Bix4 is activated directly by VegT and mediates endoderm formation in Xenopus development

The three germ layers of the amphibian embryo are established by mechanisms involving both cytoplasmic localisation and cell-cell interactions (Harland and Gerhart, 1997). The localisation of cytoplasmic determinants occurs during oogenesis, when various mRNAs and proteins become differentially distributed within the oocyte. Prominent among these are the RNAs encoding Vg1, a member of the TGF-β family (Rebagliati et al., 1985; Weeks and Melton, 1987), and VegT (Horb and Thomsen, 1997; Lustig et al., 1996; Stennard et al., 1996; Zhang and King, 1996) a member of the T-box family of transcription factors (Papaioannou, 1997; Smith, 1999). Both of these RNAs are restricted to the vegetal hemisphere of the embryo. Cell-cell interactions play a role during blastula stages, when vegetal blastomeres induce overlying equatorial cells to form mesoderm rather than ectoderm (Harland and Gerhart, 1997). This signal may derive from both maternal components, such as Vg1 (Dale et al., 1993; Thomsen and Melton, 1993) and from zygotically activated gene products, such as the nodal-related proteins (Jones et al., 1995; Joseph and Melton, 1997; Smith et al., 1995) and derrière (Sun et al., 1999).

According to this scheme, therefore, the ectoderm of the embryo represents a ‘default’ pathway – the fate that is adopted by cells that receive no positional cue, whether derived from cytoplasmic determinant or inductive signal. Endoderm, by contrast, forms either as a direct consequence of inheriting particular vegetally localised determinants, or as a result of receiving a particularly high dose of a ‘mesoderm-inducing’ signal such as Vg1 or activin. Mesoderm, as described above, forms in the equatorial region of the embryo in response to a signal derived from the vegetal hemisphere.

Transcripts encoding the vegetally localised T-box protein VegT (also known as Antipodean, Xombi and Brat) (Horb and Thomsen, 1997; Lustig et al., 1996; Stennard et al., 1996; Zhang and King, 1996) play an essential role in early Xenopus development. Depletion of maternal VegT RNA by injection of antisense oligonucleotides into the oocyte causes embryos to develop with no endoderm and greatly reduced amounts of mesoderm, and vegetal pole blastomeres lose the ability to induce mesoderm from animal pole tissue (Zhang et al., 1998). VegT is a transcription activator (Horb and Thomsen, 1997), and these results suggest that its targets must include genes involved both in the specification of endoderm and in the production of mesoderm-inducing signals. Identification of such targets would represent a significant advance in coming to understand the molecular basis of endoderm and mesoderm formation.

Members of the Bix family of paired-type homeobox-containing genes are strong candidates for direct VegT targets (Tada et al., 1998). Bix1-Bix4 were originally isolated as targets of the T-box gene Xbra, which is expressed at the onset of gastrulation in the prospective mesoderm of the Xenopus embryo (Smith et al., 1991). However, like the related genes Mix.1 (Rosa, 1989), Mix.2 (Vize, 1996) and Milk (Ecochard et al., 1998), Bix1-Bix4 are expressed in the endoderm as well as the mesoderm of the embryo, and their activation precedes that of Xbra (Tada et al., 1998). These results indicate that other genes are also involved in the regulation of Bix expression and an obvious candidate is the maternally and vegetally expressed VegT. Indeed, misexpression of VegT in Xenopus animal caps induces expression of Bix1, and the upstream regulatory region of Bix4 contains sites that bind VegT as well as Xbra (Tada et al., 1998).

In this paper, we first show that expression of Bix4 requires maternal VegT and that VegT is a potent activator both of endogenous Bix4 and of reporter constructs driven by the Bix4 upstream regulatory region. Two T-box binding sites contained within the Bix4 promoter prove to be both necessary and sufficient for mesodermal and endodermal expression of reporter genes in transgenic Xenopus embryos and, remarkably, a single T-box binding site is able to act as a mesoderm-specific enhancer when placed upstream of a minimal promoter.

These results show that Bix4 is a direct target of VegT and they raise the question of whether Bix4 mediates the effects of VegT. Our results show that Bix4 rescues the formation of endodermal markers in embryos in which VegT transcripts have been ablated but does not restore the ability of vegetal pole blastomeres to induce mesoderm. These results demonstrate that Bix4 acts directly downstream of VegT to specify endodermal differentiation in Xenopus embryos but that rescue of mesoderm induction requires additional VegT targets.

