Interactions between Nodal/Activin and Fibroblast growth factor (Fgf)signalling pathways have long been thought to play an important role in mesoderm formation. However, the molecular and cellular processes underlying these interactions have remained elusive. Here, we address the epistatic relationships between Nodal and Fgf pathways during early embryogenesis in zebrafish. First, we find that Fgf signalling is required downstream of Nodal signals for inducing the Nodal co-factor One-eyed-pinhead (Oep). Thus, Fgf is likely to be involved in the amplification and propagation of Nodal signalling during early embryonic stages. This could account for the previously described ability of Fgf to render cells competent to respond to Nodal/Activin signals. In addition, overexpression data shows that Fgf8 and Fgf3 can take part in this process. Second, combining zygotic mutations in ace/fgf8 and oep disrupts mesoderm formation, a phenotype that is not produced by either mutation alone and is consistent with our model of an interdependence of Fgf8 and Nodal pathways through the genetic regulation of the Nodal co-factor Oep and the cell propagation of Nodal signalling. Moreover,mesodermal cell populations are affected differentially by double loss-of-function of Zoep;ace. Most of the dorsal mesoderm undergoes massive cell death by the end of gastrulation, in contrast to either single-mutant phenotype. However, some mesoderm cells are still able to undergo myogenic differentiation in the anterior trunk of Zoep;aceembryos, revealing a morphological transition at the level of somites 6-8. Further decreasing Oep levels by removing maternal oep products aggravates the mesodermal defects in double mutants by disrupting the fate of the entire mesoderm. Together, these results demonstrate synergy between oep and fgf8 that operates with regional differences and is involved in the induction, maintenance, movement and survival of mesodermal cell populations.
Introduction
The concept of germ layers emerged from classical studies in experimental embryology. Ectoderm, mesoderm and endoderm layers are induced before they are shaped through gastrulation movements to give rise subsequently to specific differentiated cell types. A model for the formation of mesoderm was mainly derived from studies in amphibian embryos where mesoderm induction in the marginal zone relies on secreted molecules emanating from the vegetal pole(Nieuwkoop, 1973). This model can be readily applied to the zebrafish embryo where the extra-embryonic yolk syncytial layer and the blastoderm margin have been suggested to be the source of inducing signals (Mizuno et al.,1996; Rodaway et al.,1999; Chen and Kimelman,2000).
At the molecular level, secreted molecules of the transforming growth factor β (TGFβ) family have been implicated in the formation of mesoderm in both organisms (Harland and Gerhart, 1997). In the Xenopus embryo, TGFβ family members such as Activin and Vg1 were initially thought to act as maternal factors. However, characterization of the maternal transcription factor VegT,localized at the vegetal pole, led to a model in which the TGFβ family members Vg1-related Derriere and Nodal-type Xnr1, Xnr2 and Xnr4 are required zygotically downstream of VegT for the formation of the entire mesoderm(Sun et al., 1999; Clements et al., 1999; Kofron et al., 1999; Yasuo and Lemaire, 1999; Agius et al., 2000; Hyde and Old, 2000). Although VegT does not seem to be involved in the early steps of mesoderm formation in mouse and zebrafish, the study of mutants affected in the Nodal pathway indicates a conserved role for Nodals in vertebrate mesoderm development.
In zebrafish, genetic analysis of the Nodal-related genes squint(sqt), cyclops (cyc) and of oep, which encodes a Nodal co-factor, reveals that Nodal signalling is implicated in both mesoderm and endoderm formation (Feldman et al., 1998; Gritsman et al.,1999). Indeed, embryos lacking maternal and zygotic Oep function(MZoep), and sqt;cyc double mutants are devoid of endoderm and mesoderm in the head and trunk. However, the tail mesoderm is still induced and maintained in these mutants, which indicates that pathways other than the Oep-dependent Nodal pathway are implicated in formation of ventral mesoderm.
Fgf was identified in Xenopus as a mesoderm inducer(Kimelman and Kirschner, 1987; Slack et al., 1987) and overexpression of a dominant-negative form of a Fgf receptor impairs mesoderm induction in Xenopus and zebrafish gastrulae(Amaya et al., 1991; Griffin et al., 1995), which indicates that Fgf activity could be involved in mesoderm induction in vertebrate embryos. However, Fgf signalling is not sufficient for mesoderm induction in Xenopus, because embryos expressing a dominant-negative form of the activin receptor lack mesoderm but have an intact Fgf signalling pathway (Hemmati-Brivanlou and Melton,1992). The current model, based largely on experiments in Xenopus, proposes that Fgf is required either in parallel to or downstream of Nodal/Activin signals to induce and maintain mesodermal fates(Cornell and Kimelman, 1994; Labonne and Whitman, 1994; Schulte-Merker et al., 1994; Labonne et al., 1995). However, the molecular basis of the interaction between the two pathways is largely unknown. One possibility is that Fgf acts downstream of Nodal signalling as a relay mechanism required for mesoderm induction(Rodaway et al., 1999).
In addition to their role in cell-fate induction, Nodal and Fgf signalling have been implicated in cell movements, but their possible cooperation in this matter has not been studied. In zebrafish, MZoep cells do not ingress at the margin of the embryo (Carmany-Rampey and Schier, 2001), however they are able to involute in a wild-type environment (Aoki et al.,2002a). In addition, activation of Nodal signalling promotes cell-autonomous movements (David and Rosa,2001). Analysis of fgf8 and fgfR1-mutant mice revealed that Fgf signalling is involved in gastrulation movements(Ciruna and Rossant, 2001). In Xenopus, inhibiting Fgf signalling by overexpressing either a dominant-negative receptor or a MAP kinase phosphatase severely affects gastrulation (Isaacs et al.,1994; Labonne et al.,1995; Labonne and Whitman,1997). The regulation of gastrulation movements through Fgf signalling is probably a common trait of amphibians and teleosts, because overexpression of either Fgf or a dominant-negative FgfR dramatically affects gastrulation in zebrafish (Griffin et al.,1995; Rodaway et al.,1999).
