Positive and Negative Regulation of the Transforming Growth Factor β/Activin Target Gene goosecoid by the TFII-I Family of Transcription Factors

Abstract

Goosecoid (Gsc) is a homeodomain-containing transcription factor present in a wide variety of vertebrate species and known to regulate formation and patterning of embryos. Here we show that in embryonic carcinoma P19 cells, the transcription factor TFII-I forms a complex with Smad2 upon transforming growth factor β (TGFβ)/activin stimulation, is recruited to the distal element (DE) of the Gsc promoter, and activates Gsc transcription. Downregulation of endogenous TFII-I by small inhibitory RNA in P19 cells abolishes the TGFβ-mediated induction of Gsc. Similarly, Xenopus embryos with endogenous TFII-I expression downregulated by injection of TFII-I-specific antisense oligonucleotides exhibit decreased Gsc expression. Unlike TFII-I, the related factor BEN (binding factor for early enhancer) is constitutively recruited to the distal element in the absence of TGFβ/activin signaling and is replaced by the TFII-I/Smad2 complex upon TGFβ/activin stimulation. Overexpression of BEN in P19 cells represses the TGFβ-mediated transcriptional activation of Gsc. These results suggest a model in which TFII-I family proteins have opposing effects in the regulation of the Gsc gene in response to a TGFβ/activin signal.

Development of vertebrates proceeds through a series of inductive events in which signaling molecules produced by one cell influence the developmental fate and morphogenesis of neighboring cells. Morphogenesis and the development of pattern are often presaged by the expression of spatially restricted genes. Thus, it is essential to understand how expression of these spatially restricted genes is regulated in vivo. One signaling molecule shown to play an important role in the formation of mesoderm belongs to the TGFβ superfamily, activins and nodals (20, 51). Activin and nodal family members have been shown to initiate the formation of mesoderm in ectodermal explants (animal caps) and mimic the function of Spemann's organizer in Xenopus (1, 23, 40, 43). Gene responses to the TGFβs, activins, and nodals are mediated by Smad transcription factors, Smad2 and Smad3. Upon being phosphorylated, Smad2 and/or Smad3 form a complex with Smad4 and translocates to the nucleus to regulate transcription of various target genes both positively and negatively (28, 29). Although Smad complex can directly bind to DNA and can activate gene transcription, the affinity between Smad complex and their binding sites is relatively weak and less specific. By assembling a higher order complex with nuclear cofactors, Smad complex is able to bind to a target gene promoter with higher affinity and specificity (28).

Gsc is a homeobox-containing protein present in a variety of vertebrate species. Gsc is a transcriptional repressor that regulates formation and patterning of vertebrate embryos (19, 26, 47). Xenopus Gsc is strictly expressed at the dorsal lip of embryos (6, 9), which is considered as an organizer because of its ability not only to autonomously differentiate into notochord but also to change the fate of neighboring mesodermal cells into cells with dorsal characteristics and to induce a neural axis in the overlying ectoderm. Ectopic expression of Gsc is able to duplicate axis and mimic the organizer phenomenon in Xenopus (9). Therefore, Gsc has been implicated in organizer-dependent developmental processes such as cell migration and dorsalization of the mesoderm. In mice, Gsc is expressed in the developing primitive streak region, and cells expressing Gsc are fated to form the head, which later gives rise to the anterior notochord and endoderm and the head mesoderm (5, 18). Hence, it is of fundamental interest to understand how Gsc expression is regulated.

Two regions of the Gsc gene promoter, the distal element (DE) and the proximal element (PE), are critical for its proper expression pattern in Xenopus embryos (47). The DE is activated by members of activin/nodal signals, and the PE is activated by Wnt signals (47). Recently, the Xenopus homolog of BEN (also known as a product of genes called GTF2IRD1, MusTRD1, WSCR11, or GTF3) has been isolated by a yeast one-hybrid screen as a protein which interacts with the DE of Xenopus Gsc (XGsc) gene promoter (37). Downregulation of XBEN by morpholino antisense oligonucleotides (MO) in Xenopus embryos exhibited an inhibition of activin-mediated induction of XGsc, which led to the conclusion that BEN is a positive transcriptional regulator of the Gsc gene in Xenopus embryos (37). BEN is a member of the TFII-I family of transcription factors. TFII-I and BEN share multiple helix-loop-helix (HLH) domains and a leucine zipper domain (3). Despite the structural similarity, TFII-I often acts as a transcriptional activator, and BEN often acts as a transcriptional repressor among the systems studied previously (35, 42, 44).

Here we show that the TFII-I/Smad2 complex is recruited to the Gsc gene promoter upon TGFβ stimulation, which leads to transcriptional activation of the Gsc gene in P19 cells. Downregulation of TFII-I in P19 cells completely abolishes the induction of Gsc by TGFβ. BEN, on the contrary, is constitutively recruited to the Gsc gene promoter and downregulates Gsc expression, possibly by recruiting a histone deacetylase enzyme. These results led us to propose a model in which TFII-I family transcription factors have opposing effects in the regulation of the Gsc gene in a TGFβ signal-dependent fashion.

MATERIALS AND METHODS

Plasmid constructs.

A full-length human TFII-I (delta isoform) and BEN tagged with either the glutathione S-transferase (GST) or the green fluorescent protein (GFP) and Flag-tagged HDAC3 cDNA expression constructs were described previously (8, 44, 49).

Cell culture and transfection.

Both P19 and Cos7 cell lines were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (HyClone). P19 and Cos7 cells were transfected using Fugene6 (Roche) according to the manufacturer's instructions.

Reverse transcription-PCR (RT-PCR).

P19 cells were treated with 100 pM of TGFβ (R&D Systems) in 0.2% fetal calf serum-DMEM. Total RNA was extracted by TRIzol (Invitrogen), and 1 μg of total RNA was subjected to reverse transcription using first-strand cDNA synthesis kit (Invitrogen) according to manufacturer's instructions. One-tenth of the reaction mixture was used as template for each PCR. Primers used are 5′-CCACTTGGAGACTTCTTCTTCTTCG-3′ and 5′-ATGTAGGGCAGCATCTGGTGCG-3′ for mouse Gsc, 5′-GGCAACAAAGCAACAGAAGACG-3′ and 5′-ACCTCATCCAAACACTGAAGTTCC-3′ for mouse Parp2, 5′-CATTGCTGTGTATGAGACCGACG-3′ and 5′-CAAAATCCACTTCAAGGTTGCGAG-3′ for mouse TFII-I, and 5′-GCTCTGTTTGATGAGAAGTGCGG-3′ and 5′-GTTCCTGCCCTTTTTGGTGC-3′ for Xenopus TFII-I. Mouse hypoxanthine-guanine phosphoribosyl transferase (HPRT) and Xenopus fibroblast growth factor receptor (FGFR) primers were described before (22). Semiquantitative PCR was performed in the presence of [³²P]dCTP, and products were separated on a sodium dodecyl sulfate (SDS)-polyacrylamide gel followed by autoradiography and quantitation by a phosphorimager.

