Mesoderm induction in Xenopus is a zygotic event regulated by maternal VegT via TGFβ growth factors

The mechanism of mesoderm formation has been more closely studied in the Xenopus laevis embryo than in any other vertebrate. The morphological event establishing mesodermal tissue is the process of gastrulation, but the molecular determination of mesoderm begins earlier in development. Explants of the marginal zone dissected from blastula stage embryos differentiate into mesodermal derivatives (Nakamura et al., 1970), and cell transplantation experiments indicate that groups of cells are determined to form mesoderm by the early gastrula stage (Kato and Gurdon, 1993). Explant recombination experiments suggest that mesoderm forms in the marginal zone in response to inductive signals passing from adjacent vegetal cells (Boterenbrood and Nieuwkoop, 1973; Nieuwkoop, 1969). However, there is also evidence that equatorial cells may be specified autonomously to form mesoderm, without requiring inducing signals (Gimlich, 1986; Sakai, 1996). Several members of the FGF and TGFβ family of growth factors are expressed in early Xenopus embryos, and have been implicated in mesoderm formation. eFGF, bFGF, activin B, Xnr1, Xnr2, Xnr4, derrière and bVg1 can all act as mesoderm inducers (reviewed in Harland and Gerhart, 1997; Slack, 1994), in that their expression in animal caps causes animal cells to adopt mesodermal fates. Furthermore, the expression of dominant negative forms of the receptors for FGF or activin disrupt gastrulation and inhibit mesoderm formation (Amaya et al., 1991, 1993; Dyson and Gurdon, 1996; Hemmati-Brivanlou and Melton, 1992) and dominant inhibitory forms of derrière (Sun et al., 1999), Vg1 (Joseph and Melton, 1998), and Xnr2 (Osada and Wright, 1999) also affect gastrulation and both mesodermal and endodermal development. However, the relative importance of each of these candidate growth factors in mesoderm formation and the location of their activity remains to be established.

It has been generally assumed that mesoderm-inducing signals are maternal growth factors, regulating mesodermal specification before the onset of zygotic transcription. This view has been challenged, however, by two recent observations. Firstly, heterochronic Nieuwkoop recombination experiments indicate that vegetal cells are competent to induce mesodermal tissue only after the mid-blastula transition, not before zygotic transcription starts (Wylie et al., 1996; Yasuo and Lemaire, 1999). Secondly, loss-of-function experiments have identified the maternal transcription factor VegT as a key initiator of mesoderm induction and endoderm formation (Zhang et al., 1998). Since VegT has been shown to act as a transcriptional activator (Zhang and King, 1996), it is likely to regulate the expression of zygotic genes, including growth factors. Overexpression experiments and dominant negative studies also show the likely involvement of VegT in blastopore formation and mesodermal patterning (Horb and Thomsen, 1996; Lustig et al., 1996; Stennard et al., 1996; Zhang and King, 1996).

In this paper, we aimed to establish whether this latter view is correct. In previous work, we showed that depletion of maternal VegT mRNA from Xenopus embryos affected mesoderm formation. Mesodermal markers were delayed in their expression, mesodermal differentiation occurred ectopically, in cells derived from the vegetal mass of the blastula rather than from the equatorial origin, and the vegetal cells lost their ability to produce mesoderm-inducing signals as measured by Nieuwkoop assays (Zhang and King, 1996). Here we have asked, firstly, to what extent maternal VegT is essential for mesoderm formation. Secondly, we establish which of the candidate growth factors are downstream of VegT, by studying the initiation of their expression in VegT-depleted embryos. Although mesoderm formation may be the result of both cell autonomous and inducing activity, we chose to concentrate on the latter aspect, asking the question ‘which growth factors downstream of VegT are responsible for mesoderm induction by the vegetal mass?’ We test this by injecting the candidate mRNAs into vegetal cells of the 32-cell stage VegT-depleted embryos, such that they are expressed only in the vegetal mass at the blastula stage. We then assess their ability to rescue the VegT-depleted phenotype and to rescue mesoderm formation. We find that maternal VegT is responsible for the formation of 90% of mesodermal tissue as measured by the expression of MyoD, cardiac actin, Xbra, Xwnt8 and alphaT4 globin. eFGF, FGF3, FGF8, and the TGFβs, Xnr1, Xnr2, Xnr4 and derrière, are directly or indirectly downstream of VegT, while BMP4, BMP7, Xnr3 and activin B are not. Xnr1 has potential VegT-binding sites in its promoter, and VegT activates the expression of an Xnr1 promoter-Luciferase construct, suggesting it is directly regulated by VegT. mRNAs for Derrière, Xnr1, Xnr2 and Xnr4 rescue blastopore formation and mesodermal tissue. Derrière rescues trunk and tail formation, while Xnr1, Xnr2 and Xnr4 rescue head, trunks and tails. Furthermore, Xnr1 expression rescues the mesoderm-inducing activity of VegT-depleted vegetal masses. This establishes that mesoderm formation in Xenopus depends on zygotic rather than maternal growth factors, and that mesoderm induction involves the TGFβ growth factors, Xnr1, Xnr2, Xnr4 and derrière.

