FRL-1, a member of the EGF-CFC family, is essential for neural differentiation in Xenopus early development

The EGF-CFC family of proteins are secreted signaling molecules important in early vertebrate development. They have a variant Epidermal Growth Factor(EGF)-like domain and a CFC domain that is conserved among mouse Cripto, XenopusFRL-1 and mouse Cryptic(Shen and Schier, 2000). EGF-CFC proteins contain a hydrophobic C terminus and are connected to the cell membrane via a glycosylphosphatidylinositol (GPI) linkage(Zhang et al., 1998; Minchiotti et al., 2000). The EGF-CFC family of proteins include human Cripto and Cryptic, mouse Cripto and Cryptic, chicken Cripto, Xenopus FRL-1 and zebrafish One-eyed pinhead(Oep) (Ciccodicola et al.,1989; Dono et al.,1993; Shen et al.,1997; Colas and Schoenwolf,2000; Kinoshita et al.,1995; Zhang et al.,1998). Loss-of-function experiments in zebrafish show that the one-eyed pinhead (oep) mutant phenotype is similar to that of the zebrafish double mutants of the nodal-related genes, squint and cyclops, indicating that the EGF-CFC gene is an essential cofactor in nodal signaling(Gritsman et al., 1999; Schier and Shen, 2000).

The Xenopus EGF-CFC protein FRL-1 was isolated as a ligand of the fibroblast growth factor (FGF) receptor and is considered a member of the EGF-CFC protein family (Kinoshita et al.,1995; Shen et al.,1997). FRL-1 is expressed ubiquitously during gastrulation and can induce neural and mesodermal markers in presumptive ectoderm (Kinoshita et al.,1995). Studies of FGF signaling suggest the existence of two signal transduction pathways mediated by the FGF receptor. First, activation of the FGF receptor can activate the phospholipase Cγ (PLCγ)pathway to produce inositol triphosphate (InsP₃) and facilitate Ca²⁺ release (reviewed by Powers, 2000). Second, the FGF receptor also signals to Ras, which subsequently activates the mitogen-activated protein kinase (MAPK) signaling pathway (reviewed by Powers,2000). It remains unclear which of these pathways interacts with FRL-1.

In vertebrate early development, bone morphogenetic protein (BMP) signaling stimulates epidermal induction of undifferentiated cells in the presumptive ectoderm region and inhibits neural induction(Hemmati-Brivanlou and Melton,1997). Neural cells are induced in the presumptive ectoderm when BMP signaling is inhibited by factors such as Noggin, Chordin and Follistatin,which are secreted from the organizer region that differentiates into axial mesoderm. Initially, neural-inducing signals from the organizer are thought to act in an instructive manner. They can, however, also bind directly to BMP proteins. This leads to the `default model' of neural induction, which proposes that in the absence of cell-cell signaling, ectodermal cells will adopt a neural fate (Munoz-Sanjuan and Hemmati-Brivanlou, 2002).

In this study, we have examined FRL-1 function in early Xenopusdevelopment and find that it acts as a neural inducer. FRL-1 inhibits BMP signaling via the activation of MAPK signaling, implicating it early in neural induction. These data indicate that FRL-1-induced neural differentiation may occur via a nodal-independent mechanism.

Eggs were obtained by injecting human chorionic gonadotropin (Gestron:Denka Seiyaku) into Xenopus laevis. Fertilized eggs were obtained by artificial insemination and dejellied using 1% thioglycolate in Steinberg's solution. Microinjection was carried out according to the method of Tanegashima et al. (Tanegashima et al.,2000). All injection experiments were performed more than two times independently. In the animal cap assay, mRNAs were injected into each blastomere of 4-cell-stage embryos, whose animal caps were dissected with fine needles on a 0.5 mm square and cultured in 100% Steinberg's solution with 0.1%bovine serum albumin until sampling. For western blot analysis of MAPK, animal caps were dissected in Ca²⁺- and Mg²⁺-free modified Barth's solution to prevent wounding. Embryos were staged according to Nieuwkoop and Faber (Nieuwkoop and Faber,1956).

