The Identification of Two Novel Ligands of the FGF Receptor by a Yeast Screening Method and Their Activity in Xenopus Development

Cell, Vol. 63, 621-630, November 17, 1995, Copyright 0 1995 by Cell Press The Identification of Two Novel Ligands of the FGF Receptor by a Yeast Screening Method and Their Activity in Xenopus Development

Noriyuki Kinoshita, l Jeremy Minshull,f To identify potential ligands for orphan tyrosine kinase and Marc W. Kirschner’ receptors, we have developed a novel yeast screening *Department of Cell Biology method based on the functional coexpression of a receptor Harvard Medical School and itscorresponding ligand in thesameyeast cell. Recep- Boston, Massachusetts 02115 tor-ligand interactions in principle could take place either tDepartment of Physiology on a cell surface or in intracellular compartments. In either University of California case, the cytoplasmically oriented kinase of the receptor San Francisco, California 94143-0444 would be activated and phosphorylate cytoplasmic proteins; this activity could be detected by an increase in total intracellular tyrosine phosphate. By using this approach, Summary it should be possible to screen for the ligand of an orphan receptor by coexpressing a cDNA library and the orphan We have developed a functional screen in yeast to receptor in yeast and detecting those colonies with ele- identify ligands for receptor tyrosine kinases. Using vated tyrosine kinaseactivity. Thefactthat higher eukaryo- this method, we cloned two Xenopus genes that acti- tic genes can be functionally expressed in yeast, and that vate the fibroblast growth factor (FGF) receptor. These yeast has negligible tyrosine kinase activity, supports the encode novel secreted proteins, designated FRLl and feasibility of the approach. FRLP, distantly related to the epidermal growth factor We show that yeast cells expressing the Xenopus laevis and angiogeniruribonuclease families, respectively. fibroblast growth factor (FGF) receptor gene exhibit a large Both genes activate the FGF receptor in Xenopus oo- increase in tyrosine phosphorylation upon cotransforma- cytes as well as in yeast. Overexpression induces tion with the FGF gene. This increase in tyrosine phos- mesoderm and neural-specific genes in Xenopus ex- phate occurs only if the ligand has a functional signal se- plants; induction is blocked by a dominant negative quence. We have adapted this procedure to screen a inhibitor of the FGF receptor. FRLl is broadly ex- Xenopus cDNA library by using a yeast strain expressing pressed during gastrulation and neurulation, while an FGF receptor. Two genes encoding novel secreted pro- FRL2 is expressed principally in the axial mesoderm teins were identified with the yeast assay and subse- and brain at later stages. Our results indicate that de- quently characterized by other assays. These ligands, spite their lack of similarity with FGF, FRLl and FRL2 designated FRLl and FRL2 (for FGF receptor ligands 1 are ligands for the FGF receptor that play distinct roles and 2) are very distantly related to two other proteins: an in development. epidermal growth factor (EGF)-related protein, cripto (also referred to as teratoma-derived growth factor 1, or Tdgfl), and another secreted protein, angiogenin (Ciccodicola et Introduction al., 1989; Kurachi et al., 1985; Beintema et al., 1988). Our results indicate that despite their lack of similarity to the Peptide growth factors, mediated by specific cell surface primary sequences of the FGF family, FRLl and FRL2 are receptors, regulate many cell processes, such as growth, novel ligands for the FGF receptor. Furthermore, these division, and differentiation. Many of these activate mem- ligands seem to play distinct roles in Xenopus development. bers of a superfamily of receptor tyrosine kinases that contains an extracellular ligand-binding domain, a transmembrane domain, and a cytoplasmic domain harboring a regulated protein tyrosine kinase activity. Upon ligand Results binding to the receptor, the cytoplasmic kinase is activated, leading to the activation of the receptor-specific Detection of FGF Receptor Activation by FGF in yeast intracellular signaling pathways. We tested whether coexpression of a receptor tyrosine In the past 10 years, many tyrosine kinase receptors kinase and its ligand leads to the activation of the kinase in have been identified. Because the tyrosine kinase do- Yea54 (Saccharomyces cerevisiae) cells. Three plasmids mains of this receptor family are highly conserved, tech- were constructed. The first, pG-FGFR, harbors the Xeno- niques dependent on sequence similarity can be used to Pus FGF receptor 1 gene (Musci et al., 1990). Thesecond, identify family members. By contrast, methods for identi- pFGF, contains the Xenopus FGF2 (basic FGF) gene (Ki- fying the ligands for tyrosine kinase receptors have lagged melman et al., 1988), which does not have an endogen- behind, and for many cloned receptors, the peptide ligand ous Signal sequence. The third, pssFGF, encodes FGF2 has not yet been identified. For the most part, identification fused downstream to a yeast SUC2 signal sequence (Carl- of ligands involves biochemical fractionation, which relies son et al., 1983). All these genes are under the control of on having sufficient quantities and sensitive assays; how- the GAL7 promoter. ever, other methods using receptor affinity have recently TO determine whether the receptor tyrosine kinase is been introduced (Flanagan and Leder, 1990; Lyman et activated in these strains, these transformants were cul- al., 1993; Cheng and Flanagan, 1994). tured in the presence or absence of galactose, and whole- Cell 622

served after expression of the FGF receptor alone. The level of FGF2 was unaffected by the presence of the signal sequence (data not shown). Since ssFGFs were not detected in the medium (data not shown), FGF is probably localized in intracellular compartments, in the periplasmic space, or both. These results indicate that the FGF receptor is activated by the expression of its corresponding ligand in yeast and that the ligand must be capable of enter- ing the secretory pathway for receptor activation to occur.