The analysis of a putative interaction between Nodal and Fgf signalling pathways is complicated by the fact that both are regulated through positive and negative feedback loops. It was shown in Xenopus, zebrafish and mouse, that Nodal promotes its own expression as well as the expression of its antagonist Lefty/Antivin (Meno et al.,1999; Thisse and Thisse,1999; Hyde and Old,2000). Similarly, Fgf promotes the expression of its antagonists sprouty and sef(Furthauer et al., 2001; Furthauer et al., 2002; Tsang et al., 2002), and eFgf positively regulates its own expression through the T-box transcription factor Xbra (Isaacs et al., 1994; Labonne et al., 1995; Schulte-Merker and Smith,1995; Umbhauer et al.,1995; Casey et al.,1998).
Together, these data indicate the existence of a gene network involving Fgf and Nodal signalling components, the connectivity and function of which remain to be understood. In the present study, we show that intact Fgf signalling is required for the cell-autonomous and cell-nonautonomous induction of oep downstream of the Nodal pathway, and that intact Oep signalling is required for the cell-nonautonomous induction of the Activin/Nodal type I receptor Taram-A (Renucci et al.,1996). This regulation circuit provides a model for the involvement of the Fgf pathway in the cellular response to Nodal signals, and for the maintenance, amplification and cell-to-cell propagation of Nodal activity after the activation of the zygotic genome. Such a model predicts that lowering Fgf and Nodal levels will have cooperative, deleterious effects on the cell-nonautonomous induction of genes downstream of the Tar*/Nodal pathway. Indeed, this is what we observe for the regulation of the mesoderm marker no tail (ntl) expression in chimeric embryos. In addition, overexpression data and analysis of the oep;acedouble-mutant phenotype show that Fgf8 takes part in this process. The genetic interaction between oep and ace differentially affects cell movements and survival of distinct mesodermal cell populations and is consistent with a synergistic activity of Nodal and Fgf8 in mesoderm cells through the genetic regulation of the Nodal co-factor Oep.
Materials and methods
Fish strains
Embryos were obtained from natural spawning of wild-type (TL) or mutant fish lines. The following mutant alleles were used: oeptz57 (Hammerschmidt et al., 1996), aceti282a(Brand et al., 1996). Embryos lacking both maternal and zygotic oep function were obtained by crossing oep–/– adults that had been rescued to viability by injection of oep RNA at the one-cell stage.
Microinjection and transplantation
Synthetic mRNAs were transcribed in vitro using the SP6 mMessage mMachine™ transcription kit (Ambion). Embryos were injected at the 1-4-cell stage with tar* (2 pg), sqt (5 pg), cyc (5 pg), fgf8 (40 pg) and fgf3 (40 pg) synthetic mRNAs. fgf3 (CATTGTGGCATGGCGGGATGTCGGC 0.2 mM) and fgf8(GAGTCTCATGTTTATAGCCTCAGTA 0.5 mM) morpholino modified antisense oligonucleotides (GeneTools) were prepared and injected as described(Furthauer et al., 2001). For transplantation experiments, donor embryos were injected at the 1-4-cell stage with LacZ RNA (50 pg) and/or gfp RNA (50 pg) as lineage tracers in combination with tar* (2 pg).
Labelling hypoblast cells
A solution of 10-kD DMNB-caged fluorescein (5 mg ml–1)(Molecular Probes) was injected at the one-cell stage. When embryos reached the shield stage, the dye was activated in a few blastomeres by a microlaser beam, as described (Serbedzija et al.,1998). At the tail-bud stage, the location of the labelled cells was assessed by immunocytochemistry against fluorescein following in situ hybridization.
SU5402 treatment
Embryos were treated in the dark with either 10 μM or 30 μM SU5402(Calbiochem) from the two-cell stage until they were fixed or rinsed with embryo medium to allow development to proceed.
Acridine Orange treatment
Live embryos were treated with 5 μg ml–1Acridine Orange(Sigma) in embryo medium for 30 minutes, then washed with embryo medium and observed in epifluorescence.
Annexin V injection
Annexin-V-Alexa 488 (Nexins Research B.V., 2-5 nl, injected pure) was injected directly into the posterior hypoblast of embryos at the tail-bud stage with an eppendorf transjector (5246) and a pulled glass needle. 15 minutes after the injection, embryos were mounted in methylcellulose 3% in embryo medium and observed with a confocal microscope (Leica TCS SP2) with a 40× objective (Leica 40×/0.80 W).
In situ hybridisation and immunocytochemistry
In situ hybridisation was carried out as described(Hauptmann and Gerster, 1994)using the following mRNA: myoD(Weinberg et al., 1996); ntl (Schulte-Merker et al.,1992); tbx6 and fgf3(Hug et al., 1997); fgf8 (Furthauer et al.,1997); tar (Renucci et al., 1996); oep(Zhang et al., 1998); cyc (Rebagliati et al.,1998b); sqt (Feldman et al., 1998; Rebagliati et al., 1998a); her5(Müller et al., 1996);spt (Griffin et al.,1998); hgg1 (Thisse et al., 1994); sprouty4(Furthauer et al., 2001); and myhz1 (Xu et al.,2000). β-galactosidase was revealed with a rabbit polyclonal antibody (Cappel) at 1:1000 dilution and diaminobenzidine staining. Green fluorescent protein was revealed with a rabbit polyclonal antibody (Molecular Probes) at 1:1000 dilution and anti-rabbit CY3 (Jackson Immuno Research) at 1:500.
