. Laboratoire d′Embryologie Moléculaire, Université Libre de Bruxelles, Institut de Biologie et de Médecine Moléculaires (IBMM), Gosselies, Belgium
    Search for more papers by this author

Abstract

The zinc finger/homeodomain transcription factor (zfhx1) family in vertebrates consists of two members, δEF1 and SIP1. They have been proposed to display antagonistic activities in the interpretation of Smad-dependent TGFβ signaling during mesoderm formation. We cloned Xenopus δEF1 cDNA, analyzed the expression profile of the gene, and compared the inducing and interacting properties of the protein to that of XSIP1. Whereas XSIP1 RNA is selectively expressed in the early developing nervous system, we show that XδEF1 gene transcription is only activated during neurulation and that its expression is restricted to the paraxial mesoderm. From early tail bud stage, XδEF1 and XSIP1 are coexpressed in migratory cranial neural crest, in the retina, and in the neural tube. Overproduction of XδEF1 in RNA-injected embryos, like that of XSIP1, reduced the expression of BMP-dependent genes but only XSIP1 has the ability to induce neural markers. We find that XδEF1 and XSIP1 can both form complexes, although with different efficiency, with Smad3, with the coactivators p300 and pCAF, and with the corepressor CtBP1. Together, these results indicate that δEF1 and SIP1 do not function as antagonists during Xenopus early embryogenesis but do display different repression efficiencies and interaction properties. Developmental Dynamics 235:1491–1500, 2006. © 2006 Wiley-Liss, Inc.

INTRODUCTION

One classification of transcription factors with C2H2 type zinc fingers is based on the number and organization of the fingers (Iuchi,2001). The zinc finger homeobox 1 (zfhx1) proteins are characterized by two clusters of multiple zinc fingers: one located in the N-terminal part of the protein and one in the C-terminal part, separated by a linker domain containing a homeodomain-like sequence. This family has two members only, δEF1 (ZEB1, Zfhx1a) and SIP1 (ZEB2, Zfhx1b), which exist in zebrafish, Xenopus, chick, mouse, and human (van Grunsven et al.,2001).

δEF1 null mutant mice have multiple skeletal malformations, absence of joints, a severe T-cell deficiency, and they die immediately after birth (Takagi et al.,1998). SIP1 null mouse embryos show multiple defects from embryonic day (E) 8.5 onward, do not undergo embryonic turning, are severely growth retarded by E9.5, and die shortly thereafter. Analysis of these embryos showed that SIP1 is essential for the correct specification of the neuroepithelium and delineation of the neural plate, for the development of vagal neural crest cells, and for the delamination and migration of cranial neural crest cells (Van de Putte et al.,2003). Patients with heterozygous deletions or truncations of the ZFHX1B gene (chromosome 2q22) suffer from Mowat–Wilson syndrome (Wakamatsu et al.,2001; Mowat et al.,2003). In Xenopus, XSIP1 is expressed in the prospective neurectoderm during gastrulation (Eisaki et al.,2000; van Grunsven et al.,2000). Overproduction of XSIP1 induces neural-specific gene expression in explants of ectoderm and represses in the embryo the transcription of the pan-mesodermal and nodal-induced brachyury (Xbra) gene, of BMP4, and of other genes in the presumptive epidermis (Eisaki et al.,2000; Lerchner et al.,2000; Papin et al.,2002; Postigo et al.,2003). Loss-of-function studies in Xenopus showed that XSIP1 is essential for neural induction in amphibian embryos (Nitta et al.,2004).

Studies aiming at elucidating the function(s) and mechanism(s) of action of the zfhx1 proteins mainly have been carried out in cultured cells. The DNA binding activity and specificity of the zinc finger clusters of δEF1 and SIP1 and a model in which both zinc finger clusters bind preferentially to two spaced binding sites (primarily CACCT(G), and alternatively CACANNT) have been proposed (Sekido et al.,1997; Remacle et al.,1999; Verschueren et al.,1999). Spaced CACCT(G) sites are found in the upstream regulatory region of many potential target genes (van Grunsven et al.,2001). Recently, the E-cadherin gene has been identified as a target for transcriptional repression by δEF1 and SIP1 in vitro (Grooteclaes and Frisch,2000; Comijn et al.,2001) and in vivo (Van de Putte et al.,2003).