Xenopus embryos were obtained and fertilised as described (Smith, 1993) and staged according to Nieuwkoop and Faber (1975).

Antisense ablation of maternal VegT RNA was performed as described (Zhang et al., 1998). RNAase protection was carried out as described (Tada et al., 1998). A Bix4 probe was made by linearising pBix4 with EcoRV and transcribing with T7 RNA polymerase. An Xwnt11 probe was made from the plasmid pPCR121, isolated from the subtracted cDNA library previously described (Tada et al., 1998). The plasmid was linearised with HinfI and transcribed with T3 RNA polymerase. Other probes, including Xwnt8 (Christian et al., 1991), Xbra (Smith et al., 1991), ornithine decarboxylase (Isaacs et al., 1992), Xvent1 (Gawantka et al., 1995), Xsox17α (Hudson et al., 1997), muscle-specific actin (Mohun et al., 1984), αT4-globin (Walmsley et al., 1994), IFABP (Shi and Hayes, 1994) and endodermin (Sasai et al., 1996) were as described (Tada et al., 1998).

A Green Fluorescent Protein cDNA (Zernicka-Goetz et al., 1997) was cloned into the HindIII/XbaI sites of pGL3-basic (Promega). The Bix4 promoter (Tada et al., 1998) was then cloned into the NheI and HindIII sites of this construct and of pGL3-basic to generate GFP and luciferase reporter gene constructs. Point mutations were generated with the Stratagene QuickChange site-directed mutagenesis kit. The eFGF and Bix4 T-box binding sites (see Fig. 3A) were cloned as annealed oligonucleotides into the XbaI or SalI sites of pA48.pBLCAT3T. All constructs were sequenced. Luciferase assays used the Promega Dual Luciferase kit. Five animal caps were lysed in 50 μl of passive lysis buffer (Promega) and 5 or 10 μl were assayed for luminescence.

Transgenic Xenopus embryos were generated as described (Kroll and Amaya, 1996). Typically, 5-35 copies of the reporter gene are integrated into the host genome, with 2-6 copies at a single site (Kroll and Amaya, 1996). In situ hybridisations were as described (Tada et al., 1998). To make in situ hybridisation probes, GFP and luciferase plasmids were linearised with NcoI and CAT plasmids were linearised with SalI or PstI.

Expression of Bix4 was analysed in embryos in which maternal VegT transcripts were depleted by injection of antisense oligonucleotides into the oocyte. Previous work indicates that endodermin, a pan-endodermal marker (Sasai et al., 1996), Xsox17α, an endoderm-specific transcription factor (Hudson et al., 1997), and IFABP, a marker of small intestine (Shi and Hayes, 1994) are not expressed in such embryos, and that expression of mesodermal markers is delayed (Zhang et al., 1998). Fig. 1 shows that Bix4 (and Bix1, not shown) is also undetectable in VegT-depleted embryos, indicating that maternal VegT is essential for initiation of transcription of this gene. Consistent with this observation, VegT and (to some extent) the related T-box gene Xbra, as well as a combination of the two, induce expression of Bix4 in animal caps (Fig. 2A). They also activate expression in animal caps of a luciferase reporter gene driven by 1.6 kb of sequence upstream of the Bix4 transcription start site (Fig. 2B and Tada et al., 1998).

Although Xbra and VegT are both eventually expressed in maternal VegT-depleted embryos (Fig. 1 and Zhang et al., 1998), their activation at stage 14 is apparently too late to induce Bix4; previous work has demonstrated that blastomeres lose the ability to respond to Xbra by stage 13 (Tada et al., 1997).

The Bix4 regulatory region contains two 10 bp T-box binding sites within 70 bp of the transcription start site, and a related sequence is positioned a further 15 bp upstream (Fig. 3A and Tada et al., 1998). We refer to these sites as ‘distal’, ‘middle’ and ‘proximal’ (Td, Tm and Tp). In vitro binding assays indicate that both Xbra and VegT bind to the middle site, while only Xbra binds strongly to the proximal site (Tada et al., 1998). Interaction between VegT and the proximal site can be detected, however, with increased levels of MgCl2 and VegT protein in the binding reaction (not shown).

