Two FGF-receptor homologues of Drosophila: one is expressed in mesodermal primordium in early embryos

Fibroblast growth factors (FGFs) constitute a growth factor family of at least seven members (Burgess and Maciag, 1989). In mammals, FGFs are not only capable of inducing cell proliferation and chemotaxis (Burgess and Maciag, 1989), but also are potent inducers of angiogenesis (formation of new blood vessel; Cross and Dexter, 1991). Response to FGFs is mediated by high-affinity cell-surface receptors (FGF-Rs) through activation of intracellular tyrosine kinase (Lee et al., 1989). FGF-Rs are widely distributed in various embryonic cells in mammals so that mammalian embryogenesis may require pleiotropic functions of FGFs (Wanaka et al., 1991).

The FGF/FGF-R signaling system has also been shown quite likely essential for mesoderm formation in Xenopus (Jessell and Melton, 1992; Whitman and Melton, 1992). Most, if not all, FGFs can mimic the mesoderm-inducing activity of vegetal pole cells in the early blastula and, hence, may possibly be natural inducers of mesoderm formation (Slack et al., 1987; Kimelman et al., 1988). An injection of mRNA encoding a dominant-negative form of FGF-R into embryos has been demonstrated to cause significant deformation in mesodermal structures of tadpoles (Amaya et al., 1991).

Molecular analysis of FGF-Rs in chicken, humans and Xenopus has indicated that FGF-Rs constitute a family of receptor-tyrosine kinases (Lee et al., 1989; Ruta et al., 1989; Seno et al., 1991; Friesel et al., 1991). Each FGF-R mole-cule has more than two extracellular immunoglobulin-like domains (Ig-domains) as is also the case for receptors of the platelet-derived growth factor (PDGF) and colony stimulating factor-1 (CSF-1; Williams, 1988). However, unlike receptors for PDGF and CSF-1, receptors for FGFs possess in common additional characteristics such as a relatively long juxtamembrane region, kinase catalytic sequences split by a short stretch of amino acids and a short carboxyl terminal tail.

Understanding the physiological roles of the FGF/FGF-R signaling system may be facilitated by clarification of the roles of homologous genes in Drosophila, in which both genetic and developmental analyses are applicable. In a previous experiment (Shishido et al., 1991), we searched for genomic DNA fragments of Drosophila, encoding novel tyrosine kinases, by polymerase chain reaction (PCR). Two of the seven DNA fragments isolated (dtk1 and dtk2) were found partly to encode polypeptides highly homologous in amino acid sequence to vertebrate FGF-Rs. Using a different approach, Glazer and Shilo (1991) independently identified a Drosophila FGF-R homologue (DFGF-R). Unlike our dtk1 and dtk2 products, the D-FGFR protein appeared to have a long substitution at or near subdomain VIII, a tyrosine kinase subdomain conserved in all tyrosine kinases so far examined (Hanks et al., 1988).

For clarification of the above and further extension of our analysis, complete nucleotide sequences of the coding regions of dtk1 and dtk2 genes were determined and their expression patterns in embryos were examined. dtk1 and dtk2, respectively, were found to be Drosophila FGF-R homologues having two and five Ig-domains. dtk1 expression was specific to mesoderm in early embryos, while dtk2, seemingly corresponding to DFGF-R, was expressed in endodermal precursors.

Canton-S (wild type) was obtained from Y. Hotta (University of Tokyo). Fly strains with twi^ID96,sna^IIG were obtained from C. Nüsslein-Volhard (Max Plank Institute, Tübingen). Strain Df(1)260-1/FM4 was obtained from UMEÅ Stock Center. Df(1)260-1 is a mutation, in which the whole AS-C region is deficient (Gonzalez et al., 1989). Strain sr16 /TM1 (De La Pompa et al., 1989) and a Dmyd-lacZ transformant line, 14.1(II) (Paterson et al., 1991) were obtained from A. Ferrus (Instituto Cajal, Madrid) and W. J. Gehring (University of Basel), respectively.

Genomic DNA and cDNA clones were isolated from our genomic DNA and pupal cDNA libraries, respectively (Kojima et al., 1991). Nucleotide sequences were determined as described previously (Emori et al., 1985). All other procedures for molecular analyses were essentially as described by Sambrook et al. (1989).