Small inhibitory RNA (siRNA) treatment.

Synthetic oligonucleotides were purified and annealed according to the manufacturer's instructions (Dharmacon). The siRNA sequence for targeting TFII-I is 5′-UCAGCUCCAUGAGGAGGAUCU-3′. The siRNA sequence for BEN was described before (42). As a negative control, siRNA with a scrambled sequence (Dharmacon) was used. The siRNAs were transfected to P19 cells by OligofectAMINE (Invitrogen) according to the manufacturer's instructions. Forty-eight hours after transfection, cells were treated with 100 pM TGFβ and harvested.

Luciferase assay.

Luciferase assays were carried out essentially as described previously (22). After transfection the cells were reseeded onto 12-well plates and treated with 100 pM TGFβ for 24 h in 0.2% fetal calf serum-DMEM.

Immunoprecipitation and immunoblot assays.

Cos7 cells were transfected with GST fusion of wild-type TFII-I (WT), TFII-I (ΔBR), TFII-I (p70), or BEN. Cells were then treated with 100 pM of TGFβ for 2 h and lysed in TNE buffer (1% NP-40, 10 mM Tris-HCl [pH 7.5], 1 mM EDTA, 150 mM NaCl). Lysates were incubated with glutathione agarose beads (Amersham Pharmacia). Protein complexes were separated on a SDS-polyacrylamide gel electrophoresis (SDS-PAGE) gel and transferred to polyvinylidene difluoride membranes. Anti-Smad2 polyclonal antibodies (Zymed), anti-glyseraldehyde-3-phosphate dehydrogenase (GAPDH; Research Diagnostic), and anti-GST monoclonal antibody (Sigma) were used to detect Smad2, GAPDH, and GST-TFII-I or GST-BEN fusion proteins. Anti-TFII-I and anti-BEN rabbit polyclonal antibodies were described previously (8a, 42).

Xenopus embryo culture and microinjections.

Eggs were obtained by injecting Xenopus laevis females with 700 U of human chorionic gonadotropin. In vitro fertilization and embryo culture were done as described previously (32). Staging was according to Nieuwkoop and Faber (33). MO to Xenopus TFII-I had the following sequence: 5′-AGTCAGGCATCTGGAAACGGGCCAT-3′. Control MO had the following sequence: 5′-AGAGACTTGATACAGATTCGAGAAT-3′. Capped synthetic RNAs were generated from a full-length human TFII-I cDNA in pcDNA3 (49) and pCS2-LacZ (41) by in vitro transcription with SP6 or T7 RNA polymerase (24) using the MessageMachine kit (Ambion). Embryo microinjections were performed as described previously (41). Fifty nanograms (high dosage) or 5 ng (low dosage) of MO and/or 4 ng of synthetic RNAs were injected into each embryo.

In vitro transcription an translation.

Capped synthetic RNAs were generated from Xenopus TFII-I and BEN cDNA in pCS2 by in vitro transcription with SP6 or T7 RNA polymerase using the MessageMachine kit (Ambion). Synthetic RNAs were incubated with MO at room temperature for 30 min, followed by in vitro translation using the PROTEINscript II kit (Ambion) in the presence of ³⁵S-labeled methionine (Amersham). In vitro translation products were separated on an SDS-PAGE and visualized by autoradiography.

RNA isolation and Northern blot analysis.

Northern analysis of total RNA isolated from stage 10 embryos was done essentially as described previously (41). RNA from the equivalent of 1.5 embryos was loaded per lane. Endogenous fibronectin RNA served as a control for loading. Antisense RNA probes were prepared by in vitro transcription from plasmids containing fibronectin and goosecoid cDNA fragments, using [³²P]UTP and SP6 or T7 RNA polymerase (41).

Chromatin immunoprecipitation assay.

Chromatin immunoprecipitation (ChIP) assay was performed as described previously (39). Briefly, Cos7 cells were transfected with GST-TFII-I (WT, ΔBR, or p70), GST-BEN, or GFP-BEN expression plasmid with or without the (DE)x6-luc reporter plasmid. After TGFβ treatment, genomic DNAs were sonicated to an average length of 500 bp. Soluble chromatin was then incubated with either anti-GST anti-GFP monoclonal antibody or rabbit nonspecific immunoglobulin G as a negative control. Immunoprecipitated DNA fragments were amplified with PCR primers (21 cycles) specific for the Xenopus Gsc gene, 5′-GAATACAAGCTAGCTTGCATGCCTG-3′ and 5′-CAGACTGCAGTCCTCTCCCATCTGTG-3′, or the murine Gsc gene, 5′-TTAGTTGAGGGAGGACACAG-3′ and 5′-CCTAATCTTTTCCATTCGGCAG-3′.

PCR products were separated on an agarose gel and visualized by ethidium bromide staining.

Immunofluorescence staining.

Cos7 cells were grown on coverslips, transfected with GST-TFII-I or GFP-BEN or both constructs together. Immunostaining was performed as described previously (45). Nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI).

RESULTS

Gsc as a target of TGFβ signaling in P19 cells.

The TGFβ superfamily of growth factors, activins and nodals, are known to regulate expression of the homeobox-containing gene, Gsc, in the organizer region of Xenopus embryos (1, 23, 43). To examine whether or not the TGFβ signaling pathway positively regulates transcription of the Gsc gene in mammalian cells, we measured Gsc mRNA levels in P19 cells, in which we have previously confirmed the expression of transcription cofactors essential for activation of early embryonic genes such as Vent-2 and Tlx-2 (22). P19 cells express low levels of Gsc mRNA. TGFβ stimulation led to a 4.9-fold increase in expression and persisted for at least 4 h (Fig. 1A). Treatment of cells with the protein synthesis inhibitor cycloheximide prior to TGFβ stimulation did not abolish the induction of Gsc transcripts, suggesting that the induction of Gsc mRNA is at the transcription level (data not shown). These results confirm that P19 cells contain a transcription factor(s) required for the induction of Gsc by the TGFβ/activin signaling pathway.

FIG. 1.

