The Xenopus T-box gene, Antipodean, encodes a vegetally localised maternal mRNA and can trigger mesoderm formation

Two fundamental mechanisms are able to establish different regions within the developing embryo (reviewed in Gurdon, 1992). One involves the asymmetric distribution of determinants in the egg, whereby factors that are able to confer a specific developmental fate are inherited by only a subset of cells after the onset of cleavage. The other mechanism is that of embryonic induction, where an inducing cell is able to influence one or more nearby responding cells to change developmental fate. Mesoderm induction in Xenopus is a classical example of the latter, where a signal originating from the vegetal pole is believed to induce mesoderm in the equatorial region of the gastrula. As the initial stages of this induction occur well before the onset of zygotic transcription (Jones and Woodland, 1987), it therefore appears that mesoderm induction initially involves factors that are maternally inherited.

We have been interested in tracing the formation of mesoderm back through the sequential series of inductions arising from the earliest signal. We have chosen to search for novel genes that are activated by the TGF-β molecule, activin, whose likely developmental importance is indicated because: (1)activin very efficiently induces a wide variety of mesoderm derivatives in Xenopus ectodermal (animal cap) cells in vitro (Green et al., 1992); (2) the maternally provided activin protein is required for mesoderm and axis formation in fish in vivo (Wittbrodt and Rosa, 1994) and (3) a dominant negative activin receptor that abolishes TGF-β signalling blocks mesoderm induction (Hemmati-Brivanlou and Melton, 1992).

Our cloning strategy used the PCR-based technique known as differential display (Liang and Pardee, 1992) to compare the genes expressed in untreated or activin-treated animal cap cells. We have consequently cloned a novel gene belonging to the Xenopus T-box gene family that also includes Xbrachyury (Xbra) (Smith et al., 1991) and, most recently, Eomesodermin (Eomes).

Our results indicate that Apod is unique in this T-box gene family in that it is maternally expressed at a high level in addition to being zygotically expressed throughout the mesoderm. Most importantly, maternal transcripts are localised to the vegetal hemisphere from very early in oogenesis and we have therefore named it Antipodean (Apod). Since the vegetal pole gives rise to cells that are directly involved with normal mesoderm induction, Apod may act as an asymmetrically distributed determinant contributing to this process. Overexpression of Apod mRNA in ectoderm (animal cap) cells activates mesoderm formation as well as the expression of Xbra and Eomes, two key genes in mesoderm cell differentiation (Smith et al., 1991; Ryan et al., 1996). We suggest that Apod expression from maternal and zygotic Apod mRNA may make an important contribution to mesoderm formation in Xenopus.

Eggs were in vitro fertilised then dejellied and reared in 0.1× Modified Barth Saline (MBS) as described previously (Gurdon et al., 1985a). Embryos were staged according to Nieuwkoop and Faber (1967). Dissection and culturing of explants was performed in 1× MBS. Oocytes were defolliculated in 0.2% collagenase in 0.1 M sodium phosphate buffer at room temperature for approximately 2 hours.

Embryos for microinjection were transferred to 1× MBS; 4% Ficoll at the 2-cell stage and injected with a Drummond Nanoject variable microinjector. The injected volume was between 10 and 20 nl. At stage 7, embryos were transferred to 50% (0.1× MBS):50% (1× MBS, 4% Ficoll), then 30 minutes later transferred to 0.1× MBS.

2-cell stage Xenopus embryos were injected in the animal pole region with 20 pg activin mRNA. Animal caps were cut at stage 9 and cultured until sibling embryos had reached stage 10.25. Total RNA was extracted from activin mRNA-injected and uninjected animal caps by the NETS method.