Despite the vast number of studies on potential target genes for zfhx1 family proteins, the mechanism(s) by which these proteins regulate target genes remains unclear. They do bind to the MH2 domain of activated Smad proteins (Verschueren et al.,1999; Postigo,2003), the intracellular effectors of transforming growth factor-beta (TGFβ) family signalling (Hill,2001; Shi and Massague,2003). δEF1 and SIP1 interact with the corepressor CtBP (carboxyl terminal binding protein; Chinnadurai,2003), which contributes to their repressive activity in vitro (Furusawa et al.,1999; Postigo and Dean,1999; Grooteclaes and Frisch,2000; Shi et al.,2003; van Grunsven et al.,2003). However, when full-length proteins, which are defective for CtBP binding, are used in these in vitro assays, no clear correlation between the repressive activity of the zfhx1 proteins and CtBP binding can be established for the target genes studied thus far (van Grunsven et al.,2003). Different repressive potentials have been attributed to the different domains of δEF1 and SIP1, but evidence showing a correlation between these domains and the function of δEF1 and SIP1 in vivo has been lacking. Postigo and coworkers (Postigo et al.,2003) have shown that δEF1 and SIP1 can regulate TGFβ/bone morphogenetic protein (BMP) signalling in opposite ways, with δEF1 synergizing with Smad-mediating transcriptional activation and SIP1 repressing it. It was suggested that this differential effect is because SIP1 functions only as a transcriptional repressor by recruiting CtBP to the Smads, whereas δEF1 also binds to the coactivators p300 and pCAF, which displace CtBP and allow it to function as an activator (Postigo,2003; Postigo et al.,2003). In Xenopus, XSIP1 has been shown to repress directly Xbra expression during gastrulation, thereby playing a role in the differentiation between mesodermal and ectodermal layers (Lerchner et al.,2000). In contrast, overproduction of human δEF1 induces Xbra expression ectopically, suggesting that both proteins compete for the same binding sites on the Xbra promoter and act as antagonists (Postigo et al.,2003).

Here, we cloned a Xenopus δEF1 cDNA and show that the gene is activated later than XSIP1 and that it is predominantly expressed in the embryonic paraxial mesoderm. By using an identical RNA injection approach used by others, we show that XδEF1, like XSIP1, represses the expression of BMP-dependent genes, the inhibitory effect observed being less dramatic, however, than that of XSIP1, which may explain the inability of XδEF1 to neuralize ectodermal cells. We show that both XδEF1 and XSIP1 can interact with the coactivators pCAF and p300 and the corepressor CtBP1. In contrast to XSIP1, XδEF1 needs p300 binding to be able to bind to Smads. Together, these data indicate that XδEF1 and XSIP1 do not function antagonistically but have overlapping, yet slightly different activities during early Xenopus development.

RESULTS

δEF1 and SIP1 Are Differentially Expressed During Xenopus Embryogenesis

We and others have found that Xenopus SIP1 (XSIP1) is expressed in the early embryo in the neurectoderm and plays a role in neural differentiation (Eisaki et al.,2000; van Grunsven et al.,2000; Nitta et al.,2004). In this manuscript, we wanted to know whether SIP1 and δEF1 have overlapping, counteracting, or distinct functions in the regulation of early embryogenesis. Therefore, we cloned the Xenopus δEF1 (XδEF1) cDNA and compared its expression and activities in the embryo to that of XSIP1. Figure 1A shows an amino acid alignment of human, mouse, and Xenopus laevis δEF1. The domains that were identified in earlier studies in mouse and human δEF1 are conserved in Xenopus δEF1. An exception is the C3H zinc finger that is present in mδEF1 (from amino acid 499 to 519), and in mouse and Xenopus SIP1 but is absent from human and Xenopus δEF1 (Fig. 1B). This zinc finger has been shown to be necessary for the binding of the POU-domain of Oct1 to rat δEF1 (Smith and Darling,2003). In addition, the first CtBP binding site that is present in mouse and human δEF1 and in SIP1 proteins is not conserved in Xenopus. During the course of this study, the NIH Xenopus initiative reported a full-length Xenopus laevis clone very similar to our Xenopus δEF1 gene (GenBank accession no. BC073606, further referred to as XδEF1b; 92% homology) that has the first CtBP site intact (Klein et al.,2002). A schematic representation of Xenopus laevis and mouse δEF1 and SIP1 is given in Figure 1B, showing that the overall similarity is low between SIP1 and δEF1, except for the well-known domains. In particular, the amino-terminal (NZF) and carboxy-terminal (CZF) zinc fingers are highly conserved as well as the homeodomain-like domain (HD; Genetta and Kadesch,1996). Striking is the low homology between the respective Smad-binding domain (SBD) regions, i.e., 15% between XδEF1 and XSIP1, and 28% between mδEF1 and mSIP1.

Figure 1.

