Ligand-independent activation of fibroblast growth factor receptors by point mutations in the extracellular, transmembrane, and kinase domains.

The fibroblast growth factor receptors (FGFRs) are a family of receptor protein tyrosine kinases that have been shown to mediate a variety of cellular processes including angiogenesis, wound healing, tumorigenesis, and embryonic development. Distinct FGFR mutations in individuals with autosomal dominant disorders of bone growth and development provide a unique opportunity to determine the function of FGFRs during embryonic development. To determine the consequences of these mutations on receptor function, we have made mutations in Xenopus FGFR1 (XFGFR1) and FGFR2 (XFGFR2) that correspond to several of the mutations identified in these dysmorphic syndromes. Analysis of mutant receptor proteins expressed in Xenopus oocytes indicates that all but one have elevated tyrosine kinase activity relative to their wild-type counterparts. Those mutations that give an unpaired cysteine residue in the extracellular domain result in intermolecular disulfide bond formation and covalent receptor dimerization. Microinjection of Xenopus embryos with RNA encoding mutant receptors with elevated tyrosine kinase activity results in ligand-independent induction of mesoderm in animal pole explants. Wild-type XFGFR1 and XFGFR2 do not induce mesoderm when injected at similar doses. Co-injection of RNA encoding a dominant negative FGF receptor, lacking the tyrosine kinase domain, together with RNA encoding various activated FGFRs inhibits mesoderm induction by a receptor activated by a transmembrane domain mutation or extracellular mutations that introduce an unpaired cysteine residue into the extracellular domain but does not inhibit mesoderm induction by receptors bearing a tyrosine kinase domain mutation. These results indicate that different point mutations may activate FGFRs by distinct mechanisms and that ligand-independent FGFR activation may be a feature in common to many skeletal disorders.

The fibroblast growth factor receptors (FGFRs) are a family of receptor protein tyrosine kinases that have been shown to mediate a variety of cellular processes including angiogenesis, wound healing, tumorigenesis, and embryonic development. Distinct FGFR mutations in individuals with autosomal dominant disorders of bone growth and development provide a unique opportunity to determine the function of FGFRs during embryonic development. To determine the consequences of these mutations on receptor function, we have made mutations in Xenopus FGFR1 (XFGFR1) and FGFR2 (XFGFR2) that correspond to several of the mutations identified in these dysmorphic syndromes. Analysis of mutant receptor proteins expressed in Xenopus oocytes indicates that all but one have elevated tyrosine kinase activity relative to their wild-type counterparts. Those mutations that give an unpaired cysteine residue in the extracellular domain result in intermolecular disulfide bond formation and covalent receptor dimerization. Microinjection of Xenopus embryos with RNA encoding mutant receptors with elevated tyrosine kinase activity results in ligand-independent induction of mesoderm in animal pole explants. Wild-type XFGFR1 and XFGFR2 do not induce mesoderm when injected at similar doses. Co-injection of RNA encoding a dominant negative FGF receptor, lacking the tyrosine kinase domain, together with RNA encoding various activated FGFRs inhibits mesoderm induction by a receptor activated by a transmembrane domain mutation or extracellular mutations that introduce an unpaired cysteine residue into the extracellular domain but does not inhibit mesoderm induction by receptors bearing a tyrosine kinase domain mutation. These results indicate that different point mutations may activate FGFRs by distinct mechanisms and that ligand-independent FGFR activation may be a feature in common to many skeletal disorders.
Receptor tyrosine kinases (RTKs) 1 have been shown to play a key role in developmental processes such as proliferation, migration, and differentiation. Many RTKs have been shown to be expressed at high levels during embryonic development (1). Evidence also indicates that when normal RTK function is disrupted during embryogenesis, major phenotypic changes and embryonic lethality can occur (2)(3)(4)(5)(6). RTK signal transduction pathways are highly conserved in both vertebrates and invertebrates (1,7). RTK signaling is normally initiated by ligand binding followed by receptor dimerization that results in auto/trans-phosphorylation of specific tyrosine residues (7). Mutations in RTKs have been shown to result in deregulation of tyrosine kinase activity, resulting in ligand-independent activation. Many of the activating mutations identified in RTKs have been shown to be oncogenic (7).