Unless otherwise indicated, reagants used for differential display were components of an RNAimage kit (GenHunter). Total RNA (1μg) was treated with 1 U DNase (Gibco BRL) in a volume of 10 μl according to the manufacturer’s instructions. An aliquot (0.2 μg) was reverse transcribed in a 20 μl reaction (25 mM Tris-Cl, pH 8.3, 37.6 mM KCl, 1.5 mM MgCl₂ and 5 mM DTT) with 20 μM dNTPs, 0.2 μM anchored oligo(dT) primer (5′-AAGCTTTTTTTTTTTM-3′ where M is A, G or C) in a thermocycler: 65°C, 5 minutes; 37°C, 60 minutes; 75°C, 5 minutes. After 10 minutes at 37°C, 100 U MMLV Reverse Transcriptase was added to the reaction. Either 2 or 4 μl of the reverse transcription reaction was transferred to a 20 μl PCR reaction (10 mM Tris-Cl, pH 8.4, 50 mM KCl, 1.5 mM MgCl₂ and 0.001% gelatin) containing 1 U AmpliTaq DNA polymerase (Perkin-Elmer) and 0.25 μl α-[³³P]dATP (2,000 Ci/mmole) (DuPont-NEN), 20 μM dNTPs, 0.2 μM of the same anchored oligo(dT) primer used in the reverse transcription and 0.2 μM of a 10-mer of arbitrary sequence. Tubes were placed in a thermocycler: 94°C, 30 seconds; 40°C, 2 minutes; 72°C, 30 seconds for 40 cycles then 72°C for 5 minutes. Different combinations of primers were used to examine the entire mRNA population. An aliquot from the PCR reaction (3.5 μl) was electrophoresed through 6% denaturing polyacrylamide gels. The gel was blotted onto 3MM paper (Whatman) and dried at 80°C for 1 hour under vacuum. Gels were exposed to X-ray film at −80°C for 24 to 48 hours. Gels and autoradiographs were aligned and differentially expressed bands cut from the gel and the DNA eluted and reamplified. Fragments were gel purified before further use.

The nucleotide sequence of Apod was determined by Sanger dideoxy sequencing of overlapping double-stranded clones in RN3 using designed oligo primers and Sequenase version 2 (USB), according to the manufacturer’s instructions. Both strands of the Apod open reading frame were sequenced.

RNase protections were performed as described previously (Gurdon et al., 1985a). An Apod RNase protection probe was prepared by cloning a PCR amplified 280 bp fragment of the Apod 3′-UTR into SmaI restricted BlueScript. This construct (pApodRP/BS) was linearised with BamHI and transcribed with T7 RNA polymerase to produce a 360 bp unprotected probe that protects a 280 bp fragment of Apod mRNA. All other probes were as described previously: Xbra, gsc, FGFr, Xwnt-8 (Lemaire and Gurdon, 1994), Mix.1 (Lemaire et al., 1995), Siamois (Carnac et al., 1996) and Eomes (Ryan et al., 1996).

Synthetic capped mRNA was prepared as described in Lemaire et al. (1995). Synthetic Apod mRNA was prepared by cloning the 2562 bp Apod cDNA, containing the open reading frame and untranslated regions, into EcoRI/NotI-restricted pBluescript-RN3 (Lemaire et al., 1995). This plasmid (pApodR/RN3) was linearised with SfiI and transcribed with T3 RNA polymerase.

Activin mRNA was prepared from pSP64T Activin βB (a gift from Doug Melton) as previously described (Thomsen et al., 1990). BMP-4 mRNA was prepared from plasmid pSP64T-XBMP4II (a gift from Leslie Dale) as described by Dale et al. (1992). bFGF mRNA was made from plasmid containing bFGF in 64T(gift from Betsy Pownall) as described by Thompson and Slack (1992). Xnr-2 mRNA was prepared from plasmid Xnr-2 (a gift from Jim Smith) as described by Jones et al. (1995). bVg1 mRNA was prepared from plasmid pSP64TBVg1 (a gift from Doug Melton) as described by Thomsen and Melton (1993). Xwnt-8 mRNA was prepared from an RN3-based vector as described by Lemaire et al. (1995). Xbra mRNA was prepared from a construct called pSP64RN3XbraR (Lemaire unpublished), which was constructed with the coding region of Xbra from pSP73Xbra (a gift from J. Smith), cloned into pBluescipt-RN3 (Lemaire et al., 1995). Eomes was prepared from pEomes/RN3-3 as described in Ryan et al. (1996). XFD mRNA was prepared from the construct XFD/Xss (a gift from Enrique Amaya) as described in Amaya et al. (1991).