Measuring Receptor Activation by Colony lmmunoblotting We screened large numbers of yeast transformants by using colony immunoblotting (Lyons and Nelson, 1984) 12345670 -1 with anti-phosphotyrosine antibody. To test the reliability glucose galacto~e of receptor activation by this method, we used yeast strains expressing the FGF2 and the FGF receptor genes under the control of the GAL7 promoter. FGF or ssFGF was coexpressed with the receptor, and the yeast cells were streaked on a glucose-containing plate. Colonies were transferred to two nitrocellulose filters; one filter was placed on a galactose-containing plate to induce FGF expression, and the other was placed on a glucose-containing plate as a negative control, and the level of tyrosine phosphorylation in each colony was determined with anti- phosphotyrosine antibody. The results confirmed the immunoblotting data in Figure 1A. The expression of the FGF receptor led to an increase in the level of tyrosine glucose galactose phosphorylation that was substantially augmented when ssFGF was coexpressed, but not when FGF lacking a sig- Figure 1. Activation of the FGF Receptor in Yeast by Coexpression of the FGF2 Gene nal sequence was coexpressed (Figure 16). (A) Phosphotyrosinecontent of endogenousyeast proteins in response There was an unsuitably high background from the FGF to expression of FGF and its receptor. Yeast strains transformed with receptor under control of the GAL7 promoter in the ab- the vectors (lanes 1 and 5), pG-FGFR (lanes 2 and 6) pG-FGFR plus sence of FGF expression. We therefore tried placing the pFGF (lanes 3 and 7), and pG-FGFR plus pssFGF (lanes 4 and 6) FGF receptor under the control of the ACT7 promoter (ac- were cultured in glucose-containing media (lanes l-4) and in galactose-containing media (lanes 5-8). Whole-cell extracts were prepared tin gene; Gallwitz et al., 1981) in an attempt to optimize by cell disruption with glass beads and were analyzed by immunoblot- the assay. We found the ACT7 promoter more suitable. It ting with anti-phosphotyrosine antibody. gave similar levels of FGF receptor gene expression in (6) The detection of the FGF receptor activation in yeast cells by colony galactose- and in glucose-containing media, and levels of immunoblotting. FGF and the FGF receptorwere coexpressed in yeast tyrosine phosphorylation were low in the absence of FGF, and analyzed by colony immunoblotting with anti-phosphotyrosine antibody. The FGF receptor was under the control of the GAL7 promoter but significantly increased by expression of FGF (Fig- (G-FGFR) or the ACT7 promoter (A-FGFR). FGF with or without signal ure 16). sequence (ssFGF or FGF) was under the control of the GAL I promoter. Yeast strains transformed by the indicated genes were grown on a plate. Colonies were transferred to two membranes and cultured on Screening cDNA Libraries with a Yeast Strain a glucose-containing or a galactose-containing plate. Expressing the Xenopus FGF Receptor Gene The above results suggested that the method could be used to identify novel ligandsfortyrosine kinase receptors. cell extracts were analyzed by immunoblotting with an- As a first step, we used the method to identify ligands for ti-phosphotyrosine antibody (Figure 1A). The following the FGF receptor. Yeast cells expressing the FGF receptor results were obtained when expression was induced by were transformed with two cDNA libraries: one was made galactose: first, expression of the FGF receptor gene alone from mRNA isolated from Xenopus unfertilized eggs and led to a substantial increase in tyrosine phosphorylation gastrula embryos (egg library), and the other was made of several endogenous proteins; second, the coexpression from Xenopus tissue culture (XTC) cells (XTC library). of the FGF receptor and FGF2 lacking a signal sequence The egg and XTC libraries yielded 150,000 and 25,000 (pFGF) did not increase phosphorylation above that of the transformants, respectively. In the first screening by FGF receptor alone; and third, coexpression of the FGF colony immunoblotting with an anti-phosphotyrosine anti- receptor and FGF2 with a signal sequence (pssFGF) dra- body, 65 and 29 candidates were identified, respectively, matically increased tyrosine phosphorylation. The phos- and by the second screening, nine and two transformants phorylation level was five times higher than the level ob- were found to be positive (egg and XTC libraries, respec- Novel FGF Receptor Ligands in Xenopus Development 623

Table 1. Genes That increase Protein-Tyrosine Phosphorylation in Yeast Ceils Expressing FGF Receptor I FGF Receptor Frequency Gene Dependency Gene Product ’ of Isolation

FRL 1 + Cripto (EGF-like growth factor) 4 FRLP + Homologous to angiogenin and RNase A 1 XT2 + 58% identical to human cathepsin L 1 EGl + Heterogeneous ribonucleoprotein 2 EG3 + RNA-binding protein 1 EG4 + Novel 96 kDa protein 1 EG5 Cytoplasmic tyrosine kinase FER 1

tively). To test the dependence on expression of the FGF Of particular interest were the two genes FRL7 and receptor, plasmid DNA in each transformant was rescued FRLP, which encoded polypeptides with putative signal and retransformed into control yeast strains or strains ex- sequences. Figure 2 shows that phosphotyrosine levels pressing the FGF receptor gene. Only one plasmid was increase dramatically in yeast strains expressing the FGF found to be positive in the absence of the receptor. This receptor together with either FRLl or FRL2. FRLl is dis- plasmid encoded a putative tyrosine kinase homologous tantly related to cripto, which is an EGF family member to the mouse cytoplasmic tyrosine kinase FER (for fpsl identified in the mouse and human (Figure 3A; Ciccodicola fes-related; Hao et al., 1989; data not shown). The other et al., 1989; Dono et al., 1993). Though the predicted genes increased tyrosine phosphorylation only when the amino acid sequence identity is very low (24%) compared FGF receptor gene was coexpressed. with human and mouse cripto, the EGF motif is completely conserved. FRLP encodes a protein weakly homologous Sequences of the Cloned Genes to the angiogeninlribonuclease superfamily (Figure 36; In the ten plasmids that activated the FGF receptor, there Beintema et al., 1988). The identity again was very low: were six unique sequences (Table 1). These genes encode 21% to both bovine angiogenin (Maes et al., 1988) and two putative secreted proteins (FRLI and FRL2), one pro- bovine pancreatic ribonuclease A (Smyth et al., 1963; Car- tease (XT2), two RNA-binding proteins (EGl and EG3), sana et al., 1988), and 26% to the bullfrog ribonucleasel and one novel protein (EG4) lacking a signal sequence. sialic acid-binding lectin (Titani et al., 1987). Phylogenetic FRLl and XT2 were isolated from the XTC library, and analysis of these genes shows that FRLP does not seem the others are from the egg library. Genes related to to belong to either the angiogenin or the ribonuclease the traditional FGF gene family were not isolated in this branches (Figure 3D). screen. Both FRLl and FRLP have hydrophobic regions at the C-terminus; the N-terminal hydrophobic domain probably is the signal sequence (Figure 3C). These proteins could be anchored to the cell membrane by their C-termini. To test this possibility, we expressed these genes and the C-terminal deletion mutants of these genes in Xenopus oocytes and found that the deletion mutants were secreted into the media, whereas the full-length gene products were not (data not shown). This suggests that FRLl and FRL2 retain an association with membranes or vesicles and that the C-terminal regions are required for this association. The XT2 sequence was found to be similar to human cathepsin L, sharing 58% amino acid identity in the region we sequenced (90 out of 155 amino acids; data not shown; Joseph et al., 1988; Gal and Gottesman, 1988). EGl was FRL-2 FRL-P previously identified in Xenopus as a heterogeneous ribo- glucose galactose nucleoprotein (Kay et al., 1990). EG3 has an RNA recogni- tion motif found in many RNA-binding proteins (data not Figure 2. The Activation of the FGF Receptor by FRL7 and FRLP in shown; Kim and Baker, 1993). EG4 encodes a novel 96 Yeast kDa protein. Recently, a gene similar to EG4 was found A yeast strain expressing the FGF receptor was cotransformed with by the complete sequencing of the chromosome III in ssFGF, FRLl, and FRLPgenesunderthecontroloftheGALl promoter. Caenorhabditis elegans; it was designated Rl OE12.1 The transformants were streaked, and the colonies were transferred (39% identity), but its function is unknown (Wilson et al., to two membranes and cultured on a glucose-containing or a galactose-containing plate. The membraneswere analyzed by colony immu- 1994). No putative signal sequence was found in this gene noblotting method with anti-phosphotyrosine antibody. (data not shown). Cell 624

