The T-box transcription factor Brachyury regulates expression of eFGF through binding to a non-palindromic response element

The mesoderm of Xenopus laevis is formed through an inductive interaction in which cells of the vegetal hemisphere of the embryo act on overlying equatorial cells (Sive, 1993; Slack, 1994). Members of the FGF and the transforming growth factor type β (TGF-β) families, such as eFGF and activin, can mimic the effect of vegetal tissue, and several lines of evidence suggest that they are indeed involved in mesoderm induction (Slack, 1994; Dyson and Gurdon, 1996). One of the genes activated in an immediate-early fashion by these factors, as well as by vegetal pole blastomeres, is the T-box family member Brachyury, a sequence-specific transcription activator required for formation of posterior mesoderm and notochord (Smith et al., 1991; Kispert and Herrmann, 1993; Herrmann and Kispert, 1994; Kispert et al., 1995; Smith, 1997).

Xenopus Brachyury (Xbra) is expressed throughout the mesoderm of Xenopus embryos at the early gastrula stage (Smith et al., 1991) and misexpression of the gene in prospective ectodermal tissue causes the formation of ectopic mesoderm (Cunliffe and Smith, 1992). Interference with Xbra function, by overexpression of a dominant-negative construct in which the activation domain of the protein is replaced by the repressor domain of Drosophila engrailed, results in loss of notochord and posterior structures (Conlon et al., 1996). This phenotype resembles that of mouse and zebrafish embryos with mutations in the Brachyury gene (Herrmann et al., 1990; Halpern et al., 1993; Schulte-Merker et al., 1994). Xbra is therefore both necessary and sufficient for normal formation of mesoderm, and identification of its primary gene targets is essential for an understanding of how mesoderm-inducing signals are interpreted and transmitted.

One potential target of Xbra is embryonic FGF (eFGF). The expression patterns of these two genes are similar (Isaacs et al., 1995), and they are components of an indirect autoregulatory loop in which Xbra activates expression of eFGF and eFGF, via the MAP kinase pathway (Gotoh et al., 1995; LaBonne et al., 1995; Umbhauer et al., 1995), maintains expression of Xbra (Isaacs et al., 1994; Schulte-Merker and Smith, 1995).

In this paper, we first show that inhibition of Xbra function disrupts expression of eFGF in notochord and dorsal mesoderm, suggesting that the autoregulatory loop functions primarily in these tissues. We then use a hormone-inducible Xbra construct (Tada et al., 1997) to demonstrate that Xbra is able to activate expression of eFGF in dispersed cells, in the absence of protein synthesis, and in the presence of a truncated FGF receptor, suggesting that its effects are direct. We go on to show that the eFGF promoter contains an Xbra response element which comprises half of the 20 bp palindromic Brachyury-binding sequence previously identified by a PCR-based selection procedure (Kispert et al., 1995). This half-site is approximately 1 kb upstream of the transcription start site and is also present in the same position in the upstream regulatory regions of the mouse and human homologues of eFGF. Gel-shift analysis demonstrates that this sequence is specifically bound by monomers of Xbra and experiments in Xenopus oocytes show that it is sufficient for Xbra-dependent reporter gene expression. Additional experiments confirm that the marginal zone of the Xenopus embryo contains factors that interact with the Brachyury half-site. eFGF is thus the first-identified target of a T-box protein and our data show that Brachyury can function by binding to a non-palindromic sequence.

Fertilisation, culture and microinjection of Xenopus embryos were as described (Smith, 1993). Embryos were staged according to Nieuwkoop and Faber (1975). Animal caps were dissected as described (Smith, 1993) and cultured in 75% NAM (Slack, 1984), or in 75% NAM containing 0.1% bovine serum albumin if Xenopus FGF-2 was included in the medium. Dexamethasone was dissolved in ethanol at 2 mM and diluted to a working concentration of 10⁻⁶ M in 75% NAM.