Full-grown oocytes were manually defolliculated and cultured in oocyte culture medium, as described in Zuck et al. (1998). Oocytes were injected at the vegetal pole with oligo using a Medical Systems picoinjector, in oocyte culture medium (OCM) and cultured a total of 24-48 hours at 18°C before fertilization. In preparation for fertilization, they were stimulated to mature, by the addition of 2 μM progesterone to the culture medium, and cultured for 12 hours.

Oocytes were then labelled with vital dyes and fertilized using the host-transfer technique described previously (Zuck et al., 1998). 3 hours after placing in the frog’s body cavity, the eggs were stripped and fertilized along with host eggs using a sperm suspension. Embryos were maintained in 0.1× MMR, and coloured, experimental embryos were sorted from host embryos. Unfertilized eggs and abnormally cleaving embryos were removed from all batches.

For injections of mRNAs, embryos were transferred to 1% Ficoll in 0.5× MMR at the 16-cell stage. mRNAs were injected into blastomeres as described in the text. Embryos were washed thoroughly and returned to 0.1× MMR during the blastula stage.

The antisense oligodeoxynucleotide (oligo) used was an 18 mer: C*A*G*CAGCATGTACTT*G*G*C where * indicates a phosphorothioate bond and was HPLC purified before use (Genosys/Sigma). Oligos were resuspended in sterile, filtered water and injected in doses of 5-8 ng per oocyte, and cultured immediately at 18˚C. Capped RNAs were synthesized using the mMessage mMachine kit (Ambion), ethanol precipitated and resuspended in sterile distilled water for injection. Isolation of the Xnr1 genomic DNA, construction of Pro-Luc, and methods for luciferase assays are described elsewhere (S.-I. Osada et al., unpublished data).

Embryo RNA was extracted as described (Gurdon et al., 1985). Electrophoresis and northern blotting were performed as described (Hopwood et al., 1989) using two embryo equivalents per lane. The probe was synthesized by random priming of the excised insert of VegT (EcoRI). Blots were stripped and rehybridized with a probe for plakoglobin as a loading control.

Total RNA was prepared from oocytes and embryos using the proteinase K method and treated with RNase-free DNase 1 (10 μg/μl Boehringer Mannheim) prior to cDNA synthesis. cDNA was synthesized from 0.5 to 1.0 μg RNA according to Zhang et al. (1998) in a volume of 20 μl. After reverse transcription, 1 μl 0.5 M EDTA, 30 μl H2O, 1 μl glycogen (20 μg/μl) and 10 μl 5 M ammonium acetate, were added to each RT-reaction. Each sample was extracted once with phenol/chloroform/isoamyl alcohol (25:24:1) and precipitated overnight at −20°C with 2.5 volumes 100% EtOH. Samples were centrifuged at 4°C, 16,000 g for 15 minutes, washed with 70% EtOH, dried in a speedvac and resuspended in 150 μl H₂O per 1/6th embryo equivalent of RNA used for cDNA synthesis.

