Developmental expression of two murine fibroblast growth factor receptors, fig and bek

Fibroblast growth factors (FGF) and their receptors (FGF-R) have numerous important biological activities. They contribute to the control of normal and transformed cell growth and they are involved in tumor angiogenesis as well as in atherosclerosis (for reviews, see: Rifkin and Moscatelli, 1989; Folkman and Klags-brun, 1987; Klagsbrun and Edelman, 1989). They are expressed in the developing embryo, where they are active in mesoderm induction (reviewed by Smith, 1989; Whitman and Melton, 1989). Hence the FGF system is important both for understanding early vertebrate development and for the analysis of medically important processes.

A number of FGF-related and other polypeptide growth factors can induce mesoderm in Xenopus animal caps. FGF mainly affects the posterior and ventral mesoderm, where it interacts with homeobox containing genes, whereas activin and members of the TGFβ family display stronger effects on the development of the anterior axis and head (Rosa, 1989; Ruiz i Altaba and Melton, 1989; Smith et al. 1990; Thomsen et al. 1990). It has been shown that oocytes and early Xenopus embryos contain FGF peptides (Kimmelman et al. 1988; Slack and Isaacs, 1989). Hence FGF and FGF-R are thought to be intrinsic to amphibian development. Less is known however, about their contribution to comparable events in mammahan embryogenesis (Wilkinson et al. 1988; Joseph-Silverstein et al. 1989; Gonzalez et al. 1990; Curatola and Basifico, 1990; Wanaka et al. 1990; Moore et al. 1991) and the information regarding FGF receptors falls short of what is known about their ligands.

The FGF family contains seven polypeptides, which are distinguished from other polypeptide growth factors by their affinity to heparin-like extracellular matrix components (for review, Burgess and Maciag, 1989). High affinity cell surface receptors for the FGFs are membrane-spanning protein tyrosine kinases. Their cDNA sequence places them, as a separate subgroup (Lee et al. 1989; Dionne et al. 1990; Avivi et al. 1991; Raz et al. 1991), into the receptor tyrosine kinase gene family (RTK, Yarden and Ullrich, 1988). Recently numerous FGF-Rs have been cloned, including three from chicken (cekl, cek2 and cek3; Lee et al. 1989; Pasquale, 1990), three from the mouse (flg, Safran et al. 1990 and Reid et al. 1990; bek, Kombluth et al. 1988; Raz et al. 1991 and KGF-R, Miki et al. 1991) and seven from man (flg and bek, Dionne étal. 1990; Johnson étal. 1990; Reid et al. 1990; Mansukhani et al. 1990; flg-2, Avivi et al. 1991; k-samHattori et al. 1990; FGFR3 and FGFR4, Partanen et al. 1991; as well as TK14, Houssaint et al. 1990). The cek-1 chicken cDNA shows highest homology to the human and murine flg sequence, cek-3 of the chicken appears to be homologous to the mammalian bek and k-sam and KGF-R cDNA clones (Pasquale, 1990), whereas cek-2 may be the cognate of murine flg-2 (Avivi et al. 1991) and human FGFR3 cDNA-s (Partanen et al. 1991).

FGF receptors are expressed with considerable variation. Fig transcripts were described both with two and with three immunoglobulin-like domains, which however, had similar ligand-binding specificity. Putative secreted forms have been observed also (Johnson et al. 1990; Reid et al. 1990; Mansukhani et al. 1990) and a recent report by Hou et al. (1991) suggests that cultured liver cells may express as many as twelve FGF-R variants, different in specific structural domains, but homologous along most of their sequence. Similar local nucleotide sequence variation separates the otherwise very closely related bek, k-sam and KGF-R cDNAs (Hattori et al. 1990; Miki et al. 1991; Raz et al. 1991). Therefore, it is not clear whether certain FGF-R cDNA clones represent separate loci or allelic variants.