To identify eFGF regulatory sequences, Xenopus genomic DNA was digested with PstI, fractionated by agarose gel electrophoresis and a Southern blot was probed with a 293 bp XeFGF(i) PCR fragment (Isaacs et al., 1992). Size fractions of approximately 3.5 kb were then cloned into pBluescript SK+ (Stratagene) and screened with the same probe. Five clones were obtained, four of 3 kb and one of 3.5 kb. All five contained the first exon of eFGF(i) and approximately 2.5 kb of sequence upstream of the translation start site. The larger size of the 3.5 kb clone was due to an extra copy of a 500 base pair repeat, which was located in the upstream region (Fig. 3A) Automated sequencing was carried out using a series of exonuclease deletions (Henikoff, 1990). The eFGF transcription start site was identified by 5′ RACE using total RNA from embryos and nested primers AACTCACGACTCCAACTTCCACTG and TCCCCCATGGACC-TGGAAAAGAG (Marathon cDNA Amplification Kit, Clonetech). pSP64T-Xbra (Cunliffe and Smith, 1992), XFD (Amaya et al.,1991), Xbra-GR (Tada et al., 1997), Xbra-En^R (Conlon et al., 1996) and En^R (Conlon et al., 1996) are as described. Xbra▵AD contains amino acids 1-303 of Xbra cloned into pSP64T and Xbra▵DBD was generated by religation of a SacI digest of pSP64T-Xbra which deletes amino acids 26 to 56. RNA was transcribed from these plasmids as described (Smith, 1993). All constructs were sequenced and proteins of the correct size were produced after in vitro translation.

To generate Brachyury half-site reporter constructs, oligonucleotides containing the distal half-site in the eFGF regulatory region were annealed to produce XbaI compatible ends, and cloned into a CAT reporter construct (pA48.pBLCAT3T) which contains 50 base pairs of the cytoskeletal actin (CSKA) promoter in which the serum response element (SRE) has been replaced by an XbaI site (Mohun et al., 1987, and Tim Mohun, personal communication). The sequence of the Xenopus half-site is CTAGGTTTTCACACCTAT. The mouse half-site was CTAGCATTCACACCTAGGAG. The following oligonucleotide (Xen mut), in which the italicised bases were mutated, was used as a negative control: CTAGGTTTGCACGCTTAT.

All eFGF-CAT reporter constructs were generated by cloning BamHI-XhoI fragments into pBLCAT3T (Modak et al., 1993). Proximal and distal deletions of the eFGF upstream regulatory region were generated by PCR and internal deletions were generated by mutagenesis (Quick Change kit, Stratagene). Details are available on request. To generate −2250▵B, the Brachyury half-site 936 nucleotides upstream of the transcription start site was replaced by an EcoRI site. To generate proximal deletions the related site in the transcribed 5′ untranslated region was removed by insertion of an XhoI site directly upstream.

Cell dispersal and protein synthesis inhibition experiments were carried out as described (Smith et al., 1991), except that cycloheximide was applied continuously at a concentration of 10⁻⁵ M.

RNA extraction and RNAase protection analyses were essentially as described (Smith, 1993). Samples were analysed with probes specific for ornithine decarboxylase (ODC) (Isaacs et al., 1992), XeFGF(i) (Isaacs et al., 1992), and the 3′ untranslated region of Xbra (Smith et al., 1991). RT-PCR was performed using Xbra and EF-1α primers described in the Xenopus Molecular Marker Resource (http://vize222.zo.utexas.edu/). eFGF RT-PCR primers were: CTTTCTTTCCACACAAACGACACCG and AACTCACGACT-CCAACTTCCACTG.

Xenopus ovaries were disassociated by treatment with collagenase and healthy stage VI oocytes were maintained in OR2 (Kay, 1991). Test oocytes were injected with 2.5-5 ng RNA and incubated at 18°C until the following day when circular DNA reporter constructs (1 ng) were injected into the germinal vesicle. A control group for each reporter construct consisted of injection of DNA alone. Oocytes were cultured for a further 20 hours at 18°C and then assayed for CAT activity by thin layer chromatography. Per cent acetylation was measured using a Molecular Dynamics phosphorimager.

One A or C tier cell of Xenopus embryos at the 32-cell stage was injected with 20 pg of reporter construct together with 10 pg of pRL-CMV (Promega) in which Renilla luciferase is driven by the CMV promoter. Groups of 10 embryos were harvested at stage 12 and they were lysed in 50 μl reporter lysis buffer (Promega). 40 μl of extract was used to assay for CAT activity and 5 μl of extract for Renilla luciferase activity. CAT activities were normalised by reference to luciferase measurements.