All samples were fixed in MEMFA (Hemmati-Brivanlou and Harland, 1989) for 2 hours, 30 minutes in methanol, then stored at −20°C. DIG-labelled riboprobes were synthesized essentially as described by Harland (1991). The construct pApodI/BS, consisting of a 1.3 kb PstI- EcoRI fragment covering the 3′-UTR and coding sequence of the C- terminal region of Apod cloned into Bluescribe, was cut with EcoRI and transcribed with T3 RNA polymerase to make antisense probes, or cut with HindIII and transcribed with T7 RNA polymerase to make sense probes. The Xbra probe was synthesised as described in Lemaire and Gurdon (1994). In situ hybridisation to sectioned material was performed as described previously (Lemaire and Gurdon, 1994), but without hydrolysing probes and using Boehringer BM purple substrate (Cat. No. 1442074) to visualise DIG-labelled hybrids. Whole-mount in situ hybridisation was performed essentially as described previously (Hemmati-Brivanlou et al., 1990) with the same method of visualisation as the whole-mount procedure. In some cases embryos were fixed overnight in Bouin’s fixative then bleached (70% methanol:30% hydrogen peroxide) to remove pigment.

Immunohistochemistry was performed as previously described (Carnac et al., 1996).

Partial cDNA clones from the 3′ ends of activin-induced mRNAs were identified using differential display and sequenced. Clones showing no similarity to sequences in the database were radioactively labelled and hybridised to northerns containing RNA from different developmental stages of Xenopus embryos. Of these, a 600 bp partial cDNA clone hybridised to an approximately 2.7 kb transcript that was highly expressed in both the egg and gastrula (data not shown). This was used to extract full-length clones from a stage 10 Xenopus embryo cDNA library. A 2562 bp cDNA with stops in all reading frames except one was isolated and sequenced (EMBL Accession number: X99905). Conceptual translation of the 1.3 kb open reading frame gave a 435 amino acid sequence, shown in Fig. 1A. The clone contained 17 bp of the 5′- and 1.24 kb of the 3′-untranslated regions.

A comparison of the deduced amino acid sequence of Apod with other proteins using Blast program analysis (Altschul et al., 1990) shows a region in the Apod N-terminal domain homologous to a T-box sequence motif (Bollag et al., 1994). Recently, mouse brachyury has been shown to act as a tissue-specific transcription factor with the T-box defining a DNA-binding domain (Kispert et al., 1995). Thus, the conservation of this motif suggests that the role of Apod in the embryo may be that of a transcription factor controlling the expression of developmentally regulated genes. There is no sequence homology at the amino acid or nucleotide level outside the T-box; Apod therefore encodes a new member of the T-box family of genes, that within Xenopus, includes Xbra (Smith et al., 1991) and most recently Eomes (Ryan et al., 1996). An amino acid sequence comparison between the Apod T-box and related T-boxes is shown in Fig. 1B. Within the T-box, Apod is most similar to the human and mouse homologs of Tbx2 (Bollag et al., 1994; Campbell et al., 1995) and to Drosophila optomotor-blind (omb) (Pflugfelder et al., 1992). It shows a weaker similarity to Xenopus Xbra (Smith et al., 1991) and to Eomes (Ryan et al., 1996). A diagrammatical representation of these proteins is shown in Fig. 1C, and demonstrates that the T-box’s position in Apod is most like that in Xbra.

Initial screening for activin response genes had shown that Apod mRNA was present at high levels in both the egg and gastrula (data not shown). To further characterise the temporal expression of this gene, we examined Apod mRNA levels throughout oogenesis and early embryonic development by RNase protection analysis. In parallel, we compared the Apod expression profile to that of Xbra, the expression of which has been previously analysed by Smith et al. (1991).

As shown in Fig. 2, Apod mRNA accumulates during stage 1 of oogenesis. It is then present during oogenesis from stage 2 and thereafter at similar levels. After fertilisation, levels of Apod mRNA again remain relatively constant until there is an increase at the onset of zygotic transcription. The peak of Apod expression starts between stages 9 and 10, just before the onset of gastrulation. This is much earlier than the peak of Xbra expression, which is beginning to reach maximal levels between stages 10.5 and 12. Similar to Xbra, Apod mRNA levels decline in the stage 13, late gastrula embryo. In stage 16 neurula embryos, expression of Apod and Xbra mRNA is noticeably reduced and is barely detectable in the tailbud embryo. Apod transcripts are therefore present continuously from the egg to late gastrula stages. Apod is expressed prior to, and during, the onset of Xbra mRNA expression. The expression of Apod mRNA overlaps with that of Eomes mRNA, which is also first expressed soon after the onset of zygotic transcription and prior to the onset of Xbra expression (Ryan et al., 1996).