Induction of Gsc by TGFβ is mediated by TFII-I. A. P19 cells were stimulated with TGFβ for various lengths of time. Total RNA was extracted and subjected to RT-PCR using primers of Gsc and HPRT as the loading control. Relative Gsc expression during the time course is shown in the right panel by quantitation of the Gsc bands with a phosphorimager followed by normalizing the expression with the HPRT level. The −RT lane represents the sample of lane 5 that is directly applied to PCR without reverse transcriptase treatment. B. P19 cells were transfected with TFII-I-specific siRNA (left panels) or scrambled siRNA as a control (right panels) prior to TGFβ stimulation. Total RNAs were prepared after TGFβ treatment, and expression of TFII-I, Gsc, and Parp2 (loading control) was measured by RT-PCR. Gsc expression was quantitated by a phosphorimager followed by normalization with Parp2 expression and is presented as a graph at the right. Closed circles indicate Gsc expression in cells treated with siRNA of TFII-I, and open circles are samples from control (scrambled) siRNA-treated cells. C. Total cell lysates were prepared from P19 cells transfected with TFII-I-specific siRNA (TFII-I), BEN-specific siRNA (BEN), or scrambled siRNA (control). Each lysate was subjected to immunoblot analysis using anti-TFII-I and anti-BEN antibodies. Anti-GAPDH monoclonal antibody was used as a loading control.

TFII-I is essential for transcriptional activation of Gsc by TGFβ.

Because it has been reported that XBSCR11, a Xenopus homolog of the TFII-I-family protein BEN, is able to bind to the distal element (DE) of a Xenopus Gsc promoter and positively modulate its transcription (37), we examined a possible role of its closest family member, TFII-I, in the regulation of Gsc in murine P19 cells. We addressed the physiological significance of TFII-I in the induction of Gsc by TGFβ in P19 cells by downregulating endogenous TFII-I expression using small inhibitory RNA (siRNA). Both TFII-I mRNA (Fig. 1B, mTFII-I) and protein (Fig. 1C) expression were significantly downregulated in cells transfected with TFII-I siRNA compared to the control (scrambled) siRNA-transfected cells. The specificity of the effect of TFII-I siRNA was also confirmed by using BEN siRNA (42) (Fig. 1C). TFII-I siRNA showed no effect on expression of its family protein BEN (Fig. 1C). Similary, BEN siRNA, which efficiently downregulates BEN protein expression, had no effect on TFII-I expression (Fig. 1C). The induction of Gsc by TGFβ was completely abolished when TFII-I expression was downregulated (Fig. 1B, mGsc, left panel). The level of TFII-I mRNA was not altered by TGFβ treatment (Fig. 1B, mTFII-I). The effect of TFII-I siRNA on Gsc is specific because the level of expression of unrelated genes such as Parp2 (Fig. 1B, mParp2) or HPRT (data not shown) was similar to that of control siRNA-treated cells. These results suggest that TFII-I is essential for induction of Gsc by TGFβ.

Downregulation of TFII-I blocks Gsc expression in Xenopus embryos.

The role of TFII-I in the transcriptional regulation of Gsc expression was further examined in Xenopus embryos. The Xenopus homologue of TFII-I (XTFII-I) is expressed at relatively constant levels throughout Xenopus development (Fig. 2A). We then injected MO specific for XTFII-I into four-cell-stage Xenopus embryos to downregulate endogenous TFII-I protein expression. Inhibition of XTFII-I protein synthesis by XTFII-I MO was demonstrated by in vitro translation analysis using in vitro-transcribed XTFII-I mRNA. XTFII-I protein synthesis was efficiently inhibited when XTFII-I MO was added to the reaction (Fig. 2B). Control MO had no effect on XTFII-I protein synthesis (Fig. 2B). MO against XBEN, which has been reported previously (37), inhibited XBEN protein synthesis and reduced XTFII-I protein synthesis by 50% (Fig. 2B). Northern blot analysis demonstrated that the endogenous expression of XGsc was localized in the dorsal side of embryos at stage 10, and no expression was detected in the ventral side of embryos as previously reported (Fig. 2C, lanes 1 and 2) (26, 47). Despite similar expression levels of fibronectin mRNA (Fig. 2C, XFN, lanes 3 and 4), significant reduction of XGsc mRNA was observed in the embryos in which XTFII-I MO was injected at the dorsal side (Fig. 2C, XGsc, lanes 3 and 4). RNA samples from embryos injected with TFII-I MO into the ventral side did not show alteration of XGsc expression at the dorsal side (Fig. 2C, XGsc, lane 5), suggesting that TFII-I is essential for the expression of Gsc at the dorsal side of Xenopus embryo, presumably under the control of activin/nodal signaling pathways.

FIG. 2.

Role of TFII-I in Xenopus embryos. A. RT-PCR analysis with RNA isolated from whole embryos demonstrates the expression of XTFII-I throughout Xenopus embryogenesis. XFGFR expression is shown as a loading control. B. Effect of XTFII-I MO on XTFII-I protein synthesis was examined in vitro. In vitro-transcribed XTFII-I mRNA was incubated with XTFII-I MO prior to in vitro translation reaction. Control MO or XBEN MO was used as a control. In vitro-synthesized protein was separated on SDS-PAGE and visualized by autoradiography. Effect of XBEN MO on XBEN protein synthesis was examined in parallel. C. RNA samples from the stage 10 embryos injected with XTFII-I or control MO at dorsal or ventral side were subjected to Northern blot analysis using probes for XGsc and XFibronectin (XFN). Ribosomal RNAs stained with ethidium bromide are shown as loading controls. D. XTFII-I MO-injected embryos lack dorsal lip structure at stage 10. Statistical analysis of dorsal lip phenotype at stage 10 is shown as a bar graph. Total numbers of embryos subjected to the analysis are shown above the graph. Hi and Lo indicate high dosage (50 ng/embryo) and low dosage (5 ng/embryo) of MO injection, respectively. Coinjection of human TFII-I RNA rescues the TFII-I MO phenotype. St., stage.

More than 95% of embryos in which a high dose of XTFII-I MO is injected at the dorsal side appeared to lack a dorsal lip structure, whereas control MO-injected embryos or uninjected embryos (data not shown) developed normal dorsal lips at stage 10 (Fig. 2D, dorsal injection). This effect is dose dependent because only 32% of embryos in which a low dose of XTFII-I MO is injected failed to develop a dorsal lip structure at stage 10 (Fig. 2D). Despite a lack of dorsal lip structure, the XTFII-I MO-injected embryos appeared to go through normal gastrulation, suggesting that dorsal lip formation was delayed but not inhibited. This is consistent with the observation that mice in which the Gsc gene is inactivated undergo normal gastrulation (52). Eighty-eight percent of embryos injected with a high dose of XTFII-I MO at the ventral side showed no delay in dorsal lip formation, suggesting that the effect of XTFII-I MO is specific to inhibition of XTFII-I at the dorsal side of the embryo where Gsc is expressed (Fig. 2D, ventral injection). At later stages, however, embryos dorsally injected with XTFII-I MO appeared to have defects in head structure and reduced anterior-posterior axis (data not shown). Fifty percent of the embryos injected with human TFII-I mRNA together with XTFII-I MO appeared to develop a dorsal lip structure normally at stage 10 (Fig. 2D). All embryos coinjected with β-galactosidase RNA and XTFII-I MO showed delayed dorsal lip formation (Fig. 2D). Therefore, we conclude that the XTFII-I MO phenotype observed in Xenopus embryos is specific and caused by downregulation of the Gsc gene. Taken together, these results suggest that TFII-I plays an important role in early vertebrate developmental processes by regulating the expression of the Gsc gene and possibly another homeotic gene(s).