Figure 4. The Activation of the FGF Receptor in Xenopus Oocytes by FRLI and FRL2 Indicated RNAs were transcribed in vitro and injected into oocytes. After 48 hr of incubation, oocytes were labeled with Wa2+ for 3 hr, and released Ca2+ from oocytes was measured over 10 min.

add soluble ligand to the culture media and then assay release of %a’+. We have not succeeded in obtaining active purified preparations of FRLl and FRL2 proteins in soluble form. Instead, we performed the calcium release assays under conditions of continuous ligand exposure by coexpressing the ligand and receptor in the same cell. The mRNAs for FRLl and FRLP were coinjected with FGF receptor mRNA into oocytes, and the basal level of Ca2+ efflux was measured. We have found this assay to be accurate and highly reproducible. As shown in Figure 4, we found that coinjection of FGF and FGF receptor mRNAs increased Ca*+ release to a level 3-fold higher than that for the injection of FGF receptor message alone. Coinjection of FM7 and FRL2 mRNAs with the FGF receptor mRNA increased Ca*+ release 2-fold and 2.5fold, respectively. FM7 or FRL2 mRNA alone did not increase Figure 3. Ammo Acid Sequences of FRLl and FRLZ Ca*+ release. To test whether the stimulation of the Ca*+ (A) Sequence alignment of FRLI with mouse cripto. Stars indicate release by FRLl and FRL2 was FGF receptor-specific, identical amino acid residues. The putative signal sequence and the C-terminal hydrophobic region of FRLl are underlined. Xenopus platelet-derived growth factor (PDGF) receptor (B) Sequence alignment of FRL2 with bovine angiogenin and pancre- was coexpressed with FRLl and FRLP in oocytes. There atic ribonuclease A. Asterisks indicate amino acid residues identical was no increase in basal Ca*+ release by FRLl or FRL2 among the three. The putative signal sequence and the C-terminal in the oocytes expressing the PDGF receptor (data not hydrophobic region of FRLZ are underlined. (C) Hydrophobicity plots of FRLI and FRL2 (Kyte and Doolittle, 1982). shown). These results suggest that FRLI and FRLP can (D) Phylogenetic analysis of FRL2 and some of the angiogeninlribo- work as activators of the FGF receptor in oocytes as well nuclease family members: human angiogenin (Kurachi et al., 1985), as in yeast cells. mouse angiogenin (Bond and Vallee, 1990) rabbit angiogenin (Bond et al., 1993) bovine angiogenin (Maes et al., 1988) mouse ribo- FRL2 Binds to the Extracellular Domain of the FGF nuclease A (Schuller et al., 1990) bovine ribonuclease A (Carsana et al., 1988), and bullfrog ribonuclease (Titani et al., 1987). Receptor In Vitro To test whether FRLl and FRL2 bind directly to FGF receptor, we constructed a plasmid (XFR-AP) in which the extracellular domain of Xenopus FGF receptor 1 was FRLl and FRL2 Activate the FGF Receptor fused to alkaline phosphatase (AP). This plasmid was in Xenopus Oocytes transfected into COS cells, and the secreted fusion protein Since FRLl and FRL2 were potential ligands for the FGF was precipitated with anti-AP antibodies and protein A- receptor, we tested whether they could also activate the Sepharose beads. The beads were mixed with in vitro FGF receptor expressed in higher eukaryotic cells. Since it translated FRLl or FRL2 proteins, which had been sol- is known that the activation of the FGF receptor in Xenopus ubilized with Triton X-100, cross-linked with 3,3’-dithiobis oocytes is linked to a rapid Ca2+ release from internal (sulfosuccinimidylpropionate) (DTSSP), and washed. The stores (Johnson et al., 1990; Musci et al., 1990), we per- cross-link was cleaved by 2-mercaptoethanol treatment formed a Ca’+ release assay with Xenopus oocytes. The and boiling. As shown in Figure 5, FRL2 was coprecip- conventional way to measure agonist stimulation of recep- itated with the receptor protein. The level of FRL2 that tors that have been expressed in Xenopus oocytes is to was immunoprecipitated increased 4-fold as compared Novel FGF Receptor Ligands in Xenopus Development 625

vector XFR-AP + FGF

< FRL-2

Figure 5. FRLP Binding to the FGF Receptor The fusion protein XFR-AP, composed of the extracellular domain of the FGF receptor and alkaline phosphatase, was produced in COS ceils. %-labeled FRL2 protein synthesized in the reticulocyte lysate was mixed with the XFR-AP in the presence or absence of unlabeled FGF. The proteins precipitated with anti-alkaline phosphatase antibody were analyzed by autoradiography. The vector-transfected COS cells was used as a control (vector). The receptor brings down the 18 kDa ligand at a level of 3.9x over background. Excess cold FGF reduces this to 1.9x over background.

EF-la L with the vector alone. Binding was completely competed by unlabeled FGF. These results indicate that FRL2 binds C to the extracellular domain of the FGF receptor, and that FRL-1 FRL-2 FGF and FRL2 share a common binding site on the FGF receptor. We were unable to coprecipitate FRLl with the same binding assay (data not shown).