In situ hybridisation with antisense DIG- labelled riboprobes was used to determine the localisation of Apod mRNA during oogenesis. Whole-mount in situ hybridisation reveals that maternal Apod mRNA is distributed throughout the stage 1 oocyte and begins its localisation during stage 2 of oogenesis (data not shown). Hybridisation to sectioned stage 3 oocytes (Fig. 3A) clearly demonstrates that Apod mRNA is deposited at both the vegetal cortex and in the vegetal yolk mass. In stage 4 oocytes (Fig. 3B), Apod mRNA has a predominantly vegetal localisation, and is still found at the vegetal cortex as well as in the vegetal yolk mass. Staining for Apod mRNA in these earliest stages is very granular in appearance, suggesting that the mRNA is aggregated into clumps. In mature stage 6 oocytes (Fig. 3C), there is a significantly reduced distribution of Apod mRNA at the cortex. It is predominantly in the vegetal yolk mass and is still particulate in its appearance. By this stage, we have noted that mRNA particles have moved towards the equator and are now in a more subequatorial position.

In situ hybridisation to egg sections shows that Apod mRNA has a similar distribution tothat observed in the mature oocyte; however, the pattern of staining is diffuse rather than particulate (Fig. 3D). Probes consisting of a labelled sense RNA strand which are unable to bind mRNA were used to detect non-specific staining in our protocol. Hybridisation with a sense Apod probe (Fig. 3E) to sectioned eggs shows no non-specific staining in the vegetal pole. A distribution of Apod mRNA similar to that found in the egg is maintained after fertilisation and is mainly confined to vegetal cells upon formation of the horizontal cleavage plane that forms at the 8-cell stage (data not shown).

The distribution of Apod throughout the embryo after the onset of zygotic transcription was determined by in situ hybridisation to either whole-mount or sectioned material. Both types of in situ hybridisation techniques were used so that any problems that might arise from poor penetration of the probe into yolky endodermal cells associated with whole-mount staining would be detected by use of sectioned material. Whole-mount and section in situ hybridisations with Apod antisense DIG-labelled riboprobes to stage 10.5 gastrulae are shown in Fig. 4A,C. These reveal mesodermal Apod mRNA expression throughout the equatorial region and a greater area of Apod mRNA- expressing cells on the dorsal side, compared to the ventral side, at this stage of development. Interestingly, Apod mRNA is expressed right up to the edge of the dorsal lip (Fig. 4C). Its expression may also extend into some of the future dorsal and ventral endoderm. In contrast, the whole-mount and sectioned material stained for Xbra mRNA (Fig. 4B,D, respectively) shows a less widespread distribution. The domain of Xbra mRNA expression is separated by several cell diameters from the dorsal lip and is entirely restricted to the mesoderm (Fig. 4D, see also Smith et al., 1991).

In late gastrula embryos, Apod mRNA is expressed pre-dominantly around the ventral blastopore (Fig. 4E) and extends into the posterior paraxial mesoderm (Fig. 4F). A particularly interesting feature of Apod expression is that, whereas Xbra at this stage is expressed around the blastopore and in the notochord (Smith et al., 1991), expression of Apod mRNA is excluded from the notochord (Fig. 4E,F) and thus presents a reciprocal expression pattern to that of Xbra.

In situ hybridisation shows that Apod mRNA is not expressed in the animal cap region of the embryo. To determine whether ectopic expression of Apod in the whole embryo can affect development, we injected Apod mRNA into the animal pole of 2-cell- stage embryos and allowed them to develop until sibling embryos had reached tailbud stage. Invagination failed to occur in these embryos after formation of the dorsal lip and they developed as exogastrulae (Fig. 5A). This defect also occurs upon injection of the mesoderm inducer, Xbra, into the animal pole region of the embryo (Cunliffe and Smith, 1992). To determine if injection of Apod mRNA into the animal pole of Xenopus embryos also induces ectopic mesoderm, we looked for expression of the mesodermal marker, Xbra, by in situ hybridisation. Expression of Xbra extends abnormally into the animal cap region of Apod mRNA-injected embryos (Fig. 5B) relative to controls (Fig. 5C), indicating that Apod mRNA is able to direct cells normally fated to become ectoderm towards a mesodermal fate.

