FGFR-4, a new member of the fibroblast growth factor receptor family, expressed in the definitive endoderm and skeletal muscle lineages of the mouse

Although it has been known for some time that cell-to-cell interactions are important in the development of all complex organisms, candidate molecules that may mediate these regulatory interactions have only recently been identified. Not surprisingly, peptide growth factors, and their receptors, are prime candidates. One group of molecules, those related to the acidic (a) and basic (b) fibroblast growth factors (FGFs), and their cognate receptors, have many features that suggest that they normally function in the control of development (for recent reviews, see Rifkin and Moscatelli, 1989; Goldfarb, 1990).

There are currently seven members of the FGF family. In addition to a-FGF and b-FGF, they include three genes isolated by their ability to transform various cell types, int-2 (Moore et al. 1986), K-FGF (Delli Bovi et al. 1987; Taira et al. 1987) and FGF-5 (Zhan et al.1988. FGF-6 was cloned by low-stringency hybridization to K-FGF (Marics et al. 1989). Finally, KGF, a mitogenic factor originally identified in conditioned medium from human embryonic fibroblasts, was cloned from expressing cells (Finch et al. 1989).

In the mouse, several FGFs are known to be expressed in development. The most extensive analysis has been performed on the spatial and temporal distribution of int-2 RNA (Wilkinson et al. 1988, 1989); however, it is also clear that a-FGF, b-FGF, K-FGF and FGF-5 are also expressed at different embryonic and fetal stages (Gonzalez et al. 1990; Hebert et al. 1990; Haub and Goldfarb, 1991). From the localization of int-2 transcripts, we have suggested that int-2 may have many different roles depending upon the context in which the gene is expressed (Wilkinson et al. 1988,1989). These roles may include tissue induction, stimulation of migration and neurotrophic activity. As yet, no definitive role has been established for any member of this family. However, the recent demonstration that several FGFs are capable of inducing mesoderm in Xenopus (Slack et al. 1987; Patemo et al. 1989), and that b-FGF is a normal component of the egg and early embryo (Kimelman et al. 1988) suggests a function in some of the earliest cellular events in the gastrulating Xenopus embryo.

Understanding the developmental roles of FGF ligands clearly requires a detailed knowledge of downstream aspects of the signalling pathway, in particular their cognate receptors. Recently, an FGF receptor (FGFR) was cloned in the chick (Lee et al. 1989). This receptor, termed FGFR-1 (as suggested by Keegan et al. (1991) consists of three immunoglobulin (Ig)-like external domains, a transmembrane domain and an interrupted intracellular protein tyrosine kinase (PTK) domain, similar to other protein tyrosine kinase receptors (Hanks, 1991). In addition, two other family members are known; FGFR-2 (bek, Kornbluth et al. 1988 and cek-3, Pasquale, 1990) and FGFR-3 (cek-2, Pasquale, 1990 and FGFR-3, Keegan et al. 1991). In addition to the diversity generated by these three distinct genes, different FGFRs are also generated by complex splicing events. Analysis of FGFR-1 cDNA clones predicts isoforms of this receptor that may be secreted or that lack the first Ig-domain (Johnson et al. 1990; Reid et al. 1990; Hou et al. 1991). Expression cloning of a KGF receptor identified a new variant of FGFR-2 with only two Ig-domains (Miki et al. 1991). Thus, it seems reasonable to conclude that different receptor types are used in vivo for reception of specific ligands.

Little is known about the in vivo receptors for any of the reported ligands. a-FGF and b-FGF bind to isoforms of FGFR-1, 2 and 3, although the recently identified KGFR variant of FGFR-2 binds bFGF poorly when compared to a-FGF and KGF (Miki et al. 1991). On the basis of receptor distribution in cell-lines, this modified form may indeed encode, in vivo, a receptor for KGF (Miki et al. 1991).

There is no reason to believe-that all ligands of the FGF family have been isolated, nor is there good evidence that the FGFR molecules so far described include receptors that are capable in vivo of binding all known FGFs. Indeed, it is likely to be difficult in the absence of detailed knowledge of the in vivo distribution of receptors and ligands to match the appropriate receptor with the appropriate ligand.