The fibroblast growth factors (FGFs) constitute a family of at least nine structurally related but genetically distinct members, all of which have been implicated in a variety of biological processes including mitogenesis, angiogenesis, migration, differentiation, and mesoderm induction (8). A family of at least four high affinity transmembrane tyrosine kinases comprise the FGF receptor (FGFR) family (8 -10). These receptors share several structural features in common, including three extracellular immunoglobulin-like (Ig) domains, a hydrophobic transmembrane domain, and an intracellullar split tyrosine kinase domain with a 14-amino acid insertion (8,9). Several isoforms of each FGFR have been identified and are the result of alternative splicing of their mRNAs (11,12). Some of the receptor variants that result from this alternative splicing give rise to isoforms with different ligand binding specificities and affinities (13)(14)(15).
In several vertebrate species, the spatiotemporal expression patterns of FGFs and FGFRs indicates that these polypeptides may have tissue-specific as well as developmental-specific functions (16 -19). Targeted disruption of fgfr1 and fgf4 have demonstrated that FGF signaling pathways are crucial to vertebrate embryogenesis. Mice homozygous for the fgfr1 null mutation die shortly after gastrulation with defects in mesodermal patterning (4,6). Homozygous mutant fgf4 mice die at embryonic day 6.5 presumably due to failure of proliferation of the inner cell mass (20). More recently, targeted disruption of fgfr3 demonstrates that this receptor is a negative regulator of endochondral ossification (21). These and other FGF mutations (9) indicate that FGFs and FGFRs play multiple roles in development.
Other advances in our understanding of the developmental roles of FGFRs have resulted from the recent identification of mutations in human FGFR genes associated with dysmorphic syndromes of bone growth and development (reviewed in Refs. 9 and 22). Mutations in FGFR3 have been shown to be responsible for achondroplasia (23,24), thanatophoric dysplasia (25), and hypochondroplasia (26), which are all forms of dwarfism with varying degrees of severity. Mutations in the extracellular domain of FGFR1 have been associated with Pfeiffer syndrome (27), an autosomal dominant disorder characterized by craniosynostosis, an abnormality in which the cranial sutures fuse prematurely. Individuals with Pfeiffer syndrome also display hand and foot abnormalities (27). Mutations in FGFR2 have been identified in Pfeiffer syndrome (28) and in three other craniosynostotic syndromes referred to as Crouzon (29), Apert (30), and Jackson-Weiss syndromes (31). Like Pfeiffer syndrome, Apert and Jackson-Weiss syndromes have associated hand and foot abnormalities, whereas individuals with Crouzon syndrome have normal hands and feet (22,29). The identification of mutations in FGFRs associated with these syndromes provides genetic evidence that FGF signaling may be involved in bone and limb growth and morphogenesis.
To determine the functional consequences of the mutations identified in these dysmorphic syndromes, we have introduced analogous missense mutations into Xenopus FGFR1(XFGFR1) and FGFR2 (XFGFR2). Expression of all but one of these mutant receptors in Xenopus oocytes results in enhanced tyrosine kinase activity relative to wild-type receptors. In addition, expression of the mutant receptors with elevated tyrosine kinase activity in Xenopus animal caps results in ligand-independent induction of mesoderm as determined by both morphological and molecular criteria. These results indicate that the majority of mutations in FGFRs associated with human disorders of bone growth and development are likely to result in constitutive activation of these mutant receptors; thus, aberrant FGFR signaling is likely to contribute to the etiology of these syndromes.

EXPERIMENTAL PROCEDURES
Materials-A murine monoclonal antibody to phosphotyrosine, PY20, was obtained from Transduction Laboratories. Affinity-purified rabbit antibodies to FGFR were prepared as described previously (32). For some experiments, an anti-XFGFR1 monoclonal antibody was used (33). C4-Raf cDNA and N17-Ras cDNA were obtained from U. Rapp (National Cancer Institute) and T. Sargent (National Institutes of Health), respectively. XFD, a dominant negative FGFR1, was a gift from E. Amaya (University of California). Recombinant FGF-1 was a gift from W. Burgess (Holland Laboratory) and recombinant FGF-2 was obtained from Bachem. Both FGF-1 and FGF-2 were iodinated according to published procedures (34).