FRLl and FRL2 Induce Mesoderm through the FGF Receptor Signaling Pathway in Animal Cap Explants from Xenopus Embryos uninjected It is known that exposure to FGF and other related growth factors induces mesoderm in animal caps isolated from Figure 6. Effects of the Overexpression of FRLl and FRLZ on Animal Caps and Embryos Xenopus embryos (Kimelman and Kirschner, 1987; Slack (A) Morphological changes in animal caps. FRL7 RNA (5 ng) and FRLZ et al., 1987; lsaacs et al., 1992). We tested whether FRLl RNA (1 ng) were injected into both blastomeres at the 2-cell stage. and FRL2 might also be capable of inducing mesoderm Animal caps were isolated at stage 8, cultured for 20 hr at 16OC, and through activation of the FGF receptor. We injected mRNA photographed. encoding FRLl and FRL2 into both blastomeres of 2-cell (6) The effect of FRLI and FRL2 on muscle actin and N-CAM expression. Uninjected animal caps and animal caps injected with 1 ng and embryos, isolated the animal caps at stage 8, and incu- 5 ng of FRLI, 5 ng of FRLl plus 1 ng of HAV@, 5 ng of FRLl plus 1 bated them until sibling embryos had reached stage 20. ng of XFD, 1 ng of FRLZ RNA, 1 ng of FRLP plus 1 ng of HAVQ, or Uninjected caps rounded up, giving the uninduced pheno- 1 ng of FRLP plus 1 ng of XFD RNA were cultured at 16OC for 20 hr. type of epidermis. Animal caps injected with FRLI or FRLP RNA was isolated, and induction of muscle actin (m. actin) and N-CAM genes was assayed by RT-PCR. EFl o was used as an internal control. RNAs exhibited striking morphological changes, with (C)Common phenotypes of the tailbud stage embryos overexpressing FRLl-injected animal caps elongating more than FRL2- FRLl and FRLP. FRLi or FRL2 RNA (1 ng) was injected into both injected animal caps (Figure 6A). Both showed more ex- blastomeres at the 2-cell stage. The left side is anterior. treme morphological responses than animal caps treated with FGF. To demonstrate that these animal caps had responded to the ligands by changing the state of differentiation, we enous FGF receptor. As a control in these experiments, assayed them by reverse transcription-polymerase chain we used HAV@, which is a mutated form of XFD with a reaction (RT-PCR) for expression of muscle actin, which is deletion of three amino acids in its extracellular domain; a mesodermal marker, and neural cell adhesion molecule HAV@ does not inhibit the activation of the FGF receptor (N-CAM), a neural marker. As shown in Figure 6B, FRL2 by FGF (Byers et al., 1992; Amaya et al., 1993). mRNA induced muscle actin, but did not induce N-CAM. By con- encoding XFD or HAV@ mutants was coexpressed with trast, FRLl induced both muscle actin and N-CAM at FRLl or FRL2 in animal caps, and the induction of muscle higher doses of injected RNA (5 ng), but at lower doses actin and N-CAM genes was assayed by RT-PCR. Sur- (1 ng), FRLl induced only N-CAM. At 5 ng, FRL2 RNA prisingly, as shown in Figure 6B, the induction of muscle was toxic. actin by FRLl was inhibited by both XFD and HAV@. Simi- We tested the dependence of FRLl and FRL2 activities larly to FGF, the induction by FRL2 was inhibited by XFD on the FGF receptor by coexpressing a dominant negative but not by HAV@. Inhibition of both FRLl and FRL2 by FGF receptor (XFD; Amaya et al., 1991). XFD is a domi- XFD indicates that the activation of the FGF receptor is nant negative mutant in which almost all the cytoplasmic required for the function of FRLl and FRL2. This strongly domain of the receptor is deleted; it is thought to inhibit suggests that FRLl and FRL2 bind to the extracellular the FGF receptor by forming a heterodimerwith the endog- domain of the dominant negative receptor and form a het- Cdl 626

D Emor V” “4 As L

FRL-1

EFla

Figure 8. The Temporal and Spatial Expression of FRL7 and FRLP Genes during Embryogenesis (A) Temporal expression. RNA was isolated from unfertilized eggs and the indicated stages of embryos. The expression of the FRLl and FRLP Figure 7. The Paracrine Signaling Assay genes was assayed by RT-PCR. Ornithine decarboxylase (ODC) is present as an internal control. (A) The procedure. mRNA encoding FRLI or FRL2 (10 ng) was injected (6) Spatial expression. Stage 10 embryos were dissected into the dor- into Xenopus oocytes. After 1 day of incubation, animal cap explants sal half (Dr) and the ventral half (Vn), or into the vegetal half including isolated from Xenopus embryos at stage 8 were grafted onto the in- the marginal zone (Vg) and the animal half (An), and RNA was isolated jected or uninjected oocytes. from these tissues or from stage 10 embryos (Em). The expression (B) The effect of FRLl and FRL2 on muscle actin and N-CAM expres- of the FRLl gene was assayed by RT-PCR. goosecoid (gsc) is known sion. RNA was isolated from the grafted explants when the sibling to be expressed in the dorsal marginal zone (Cho et al., 1991). EFlu embryos reached stage 25. Induction of muscle actin (m. actin) and is a ubquitously expressed control (Krieg et al., 1989). N-CAM genes was assayed by RT-PCR. EFI a was used as an internal (C) Whole-mount in situ hybridization in the tailbud stage embryo of control. FRLZ RNA. Purple staining indicates the localization of the antisense FRLZ probe. erodimer with the full-length receptor to inhibit the receptor activation. The differences between the response of XFD jetted into 2-cell embryos, and the phenotype was ob- and HAV@ to FFiLl and both FRLP and FGF suggest possi- served. As shown in Figure 6C, a common phenotype in ble differences in the interactions of these ligands with both FRL7-injected and FRL2-injected embryos was re- the extracellular domain of the FGF receptor. duced anterior head morphology. Embryos without eyes, cement gland, or both were often found (11 out of 30 em- FRLI and FRL2 Can Act in a bryos injected with 1 ng of FRL7 and 11 out of 32 embryos Non-Cell-Autonomous Manner injected with FRLP RNA at 2-cell stage). In addition, FRL7 To test whether FRLl and FRL2 act as intercellular signal- induced tail-like appendages (11 out of 30 embryos). At ing molecules, we have employed a paracrine signaling tadpole stage, a fin and melanocytes were formed, but assay developed by Lustig and Kirschner (1995; Figure neither notochord nor somites could be identified in these 7A). In this assay, mRNA encoding FRLl or FRL2 was appendages by immunohistochemistry (data not shown). injected into oocytes and cultured for 1 day, after which The FRLP-injected embryos showed severe gastrulation animal caps from stage 8 embryos were grafted onto the defects as well as head defects. The head defect pheno- oocytes. The animal caps were harvested when the sibling type suggests that ectopic expression of FRLl and FRL2 embryos reached stage 25 and were analyzed by RT- might cause posteriorization of anterior tissue, or loss of PCR. If the proteins act intercellularly, the proteins pro- anteriorization. duced in oocytes are expected to activate the receptors of the grafted tissues, resulting in the induction of the muscle The Temporal and Spatial Expression Pattern actin, N-CAM genes, or both. As shown in Figure 78, both of FRLI and FRLP in Xenopus Embryo FRLl and FRL2 expressed in oocytes induced muscle The temporal expression patterns of FRL7 and FRL2 were actin and N-CAM in the animal caps. This induction was mapped with RT-PCR, using ornitihine decarboxylase as comparable to that for whole embryos when normalized for an internal standard. As shown in Figure 8A, both FRL7 elongation factor 1 a (EFl a). This result strongly suggests and FRLP mRNAs were of low abundance maternally, but that FRLl and FRL2 can act in a non-cell-autonomous were clearly detected by stage 10. The FRLP transcript manner. gradually increased during embryogenesis through stage 38, whereas the FRL7 transcript was detectable only dur- Tail-like Appendages from the Ectopic Expression ing gastrula stages and disappeared rapidly at early neu- of FRL1 rula stage. To investigate the activity of FRLl and FRLP in em- The spatial distribution of FRL7 expression was mapped bryogenesis, mRNAs encoding these proteins were in- with whole-mount in situ hybridization on gastrula stage Novel FGF Receptor Ligands in Xenopus Development 627