δEF1 amino acid and XZfhx1 family comparison. A: Amino acid sequence comparison between human (hs), mouse (mm), and Xenopus laevis (xla, this study; xlb, δEF1 submitted to GenBank by an NIH Xenopus initiative; GenBank accession nos. U12170, D76432, DQ319867, and BC073606, respectively). The boxed sequences are previously determined functional domains in mouse or human δEF1. C2H2 and C3H indicate the type of zinc finger; SBD, Smad binding domain; CtBP, CtBP binding site; HD, homeo-like domain. B: Schematic representation of XδEF1 and its similarities with mδEF1 and XSIP1. As a reference the homology between XSIP1 and mSIP1 is also represented. Dark gray boxes represent C2H2-type zinc fingers; light gray boxes C3H-type zinc-fingers; stripes represent individual CtBP binding sites (CBS); NZF, N-terminal zinc finger domain; CZF, C-terminal zinc finger domain. Comparisons at the amino acid level were carried out using ClustalW analysis.

Expression profiling showed that XδEF1 transcripts can be detected from the mid-neurula stage, while XSIP1 transcripts are already detected at early gastrula stage (Fig. 2A). By in situ hybridization, XδEF1 expression is observed first at stage 13–14 in the paraxial mesoderm, while XSIP1 expression is restricted to the neurectoderm (Fig. 2B,C). By late neurula to early tail bud stage, XδEF1 expression is also detected in migratory cranial neural crest, in the retina and in the neural tube. Within the spinal cord, XδEF1 mRNA is confined to the dorsal lateral edges (Fig. 2D,F,H,J,L). By late tadpole, XδEF1 RNA levels remain high in the head and the tail (Fig. 2N,O). The expression profile of XδEF1b is identical to that of XδEF1a (see Supplementary Figure S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat). From late neurula stage, expression of XδEF1 overlaps that of XSIP1 in the eye and cranial neural crest (Fig. 2D,E,H–M) but not in the dorsal midline where XδEF1 expression is excluded from presumptive premigratory trunk neural crest cells, which strongly express XSIP1 (Fig. 2F,G). Thus, δEF1 and SIP1 show distinct temporal and spatial expression patterns during Xenopus embryogenesis, although substantial overlap of expression also exists. Besides the expression of XδEF1 in the eye, these results corroborate the expression domain of δEF1 observed in mouse and chick embryos (Funahashi et al.,1993; Takagi et al.,1998).

Figure 2.

Expression of XδEF1 and XSIP1. A: Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of XδEF1 expression at the indicated embryonic stages, compared with that of XSIP1. Control RT-PCR of histone H4 was used as a reaction and loading control. BO: Whole-mount in situ hybridization of XδEF1 and XSIP1 expression at the indicated stages. Sections through the spinal cord of the embryos in D,E,J, and K are shown in F,G,L, and K, respectively. ba, branchial arches; ev, eye vesicle; n, notochord; nc, neural crest; nt, neural tube; s, somites; tt, tail tip. B–E: Dorsal view, anterior toward the right. J,K: Dorsal view, anterior toward the left. (H,I,L,M) Lateral view, anterior toward the right.

XδEF1 Represses, Less Efficiently Than XSIP1, BMP-Dependent Genes and Does Not Cause Induction of Neural Markers in Ectodermal Cells

Injection of RNA encoding XSIP1 or human SIP1 in blastomeres of one- to four-cell stage Xenopus embryos has been shown to cause repression of Xbra in the mesoderm and of the ectodermal marker epidermal keratin (EK), and induction of the neural plate marker Sox2, a gene critical for early neurogenesis (Papin et al.,2002; Nitta et al.,2004). To determine whether XδEF1 has transcription regulatory effects similar to XSIP1 in this type of experiments, we injected RNA encoding a myc-tagged XδEF1 or XSIP1 into Xenopus embryos (250–500 pg RNA). Western blot detection of protein extracts from XSIP1 and XδEF1 injected embryos shows that the proteins are equally well expressed (Fig. 3F). Overproduction of XδEF1, like that of XSIP1, was found to repress Xbra (Fig. 3A). At these doses, the expression of the BMP-targets Msx1 (Suzuki et al.,1997), Gata2 (Walmsley et al.,1994), and E.K. was also repressed by both proteins (Fig. 3B–D), the repression observed in the XδEF1-injected embryos appearing often less dramatic than that observed in XSIP-injected embryos. At lower doses of XδEF1 (25–50 pg RNA), the expression of these marker genes was unaltered, whereas their expression was still totally abolished by XSIP1 (Fig. 3E, and data not shown). In XδEF1-injected embryos, no ectopic Sox2 or Sox3 expression was observed, whereas it was found expanded in embryos injected with XSIP1 RNA. Identical results were obtained using non–myc-tagged versions of the constructs (Fig. 4A,B, and data not shown).

Figure 3.