RT-PCR was carried out using a LightCycler^TMSystem (Roche), which allows amplification and detection (by fluorescence) in the same tube, using a kinetic approach. LightCycler PCR reactions were set up in microcapillary tubes using 5 μl cDNA with 5 μl of a 2× SYBR Green I (Roche Molecular Biochemicals, Wittwer et al., 1997) master mix containing upstream and downstream PCR primers, MgCl₂ and SYBR Green. The final concentrations of the reaction components were 1.0 μM each primer, 2.5 μM MgCl₂, and 1× SYBR Green master mix. The primers used and cycling conditions are listed in Table 1. In order to compare expression levels of depleted and rescued embryos relative to controls, a dilution series of uninjected control cDNA was made and assayed in each LightCycler run. Undiluted control cDNA = 100%, 1:1 cDNA: H₂O = 50% and 1:10 cDNA: H₂O = 10% (shown only in Fig. 2). In experiments where multiple embryonic stages were examined, the dilution series was used from cDNA of the uninjected control stage of development predicted to give the highest expression of the gene product being amplified. These values were entered as concentration standards in the LightCycler sample input screen. Other controls included in each run were −RT and water blanks. These were negative in all cases but not included in the figures for lack of space.

After each elongation phase, the fluorescence of SYBR green (a dye that binds double-stranded DNA giving a fluorescent signal proportional to the DNA concentration) was measured at a temperature 1°C below the determined melting point for the PCR product being analyzed. This excluded primer-dimers, which melt at a lower temperature, from the measurement. The fluorescence level is thus quantitated in real-time, allowing the detection and display of the log-linear phase of amplification as it happens. LightCycler quantification software v1.2 was used to compare amplification in experimental samples during the log-linear phase to the standard curve from the dilution series of control cDNA. The comparisons are displayed as histograms (Figs 2, 3, 7, 9 and 10).

For each primer pair used, we optimized conditions so that melting curve analysis showed a single melting peak after amplification, indicating a specific product. Some published primers used for radioactive PCR always gave multiple peaks in all conditions tried and so were not used.

For X-gal staining, embryos were fixed in MEMFA for 2 hours, rinsed in PBS and stained using X-gal (Hemmati-Brivanlou and Harland, 1989). Embryos were washed in PBS after staining and photographed before clearing with Murray’s clear (2:1 butyl alcohol and butyl butyrate).

Mid-blastula stage uninjected, VegT-depleted and VegT-depleted/Xnr1 mRNA-injected embryos were devitellined and dissected on agar-coated dishes in 1× MMR, into animal, equatorial and vegetal segments. The VegT-depleted embryos used to provide either equators or vegetal masses were dyed different colors as oocytes, so that they could be distinguished and separated after recombination. After washing away dead cells, equatorial pieces were placed, each over two vegetal masses, and cultured on agar in OCM for 3 hours. The recombinants were separated using tungsten needles and stray vegetal cells were identified by their different vital dye colouring and removed from the equators. The equators were cultured in OCM until sibling embryos reached stage 21 and frozen for analysis.

To study further the role of maternal VegT in mesoderm formation, we enhanced the effectiveness of the antisense oligo used previously, by both increasing the dose and by HPLC purifying the oligo (Fig. 1a,b). Embryos derived from oocytes injected with 5-7 ng of this HPLC-purified oligo have a more severe phenotype than in the previous study (Fig. 1c,d), that is specifically due to the loss of maternal VegT mRNA, since it is rescued by the reintroduction of synthetic VegT mRNA (Fig. 1c,d). The VegT-depleted embryos do not form blastopores and remain only with an animal/vegetal axis when siblings are at the swimming tadpole stage (Fig. 1d).

We studied the effect of maternal VegT depletion on mesoderm formation, by comparing the expression of mesodermal markers MyoD (dorsal mesoderm), cardiac actin (dorsal mesoderm), Xwnt8 (ventral mesoderm), T4 globin (ventral mesoderm) and Xbra (early general mesoderm) in a developmental series of control and experimental embryos, using RT-PCR. To make these comparisons as accurate as possible and to ensure they were made in the linear range, we used the LightCycler^TM (Roche) amplification and detection system for real-time RT-PCR. The program compares the degree of amplification of each sample with that of a dilution series (100%, 50% and 10%) of uninjected control cDNA in each RT-PCR run (see Materials and Methods). While control embryos expressed increasing amounts of MyoD, cardiac actin and alphaT4 globin over time, VegT-depleted embryos expressed very low levels (10% or less of control) at all stages (Fig. 2). The early mesodermal markers, Xbra and Xwnt8 were similarly reduced to below the 10% level in VegT-depleted embryos compared to uninjected controls. The experiment was repeated, with the same result (data not shown). This indicates that maternal VegT regulates the formation of mesodermal tissue in the embryo.