Several FGF-R bind more than one polypeptide growth factor with high affinity. The human bek and fig receptors bind both acidic (aFGF) and basic FGF (bFGF) with similar affinities (Dionne et al. 1990). Mansukhani et al. (1990) demonstrated that a fig variant with two immunoglobulin-like loops binds both bFGF and hst/K-fgf. The KGF (kératinocyte growth factor) receptor, which is closely related to bek (Kombluth et al. 1988; Miki et al. 1991; Raz et al. 1991), recognizes KGF and aFGF preferentially (Rubin et al. 1989). Moreover FGFR4 specifically binds aFGF rather than bFGF (Partanen et al. 1991). One interpretation for the multiplicity of the FGF -FGF-R system is based on the closely similar binding specificity of some FGF-R, like bek and fig (Dionne et al. 1990) and suggests that redundant sets of FGF-R may exist.

In this study we used in situ hybridization to compare the developmental expression of bek and fig. aimed at an initial assessment of the role of FGF-R in mammalian embryogenesis. It was also expected that the developmental localization of bek and fig would reveal whether they are indeed redundant, and if they are functionally distinct despite their similar ligand binding specificity.

C57BL/6J mice were kept under a 14 h light, 10 h dark (from 7.30p.m. to5.30a.m.) regime and the time of pregnancy was established by the presence of vaginal plugs the morning following mating. This time was regarded as day 0.5 post coitum (p.c.). Early embryos (6.0 to 8.0 days p.c.) were prepared for sectioning within the decidua. Later embryos were carefully dissected from their embryonic membranes. All embryos were prefixed in 4% paraformaldehyde at 4°C.

The probes used in this study derived from a mouse brain cDNA library and they were cloned into Bluescriptll KS+. The bek probe was a 281 bp long NlalV-Smal fragment starting 121 bp upstream from the assumed initiator methionine and included the hydrophobic leader sequence and part of the first immunoglobulin-like loop of bek (Raz et al. 1991). The hydrophobic leader sequence of bek and KGF-R (Miki et al. 1991) are identical (A. Yayon et al. unpublished data), therefore our bek probe, recognizes both the bek and the KGF-R transcripts. The/flg probe was a 220 bp long Styl-Nrul fragment including the leader sequence and entering 24 nucleotides into the first immunoglobulin-like domain (Safran et al. 1990). T3 and T7 polymerase-catalyzed transcripts were synthesized in the presence of ³⁵S-UTP for in situ hybridization. In each experiment both sense (control) and antisense transcripts were used for hybridization. No significant signals were obtained with sense (control) transcripts of thesè probes.

Embedding, sectioning (cryostat), post-fixation, hybridization and washing was as previously described (Orr-Urtreger et al. 1990). Post-hybridization washing was performed at high stringency in 2×SSC, 50% formamide and 0.1 % dithiothrei-tol at 65 °C, followed by RNAase digestion and by a further high-stringency wash. The slides were inspected and photographed in a Wild/Leitz Macroscope or in a Zeiss Photomicroscope III.

To test whether bek and fig are active in oogenesis, we analyzed their expression in adult ovaries. Both genes displayed low transcript levels in the ovarian stroma (not shown), whereas the ovarian follicles, the corpora lútea and various oocytes contained no detectable FGF-R transcripts. The apparent absence of bek and fig expression in the oocyte suggested that they may lack maternal regulation. This was in contrast to other RTK genes, like c-kit of the mouse, which is active both in male and female gametogenesis (Orr-Urtreger et al. 1990), or torso and faint little ball, which are maternal mutations of Drosophila (Sprenger et al. 1989; Schejter and Shilo, 1989).

With the appearance of the proamniotic cavity both bek and fig were expressed in the primitive ectoderm (Fig. 1A-D). The fig probe hybridized to most cells along the proamniotic cavity, suggesting that it is active both in the extraembryonic and in the embryonic ectoderm (Fig. 1A and B). Bek displayed higher transcript concentrations than fig. In the midsagittal plane bek transcripts were confined to the embryonic ectoderm through its whole extent and showed no anterior or posterior preference (Fig. 1C and D). In more parasagittal planes, bek transcripts accumulated in the extraembryonic ectoderm (not shown). Thus, it seems that bek and fig are expressed both in the ICM (inner cell mass)-derived and in the trophectoderm-derived ectoderm of the egg cylinder.