Overexpression of XδEF1 in embryos represses less efficiently than XSIP1 bone morphogenetic protein (BMP) -dependent genes. A: Four-cell stage embryos were injected with XδEF1 or XSIP1 encoding RNA into one blastomere (500 pg) in the equatorial region and analyzed for Xbra expression. Both XδEF1 and XSIP1 repress Xbra expression. BE: Eight- to sixteen-cell embryos injected into one animal blastomere with low (50 pg) or high (500 pg) doses of the indicated mRNA and analyzed for Gata2, Msx1, and EK expression. F: Western blot analysis of XδEF1- and XSIP-injected embryos showing that the proteins are equally expressed. β-Tubulin antibody is used as a loading control. A–C: Respective repression 100%, n = 54 for XSIP1 and 100%, n = 50 for XδEF1 (A); (B) 100%, n = 24 for XSIP1 and 100%, n = 34 for XδEF1; (C) 100%, n = 16 for XSIP1 and 100%, n = 14 for XδEF1. (D) 100%, n = 44 for XSIP1 and 100%, n = 36 for XδEF1; (E) all none or only partially inhibited, n = 30 for XδEF1 injected and all repressed, n = 32 for XSIP1. A: Embryos at gastrula stage are viewed from the vegetal pole. Embryos in D and E are viewed laterally, anterior to the right; B is a ventral view; C a dorsal view, anterior to the right. The injected area (indicated by a black arrow) is marked by blue staining for nuclear LacZ expression obtained after coinjection of LacZ RNA. All control embryos are viewed dorsolaterally except Xbra, which is a vegetal view.

Figure 4.

Overexpression of XδEF1, in contrast to XSIP1, cannot induce Sox2 expression in the embryo. A,B: The 8- to 16-cell embryos injected into one animal blastomere with 500 pg of the indicated mRNA and analyzed for Sox2 expression. Respective induction: none, n = 30 for XδEF1; 100%, n = 28 for XSIP1. CJ: The 16- to 32-cell embryos were injected in one ventral-most animal blastomere with the indicated mRNAs and analyzed for Sox2 expression at stage 15. Respective induction of Sox2: 0%, n = 34 (C); 0%, n = 11 (D); 0%, n =17 (E); 92%, n =26 (F); 0%, n = 11 (G); 92%, n =26 (H); 0%, n =22 (I); 0%, n =15 (J). A,B: Anterior view. C–J: Ventral view, anterior toward the right. The injected area (arrow) is marked by blue staining for nuclear LacZ expression obtained after coinjection of LacZ RNA.

The acquisition of neural fate in Xenopus, like in chick embryos, does not occur solely by inhibition of BMP ligand activity (the “default” model) but involves BMP-independent fibroblast growth factor (FGF) signaling (Linker and Stern,2004; Delaune et al.,2005). We asked whether increasing FGF signaling in the embryo would help XδEF1 and XSIP1 to induce neural tissue. As reported, injection of RNA encoding a dominant-negative form of the BMP receptor (tBR, 200 pg/blastomere) together with 0.16 pg of eFGF RNA into ventral animal cells in 16-cell stage embryos induces Sox2 in ventral ectodermal cells. At these doses, neither tBR nor eFGF alone activates mesoderm formation or induces Sox2 (Fig. 4D–F; Delaune et al.,2005). XδEF1 and XSIP1 RNA were injected alone or together with 0.16 pg of eFGF RNA. We found that, similar to the coinjection of tBR+eFGF RNA, a combination of XSIP1+eFGF RNA induced Sox2 in ventral ectodermal cells, whereas neither XSIP1 nor eFGF could do this on their own (respectively, Fig. 4E,G,H). In identical conditions, no activation of Sox2 was observed in embryos injected with XδEF1+eFGF RNAs (Fig. 4I,J).

We next used animal cap assays for documenting the effects of XδEF1. Animal caps, cut from embryos injected at the two-cell stage in each blastomere with 250 pg of myc-tagged XSIP1 RNA, revealed a dramatic increase in the levels of Sox3, Sox2, and NCAM transcripts. We did not observe this increase in myc-tagged XδEF1 RNA injected caps (Fig. 5A–C) or XδEF1b injected caps (see Supplementary Figure S2). Identical results were obtained with non–myc-tagged constructs (data not shown). These dramatic increases of the expression of neural markers (not seen in intact embryos) is most probably due to an increased level of FGF signaling in caps compared with the corresponding domain in intact embryos (Delaune et al.,2005). Indeed, inhibition of FGF signaling either by ΔR4 RNA, which encodes a dominant-negative membrane-bound form of the FGF receptor FGFR4 (Hongo et al.,1999), or treatment of the explants with SU5402, a pharmacological inhibitor of FGFRs (Mohammadi et al.,1998), reduced XSIP1-mediated induction of Sox2 (Fig. 5D). Thus, XδEF1 and XSIP1 appear to share similar regulatory properties, XSIP1 being, however, a more potent inhibitor than XδEF1 of the known anti-neuralizing effects of BMPs.