PBluescript FRL-1 (pBS-FRL-1) was obtained by screening our cDNA library containing maternal genes. The construction of pCS2-FRL-1Δ5′UTR was performed by amplifying the FRL-1 ORF using PCR. For the construction of pCS2-FRL-1,which contains the 5′UTR region, EcoRI- and XhoI-digested pBS-FRL-1 was ligated into pCS2 vector. To test FRL-1 function (Figs 1, 10 and 11), we used pCS2-FRL-1Δ5′UTR as FRL-1 because of its strong activity. For construction of pCS2-FRL-1Δ5′UTR-6myc and pCS2-FRL-1-6myc, FRL-1 cDNA fragments were amplified by PCR from pCS2-FRL-1Δ5′UTR and pCS2-FRL-1,respectively and the amplified products were ligated into pCS2+6myc vector. pCS2-FRL-1ΔCFC-6myc and pCS2-FRL-1ΔEGF-6myc constructs coded FRL-1-6myc protein without CFC domain (116Cys-stop codon),and EGF domain (77Lys-112Arg), respectively. For construction of pCS2-oep,oep ORF was amplified from pCDNA-oep-flag(Zhang et al., 1998) and ligated into pCS2+ vector. For construction of pCS2-chickCFC,pKS-chickCFC (Andree et al.,2000) was digested with PstI. The released fragments were blunt-ended and ligated into pCS2 vector. For the preparation of mRNA,pCS2-chordin, pCS2-lacZ(Takahashi et al., 2000),pCS2-FRL-1, pCS2-FRL-1Δ5′UTR,pCS2-FRL-1-6myc, pCS2-FRL-1Δ5′UTR-6myc,pCS2-FRL-1ΔCFC-6myc, pCS2-FRL-1ΔEGF-6myc,pCS2-cripto (Yeo and Whitman,2001), pSP64T-BMP-4, pSP64T-XFD(Amaya et al., 1991) and pSP64T-SESE-MAPKK (Gotoh et al.,1995) were linearized and transcribed using the mMASSAGE mMACHINE SP6 kit (Ambion).

The FRL-1 morpholinoantisense oligonucleotides (FRL-1MO), FRL-1MO-second and FRL-1-4misMO were designed as follows; FRL-1MO:5′-AAACTGCATTGTTTTCTGCAAAGGC-3′; FRL-1MO-second:5′-ATTGAATGTGTCCTTAGCAAAAACC-3′, FRL-1-4misMO:5′-AAACaGCATaGTTTTgTGCAgAGGC-3′, lower case letters indicate changes to the FRL-1MO-first oligo sequence (Gene Tools LLC).

Total RNA isolation, RT-PCR and histological analyses were performed as described previously (Tanegashima et al.,2000). The primer pairs used were as follows; soxD(forward, 5′-TCAGCAACAGGTCCAGTACC-3′; reverse,5′-TCTAACAAGATCCGACGCC-3′), FRL-1 (forward,5′-ATGCAGTTTTTAAGATTT-3′; reverse,5′-TTAAAGTCCAATATT-3′), endodermin (edd) (forward,5′-TATTCTGACTCCTGAAGGTG-3′; reverse,5′-GAGAACTGCCCATGTGCCTC-3′), collagen type II(col II) (forward, 5′-ATTCAGTTGACCTTCCTGCG-3′; reverse,5′-TCCATAGGTGCAATGTCTACG-3′), Xngnr-1 (forward,5′-CGCCGCAACCCGACTCACCT-3′; reverse,5′-CCTGCATCGCGGGCTGTTCTC-3′), Xvent-1 (forward,5′-TTCCCTTCAGCATGGTTCAAC-3′; reverse,5′-GCATCTCCTTGGCATATTTGG-3′). The primer pairs for chordin,Xmsx-1, β-crystallin, ms-actin, EF1-α, otx2,zic3, N-tubulin, Xbra and N-CAM were as described in Xenopus Molecular Marker Resource(http://www.cbrmed.ucalgary.ca/pvize/html/WWW/Welcome.html). EF1-α was used as a loading control. Reverse transcriptase negative (RT–) reactions were performed to indicate the absence of contaminating genomic DNA.