The DNA-binding domain of Xbra (XbraDBD), comprising amino acids 1-234, was cloned into the vector pET-22B (Novagen), which introduces a C-terminal His-tag, and expressed in E. coli. Purified XbraDBD (0.05 pmole) was incubated for 20 minutes at room temperature with 0.25 ng labelled probe in 75 mM KCl, 0.25 mM EDTA, 10 mg/ml BSA, 1 mM DTT, 0.1 % NP-40, 1 mM MgCl₂, 1 mM PMSF, 10% glycerol, 25 mM Hepes pH 7.0. Some reactions included a 100-fold molar excess of unlabelled probe as a control. The distal Brachyury half-site sequence (→) was GATCCCTCGGT-TTTCACACCTATAGACCTCGA, the mutated Brachyury half-site sequence (Mut) was GATCCCTCGGTTTGCACGCTTATAGA-CCTCGA, the palindromic Brachyury sequence (→←) was ATTAGTCACACCTAGGTGTGAAGAGCC and the proximal Brachyury-like half-site sequence was GATCGAGAGCAAACCA-CACCTCCTCTGGAGCT. Super-shift assays were carried out by incubating DNA/protein complexes with rabbit preimmune (PI) serum or Xbra (α-X) polyclonal antibody for 10 minutes on ice.

The expression pattern of eFGF closely resembles that of Xbra (Isaacs et al., 1995; Fig. 1A,D) and embryological experiments suggest that the two genes are components of an indirect autoregulatory loop (Isaacs et al., 1994; Schulte-Merker and Smith, 1995). Consistent with this idea, inhibition of Xbra function in the early Xenopus embryo by injection of RNA encoding Xbra-En^R inhibits dorsal and notochord-specific expression both of Xbra itself (Conlon et al., 1996; Fig. 1E-G) and of eFGF (Fig. 1B,C,G). We note that expression of Xbra in embryos injected with RNA encoding Xbra-En^R extends more dorsally than expression of eFGF (Fig. 1B,C,E,F). It is possible that Xbra expression is maintained in these dorsolateral cells in response to eFGF derived from neighbouring lateral cells. The decrease in eFGF expression is also detectable by RNAase protection analysis (Fig. 1G). Xbra-En^R likewise inhibits expression of both Xbra and eFGF in Xenopus animal caps in response to FGF (Fig. 1H). Xbra function is, therefore, required for maintenance of eFGF expression both in dorsal mesoderm and in FGF-treated animal pole tissue.

Activation of Xbra by FGF occurs through the MAP kinase pathway (Gotoh et al., 1995; LaBonne et al., 1995; Umbhauer et al., 1995) and does not require cell-cell communication or protein synthesis (Smith et al., 1991). We have used the hormone-inducible Xbra construct Xbra-GR (Tada et al., 1997) to ask whether, according to the same criteria, activation of eFGF by Xbra is direct. Our results show that Xbra-GR does induce expression of eFGF in dispersed cells (Fig. 2A), and in the presence both of cycloheximide (Fig. 2B) and of a dominant-negative FGF receptor (Fig. 2C). Thus, induction of eFGF by Xbra is cell autonomous, does not involve synthesis of an intermediate transcription factor and does not require an intact FGF signalling pathway. In this last respect, induction of eFGF by Xbra contrasts with autoinduction of Xbra, which is inhibited by cell dispersion and by a dominant-negative FGF receptor (Tada et al., 1997) and presumably occurs via eFGF.

The data above indicate that Xbra activates expression of eFGF directly. To investigate this question, we isolated 2.5 kb of the upstream regulatory region of eFGF (Fig. 3A). Sequencing revealed a single 10 base pair element TTTCACACCT located 936 nucleotides upstream of the transcription start site. This sequence is identical to half of the 20 base pair palindromic Brachyury site previously identified (Kispert and Herrmann, 1993). A related sequence, AACCACACCT, is located 123 nucleotides downstream of the transcription start site.