To further characterise the ability of Apod mRNA to induce mesoderm, we injected different amounts of Apod mRNA into the animal pole region of 2-cell-stage embryos. Animal caps were dissected at stage 9, cultured until stage 10.25 and then analysed by RNase protection assay for the expression of mesodermal genes (see Fig. 5D). Injection of as little as 25 pg Apod mRNA is able to strongly induce ventral (Xwnt-8) (Christian et al., 1991) or pan mesodermal (Xbra, Eomes, Mix.1) genes (Smith et al., 1991; Ryan et al., 1996; Rosa, 1989). At higher concentrations of Apod mRNA, there is a weak activation of dorsal mesodermal markers such as gsc (see Dawid, 1994). Another dorsal gene, Siamois (Lemaire et al., 1995) is not activated at all.

Histological examination of animal cap explants injected with 25 pg Apod mRNA and cultured until stage 26 shows that, whereas uninjected controls form atypical epidermis (Fig. 5E), Apod-injected animal caps form mesoderm of ventral character as judged by presence of mesenchyme and vesicles (Fig. 5F). To further characterise the type of mesoderm induced by Apod, animal caps previously injected with increasing amounts of Apod mRNA were cultured until stage 26 and immunostained with the monoclonal antibody 12/101, which detects muscle cells. Doses from 25 to 500 pg of Apod mRNA do not induce muscle formation (Table 1). However, at a high dose (1 ng), there is a low percentage of explants (5 out of 27) that express the muscle marker in a small number of cells. Injections of higher concentrations of Apod mRNA are toxic.

Thus, Apod can induce ventral mesoderm and only at high doses can it weakly produce muscle at low frequencies. Xbra induces ventral-lateral mesoderm when expressed in animal caps, as characterised by vesicles, mesenchyme and muscle (Cunliffe and Smith, 1992). Eomes is able to induce more dorsal structures such as muscle and notochord (Ryan et al., 1996). These data demonstrate that different T-box genes may be involved with the production of different types of mesoderm.

It has been previously reported that small changes in activin concentration induce the expression of different mesodermal genes in Xenopus animal caps (Green et al., 1992). As Apod was identified on account of its induction in animal caps following treatment with activin, we decided to further characterise Apod’s response to this mesoderm inducer. Increasing amounts of activin mRNA were injected into the animal pole region of 2-cell embryos. Animal caps were dissected at stage 9, cultured until stage 10.25 and then analysed by RNase protection for Apod mRNA expression. We also determined the expression of genes induced at high or low doses of activin, namely the dorsal gene gsc and the pan-mesodermal gene Xbra respectively (Green et al., 1992). FGFr mRNA expression was used as an internal loading control. As shown in Fig. 6, Apod mRNA expression increases with increasing activin concentrations. Interestingly, this is different to the Xbra mRNA induction profile where levels of mRNA peak and then decline as the amount of injected activin mRNA increases (see also Green et al., 1992; Gurdon et al., 1994). Thus, Apod responds to activin at low doses like Xbra, but with a profile that is more similar to that of gsc in that its expression is not downregulated at high concentrations of activin.