We have used reverse transcription, coupled with polymerase chain reaction (RT-PCR) to attempt to identify new FGFR members in cerebellar RNA from 17.5 day post coitum (p.c.) fetuses. At this time, the cerebellum expresses high levels of int-2 RNA (Wilkinson et al. 1989). Thus, we reasoned this tissue might be a source of an int-2 specific FGFR. We describe the isolation, full-length cDNA sequence and in vivo expression of a new member of the FGFR family, tentatively named FGFR-4, which ‘shares extensive homology to previously identified FGFRs. We argue that whereas FGFR-4 is likely to encode an active FGFR, it is unlikely to be the receptor for the int-2 protein, but is presumably the receptor for another FGF-related ligand.

Cerebellums were dissected from 17.5 days p.c. mouse fetuses and total RNA prepared by the LiCl/urea procedure (Auffray and Rougeon, 1980). First strand cDNA was synthesized using 10μg of total cerebellar RNA as template and oligo(dT) as primer (InVitrogen Red Module, according to manufacturer’s conditions).

PTKs were amplified in 20 μl of a PCR reaction containing 50mM KC1, 10mM Tris –HCl (pH8.4), 1.5mM MgCl₂, 20 μgml^-1 gelatin, 0.2mM dNTPs (Pharmacia), one-twentieth of the cDNA reaction, 0.6 units Taq polymerase (Perkin-Elmer/Cetus) and 200 ng each of the PTK-specific degenerate oligonucleotide sets (Wilks, 1989). The 5 ’ and 3 ’ primer each contained unique cloning sites as indicated below (bold indicates nucleotides which encompass the 5’[IHRDL] and 3 ’ [DVWSFG] conserved stretches of amino acids).

Samples were overlaid with mineral oil and amplified for 30 cycles using an MJ Research thermal cycler according to the following conditions: 1.5 min at 94°C, 1.5 min at 50°C, 1.5 min at 72 °C. An aliquot of the reactions was electrophoretically separated on a 2 % agarose gel and the amplification products visualized by ethidium bromide staining under ultraviolet light to confirm amplification of a band of the expected size (210 –220bp). A second round of amplification was then performed using the same conditions and 0.5 μl of the primary reaction as template in a 100 μl reaction mix. The second PCR reaction was extracted once with phenol, once with Sevag, and precipitated by the addition of a half volume of 7.5 M ammonium acetate and 2.5 volumes of 100% ethanol. Following digestion with EcoRI and BamHI, PCR reaction products were subcloned into the plasmid pGEM4 (Promega) and recombinant colonies identified by EcoRI/BamHI double digests of alkaline lysis prepared plasmid mini preps (Maniatis et al. 1982). Since the initial screening indicated that almost every colony contained an insert of the expected size, the restriction enzyme analysis was later excluded. DNA obtained from mini preps was directly sequenced using a modification of the chain termination method (Sanger et al. 1980) with T7 polymerase (Pharmacia) according to the manufacturer’s specifications.

The PCR generated PTK61 BomHI –EcoRI fragment was random primed using an oligonucleotide priming kit with [α-³²P]dCTP (Multiprime, Amersham). Labelled DNA was hybridized to 10⁶ plaques of an 8.5 day p.c. mouse fetal cDNA library (Fahmer et al. 1987). Filters were hybridized overnight at 42°C in a mixture containing 5×10⁵ctsmin^-1ml^-1 of probe, 50% formamide, 0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin, 0.2% Ficoll, 0.05 M Tris –HCl (pH7.5), 1.0M NaCl, 0.1 % sodium pyrophosphate, 1% SDS and 100 μgml^-1 yeast tRNA. Filters were then washed 3 ×15min at room temperature (2 ×SSC, 0.2% SDS), 3 ×30min at 68°C (0.2 ×SSC, 0.1% SDS) and exposed to X-ray film overnight at –70°C in the presence of an intensifying screen. Positive plaques were picked to 1ml of SM, then plaque purified through four additional rounds of screening. Insert DNA was isolated by EcoRI digestion of Lambdasorb (Promega) isolates of phage DNA and subcloned into pGEM 7Z (Promega): All cDNA inserts were sequenced on both strands as above using directed subclones and internal oligonucleotide primers.