Analysis of whole-mount in situ hybridization was carried out as described previously (Harland, 1991)using digoxigenin-labeled antisense probes. For cell-lineage tracing, 20 ng of FRL-1MO or 20 ng of FRL-1-4misMO were co-injected with 250 pg of lacZinto one animal blastomere of albino embryos. Injected regions were stained with red gal (Research Organics, Inc, USA). Probes were synthesized using pBS-chordin (Sasai et al.,1994), pBS-zic3, pBS-soxD, pGEM-otx2,pBS-Xngnr-1 (Ma et al.,1996) and pBS-FRL-1 as templates. PBS-soxD and pBS-zic3 were obtained by our own screening.

FRL-1 proteins with 6myc-epitope tags were detected using the 9E10 monoclonal antibody (Santa Cruz). α-actin served as the loading control and was detected by the AC-40 monoclonal antibody (Sigma). Activated MAPK was detected by monoclonal anti-MAPK, activated (diphosphorylated ERK-1 and 2)clone MAPK-YT (Sigma) and total MAPK protein were detected by ERK2, rabbit polyclonal IgG (Santa Cruz). Anti-mouse IgG, HRP-linked antibody (Cell Signaling Technology, Inc.) and peroxidase-conjugated goat anti-rabbit IgG(Sigma) were used as the secondary antibodies.

Neu-1 monoclonal antibody was described previously(Itoh and Kubota, 1989). After detection, wild-type embryos with Neu-1 staining were bleached. For the detection of activated MAPK, whole-mount in situ immunohistology was done using monoclonal anti-MAPK, activated (diphosphorylated ERK-1 and 2) clone MAPK-YT (SIGMA). The signal was detected using BM purple (Roche).

Purified bFGF (Amersham Biosciences), PD98059 (Biomolecules for Research Success) and LY294002 (Biomolecules for Research Success) were purchased commercially. PD98059 and LY294002 were dissolved in DMSO and stored at 10 mM concentration.

FRL-1 induces expression of the pan-neural marker, N-CAM(Kintner and Melton, 1987) in animal caps and the mesodermal marker, muscle specific-actin(ms-actin) (Stutz and Spohr,1986) at high doses (Kinoshita et al., 1995). However, it is unclear how FRL-1 functions in the earlier stages of neural induction. To analyze this, we investigated whether FRL-1 induces the expression of early neural and mesodermal markers in animal caps. RT-PCR analysis indicated that 1 ng of FRL-1induced the early neural markers, zic3(Nakata et al., 1997), soxD (Mizuseki et al.,1998) and Xngnr-1 (Ma et al., 1996) but not the pan-mesodermal marker Xbra(Smith et al., 1991) or the dorsal mesodermal marker chordin(Sasai et al., 1994) in animal caps at the early neurula stage (Fig. 1A, lane 2). In addition, FRL-1 suppressed the expression of BMP responsive genes, Xvent-1(Gawantka et al., 1995) and Xmsx-1 (Maeda et al.,1997; Suzuki et al.,1997) (Fig. 1A,lane 2). These results indicate that overexpression of FRL-1 causes neural induction without mesoderm induction as does the BMP antagonist, chordin (Fig. 1A, lane 3), suggesting that FRL-1 acts as a neural inducer by inhibiting BMP signaling.

To test this hypothesis, we used the animal cap assay to examine whether FRL-1 inhibited early BMP response genes induced by BMP-4(Koster et al., 1991). BMP-4 induced the expression of Xmsx-1, Xvent-1 and Xbra at the early gastrula stage(Fig. 1B, lane 2), whereas FRL-1 completely inhibited the expression of BMP response genes induced by BMP-4 (Fig. 1B, lane 3).