Members of the FGF and TGFβ families have been implicated in mesoderm formation in Xenopus embryos. We tested whether maternal VegT is necessary for the initiation of expression of these genes by comparing their zygotic expression in a staged series of control and maternal VegT-depleted embryos. The effect of VegT depletion on early endoderm formation will be considered elsewhere (J. X. et al., unpublished data). The expression of Xnr1, Xnr2, Xnr4 and derrière, as well as FGF3 and FGF8 was dramatically reduced to below the 10% level in VegT-depleted embryos compared to controls (Fig. 3). In contrast, BMP4 and BMP7 expression was either unaffected or increased in VegT-depleted embryos compared to control levels (Fig. 3a and b). eFGF, activin B and Xnr3 expression was reduced but only to 30-50% of control levels. These results show that maternal VegT regulates, either directly or indirectly, the transcription of several TGFβ and FGF class growth factors.

To study whether VegT activates TGFβ genes directly, we scanned the regulatory regions of Xnr1 (Fig. 4a; S.-I. O. et al., unpublished data) for potential VegT binding sites (CTTCACACCT; Tada et al., 1998; Smith, 1999). Two such possible sites were found, a distal 10/10 match in reverse orientation and proximal 7/10 match (underlined in Fig. 4a).

A promoter-Luciferase construct containing these sites was injected either alone or together with VegT mRNA into both cells of 2-cell stage wild-type embryos. VegT dose-dependently activates the reporter gene expression from the Xnr1 promoter (Fig. 4b) suggesting that Xnr1 is directly regulated by VegT. Injection of reporter construct alone into vegetal regions led to activation of luciferase by endogenous factors, with a higher activity in vegetal compared to animal injections (data not shown).

To test the likely importance of each of these factors downstream of VegT in mesoderm formation, we studied their ability to rescue VegT-depleted embryos. Since mesoderm-inducing signals are known to be produced by vegetal masses, we injected Xnr1, Xnr2, Xnr4, derrière, eFGF and VegT mRNAs into four of the D tier cells of VegT-depleted embryos at the 32-cell stage. We confirmed that this site of injection restricted the injected mRNAs to cells of endodermal fate, by injecting the lineage label, β-galactosidase mRNA along with the growth factor mRNAs. Fig. 5 shows a representative experiment in which the embryos were cleared and stained for X-gal. Control embryos have X-gal staining in the embryonic endoderm and not in mesodermal structures (Fig. 5a).VegT-depleted embryos injected with VegT (Fig. 5b), Xnr1 (Fig. 5c) and derrière mRNA (Fig. 5d) also have staining in the rescued gut region, while VegT-depleted embryos have blue cells in one hemisphere (Fig. 5e).

The degree to which VegT-depleted embryos were rescued by growth factor mRNAs was examined by three criteria: by the formation of the blastopore, by the rescue of mesodermal markers using RT-PCR and by the phenotypic rescue of head, trunk and tail. Rescue of head organizer and endoderm specific genes will be considered elsewhere (J. X. et al., unpublished data).

Rescue of blastopore formation

Embryos severely depleted of maternal VegT do not form blastopores or undergo gastrulation movements (Fig. 6a,b; data not shown).

The expression of all of the TGFβ class mRNAs, with the exception of Xnr3, rescued the formation of blastopores in VegT-depleted embryos (Table 2). Xnr2 and Xnr1 were most effective in this respect (95% of VegT-depleted embryos have and Xnr4 were less active (68% for derrière, 43% for Xnr4). Blastopores were generally delayed in their appearance, and frequently more vegetally placed than those of control embryos, but bottle cell formation was followed by tissue involution (Fig. 6c), as in control embryos. For all these mRNAs (except Xnr4) an upper limit of concentration was found (activin B: 4 pg, Xnr1: 600 pg, Xnr2: 150 pg, derrière: 400 pg), above which the blastopores formed were extremely large and the embryos developed with the same abnormal appearance. eFGF did not rescue blastopore formation over a range of doses (0.5-24 pg), and higher doses (12 pg) caused cell disaggregation at the gastrula stage (data not shown).

Embryos from these experiments were frozen at the mid-gastrula stage and analysed by RT-PCR for the expression of mesodermal markers, chordin, Xbra and Xwnt8 (Fig. 7). The expression of VegT mRNA itself, as well as Xnr1, Xnr2 (Fig. 9), Xnr4 and derrière in vegetal cells of VegT-depleted embryos caused a partial rescue of these markers.