Total RNA was prepared from fetal and adult tissues and P19 cells as described earlier; poly(A) RNA from fetal and adult tissues was prepared by oligo(dT) selection using a Fast Track (InVitrogen) mRNA isolation kit, according to the manufacturer’s instructions. For P19 cells, poly(A) RNA was prepared by a single passage through a Stratagene oligo(dT) push column. Approximately 2.5 μg of poly(A) RNA was electro-phoretically separated on a 1.2% formaldehyde denaturing agarose gel (Maniatis et al. 1982) transferred to a nylon membrane (Duralon, Stratagene), crosslinked by ultra-violet light (Church and Gilbert, 1984) and hybridized to the 5 ’ EcoRI fragment (1 –1481) which encodes the more variable extracellular region of the receptors. Southern analysis indicated that this probe only detects a single gene at high stringency. Hybridizations were performed at 45°C with 2 ×10⁶ctsmin^-1 ml^-1 of [α-³²P]dCTP-labelled probe in 50% formamide, 0.25M sodium phosphate (pH7.0), 0.25M NaCl, 7% SDS, 1mM EDTA and 10% polyethylene glycol-20000 (Sigma). Washes were as follows: 3 ×20 min at room temperature (2 ×SSC, 1.0% SDS), 2 ×30min at 68°C (2 ×SSC, 0.2% SDS) and 2 ×60min at 65°C (1 ×SSC, 0.2% SDS).

In situ hybridization was performed according to our previously published procedures (Wilkinson et al. 1987) using a T7 RNA polymerase-derived [³⁵S]UTP-labelled singlestranded antisense RNA probe which encompasses a region of the 3’untranslated region of PTK61 (3026–2459). This probe shows typical single copy hybridization on Southern blot analysis (data not shown). An α-cardiac actin specific 5’untranslated region probe (Sassoon et al. 1988) was used to identify developing skeletal and cardiac muscle. Exposure times ranged from two to ten days. Sections were photographed using a Leitz Aristoplan microscope and Fuji Velvia 50 ASA Film. Silver grains were initially photographed under dark-field illumination using a red filter. A second bright-field exposure using a blue filter was then superimposed on the dark-field image to obtain a double exposure representation of silver grains and tissue histology.

High-affinity FGFRs are members of the protein tyrosine kinase (PTK) super-family. To attempt to obtain an FGFR which might, in vivo, recognize int-2 as a ligand, we decided to clone PTK cDNAs directly from fetal RNA. int-2 is not expressed at appreciable levels in any adult tissue (Jakobovits et al. 1986). In the embryo and fetus, however, int-2 RNA accumulates at several specific sites including the mesoderm of the primitive streak, the hindbrain opposite the otocyst (rhombomeres 5 and 6), neuroblasts of the eye, sensory epithelium and supporting cells of the inner ear, and Purkinje cells of the fetal cerebellum (Wilkinson et al. 1988, 1989). On the basis of in situ hybridization, the highest levels of int-2 RNA are in the cerebellum from 14.5 days p.c. to term. Thus, we decided to attempt to clone FGFRs from cerebellar RNA at 17.5 days p.c.

To isolate cerebellar PTKs specifically, we utilized the PCR strategy of Wilks (1989). Comparison of the amino acid sequences from various PTKs has identified a highly conserved region within the catalytic domain of these proteins (for review see Hanks et al. 1988). This conservation, in combination with PCR using degenerate oligonucleotide primers, has been used to clone new members of the PTK gene family (Wilks, 1989; Partanen et al. 1990; Reid et al. 1990). The 5’ oligonucleotide encodes the amino acid sequence IHRDL and the 3’ oligonucleotide DVWSFG. Approximately sixty amino acids with several residues highly diagnostic of PTKs are encoded in the DNA sequence which lies between these two conserved regions. In addition, areas of more divergent amino acid sequences are present, which allow unambiguous assignment of novel genes to specific PTK families.