Under our experimental conditions, FRL-1 induced neural induction without mesoderm induction at an early stage. We also tested whether FRL-1 induced the expression of late neural markers and mesodermal markers. FRL-1 induced the expression of N-CAM and the anterior neural marker otx2(Pannese et al., 1995) in a concentration-dependent manner, but not ms-actin and collagen type II (Su et al.,1991), which were expressed in muscle and notochord, respectively,even when FRL-1 was used at high doses(Fig. 1C). These results indicate that FRL-1 can directly induce anterior neural tissue without the induction of mesoderm.

EGF-CFC proteins have two conserved motifs, the EGF-like and the CFC motif(Shen and Schier, 2000). To examine the function of these motifs in neural induction, two cDNA constructs encoding an EGF-like motif-deleted form with the 6myc-epitope tag(FRL-1ΔEGF domain) and a CFC motif-deleted form with the 6myc-epitope tag (FRL-1ΔCFC domain) were generated and used for the animal cap assay. Western blotting confirmed expression of these protein products in the injected animal caps. Overexpression of mRNA encoding FRL-1 with 6myc-epitope tag in ectoderm induced the neural marker N-CAM.However, the FRL-1ΔEGF domain or FRL-1ΔCFC domain showed no such inductive ability(Fig. 2A). This suggests that both the EGF-like and CFC domains of FRL-1 are required for neural induction.

We next examined whether the neural induction activity of FRL-1was conserved among the other EGF-CFC genes(Fig. 2B). Overexpression of oep (a zebrafish EGF-CFC gene) neuralized animal caps, as previously reported (Kiecker et al.,2000), while overexpression of other EGF-CFC genes, mouse cripto, and chick CFC (chick cripto) induced the expression of neural marker, N-CAM without the induction of mesodermal marker, ms-actin. The N-CAM expression induced by these EGF-CFC genes was weaker than that induced by FRL-1 (data not shown), which may reflect receptor differences between species. Overall, our results suggest that the function in neural induction is conserved among EGF-CFC genes.

We clearly show that FRL-1 acts as a neural inducer by inhibiting BMP signaling in the animal cap. To study the role of FRL-1 in neural induction in Xenopus embryos, we generated an antisense morpholino oligonucleotide against FRL-1 (FRL-1MO) and then tested whether it specifically inhibited the translation of FRL-1 using the Xenopus oocyte expression system(Fig. 3). First, FRL-1-6myc proteins, consisting of FRL-1 tagged with the 6myc-epitope at the C-terminal region were detected at approximately 35 kDa and 25 kDa(Fig. 3, lanes 2, 5). A 25 kDa protein is consistent with the predicted molecular size of the FRL-1-6myc protein. The 35 kDa band may represent a glycosylated form of FRL-1-6myc since the mouse EGF-CFC protein, Cripto is known to be fucosylated(Schiffer et al., 2001). FRL-1MO was shown to inhibit the translation of FRL-1(Fig. 3, lane 3), but not FRL-1-4misMO, which contains four nucleotide substitutions to exclude the toxicity of MO or FRL-1Δ5′UTR that lacks the target sequence of FRL-1MO (Fig. 3,lanes 4, 6). These results imply that the FRL-1MO specifically inhibited FRL-1 translation.