Rescue of the VegT-depleted phenotype

Since previous studies have shown that there is a dose-dependent activity of TGFβs and FGFs, we carried out rescue experiments using a dose range of each of the growth factors regulated by VegT. Fig. 8a-d shows a typical experiment, where embryos were either VegT-depleted (Fig. 8a), or VegT-depleted and injected with derrière RNA at the 32-cell stage in doses of 200 (Fig. 8b), 400 (Fig. 8c) and 800 pg (Fig. 8d). In three experiments, derrière RNA rescued the formation of trunk and tail at doses above 200 pg but did not rescue head formation. 800 pg caused abnormalities in control embryos.

Similar experiments with Xnr1 (Fig. 8e), Xnr2 (Fig. 8f) and Xnr4 (Fig. 8g), showed that all these nodal-related proteins rescued head as well as trunk and tail formation. Each showed a dose-responsiveness of activity, with Xnr2 having most rescue activity at lowest doses (60 pg), while Xnr1 was most active at 100-200 pg and Xnr4 showed the least rescuing activity, particularly of head formation. In higher doses of Xnr1 (200 pg), embryos often had enlarged heads and reduced tails (e.g. Fig. 5c). In comparison, eFGF was unable to rescue formation of the normal body axis, even at doses (3 pg) that caused tail abnormalities in control embryos (Fig. 8h). Higher doses of eFGF caused abnormality and cell disaggregation in both control and experimental embryos. Even though activin expression is not solely regulated by VegT (Fig. 3), we also tested its ability to rescue the VegT-depleted phenotype. Doses above 3 pg caused abnormalities in both experimental and control embryos, while lower doses rescued trunk and tail structures but not heads (data not shown). Perhaps not surprisingly, given our limited way of introducing the rescuing mRNAs compared to endogenous mechanisms, VegT-depleted embryos never developed completely normally when rescued in this way, with either VegT or the growth factor mRNAs.

To study the extent to which the growth factor mRNAs were able to rescue mesodermal markers in these embryos, sibling embryos were frozen at the gastrula and neurula stages and analysed by RT-PCR. Fig. 9 shows the rescue of the expression of MyoD, Xwnt8, Xbra, eFGF, chordin and cardiac actin mRNAs in VegT-depleted embryos with increasing doses of Xnr2 mRNA. Doses of 300-600 pg caused overexpression of all of these markers in VegT-depleted embryos (with the exception of Xbra) compared to wild-type levels. Xnr1, Xnr4 and derrière rescued mesodermal markers in a similar fashion (data not shown). For all the TGFβ growth factors, there was a close correlation between the degree of phenotypic rescue and the degree of rescue of mesodermal markers.

The experiments described above suggest that TGFβs downstream of VegT are released by vegetal cells to induce mesoderm. To confirm this, we tested whether one of the factors with most rescuing ability, Xnr1, was able to rescue the mesoderm-inducing activity of VegT-depleted vegetal masses in recombination experiments. In previous experiments, we showed that VegT-depleted equatorial regions are unable to form mesoderm (Zhang et al., 1998; Fig. 10). Here we co-cultured such equatorial regions with the vegetal masses from wild-type, or VegT-depleted embryos, or with VegT-depleted/Xnr1-expressing vegetal masses, where Xnr1 mRNA had been injected into four vegetal cells at the 32-cell stage. Recombinants were combined at stage 8 (Fig. 10a), separated after 3 hours and cultured until stage 21. While control VegT-depleted equators cultured alone or in combination with VegT-depleted vegetal masses failed to elongate or express mesodermal markers, those cultured with wild-type vegetal masses or VegT-depleted/Xnr1-rescued vegetal masses underwent shape changes (Fig. 10b) and formed mesoderm, as evidenced by the expression of Xbra, Xwnt8, MyoD, cardiac actin, VegT and eFGF (Fig. 10c). Thus Xnr1 rescues the ability of VegT-depleted vegetal masses to induce mesoderm in the equatorial region and to initiate the expression of eFGF in this region.

The loss-of-function approach used here allowed us to study the importance of maternal VegT for mesoderm formation in Xenopus development and to identify zygotic mesoderm-inducing growth factors regulated by VegT. We find that when VegT is depleted, ventral, general and dorsal mesodermal markers are reduced to less than 10% of control uninjected levels at gastrula and neurula stages. It is unclear whether residual mesoderm formation occurs because there is some VegT protein still active in these embryos or whether a second pathway is involved. Other studies and reviews have suggested the existence of a second “early weak mesoderm-inducing signal”(Clements et al., 1999; Kimelman and Griffin, 1998; Yasuo and Lemaire, 1999). If this signal exists, it is unable to cause blastopore formation and can only induce mesodermal markers to the 10% level, in VegT-depleted embryos.