Members of the TGF-β family of molecules with mesoderm-inducing capacity can be classified according to their abilities to dorsalise the mesoderm: activin, bVg1 and Xnr-2 can give rise to dorsal mesoderm tissues such as muscle and notochord, whereas Bmp-4 differentiates the mesoderm into ventral derivatives such as blood (Asashima et al., 1990; Dale et al., 1992; Kessler and Melton, 1995; Jones et al., 1995). To determine which of these factors are able to influence Apod expression, different doses of activin, bVg1, Xnr-2 or Bmp-4 mRNAs were overexpressed in animal caps and then Apod and Xbra mRNA levels determined (Fig. 7A, B respectively). Values were expressed as a percentage of mRNA levels found in the whole embryo. Apod mRNA was strongly induced by activin, bVg1, Xnr-2, and Bmp-4 to levels comparable to, or greater than, those found in the whole embryo. This suggests that the expression of Apod does not depend of the degree of dorsalisation promoted by these different TGF-β molecules. Xbra also responded to all mesoderm inducers tested. The most significant difference in the response of the two genes was that Xbra was considerably more responsive to BMP-4 than was Apod. At a dose of 100 pg, Apod was barely induced whereas Xbra mRNA levels were over twice that observed in the whole embryo.

There is evidence to suggest that the FGF signalling pathway is required both for the formation of mesoderm with posteroventral characteristics and for the initiation and/or the maintenance of Xbra gene expression (Amaya et al., 1991; Isaacs et al., 1994; Schulte-Merker and Smith, 1995). In an attempt to determine the effect of FGF on Apod gene expression, we first overexpressed increasing doses of bFGF mRNA in animal caps. Surprisingly Apod was not induced by bFGF, even at high doses, whereas Xbra was induced by bFGF to levels several times greater than that found in the whole embryo (Fig. 8A).

To further examine the dependence of Apod on FGF signalling, we radially injected mRNA for the the dominant negative FGF receptor (XFD) into the equatorial region of 2-cell-stage embryos to abolish FGF signalling (Amaya et al., 1991). We subsequently assayed the level of Apod and Xbra mRNA by RNase protection assay at different stages of development to see how the expression of these genes was affected. As shown in Fig. 8B, Apod mRNA levels are relatively unaffected by XFD expression in the late blastula (stage 9) and early gastrula (stage 10) whereas Xbra mRNA levels are severely diminished at equivalent stages. Quantitation using a phoshorimager shows that the level of Xbra mRNA in XFD-injected, stage 10, early gastrula embryos is reduced to half that found in equivalent uninjected controls. In contrast, the level of Apod mRNA is still approximately 90% that observed in control embryos at the same time. From the mid gastrula stage onwards, Apod mRNA levels are then reduced to approximately 60% of control values in XFD-injected embryos. In contrast, Xbra mRNA levels are markedly reduced to 20 to 30% that of controls. These results are also consistent with our observation that Apod mRNA induced in animal cap explants following injection with activin mRNA is only partly diminished by coinjection of XFD mRNA, whereas Xbra induction is severely blocked (data not shown).

The relative insensitivity of Apod to XFD expression compared to Xbra suggests that the requirement of these two T-box genes for FGF signalling is quite different. One explanation is that, whereas Xbra requires FGF signalling in an autoregulatory loop to maintain its later expression (Isaacs et al., 1994; Schulte-Merker and Smith, 1995), Apod does not.

The T-box gene, Xbra, is considered to be a marker that defines mesoderm (Smith et al., 1991). Now two more T-box genes, Eomes (Ryan et al., 1996) and Apod, have been cloned that are expressed pan-mesodermally and can specify a mesodermal cell fate. It is a likely possibility that these and potentially other T-box genes interact in overlapping pathways that bring about mesoderm differentiation and we have indeed already shown that Apod is able to induce the expression of Eomes and Xbra (Fig. 6D). We therefore examined the ability of Eomes and Xbra to regulate other Xenopus T-box genes by ectopically expressing each in the animal pole region of embryos and analysing gene expression by RNase protection in animal cap explants. The expression of Xwnt-8 was used as a positive control since Eomes and Xbra have been shown to elevate its mRNA levels (Ryan et al., 1996).

We first examined what effect Eomes has on the expression of Apod. Fig. 9A shows that injection of 4 ng Eomes mRNA into the animal pole can strongly induce the expression of Apod mRNA in animal cap explants at stage 10.25. Thus, Apod and Eomes are able to induce each other’s expression and may act in a cross-regulatory loop. Eomes is also able to activate Xbra (Ryan et al., 1996), suggesting that both Apod and Eomes are upstream components of the mesoderm induction pathway that is mediated by Xbra.