To confirm our results with misexpression of FRL-1, we performed a loss-of-function study by microinjecting FRL-1MO into Xenopus embryos(Fig. 4). Embryos that had been injected animally into each blastomere at the 8-cell stage with 15 ng of FRL-1MO showed head defects with no eye structures(Table 1, Fig. 4B). Injection of 15 ng of FRL1-4misMO animally or with 15 ng of FRL-1MO vegetally had no apparent inhibitory effects (Table 1). We also generated a second antisense MO against FRL-1, which was designed against another region of the 5′UTR of the FRL-1 mRNA(FRL-1MO-second). Embryos injected with 40 ng of FRL-1MO-second showed a phenotype similar to those injected with FRL-1MO(Table 1, Fig. 4G). Head structure defects with lack of eyes, caused by microinjection of either FRL-1MO or FRL-1MO-second, could be rescued by co-injection of 20 pg of FRL-1Δ5′UTR (Table 1, Fig. 4D,H). These results suggest that the phenotypes caused by microinjection of the MOs against FRL-1 were due specifically to the depletion of FRL-1. Next, we tested the possibility that other EGF-CFC genes or BMP antagonists, such as chordin, noggin or FRL-1ΔEGF and ΔCFC domain could rescue the FRL-1-depleted phenotype. The phenotypes of the embryos injected with FRL-1MO were rescued by co-injection with oep mRNA (Table 1, Fig. 4E) and cripto (Table 1, Fig. 4F), but not with chordin, noggin, FRL-1ΔEGF domain or FRL-1ΔCFC domain (Table 1), suggesting that other EGF-CFC genes could compensate for the depletion of FRL-1.

Histological analysis was performed to investigate these phenotypes further. The embryos injected with FRL-1MO had neural tissue defects without the loss of mesodermal tissue, such as notochord and pronephric tube(Fig. 5G). However,FRL-1-4misMO-injected embryos showed no significant defects(Fig. 5H). Whole-mount in situ hybridization experiments showed that the expression of otx2 was completely suppressed by FRL-1MO-injected embryos, confirming that FRL-1MO blocked the formation of anterior neural tissue in Xenopus embryo(Fig. 5A). The inhibition of neural induction was also confirmed by whole-mount in situ immunostaining using Neu1 antibody that specifically recognizes neural tissue. The staining showed that the injection of both FRL-1MO and FRL-1MO-second inhibits neural induction but not FRL-1-4misMO (Fig. 5D-F). RT-PCR analysis also showed that FRL-1Δ5′UTR could rescue the FRL-1MO-induced suppression of neural markers, N-CAM, otx2, β-crystallin(Altmann et al., 1997) and N-tubulin (Richter et al.,1988; Good et al.,1989; Hartenstein,1989), but not the mesodermal markers, ms-actin and collagen type II and endodermal marker, endodermin(Sasai et al., 1996)(Fig. 6).

To investigate whether FRL-1 is involved in neural induction at an early stage, we observed the expression of early neural genes in FRL-1-depleted embryos (Fig. 7). FRL-1MO (20 ng) or FRL-1-4misMO (20 ng) was co-injected with cell lineage tracer, lacZ into one blastomere of 4-8 cell-stage embryos, which were cultured until early gastrula stage for chordinand early neurula stage for zic3, soxD and Xngnr-1. Whole-mount in situ hybridization analysis showed that the expression of early neural markers, zic3, soxD and Xngnr-1 was suppressed in the FRL-1MO-injected regions that were stained red for β-galactosidase activity (Fig. 7C-H),suggesting that FRL-1MO inhibits the early step of neural induction. However,the expression of chordin, which acts as a neural inducing signal from dorsal mesoderm, was not suppressed in the FRL-1MO-injected region(Fig. 7A,B). Therefore, these results indicate that FRL-1 is required for neural differentiation at an early stage, and does not affect the expression of the neural inducer, chordin.

Neural induction occurs autonomously when BMP signaling is inhibited in the presumptive ectoderm of Xenopus embryos(Munoz-Sanjuan and Hemmati-Brivanlou,2002), suggesting that ectodermal cells originally are competent to adopt a neural fate and that a set of genes to drive that fate is stored in the presumptive ectoderm maternally. FRL-1 is expressed maternally and ubiquitously in Xenopus early gastrulae(Kinoshita et al., 1995) and is required for neural induction during the early stages of embryogenesis. Therefore, it was thought that FRL-1 might confer the neural fate competence in the ectoderm. To test this hypothesis, we examined the response to neural inducing signals in FRL-1-depleted ectoderm. Chordin induced the expression of the neural markers, N-CAM,otx2 and N-tubulin in animal caps and the induction was not inhibited by FRL-1-4misMO (Fig. 8, lanes 2, 3). However, FRL-1MO did inhibit the expression of N-CAM, Otx2 and N-tubulin in response to the neural inducing signal of chordin (Fig. 8, lane 4). The attenuation of neural competence was rescued by FRL-1Δ5′UTR (Fig. 8, lane 5). These results indicate that FRL-1 confers a competence in the ectoderm to respond to neural inducing signals.