One question is the degree to which mesoderm formation is an autonomous property of cells of the equatorial region. Although VegT mRNA is located throughout the vegetal hemisphere of oocytes (Zhang et al., 1996; Stennard and Gurdon, 1996), recent studies on VegT protein localization show that it is concentrated in cells of the vegetal mass at the gastrula stage, and not in equatorial cells (Stennard et al., 1999). This suggests that the principle mechanism by which mesoderm forms is one of induction by zygotic growth factors initiated by vegetally localized VegT. It is also possible that low levels of VegT protein could activate mesoderm formation autonomously in equatorial cells (eg. by initiating FGF expression directly in this region).

The lack of expression of mesodermal markers in VegT-depleted embryos is explained here by the fact that many of the known candidate zygotic mesoderm inducers are regulated by maternal VegT. As a result, VegT-depleted embryos have a more extreme phenotype than that caused by the inhibition of FGF, Xnrs or derrière individually (Amaya et al., 1991; Osada and Wright, 1999; Sun et al., 1999). Since the expression of Xnr1, Xnr2, Xnr4, derrière, FGF3 and FGF8 are almost completely absent in VegT-depleted embryos from the early gastrula stage, it is likely that maternal VegT is the sole transcription factor initiating their expression whether directly or indirectly. We disagree with previous overexpression studies that conclude that the primary signal downstream of VegT consists of activin B, derrière and Xnr4 (Clements et al., 1999), since activin B is still expressed in VegT-depleted embryos, and Xnr1 and Xnr2 are at least as active in germ layer formation as Xnr4 and derrière. Similarly, our work does not support an overexpression study suggesting that VegT is insufficient to activate Xnr1 expression (Yasuo and Lemaire, 1999).

VegT might regulate this gene expression directly, or initiate a chain, or, more likely, a network of signals. For Xnr1 at least, there is evidence from the work presented here (Figs 4a, 9) for both direct activation by VegT and indirect regulation via an Xnr2-initiated pathway. Also we show that eFGF expression is initiated by the reintroduction of Xnr1 and Xnr2 mRNAs into VegT-depleted embryos (Figs 9, 10c), while others have found T-box response elements in the promoter of eFGF (Casey et al., 1998). Verification of the direct targets of VegT may require their functional analysis by the transgenic approach, as recently carried out for the Bix 4 gene (Casey et al., 1999), a direct target of VegT involved in endoderm formation.

Although activin B is induced by VegT in animal cap assays (Clements et al., 1999), we find here that activin is not solely regulated by VegT in the embryo. It is present at 50% control levels in VegT-depleted embryos, but this level of expression does not allow embryos to form mesoderm or to undergo gastrulation. This supports the body of evidence that activin does not play a major role in mesoderm formation (Schulte-Merker et al., 1994). In contrast, when we inject synthetic activin mRNA into VegT-depleted embryos, it has rescuing ability, similar to that of derrière. One interpretation of this is that all Smad2-activating TGFβs can mimick the endogenous pathway when overexpressed.

This work adds further evidence to the loss-of-function studies on mouse nodal (Conlon et al., 1994) and zebrafish nodal-related genes Znr1 and Znr2 (Erter et al., 1998; Feldman et al., 1998; Sampath et al., 1998), and to Xenopus overexpression and dominant negative studies (Jones et al., 1995; Osada and Wright, 1999; Clements et al., 1999; Yasuo and Lemaire, 1999), indicating that the nodal-related genes are essential for both gastrulation movements and mesodermal tissue formation. The fact that Xnr1, Xnr2 and Xnr4 mRNAs show similar abilities to rescue mesoderm, makes it difficult to assess whether each has separate roles in the embryo; this could only be assessed by loss-of-function studies on each gene. Zebrafish nodal-related genes have been shown to have functional redundancy (Erter et al., 1998; Feldman et al., 1998; Sampath et al., 1998).