Fixation and pretreatment of embryos were performed as described by Tautz and Pfeifle (1989). RNA probes were prepared as follows. cDNA inserts in cFR4-6 (DFR1), cFR5-1 ( DFR2) and PCR-amplified twist genomic DNA (Thisse et al., 1988) were cloned into Bluescript plasmid vectors (Stratagene). Purified plasmid DNA was linealized by a suitable restriction emzyme and used as a template. Antisense RNA was prepared by in vitro transcription using digoxigenin-11-UTP (Boehringer Mannheim) and T7 or T3 RNA polymerase. The reaction was carried out according to the manufacturer’s standard protocol. After fragmentation into an average size of 150 nucleotides by incubation at 60°C in a buffer containing 40 mM NaHCO₃, 60 mM Na₂CO₃, and 50 mM DTT, RNA was used as a probe for hybridization.

Hybridization was carried out essentially as described by Yokouchi et al. (1991). Embryos were incubated overnight at 50°C in a solution containing 20 mM Tris-HCl (pH 8.0), 2.5 mM EDTA (pH 8.0), 1× Denhardt’s solution (0.02% BSA, 0.02% Ficoll,

0.02% polyvinylpyrrolidone), 300 mM NaCl, 1 mg/ml Escherichia coli tRNA, 50% formamide, 10% dextran sulfate and 0.5 μg/ml of digoxigenin-labeled RNA probe. After hybridization, embryos were washed with 2× SSC containing 50% formamide for 1 hour at 50°C, followed by RNase treatment in 10 mM Tris-HCl (pH 8.0) with 500 mM NaCl and 20 μg/ml RNaseA for 30 minutes at room temperature. Then, embryos were washed again with 2× SSC containing 50% formamide for 1 hour at 50°C and then with 1× SSC in 50% formamide for 1 hour at 50°C. Immunological detection was carried out essentially as described by Tautz and Pfeifle (1989).

To examine DFR1 expression in AS-C embryos, crosses were made with Df(1)260-1/+ females derived from Df(1)260-1/FM4 and wild-type flies, and wild-type males. Consistent aberrations observed in one quarter of the embryos were noted as mutant phenotypes. P[Dmyd-lacZ]/+; sr16/+ flies were obtained by crossing flies homozygous for the P insertion (on the second chromosome) and sr16/TM1 flies. P[Dmyd-lacZ]/+; sr16/+ flies were crossed to each other and the resultant embryos (progeny) were stained with anti-β-galactosidase antibody as described (Higashijima et al., 1992). Consistent aberrations observed in one quarter of the lacZ- expressing embryos were noted as mutant phenotypes.

After dehydration through an ethanol series, stained embryos were embedded in polyester wax (BHD, England), incubated at 37°C and sectioned at 10°C. Samples thus obtained were treated with 100% ethanol and mounted in Euparal (Chroma-gesellschaft, Germany).

Using dtk1 and dtk2 fragements as probes, genomic DNA clones were isolated. cDNA clones were isolated using suitable genomic DNA probes. Since, as described below, polypeptides encoded by cDNA inserts are very similar in sequence and organization to vertebrate FGF-Rs, dtk1 and dtk2, respectively, were renamed DFR1 and DFR2 (Drosophila FGF-receptor homologues 1 and 2). In situ hybridization to polytene chromosomes showed DFR1 and DFR2 to be mapped at 90D-E and 70C, respectively (data not shown).