FRL-1 is expressed ubiquitously in early gastrula(Kinoshita et al., 1995). However, no spatial expression of FRL-1 in late gastrula has been reported. We examined the relationship between FRL-1 expression and neural induction using whole-mount in situ hybridization. Our analysis showed that FRL-1 was ubiquitously expressed in the early gastrula(Fig. 9A), as previously reported (Kinoshita et al.,1995), whereas it was confined to the presumptive neuroectoderm in late gastrula (Fig. 9B). RT-PCR confirmed the restricted expression of FRL-1 to presumptive neuroectoderm that was induced by chordin. Expression levels of FRL-1 were higher in chordin-injected ectoderm(Fig. 9C, lane 2), which differentiates into neural tissue without mesoderm(Fig. 1A, lane 3; Fig. 9C, lane 2), than in the uninjected ectoderm, when sibling embryos grew to late gastrula. It was previously reported that transcripts of FRL-1 were expressed until the gastrula stage and that these were rapidly reduced by the end of gastrulation (Kinoshita et al.,1995). Our results suggest that the expression of FRL-1is attenuated in presumptive epidermis ahead of the neuroectoderm and is needed to ensure a neural cell fate.

We showed here that FRL-1 regulates the early step of neural induction by inhibiting BMP signaling (Fig. 1A,B). FRL-1 in animal caps can be blocked by a dominant-negative FGF receptor-1 construct (XFD)(Kinoshita et al., 1995). Our results also confirm that the N-CAM-inducing activity of FRL-1 is required for the FGF receptor signaling pathway(Fig. 10A). Since FGF receptor signaling can activate the MAPK pathway(Powers et al., 2000), we therefore examined whether FRL-1 could activate the MAPK pathway using a monoclonal antibody against activated MAPK. Western blot analysis showed that MAPK activation was upregulated in FRL-1-injected animal caps compared with uninjected controls (Fig. 10B). In addition, MAPK activation was detected in the animal caps treated with bFGF, a known activator of MAPK signaling, to a similar extent to those injected with FRL-1 (Fig. 10B). It is known that MAPK is activated in response to wounding(such as dissection) of animal caps. To exclude the possibility that the result was a response to wounding, we tested for MAPK activation by whole-mount immunostaining. lacZ with or without FRL-1 (1 ng) was injected into one blastomere of 2 cell-stage embryos, which were cultured until late blastula stage for whole-mount immunostaining using anti-activated MAPK. Staining for activated MAPK was seen in the region with FRL-1 (Fig. 10D) but not in the regions with lacZ alone(Fig. 10E). We also examined,by western blot analysis, whether MAPK activity was inhibited in the FRL-1-depleted region. In 2 of 3 experiments, MAPK activation was strongly downregulated in FRL-1MO-injected animal caps(Fig. 10C), but was unaffected in the third experiment (data not shown). This result suggested that the phenotype seen with FRL-1MO injection may be caused by downregulation of MAPK activation. To confirm this, we attempted to rescue this phenotype by expression of the constitutively activated form of MAPKK, an intracellular activator of MAPK. Indeed, this did rescue the FRL-1MO phenotype, suggesting that FRL-1 regulates ectodermal MAPK activity to induce neural induction(Fig. 10F,G).