embryos. Expression was detected throuighout the em- of endogenous FGF molecules in oocytes or anirnal caps, bryo at gastrula stage (data not shown). When we dis- or they might stimulate both activities. Indeed, it is known sected embryos into dorsal and ventral or animal and veg- thatXFD inhibitsthe mesoderm induction byactivin, which etal pieces at gastrula stage, we found broad and uniform does not interact with the FGF receptor (LaBonne and expression of FRL7 by RT-PCR, confirmiing the lack of Whitman, 1994; Cornell and Kimelman, 1994). However, localization by in situ hybridization (Figure 8B). As a con- one unusual experimental result suggests that this is not trol, we found goosecoid expression limited to the dorsal the case for FRLl. Its ability to induce muscle actin was lip, and EFla uniformly expressed. In situ hybridization inhibited, and its ability to induce N-CAM was partially of FRLP at tailbud stage shows expression in somites, otic inhibited not only by XFD, but also by a mutant form of vesicles, brain, eyes, cement gland, and visceral arches, the dominant negative inhibitor, HAV@, that does not in- suggesting that FRl.2 might be involved in the develop- hibit FGFfunction (Byers et al., 1992, Amaya et al., 1993). ment of these tissues (Figure 8C). This indicates that FRLl does not function through endogenous FGF and may suggest that FRLl interacts with dif- Discussion ferent binding determinants in the FGF receptor than does FRL2 or FGF. Taken together, these data suggest a direct There are several transmembrane tyrosine kinase recep- interaction for both FRLl and FRL2 with the FGF receptor tors known to be expressed at gastrula and neurula stage but provide strict proof only for FRLP, where direct and in Xenopus embryos, which suggests that several undis- FGF-competable binding with the extracellular domain covered ligands might be present that would play a role can be demonstrated. in early patterning and tissue differentiation To identify Finding a new member of the FGF family of ligands these molecules, we have developed a functional assay would be interesting but not surprising. Finding two novel in yeast for ligands for tyrosine kinase receptors. Initially putative ligands with no homology to the FGF familly mem- we demonstrated that we can reproduce the activation of bers is more unusual, since in general, receptor tyrosine the kinase by coexpression of the receptor and ligand by kinases have been thought to interact with specific ligand use of the FGF receptor and FGF2. We tested this assay family members. However, FGF receptor family members on two Xenopus cDNA libraries and identified two novel differ in their extracellular domains, which could be the secreted proteins, FRLl and FRLP, that activate the FGF basis of differences in receptor binding. Indeed, we have receptor in the yeast assay. Surprisingly, lthese ligands found that the FGF receptor 4 from Xenopus can be acti- are not members of the homologous class of FGF ligands vated in the yeast assay with FGFP but not with FFlLl and and therefore constitute putative novel extracellular acti- FRL2 (N.K., unpublished data), while the FGF receptor 1 vators of the FGF receptor. can be activated by all three ligands. Interactions of a Several lines of evidence suggest that FRLl and FRLP receptor with more than one ligand family are not unprece- can activate the FGF receptor extracellularly, although dented. For example, the insulin-like growth factor (IGF)-II proof that the FGF receptor is the appropriat’e receptor for receptor interacts not only with IGF-II but also with man- these or any of the FGF family of ligands is slill inferential. nose8-phosphate, and IGF-II also interacts with the IGF-I In addition to the yeast assay, both of these lllgands stimu- receptor (Morgan et al., 1987; Rechler et al., 1980). late calcium release in Xenopus oocytes in the presence of a functional FGF receptor. The paracrine assay shown The Nature of FRLI and FRLP and Their Role in Figure 7 shows that both FRLI and FRL2 can be ex- in Development pressed in one cell (the oocyte) and act on another (animal FRLl contains an EGF motif, with only 24% identity to caps); hence, FRLl and FRL2 can act in a non-cell- mouse cripto. In the mouse embryo, cripto is expressed autonomous manner. in the primitive streak and mesodermal cells, and llater in In the cake of FRL2, we were able to demonstrate that the heart (Dono et al., 1993), suggesting that cripto might the ligand binds to the extracellular domain of the FGF be involved in mesoderm formation. In Xenopui, FRLl is receptor and that this binding can be competed by FGFP. expressed throughout gastrulation and early neurulation, Our inability to demonstrate this interaction with FRLl a period of about 4 or 5 hr, and is not localized. FRLl could be due to weaker affinity or lack of activity of the could be regulated posttransriptionally or posttransla- FRLl in the presence of the detergents needed to solubi- tionally. lize it. FRLl shows different inductive effects compared with Further evidence that FRLl and FRL2 acts through the those of FGFor FRLP. A high dose of FRL7 mRNA induced FGF receptor has come from induction assays in frog em- both muscle actin and N-CAM, while a lower dose induced bryos. Both of these molecules gave potent and interesting only N-CAM, but not muscle actin. Hence, FRLl can in- biological effects (discussed below), and these effects, duce neural tissue without inducing appreciable meso- which include expression of muscle actin and morphologi- derm or at least muscle in this assay. Though other meso- cal changes, were completely blocked by c’oexpression derm inducers, such as activin, can also induce rneural of the dominant negative mutant of the FGF receptor, XFD. tissues, this is thought to be an indirect induction meldiated Formally, these experiments show that the biological ef- by the dorsal mesoderm induced by activin. fects of FRLl and FRL2 require FGF signaling, but do not FRL2 shows no homology to typical FGF family mem- prove that these ligands act directly on the FGF pathway. bers but is distantly related to the ribonuclease/angialgenin FRLl and FRLP might stimulate production Norsecretion family (identities of 21%-26%). These molecules share Cdl 628