Oocyte Injections-Stage VI oocytes were obtained from mature fe-male Xenopus laevis (Nasco). Oocytes were defolliculated by mild collagenase treatment and maintained in 1 ϫ MBS (39) containing 1 mg/ml bovine serum albumin (BSA) and 50 g/ml gentamicin at 18°C. Oocytes were injected with the indicated amounts of RNA in a volume of 10 -50 nl and incubated for 2 days before immunoblot analysis, in vitro immune complex kinase assays or cross-linking. Immunoblot Analysis-Injected oocytes were extracted in cold lysis buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1.5 mM MgCl, 1.0 mM EGTA) containing 0.1 mM NaVO 4 . Lysates were extracted with 1 volume of freon to remove yolk proteins and centrifuged at 10,000 ϫ g for 10 min at 4°C. Equal amounts of protein were mixed with an equal volume of 2 ϫ SDS sample buffer, with or without ␤-mercaptoethanol, and proteins were separated by electrophoresis on SDS-8.5% or 6% polyacrylamide gels. Proteins were transferred electrophoretically to nitrocellulose filters (Schleicher and Schuell). Filters were blocked in 50 mM Tris-HCl, 150 mM NaCl, pH 7.4, and 0.05% Tween 20 (TBST) containing 5% BSA at 37°C for 1 h. Bound primary antibodies were visualized with horseradish peroxidase-conjugated secondary antibodies and enhanced chemiluminescence (ECL) procedures (DuPont NEN).
In Vitro Kinase Assays-Oocyte lysates were prepared as described above, and equal amounts of protein were subjected to immunoprecipitation with anti-FGFR antibodies (32). Immune complexes were immobilized on protein G-Sepharose beads (Pharmacia Biotech), washed extensively with lysis buffer, washed once in kinase buffer (50 mM Tris-HCl, pH 7.4, 10 mM MnCl 2 ) and finally resuspended in 50 l of kinase buffer containing 20 Ci of [ 32 P]ATP (3000 Ci/mmol) (DuPont NEN). Reactions were allowed to proceed at room temperature for 20 min. Reactions were terminated by washing once in lysis buffer followed by solubilization with 2 ϫ SDS sample buffer. Samples were separated on SDS-8.5% polyacrylamide gels, fixed, dried, and subjected to autoradiography as described previously (32).
FGF Receptor Cross-linking-Oocytes were injected with 5 ng of wild-type or mutant receptor RNA and incubated as described above for 2 days prior to cross-linking analysis. Oocytes expressing either wildtype or mutant FGFRs were incubated in 0.5 ml of 1 ϫ MBS containing 1 mg/ml BSA and 10 units/ml heparin (Upjohn) (binding buffer) in the presence or absence of 20 ng/ml 125 I-FGF-1 (6 -10 ϫ 10 4 cpm/ng) or 125 I-FGF-2 (8 -12 ϫ 10 4 cpm/ng) for 2 h at 4°C, followed by extensive washing in binding buffer. Receptor-ligand complexes were cross-linked by adding disuccinimidyl suberate (Pierce) to oocytes at a final concentration of 0.3 mM for 20 min at 4°C. Oocyte lysates were prepared as described above, and ligand-receptor complexes were analyzed by electrophoresis on SDS-7.5% polyacrylamide gels. Cross-linked complexes were visualized by autoradiography.
Embryo Injections-Eggs were collected from mature Xenopus laevis females and fertilized in vitro as described (40). Eggs were dejellied 60 min after fertilization with 2% cysteine, pH 7.8 -8.1, and washed extensively in 0.1 ϫ MBS. At the two cell stage, embryos were transferred to 1 ϫ MMR (5 mM HEPES pH 7.8, 100 mM NaCl, 2 mM KCl, 1 mM MgSO 4 , 2 mM CaCl, and 0.1 mM EDTA) containing 5% Ficoll. Each blastomere at the two-cell stage was injected with 5-10 nl containing the indicated amount of RNA. Injected embryos were incubated in 1 ϫ MMR containing 5% Ficoll until stage 7-8 (41) at which time they were transferred to 0.5 ϫ MMR for animal cap dissection.
Animal Cap Assays and RNA Gel Blot Analysis-Embryos were staged according to Nieuwkoop and Faber (41). Animal caps were dissected from embryos at stages 8 -9 and incubated in 0.5 ϫ MMR containing 1 mg/ml BSA and 50 g/ml gentamicin in the presence or absence of recombinant FGF-1 (50 ng/ml) at 18°C. Induction was scored by morphological evidence of convergent extension of animal cap elongation at neurula stages (37,42). Animal caps were harvested when control embryos reached stage 22-30 for analysis of muscle ␣-actin mRNA expression. Total RNA was isolated and analyzed by RNA gel blot analysis as described previously (43).