Genotyping
Embryos were genotyped after in situ hybridization using a PCR-based method. Single embryos were boiled in 10 μl lysis buffer (Tris-HCl 10 mM pH 7.3, KCl 50 mM, MgCl2 1.5 mM, Tween-20 3%, NP40 3%). They were then digested for 4 hours at 56°C with 1 mg ml-1 proteinase K. The latter was inactivated by boiling for 5 minutes. PCR reaction was performed using 2.5 μl of the embryo extract in a 25 μl reaction volume. The oeptz57 allele was identified using the primers tzup,5′-AGATGGAGATGTTCTAATGGTGTTTTTGGG-3′, and tzdown,5′-TGACAAATAATCACAGCAAACATCAAGAAC-3′. The PCR product was digested with MaeIII, which cuts the mutant allele only. The aceti282a allele was identified using the primers EcoRVup,5′-CTTCGGATTTCACATATTTATGCCCGTATGTATGCATATC-3′, and EcoRVdown,5′-CAGTTTTAGTAAGTCACAAAAGTGATGACTTTTTCAGATA-3′. The PCR product was digested with Eco RV, which cuts the mutant allele only.
Results
The Fgf signalling pathway is a relay in the induction of oep by the Tar*/Nodal pathway
It was shown that the Nodal pathway might be regulated through positive and negative feedback (Meno et al.,1999; Thisse and Thisse,1999; Hyde and Old,2000; Dickmeis et al.,2001). To further elucidate the genetic interactions and cellular mechanisms involved, we examined the cell-autonomous and cell-nonautonomous induction of Nodal pathway components in cells with constitutively active intracellular Activin/Nodal signalling. This was achieved by transplanting wild-type cells expressing Tar*, a constitutively active form of the Activin/Nodal type I receptor Taram-A(Renucci et al., 1996), after the injection of tar* RNA at the one-cell stage into the animal pole of wild-type embryos at the sphere stage. Assessed by in situ hybridization in the resulting chimeric embryos at the shield stage, Tar* activates cell-autonomously the expression of the genes encoding the Nodal ligands sqt and cyc, and activates both cell-autonomously and cell-nonautonomously the expression of tar and oep(Fig. 1A-D, Table 1). The type I receptor Tar was previously described as a potential receptor for the Nodal ligands Sqt and Cyc (Aoki et al., 2002b). In accord with this, we found that wild-type cells that express the Nodal ligands Sqt or Cyc have the same properties as Tar* expressing cells,activating the expression of tar and oep both cell-autonomously and cell-nonautonomously (Fig. S1A,B,G,Hand Table 1). These observations indicate that the Tar*/Nodal signalling pathway can be amplified through a positive-feedback loop and that there is a relay mechanism for the cell-nonautonomous induction of oep and tar that is downstream of the Tar*/Nodal signalling pathway.
Injected RNA . | Probe . | Wild type . | Wild type + SU5402 . | MZoep . | MZoep + SU5402 . | Wild type fgf8 MO . | MZoep fgf 8MO . |
---|---|---|---|---|---|---|---|
tar* | cyc | 55% (n=65) (auto) | 41% (n=46) (auto) | ||||
tar* | sqt | 52% (n=38) (auto) | 75% (n=24) (auto) | ||||
tar* | tar | 68% (n=38) | 86% (n=45) | 100% (n=17) (auto) | 78% (n=18) (auto) | 89% (n=19) | 87% (n=38) (auto) |
tar* | oep | 75% (n=87) | 13% (trace) (n=62) (auto) | 93% (n=42) | 20% (trace) (n=35) (auto) | 100% (weak) (n=18) | 90% (weak) (n=44) (auto) |
tar* | fgf3 | 39% (n=31) | 37% (n=17) | ||||
tar* | fgf8 | 71% (n=45) | 68% (n=19) | 35% (weak) (n=20) (auto) | 33% (weak) (n=18) (auto) | ||
tar* | ntl | 73% (n=41) | 0% (n=14) | 54% (n=15) (auto) | 0% (n=12) | ||
sqt | tar | 90% (n=11) | 100% (n=8) | 0% (n=8) | 0% (n=6) | ||
sqt | oep | 93% (n=15) | 44% (trace) (n=9) (auto) | 0% (n=8) | 0% (n=8) | ||
sqt | ntl | 92% (n=14) | 72% (trace) (n=11) (auto) | 0% (n=6) | 0% (n=7) | ||
cyc | tar | 100% (n=17) | 100% (n=11) (auto) | ||||
cyc | oep | 100% (n=23) | 60% (trace) (n=17) | ||||
cyc | ntl | 95% (n=23) | 70% (trace) (n=11) (auto) |
Injected RNA . | Probe . | Wild type . | Wild type + SU5402 . | MZoep . | MZoep + SU5402 . | Wild type fgf8 MO . | MZoep fgf 8MO . |
---|---|---|---|---|---|---|---|
tar* | cyc | 55% (n=65) (auto) | 41% (n=46) (auto) | ||||
tar* | sqt | 52% (n=38) (auto) | 75% (n=24) (auto) | ||||
tar* | tar | 68% (n=38) | 86% (n=45) | 100% (n=17) (auto) | 78% (n=18) (auto) | 89% (n=19) | 87% (n=38) (auto) |
tar* | oep | 75% (n=87) | 13% (trace) (n=62) (auto) | 93% (n=42) | 20% (trace) (n=35) (auto) | 100% (weak) (n=18) | 90% (weak) (n=44) (auto) |
tar* | fgf3 | 39% (n=31) | 37% (n=17) | ||||
tar* | fgf8 | 71% (n=45) | 68% (n=19) | 35% (weak) (n=20) (auto) | 33% (weak) (n=18) (auto) | ||
tar* | ntl | 73% (n=41) | 0% (n=14) | 54% (n=15) (auto) | 0% (n=12) | ||
sqt | tar | 90% (n=11) | 100% (n=8) | 0% (n=8) | 0% (n=6) | ||