conserved cysteine positions and a conserved histidine tify these ligands, screening by structural homology might bounded by valine and phenylalanine in the C-terminus. need to be augmented by independent screens based on Angiogenin has strong angiogenic activity, which is also functional assays. true of FGF (Folkman and Klagsburn, 1987; Risau, 1990). Experimental Procedures There is no known receptor for angiogenin, and it is tempt- ing to speculate that angiogenin, like FRLP, might activate Yeast Strain and Plasmids a member of the FGF family of receptors, rationalizing The yeast (S. cerevisiae) strain used in this study was PSY315 (Mata, why it is effective as an angiogenic agent. FM2 mRNA /euZ-3.122, ura3-52, his3-d200, /ysZ-802). was detected at gastrula stage and gradually increased The vector plasmids pTS210 and pTS249 carry URA3 and LEUP, respectively, and both carrythe CfN4, GAL1 promoter and ACT7 termi- through tadpole stage. It is expressed broadly in somites, nator (gifts from Dr. T. Stearns, Stanford University). pKNA1 is the nervous system, eyes, cement gland, and visceral arches, same as pTS249, except that the GAL promoter is replaced with ACT7 where the FGF receptors 1 and 4 are expressed (Brandli promoter. The promoter of ACT1 (Gallwitz et al., 1981) was cloned by and Kirschner, 1995). PCR using yeast genomic DNA as a template. The length of the frag ment is 678 bp. Two plasmids for expression of Xenopus FGF2 in yeast were con- The Yeast Assay for Tyrosine Kinase Function structed: one plasmid was constructed by cloning FGF2 into pTS210 In the initial studies of the FGF receptor and basic FGF, (pFGF), and the other plasmid is identical to the first one, except that we found that the receptor was activated only if FGF pos- a signal sequence of S. cerevisiae SUC2 (Carlson et al., 1983) was sessed a signal sequence (ssFGF). We did not determine inserted at the initiation codon of the FGF2 gene (pssFGF). For FGF receptor expression, the Xenopus FGF receptor 1 gene (Musci et al., cytological localization of the receptor or the ligand. How- 1990) was subcloned into pTS249 and pKNA1 (pG-FGFR and PA- ever, ssFGF expressed in yeast was not detected in the FGFR, respectively). medium by immunoblotting with anti-FGF antibody, sug- The full-length construct XFR and the dominant negative construct gesting that the FGF receptor might be activated in intra- XFD of the FGF receptor have been described by Amaya et al. (1991). The deleted mutant construct HAV@ has been described by Amaya cellular compartments. This property made the assay et al. (1993). strictly cell autonomous. Theoretically, there is no reason The vector plasmid pCS2+ (Turner and Weintraub, 1994) was used why ligands and receptors could not meet in intracellular for in vitro transcription of the FRL7 and FRLP genes. The coding vesicles and signal in the cytoplasm. In mammalian cells, regions were amplified and tagged with the FLAG sequence (Hopp the v-sis oncogene, a dimer of PDGF 0 chains, has been et al., 1986) at the C-terminus by PCR and inserted to pCSP+. shown to transform cells without secretion by activating lmmunoblotting the PDGF receptor in intracellular compartments (Keating Anti-phosphotyrosine antibody 4G IO was purchased from Upstate Bio- and Williams, 1988; Westermark and Heldin, 1991). The technology, Incorporated. Anti-alkaline phosphatase was purchased receptor may be activated by FGF in yeast and higher from Medix Biotech. Anti-FLAG M2 antibody was purchased from Ko- organisms in the similar manner, where activation of the dak. Polyclonal rabbit serum against FGF2 was raised against bacterially expressed Xenopus FGF2 protein (Kimelman et al., 1988). receptor could occur by an encounter with the ligand in For colony immunoblotting, we followed the procedure described the endoplasmic reticulum, Golgi, or other vesicular com- by Lyons and Nelson (1984). Colonies were transferred onto two nitro- partment. cellulose membranes (Millipore HATF 082) and incubated on SD and To our surprise, the initial screen for FGF ligands did SG plates overnight at 30%. The cells were lysed with 0.1% SDS, 0.2 M NaOH, 35 mM DTT. Chemiluminescence reagents (ECL, Amer- not produce typical FGF ligands, despite the demonstra- sham) were used for detection. tion that FGFP is active in the assay. Part of that reason may be that the level of maternal mRNA encoding FGF cDNA Libraries is greatly reduced after oocyte maturation (Kimelman and Two cDNA libraries, the egg and embryo library and the XTC library, were used. The vector plasmid of the cDNA library was hYES (Elledge Kirschner, 1989), and the other FGF forms may be quite et al., 1991). rare. It might be that the signal sequence of these FGFs cDNA for the XTC library was isolated from XTC cells and XTC might not be functional in yeast. Finally, it could be that cells treated with 100 nglml nocodazole for 24 hr. Untreated and noco- owing to glycosylation differences, FGFs synthesized in dazole-treated mRNAs were used in the ratio 4:l. cDNA for the egg yeast may have low affinities for the receptor. Three other and embryo library was isolated from eggs, 10 hr embryos (gastrula), and 10 hr embryos treated with IO pglml nocodazole between 8 and proteins activated the receptor. There are avariety of ways IO hr. mRNA from each was used in the ratio 1:3:1. a protease and RNA binding-proteins might affect the level of expression or its function. Assays Ca2+ Release Assay It has become increasingly clear that several signaling We followed the method previously described by Musci et al. (1990) systems operate in the early embryo to specify the pattern and Amaya et al. (1991). Injected oocytes were cultured at 16% for and differentiation of tissues. The existence of large fami- 48 hr and washed in Ca2+-free solution. %a*+ (Amersham) was loaded lies of receptors and ligands is not likely to reflect simply for 3 hr and then washed extensively. The radioactivity released for redundancy but different signals for different processes. the first 10 min was counted. in Vitro Binding Assay In general, it is not possible to identify which ligand and A DNA fragment encoding the extracellular domain of the FGF receptor receptor interact in vivo. The discovery of novel FGF recep- (from Met-l to Glu-372) was fused to the human alkaline phosphatase tor ligands along with the multiplicity of FGF-type receptors gene (Flanagan and Leder, 1990). COS cells were transiently transfected intensifies this issue. The results of this investigation sug- with the plasmid by the calcium phosphate method. The fusion protein was precipitated by anti-alkaline phosphatase antibody (Medix Bio- gest that receptors may be even more promiscuous than tech) and protein A-Sepharose (Pierce). [“Slmethionine-labeled FRLI simply operating within a single family of ligands. To iden- and FRL2 proteins were synthesized in reticulocyte lysate with canine Novel FGF Receptor Ligands in Xenopus Development 629