TFII-I and Smad2 synergistically upregulate Gsc transcription through the DE.

The TGFβ/activin signaling pathways are known to regulate Gsc expression through the DE (7, 26, 47). The 29-bp DE sequence is highly conserved among Gsc gene promoters from different vertebrate species (10, 12, 47). A reporter construct containing six tandem copies of Xenopus DE sequence upstream of the luciferase gene (DE-Luc) was transfected into P19 cells with increasing amounts of human TFII-I expression plasmid (Fig. 3A). Low levels of exogenous TFII-I slightly augmented the basal reporter activity, which was further enhanced by TGFβ stimulation (Fig. 3A, 0.125 and 0.25 μg). Both the basal and TGFβ-induced reporter activities were strongly augmented with increasing TFII-I expression plasmid DNA (0.5 and 1.0 μg). Immunofluorescence staining of Smad2 showed that overexpressed TFII-I facilitates nuclear translocation of Smad2 in the absence of TGFβ signal (data not shown). This observation is consistent with our finding that overexpression of TFII-I augmented the basal reporter activity and TGFβ stimulation had no further effect on the reporter activity at the highest level of TFII-I (Fig. 3A). Activation of the DE-Luc reporter by TFII-I is mediated by the DE sequence, because only a slight increase of the basal activity was observed by expression of TFII-I with the reporter construct lacking the DE sequence (Fig. 3A, −104-Luc).

FIG. 3.

TFII-I and Smad2 mediate TGFβ-dependent transcriptional activation of the Gsc. A. P19 cells were transfected with increasing amounts of GST-tagged TFII-I expression plasmid (125, 250, 500, or 1,000 ng) and the Gsc reporter plasmid containing six copies of DE motifs (DE-Luc) or the Gsc reporter plasmid without DE (−104-Luc). Luciferase activity was measured 16 h after TGFβ stimulation. Exogenous TFII-I protein was visualized by Western blot with anti-GST monoclonal antibody. B. Two Gsc reporter constructs (DE-Luc and −104-Luc) were injected in Xenopus embryos with or without XTFII-I MO, control MO, or hTFII-I mRNA, followed by measurement of luciferase activities. Xenopus Siamois promoter construct (Sia-Luc) was used as a control. C. GST-tagged TFII-I expression plasmid (125 ng) and the DE-Luc reporter construct were transfected to P19 cells alone or together with Smad2, Smad3, or Smad1 expression plasmid (100 ng). D. Cos7 cells overexpressed with GST-tagged TFII-I with the DE-Luc reporter plasmid were stimulated with TGFβ for various lengths of time and then subjected to the ChIP assay. Recruitment of TFII-I to the DE occurs within 15 min after TGFβ stimulation and persists until 4 h later. WB, Western blot; IP, immunoprecipitation.

The physiological role of TFII-I in the regulation of the DE-Luc reporter was tested in Xenopus embryos using XTFII-I MO. Coinjection of TFII-I mRNA and DE-Luc led to a threefold increase in luciferase activity (Fig. 3B). Embryos injected with XTFII-I MO and DE-Luc had reduced luciferase activity compared to that of control MO-injected or MO-uninjected embryos (Fig. 3B). The inhibitory effect of XTFII-I MO on DE-Luc is mediated specifically through the DE sequence because the reporter construct lacking the Gsc DE (−104-Luc) or the reporter containing Xenopus Siamois promoter (Sia-Luc) (15), which is known to respond to the Wnt signal, was not inhibited (Fig. 3B). This result is consistent with the observation of the TFII-I siRNA-treated P19 cells (Fig. 1B) and suggests that XTFII-I acts as a transcriptional activator of the Gsc promoter through the DE sequence. The DE-Luc reporter activity was also slightly enhanced by overexpression of the TGFβ/activin-specific signal transducer Smad2; however, synergistic upregulation of the reporter activity upon TGFβ stimulation was observed only when both TFII-I and Smad2 expression plasmids were cotransfected (Fig. 3C). Cotransfection of TFII-I with the bone morphogenetic protein-specific signal transducer Smad1 did not increase the reporter gene (Fig. 3C). Interestingly, coexpression of TFII-I with Smad3 showed no synergistic effect on the DE-Luc reporter (Fig. 3C), suggesting that TFII-I and Smad2 specifically cooperate to upregulate the transcription of Gsc upon TGFβ treatment.

Recruitment of TFII-I to the Gsc promoter is mediated by TGFβ signal.

To examine a possible recruitment of TFII-I to the DE in vivo, Cos7 cells were transfected with GST-TFII-I expression plasmid and the DE-Luc reporter construct, treated with or without TGFβ, and followed by a chromatin immunoprecipitation (ChIP) assay (Fig. 3D). Consistent with the result of an electrophoretic gel mobility shift assay (data not shown), TFII-I was recruited to the DE as early as 15 min after TGFβ stimulation and remained on the promoter for up to 4 h after stimulation (Fig. 3D). This result suggests that although the ectopic TFII-I is constitutively in the nucleus, it is unable to be recruited to the DE in the absence of TGFβ signaling.

Carboxyl-terminal domain of TFII-I is essential for the activation of Gsc.

To examine which domain of TFII-I is important for the transcriptional activation of Gsc and binding to the DE, we employed wild-type TFII-I and two key deletion mutants (Fig. 4A), the basic region (BR, amino acids 301 to 306) deletion mutant (ΔBR) and the carboxyl-terminal deletion mutant (p70) in the reporter assay (Fig. 4A) and ChIP assay (Fig. 4B). The BR in traditional HLH proteins has been shown to constitute a sequence-specific DNA binding domain (16). The ΔBR mutant was shown to lack significant DNA binding activity and fail to upregulate the c-fos and Vβ promoters (8). However, in the DE reporter assay, the ΔBR mutant, which is expressed at the same level as the wild-type TFII-I, retained about 50% of its transcription activity compared to that of the wild type (Fig. 4A). Consistent with this functional assay, the ChIP assay showed that the ΔBR is recruited to the Gsc gene in vivo upon TGFβ stimulation, although the extent of promoter occupancy seems to be less than the wild-type TFII-I (Fig. 4B, ΔBR). In contrast to ΔBR, the p70 mutant (Fig. 4A), in which the C-terminal 222 amino acids containing a transcriptional activation domain is deleted (8), failed to activate the reporter even at the highest expression level (Fig. 4A, p70). Despite its lack of transcriptional activity, the p70 mutant was constitutively recruited to the DE (Fig. 4B, p70). Therefore, the C-terminal transcriptional activation domain is essential for TGFβ-dependent regulation of Gsc by TFII-I.