We then asked whether Xbra is able to induce Apod or Eomes. Injection of 4 ng Xbra mRNA is unable to increase the level of Apod mRNA at stage 10.25 above the low maternal level sometimes found in the large uninjected animal caps (Fig. 9A). Lower or higher concentrations of Xbra were still unable to induce Apod at this stage (data not shown). However, if caps are assayed at stage 11, some 2 hours later (see Fig. 9B), then induction of Apod in response to overexpression of Xbra becomes evident. One possible explanation for these observations is that, by stage 11, complex interactions between several genes have had time to occur and the induction of Apod mRNA by Xbra may now be indirect. Another possibility is that the early mechanisms regulating the initial zygotic expression of Apod are different to those regulating the later, maintained expression of Apod, and it is the latter on which Xbra is able to exert any influence.

These observations show that (1) Apod and Eomes can induce each others’ expression and (2) Apod and Eomes can both induce Xbra. This suggests that once Apod and Eomes are induced, they may then participate in a regulatory loop where each maintains the others’ expression. The mechanism of such ‘cross-talk’ for defining mesoderm identity will require further investigation.

The new Xenopus T-box gene described here, Antipodean (Apod), encodes a putative transcription factor that is likely to be important in mesoderm formation for the following reasons: (1)a large maternal content of Apod mRNA is localised to the vegetal part of oocytes and early embryos, a region involved in mesoderm induction; (2) zygotic Apod transcription takes place in very early mesoderm cells, and can be induced by known mesoderm-inducing molecules of the TGF-β class; (3) ectopic expression of Apod in ectodermal (animal cap) cells changes their fate toward the mesoderm lineage.

The study of chordate development (ascidians and Xenopus) has provided compelling evidence for the existence of localised maternal factors that play critical roles in the determination of cell fate during early embryogenesis. In ascidian embryos, cytoplasmic factors, probably mRNA molecules, are required for mesoderm, endoderm and epidermis cell differentiation (Nishida, 1996). In amphibians, mesoderm formation is thought to be largely initiated in the equatorial region of the embryo as the result of inductive interactions between vegetal and animal regions. Such processes can be reproduced in vitro by incubating animal caps with diffusible molecules of the TGF-β and FGF families (reviewed by Slack, 1994) or can be artificially blocked by preventing cell-cell contact during the initial step of mesoderm formation in Xenopus embryos (Gurdon et al., 1984; Symes et al., 1988).

In contrast to this ‘all-induction’ model, Gurdon and colleagues (1985b) identified a subequatorial zone in fertilised eggs needed for the initiation of muscle differentiation in embryos as gastrulation proceeds. The possibility of an autonomous regulatory pathway in mesoderm formation has been further emphasised by the finding that the expression pattern of some genes involved in the definition of the dorso-ventral polarity, goosecoid and Xwnt-8 respectively, does not depend on cell-cell interactions (Lemaire and Gurdon, 1994). Thus, it has been speculated that the accurate formation of mesoderm is dependent on both cell interaction and cell-autonomous pathways.

One way of relating the results that we report here to previous work is to envisage that Apod is an important molecule of both regulatory pathways and that vegetally localised maternal Apod mRNA may act as a maternal determinant for the mesoderm lineage in a subequatorial position. Two possibilities might be envisaged: (1) a subset of cells that inherit Apod mRNA are defined as potential mesodermal cells; (2) alternatively, maternal Apod mRNA may be primarily involved in establishing the specification of the endogenous mesoderm-inducing tissue, the endoderm. Further investigation will be necessary to distinguish between these possible roles of maternal Apod expression.

Apod appears to the first example of a trancription factor mRNA localised during oogenesis to the vegetal region. Other maternal mRNAs that are first localised to the vegetal pole at a similar time as Apod include Xwnt-11 (Ku and Melton, 1993), Xcat-2 (Mosquera et al., 1993; Forristall et al., 1995), Xcat-3 (Elinson et al., 1993) and Xlsirts (Kloc et al., 1993). In contrast, Vg1, is localised to the cortex later than Apod, at stage 4 (Weeks and Melton, 1987). Apod is different from these other vegetally localised mRNAs in that it is predominantly expressed in the vegetal yolk mass of the mature oocyte, as opposed to the cortex. Furthermore, it is more concentrated in a subequatorial position, a novel location amongst characterised Xenopus mRNAs. It has been proposed that Xcat-2 and Xlsirts move to the vegetal pole with the mitochondrial cloud as it moves from near the germinal vesicle to the vegetal cortex (Forristall et al., 1995; Kloc et al., 1993). The localisation of Apod to the vegetal pole at a similar time as these mRNAs suggests that it may also be dependent on this process.