Mouse embryos at the early egg cylinder stage are made of two cell layers, the primitive ectoderm and the visceral endoderm. Around the sixth day of gestation ectodermal cells exfoliate through the primitive streak into the space between these two layers, and the mesoderm forms. Serial sections were used to find nascent mesoderm in 6.5 and 7.5 day-old embryos. At 7.5 days bek expression was easily detectable in the primitive ectoderm-neural plate and it culminated in the head fold. In these embryos the non-segmented, presomitic mesoderm was already clearly visible; it however displayed only weak hybridization signals, detectable both in the embryonic and extraembryonic mesoderm (Fig. IE and F).

In the 8 day-old mouse embryo the mesoderm differentiates into axial, paraxial and lateral mesoderm. The paraxial mesoderm then aggregates into somites and the anterior part of the neural plate develops into the cranial folds. Fig. 2 demonstrates the expression of both FGF-R genes at this stage. Ectodermal expression at this stage exhibited once again the strongest signals in the cranial folds. Bek was expressed also in the somites, in the surface ectoderm and in the heart primordium (Fig. 2A-D), whereas fig was expressed (at lower levels) in the presomitic mesoderm (Fig. 2E and F). Taken together, at the primitive streak stage the two FGF-R genes were first expressed in the extraembryonic and embryonic ectoderm, whereas later in development they appeared also in early derivatives of the mesoderm.

Transversal sections of 9.5 day-old embryos were hybridized with bek and flg probes. Bek was expressed in the lateral and splanchnic mesoderm. Its transcripts were apparent also in the cranial and dorsal neural tube and in the surface ectoderm of the embryo (Fig. 3B,E-G). Bek hybridization signals in the embryonic integument originated in the most superficial cell layer (the surface ectoderm) and not in the sectionslide interface (edge effect), as it can be judged by comparing these signals with those in consecutive sections hybridized with flg (Fig. 3C and H). Transcripts of both bek and flg were apparent along the anterior part of the body wall and a pair of sites with relatively high concentrations of transcripts could be seen in the dorsal roof of the embryonic coelom, which corresponded to the nephrogenic cords.

Bek transcripts were unevenly distributed (Fig. 3D-G) in the early (9.5 days p.c.) forelimb bud. Their concentration increased in posterior and proximal direction towards a position close to the body wall. Similar observations were also made one day later, when the first signs of mesenchymal condensations became visible (Fig. 4B). The expression of flg was more homogeneous than that of bek (Fig. 3H), but later, at day 10.5, this gene also was expressed in a nonhomogeneous manner along the limb bud (Fig. 4C). Next, at 11.5 days p.c., bek expression localized to mesenchymal aggregates corresponding to the future bones of the limb. This process was less evident in consecutive sections hybridized with the flg probe (Fig. 4D–F). Thus the expression of bek followed more closely the osteogenic mesenchymal condensation than flg. Fig. 4 demonstrates another characteristic difference between the localization of bek and flg, already observed in younger embryos. It shows bek transcripts in the surface ectoderm of the limb, as a continuous contour, which seems to be strongest in the interdigital web (at 14.5 days p.c., see Fig. 5E). In contrast flg transcripts are distributed diffusely in the mesenchyme.

To continue studying the role of FGF-R in limb development, 12.5 and 14.5 day-old embryos were investigated (Fig. 5). Ossification in the limb starts in the long bones (humerus, radius and ulna) and continues radially towards the metacarpal bones and finally reaches the phalanges and the carpal bones. Bek and flg expression followed this order. Our data demonstrate that bek transcripts occupy the chondrification centers at 12.5 days (Fig. 5B) and the body of the distal bones (phalanges and carpals) at 14.5 days p.c. Later they appear in the periphery of the developing radius, in which ossification (calcification) has already commenced (Fig. 5E; see also Rugh, 1968). This temporal order of expression suggested that the FGF system may have a genuine role in limb and bone development. The presence of FGF in the regenerating limb blastema of urodele amphibians also supports this notion (Boilly et al. 1991).