Generation and Expression of FGFR Mutants-Members
of the FGFR family possess two or three immunoglobulin-like domains in their extracellular domains, a transmembrane domain, and an intracellular tyrosine kinase domain split by a short kinase insert sequence (8, 10) (Fig. 1). Mutations in each of these three major domains have been identified in several human disorders of bone growth and development (9). We have previously described a C332Y mutation in the extracellular domain of XFGFR2 that is analogous to the C342Y mutation most commonly seen in Crouzon syndrome (37). This mutation results in constitutive activation of the receptor. In order to determine the biochemical consequences of other FGFR mutations associated with dysmorphic syndromes of bone growth and development, we have made mutations in the extracellular, transmembrane, and tyrosine kinase domains of XFGFR1 and in the extracellular domain of XFGFR2 (Fig. 1).
Achondroplasia has been shown to associated with a G380R mutation in the transmembrane domain of human FGFR3 (23,24). To determine the functional consequences of this mutation in a FGFR family member, we have made a C289R mutation in the transmembrane domain of the two Ig loop form of XFGFR1 that is identical in position to the FGFR3 G380R mutation as defined by the first basic residue that demarcates the end of transmembrane domains (23, 24) (Fig. 1). To determine the specificity of this mutation we have also created a C289K mutation as well as a kinase-deficient C289R mutant (XF-GFR1-C289R/K420A) (Fig. 1).
Thanatophoric dysplasia type II, another form of dwarfism is characterized by mutations in the extracellular and tyrosine kinase domains of FGFR3 (25). One of the most frequently identified mutations in this syndrome is a K650E mutation in the tyrosine kinase domain (25). The region of the tyrosine kinase domain in which this mutation occurs is highly conserved among all members of the FGFR family (8). To determine the consequences of this mutation, we have made a K562E mutation in the two Ig loop form XFGFR1. Many of the mutations associated with these skeletal disorders result in the loss of a cysteine residue that disrupts disulfide bond formation between the two highly conserved cysteines that form the IgIII domain (9,22). We have previously demonstrated that a C332Y mutation in XFGFR2 which is homologous to a mutation in Crouzon syndrome results in ligand-independent activation of the receptor (37). To further characterize the nature of these mutations, we have made a C268F mutation in XFGFR2 analogous to another Crouzon syndrome mutation in human FGFR2 (C278F) (22). We have also made the double mutant XFGFR2-C268F/C332Y to determine the consequences of perturbing the IgIII domain without creating an unpaired cysteine residue. In addition, we have made a mutation in XFGFR1 (C249Y) which also results in an unpaired cysteine residue, that is also likely to affect ligand binding and may also result in intermolecular disulfide bond formation. This mutation was made to determine whether analogous mutations in two FGFR family members would have similar functional consequences.
Pfeiffer syndrome has been characterized by several mutations in FGFR2, most of which have been proposed to result in the disruption of normal folding and disulfide bond formation in the IgIII domain (28). In addition, a mutation in FGFR1 has also been associated with Pfeiffer syndrome (27). The mutation in FGFR1 (P252R) occurs in the linker region between IgII and IgIII where a homologous mutation in FGFR2 (P253R) has been found in Apert syndrome (30). We have made the analogous mutation in XFGFR1 (P160R) (Fig. 1) to determine the effect of this mutation on receptor function.