sqt | oep | 93% (n=15) | 44% (trace) (n=9) (auto) | 0% (n=8) | 0% (n=8) | ||
sqt | ntl | 92% (n=14) | 72% (trace) (n=11) (auto) | 0% (n=6) | 0% (n=7) | ||
cyc | tar | 100% (n=17) | 100% (n=11) (auto) | ||||
cyc | oep | 100% (n=23) | 60% (trace) (n=17) | ||||
cyc | ntl | 95% (n=23) | 70% (trace) (n=11) (auto) |
A few cells taken from donor embryos injected at the one-cell stage with tar*, cyc or sqt synthetic mRNA were transplanted to the animal pole at the sphere stage into a host with the same genetic background. In some cases, a fgf8 MO was injected into the recipient embryo and in other cases the embryos were treated with SU5402 (+ SU5402) immediately after transplantation. The qualitative effect of the SU5402 treatment or of the genetic background compared with wild-type embryos is indicated (weak or trace staining). In some cases, the staining with the probe was only cell-autonomous to the transplanted cells (auto).
The secreted components of the Nodal pathway (Sqt and Cyc), which are themselves induced by Tar*, could act as a relay for the cell-nonautonomous induction of oep and tar. To investigate this, we assessed the induction of both genes in the MZoep-mutant background. That these cells are unable to transduce Nodal signals(Gritsman et al., 1999) was confirmed by the inability of Sqt-expressing cells to induce the expression of tar, oep and the T-box gene ntl in the MZoep mutant background (Fig. S1M-Oand Table 1). Although we observed cell-autonomous induction of tar by Tar* in the mutant background, its cell-nonautonomous induction was abolished(Fig. 1O). This result indicates that Oep-dependent signalling is involved in the cell-nonautonomous induction of tar. By contrast, as in wild-type chimeras, we observed cell-autonomous and cell-nonautonomous induction of oep by Tar* in the MZoep-mutant background (Fig. 1P, Table 1). This result indicates that a relay signal other than Nodal is involved in the cell-nonautonomous induction of oep by Tar*.
Previously, Fgf signalling was proposed to act as a relay signal from the Nodal-expressing mesendoderm into the mesoderm proper(Rodaway et al., 1999). Consistent with such a model, Tar* expressing cells, which have mesendodermal characteristics (Mathieu et al.,2002), could induce both cell-autonomous and nonautonomous expression of the mesoderm marker ntl in chimeric, wild-type embryos(Fig. 1G and Table 1). Cells that express Cyc or Sqt are also very potent inducers of ntl in wild-type chimeras(Fig. S1C,I). Therefore, we investigated whether Fgf signalling was involved in the response to Tar* and Nodals, using SU5402, a pharmacological inhibitor of FgfR activity(Mohammadi et al., 1997). The cell-nonautonomous induction of the mesoderm marker ntl by Tar*, Cyc or Sqt expressing wild-type cells was abolished by SU5402 treatment, and the cell-autonomous induction of ntl in the same chimeric embryos was strongly diminished (Fig. 1G,N,Fig. S1C,F,I,Land Table 1). In addition, the induction of ntl by Tar* expressing cells in MZoep embryos,was abolished by SU5402 treatment (Fig. 1R,X, Table 1). These observations are in agreement with previous studies in Xenopusdemonstrating that TGFβ-dependent induction of mesodermal genes such as ntl depends on FGF signalling(Cornell and Kimelman, 1994; Labonne and Whitman, 1994). In addition, they indicate that Fgf signalling might act as a relay signal from the Nodal-expressing cells into the neighbouring cells, and that Fgf signalling cooperates with Oep-dependent signalling to induce downstream genes.
The activation of oep and tar expression by Tar*/Nodal differs in their requirements for Oep and FGF. The cell-nonautonomous induction of oep by Tar*, Sqt and Cyc in wild-type cells was abolished by the SU5402 treatment and there remained only traces of oep cell-autonomous expression in the chimeric embryos(Fig. 1I, Fig. S1E,Kand Table 1). Similarly,cell-nonautonomous induction of oep by Tar* expressing cells was abolished in the presence of SU5402 in MZoep background and there only remained traces of cell-autonomous induction(Fig. 1V). In contrast, the induction of tar by Tar* was not affected by the SU5402 treatment in either background (Fig. 1H,U, Table 1) and the induction of cyc and sqt by Tar* in the wild-type background, was not affected by the SU5402 treatment either(Fig. 1J,K, Table 1). In summary, these observations indicate that at least two relays act in the positive-feedback loop of the Nodal signalling pathway: Nodal, which is needed for non-autonomous tar expression; and Fgf, which is needed for autonomous and nonautonomous oep expression.