Figure 5.

XδEF1, in contrast to XSIP1, does not induce neural markers in animal caps. AC: Animal caps explanted at stage 9 express Sox2, Sox3, and NCAM in response to XSIP1 but not to XδEF1 injection. The expression of each marker in uninjected animal caps and embryo are shown at the right of each panel. Respective induction 0%, n = 35 for XδEF1, 100%, n = 27 for XSIP1 (A); 0 %, n = 49 for XδEF1, 97 %, n = 39 for XSIP1 (B); 14 %, n = 49 for XδEF1, 94 %, n = 57 for XSIP1 (C). D: Coinjection of XSIP1 mRNA with ΔR4 mRNA or treatment of the explants from the time of their excision at blastula stage 9 with 120 μM SU5402 reduced Sox2 induction (positive caps: XSIP1, 88%, n = 27; XSIP1+ΔFGF4r, 20%, n = 29; XSIP1+SU5402, all reduced, n = 19). Control embryos are viewed from the dorsal side, anterior toward the right.

Both XδEF1 and XSIP1 Interact With the Coactivators p300 and pCAF and the Corepressor CtBP1

As our results show that XδEF1 is less effective than XSIP1 to inhibit the expression of BMP-dependent genes, we tested the possibility whether this different activity is due to differential binding to coactivators or corepressors. First, we tested whether the Xenopus proteins, like their mammalian orthologues, can interact with the corepressor CtBP1. Flag-tagged CtBP1 and Myc-tagged XSIP1 or XδEF1 were coexpressed in 293T cells and immunoprecipitations (IP) on the cell lysates were done using anti-Flag antibodies. CtBP1 immunoprecipitated XSIP1 and XδEF1 very efficiently (Fig. 6A).

Figure 6.

XδEF1, like XSIP1, interacts with the coactivators p300 and pCAF and the corepressor CtBP. Plasmids encoding the indicated proteins were transfected in 293T cells and cell lysates processed for immunoprecipitation with the appropriate antibodies and protein complexes (IP) separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and analyzed by Western blotting. A portion of each lysate was analyzed for expression of transfected proteins by Western blot (Lysate). A: CtBP can coimmunoprecipitate XSIP1 and XδEF1, immunoprecipitations were done with anti-Flag antibodies. B: p300 can immunoprecipitate XδEF1 and XSIP1. Immunoprecipitations were carried out with anti-HA antibodies. C: pCAF interacts with XSIP1 and XδEF1 and the pCAF -SIP1 and pCAF-δEF1 complexes do not contain CtBP. Here, anti-Flag antibodies were used. Note that the synthesis levels of the different Zfhx1 proteins are equal but that SIP1 forms a stronger complex with pCAF than δEF1 does (compare lanes 2–5).

Next, we tested the interaction between p300 (HA-tagged) and the full-length zfhx1 proteins of both Xenopus and mouse. Cell lysates from 293T cells were incubated with anti-HA antibodies, and the interaction between p300-HA and XSIP1/XδEF1 was visualized using anti-Myc antibodies after separation on polyacrylamide gel electrophoresis (PAGE) followed by Western blotting. In contrast with previous reports, we observed that p300 co-immunoprecipitated full-length XδEF1 and XSIP1, the interaction with XSIP1 appearing slightly stronger than that with XδEF1 (Fig. 6B). This prompted us to test pCAF binding to SIP1 proteins also. Acetylation of CtBP-interacting proteins by pCAF has been shown to regulate CtBP interaction (Zhang et al.,2000). Therefore, we decided to compare the ability of pCAF to interact with δEF1 and SIP1 in the presence of CtBP1. Cells were cotransfected with expression plasmids coding for CtBP1, Flag-pCAF, and Myc-tagged XδEF1, XSIP1, mδEF1 and mSIP1, respectively. Cell lysates were precipitated with anti-Flag antibody, and the precipitates were analyzed for the presence of pCAF, δEF1 or SIP1, and CtBP. pCAF, like p300, was capable of forming complexes with all full-length SIP1 and δEF1 proteins, the interaction with SIP1 being much more efficient than that with δEF1 proteins (the XδEF1 band becoming visible after longer exposure of the gel shown here; Figure 6C, lanes 2–5). We observed that CtBP is not present in the pCAF-δEF1 complexes but found this to be true for pCAF-SIP1 complexes as well. In the same experiment, Flag-SIP1 was used to immunoprecipitate CtBP, showing that a CtBP-SIP1 interaction occurs under these conditions (Fig. 6C, lane 6). Thus, both Xenopus δEF1 and SIP1 are found to interact, although with different efficiency, to the corepressor CtBP and the coactivators p300 and pCAF, the interaction with pCAF prohibiting the recruitment of CtBP to δEF1/SIP1.