PCR amplification of cerebellar cDNA gave a strong ethidium-bromide-stained band of the expected size (210 bp, data not shown). Reaction products were cloned and sequenced on both strands. In total, sixty-two inserts were sequenced, most of which corresponded to a single PTK species, PTK-1 (Fig. 1). In addition to PTK-1, we isolated 1 to 4 independent isolates of ten different sequences (Fig. 1). One of these, PTK-75, does not conserve several features typical of PTKs including a conserved sequence AARN immediately following the IHRDL sequence encoded by the 5’ oligonucleotide. However, the presence of the sequence KPEN at this position, which is highly conserved in protein serine/threonine kinases, suggests the clone no. 75 is a putative serine/threonine kinase, although it is not identical to any published sequence (Hanks, 1991). Since there is some degree of conservation of the terminal amino acid sequences between these two super-families of kinases, it is not wholly unexpected that a serine/threonine kinase was cloned. All other cDNAs share features of PTKs, although PTK-1 and -3 have unusual features. PTK-1 does not conserve the internal amino acid sequence DFG which is absolutely conserved in all known PTKs and most serine/threonine kinases. Thus, PTK-1 may not encode a functional kinase, although all other aspects of the sequence show typical PTK conservation (Fig. 1). It is unlikely that a sequencing error or PCR-generated error is responsible for the absence of the DFG motif in PTK-1, since all clones sequenced encoded an identical sequence in this region. Moreover, we have also isolated PTK-1 in an independent study investigating PTKs in the mouse limb bud (G. Maio and A. McMahon, unpublished observation). In addition, Partanen et al. (1990) describe a human clone JTK-5, generated by RT-PCR of K562 cells, which encodes an identical amino acid sequence to PTK-1.

PTK-3, while conserving all key amino acids, shows an unusual spacing. The conserved tyrosyl residue is normally positioned 10–12 amino acids following the conserved DFG motif, whereas in PTK-3 this tyrosyl residue occurs only 8 amino acid residues after this sequence. The significance of the altered spacing is unknown.

Comparison of the predicted amino acid sequences encoded by the PTK cDNA clones with all known sequences published as of January 1991 (Hanks, 1991), indicates that seven of the eleven isolates have been previously reported. PTK-1 has been described earlier; PTK-6 is derived from platelet growth factor receptor-α (PDGFR-α, (Claesson-Welsh et al. 1989); PTK-12 and PTK-17 correspond to clones no. 22 and no. 17, respectively, amplified from a murine hemopoietic cell line FDC-P1 (Wilks, 1989). PTK-55 is derived from murine HCK (Holtzman et al. 1987); PTK-65 is derived from insulin-like growth factor 1-related receptor (IGF1R, Ullrich et al. 1986); PTK-74 is derived from FGFR-2 (Kornbluth et al. 1988; Pasquale, 1990). A second clone, PTK-61, shared a similar high level of amino acid identity to FGFR-1 (Fig. 1) and to PTK-74. Thus, it is likely that PTK-61 is a previously unidentified member of the FGFR family. The other novel clones, PTK-3 and 37, do not share greater identity to any existing PTK family.

To attempt to isolate cDNA clones encompassing a complete PTK-61 open-reading frame, we screened 10⁶ recombinant phage plaques of an early fetal mouse cDNA library (Fahrner et al. 1987) with the PTK61 PCR product. Five positive clones were obtained, the longest of which contained three EcoRI restriction sites that generated fragments of approximately 1500, 1300 and 300 nucleotides. Individual EcoRl fragments were subcloned and sequenced in their entirety, on both strands. The complete sequence of λPTK61 is shown in Fig. 2. Analysis of this sequence clearly indicates that λPTK61 is a member of the PTK family and more specifically a new member of the FGFR subclass of this family.

A long open reading frame, preceded by an in-frame stop codon, extends from nucleotide 105 to a stop codon (TGA) at position 2504 (Fig. 2). Thus, λPTK-61 encodes an 799 amino acid (90,556 M_r) polypeptide. The region encompassing nucleotides 1920-2119 is identical to the PTK61 PCR product, indicating that λPTK-61 is indeed a full-length version of the PCR-generated clone.

Two strongly hydrophobic areas are predicted. One at the amino terminus incorporates a putative signal-peptide sequence (nucleotides 105–158). The second, an internal region, encodes a putative transmembrane domain (nucleotides 1203–1265) suggesting that PTK-61 is a membrane spanning protein. The putative intracellular PTK domain is split into two parts (nucleotides 1494–1865 and 1848–2324), separated by a stretch of 14 amino acids, a feature of several classes of PTKRs (Hanks, 1991). The PTK domain shows the typical conservation expected of a functional PTK. These features include a potential ATP-binding region at the ATP-binding consensus sequence G-X-G-X-X-G (indicated by asterisks in Fig. 2), as well as conservation of a number of key amino acid residues already discussed in the catalytic domain. The extracellular domain is made up of three immunoglobulin-like domains (Fig. 2, nucleotides 261–398, 609–767, 906–1094) bounded by cysteine residues at either end with a conserved tryptophan 10–12 amino acids from the N-terminal cysteine. The C-terminal cysteine is incorporated in the amino acid sequence, D-X-G-X-Y-X-C, a conserved motif present at the C-terminal end of Ig-domains.