To ask whether VegT activates Bix4 expression through direct interaction with these sites, we made transgenic Xenopus embryos in which the 1.6 kb Bix4 upstream regulatory sequence was used to drive expression of reporter genes. Analysis was carried out at the early gastrula stage, and three reporter genes were used: Firefly Luciferase (LUC), Green Fluorescent Protein (GFP) and chloramphenicol acetyltransferase (CAT). GFP fluorescence is difficult to visualise in gastrula-stage embryos due to yolk autofluorescence, so the spatial expression of this and other reporters was analysed by in situ hybridisation.

Transgenic Xenopus embryos carrying the Bix4-GFP or Bix4-LUC construct expressed reporter genes in both mesoderm and endoderm, in a pattern resembling that of endogenous Bix4 (Fig. 3B; Table 1). Interestingly, vegetal expression of the Bix4-GFP construct was higher than that of Bix4-LUC. This may reflect a difference in RNA stability or in the ability of the in situ probe to penetrate endoderm. Since here we are particularly interested in identifying elements required for endodermal expression of Bix4, we have concentrated on use of the Bix4-GFP construct.

To investigate directly the roles of the T-box binding sites in the regulation of Bix4, they were mutated individually and these constructs were tested for the ability to drive expression of GFP in transgenic embryos (Fig. 3B; Table 1). Mutation of the middle T-box site (Tm) resulted in the most striking change in expression pattern; transgenic embryos containing this reporter construct lacked expression in the endoderm even after prolonged staining. In contrast, mutation of the distal or proximal sites (Td or Tp) did not affect the strong endodermal expression. In addition, we note that all three mutations caused a decrease in intensity but an increase in extent of reporter gene expression in the marginal zone, that mutations in the Tp site frequently caused higher levels of reporter gene expression on the dorsal side of the embryo, and that mutations in the Td site caused elevated levels of reporter gene expression in the vegetal hemisphere.

These experiments demonstrated that the Tm site is necessary for expression of Bix4 reporter genes in the endoderm, but the functions of the Tp and Td sites were harder to define. In an effort to elucidate their roles, double mutations with the Tm site were made (Fig. 3B). Embryos with the Td/Tm double mutation construct exhibited low, diffuse reporter gene expression; there is little or no expression in the endoderm and weak expression throughout the ectoderm and marginal zone. Embryos with the Tm/Tp double mutation exhibit complete loss of reporter gene expression.

Together, our results indicate that the middle (Tm) site, to which both VegT and Xbra bind, is required for endodermal expression and for strong expression in the mesoderm. The proximal (Tp) site, which Xbra binds strongly and VegT only weakly, is required together with the middle site for expression in the mesoderm. Finally, since mutation of the Td site leads to higher and more diffuse expression of reporter genes, this distal site appears to be required to restrict expression to the correct cells and to repress levels of expression. All three T-box binding sites are necessary for correct expression of Bix4 in the mesoderm and endoderm of the early gastrula.

To address the effects of mutating the T-box sites in a more quantitative fashion, we studied the ability of VegT to activate wild-type and mutated Bix4-LUC constructs in animal caps (Fig. 4). Consistent with our observations in transgenic embryos, mutation of the middle site (Tm) essentially abolishes luciferase induction by VegT, while mutation of the proximal site (Tp) has little effect. Most significantly, mutation of the distal site (Td) causes a four-fold increase in basal reporter gene expression while retaining significant VegT inducibility. These results suggest that the levels of staining in transgenic embryos are related to the levels of expression of Bix4 reporter constructs. They also confirm that the distal T-box binding site (Td) plays a role in repressing expression of Bix4.

The importance of the T-box binding sites in the expression of Bix4 during early Xenopus development was emphasised in experiments in which a short sequence from the Bix4 promoter containing the Tm and Tp sites was placed upstream of the minimal cytoskeletal actin (CSKA) promoter driving expression of CAT. This element proved to be sufficient to drive expression of CAT in both mesoderm and endoderm (Fig. 5). Mutation of either site caused loss of expression in the endoderm and, in the case of mutations in the proximal site (Tp), a decrease in mesodermal expression. Mutation of both sites resulted in loss of specific expression in both germ layers. These results suggest, remarkably, that a single T-box site is sufficient to drive mesoderm-specific expression of reporter constructs. To verify this idea, the 10 bp T-box site identified in the eFGF promoter (Casey et al., 1998) was placed upstream of the CSKA or thymidine kinase (TK) minimal promoters and these constructs were used to make transgenic Xenopus embryos. In both cases, expression of the CAT reporter gene during gastrula stages was confined to the mesoderm (Fig. 5; data not shown for the TK minimal promoter). Together, these experiments demonstrate that T-box sites are sufficient, as well as necessary, for reporter gene expression in mesoderm and endoderm. Constructs such as those described above might be used to drive specific expression of dominant-negative or other genes in the mesoderm and endoderm, and we are presently testing T-box sites to see if they function as mesodermal or endodermal enhancers in other organisms.