XδEF1 Binding to Smad3 Requires p300

Smads are proteins that may have an influence on the transcriptional activity of δEF1 and SIP1 as well. To further investigate on a molecular interaction level, the differences in repressive capacity of XδEF1 and XSIP1, we also addressed whether there was a difference in Smad-binding capacity. SIP1 has been found to interact with Smads 1, 2, 3, and 5, and a Smad-binding domain (SBD) sufficient for interaction has been mapped in SIP1 (Verschueren et al.,1999). Using a yeast two-hybrid approach, we were unable to detect interactions between mouse δEF1 polypeptides and Smads 1 through 8 (van Grunsven et al.,2003), but a Smad-δEF1 complex has been observed by other means (Postigo et al.,2003). To explain the discrepancy between these and our results, we hypothesized that δEF1 and SIP1 both can bind to Smads but not in a similar manner. In particular, we took into account the possibility that SIP1 would bind to Smads directly, whereas δEF1 would bind Smads through p300. This would explain why in yeast the δEF1-Smad interaction could not be confirmed. First, we investigated whether the Smad3-p300 interaction was influenced by the presence of XSIP1 or XδEF1. Expression plasmids encoding Flag-Smad3, Myc-XδEF1 or Myc-XSIP1, and HA-tagged p300 were transfected in 293T cells in the presence or absence of a constitutively active TGFβ type I receptor (CA Alk4), and precipitations were done using anti-Flag antibodies. A representative Western blot analysis of the proteins that coprecipitated with Flag-Smad3 is shown in Figure 7. As shown by Pouponnot and coworkers (Pouponnot et al.,1998), p300-Smad3 interaction can be evidenced when TGFβ signaling is activated but it is absent without signaling (Fig. 7, lanes 1 and 2). The interaction was enhanced when XδEF1 or XSIP1 was coexpressed (lanes 3–4). Together with our finding that XδEF1 and XSIP1 can both interact with p300 (Fig. 6B), this suggests that both XδEF1 and XSIP1 can stabilize the Smad3-p300 interaction.

Figure 7.

In contrast to SIP1, δEF1-Smad3 binding requires p300. Lanes 1–2: Smad3 forms a complex with p300 in the presence of a constitutively active TGFβ receptor (CA-ALK4; lane 1). Lanes 3–4: In the presence of XδEF1 or XSIP1, this Smad3-p300 complex-formation is enhanced. Lane 6: Smad3 can coimmunoprecipitate XSIP1 in the presence of CA-ALK4. Lanes 7–8: p300 cotransfection results in a gain of Smad3-XδEF1 complex, while the XSIP1-Smad3 complex formation is enhanced. We used anti-Flag antibodies in these experiments to perform the immunoprecipitations.

The same Flag-Smad3 precipitates were analyzed for the presence of XδEF1 or XSIP1 in the presence or absence of p300, and under TGFβ–Smad pathway activation (Fig. 7, lane 5–8, the precipitated complexes visualized in lanes 3–4 were loaded again in 7–8). In the absence of p300, but in the presence of TGFβ signaling, only XSIP1 interacted with Smad3 (compare lane 5 to 6). However, when p300 was coexpressed, a clear interaction between Smad3 and full-length XδEF1 was also evidenced, as well as an increase in Smad3-XSIP1 interaction (lanes 7–8). These data indicate that full-length XδEF1 and XSIP1 can both interact with Smad proteins, the XδEF1–Smad interaction however being one that is strictly dependent on the presence of p300 in the complex.

DISCUSSION

The existence of families of transcription factor, like the zfhx1 family studied here, does not only point out the possibility for a different role of each member during development but also does not exclude that their mechanism of action may vary. We used Xenopus embryos to study the expression, the function, and the mechanism of action of SIP1 and δEF1 in vertebrates. We find that XδEF1 is only expressed during postgastrulation stages. We demonstrate that both δEF1 and SIP1 have, to different extends, the capacity to transcriptionally repress BMP-dependent genes but only XSIP1 has strong neuralizing activity. This activity of XSIP1 requires intact FGF signaling. Like XδEF1, XSIP1 is found to bind to the coactivators p300 and pCAF and, in addition, and unlike XSIP1, XδEF1 binding to Smad3 is strictly dependent on the presence of p300 in the complex.