Previous reports have suggested that Brachyury does not bind to a half-palindrome (Kispert and Herrmann, 1993), and that two half-palindromes, appropriately spaced, are required for transcription activation (Kispert et al., 1995). We note, however, that the 5′ regulatory regions of mouse and human FGF-4, to which eFGF is closely related (Isaacs et al., 1992), also contain a single Brachyury half-site within about 1 kb of their transcription start sites (Fig. 3B). This conservation suggests that the sequence is involved in regulation of eFGF/FGF-4 expression. To ask whether Xbra can recognise this site, a 32 base pair probe containing the upstream Brachyury half-site and flanking sequence was used in an electrophoretic mobility shift assay. The probe was incubated with a truncated Xbra protein (XbraDBD) purified from E. coli, which contains 234 amino acids and includes the entire DNA-binding domain. Competition studies and use of an anti-Xbra antibody demonstrate that Xbra binds specifically to the Brachyury half-site as well as to the complete palindrome (Fig. 4A). The mobility of the Xbra/half-site complex was greater than that of the Xbra/palindrome complex, indicating that Xbra binds to the half-site as a monomer (see below). Similar results were obtained with a truncated form of Xbra lacking the last 129 amino acids. Recombinant full-length Xbra could not be obtained in soluble form, and material obtained by translation in reticulocyte lysates showed only weak binding to the half-site. This binding was stabilised by addition of Xbra antibody (data not shown).

The ability of Xbra to activate transcription through binding to the half-palindrome described above was assessed by introducing the site into a plasmid (pA48.pBLCAT3T) in which a chloramphenicol acetyl transferase (CAT) reporter gene is driven by a cytoskeletal actin (CSKA) minimal promoter and assaying for CAT expression in oocytes. Xenopus oocytes received cytoplasmic injections of 2 ng of RNA encoding Xbra followed by nuclear injections of the reporter construct. They were assayed for CAT activity 20 hours later. A single copy of the half site from Xenopus or mouse was sufficient to drive strong CAT activity in the presence of Xbra (Fig. 4B). This activity was abolished when three bases in the half-site were mutated, demonstrating that Xbra must interact with the half-site to drive expression. Little or no activation was observed with forms of Xbra lacking either the C-terminal activation domain (Xbra▵AD) or a 30 amino acid internal deletion of the N-terminal DNA-binding domain (Xbra▵DBD) (Fig. 4C). Simultaneous expression of these two constructs did not increase activation (Fig. 4C), consistent with the idea that Xbra does not form dimers on the Brachyury half-site; formation of dimers would bring together a functional DNA-binding domain with a functional activation domain, and the two together should activate transcription in a synergistic manner.

To investigate whether one or both of the half-palindromes described above is required for Xbra-dependent expression of the intact eFGF promoter, we linked 2.5 kb of the eFGF upstream regulatory region to a CAT reporter gene (−2250.CAT) and made a series of deletion constructs. Using the oocyte assay, injection of 4 ng RNA encoding Xbra resulted in approximately 10-fold induction of CAT activity of the −2250.CAT construct, while RNA encoding Xbra▵AD caused only a 2-fold increase (Fig. 5). We note that the level of activation observed with the −2250.CAT construct was low compared with that obtained using pA48.pBLCAT3T containing the distal eFGF Brachyury half-site (Fig. 4B). This may reflect the low level of expression of eFGF in the Xenopus embryo.

Deletion analysis revealed that 1 kb of eFGF upstream sequence can be removed with very little effect on Xbra-dependent activation but that removal of additional sequences between −1050 and −970, wherein lies the distal Brachyury half-site, reduced induction by almost 80%. Replacement of this distal site with an EcoRI site yielded similar results (Fig. 5). A proximal deletion of 120 bp, which removes the downstream Brachyury-like half site, also decreased activation by 80% although, under the conditions described, we cannot demonstrate binding of Xbra to this site in vitro (data not shown). These results indicate that both half-palindromes are required for full induction of the eFGF promoter by Xbra.

Attempts using transgenic Xenopus embryos (Kroll and Amaya, 1996; Amaya and Kroll, 1998) to study the requirement of the Brachyury half-sites for normal expression of eFGF did not meet with success, because expression of the −2250.CAT reporter in such experiments was very low (ESC and Walter Lerchner, unpublished observations). This weak reporter gene expression matches that of endogenous eFGF, whose transcripts are present at much lower levels than are those of Xbra (see Fig. 2C, lane 13) and, even when the Xbra promoter is used to drive expression of green fluorescent protein (GFP) in the Xenopus embryo, it is necessary to resort to in situ hybridization to detect GFP transcripts rather than to use direct observation of GFP fluorescence (Latinkic et al., 1997).