Comparison of the PTK-61 sequence with that of three representatives of the FGFR-family, indicates that PTK-61 shares extensive homology with this family of receptor tyrosine kinases (Fig. 3 and 4). In overall structure, two features are indicative of FGFRs; an extracellular region consisting of three Ig-domains and an intracellular kinase domain split by a fourteen amino acid spacer. Amino acid identity is highest in the TK domains I and II, 74–79% and 81–84%, respectively (Fig. 4). Conservation is also high in the Ig-II and -III domains (53–75 % and 65–71 %, respectively, Fig. 4) and the spacer region between these domains (67–78%), but falls dramatically in the Ig-I domain (26–28 %). The region between Ig-I and -II domains in PTK-61, FGFR-1, FGFR-3, and the N-terminal region of KGFR (there is no Ig-I domain in the KGFR variant of FGFR-2), contains a highly conserved region encompassing 26 amino acids. Immediately upstream of this in FGFR-1 and FGFR-3 is a highly acidic region which is not conserved in PTK-61. Thus extensive identity ends in the Ig-I–Ig-II spacer region (Fig. 3 and 4), but conservation elsewhere in the molecule supports the proposal that PTK-61 is a new member of the FGFR family.

Expression of PTK-61 was initially examined by northern blot analysis of adult and fetal RNAs (Fig. 5). A single, 3.5 kb transcript was detected in adult liver, lung and, at much lower levels, in kidney RNAs. No PTK-61 RNA was detected in adult heart, brain, testes or spleen (Fig. 5). A single 3.5 kb PTK-61 transcript was detected during fetal development with maximum relative expression occurring between 13.5 and 14.5 days p.c. We also examined the CCE mouse embryo stem-cell line (Robertson et al. 1986), which expressed low levels of PTK-61, comparable to that of the adult kidney.

Since PTK-61 was originally isolated by RT-PCR from fetal cerebellar RNA, we examined various fetal organs for expression of PTK-61. Surprisingly, we failed to detect transcripts in total RNA from 17.5 days p.c. cerebellum. In addition, eye and whole brain from similar fetal stages did not express PTK-61 RNA at detectable levels (data not shown). In contrast, PTK-61 transcripts were detected in 17.5 days p.c. fetal lung and liver (data not shown).

PTK-61 expression was examined by in situ hybridization of a single-stranded ³⁵S-UTP labelled PTK-61 RNA probe to sections of 8.5, 9.5 and 14.5 day p.c. mouse fetuses. Spatially localized transcripts were detected at all stages. At 8.5 days p.c. PTK-61 RNA in the fetus was restricted to the definitive endoderm of the developing gut (Fig. 6A and B). In addition, PTK61 was also expressed in the endodermal component of the yolk sac, which surrounds the fetus (Fig. 6B); a population of cells with a different lineage from the underlying yolk sac mesoderm (Gardner and Rossant, 1979). PTK-61 transcripts remain in the yolk sac endoderm until at least 14.5 days p.c. (data not shown).

At 9.5 days, fetal PTK-61 RNA was detected in two groups of cells. As earlier, high levels of transcripts were present in the developing gut (Fig. 6C). In addition, a repeating pattern of PTK-61 hybridization was visible along the anterior–posterior body axis within the somites (Fig. 6C). Expression in the somites was entirely restricted to the myotomal component (Fig. 6C and D), the region fated to give rise to the skeletal muscle of the body.

By 14.5 days p.c., the fetus has undergone extensive organ development and not surprisingly the pattern of PTK-61 expression shows several differences (Fig. 7). At this time, RNA was detected in many distinct regions, with the notable exception of the central and peripheral nervous systems, which never expressed detectable levels of PTK-61 RNA. Much of the expression (Figs 7, 8A) occurs in groups of cells associated with the developing axial and appendicular skeleton, which resembled developing muscles in both position and histological appearance. In addition, transcripts were detected in cartilage, kidney, adrenal gland, liver, gut, pancreas and lung (Figs 7, 8).