The results described above indicate that Bix4 is a direct target of VegT and they raise the question of whether Bix4 mediates the effects of VegT. Preliminary experiments demonstrated that Bix4, like Bix1 (Tada et al., 1998) and members of the related Mix family (Ecochard et al., 1998; Henry and Melton, 1998; Latinkic and Smith, 1999; Rosa, 1989; Vize, 1996), can activate endodermal markers such as endodermin, Xsox17α and IFABP in animal caps, as well as the late ventral mesodermal marker αT4-globin (Fig. 6A). Bix4 also, in a dose-dependent manner, restored expression of Xsox17α, endodermin and Xbra to whole embryos depleted of maternal VegT RNA (Fig. 6B and data not shown), with lower doses tending to rescue Xbra and higher doses rescuing Xsox17α. We note, however, that this rescue is both delayed and incomplete; elevated expression of both Xbra and Sox17α in response to Bix4 is only detectable by stage 12.5. This indicates that Bix4 cannot be the sole mediator of the effects of VegT. Consistent with this conclusion, Bix4 cannot effect morphological rescue of VegT-depleted embryos (data not shown), and, unlike VegT itself, it does not restore to VegT-depleted vegetal pole regions the ability to induce mesoderm. Thus, wild-type animal caps do not elongate when juxtaposed with VegT-depleted vegetal bases expressing Bix4 and muscle actin is not expressed in these conjugates (Fig. 7A,B). This indicates that rescue of Xbra expression by Bix4 (Fig. 6B) is a cell-autonomous effect of Bix4 rather than an indirect effect mediated by vegetal pole blastomeres.

The results described in this paper demonstrate that the paired-like homeobox gene Bix4 is a direct target of the T-box transcription factor VegT. Depletion of maternal VegT transcripts prevents Bix4 expression and VegT proves to be a potent activator both of endogenous Bix4 and of reporter constructs driven by the Bix4 promoter. Analysis of transgenic Xenopus embryos expressing Bix4 reporter constructs shows that the Tm site (Fig. 3A), to which VegT binds strongly, is necessary for expression in the endoderm, while two other sites in the promoter (Td and Tp) are also required for correct spatial expression at the early gastrula stage.

Additional experiments show that a short sequence from the Bix4 promoter containing the Tm and Tp sites, when placed upstream of a minimal promoter, is sufficient to drive expression of reporter constructs in both mesoderm and endoderm. Furthermore, a single T-box binding site derived from the eFGF promoter, comprising just 10 nucleotides, is sufficient to drive expression of reporter constructs exclusively in the mesoderm. Finally, we observe that Bix4 can restore, to some extent, the expression of endodermal and mesodermal markers in VegT-depleted embryos but cannot rescue the ability of the vegetal pole to induce mesoderm. These and other issues are discussed below.

The four Bix genes were isolated in a screen for targets of Xbra, although it is clear from our previous work (Tada et al., 1998) and from work described here that they are also activated by other T-box gene products, including VegT. In this connection, we note that expression of Bix1 is induced equally strongly by Xbra and VegT (Tada et al., 1998), while Bix4 is much more efficiently activated by VegT (Fig. 2A). This indicates that Xbra plays a more significant role in the regulation of Bix1 than in Bix4 and, consistent with this suggestion, we observe that mesodermal expression of Bix1 is higher than that of Bix4 (M. T., unpublished observations). A comparison of the regulatory regions of Bix1 and Bix4 may therefore shed light on T-box gene specificity.