The early zygotic trancription of Apod is localised to the meso-dermal region of gastrula embryos and, with Eomes (Ryan et al., 1996), it is the earliest mesodermally localised gene expression. We have shown here that Apod expression is induced by all molecules of the TGF-β class that we have tested, with an efficiency that is independent of their dorsalising activity. Surprisingly, and in contrast to Xbra, we have been unable to activate Apod by bFGF. Activin-inducible, mesodermal genes can be classified according to the extent to which they depend on FGF signalling. Expression of Xbra, MyoD and muscle actin require FGF signalling for their expression, whereas Xlim-1, Xwnt-8, gsc and Mix.1 do not (Cornell and Kimelman, 1994; Cornell et al., 1995). The inability of Apod to respond to bFGF potentially places Apod in the latter class of genes. This may indicate that the TGF-β and FGF molecules induce/maintain mesoderm by activating different sets of T-box genes, which in turn control the expression of different subsets of mesodermal genes in Xenopus embryos.

All these activin inducible genes are expressed in mesoderm cells during gastrula stages in Xenopus development. However, only those that do not depend on FGF signalling, i.e. Xlim-1, Xwnt-8, gsc, Mix.1, are also expressed in endodermal cells at the same stages (Rosa, 1989; Cornell et al., 1995; Lemaire and Gurdon, 1994). Furthermore, their endodermal expression is dependent only on an intact activin or activin related signalling pathway (Hemmati-Brivanlou and Melton, 1992; Schulte-Merker et al., 1994; Cornell et al., 1995,). In addition, TGF-β molecules seem to be important for patterning the expression of some endodermal genes along the dorsoventral axis (Henry et al., 1996). Taken together, these observations support the view that, despite the fundamental difference in behaviour, endoderm and mesoderm cells are subject to similar TGF-β signalling regulatory pathways and dorso-ventral patterning.

Our data show that Apod is dependent on TGF-β, but not FGF signalling. It is expressed maternally in the vegetal pole from which endodermal lineages arise, as well as zygotically in what may be both the mesoderm and parts of the endoderm. The possibility exists, therefore, that zygotic Apod may be involved with co-ordinating the initiation of both mesoderm and endoderm lineages in response to a TGF-β signal.

Developmentally important genes are generally able to activate the trancription of other very early genes, by over or ectopic expression. Apod has this ability, since the injection of Apod mRNA in animal caps induces most other early mesodermal genes, notably Eomes and Xbra. These genes are all expressed in an overlapping distribution throughout most of the mesoderm, Xbra following soon after Eomes and Apod. Eomes and Apod can induce each other, and both can induce Xbra. This may reflect a mechanism by which early T-box genes maintain and amplify mesoderm gene expression in the correct region of the embryo after its initiation.

By the late gastrula stages, Apod and Xbra expression is complementary; Xbra mRNA is expressed in the notochord, whereas Apod mRNA is associated with ventral structures and is excluded from notochord cells. The mechanisms responsible for this tissue specificity are likely to be quite complex, and may involve regulation by different signalling pathways.

In view of the localised maternal content of Apod mRNA, it is possible that its translation helps to initiate the mesodermally localised transcription of other T-box genes, thereby initiating a cascade of expression of this class of gene and leading to mesodermal differentiation.

We are very grateful to Dr Aaron Zorn for invaluable suggestions and discussions. We also thank Dr Ken Ryan for use of the Eomesodermin clone and unpublished information. This manuscript was improved upon critical reading by Drs Ken Ryan and Aaron Zorn. We thank Nigel Garrett, Andrew Mitchell and Dr Nobu Kikyo for excellent technical advice and Elizabeth Tweed for keeping the frog colony. This work was supported by the Cancer Research Campaign.

F. S. is a Cancer Research Campaign Postdoctoral Fellow and G. C is a European Science Foundation and European Postdoctoral fellow.