The complete nucleotide sequence of the cDNA insert in cFR4-6 and its relevant genomic regions were determined (Fig. 1B). As schematically shown in Fig. 1A, the DFR1 gene included no intron. The cFR4-6 insert contains a single long open reading frame (ORF) of 2251 bp, followed by 523 bp of the 3′ non-coding sequence with a truncated poly(A) tail. The analysis of genomic sequences indicated the in-frame termination codon to be located at nucleotide positions –56 to –54 and no additional in-frame initiation codon to be present within the region from –1 to –53. Regions with the first and second ATG codons (nucleotide positions, 61-67 and 106-112, respectively) matched the Drosophila consensus sequence for the translational initiation (Cavener, 1987). However, only the second ATG codon may possibly be functional, since, as in the case of vertebrate FGF-R genes, it is directly followed by a nucleotide sequence coding for a hydrophobic stretch of 22 amino acids capable of serving as a signal peptide. The expected cleavage site of the putative signal peptide may be between Ala-19 and Val-20 (von Heijne, 1983). The second hydrophobic stretch was detected in the middle of the ORF (amino acid positions, 295-318), presumably representing the transmembrane domain. Thus, the extracellular and intracellular domains of the DFR1 protein were assigned to amino acid residues 20-294 and 319-714, respectively.

As in the case of vertebrate FGF-Rs (Lee et al., 1989), the intracellular region contains a relatively long juxtamembrane domain (77 amino acids), split tyrosine kinase catalytic sequence (278 amino acids) and a short carboxyl terminal tail (41 amino acids). All 11 tyrosine kinase subdomains (I to XI; Hanks et al., 1988) were found to be conserved. The DFR1 kinase insert (19 amino acids) is slightly longer than counterparts in vertebrate FGF-Rs (14 amino acids). Except for the kinase insert, the overall sequence homology of the kinase domain between DFR1 and vertebrate FGF-Rs was estimated as about 60%. Homologies with other types of vertebrate tyrosine kinases were approx-imately 40%. The carboxyl terminal tail of DFR1 includes a tripeptide sequence, Tyr-Leu-Asp, which may provide the tyrosine phosphorylation site supposedly required for binding of SH2 of phosphoinositide-specific phospholipaseC-γ (Mohammadi et al., 1991).

The extracellular domains of FGF-Rs are, in general, characterized by a series of Ig-domains, each containing two Cys residues, possibly connected by a disulfide bond, in an interval of several scores of residues capable of constituting β-sheet structures (Williams and Barclay, 1988). The consensus sequence for the Ig-domains is Vx(I/L)xC(8x-12x)W(20x-50x)DxGxYxC (x=arbitrary amino acid). As shown in Figs 2 and 3, the DFR1 protein contains two Ig-domains. (Most vertebrate FGF-Rs have two or three Ig-domains (Lee et al., 1989; Ruta et al., 1989; Seno et al., 1991).) As in the case of vertebrate FGF-Rs, DFR1 extracellular domains contain a short stretch of acidic amino acids (Fig. 3). Locations of two potential N-glycosylation sites were found conserved between DFR1 and the flg, one of human FGF-Rs (Ruta et al., 1989).

Using a human FGF-R cDNA (Seno et al., 1991) as a probe, Southern blot analysis of the total genomic DNA of Drosophila was carried out. DFR1-derived fragments were found to be hybridized in all lanes under a less stringent condition (data not shown). From the above, the DFR1 gene would appear quite likely to be a counterpart of vertebrate FGF-Rs in Drosophila.

We next determined the nucleotide sequences of the cDNA insert in cFR5-1 and its relevant genomic regions. As schematically shown in Fig. 4A, the DFR2 gene had an intron 85 bp long. The DFR2 cDNA contained a single long ORF followed by 33 bp long, 3′ non-coding sequence. Since neither possible signal peptide sequence nor probable translational initiation codon was found to be coded for by the cDNA, we sought to determine further 5′ sequences using a DFR2 genomic clone. Fig. 4B shows the composite nucleotide sequence thus obtained, which contains a 3159 bp long ORF following an in-frame termination codon of nucleotide positions, 121-123.

19 amino acids following the first methionine (amino acid positions, 2-20) may serve as a signal peptide, being highly hydrophobic. The second hydrophobic stretch, including 25 amino acids (amino acid positions, 601-625) may represent the transmembrane region, since it is followed by a putative tyrosine kinase sequence as described below.

As noted for DFR1, the intracellular region of DFR2 con-tained a split tyrosine kinase catalytic domain (Figs 3, 4B), highly homologous to those of vertebrate FGF-Rs. Homology between DFR1 and DFR2 kinase domains was 79%. The Tyr-Leu-Asp sequence was found near the carboxyl terminus of DFR2. The DFR2 extracellular region included five Ig-domains, of which two were found in the region flanked by the acidic domain and transmembrane domain (Figs 2, 3). Three and four of nine N-glycosylation sites of DFR2 were conserved in both DFR1 and flg, respectively (Fig. 3).