Fgf3 and Fgf8 are candidates for involvement in the Nodal positive-feedback loop
Because our data indicated that Fgf signalling is involved in a relay mechanism downstream of Nodal, we predicted that the expression of specific Fgf ligands should be regulated by Nodal, and that these ligands should be involved in the regulation of endogenous oep. Both fgf3 and fgf8 are expressed at early developmental stages in marginal blastomeres and could play such a role. To investigate this possibility, we tested whether fgf3 and fgf8 could be induced by overexpression of RNA encoding Nodal ligands Sqt and Cyc. Sqt and Cyc overexpression was sufficient to induce the expression of both fgf3and fgf8 (Fig. 2A-F). In chimeric embryos, Tar*-expressing cells also induced fgf3 and fgf8 cell-autonomously and cell-nonautonomously(Fig. 1E,F, Table 1). Treatment with SU5402 had little effect on the expression of these fgf genes by Tar*-expressing cells (Fig. 1L,M, Table 1). However, fgf8 expression was induced in a higher percentage of embryos that expressed Tar* than fgf3 (twice as much), which makes Fgf8 a better candidate to act downstream of Tar*. These observations indicate that activation of the Nodal pathway is sufficient to induce expression of Fgf ligands.
In addition, we found that although the cell-autonomous induction of fgf8 by Tar*-expressing cells persisted, even on SU5402 treatment(Fig. 1M), fgf8cell-nonautonomous induction was abolished in the MZoep background(Fig. 1Q). Thus, Nodal signalling is necessary for the nonautonomous induction of fgf8 by Tar* in chimeric embryos. In addition, Nodal signalling is necessary for the early expression of fgf3 and fgf8 in uninjected embryos,particularly at the shield stage (Fig. 2N,Q). However, by mid-gastrulation, expression of fgf8in the posterior mesoderm partially recovers in MZoep mutants(Fig. 2R). These observations indicate that activation of the Nodal pathway is necessary and sufficient for the early expression of the Fgf ligands in the presumptive mesoderm, whereas the later expression of Fgf3 and Fgf8 in the neuroectoderm and Fgf8 in the posterior mesoderm depends on Nodal-independent factors.
We next asked whether Fgf signalling was involved in the regulation of endogenous oep. Again consistent with the hypothesis that Fgf3 and/or Fgf8 may act downstream of the Tar*/Nodal pathway to induce the cell-nonautonomous expression of oep, we observed that fgf3and fgf8 overexpression induced the expression of oep but not tar (Fig. 2G-L) sqt and cyc (data not shown). In addition, impairment of Fgf signalling selectively affected the expression of oep but not tar as observed in embryos treated from the one-cell stage with 30μM SU5402 (Fig. 2S-U and data not shown). However, impairment of both fgf3 and fgf8using morpholino modified antisense oligonucleotides (MOs) reduced but did not abolish oep expression at early gastrula stages (Fig. S2)indicating that other Fgfs might contribute to the regulation of oep. To further investigate the contribution of Fgf8 to the induction of oep downstream of Tar*/Nodal, we performed transplantation experiments in the context of impaired Fgf8 function by injection of a fgf8 MO at the one-cell stage into wild-type or MZoep donor and host embryos (Fig. 1S,T,Y,Z, Table 1). Induction of Tar was not affected in fgf8 MO-injected embryos, and although oep expression was much weaker than in uninjected embryos, it was not lost. Thus, fgf8 MO injection does not completely mimic the effects of SU5402 treatment.
Together, these results show that Fgf signalling is necessary and sufficient to trigger the induction of oep downstream of the Nodal pathway, and indicate that Fgf3 and Fgf8 are likely to be involved in such a relay mechanism, although other Fgfs may take part in this process.
oep and fgf8 interact genetically in vivo in the formation of mesoderm
The previous experiments indicated the presence of a previously unsuspected regulatory interaction between the Oep/Nodal and Fgf pathways and indicated the involvement of Fgf8 in this process. To further address the interaction in vivo between Nodal and Fgf8, we examined the phenotype of Zoep;acedouble mutants. The presence of maternal Oep protein enables some Nodal signalling to occur in embryos lacking Zoep function. Similarly, aceti282a is a mutant allele of fgf8, but the presence of correctly spliced full-length fgf8 message renders aceti282-mutant embryos hypomorphic in Fgf8 signalling(Reifers et al., 1998). Because both these mutations reduce but do not eliminate the corresponding activities, they can be used as sensitised mutant backgrounds to uncover processes in which both pathways may be involved.
By day one of development, the double mutants were identified easily by the additive combination of brain defects, including the absence of hypothalamus,cerebellum and midbrain-hindbrain boundary. They also exhibited a profoundly altered morphology of mesoderm derivatives in the trunk and tail, contrasting with the phenotype of Zoep and ace single mutants(Fig. 3D,G,J). The posterior trunk and tail tissue of the double-mutants appeared necrotic and was heavily stained by acridine orange (Fig. 4A-D). However, remnants of notochord and somites could be detected in the anterior trunk. Furthermore, we found by in situ hybridization that a variable number of trunk cells in most double-mutant embryos expressed myoD (Fig. 3K,M,N) and myhz1 myosin heavy chain RNA (Fig. 3L,O,P), indicating that the differentiation of some muscle fibres occurred in a region extending at most until somite 8.
These observations show that mesodermal cells are affected by the Zoep;ace loss-of-function and that some do not survive until 24 hours of development in the double-mutant background, indicating a synergistic activity of oep and ace in mesoderm formation.