pancreatic microsomal membranes (Promega). PEtS (10 vol), 0.5% of functionally important residues and regions, Biochim. Bio#phys. Acta Triton X-100 was added and mixed to the receptor-clonjugated Sepha- 1162, 177-186. rose described above. To test the competition with FGF for binding Brandli, A.W., and Kirschner, M.W. (1995). Molecular clonllng of tyro- to the receptor, 1 pglml of bacterially produced Xenopus FGF2 protein Sine kinases in the early Xenopus embryo: identification of F&-related was added. After overnight incubation at 4”C, 2 mM of DTSSP (Pierce) genes expressed in cranial neural crest cells of the second (hyoid) was added and incubated for 1 hr. The Sepharose beads were washed arch. Dev. Dyn. 203, 119-140. four times in TBS, 0.5% Triton X-100. Byers, S., Amaya, E., Munro, S., and Blaschuk, 0. (1992). Fibroblast Animal Cap Assay growth factor receptors contain a conserved HAV region c#ommon to RNA was injected into both blastomeres at the 2-cell stage. Animal cadherins and influenza strain A hemagglutinins: a role in protein- caps were isolated at stage 8 (Nieuwkoop and Faber, 1994), incubated protein interactions? Dev. Biol. 152, 411-414. in 1 x MMR at 16OC until sibling embryos reached stage20, and then frozen for the RNA analysis. Carlson, M., Taussig, R., Kustu, S., and Botstein, D. (1983). The se- RT-PCR creted form of invertase in Saccharomyces cerevisiae is synthesized RT-PCR was carried out as described (Rupp and Weintraub, 1991; from mRNA encoding a signal sequence. Mol. Cell. Biol. 3, 439-447. Wilson and Melton, 1994; Lustig and Kirschner, 1995). The primer Carsana, A., Confalone, E., Palmieri, M., Libonati, M., and Furia, A. sequences used for analysis of EFl a, goosecoid, N-CAM, and muscle (1988). Structure of the bovine pancreatic ribonuclease gene: the actin were as described (Hemmati-Erivanlou et al., 1994; Hemmati- unique intervening sequence in the 5’ untranslated region contains a Brivanlou and Melton, 1994). The primer sequences for FRLl are 5’- promoter-like element. Nucl. Acids Res. 16, 5491-5502. ATGCAGTTTTTAAGATTT-3’ and 5’-TTAAAGTCCAATATTCAG-3’. Cheng, H.J., and Flanagan, J.G. (1994). Identification and cloning of The sequences for FRLP are 5’-ATGCTTGACATTATGGTG-3’ and 5’- ELF-I, a developmentally expressed ligand for the Mek4 and Sek TTATCTCATAGCAGGGAG-3’. The sequences for ODC are 5’-AATG- receptor tyrosine kinases. Cell 79, 157-168. GATTTCAGAGACCA-3’ and S-CCAAGGCTAAAGTTGCAG3’ (Bas- Cho, K.W., Blumberg, B., Steinbeisser, H., and De Robertis, E.M. sez et al., 1990). (1991). Molecular nature of Spemann’s organizer: the role of lhe Xeno- Paracrine Signaling Assay pus homeobox gene goosecoid. Cell 67, 1111-l 120. The paracrine signaling assay was carried out as described (Lustig Ciccodicola, A., Dono, R., Obici, S., Simeone, A., 20110, M., and Per- and Kirschner, 1995). sico, M.G. (1989). Molecular characterization of a gene of the “EGF Whole.Mount In Situ Hybridization family” expressed in undifferentiated human NTERA2 teratocarci- The procedure for the whole mount was carried out as described (Har- noma cells. EMBO J. 8, 1987-1991. land, 1991). The antisense strands of the FRLI and FR1-2coding region were used for the probes. Cornell, R.A., and Kimelman, D. (1994). Activin-mediated mesoderm induction requires FGF. Development 120, 453-462. Acknowledgments Dono, R., Scalera, L., Pacifico, F., Acampora, D., Persico, M.G., and Simeone, A. (1993). The murine cripto gene: expression during meso- Correspondence should be addressed to M. W. K. We thank Tim derm induction and early heart morphogenesis. Development 118, Stearns for early discussion and critical suggestions for the yeast 1157-l 168. assay. We thank Kevin Lustig, Robert Davis, and Kris IKroll for reading Elledge, S.J., Mulligan, J.T., Ramer, SW., Spottswood, M., and Davis, the manuscript and for their insightful comments, and Enrique Amaya R.W. (1991). IYES: a multifunctional cDNA expression vector for the for instructing N. K. in the frog experiments and for discussion. We isolation of genes by complementation of yeast and Escherichia co/i are grateful to John Flanagan and the members in his laboratory for mutations. Proc. Natl. Acad. Sci. USA 88, 1731-1735. the helpful advice and the gift of the AP-Tag plasmid, to Futoshi Shiba- Flanagan, J.G., ano Leder, P. (1990). The kit ligand: a cell surface saki for COS cells, to Paris Ataliotis and Mark Mercola for the Xenopus molecule altered in steel mutant fibroblasts. Cell 63, 185-194. PDGF receptor plasmid, and to Teresita Bernal and Rebecca Ramos Folkman, J., and Klagsburn, M. (1987). Angiogenic factors. Science for the technical assistance. We wish to acknowledge all the help, 235, 442-447. support, and advice of the former and current members of our laboratory. This work was supported by a fellowship from the International Gal, S., and Gottesman, M.M. (1988). Isolation and sequence of a Human Frontier Science Program to N. K., the National Institutes of cDNA for human pro-(cathepsin L). Biochem. J. 253, 303-306. Health, Department of Health and Human Services-National Heart, Gallwitz, D., Perrin, F., and Seidel, R. (1981). The actin gene in yeast Lung and Blood Institute to M. W. K., and a generous glftfrom Chiron Saccharomyces cerevisiae: S’and 3’end mapping, flanking and puta- Corporation. tive regulatory sequences. Nucl. Acids Res. 9, 6339-6350. Hao, Q.L., Heisterkamp, N., and Groffen, J. (1989). Isolation and se- Received July 26, 1995; revised October 20, 1995 quence analysis of a novel human tyrosine kinase gene. Mol. Cell. Biol. 9, 1587-1593. References Harland, R.M. (1991). In situ hybridization: an improved whole mount method for Xenopus embryos. Meth. Cell Biol. 36, 685-695. Amaya, E., Musci, T.J., and Kirschner, M.W. (1991). Expression of a Hemmati-Brivanlou, A., and Melton, D.A. (1994). Inhibition of activin dominant negative mutant of the FGF receptor disrupts mesoderm receptor signaling promotes neuralization in Xenopus. Cell 77, 273- formation in Xenopus embryos. Cell 66, 257-270. 281. Amaya, E., Stem, P.A., Musci, T.J., and Kirschner, M.W. (1993). FGF Hemmati-Brivanlou, A., Kelly, O.G., and Melton, D.A. (199,4). Fol- signaling in the early specification of mesoderm in Xenopus. Develop- listatin, an antagonist of activin, is expressed in the Spemann orga- ment 118, 477-487. nizer and displays direct neuralizing activity. Cell 77, 283-295. Bassez, T., Paris, J., Omilli, F., Dorel, C., and Osborne, H.B. (1990). Hopp, T.P., Prickett, K.S., Price, V., Libby, R.T., March, C.J., Cerretti, Post-transcriptional regulation of ornithine decarboxylase in Xenopus P., Urdal, D.L., and Conlon, P.J. (1986). A short peptide marker se- laevis oocytes. Development 170, 955-962. quence useful for recombinant protein identification and purification. Beintema, J.J., Schuller, C., Irie, M., andCarsana,A. (1988). Molecular Biotechnology 6, 1205-1210. evolution of the ribonuclease superfamily. Prog. Biophys. Mol. Biol. Isaacs, H.V., Tannahill, D., and Slack, J.M. (1992). Expression of a 51, 165-192. novel FGF in thexenopus embryo: a new candidate inducing factor for Bond, M.D., and Vallee, B.L. (1990). Isolation and sequencing of mesoderm formation and anteroposterior specification. Development mouseangiogenin DNA. Blochem. Biophys. Res. Commun. 171,988- 114, 71 l-720. 995. Johnson, D.E., Lee, P.L., Lu, J., and Williams, L.T. (1990). Diverse Bond, M.D., Strydom, D.J., and Vallee, B.L. (1993). Characterization forms of a receptor for acidic and basic fibroblast growth factors. Mol. and sequencing of rabbit, pig and mouse angiogenins: discernment Cell. Biol. 10, 4728-4736. Cell 630