FIG. 4.

Carboxyl terminus of TFII-I is essential for its interaction with Smad2 and activation of a Gsc promoter. A. Schematics of TFII-I wild type and its mutant proteins, ΔBR and p70. LZ and NLS stand for a leucine zipper domain and a nuclear localization signal, respectively. The HLH repeat domains are shown as R1 to R6. Increasing amounts (125, 250, 500, or 1,000 ng) of GST-tagged wild-type or mutant TFII-I expression plasmids were transfected into P19 cells with the Gsc reporter construct, followed by the luciferase assay. Levels of expression of each TFII-I constructs were visualized by Western blot using anti-GST antibody. B. GST-tagged wild-type or mutant TFII-I expression plasmids and DE-Luc construct were transfected into Cos7 cells, stimulated with TGFβ for 1 h, and subjected to the ChIP assay. Wild-type or mutant forms of ectopic TFII-I bound to genomic DNAs were isolated by anti-GST monoclonal antibody. DNA fragments, which contain the DE sequence, were amplified with a semiquantitative PCR. C. GST-tagged wild-type or mutant TFII-I expression plasmids were transfected into the Cos7 cells. Existence of TFII-I in the Smad2 immunoprecipitates was visualized by Western blot with anti-GST monoclonal antibody. Levels of exogenous TFII-I in total cell lysates are shown by Western blot using anti-GST antibody. IP, immunoprecipitation; IgG, immunoglobulin G; WB, Western blot.

TFII-I interacts with Smad2 upon TGFβ signaling.

To test whether TFII-I can physically associate with the signal transducers of the TGFβ/activin signaling pathway, we examined the ability of GST-TFII-I to interact with endogenous Smad2 or Smad3 in mammalian cells by immunoprecipitation of Smad2 or Smad3 followed by Western blot with anti-GST antibody. Smad2 and TFII-I formed a complex upon TGFβ stimulation (Fig. 4C, WT). Consistent with this result, addition of anti-Smad2 or anti-Smad4 antibodies in a reaction of a gel mobility shift assay abolished the TGFβ-dependent band (data not shown), suggesting that TFII-I, Smad2, and Smad4 may be recruited to the DE as a complex in response to a TGFβ signal. Because Fast1 (also known as FoxH1) has been known to bind to the activin response element (ARE) of the Mix2 gene upon TGFβ/activin stimulation (25, 27, 34, 48), we tested whether the murine homologue of Fast1 (Fast2) is present in the TGFβ-inducible DE-binding complex. Anti-Fast2 antibodies have been previously shown to supershift the band containing Fast2 in gel mobility shift assays using ARE oligonucleotides as probes (25). In our assay, the addition of anti-Fast2 antibodies had no effect on any bands, suggesting that the DE-binding complex does not contain Fast2 (M. Ku and A. Hata, unpublished observation). Consistent with this result, overexpression of Fast2 did not affect the DE-Luc reporter activity in P19 cells (data not shown).

Smad3 failed to interact with TFII-I under the same condition (Fig. 4C), supporting the result that Smad3 did not upregulate the DE-Luc activity together with TFII-I (Fig. 3C). The ΔBR mutant of TFII-I, which retained partial transcriptional activity (Fig. 4A, ΔBR), interacts with Smad2 as well as the wild-type TFII-I (Fig. 4C, ΔBR). The p70 mutant of TFII-I, on the other hand, did not interact with Smad2 even in the presence of TGFβ stimulation (Fig. 4C, p70), suggesting that the C-terminal domain of TFII-I is required for its interaction with Smad2. Despite the fact that the p70 mutant of TFII-I is constitutively recruited to the DE, its failure to interact with Smad2 may explain its lack of transcriptional activation of Gsc. Whether the interaction between TFII-I and Smad2 is direct is yet to be addressed.

BEN negatively regulates TFII-I-mediated activation of Gsc by TGFβ.

We next examined the role of transcription factor BEN, structurally related to TFII-I, in regulation of Gsc by TGFβ in P19 cells. In contrast to TFII-I, overexpressed BEN completely abolished TGFβ-dependent induction of the reporter, even at the lowest level of expression tested, suggesting that BEN might act as a negative regulator of the Gsc gene (Fig. 5A). We then tested whether TFII-I and BEN antagonize each other in the regulation of Gsc gene transcription. An increasing amount of the BEN expression plasmid was cotransfected with constant amounts of TFII-I together with the DE-Luc reporter construct (Fig. 5B). In the presence of increasing amounts of BEN (1:1, 1:2, 1:4, and 1:8 ratio of TFII-I:BEN), TGFβ-dependent activation of the DE by TFII-I was strongly reduced at the 1:1 ratio and was completely abolished at a higher ratio (Fig. 5B, 5th to 12th columns). In contrast, increasing amounts of TFII-I prevented the transcriptional inhibition by BEN (Fig. 5B). These results suggest that TFII-I and BEN have opposing transcriptional effects in regulating the Gsc promoter, and because inhibition of TFII-I activity by BEN was rescued by overexpression of TFII-I, TFII-I and BEN may compete for a common target.

FIG. 5.

BEN antagonizes the activation of Gsc by TFII-I. A. P19 cells were transfected with increasing amount of GST-tagged TFII-I or BEN expression plasmid (125, 250, 500, or 1,000 ng) and the DE-Luc reporter plasmid. Levels of expression of exogenous TFII-I or BEN were visualized by Western blot with anti-GST monoclonal antibody. B. Increasing amounts of BEN or TFII-I expression plasmid (125, 250, 500, or 1,000 ng) were transfected into P19 cells with a steady amount (500 ng) of TFII-I or BEN, respectively, together with the DE-Luc reporter plasmid. Luciferase activity was measured with or without TGFβ stimulation. C. Schematics of BEN wild type and its mutant proteins. LP and ΔNLS stands for a leucine zipper domain mutant and a nuclear localization signal-deleted mutant, respectively. The HLH repeat domains are shown as R1 to R6. Increasing amount (250 and 500 ng) of GST-tagged wild-type TFII-I, wild-type BEN, or BEN mutant expression plasmid was transfected into P19cells together with the DE-Luc reporter plasmids. D. Cos7 cells were transfected with GST-TFII-I or GFP-BEN alone or cotransfected with both GST-TFII-I and GFP-BEN (250 ng each/chamber slide). The ectopically expressed TFII-I and BEN were visualized by immunofluorescence staining using anti-GST antibody and GFP staining, respectively. Nuclei were stained with DAPI. E. GST-tagged TFII-I or BEN expression plasmid was transfected into the Cos7 cells. Complex formation between TFII-I or BEN and Smad2 was examined by immunoprecipitation of Smad2, followed by Western blot with anti-GST antibody. Total cell lysates were probed with anti-GST antibody to show expression of exogenous TFII-I or BEN. WB, Western blot; IP, immunoprecipitation.