To verify that PTK-61 was indeed expressed in developing skeletal muscle, we took sections separated by approximately 80 μm and hybridized one with PTK-61 and the other with a muscle-specific α-cardiac actin probe which detects transcripts in both cardiac and skeletal muscle (Sassoon et al. 1988). Comparison of these results (Fig. 8A and B), and of sections through the entire fetus (data not shown), clearly demonstrates that PTK-61 transcripts were present in all skeletal muscle. Moreover, expression is restricted to this muscle type. No expression was seen in the heart (Fig. 8A) or the developing smooth muscle of the gut.

A second component of the skeleton, the endochondral cartilage, also showed extensive PTK-61 expression. This is most readily apparent in the cartilage of the ribs and the developing cartilage of the olfactory and auditory regions (Figs 7, 8A and C).

Continued expression of PTK-61 was observed in the endoderm of the intestine (Fig. 8D). In addition, three other tissues with endodermal components expressed PTK-61; the pancreas (Fig. 8D), the liver (Fig. 8D), and the lung (Fig. 8F). In the liver, expression was widespread throughout the organ, whereas in the pancreas, expression was restricted to the endodermal epithelium. In the lung, much of the mesenchymal and epithelial components express PTK-61, with the exception of the epithelium and associated mesenchyme of the bronchioles and bronchi (Fig. 8F). Finally, the adrenal cortex and the collecting tubules of the kidney, which are both mesodermally derived, showed high levels of PTK-61 transcripts (Fig. 8E).

Expression of PTK-61 in the myotome, and later in skeletal muscle, suggests that PTK-61 may play some part in the development of muscle. To address this issue further, we examined PTK-61 expression in the P19 embryonal carcinoma cell line, which on aggregation in the presence of DMSO, differentiates to several cell types including both cardiac and skeletal muscle (McBurney et al. 1982; Edwards et al. 1983; Rudnicki et al. 1990). PTK-61 expression was induced in these cells 48h after treatment (Fig. 9A). In contrast, muscle specific α-cardiac actin transcripts, which are indicative of differentiated muscle, were not detected until several days later (Fig. 9B). Parallel treatment of cells with retinoic acid to induce neural cell types did not lead to activation of PTK-61 (data not shown). Thus, induction of PTK-61 in differentiating P19 cells occurs rapidly and is specific to cells stimulated to differentiate by DMSO.

We have described the cloning of an FGFR-related cDNA, PTK-61. Several features suggest that this is a new member of the FGFR family of receptor protein tyrosine kinases. PTK-61 shares a typical FGFR structure consisting of an extracellular domain containing three Ig domains, a transmembrane domain and an intracellular PTK domain with a 14 amino acid interruption. Moreover, PTK-61 displays high levels of amino acid identity, with previously identified FGFR-1, -2 and -3 sequences, particularly in the Ig-II, Ig-III and PTK domains. Indeed, the amino acid identity is comparable to the amino acid identity amongst these known FGFRs. The Ig-I domain is highly divergent, but this is a general feature of the family. In addition, PTK-61 does not conserve the acid-rich region between Ig-I and -II domains. It is not known how differences in these regions affect ligand binding, but there is evidence to suggest that variability in the Ig-I and Ig-III domains may contribute to ligand specificity (Dionne et al. 1990; Miki et al. 1991). In summary, although we have not formally proven that PTK-61 binds a member of the FGF family of putative ligands, we feel confident on the basis of the observed structural similarities to consider PTK-61 a new member of this family of receptors. Consequently, in line with the nomenclature suggested by Keeganet al. (1991), PTK-61 is renamed FGFR-4. After submission of this manuscript, Partanen et al. (1991) reported the isolation of a human clone also designated FGFR-4 which by sequence comparison and expression analysis is clearly the human counterpart of murine FGFR-4.

The initial premise on which this work was initiated was to restrict the PCR analysis to a tissue in which only one specific FGF-related ligand is known to be expressed, thereby biasing the receptor search towards a putative FGFR which might function in vivo as a receptor for the ligand of interest. Thus, in starting with fetal cerebellar RNA in which int-2 is expressed at high levels, any FGFRs isolated from this tissue become candidates for the int-2 receptor. Since any given receptor binds multiple ligands in vitro (Dionne et al. 1990; Johnson et al. 1990; Mansukhani et al. 1990; Miki et al. 1991), assigning ligand specificity to a given receptor is a difficult problem. Moreover, this problem is further compounded by the many different isoforms that exist for the FGFR-1 and FGFR-2 receptors (Dionne et al. 1990; Johnson et al. 1990; Reid et al. 1990; Hou et al. 1991; Miki et al. 1991).