In accordance with previous observations made in chick and mouse (Funahashi et al.,1993; Takagi et al.,1998), we find that XδEF1 is activated in the postgastrulation embryo only, predominantly in mesodermal tissues, and later it is also expressed in the nervous system. In contrast, XSIP1 is expressed from early gastrula onward and is highest in the dorsal epithelial cells (Fig. 2; Eisaki et al.,2000; van Grunsven et al.,2000). Consequently, in contrast to the model proposed by Postigo et al. (2003), SIP1 is the only member of the zfhx1 family to function during early Xenopus embryogenesis in the differentiation of mesodermal and neural tissues.

When we analyzed the activities of full-length XδEF1 and XSIP1 proteins, we observed that both proteins inhibit the expression of BMP-dependent genes. A difference in transcriptional repression capacity was observed when lower amounts of RNA were injected, which could explain the inefficiency of XδEF1 to induce neural tissue (Figs. 4, 5). Thus, these results differ from those obtained by Postigo and collaborators using human δEF1 and showing that overexpression of human δEF1 from injected RNA (using 250–500 pg RNA) has no effect on EK expression but induces ectopic Xbra expression. Therefore, these authors proposed a model in which δEF1 and XSIP1 function as antagonists in the control of the activation of Xbra (Postigo et al.,2003). The basis of the contradiction between their results and ours is unclear. It may reflect distinct specificity of action of the human and Xenopus δEF1 proteins. It should be pointed out however that the model of an activating function of δEF1 in the Xenopus embryo, put forth by Postigo et al. (2003), relies mainly on the effect of a hypothetical dominant-negative form of δEF1 and on the hypothesis that δEF1 is expressed earlier during Xenopus development than SIP1. However, the latter hypothesis is proven wrong (Fig. 2). Our observations rather suggest that the two proteins have overlapping functions in the embryo in those sites where their expression overlaps.

We observed that intact FGF signaling is required for XSIP1 to neuralize ectodermal cells (Fig. 4). This finding agrees with recent observations in Xenopus, like in amniote and ascidian embryos, that early FGF signaling besides BMP inhibition is required for neural induction (Linker and Stern,2004; Delaune et al.,2005). It is unknown what the FGF signaling means for XSIP1's activity. FGF has been shown to inhibit BMP activity and, hence, to promote neural development by phosphorylating the linker region of Smad1 (Pera et al.,2003). Alternatively, FGF could influence XSIP1 activity through modification of SIP1 or through cooperation with other FGF-induced proteins, such as Gata and Ets factors (Bertrand et al.,2003) or other factors that remain to be identified.

Human δEF1 has been reported to act as an activator and repressor of transcription, but human SIP1 would only be a repressor, due to the incapacity of SIP1 to bind to coactivators (Postigo et al.,2003). We have not observed this in our experiments. Our studies show that both XSIP1 and XδEF1 bind to the coactivators p300 and pCAF and that their ectopic expression decreases BMP-dependent gene transcription. Again the basis of this contradiction is unclear. It should be pointed out that most of the promoter studies performed to demonstrate the functional interaction between the coactivators and δEF1 were rather artificial and that SIP1 was never tested in these assays (Postigo et al.,2003).

The function of pCAF or p300 binding to XSIP1 may be to acetylate it, thereby leading to the release of CtBP, as has been postulated for δEF1 (Postigo et al.,2003) and shown for other proteins like E1A (Zhang et al.,2000). Our results also show that both XδEF1 and XSIP1 stabilize Smad3-p300 interaction and that the Smad3-XδEF1 interaction, but not Smad3-XSIP1 interaction, requires the presence of p300 (Fig. 7). The molecular basis of this difference is not clear. SIP1 and δEF1 proteins have quite divergent SBD regions (Fig. 1B), and it would be interesting to determine whether their SBD domain is responsible for this difference by constructing, for example, chimera proteins between XSIP1 and XδEF1.

In conclusion, we suggest that the proposed model by Postigo and collaborators (Postigo et al.,2003) in which SIP1 and δEF1 function particularly as antagonists in the regulation of TGFβ family signaling during early Xenopus development needs to be redefined. We rather suggest that the two proteins have overlapping, yet slightly different, activities in the embryo in those sites where their expression overlaps. A way to achieve a better understanding of the action of XδEF1 and XSIP1 on BMP/TGFβ signaling would be to determine whether their differential interaction with Smads and with the p300/pCAF coactivators explain their difference in transcriptional repressive activity.