The dorsal mesoderm of Zoep;ace embryos undergoes massive cell death at the end of gastrulation
We further investigated the timing of the mesodermal cell death observed in Zoep;ace mutants by inspecting live embryos during gastrulation by Nomarski video microscopy. We observed that dorsal hypoblastic cells underwent a dramatic morphological change by the end of gastrulation, losing their transparency and displaying the typical morphology of apoptotic cells(Fig. 4E-J). These cells were stained neither by acridine orange nor by TUNEL, which only marked small vesicles dispersed in the overlying ectoderm (data not shown), but they were stained by Annexin V Alexa injected directly into the extracellular space(Fig. 4G,H). The latter observation indicated that these hypoblastic cells were undergoing an apoptotic process (van den Eijnde et al.,1997). More laterally, cells with a typical hypoblastic morphology were found 100 μm from the bulk of dying cells(Fig. 4I). As development proceeded, somite condensation and notochord formation occurred in the anterior trunk and the mass of dead cells was pushed caudally(Fig. 4J and data not shown). Concomitantly, numerous cell corpses were incorporated into the yolk cell, as indicated by Nomarski images (Fig. 4J,L).
We conclude from these observations that the mesoderm is subdivided into populations that differ in their requirement for cooperation between the Oep and Fgf8 pathways. Most of the dorsal mesoderm undergoes massive cell death by the end of gastrulation in Zoep;ace mutants and this aspect of the double-mutant phenotype is fully penetrant. More lateral mesoderm cells,although somewhat scattered on the yolk syncytium, displayed a normal morphology by the end of gastrulation (Fig. 4I).
Cooperation between oep and ace is required before the onset of gastrulation to maintain axial mesoderm fate
As described above, the dorsal mesoderm underwent massive cell death by the end of gastrulation in Zoep;ace mutants. However, some cells were able to contribute to anterior notochord and somites and we hypothesized that they expressed mesoderm markers. To investigate this, we analyzed the expression of the axial and adaxial mesoderm markers ntl and myoD, respectively, at the end of gastrulation(Fig. 5A,B). In the double mutant, myoD was undetectable by this stage(Fig. 5B). Staining of ntl was detected at the blastoderm margin and was very limited and scattered in the embryonic axis (Fig. 5B). In addition, we found that the anterior limit of ntlstaining along the rostro-caudal axis did not reflect the hypoblast involution that occurred in the double mutants. The hypoblast involution was assessed in wild-type and Zoep;ace embryos by labelling cells in the embryonic shield. This was achieved by locally uncaging caged-fluorescein injected at the one-cell stage in the progeny of double-heterozygous fish. By the end of gastrulation, treated embryos were stained with ntl and uncaged fluorescein was revealed by immunocytochemistry(Fig. 5C,D). From the observation of the cell tracer, we concluded that the rostro-caudal extension of the axial hypoblast in the double-mutants(Fig. 5D) was significantly less than in wild-type (Fig. 5C) and single mutants (data not shown). However, the defects in involution do not readily explain the loss of ntl and myoDexpression in the hypoblast of Zoep;ace mutants. Rather, we hypothesize that Oep and Fgf8 cooperate in maintaining dorsal mesodermal markers.
We further investigated the timing of a putative cooperation between Oep and Fgf to maintain ntl expression earlier during gastrulation by treating the progeny of Zoep heterozygous fish with SU5402 and by varying time windows of exposure to the drug. Our experiments revealed a crucial period between sphere stage and 50% epiboly (onset of gastrulation)during which intact Fgf signalling is required for Zoep embryos to retain a normal ntl-expression domain, whereas wild-type embryos are essentially unaffected by this specific treatment(Fig. 5E-H).
The phenotype of Zoep;ace mutants reveals a cooperation between Oep and Fgf8 pathways differentially affecting the morphogenesis of various mesodermal territories. In particular, we conclude that the axial hypoblast requires cooperation between Oep and Fgf8 before the onset of gastrulation to maintain the expression of the notochord marker ntl and insure the survival of dorsal mesodermal cells.
Cooperation between oep and fgf8 acts differentially on the induction of mesoderm markers but is mainly required for the maintenance of mesodermal fates
The above results did not address whether Oep and Fgf8 act cooperatively in the initial phase of mesoderm induction. To investigate this, we analyzed the expression of the T-box mesoderm markers ntl, tbx6 and spadetail(spt) before the onset of gastrulation in Zoep;ace mutants(Fig. 6). No difference was seen between wild-type and single mutants, but the latter differed slightly from Zoep;ace double mutants in the levels of expression of ntl at 30% epiboly (Fig. 6A,G) and of tbx6 at 30% epiboly and shield stage(Fig. 6C,I,F,L). By contrast,the expression of spt was not obviously affected at either stage(Fig. 6D,E).
Correlating with their overall normal morphology(Fig. 4I and data not shown),lateral mesoderm cells expressed the T-box markers ntl, tbx6 and spt (Kimelman and Griffin,2000) during gastrulation in Zoep;ace mutants. However,their expression domains were markedly reduced in the double mutants at 80%epiboly compared to wild-type or single mutant embryos(Fig. 6M-O,S-U). At the 12-somite stage, very faint expression of the three markers was retained in the protruding tail bud of the double mutants(Fig. 6V-X,XX-XZ).
These observations indicate that oep and fgf8 act synergistically on the maintenance of all the mesoderm markers tested. Their combined action on induction of T-box genes is rather subtle but suggests a differential regulation with no effect on spt expression, a delay in the amplification of ntl and a downregulation of tbx6 that can be detected at all stages.