Although the original PCR fragment was generated from cerebellar tissue, we were unable to detect FGFR-4 expression in fetal cerebellum, whereas we were able to detect transcripts in fetal lung and liver at 17.5 days p.c. Expression in these tissues was in agreement with our more detailed in situ hybridization studies at early stages. Moreover, we did not detect expression in fetal eye, which like the cerebellum, is a normal site of int-2 expression. Thus, as these studies failed to reveal any overlap whatsoever in the expression of int-2 (Wilkinson et al. 1988, 1989) and FGFR-4 at both early and late fetal stages, it seems highly unlikely that FGFR-4 is the int-2 receptor.

The only other PCR-generated fragment isolated in our study with significant homology to FGFRs was PTK-74, a sequence identical to part of the PTK region of the FGFR-2/bek class of receptor. Thus, it is possible that, if an int-2 receptor resides in the fetal cerebellum, it may be a variant of FGFR-2. Evidence exists that different variants of FGFR-2 are capable of binding at least three ligands, b-FGF, a-FGF and KGF (Dionne et al. 1990; Miki et al. 1991) but no information on int-2 binding is available. It is therefore possible that in vivo different ligands may interact with the same receptor producing a complex arrangement of receptor-ligand interactions.

The in vivo distribution of other FGF-members has not been thoroughly studied. Immunolocalization of b-FGF in the rat (Gonzalez et al. 1990) and chick (Joseph-Silverstein et al. 1989) give conflicting results. The former study reports a widespread distribution of b-FGF at 18 days p.c. in the basement membrane of many tissue types; the latter study reports a more restricted distribution limited to cardiac and skeletal muscle. FGF-5 has been studied in more detail (Haub and Goldfarb, 1991) and shows many specific sites of RNA accumulation, which include several mesodermal populations including specific facial muscles. Thus, although aspects of the expression of these ligands shows some concordance with expression of FGFR-4, the general correlation is poor. For example, while FGFR-4 is expressed in liver, and gut endoderm, none of the above ligands are expressed in these tissues. In contrast, FGFR-4 is expressed in skeletal muscle. However, FGF-5 transcripts are restricted to discrete skeletal muscle populations and b-FGF is also expressed at high levels in cardiac muscle. Thus, from the limited data available, it would be surprising if FGFR-4 is an exclusive receptor for either of these ligands. A more reasonable explanation is that one or more of the other known members of the family, or a form not yet identified, are in vivo ligands for FGFR-4. Given that no systematic search for FGF-related molecules has been performed, the family of ligands may be considerably larger than is currently known.

Several aspects of the fetal expression of FGFR-4 are developmentally interesting. Previous evidence suggests that FGF-mediated regulation may be restricted to tissues of a neural–ectodermal, ectodermal or mesodermal origin (reviewed in Rifkin and Moscatelli, 1989; Goldfarb, 1990). However, the demonstration here that FGFR-4 is expressed in definitive endoderm cells of the gut at early somite stages and in the gut endoderm, and endoderm derivatives at mid-fetal life, points to a role for FGFR-4 in the regulation of endoderm. The exact role of FGFR-4 remains to be elucidated, but the very early expression suggests that FGFR-4 may be involved in some basic aspect of the establishment of this cell lineage.