DFR2 is virtually identical in nucleotide sequence to DFGF-R, whose sequence has been reported by Glazer and Shilo (1991). However, their original DFGF-R sequence appears to include some defects in two regions. One base deletion in an C stretch of DFGF-R (nucleotide positions, 865-868 in Fig. 4B) disrupted the coding region for the extracellular domain. Nucleotide changes in the region from 2850 to 2892 (see Fig. 4B) results in loss of the invariant glutamic acid of subdomain VIII of DFGF-R.

The spatial and temporal expression of DFR1 RNA was examined by in situ hybridization of whole embryos. The expression of DFR1 RNA was first detected at the cellular blastoderm stage (stage 5; Campos-Ortega and Hartenstein, 1985), about 150 minutes following egg fertilization, just before the onset of gastrulation. The ventral surface of the embryo, 16-17 cells wide and from 10 to 85% egg length, was hybridized with the DFR1 probe (Fig. 5A,B,D). This region of DFR1-expressing cells corresponds to mesodermal primordium, which eventually participates in the ventral furrow formation. Note that at stages 6-7 (about 3 hours after egg fertilization), all DFR1-expressing cells invaginated to form mesoderm (Fig. 5C,E). Fig. 5F shows the staining pattern of an embryo at stage 8 (3 hours 30 minutes; stage for germ-band extension). The anterior part of the DFR1-positive region appears to be separate from the main body in forming the cephalic mesoderm. This is most probably due to invagination of midgut primordium (see an arrowhead labeled am in Fig. 5F). At stage 9 (about 4 hours), the DFR1-positive region began to segment (Fig. 5G). These findings indicate the expression of DFR1 to be mesoderm-specific at least in early embryos.

The invaginated mesodermal cells mainly differentiate into somatic and visceral muscles (Campos-Ortega and Hartenstein, 1985). DFR1 expression in somatic mesoderm is described in the following.

DFR1 expression dramatically changes at stage 11 (6-7 hours), when germ-band fully extends and intersegmental furrows become much more discernible. As shown in Fig. 6A, the overall expression of DFR1 in the mesoderm is virtually extinct and, instead, a small number of cells strongly expressing DFR1 emerge in the ventrolateral region of each segment. Since such DFR1-positive cells were found in all of 50 embryos derived from AS-C /+ parents, the achaete-scute gene products may not be required for their presence. Thus, these DFR1-positive cells are suggested not to be neural cells. During germband retraction (stage 12), isolated DFR1-positive cells increase in number and scatter in the entire region of the somatic mesoderm (Fig. 6B).

At late stage12 (about 8 hours), three lows of DFR1-positive cells are distinguishable in bodywall (Fig. 6C). These cells appear to be precursor cells for ventral (v), pleural (p) or dorsal (d) groups of muscles. When the germ-band has completely retracted, cell fusions to produce the muscle pattern finish (Bate, 1991). As expected, at early stage 13 (9.5 hours), vs 2-4 (muscles 26,27,25; ventral group; Crossley, 1978) and pet 1-3 (muscles 21-23; pleural group) were clearly DFR1-positive (Fig. 6D). Thus, the majority of DFR1-positive cells appear to eventually differentiate into muscle.

DFR1 expresssion also occurred in tissues other than somatic mesoderm. In the middle of stage 12, mid-ventral, DFR1-positive cells were detected (Fig. 6E). They may be ventral nerve cord cells, since their presence requires the achaete-scute gene products (Fig. 6F). DFR1 expression in CNS persists at least until stage 16 (13 hours; Fig. 6H). Cells surrounding the hindgut and foregut were noted to be DFR1-positive at stage 12-13 (Fig. 6G).