Next we examined which domain of BEN is essential for repression of the DE-Luc in P19 cells. Increasing amounts of the wild type (WT) or various BEN mutant expression vectors were cotransfected with the DE-Luc reporter construct in P19 cells (Fig. 5C). Whereas wild-type BEN strongly repressed the reporter activity (Fig. 5C, WT), a BEN mutant in which the nuclear localization signal (amino acids 883 to 889) is deleted (Fig. 5C, ΔNLS) and is known to localize predominantly in the cytoplasm (44) showed markedly decreased inhibition of the reporter activity compared to wild-type BEN (Fig. 5C, ΔNLS). The leucine zipper domain of BEN was shown to mediate not only homodimerization of the protein (46) but also heterodimerization with TFII-I (42), and introduction of prolines in place of leucines at positions 38 and 45 (Fig. 5C, LP) has been shown to significantly decrease its ability to form both homo- and heterodimers (42). The result that BEN(LP) strongly inhibited the Gsc reporter activity suggests that formation of neither the BEN homodimers or TFII-I-BEN heterodimers is essential for negative regulatory action of BEN (Fig. 5C, LP).

Mechanism of antagonistic relationship between TFII-I and BEN.

To explore the mechanism by which BEN opposes the effect of TFII-I, we studied the cellular localization of these two factors. Immunostaining analysis of Cos7 cells overexpressing GST-TFII-I or GFP-BEN alone indicated that both TFII-I and BEN are localized predominantly in the nucleus both in the absence and presence of TGFβ (Fig. 5D, GST-TFII-I and GFP-BEN). Cotransfection of GST-TFII-I and GFP-BEN at a 1:1 ratio did not alter nuclear localization of TFII-I, suggesting that nuclear exclusion of TFII-I may not be the major mechanism of inhibition of TFII-I by BEN under these conditions (Fig. 5D, GST-TFII-I plus GFP-BEN). Coexpression of TFII-I and BEN, however, slightly augmented cytoplasmic localization of BEN (Fig. 5D, GST-TFII-I plus GFP-BEN). Therefore, it is possible that a high level of TFII-I might be able to exclude BEN from the nucleus, leading to activation of DE-Luc.

We then tested whether Smad2 interacts with BEN. Unlike TFII-I, BEN appeared to have no physical association with Smad2, suggesting that it is unlikely that BEN antagonizes TFII-I activity by competing for Smad2 (Fig. 5E).

Next, the recruitment of TFII-I and BEN to the endogenous Gsc gene promoter in the absence or presence of TGFβ signal was examined by a ChIP assay in P19 cells. Consistent with the results shown in Fig. 3D, endogenous TFII-I was recruited to the Gsc promoter upon TGFβ stimulation (Fig. 6A, lanes 1 to 3). Endogenous BEN, however, was constitutively bound to the Gsc promoter in the absence of TGFβ stimulation (Fig. 6A, lane 4), and the amount of BEN recruited to the Gsc promoter was reduced upon TGFβ stimulation (Fig. 6A, lanes 5 and 6). The same result was obtained by a ChIP assay in Cos7 cells overexpressing GST-TFII-I or GST-BEN with the DE-Luc reporter construct (data not shown). Furthermore, TFII-I recruitment to the DE was decreased by overexpression of BEN in a dose-dependent manner and was completely abolished at the highest level of expression of BEN (Fig. 6B). These observations suggest that binding of TFII-I and BEN to the DE in response to TGFβ might be mutually exclusive.

FIG. 6.

Recruitment of BEN leads to downregulation of Gsc at a steady state. A. P19 cells were stimulated with 100 pM TGFβ for 0.5 or 1 h and subjected to ChIP assay. Genomic DNA fragments bound to endogenous TFII-I or BEN were isolated by anti-TFII-I or anti-BEN polyclonal antibodies. DNA fragments containing the DE of the endogenous Gsc gene were amplified with PCR, and products were visualized on an agarose gel. B. Overexpression of BEN abolishes TFII-I binding to the DE in vivo. Cos7 cells were transfected with a steady amount of GST-tagged TFII-I expression plasmid in the absence or presence of increasing amount of GFP-tagged BEN expression plasmids. Cells were then stimulated with 100 pM TGFβ for 2 h, and the ChIP assay was performed. Total cell lysates were subjected to Western blot with anti-GFP antibodies to visualize the expression of exogenous BEN. C. P19 cells were treated with siRNA for BEN or control (scrambled) siRNA and then treated with or without TGFβ for 2 h. Total RNA was extracted and subjected to RT-PCR using primers of mGsc, mBEN, and mHPRT as the loading control. D. P19 cells were transfected with the DE-Luc reporter together with or without the indicated amount of HDAC3 expression construct. Twenty-four hours prior to analysis, the indicated amounts of TSA were added to some cells. The activity obtained in the cell without exogenous HDAC3 in the absence of TSA has been set equal to 1. E. GST-tagged TFII-I or BEN expression plasmid was cotransfected with Flag-HDAC3 expression plasmid into the Cos7 cells. Complex formation between TFII-I or BEN and HDAC3 was examined by immunoprecipitation of TFII-I or BEN by anti-GST antibody, followed by Western blot with anti-Flag antibody (M2). Total cell lysates were probed with anti-GST or anti-Flag antibody to show expression of exogenous TFII-I, BEN, or HDAC3. IP, immunoprecipitation; W, Western blot.

Finally, we addressed the role of the recruitment of BEN to the Gsc promoter in the absence of TGFβ signal by downregulating endogenous BEN expression in P19 cells. The level of Gsc mRNA, which was very low in control siRNA treated cells (Fig. 6C, lane 3), was significantly increased by downregulation of BEN (Fig. 6C, lane 1). The TGFβ-induced level of Gsc mRNA in BEN siRNA-treated cells was similar to that of control siRNA-treated cells (Fig. 6C lanes 2 and 4). The effect of BEN siRNA on Gsc is specific because the level of expression of HPRT was similar to that of control-siRNA treated cells (Fig. 6C, mHPRT). This result demonstrates that BEN is responsible for a transcriptional repression of Gsc in the absence of TGFβ signal.