Sagittal, parasagittal and transversal sections of whole embryos (Figs 6, 7F-K and 8) demonstrate that the two FGF receptors are active in most developing organs. Sagittal sections show that both are transcribed in distinct areas of the nervous system. They are present in the cerebral ganglia, in the ependymal layer and in elements of the developing cortex and spinal cord. Bek is present also in the choroid plexus (Fig. 6B,E). In the spinal cord it is initially expressed in the alar plate (9.5-10.5 days, Figs 3, 4) and later (12.5 and 14.5 days p.c.) it concentrates, together with/Zg, to the ependymal layer (Fig. 8).

Both FGF-R are active in the development of the sense organs. Fig. 7B and C show the expression of bek and flg in the otic vesicle at 10.5 days p.c. At 14.5 days of gestation only bek expression could be detected in derivatives of the otic vesicle, when it appeared in the semicircular and cochlear ducts (Fig. 7J). Bek and flg apparently also contribute to eye development. This is suggested by their presence in the lens, by bek expression in the corneal epithelium and by the expression of flg in the cornea and sclera (not shown).

Bek expression in the digestive and excretory system could be observed at the earliest in the foregut pocket of 8 day-old embryos (Fig. 2A and B). Bek transcripts were detected in the epithelium of the pharynx (Fig. 6A,B,D and E; Fig. 7A,B,D and E), the oesophagus (Fig. 4D and E and Fig. 8A,B,D and E) and the cloaca (Fig. 8G and H). The highest concentration of bek transcripts was observed in the stomach epithelium (Fig. 8G and H). This expression pattern is specially noteworthy because amplification of the human analogue of the murine bek transcript, k-sam. is associated with dedifferentiated stomach cancer (Hattori et al. 1990). Bek is also expressed in the mesodermal wall of the intestine, which is derived from the splanchnic mesoderm (Fig. 6B,E and Fig. 8B,H). Diverse derivatives of the embryonic foregut, the liver, the pancreas, the epithelium of the bronchial tree and the alveoli of the lung also express bek. No flg transcripts were, however, apparent in these structures (Figs 6 – 8). In contrast the genital eminence expressed both FGF-R at high level. Bek transcripts were more prominent in the surface ectoderm and in the invaginating epithelium at the entrance of the urethra, which derives from the ectoderm, whereas fig was more prominent in the mesenchyme (Figs 6 and 8).

Figs 6, 7 and 8 demonstrate the two FGF receptors’ activity in osteogenesis in addition to what was also seen during limb development. The most notable observation was that both receptors were apparently expressed in all prospective bones; they were present in the basicranium, scapula, the ribs and in the sternum, as well as in the long and short bones of the extremities (Figs 5–8) and in the vertebrae (Figs 6, 7). Bek transcripts are present in the body of the vertebrae at day 12.5 and by day 14.5 of embryonic development, when calcification starts in the vertebral body, they appear in the periphery of the prevertebrae. As in limb development, also in the developing vertebral column bek is expressed earlier than fig (Fig. 4).

Both bek and fig are expressed in the craniofacial area. Fig transcripts are characteristically present in the diffuse mesenchyme of the trunk (Fig. 6F; Fig. 8C,F and I), face, tongue and mandíbula (Fig. 6C and F; Fig. 7H and K). These areas however, are mostly free from bek transcripts (Fig. 6B and E; Fig. 7G and J). At 10.5 and 11.5 days of gestation both FGF-R are expressed in the maxillary process and in the mandibular and hyoid arches (Fig. 7A – E). Bek expression is not uniform along the branchial arch, and as shown in Fig. 7B, its transcripts are distributed in a gradient increasing from the base of the arch towards its distal tip. Parasagittal sections taken at 12.5 and 14.5 days of development show that the fig message is more abundant in the facial mesenchyme, whereas bek transcripts are concentrated more to the frontal bone, the maxilla, the mandíbula, and certain areas of the fore and hindbrain (Fig. 7F-K). Bek was also detected in the ossicles of the middle ear, which derive from the first and second branchial arches (Fig. 7J).