Analysis of the Bix4 promoter has identified three 10 base-pair motifs that resemble the Tsite in the upstream regulatory region of the Xbra target eFGF (Casey et al., 1998). These motifs, Tm, Tp and Td, differ from the eFGF motif with respect to 1, 2 and 3 base pairs, respectively (Fig. 3A). VegT appears to have different affinities for each site (ESC, unpublished) and each site has a distinct role in the regulation of Bix4 expression. The site with the highest affinity for VegT, Tm, is required for expression of reporter genes in the endoderm (Fig. 3B) and for activation of reporter genes in response to VegT (Fig. 4), while the lower affinity Tp site is necessary for normal expression in the mesoderm (Fig. 3B). Mutation of the latter site does not affect the response of Bix4 to VegT (Fig. 4).

The Td site, which differs from the consensus T site with respect to three nucleotides, does not appear to bind VegT or Xbra (E. S. C., unpublished observations) and mutation of this site suggests that it is required to restrict or repress expression of Bix4. One possibility, therefore, is that the Td site binds another member of the T-box family which acts as a transcriptional repressor and, indeed, Tbx2 has recently been shown to repress the melanocyte-specific TRP-1 promoter (Carreira et al., 1998). We note, however, that the CACCT sequence contained within all the T sites is also the binding site for a family of two-handed zinc finger/homeodomain proteins which includes δEF1 (Higashi et al., 1997; Takagi et al., 1998), Zfh-1 (Lai et al., 1993) and, most recently, SIP1 (Verschueren et al., 1999). Members of this family act as transcriptional repressors, and it will be of interest to discover whether such proteins are expressed during early Xenopus development.

To simplify analysis of the roles of the T-box binding sites in the regulation of Bix4 expression, we have placed them upstream of a minimal promoter. Our results demonstrate that one T-box site linked to a minimal promoter, whether it be Tp, Tm, or the site derived from the eFGF promoter, is sufficient for activation of a reporter gene in the mesoderm, while the combined action of the Tm and Tp sites results in expression in the endoderm (Figs 3B, 5).

These observations provide no simple rationalisation of a model that proposes that endoderm and mesoderm are specified by a gradient of VegT, with high concentrations of VegT specifying endoderm and low concentrations mesoderm (Kimelman and Griffin, 1998; Stennard, 1998; Zhang et al., 1998). Indeed, from the model one might predict the opposite of what we observe – that is, a reporter driven by a single T-box site should be expressed more strongly in the endoderm, where levels of VegT are highest, than in the mesoderm.

One interpretation of the inability of a single T-box binding site to drive expression of reporter genes in the endoderm is that the vegetal hemisphere contains repressor molecules that compete with VegT for the T-box site. As discussed above, these might include T-box family members such as Tbx2 (Carreira et al., 1998) or members of the δEF1 family (Higashi et al., 1997; Lai et al., 1993; Takagi et al., 1998; Verschueren et al., 1999). According to this idea, the ability of two T-box binding sites to drive expression in the endoderm might be due to more efficient binding of VegT to two sites than to one. An alternative view is that the ability of a single T-box binding site to drive expression of reporter genes in the mesoderm and not the endoderm is that equatorially expressed T-box genes, such as Xbra and eomesodermin (Ryan et al., 1996), also play a role in the activation of Bix4 through this site and that their gene products are more efficient activators than VegT. This is under investigation.

Maternal VegT is required for proper formation of both the endoderm and the mesoderm of the Xenopus embryo (Zhang et al., 1998). Our data demonstrate that, in rescuing endoderm formation in VegT-depleted embryos, Bix4 can mediate at least some of the effects of VegT. The fact that Bix4 is unable to restore mesoderm-inducing activity to vegetal pole regions derived from VegT-depleted embryos, however, suggests that there are additional VegT targets, which may include other members of the Bix and Mix families and even the inducing factors themselves. In the future, we plan to isolate such targets, investigating first those genes that restore mesoderm-inducing activity to VegT-depleted vegetal pole regions. Candidates under investigation include the nodal-related genes (Jones et al., 1995; Joseph and Melton, 1997; Smith et al., 1995) and derrière (Sun et al., 1999). We also plan to investigate the extent to which inducing factors such as these also contribute to the regulation of Bix gene expression (Tada et al., 1998).

We are grateful to Sara Mercurio, Tristan Rodriguez and Josh Brickman for critical analysis of this manuscript. This work was supported by the Medical Research Council, the Human Frontiers Science Programme, the National Institutes of Health (RO1-HD38272) and the Harrison Fund. E. S. C. was the recipient of a Hitchings-Elion Fellowship. We thank Enrique Amaya for advice on making transgenic Xenopus embryos.