Because BEN has been shown to interact with histone deacetylase enzymes (HDACs) (45), we examined the possible involvement of HDACs in transcriptional repression of Gsc by BEN in the absence of TGFβ stimulation. As shown in Fig. 6D, the basal activity of DE-Luc reporter was increased to 1.4-fold with the highest concentration of the inhibitor of HDAC activity, trichostatin A (TSA). Expression of exogenous HDAC3 resulted in a 2.5-fold decrease of DE-Luc activity (Fig. 6D). Constitutive interaction between BEN and HDAC3 was observed when both BEN and HDAC3 were overexpressed in Cos7 cells (Fig. 6E). Under the same conditions, interaction between TFII-I and HDAC3 was not detected (Fig. 6E). This may reflect a weaker binding affinity between TFII-I and HDAC3 relative to BEN and HDAC3 as observed previously (45). These results suggest that recruitment of HDAC3 to the DE by BEN may play a role in the downregulation of Gsc in the unstimulated state.

DISCUSSION

TFII-I as a novel cofactor for TGFβ/activin-specific Smad.

Our study demonstrates that the HLH and leucine zipper domain transcription factor TFII-I is a novel nuclear cofactor of the TGFβ/activin signal transducer Smad2; its role is to recruit Smad2 and Smad4 to the DE site of the Gsc promoter in a TGFβ-dependent manner. Overexpression of TFII-I facilitated a translocation of Smad2 to the nucleus in the absence of TGFβ signal (M. Ku and A. Hata, unpublished observation). This observation is consistent with the result that overexpression of TFII-I augmented the basal reporter activity, and TGFβ stimulation had little effect on the reporter activity at the highest level of TFII-I. A similar observation was obtained by overexpression of other Smad cofactors, such as OAZ and Fast2 (22, 27). Alternatively, it is possible that the effect of TFII-I on the reporter construct is due to its lack of chromatin structure. TFII-I alone without Smad2/Smad4 may not be able to be recruited stably to endogenous Gsc gene promoter.

Interestingly, another TGFβ/activin signal transducer, Smad3, did not interact with TFII-I under the same conditions. Despite the structural homology, Smad2 and Smad3 are known to have different DNA binding specificities (53). The differential interaction between Smad2 or Smad3 and TFII-I may contribute to their difference in recognition of target genes.

Deletion analysis indicated that the C-terminal domain of TFII-I is essential for its interaction with Smad2. The lack of sequence homology between the C-terminal domains of TFII-I and BEN is consistent with our finding that only TFII-I can interact with Smad2 upon TGFβ stimulation.

Although the ΔBR mutant of TFII-I is deleted in the region that contacts DNA, it is recruited to the DE upon TGFβ stimulation. This observation raises two possibilities: (i) the BR mutant may retain residual DNA binding ability through a region other than BR, and (ii) another cofactor(s) in the TGFβ signal-dependent DE-binding complex might be facilitating the interaction between TFII-I and the DE. The gel mobility shift analysis showed no retarded band when ΔBR is ectopically expressed (data not shown). Because ΔBR is able to interact with Smad2, it is likely that they are recruited to the DE as a complex possibly with other DNA binding factor, resulting in partial activation of DE-Luc.

Previously, a short sequence [PPNK(X_3-4)D] shared among transcription cofactors for Smad2, including Fast1, Fast2, Mixer, and Milk, was identified as a Smad2 interaction motif (19, 36). We found a sequence similar to the PPNK motif in human (PPSKRPKANE), mouse (PPTKLKSTE) and Xenopus (PPNKKPKSTE) (amino acid residues conserved in the Smad2 interaction motif are underlined) TFII-I protein sequence, but not in BEN. It is of note that we found no evidence that Fast2 or Mixer/Milk is involved in TFII-I-mediated upregulation of the DE in the reporter assay (M. Ku and A. Hata, unpublished observation). The p70 mutant of TFII-I, which contains an intact PPNK-like motif but is deleted in the C-terminal domain, failed to interact with Smad2. It is possible that the PPNK-like motif in TFII-I alone is not sufficient for interaction with Smad2. Alternatively, it is possible that the PPNK-like motif (amino acid 279 to 288) of the p70 mutant, which is located adjacent to the DNA binding domain (amino acid 301 to 306), may not be available for interaction with Smad2 because it constitutively binds to DNA. It is noteworthy that the p70 mutant of TFII-I is able to bind to the DE even in the absence of a TGFβ signal in contrast to the wild-type TFII-I. This observation leads us to speculate that the C-terminal region of TFII-I might be inhibitory for DNA binding, possibly by masking a region required for contact with DNA, such as the basic region. When Smad2 interacts with TFII-I through the C-terminal region in a TGFβ-dependent fashion, the DNA contact region of TFII-I might become exposed. The C-terminal region of BEN has no sequence homology to TFII-I, therefore we speculate that the DNA binding domain of BEN might not be masked by the C-terminal region, rendering BEN constitutively able to bind to the DE.

BEN is a negative regulator of Goosecoid gene transcription.

Our results in P19 cells indicate that BEN is a negative regulator which occupies the DE only in the absence of TGFβ, resulting in downregulation of Gsc expression. A similar inhibitory effect of BEN was observed in the DE-Luc reporter assay in other mammalian cell lines, such as HepG2, Cos7, and Mv1Lu cell lines (data not shown). Opposing effect of TFII-I and BEN in regulation of a common target gene promoter has been reported in regulation of c-fos (44). Because BEN does not bind to the c-fos promoter, it was speculated that BEN might compete with TFII-I for a nuclear shuttling machinery which results in exclusion of TFII-I from the nucleus and inhibition of its transcriptional activity (44). Immunohistochemical analysis indicated that coexpression of TFII-I and BEN did not alter nuclear localization of TFII-I, suggesting that inhibition of TFII-I by BEN is not due to nuclear exclusion of TFII-I. Furthermore, a BEN mutant deleted in the serine stretch sequence localized at the C terminus (ΔSS) loses its ability to exclude TFII-I from the nucleus but still strongly inhibits TFII-I-mediated DE-Luc activity (data not shown). Therefore, we propose a novel mechanism of regulation of the Gsc gene by TFII-I and BEN (Fig. 7). In the unstimulated state, BEN constitutively occupies the DE of the Gsc promoter and represses its transcription. Upon TGFβ stimulation, Smad2, which is translocated to the nucleus as a complex with Smad4, interacts with TFII-I, binds to the DE presumably with a higher affinity than BEN, and dislodges BEN, leading to upregulation of Gsc gene transcription (Fig. 7). This is the first observation that TFII-I and BEN compete for the same DNA binding site and antagonize each other's function in a growth factor signal-dependent manner.

FIG. 7.