To further examine whether MAPK activation is required for FRL-1 regulation of neural induction, we used a MEK inhibitor, PD98059, to inhibit MAPK activation by MEK. In animal caps, the expression of N-CAM induced by the overexpression of FRL-1 was inhibited by treatment with PD98059,suggesting that FRL-1 is required for MAPK activation to induce the expression of neural markers (Fig. 11A). In contrast, a specific inhibitor of PI3K, LY294002 did not affect the N-CAM induction by FRL-1(Fig. 11A), even though the FGF receptor is known to activate PI3K signaling(Powers et al., 2000). We also showed that FRL-1 acts as a neural inducer by inhibiting BMP signaling. We therefore examined whether inhibition of BMP signaling by FRL-1 was affected by the MEK inhibitor, PD98059(Fig. 11B). RT-PCR analysis showed that the expression of BMP responsive genes, Xmsx-1 and Xvent-1, which was suppressed by FRL-1 in animal caps, were rescued by treatment with PD98059 (Fig. 11B). These data suggest that inhibition of BMP signaling by FRL-1 is required for MAPK activation.

Members of the EGF-CFC protein family play an essential role in vertebrate development. In particular, they are required for nodal signaling, which controls mesendoderm induction, anterior-posterior pattering and left-right asymmetry (Schier and Shen,2000). Biochemical studies have revealed that EGF-CFC proteins are involved not only in nodal signaling but also in another signaling pathway(Shen and Schier, 2000). In this study, we show that overexpression of FRL-1 induces early neural genes by inhibiting BMP signaling, and that loss of FRL-1 function causes neural defects during development. These results suggest that FRL-1 is essential in neural development.

The expression of early neural genes such as soxD, sox2, and zic-related genes is activated in the presumptive neuroectoderm of gastrulae, and signals the start of neuralization(Sasai, 2001). In animal cap assays, we now show that overexpression of FRL-1 induces soxD and zic3, and FRL-1MO inhibits the induction of these genes in neuroectoderm, suggesting that FRL-1 regulates an early step of neural induction. Our studies also suggest that FRL-1MO-injected ectoderm loses competence to respond to the neural inducer, chordin, which acts as an antagonist of BMP. In Xenopus, the `default model'proposes that neural induction occurs only by BMP inhibition in the presumptive ectoderm (Munoz-Sanjuan and Hemmati-Brivanlou, 2002). However, in chick, misexpression of chordin does not induce the expression of neural markers in non-neural ectoderm (Streit et al.,1998), but does induce Hensen's node, which is equivalent to the Xenopus organizer (Streit et al.,1998; Streit et al.,2000). These observations indicate that the inhibition of BMP signaling is not sufficient for neural induction and extracellular factors are required in Hensen's node in chick. FGFs are good candidates since FGFs induce neural tissue (Storey et al.,1998; Wilson et al.,2000) and the FGF receptor inhibitor SU5402 inhibits early neural induction in chick (Streit et al.,2000). In Xenopus, FGFs also induce neural tissue(Kengaku and Okamoto, 1993; Lamb and Harland, 1995; Launay et al., 1996) and dominant-negative mutants of FGF receptor type 4a efficiently inhibit anterior neural tissue (Hongo et al.,1999). Although FGF signaling is therefore required in both Xenopus and chick, FGFs induce only posterior neural tissue,suggesting that other ligands are responsible for inducing anterior neural tissue. We propose that EGF-CFC proteins are strong candidate neural inducers that can activate FGF signaling. We showed that FRL-1 induces the expression of anterior neural markers and that FRL-1MO inhibits the induction of anterior neural tissue. In addition, FRL-1 activates the FGF receptor/MAPK signaling pathway. We note that the expression of chick CFC is condensed in Hensen's node region (Colas and Schoenwolf, 2000), in contrast to the ubiquitous expression of Xenopus FRL-1 (Kinoshita et al.,1995). The expression patterns of chick CFC may correspond to a neural inducer in Hensen's node and may reflect the different competence in responding to neural inducers between Xenopus and chick.