A common feature is that bek is frequently expressed in the epithelium, whereas flg is expressed mainly in the mesenchyme, in the majority of the developing organs surveyed by us. Two examples, kidney and skin development, illustrate this issue further.

Both bek and flg are expressed already in the nephrogenic cords of 9.5 day-old embryos (Fig. 3). They are also present in the mesonephros (Fig. 4 and unpublished data) and their relative expression can be compared well in the developing permanent mammalian kidney, the metanephros (Fig. 9). The kidney develops through interactions between the epithelium of the metanephric diverticulum (ureteric bud) and the metanephric (intermediate) mesoderm. Fig. 9A – C demonstrates the expression of bek and flg in the embryonic kidney at 12.5 days of development. Bek is mostly expressed in the epithelium of the ureteric bud and its branches (Fig. 9B), whereas abundant flg transcription can be seen in the surrounding mesenchymal condensations and in their derivatives, the metanephric vesicles (Fig. 9E). In the 14.5 day-old embryonic kidney bek is present in the epithelium of the collecting tubules (Fig. 9F), which is derived from the metanephric diverticulum, while flg transcripts occupy mesenchymal elements in the cortical areas of the embryonic kidney (Fig. 9F).

Skin and its appendages (hair, teeth and certain membrane bones) develop from interactions between ectodermal and mesodermal elements. The integument of the early embryo is covered by a single ectodermal cell layer, the surface ectoderm. This cell layer displayed one of the most characteristic developmental expression patterns of the bek gene. It can be seen, in most of the figures presented in this report, that the hybridization signals of bek trace the contour of the embryo (Figs 3-8). Higher magnification revealed that this signal originates in the surface ectoderm (Fig. 10B). The latter develops into the actively dividing layer of the adult epidermis, the stratum germinativum, and gives rise to kératinocytes, which produce the keratinized protecting layer of the adult skin. In 14.5 day-old embryos bek is present in the epidermis (Fig. 10A,B) and flg transcripts accumulate immediately beneath it, in the mesenchyme of the prospective dermis (Fig. 10C,E). This relationship between the two receptors’ expression pattern demonstrates that bek and flg occupy adjacent positions at the epithelial-mesenchymal interface of the developing skin.

Hair follicles develop from invaginations of the epidermis. Mesodermal cells then penetrate into the base of these invaginations and form the dermal papilla. Longitudinal and transversal sections through the hair root, shown in Fig. 10, demonstrate that mesodermal derivatives, the dermal papilla and sheath, express the flg receptor, whereas the ectodermal components, the epithelial root sheath and bulb, contained bek transcripts. In hair and skin development, as in the development of most organs, bek occupies epithelial positions, whereas flg occupies mesenchymal positions. Additional skin derivatives, teeth, and mammary glands, like hair, also develop by epithelial-mesenchymal interactions and bek and flg here too occupy epithelial or mesenchymal components respectively (Fig. 6E and F and unpublished).

These patterns of expression demonstrate that bek and flg occupy distinct locations within different organs. Unique expression patterns were seen in the diffuse mesenchyme of the face, in the extremities and in the trunk. This domain of flg expression, which precedes myogenesis, may be connected with the expression of flg in myoblasts (Moore et al. 1991). No bek transcripts were found in this domain, which emphasizes that the two receptors and their possible transcriptional variants, are functionally distinct throughout development.

The overall nucleotide sequence homology of the bek and flg cDNA-s is 71 %. Their kinase domains are 90 % homologous and the second and third immunoglobulin-like domains are close to 80 % similar. In contrast, their first immunoglobulin-like loops and the respective interkinase segments are more divergent (Raz et al. 1991). In agreement with this structural homology, the two receptors display almost identical binding specificity. Because of this similarity it was argued that bek and flg may be functionally redundant (Dionne et al. 1990). Our localization studies, however, reveal that bek and flg have clearly distinguishable spatial specificity during development. Not only the characteristic epithelial-ectodermal preference of bek distinguishes it from flg, but also their differential expression during osteogenesis, kidney and hair follicle development, all point to the functional distinction of these two receptors.