As an alternative approach to investigating whether the Brachyury half-sites in the eFGF 5′ regulatory region can respond to factors present in the marginal zone of the Xenopus embryo, we injected pA48.pBLCAT3T containing the distal Brachyury half-site, or a mutated version of it, into tier A (prospective ectoderm) or tier C (prospective mesoderm) of Xenopus embryos at the 32-cell stage. The plasmid pCMV-RL was injected at the same time to allow normalisation of measured CAT activities. Fig. 6 shows that the construct containing the wild-type Brachyury half-site, but not the mutated version, is expressed at significantly higher levels following injection in tier C than in tier A. This indicates that the marginal zone of the Xenopus embryo, but not the prospective ectoderm, expresses factors that interact with the distal Brachyury half-site in the eFGF 5′ regulatory region.

The work described in this paper demonstrates that activation of eFGF expression by Xbra is direct, and that this activation occurs by interaction of monomeric Xbra with two non-palindromic Brachyury-binding sites, one of which is also present in the regulatory regions of mouse and human FGF-4. The experiments identify the first transcriptional target of a T-box gene and are of particular interest because previous experiments have indicated that Brachyury functions as a dimer, either to inverted palindromes (Kispert and Herrmann, 1993), or to direct repeats (Kispert et al., 1995; Muller and Herrmann, 1997; Papapetrou et al., 1997).

Our experiments are consistent with the idea that Xbra and eFGF are components of an indirect autoregulatory loop in which each maintains expression of the other (Isaacs et al., 1994; Schulte-Merker and Smith, 1995). The data also indicate that this loop functions predominantly in notochord and dorsal mesoderm, because inhibition of Xbra function results in loss of expression of eFGF (Fig. 1B,C) and Xbra itself (Fig. 1E,F; Conlon et al., 1996) in these tissues. The same may be true in the mouse embryo, where there is no evidence for direct interactions between Brachyury and FGF family members in the primitive streak, but it remains possible that an autoregulatory loop functions in the notochord and head process (Schmidt et al., 1997; Smith, 1997).

Our data indicate that Xbra regulates expression of eFGF by binding to sites that correspond to half of the palindromic sequence originally identified by PCR-based binding site selection (Kispert and Herrmann, 1993). One of these sites is positioned approximately 1 kb upstream of the eFGF transcription start site. Xbra interacts with this sequence in electrophoretic mobility shift assays and through such an interaction is capable of activating transcription in Xenopus oocyte assays (Fig. 4). Furthermore, deletion of this site impairs significantly the ability of the eFGF regulatory region to respond to Xbra (Fig. 5). A related sequence is present within the 5′ untranslated region of eFGF and, although we cannot demonstrate that Xbra binds this site in vitro, its deletion also impairs Brachyury-dependent activation of eFGF reporter constructs (Fig. 5), suggesting that interaction with Xbra can occur in vivo. Further work is required to understand whether Xbra binds co-operatively to these two sites in the eFGF promoter, thus activating transcription in a synergistic manner, and whether the spacing of the two sites affects levels of expression. Our data, however, clearly demonstrate that a single half-palindrome is sufficient to elicit Brachyury-dependent transcription when positioned close to the transcription start site (Fig. 4). Although we have been unable to use transgenic embryos to study the requirement of the Brachyury half-sites for normal expression of eFGF, we have demonstrated that the marginal zone of the Xenopus embryo expresses factors that interact in a specific fashion with the distal Brachyury half-site (Fig. 6).

Finally, it is likely that other transcription factors are also involved in eFGF/FGF-4 regulation. First, it is clear that inhibition of Xbra function does not inhibit all aspects of eFGF expression (Fig. 1B,C,G). Second, other transcription factors such as Sox2 and Oct-3 have been shown to be involved in the regulation of FGF-4 in mouse ES cells (Yuan et al., 1995; Ambrosetti et al., 1997). The presence of the Brachyury half-palindrome in the regulatory regions of Xenopus, mouse and humans suggests that the regulation of eFGF/FGF-4 has been conserved, and we are now investigating whether members of the Sox and Oct families, or other transcription factors, co-operate with Brachyury in the regulation of eFGF expression.

This work is supported by the Medical Research Council, of which M.-A. O’R. was a training fellow. E. S. C. is supported by a Burroughs-Wellcome Hitchings-Elion fellowship and F. L. C. by the Howard Hughes Medical Institute, of which J. C. S. was an International Scholar. We are grateful to Tim Mohun for his CSKA-CAT construct, Norma Towers for teaching us CAT assays, Steve Smerdon for XbraDBD protein, Brenda Price and Lynne Fairclough for characterising the Xbra antiserum and Kathy Weston, Mike Jones, Larysa Pevny, Masa Tada and Josh Brickman for numerous helpful comments.