Although it has been shown that FRL-1 activates FGFR-dependent Ca²⁺ release in oocytes and that XFD inhibits FRL-1 function in animal caps, we show here that FRL-1 activates MAPK in blastula embryos and its activation is required for BMP inhibition and neural induction by FRL-1. Furthermore, we showed both the EGF and CFC domains are necessary for neural induction of FRL-1 (Fig. 2A). However, the EGF-like domain alone of the EGF-CFC protein Cripto is sufficient to activate the MAPK pathway in mouse mammary epithelial cells(Kannan et al., 1997). This difference may be explained by these proteins binding different receptors:FRL-1 function may depend on FGF receptor signaling(Kinoshita et al., 1995)(Fig. 10A), whereas in mouse mammary epithelial cells, human Cripto indirectly activates ErbB4(Kannan et al., 1997; Bianco et al., 1999). These results suggest that both the EGF and CFC domains may be required for FGF receptor-mediated functions of EGF-CFC proteins.

While it is known that EGF-CFC genes are involved in nodal signaling, in this study, overexpression of FRL-1 inhibited BMP signaling in animal caps, where nodal-related genes are not expressed, suggesting that BMP signaling was blocked in the absence of nodal signaling. We also found that oep, mouse cripto and chick CFC induced the expression of neural markers in Xenopus animal caps. Our observations suggest that BMP signaling may not only be inhibited by FRL-1 but also by EGF-CFC genes in a nodal-independent fashion. Although EGF-CFC proteins are required for nodal signaling, the mouse EGF-CFC protein Cripto activates MAPK signaling, independent of nodal and its receptor, ALK-4(Bianco et al., 2002). These results suggest that EGF-CFC protein is able to act via MAPK activation, which is independent of nodal. Previous studies showed that activated Rascould not rescue oep mutant in zebrafish(Gritsman et al., 1999),suggesting that Oep function is not required for MAPK activation. oepmutants have defects in mesendodermal tissue and are therefore thought to lack a cofactor for nodal. Our results showed that FRL-1 acts via a MAPK pathway in neural induction (Fig. 11A). These results suggest that a functional EGF-CFC protein is required for MAPK activation in neural induction but not in nodal signaling.

It is known that some extracellular ligands that activate MAPK signaling have opposite effects to BMP. For example, FGF opposes the anti-proliferative effect of BMP-2 during limb bud outgrowth(Niswander and Martin, 1993)and epidermal growth factor (EGF) antagonizes the induction of osteogenic differentiation markers by BMP-2 (Bernier and Goltzman, 1992). Previous reports have suggested that MAPK-mediated phosphorylation negatively regulates the function of Smad1, an intracellular mediator of BMP signaling(Kretzschmar et al., 1997). MAPK phosphorylates the linker domain of Smad1, thereby inhibiting its nuclear translocation and subsequent BMP signaling. This result seems to be applicable for FRL-1 function in inhibiting BMP signaling and sustaining neural competency. Furthermore, Sater and colleagues showed that MAPK activation is required for the induction of Xenopus neuroectoderm(Uzgare et al., 1998; Goswami et al., 2001). Their results suggest that upregulation of MAPK activity is detected in the neuroectoderm of dominamt negative BMPR- or noggin-injected ectoderm and that the overexpression of MAPK phosphatase, which inactivates activated MAPK, inhibits neural induction in Xenopus(Goswami et al., 2001). Our results could account for upregulation of MAPK activity by the observed enrichment of FRL-1 transcripts in neuroectoderm and the induction by chordin. Conversely, it was proposed that BMP signaling might inhibit MAPK activity via the TAK1 pathway(Goswami et al., 2001),indicating mutual antagonisms. Taken together, our results suggest that antagonistic effects between the FRL-1/FGFR/MAPK pathway and BMP signaling are involved in the establishment of neural versus epidermal cell fate.

We would like to thank E. M. De Robertis, M. Whitman, N. Ueno, E. Nishida,R. Burdine, R. M. Harland and D. J. Anderson for their kind gifts of plasmids and H. Y. Kubota for Neu-1 antibody. This work was supported in part by grants from the ministries of Education, Culture, Sport, Science and Technology and by SORST projects of the Japan Science and Technology Corporation. K.T. and S.T. are also supported by the Japan Society for the Promotion of Science.