The bek probe, which was used in the present study, includes the hydrophobic leader sequence and part of the first immunoglobulin-like loop of the bek cDNA (Raz et al. 1991). This area is identical in bek and KGF-R. Therefore, our probe does not discriminate between the probably allelic bek and KGF-R transcripts (Miki et al. 1991 ; Raz et al. 1991 ; Dionne et al. 1990; A. Yayon et al. unpublished). The present results however, are consistent with the suggestion that bek and flg have independent patterns of expression. The strongest argument pointing to the independent expression of bek and flg is the absence of bek transcripts in the undifferentiated mesenchyme, which appears to be a specific domain for flg expression.

It is unusual that two receptors with similar structures and binding specificities should be conserved through evolution in man, mouse and chicken (Dionne et al. 1990; Pasquale, 1990; Raz et al. 1991). Our data suggest that these genes may acquire functional specificity through their developmental localization. An additional influence on the specificity of this system could be the localization of the heparin-binding sites for FGF in the extracellular matrix (Vainio et al. 1989; Trautman et al. 1991; Yayon et al. 1991). Hence, the evolution of the FGF system, because its diversity, which seems to derive from the spatial restriction of its receptors and ligands, may not have required the complex binding site polymorphism of other receptor-ligand systems. A system of specifically localized receptors, with redundant binding specificity, however, could be highly sensitive to quantitative and spatial changes. The transforming activity of hst /K-fgf, isolated from Kaposi sarcoma and from stomach cancer (Delfi Bovi et al. 1987; Taira et al. 1987), as well as the association of the k-sam protooncogene with stomach cancer (Hattori et al. 1990), may be related to such mechanisms.

Our data suggest that bek and flg are probably devoid of maternal expression. They were however, readily detectable in the egg cylinder (6 – 6.5 days p.c.). Transcripts of the two FGF-R were concentrated in the ectoderm along the proamniotic cavity. We assume that FGF-R expression in the primitive ectoderm may be connected to the role of FGF in mesoderm induction. In Xenopus FGF affects posterior and lateral rather than anterior areas of the rostrocaudal axis (for reviews, Ruiz i Altaba and Melton, 1989; Smith, 1989), whereas bek and flg are expressed along the whole anterioposterior extent of the primitive ectoderm. Whether in mammals, as in amphibians, FGF affects the posterior part of the axis, remains to be determined.

Little is known about the expression of the various members of the FGF family in the early mouse embryo. Int-2, an FGF-like protooncogene, somewhat similarly to our FGF receptors, is expressed in the early migrating mesoderm and neuroectoderm (Wilkinson et al. 1988). Another protooncogene of the FGF family, hst/K-fgf, is selectively transcribed in embryonic carcinoma cells (Brookes et al. 1989; Curatola and Basifico, 1990). Quite recently Haub and Goldfarb (1991) as well as Hebert et al. (1991) reported that a member of the FGF family, FGF-5, is first expressed, after implantation, in the primitive ectoderm of the egg cylinder and later in the lateral mesoderm including the limb bud. Although FGF-5 is only one of seven members of the FGF family and its specific receptor is not known, there appears to be a basic similarity between its expression pattern and that of the two FGF receptors described here.

Other polypeptide growth factors, like members of the TGF β gene family, which are also potent mesoderm inducers (Whitman and Melton, 1989), are expressed in the oocyte (Rappolee et al. 1988; Lyons et al. 1989). In the early embryo TGF/51 was already present in the heart anlage at 8 days of gestation (Akhurst et al. 1990), whereas BMP-4 transcripts first became distinctive in the posterior mesoderm of the 8.5 day neurula (Jones et al. 1991).

Bek and flg expression in late presomitic-early somitic stage embryos coincides with the expression of the TGF/J family. This temporal and spatial coincidence supports the notion that a multiplicity of polypeptide growth factors and receptors, are localized together at the time of gastrulation both in amphibia and in mammals. Further relevant connections can be made between the pattern-forming activity of homeoboxes and polypeptide growth factors in the amphibian embryo (Ruiz i Altaba and Melton, 1989) and between the localization of retinoic acid receptors and retinoid-binding proteins (Ruberte et al. 1991) with our two FGF receptors in late presomitic and early somitic mouse embryos.