Joseph, L.J., Chang, L.C., Stamenkovich, D., and Sukhatme, V.P. tion of nucleic acids coding for ribonucleases: the mRNA sequence (1988). Complete nucleotide and deduced amino acid sequences of of mouse pancreatic ribonuclease. Mol. Biol. Evol. 7, 29-44. human and murine preprocathepsin L: an abundant transcript induced Slack, J.M., Darlington, B.G., Heath, J.K., and Godsave, S.F. (1987). by transformation of fibroblasts. J. Clin. Invest. 81, 1621-1629. Mesoderm induction in early Xenopus embryos by heparin-binding Kay, B.K., Sawhney, R.K., and Wilson, S.H. (1990). Potential for two growth factors. Nature 326, 197-200. isoforms of the Al ribonucleoprotein in Xenopus laevis. Proc. Nab. Smyth, D.G., Stein, W.H., and Moore, S. (1963). The sequence of Acad. Sci. USA 87, 1367-1371. amino acid residues in bovine pancreatic ribonuclease: revisions and Keating, M.T., and Williams, L.T. (1988). Autocrine stimulation of intra- confirmations. J. Biol. Chem. 238, 227-237. cellular PDGF receptors in v-s&transformed cells. Science 239,914- Titani, K., Takio, K., Kuwada, M., Nitta, K., Sakakibara, F., Kawauchi, 916. H., Takayanagi, G., and Hakomori, S. (1987). Amino acid sequence Kim, Y.J., and Baker, B.S. (1993). Isolation of RRM-type RNA-binding of sialic acid binding lectin from frog (Rana catesbeiana) eggs. Bio- protein genes and the analysis of their relatedness by using a numeri- chemistry 26, 2189-2194. cal approach. Mol. Cell. Bioi. 73, 174-183. Turner, D.L., and Weintraub, H. (1994). Expression of achaete-scute Kimelman, D., and Kirschner, M. (1987). Synergisticinductionof meso- homolog 3 in Xenopus embryos converts ectodermal cells to a neural derm by FGF and TGF6 and the identification of an mRNA coding for fate. Genes Dev. 8, 1434-1447. FGF in the early Xenopus embryo. Cell 57, 869-877. Westermark, B., and Heldin, C.-H. (1991). Platelet-derived growth fac- Kimelman, D., and Kirschner, M.W. (1989). An antisense mRNA di- tor in autocrine transformation. Cancer Res. 57, 5087-5092. rects the covalent modification of the transcript encoding fibroblast Wilson, P.A., and Melton, D.A. (1994). Mesodermal patterning by an growth factor in Xenopus oocytes. Cell 59, 687-696. inducer gradient depends on secondary cell-cell communication. Kimelman, D., Abraham, J.A., Haaparanta, T., Palisi, T.M., and Curr. Biol. 4, 676-686. Kirschner, M.W. (1988). The presence of fibroblast growth factor in Wilson, R., Ainscough, R., Anderson, K., Baynes, C., Berks, M., Bon- the frog egg: its role as a natural mesoderm inducer. Science 242, field, J., Burton, J., Connell, M., Copsey, T., Cooper, J., et al. (1994). 1053-l 056. 2.2 Mb of contiguous nucleotide sequence from chromosome Ill of Krieg, P.A., Varnum, S.M., Wormington, W.M., and Melton, D.A. C. elegans. Nature 368, 32-38. (1989). The mRNA encoding elongation factor 1 CI (EF-1 a) is a major transcript at the midblastula transition in Xenopus. Dev. Biol. 133, 93- 100. Kurachi, K., Davie, E.W., Strydom, D.J., Riordan, J.F., and Vallee, B.L. (1985). Sequence of the cDNA and gene for angiogenin, a human angiogenesis factor. Biochemistry 24, 5494-5499. Kyte, J., and Doolittle, R.F. (1982). A simple method for displaying the hydropathic character of a protein. J. Mol. Biol. 757, 105-132. LaBonne, C., and Whitman, M. (1994). Mesoderm induction by activin requires FGF-mediated intracellular signals, Development 120, 463- 472. Lustig, K.D., and Kirschner, M.W. (1995). Reconstruction of inductive signaling using an oocyte expression assay, Proc. Natt. Acad. Sci. USA 92, 6234-6238. Lyman, SD., James, L., Vanden Bos, T., de Vries, P., Brasel, K., Gliniak, B., Hollingsworth, L.T., Picha, K.S., McKenna, H.J., Splett, R.R.,etal.(1993). Molecularcloningofaligandfortheflt3/flk-2tyrosine kinase receptor: a proliferative factor for primitive hematopoietic cells. Cell 75, 1157-I 167. Lyons, S., and-Nelson, N. (1984). An immunological method for detecting gene expression in yeast colonies. Proc. Natl. Acad. Sci. USA 87, 7426-7430. Maes, P., Damart, D., Rommens, C., Montreuil, J., Spik, G., andTartar, A. (1988). The complete amino acid sequence of bovine milk angiogenin. FEBS Lett. 247, 41-45. Morgan, D.O., Edman, J.C., Standring, D.N., Fried,V.A., Smith, M.C., Roth, R.A., and Rutter, W.J. (1987). Insulin-like growth factor II receptor as a multifunctional binding protein. Nature 329, 301-307. Musci, T.J., Amaya, E., and Kirschner, M.W. (1990). Regulation of the fibroblast growth factor receptor in ear1yXenopu.s embryos. Proc. Natl. Acad. Sci. USA 87, 8365-8369. Nieuwkoop, P.D., and Faber, J. (1994). Normal table ofXenopus/aevis (New York: Daudin and Garland Publishing). Rechler, M.M., Zapf, J., Nissley, S.P., Froesch, E.R., Moses, A.C., Podskalny, J.M., Schilling, E.E., and Humbel, R.E. (1980). Interaction of insulin growth factors I and II and multiplication-stimulating activity with receptors and serum carrier proteins, Endocrinology 707, 1451- 1459. Risau, W. (1990). Angiogenicgrowth factors. Prog. Growth Factor Res. 2, 71-79. Rupp, R.A., and Weintraub, H. (1991). Ubiquitous MyoD transcription at the midblastula transition precedes induction-dependent MyoD expression in presumptive mesoderm of X. laevis. Cell 65, 927-937. Schuller, C., Nijssen, H.M., Kok, R., and Beintema, J.J. (1990). Evolu-