Aside from endoderm-associated expression, the second major site of FGFR-4 transcript accumulation was skeletal muscle. Expression was detected at 9.5 days in the myotomal component of the somite. By 14.5 days, all skeletal muscle expressed high levels of FGFR-4. In contrast, no expression was detected in cardiac muscle. Recently, a number of myogenic regulatory genes have been described, all of which are members of a helix-loop-helix (HLH) family of DNA-binding proteins (Davis et al. 1987; Wright et al. 1989; Edmondson and Olson, 1989; Braun et al. 1989; Rhodes and Konieczny, 1989; Miner and Wold, 1990; Braun et al. 1990). A common property of the HLH-myogenic genes is the ability to trigger skeletal muscle differentiation (Davis et al. 1987; Rhodes and Konieczny, 1989; Wright et al. 1989; Edmondson and Olson, 1989; Braun et al. 1989). Interestingly, many of these genes are also expressed at early somite stages exclusively in myotomal cells (Wright et al. 1989; Sassoon et al. 1989; Hinterberger et al. 1991). Moreover, all HLH-myogenic genes described are expressed exclusively in skeletal muscle. Thus, it seems reasonable to suggest that FGFR-4 may play a role, along with the myogenic genes, in the regulation of skeletal muscle development. This view is substantiated by the P19 experiments. On aggregation of P19 cells in the presence of DMSO, FGFR-4 shows a rapid induction (48 h) which precedes the development of differentiated muscle cells (7 days), suggesting that FGFR-4 may act early in the pathway of skeletal muscle differentiation. FGFR-4 expression in the P19 cultures is weak, but this may reflect the relative proportions of muscle precursors, as cardiac muscle is the main muscle type induced in these experiments (McBurney et al. 1982; Edwards et al. 1983; Rudnicki et al. 1990). Alternatively, FGFR-4 may be expressed in other cell types as differentiation under the described conditions results in the formation of some non-muscle cell lineages. The P19 system should provide a useful cell culture model for examination of FGFR-4 induction.

Several lines of evidence link the FGF-family to muscle differentiation. In Xenopus embryo explants, multiple FGF ligands are capable of inducing muscle development (Slack et al. 1987; Paterno et al. 1989) in addition to other mesodermal cell types. In the chick, differentiation of early limb-bud cells, which contain skeletal muscle precursors, is suppressed by FGFs (Seed and Hauschka, 1988). Further, cultures of several mouse myoblast lines are responsive to both a-FGF and b-FGF. In the presence of these molecules, cells remain as undifferentiated myoblasts, possibly as a direct result of suppression of myogenic genes such as myogenin by FGF-mediated signalling (Brunetti and Goldfine,1990). However, when FGF is withdrawn, cells differentiate rapidly to mature muscle within a few hours (Linkhart et al. 1981; Clegg et al. 1987). Terminal differentiation is accompanied by a concomitant downregulation of FGFRs on the cell surface (Olwin and Hauschka, 1988; Moore et al. 1991).

Thus, the available evidence implicates FGF ligands and receptors in the regulation of early muscle cell types, observations consistent with the expression of FGFR-4 in myotomal cells reported here. Moreover, close examination of hybridization of FGFR-4 and α-cardiac actin probes to 14.5 day muscle suggests that, within the developing muscle block, the two probes recognize different cell types. FGFR-4 hybridization is punctate within the block, indicative of scattered single cells. In contrast, α-cardiac actin, which is expressed on differentiation of myoblasts, shows linear arrays of hybridizing cells, indicative of developing myotubes. Thus, it is likely that FGFR-4 expression is restricted to early myotomal cells and myoblasts and is not expressed in myotubes.

In summary, we have described the cloning and developmental expression of a new member of the FGFR-family. Our data suggest that this putative receptor plays a role in the development of endodermal and muscle lineages. From this study and previous reports, it is clear that the FGFR-family consists of multiple genes, as well as multiple isoforms. Identifying the in vivo ligands for these receptors and determining their normal biological role(s) is becoming an increasingly demanding challenge. The recent discovery of a Drosophila member of the FGFR-family (Glazer and Shilo, 1991) suggests that these studies will also be pertinent to non-vertebrate species.

We would like to thank René St-Arnaud for the gift of P19 RNA blots, Steve Hanks for discussion on PTKs, Barbara Kerr for typing of the manuscript and Alice O’Connor for help with several figures.

PCR clones encoding the two novel PTKs, PTK-3 and PTK-37, isolated in the initial cerebellar screen have recently been cloned by two other groups. PTK-3 is identical to TK3 (Wilkie, T. M. and Simon, M. I. [1991], Cloning multigene families with degenerate PCR primers. In Methods: A Companion to Methods in Enzymology, Vol. 2, in press). PTK-37 is identical to YK5 (Wilkie and Simon, 1991) and tyro-3 (Lai, C. and Lemke, G. [1991]. An extended family of proteintyrosine kinase genes differentially expressed in the vertebrate nervous system. Neuron6, 691-704).