The expression of DFR2 RNA is first observed in primordia of anterior midgut and posterior midgut in the begining of germ-band extension (stages 6-7, about 3 hours; Fig. 7A). This expression becomes much more prominent at stage 8 (about 3.5 hours), when anterior and posterior midguts invaginate as shown in Fig. 7B. Other expression patterns of DFR2 RNA are essentially identical to those described by Glazer and Shilo in the case of DFGF-R (1991). At stages 8, DFR2-positive regions 1-2 cells wide become discernible (Fig. 7C). They appear to correspond to CNS midline cells and/or their precursors. At stage 10 (about 5 hours), cells contributing to tracheal pit formation begin to express DFR2 RNA (Fig. 7D). At stage 13, DFR2- positive signal can also be observed in cells correponding to the presumptive salivary duct and salivary tubes (data not shown). These observations suggest that DFR2 is involved in the formation of the endodermal layer (midgut), and development of mesectoderm and tubular organs derived from ectoderm.

Embryos of null mutations in twist or snail fail to form a normal ventral furrow; and cannot form mesodermal cell layer (Nüsslein-Volhard et al., 1984; St. Johnston and Nüsslein-Volhard, 1992; Leptin, 1991). Examination was made of the distribution of DFR1 RNA in twist and snail embryos. A quarter of progeny derived from either twist/+ or snail/+ parents was found negative to the DFR1 probe (255 and 179 embryos were examined, respectively; Fig. 8A,B,D). The morphology of DFR1-negative embryos was abnormal, showing twist or snail phenotypes (Fig. 8B,D). Thus, concomitant expression of twist and snail gene products may be required for mesoderm-specific DFR1 expression in early embryos. At later stages, a small number of DFR1-positive cells were detected in the ventral region of twist^– and snail^– embryos (Fig. 8C,E). From their positions and size, these cells appeared to be neuronal cells.

During in situ hybridization experiments, we noticed that the DFR1 gene is located within a short deletion of sr16 at the 90D-E region (De La Pompa et al., 1989; Bellen et al., 1992). This deletion was estimated to be 70-300 kb in length; it is known to include at least two genes couch potato (cpo) and stripe (sr). cpo is apparently unrelated to DFR1 (Bellen et al., 1992). We first examined the gastrulation of 3-5 hour embryos derived from sr16/+ parents, by following twist RNA expression in presumptive mesoder-mal cells (Thisse et al., 1988). No abnormal invagination was observed in more than 100 embryos at least until stage 8. Germ-band-extending embryos from the same parents were also hybridized with a mixture of DFR1 and DFR2 probes. DFR1 DFR2⁺ embryos should correspond to those homozygous for sr16. As shown in Fig. 9A, sr16/sr16 embryos were essentially normal in DFR2 expression in midline cells. In contrast, aberrant DFR2-expression was observed in twist embryos defective in gastrulation (Fig. 9B). Although we do not know possible morphological changes in 5-to 7-hour-old DFR1-defective embryos, it should thus be possible for normal gastrulation to occur without the DFR1 gene product.

To examine the muscle formation in sr16 embryos, we utilized P[Dmyd-lacZ], by which Dmyd expression can be detected as lacZ gene expression (Paterson et al., 1991). Dmyd, also called nautilus, is a Drosophila homologue of the vertebrate MyoD-family genes, specifically expressed in a subset of muscle precursors (Michelson et al., 1990; Paterson et al., 1991). The effect of sr16 on Dmyd expression was examined. At stages 12-13, the number of Dmyd-expressing cells was found extensively reduced, particularly, in the dorsal region (Fig. 9C-F). At later stages, reduction of muscle fiber content was evident under a microscope with Nomarski optics (data not shown). These results show DFR1 possibly to be required for the formation or extension of muscle precursor cells.

The Drosophila genome includes two homologues of mammalian FGF-R genes, DFR1 and DFR2. Their expression dynamically changes during embryogenesis. At earlier stages, DFR1 is expressed in mesodermal primordium, and DFR2 in the primordia of the anterior and posterior midguts. At later stages, DFR1 RNA is expressed in developing tissues derived from mesoderm and neuroectoderm, including muscle precursors and CNS. DFR2 RNA appears to be expressed in those tissues derived from ectoderm and mesectoderm, including the trachea, salivary tubes and CNS.