EXPERIMENTAL PROCEDURES

Expression Plasmids

All plasmids used were cloned using standard cloning techniques. XδEF1 (GenBank accession no. DQ319867) was cloned using degenerate primers and standard PCR techniques from neural stage Xenopus poly(A) RNA and subsequently subcloned in pCS2 in frame with 6xMyc (pCS3). The XδEF1b clone was obtained from Open Biosystems (Biocat, Heidelberg, Germany) and recloned in pCS3 using the gateway system (Invitrogen, Merelbeke, Belgium). pCS2 with inserts of Myc-tagged wild-type mouse SIP1 (Verschueren et al.,1999), Xenopus SIP1 (van Grunsven et al.,2000), and mouse δEF1 (Funahashi et al.,1993) cDNA and pCDNA3 with inserts of Flag-tagged wild-type mouse SIP1 have been described previously (Remacle et al.,1999; Papin et al.,2002). Flag-CtBP1, Flag-Smad3, p300HA, and Flag-pCAF expression plasmids have been described elsewhere (Eckner et al.,1994; Verschueren et al.,1999; van Grunsven et al.,2003).

Xenopus Embryo Manipulations and Injections

Xenopus embryos were obtained from adult frogs by hormone-induced egg-laying and in vitro fertilization using standard methods (Sive et al.,2000) and staged according to Nieuwkoop and Faber (1997). Capped mRNAs were synthesized in vitro by using the mMessage mMachine kit (Ambion, Austin, TX). mRNA was injected as described at the indicated stage. Synthetic nuc-LacZ RNA (from 100 to 500 pg) was used as a lineage tracer. Initial references for the previously described expression construct used are ΔFGFR4 (Hongo et al.,1999) and eFGF (Delaune et al.,2005). Animal cap explants were prepared at late blastula (stage 9) and cultured in 1× Steinberg medium supplemented with 0.1% bovine serum albumin until early (Sox2, Sox3) or late (NCAM) neurula stage. SU5402 (Calbiochem) was dissolved in dimethyl sulfoxide (120 mM), and diluted 1,000-fold in 0.1× MBS for whole-embryo treatments, and in 1× MBS for animal cap explants. All injections were performed at least twice, and the result of one representative experiment is shown.

In Situ Hybridization

Embryos were fixed in MEMFA, stained for β-galactosidase activity with 5-bromo 4-chloro-3-indolyl-β-galactopyranoside (X-Gal), and processed for in situ hybridization using digoxigenin-labeled antisense RNA probes according to (Sive et al.,2000). To produce the XδEF1 probe, XδEF1 was cloned in PCRscript, linearized with NotI and transcribed with T7. Plasmids used for generating the other in situ probes were as described previously: Xbra (Smith et al.,1991), EK (Jonas et al.,1985), Msx1 (Suzuki et al.,1997), Gata2 (Walmsley et al.,1994), Sox2 (Kishi et al.,2000), Sox3 (Penzel et al.,1997), and NCAM (Krieg et al.,1989). For sections, embryos after completion of the whole-mount procedure were gelatin-embedded and vibratome-sectioned at 30 μm thickness.

Cell Culture, Transient Transfections, and Immunoprecipitations

All immunoprecipitations of SIP1 and δEF1 complexes, and PAGE and Western analysis, were carried out as described (van Grunsven et al.,1996) using the following buffer for lyses and at least 3 × 1 ml 10 min washing after incubation of the cell extracts with antibody and protein A/G Sepharose beads: 50 mM Tris/HCl pH 7.6; 170 mM NaCl; 0.1% (v/v) NP40; 50 mM NaF; 2 mM Na3VO4; 0.2 mM dithiothreitol; complete protease inhibitor cocktail; 10% (v/v) glycerol. Mouse anti–Myc-tag antibody and rabbit anti-CtBP antiserum were from Santa Cruz. Anti-HA (clone 12CA5) antibodies were donated by Innogenetics S.A. (Zwijnaarde, Belgium). Anti-Flag antibodies were from SIGMA.

Acknowledgements

We thank Kristin Verschueren and other members of her team for discussion of our results. We also thank Horst Grunz, Richard Harland, Laurent Kodjabachian, Roger Patient, and Thomas Sargent for providing us with various constructs; and we are grateful to Sadia Kricha for technical assistance. E.B. and D.H. was funded by the Belgian program Inter-Universitary Attraction Poles from the Prime Minister's Office, Science Policy Programming. E.B. was also supported by the Communauté Française de Belgique and the Fund for Scientific Medical Research. The work in the D.H. lab was supported by VIB, the University of Leuven (OT/00/41), and the Fund for Scientific Research-Flanders. L.A.v.G was a postdoctoral fellow from FWO-V and V.T. a fellow of the Fonds pour la Formation a la Recherche dans l′Industrie et l′Agriculture.

Ancillary