The roles of the other TGFβs, derrière and Xnr3, are also difficult to interpret. Derrière is clearly regulated by VegT but does not rescue head formation, which is consistent with studies using a dominant negative derrière construct, where posterior but not head structures were disrupted (Sun et al., 1999). This suggests that derrière’s function is non-overlapping in this respect with the Xnrs, and that it has a divergent signalling pathway, activating axial and tail genes but repressing or not regulating head genes. Xnr3 differs from the other Xnrs in continuing to be expressed in VegT-depleted embryos, although at lower levels than in controls and in that it is unable to rescue VegT-depleted embryos even in doses that cause abnormalities in control embryos (600 pg-1.2 ng, data not shown). This is consistent with its identification as a target of the β-catenin/XTcf signalling pathway (McKendry et al., 1997), its divergent structure and distinct inductive properties compared to other Xnrs (Smith et al., 1995), and supports its suggested role in dorsal rather than mesodermal patterning.

The zygotic FGFs, FGF3 and FGF8 are also regulated by VegT. The level of eFGF expression in VegT-depleted embryos was more variable than any of the other growth factors. For example, in Fig. 3a, eFGF is only slightly reduced in VegT-depleted compared to control embryos. However, in Figs 9, 10c, eFGF is not expressed in VegT-depleted embryos and is rescued by Xnr1 and Xnr2 expression. Radioactive RT-PCR also confirmed that eFGF is expressed only at very low levels in VegT-depleted embryos (data not shown). We conclude that eFGF is largely regulated by VegT and that the relatively high expression in Fig. 3a is due to a less severe depletion of VegT mRNA than in the other experiments.

In rescue experiments, the injection of eFGF mRNA into the vegetal masses of VegT-depleted embryos fails to initiate mesoderm induction (Fig. 6). However, eFGF is expressed in VegT-depleted equatorial regions in response to mesoderm-inducing signals from the vegetal mass (Fig. 10). This finding correlates with studies in wild-type embryos showing that eFGF mRNA is not present in vegetal masses at the blastula stage, but is expressed in the mesodermal ring of the gastrula (Isaacs et al., 1995). Furthermore, activation of the signalling pathway downstream of eFGF at the gastrula stage, as measured by phosphorylated ERK activity, occurs in cells of equatorial origin, not in vegetal cells (Christen and Slack, 1999). Taken together, these results show that eFGF is not a vegetal mesoderm inducer, but is activated in the equatorial region by Xnrs downstream of VegT.

The BMP class of TGFβ growth factors are expressed and even upregulated in VegT-depleted embryos (Fig. 3), and the homeobox gene regulated by BMP4, Xvent 2 (Friedle et al., 1998), is also maintained (data not shown). It is also likely, since Xnr3 is still expressed, that the dorsal (β-catenin/XTcf) signalling pathway is active. Since mesoderm fails to form in VegT-depleted embryos, it may be that the only cells available to be influenced by the remaining BMP and the Xwnt pathways are of ectodermal origin. Alternatively, the distinction of ectoderm, mesoderm and endodermal lineages may be too simplistic a model of lineage regulation: it seems likely that more than three distinct lineages are activated at the onset of zygotic transcription. We find here and in previous work that both epidermal and neural lineages are still present in VegT-depleted embryos (Zhang et al., 1998; data not shown). The pathways initiating their formation remain uncharacterized.

One unresolved question is the extent to which the VegT-initiated mesoderm-inducing pathway and the dorsal signalling pathway interact. Xnr1, Xnr2 and Xnr4 are all expressed initally throughout the vegetal mass (Jones et al., 1995; Clements et al., 1999), and are shown here to be regulated by VegT and responsible for mesoderm induction. They have also been implicated in dorsal signalling, since they are expressed later in the Spemann organizer and rescue the dorsal axis of irradiated embryos to various extents (Jones et al., 1995; Joseph and Melton, 1997). It seems likely that the final location of the Xnrs is the result of the interplay of the dorsal (β-catenin/XTcf), ventral (BMP) and vegetal (VegT-initiated) signalling pathways. The analysis of the zygotic genes regulated by β-catenin/XTcf3 and of embryos depleted of both VegT and β-catenin or XTcf3 protein should help to resolve this issue.

We are grateful to the NIH (J. H.: NICHD 38272; C. Wright: NIH/NIGMS GM56238) for financing this work. H. L. S. was supported by the Human Frontiers Science Program, and by the Genetics Institute. Thanks to Larry Richmond for technical assistance.