Bek and flg are expressed in ectoderm and mesoderm derivatives during the development of many organs. A general characteristic of their developmental regulation appeared to be the constancy of their localization. Bek and flg transcripts usually appeared very early within the first anlage of an organ and accompanied it, in progressively differentiated form, throughout embryogenesis, as in limb, kidney, heart, sense organ, or skin development. Similar constancy of developmental expression was observed also with two other receptor tyrosine kinases, c-kit (Orr-Urtreger et al. 1990) and the a receptor of PDGF (Orr-Urtreger et al. unpublished data) and may be characteristic for the PDGF-R and FGF-R subfamilies of receptor tyrosine kinases.

There is little information regarding the developmental expression of the ligands of bek and flg. Basic FGF, a ligand shared by both, appears at sites similar to the localization of the two receptors during late embryogenesis, as shown by Gonzalez et al. (1990) for rat embryos and by Joseph-Silverstein et al. (1989) for chicken embryos. They report that these polypeptides are present in skeletal and heart muscle precursors, in bone and cartilage differentiation and in the developing hair root. A recent report describes the expression of FGF-5 in the splanchic mesoderm and in muscle precursors (Haub and Goldfarb, 1991). Much insight can be expected into the FGF system’s role in organogenesis from comprehensive localization studies, which will compare the expression of different FGF-R with that of their various putative ligands.

During gastrulation and also during organogenesis bek and flg appeared in domains shared with other polypeptide growth factors and transcriptional regulators. In the heart TGF/31 and 2 displayed expression patterns related to those of bek (Potts and Runyan, 1989; Choy et al. 1990; Millan et al. 1991). Moreover Vgr-1 expression was predominantly detected in squamous epithelia of the developing skin, oesophagus and stomach, areas of expression shared with the bek receptor. Similarly to flg TGF-/32 and BMP-2 are also expressed in the dermis (Pelton et al. 1989; Lyons et al. 1989; Lyons et al. 1990; Jones et al. 1991). Like the two FGF-R, members of the TGF/3 family are also expressed during cartilage and bone differentiation (Pelton et al. 1990; Schmid et al. 1991; Millan et al. 1991). A notable example suggesting complex gene interactions involving FGF receptors may be found in early limb development. Concentration gradients along the limb bud, which are formed both by bek and by TGF/7 (BMP-2A, Lyons et al. 1990) transcripts, invite comparison with the gradients of homeobox gene transcripts in the limb bud (Oliver et al. 1989; Dolle et al. 1989a). It is possible that cellular interactions mediated by the TGF β and FGF systems in limb development are connected with the pattern-forming activity of homeoboxes, retinoic acid derivatives and their receptors (Dolle et al. 1989b ; reviewed by Eichele, 1989; Brockes, 1990).

Colocalisation of bek and flg with other putative transcriptional regulators indicate the FGF system’s possible involvement in additional interactions. Bek expression in the alar plate of the spinal cord is coincident with that of a transcriptional regulator, Pax-3 (Goulding et al. 1991), whereas the expression of flg in the intervertebral disc is coincident with the expression of Pax-1 in the same structure. A mutant allele of Pax-1, undulate (un), displays abnormal vertebral column development (Balling et al. 1988). To test the interaction between Pax-1 and the two FGF-R we investigated their expression in homozygous un embryos, but no change could be detected. This suggests that bek and flg may occupy a position upstream to Pax-1 in a developmental hierarchy.

Analysis of spontaneous mutants at the relevant loci, gene targeting or other gene transfer experiments are elements of a comprehensive approach to explore the place of the FGF-R in the molecular interactions of early mammalian development. Their role in later development, especially in the epithelial-mesenchymal interactions of organogenesis, could be amenable to a molecular approach using various organ culture systems.

This research was supported by the Raphael and Regina Levy Fund, by the Pasteur-Weizmann Joint Research Programme and by the Centers for the Biology of Aging, for Molecular Genetics and for Neurobiology of the Weizmann Institute.