However, the induction of mesoderm still occurs in the absence of the zygotic contribution of oep and fgf8. We hypothesized that Oep maternal contribution rescued earlier aspects of the formation of mesoderm. To assess this, we investigated the phenotype of MZoep;aceembryos. By 30 hours of development they were devoid of somites in the tail(Fig. 7B) unlike MZoepmutants (Fig. 7A), confirming that Oep and Fgf8 also cooperate in the maintenance of the presumptive tail mesoderm and consistent with our observation that fgf8 is expressed in the posterior mesoderm during gastrulation(Fig. 2R).
Regarding early steps of mesoderm formation, ntl expression was diminished by 30% epiboly in MZoep;ace embryos compared to MZoep siblings (Fig. 7C,H). Although the difference was stronger than that observed at the same stage comparing Zoep;ace embryos and their wild-type siblings (Fig. 6), the enhanced effect of MZoep;ace on ntl was not detected by the shield stage (Fig. 7D,I). Expression of tbx6 was highly variable in MZoep mutants throughout epiboly (Fig. 7M,N and data not shown), and removing zygotic ace activity did not obviously affect this aspect of the mesodermal phenotype at earlier stages(Fig. 7R,S and data not shown). The strong, ubiquitous expression of spt prior gastrulation did not allow us to distinguish differences between MZoep mutants and MZoep;ace siblings (Fig. 7F,K). By the onset of gastrulation, we found that sptwas expressed at slightly lower levels in MZoep;ace mutants compared to MZoep embryos (Fig. 7G,L). In summary, analysis of the expression of T-box genes in MZoep;ace mutants at the end of gastrulation reinforced our conclusion that cooperation between oep and fgf8 acts mainly to maintain mesodermal fates rather than initiate mesoderm induction. Indeed,expression of ntl, spt and tbx6 was abolished by 90% epiboly in MZoep;ace mutants (Fig. 7J,Q,V).
Discussion
Several transduction pathways operate during early embryogenesis to generate cell diversity and coordinate cell behaviours. In particular,Activin/Nodal and Fgf are involved in the formation of mesoderm. It has been proposed that Fgf signalling renders cells competent to respond to Activin(Cornell and Kimelman, 1994; Labonne and Whitman 1994), but the underlying mechanisms have remained elusive. Here we propose a mechanism accounting for this aspect of the interaction between Fgf and Nodal pathways. In our model, Fgf serves as a relay downstream of Nodal, which regulates the expression of oep. Such a process is involved in our transplantation experiments and is also likely to act in normal embryos. This would mean that Fgf signalling is required for the maintenance, amplification and cell-to-cell propagation of the Nodal pathway as part of its positive-regulatory loop. In addition, a parallel can be made between the mesendoderm-mesoderm interaction and chimeric embryos where tar*-expressing cells act on their neighbours. Indeed, cells that express tar* are fated to mesendoderm(Peyrieras et al., 1998; Mathieu et al., 2002) and are likely to mimic some of its properties(Rodaway and Patient, 2001). This is illustrated by the requirement for Fgf and Nodal signalling as relays for the cell-nonautonomous induction by Tar*-expressing cells of oepand tar respectively that parallels the simultaneous requirement for Fgf and Nodal signalling in the mesoderm proper. Such a model is consistent with the phenotypic defects observed in oep and ace single and double mutants, which show that Nodal and Fgf8 act synergistically in mesoderm formation. We further demonstrate that mesodermal cell populations differ in their requirement for the cooperation between oep and fgf8, indicating regional differences in the dynamics of the Nodal network.
Fgf is required for the amplification and propagation of Nodal signalling through a positive regulatory loop
Our overexpression data indicates a model in which Fgf signalling is required both cell-autonomously and cell-nonautonomously for the activation of oep expression downstream of Tar*/Nodal. In this model, the mechanisms and dynamics of the positive regulation loop of the Nodal pathway depend on the role of Oep. Analysis of zebrafish and mouse mutants indicates that this EGF-CFC co-factor is essential for Nodal function during early embryogenesis (Whitman, 2001). However, it has been suggested recently that Nodal signalling might function in the absence of EGF-CFC co-factors(Reissmann et al., 2001). Such a hypothesis is not supported by our transplantation experiments. Indeed,cells that expressed sqt had no detectable inducing activity in the MZoep background. In any case, the cell-nonautonomous induction of both tar and oep by Tar*-expressing cells indicates that Nodal signalling might propagate to the neighbouring cells via the secreted ligand Sqt because the latter can act 10 cells away from its source(Chen and Schier, 2001). However, Tar*-expressing cells induce the expression of cyc and sqt cell-autonomously only, indicating that the activity of Nodal signalling in neighbouring cells is modulated, and that the propagation of Nodal signalling is restricted. This could result from the negative regulation of the pathway. Indeed, the Nodal antagonist Antivin is induced downstream of Tar*/Nodal (Meno et al., 1999; Thisse and Thisse, 1999; Hyde and Old, 2000) and has been shown to limit the range of action of Sqt(Meno et al., 2001).
We assessed the potential role of Fgf signalling downstream of Nodal by using SU5402, inhibitor of the Fgf pathway(Mohammadi et al., 1997). Any Fgf molecules present between the sphere stage and 50% epiboly at the margin of the blastoderm in the mesodermal precursors could, potentially, account for the drug effects. Although we have reasons to implicate Fgf8, the high doses of SU5402 required to inhibit oep expression at the onset of gastrulation indicate contributions by other Fgfs. We explored the possibility of a combined role of FGF3 and FGF8 by using specific morpholinos(Maroon et al., 2002) and concluded that Fgf3 and Fgf8 account for at least part of the Fgf activity in early steps of mesoderm formation, but that other, as yet unidentified, Fgfs might be involved.