A comparison of the amino acid sequences of kinase domains indicated DFR1 to have about 80% and 60% homology with DFR2 and any member of vertebrate FGF-Rs, respectively. DFR1 and DFR2 genes may thus quite possibly be derivatives of a common ancestral arthropod gene, diverging from an ancient vertebrate FGF-R gene at the time of the major phylogenetic branch leading to arthropods and chordates.

Extracellular regions of Drosophila and vertebrate FGF-Rs are much more diversified than kinase regions. DFR1 and DFR2 have two and five Ig-domains, respectively. (Most members of the vertebrate FGF-R family contain three Ig-domains, while the remaining have two.) DFR1 and DFR2 may recognize different ligands. Our sequence analysis showed the extracellular region flanked by the acidic region and transmembrane domain to be considerably conserved in both sequence and organization. In particular, some N-glycosylation sites were found invariant in position among DFR1, DFR2 and flg protein. (see Fig. 3). This region may possibly be essential for common functions of the FGF-R family members.

FGFs secreted by non-mesodermal cells may induce mesoderm formation in vertebrates (Jessell and Melton, 1992; Whitman and Melton, 1992) but, so far, nothing has been reported as to the possible involvement of FGF or FGF-like factors in mesoderm formation in Drosophila. Only the nuclear concentration of the dorsal protein is considered important in determining mesodermal fate in Drosophila (Roth et al., 1989; Thisse et al., 1991). Thus, our finding of specific expression of DFR1 in mesodermal cells is of particular importance. Recent molecular-cloning experiments show that other mesodermal genes, which include twist, snail and MyoD, are also shared in common by Drosophila and vertebrates (St. Johnston and Nüsslein-Vol-hard, 1992; Michelson et al., 1990; Paterson et al., 1991) and, thus, it is reasonable to consider that Drosophila and vertebrates share in common some essential mechanisms underlying mesoderm formation and/or subsequent differentiation of mesodermal cells.

Our results suggest that DFR1 expression is not necessarily mandatory for ventral furrow formation. In its place, DFR1 would rather appear required for patterning of muscle precursor cells (somatic mesoderm), since DFR1 RNA is predominantly expressed in embryonic muscle precursor cells and its absence results in extensive reduction in the number of Dmyd/nau-expressing muscle precursor cells, particularly, in the dorsal region (see Fig. 9D,F). During the early stages of muscle formation, some muscle precursor cells migrate from the ventral to lateral or dorsal region so as to produce the muscle pattern (Dohrman et al., 1991). Although there is no direct evidence to date, muscle patterning is supposed to be specified by the underlying ectoderm (Bate, 1991). The above findings suggest the possible involvement of DFR1 (and maybe its ligand) in cell migration during muscle formation.

The stripe gene included in the sr16 deletion is required for the normal development of adult flight muscles and CNS (De La Pompa et al., 1989). Although, at present, the relation between stripe and DFR1 at the molecular level is unknown, similarity in expression/(mutant phenotype) and chromosomal locus suggest that they are genes intimately related, if not identical, to each other.

In Drosophila, various members of an immunoglobulin superfamily are expressed in developing nervous system; they include fasciclins, neuroglian, receptor-type phosphatases and a Drosphila homologue of mammalian trk (Grenningloh et al., 1990; Yang et al., 1991; Tian et al., 1991; Pulido et al., 1992). Some are homologous in amino acid sequence to vertebrate neural cell adhesion molecules and proteins with Ig-domains may possibly essential for neural development. As seen in Fig. 7C, DFR2 RNA is expressed in most midline cells, which probably play a key role in the formation of axon commissures. DFR1 RNA expression may occur along a pair of logitudinal connectives (see Fig. 6H). Since neurotrophins and FGFs are possibly required for neural cell differentiation in vertebrates (Jessell and Melton, 1992), the above findings imply important functions of DFR1 and DFR2 proteins in CNS formation in Drosophila.

We thank C. Nüsslein-Volhard, W. J. Gehring, A. Ferrus and Y. Hotta for supplying Drosophila strains. We also thank K. Igarashi for human FGF-R cDNA clones, and M. Asano and K. Shiokawa for discussion. This work was supported in part by grants from the Ministry of Education, Science and Culture of Japan.