Proposed model for TGFβ signal-dependent transcriptional regulation of Gsc at a steady state. DE is occupied by BEN, which suppresses the transcription of Gsc possibly by recruitment of HDAC3. Upon TGFβ/activin stimulation, Smad2 translocates to the nucleus, interacts with TFII-I, displaces BEN, and activates Gsc transcription.

The Xenopus homolog of BEN was identified in a yeast one-hybrid screen for proteins able to bind to the DE of the Xenopus Gsc gene (37). This previous study suggested that XBEN acts as a positive regulator of Gsc based on the results that (i) the VP16 fusion protein of XBEN activates the DE-Luc reporter and endogenous Gsc expression in Xenopus embryos and (ii) injection of XBEN MO resulted in downregulation of Gsc in Xenopus embryos (37). The fusion of the potent transcriptional activator domain of VP16 to a DNA binding protein is able to switch a transcriptional repressor to an activator (17). Therefore, previous observation of the activation of DE-Luc by VP16-BEN fusion protein only confirms that XBEN binds to DE in vivo. When we injected the same XBEN MO into Xenopus embryos together with various luciferase reporter constructs, not only the DE-Luc reporter but also the reporter lacking the DE and a reporter for the Wnt signaling pathway (15) were strongly inhibited (M. Ku and A. Hata, unpublished observation), which leads us to conclude that the effect of XBEN MO may not be limited to the DE of the Gsc promoter. Furthermore, an inhibitory effect of XBEN MO was observed on the in vitro protein synthesis of XTFII-I. This observation raises the possibility that downregulation of Gsc by XBEN MO (37) might be due, at least in part, to downregulation of XTFII-I. Alternatively, XBEN might have a role in the basal transcription machinery or in the Wnt signal-dependent transcription complex. In fact, it is known that Gsc expression is regulated not only by activin/TGFβ-like signal but also by maternal Wnt-like signals through the PE sequence of the Gsc promoter (47). The difference between our observation and the previous report is not due to the species specificity of BEN, because when XBEN was examined by the DE-Luc reporter assay in P19 cells in parallel with human BEN (hBEN), XBEN inhibited the reporter activity as well as hBEN (data not shown). Recently it has been reported that hBEN can both positively and negatively regulate the immunoglobulin-variable region promoter depending on the cell line (42). This observation leads us to speculate that the difference of transcriptional activity of BEN in P19 cells and Xenopus embryos may be due to a formation of different complexes with a different cofactor(s) such as HDAC3. It will be interesting to examine the expression pattern of the Xenopus homolog of HDAC3 in early Xenopus embryos.

Regulation of homeodomain genes by TFII-I family member proteins during embryogenesis.

The Gsc gene belongs to a family of Hox genes (5) that encode transcription factors involved in positional specification during embryonic development in diverse organisms. Therefore, Gsc expression should be critically regulated both positively and negatively to avoid misexpression that might lead to serious developmental defects. Early and ubiquitous expression of TFII-I and BEN in both Xenopus and mouse embryos (2, 11, 14) is consistent with our finding that expression of Gsc, which is expressed early in vertebrate embryos, is regulated by these two transcription factors.

The TFII-I and BEN genes are localized in tandem on a region of chromosome 7 (7q11.23) which is hemizygously deleted in individuals with Williams-Beuren syndrome (WBS) (30). Among the clinical phenotype of WBS patients, craniofacial abnormalities and cognitive defects appear to have a close association with deletion of TFII-I and BEN (30, 31). Based on the observations that mice homozygous for a targeted deletion of the Gsc gene exhibited craniofacial abnormalities in the lower mandible and nasal cavity which are similar to craniofacial phenotypes of individuals with WBS (4, 38, 55), it is tempting to speculate that the transcriptional regulation of the Gsc gene by TFII-I and BEN might be a molecular basis of why individuals with WBS have a developmental disorder including craniofacial abnormalities (30). We speculate that the defects in head structure observed in Xenopus embryos at tadpole stage upon dorsal injection of XTFII-I MO are perhaps another indication of the involvement of TFII-I gene product in craniofacial defects in WBS.

The role of TFII-I and BEN in the regulation of embryonic genes regulated by activin/nodal signals.

The role of the TGF superfamily of growth factors in early development has been studied in various organisms. Signals produced by activins and activin-related molecules such as nodal are crucial for the patterning of mesoderm and endoderm (50). It is believed that these signals trigger precise transcriptional responses in order to control specific sets of genes both positively and negatively to induce differentiation into specific cell types. In the Xenopus embryo, the major activin-like mesoendoderm-inducing activity that activates Smad2 and Smad4 is zygotic and requires the maternal transcription factor VegT for its production (21, 54). It has been shown that BEN and TFII-I are expressed both maternally and zygotically at relatively consistent levels throughout both Xenopus and mouse development (2, 11, 14, 37). Thus, both TFII-I and BEN are available in the embryos to bind the Smads complex (Smad2 and Smad4) activated by the zygotic activin-like signals, thereby initiating transcription of downstream genes like Gsc.

Although a previous report has shown that Mixer/Milk proteins are able to bind to the DE in early Xenopus embryos (19), these proteins are not expressed maternally. Therefore, we speculate that TFII-I-Smad complex is capable of inducing Gsc in response to activin-like signals in the absence of Mixer/Milk in early Xenopus embryo. It is important, however, to examine the role of mammalian homologs of Mixer or Milk, which have not been cloned, in the regulation of Gsc by TFII-I/BEN in mammalian cells.

It is noteworthy that we were unable to rescue the dorsal lip phenotype of XTFII-I MO-injected embryos by coinjection of XGsc mRNA (data not shown), while rescue with hTFII-I mRNA was effective. TFII-I may be able to interact with a variety of factors in the early embryo and affect transcription of a wide range of genes, other than Gsc, which can contribute to the phenotype when expression of XTFII-I is downregulated. It is interesting to note that BEN protein changes its subcellular localization depending on the developmental stage. In the zygotic stage, BEN is found both in the cytoplasm and pronuclei. Between the two-cell stage and the early blastocyst stage, it is found in the nucleus. By embryonic days 5.5 to 6.5, BEN localizes in the cytoplasm and then translocates to the nucleus again at embryonic days 7.0 to 8.5 (2). For example, BEN is known to localize in the cytoplasm of cells in the developing primitive streak region, where Gsc is expressed (2). Therefore, it is possible that activin/nodal signals might regulate not only the TFII-I-Smad complex formation but also subcellular localization of BEN, which in turn may regulate the expression of target genes positively or negatively. We speculate that the TFII-I family of transcription factors may be involved in the regulation of other genes downstream of activins or nodals. Identifying novel genes that are regulated by TFII-I and/or BEN in response to developmental cues will help us better understand the molecular mechanisms underlying vertebrate development and presents a challenge for the future.