Oep/Nodal and Fgf8 signalling pathways act synergistically and their cooperation differentially affects the expression of mesoderm-marker genes
We discussed above a role for Fgf downstream of Nodal and suggested that tar, fgf and oep are connected directly through a cascade of gene activation. Several mechanisms operating at the genetic and epigenetic level might account for a strict requirement of fgf8 downstream of,or in parallel to oep. By contrast, analysis of the zebrafish oep and ace single and double-mutant phenotypes demonstrates that Oep and Fgf8 cooperate in the formation of mesoderm. This means that both pathways also act in parallel, but in a redundant way rather than with a strict requirement for each other. We suggest that both types of interaction between oep and fgf8 coexist within the nodal/fgfmolecular network. Cooperation between Nodal and Fgf pathways might be achieved at the genetic level, through the interaction of their downstream components with the regulatory sequences of common target genes. This might happen for the regulation of the T-box genes. This hypothesis fits with our expression data in the Zoep and ace single and double-mutant backgrounds.
Analysis of single and double mutants of oep and aceindicates that several aspects of mesoderm formation depend on the Nodal/Fgf8 cooperation. However, the latter is not crucial for induction of mesoderm and initiation of the expression of the T-box genes ntl, spt and tbx6, although it does modulate their level of expression, probably through an amplification step. However, the Nodal/Fgf8 interaction is essential for the maintenance of mesodermal cell populations. This is shown by the loss of T-box gene expression during somitogenesis in Zoep;acemutants, and even earlier in MZoep;ace mutants. The timing of oep and fgf8 interaction, determined from our SU5402 incubation experiments, reveals an early requirement, which correlates with the later maintenance of ntl expression in the embryonic axis. It fits with the detection of fgf8 RNA at the margin of the blastoderm before gastrulation. Later, fgf8 staining is excluded from the embryonic axis during gastrulation and retained in the presomitic mesoderm. At these stages, the Oep-Fgf8 cooperation might occur at the blastoderm margin where both are expressed, and act on the maintenance of nonaxial mesoderm.
Zoep;ace and oep;ntl double mutants have a dramatic deficit in the formation of mesoderm(Schier et al., 1997). The similarities between the phenotypes of the two mutants indicate that ntl has a major role downstream of fgf8 and probably also parallel to it through a positive regulatory loop, as shown for eFgf and Xbra(Cornell and Kimelman, 1994; Labonne and Whitman, 1994; Schulte-Merker et al., 1994; Labonne et al., 1995). This also fits with the hypothesis that the T-box genes ntl, spt and tbx6 are interconnected, either as common targets of oep and fgf8, or because they interact at the genetic level to regulate each others expression (Kimelman and Griffin,2000).
Expression of myoD reveals other aspects of the cooperation between Oep and Fgf8
As in ntl mutants, the early phase of myoD expression in the adaxial cells (Weinberg et al.,1996), which extends from midgastrula to prior to somite formation, is abolished on combined zygotic reduction of function of oep and ace. However, myoD expression occurs at later stages in Zoep;ace mutants. Notably, by one day of development the myoD-expression domain reveals a transition in the trunk mesoderm at the 6–8-somite level. This rostro-caudal transition is distinct from the trunk-tail transition at the level of somite 18, which has been discussed previously (Kimelman and Griffin,2000), and is reminiscent of that described in mouse embryos(Soriano, 1997). Recently, it has been suggested (Holley et al.,2002) that in both mouse and zebrafish mutants affected in the formation of somites show an anterior to posterior polarity in phenotype, with anterior regions less severely affected than posterior regions. This rostro-caudal transition may reflect profound morphogenetic differences between the anterior and posterior somites. The first six somites are known to form more rapidly and display a more synchronous morphogenesis than the posterior somites and it has been suggested that their formation might be independent of oscillating gene expression(Holley et al., 2002). Here,we suggest that this rostro-caudal transition also affects the axial mesoderm because the notochord does not form at more posterior rostro-caudal levels in Zoep;ace mutant embryos.
Differential requirement for the synergy between ace and oep defines distinct mesodermal territories
Our data point to regional differences in the behaviour of mesodermal cells that correlate with their requirement for the cooperation between oepand ace. The fate of lateral mesoderm is maintained until mid-somitogenesis in the absence of oep and fgf8 zygotic components, but it is disrupted during gastrulation when maternal oepfunction is also removed.
Except for a few cells that either form notochord or undergo somite condensation and myogenic differentiation, cells of the dorsal mesoderm require cooperation between zygotic oep and fgf8 before the onset of gastrulation in order to survive. Indeed, we observed that in a domain that probably encompasses the notochord and paraxial mesoderm territories, cells undergo a dramatic morphological change at the time of epiboly completion. Our Annexin V binding data indicates that these hypoblastic cells might undergo an apoptotic process with the exposure of phosphatidyl-serine on the outer lipid leaflet of the cell membrane. The link between this cell behaviour and disruption of the Oep and Fgf8 pathways remains to be elucidated. Subsequent morphogenesis in the Zoep;acemutants is profoundly affected by the presence of this mass of dying cells,even though the yolk cell can engulf numerous cell corpses.
Supplemental data available online
Acknowledgements
We are grateful to Jacqueline Becker for excellent technical assistance. We thank Samuel Sidi for critical reading of the manuscript and Sophie Vriz for helpful discussions. We are grateful to our colleagues in the community for reagents and to F. Bouallague for fish care. This study was supported by grants from CNRS (NP,US), MNHN (NP), INSERM and ARC (NP,FR,US), AICR (US), and the National Science Foundation (KG,DK). JM received a fellowship from the Ligue Nationale Contre le Cancer and TD was supported by Boehringer Ingelheim and AFM.