To assess the biochemical consequences of these mutations, in vitro transcribed RNAs encoding wild-type or mutant XF-GFR1 or XFGFR2 were injected into Xenopus oocytes. Lysates from uninjected control oocytes or RNA injected oocytes were analyzed by SDS-polyacrylamide gel electrophoresis and immunoblotting with a monoclonal antibody to phosphotyrosine. Oocytes expressing either wild-type XFGFR1 (XFGFR1-WT) or wild-type XFGFR2 (XFGFR2-WT) exhibited tyrosine phosphorylation of several polypeptides including autophosphorylation of the receptors themselves (Fig. 2). Activation of the wild-type receptor tyrosine kinase in the absence of ligand is most likely due to the high density of receptor expression on the surface of the oocyte as described previously (37). Tyrosine phosphorylation of several proteins was significantly increased in lysates from oocytes expressing XFGFR1 mutants C289R, K562E, and C249Y ( Fig. 2A). Noteworthy are the tyrosine-phosphorylated polypeptides of 240, 160, 120, and 90 kDa that are not evident in XFGFR1-WT lysates. In contrast, XFGFR1-C289K and P160R did not exhibit increases in tyrosine phosphorylation above that seen in XFGFR1-WT lysates ( Fig. 2A). Lysates prepared from oocytes injected with XFGFR1-C289R/K420A RNA did not exhibit any tyrosine phosphorylation, and were comparable to lysates prepared from uninjected oocytes ( Fig.  2A). Lysates from the XFGFR2-C268F also exhibited increased tyrosine phosphorylation when compared to XFGFR2-WT lysates, and is similar to the increases seen with XFGFR2-C332Y as previously reported (Fig. 2C) (37). Particularly prominent are phosphotyrosine-containing polypeptides of 85-95 kDa seen in XFGFR2-C268F lysates but not in XFGFR2-WT ly- sates. The XFGFR2-C268F/C332Y mutant exhibited increases in tyrosine-phosphorylated polypeptides similar to that of XFGFR2-C268F. Antiphosphotyrosine immunoblots were stripped and reprobed with anti-FGFR antibodies to demonstrate equivalence of receptor protein expression in each lysate (Fig. 2, B and D).
To confirm that increases in tyrosine phosphorylation seen with XFGFR mutants are the result of increased receptor tyrosine kinase activity, we performed immune complex kinase assays (32). Equivalent amounts of receptor protein (as shown in Fig. 2, C and D) were immunoprecipitated from oocyte lysates with anti-FGFR antibodies. Both XFGFR1-WT and XF-GFR2-WT exhibit autophosphorylating activity in this assay (Fig. 3). In vitro tyrosine kinase activity was significantly increased for XFGFR1-C289R, K562E, and C249Y consistent with the antiphosphotyrosine immunoblot results. The XF-GFR1-C289K and XFGFR1-P160R possess in vitro tyrosine kinase activity similar to the wild-type receptor. XFGFR1-C289R/K420A was devoid of tyrosine kinase activity in this assay. This is consistent with antiphosphotyrosine immunoblotting results. XFGFR2-C268F has elevated in vitro tyrosine kinase activity relative to the wild-type receptor and similar to that previously reported for the XFGFR2-C332Y mutant (37) (Fig. 3). Similarly, the double cysteine mutant XFGFR2-C278F/C332Y has increased in vitro tyrosine kinase activity relative to the wild-type receptor which is consistent with the antiphosphotyrosine immunoblotting data.
Ligand Binding by Mutant FGFRs-To determine the qualitative effects of each mutation on FGF-1 binding, wild-type and mutant receptors were expressed in Xenopus oocytes and subjected to ligand binding and cross-linking analysis (Fig. 4). XFGFR1-WT and XFGFR1-K562E bound 125 I-FGF-1 to a similar extent when normalized for receptor protein expression. In contrast, XFGFR1 mutants C289R, C289R/K420A, and C289K bound 125 I-FGF-1 somewhat less efficiently than XFGFR1-WT. XFGFR1-P160R consistently bound more 125 I-FGF-1 than XF-GFR1-WT as determined by cross-linking analysis. Mutations in XFGFR1 and XFGFR2 which result in a loss of cysteine residues from the IgIII domain (XFGFR1-C249Y, XFGFR2-C268F, and XFGFR2-C268F/C332Y) abrogate ligand binding by these receptors and this is consistent with our previous report on the C332Y mutation (37) (Fig. 4). Similar results were obtained when 125 I-FGF-2 was used as the ligand (data not shown).
Covalent Receptor Dimer Formation-Several of the mutations identified in FGFRs that are associated with skeletal disorders result in the loss of one of the conserved cysteine residues that form a disulfide bond that gives the IgIII loop of the extracellular domain structural integrity (9). These mutations result in the creation of an unpaired cysteine residue that may form an intermolecular disulfide bond. We have created such mutants in XFGFR1-C249Y and XFGFR2-C268F. We have also prepared a mutant in which both Cys residues of IgIII have been mutated, XFGFR2-C268F/C332Y. Lysates from oocytes injected with XFGFR1-C249Y, XFGFR2-C268F, and XFGFR2-C268F/C332Y were analyzed by SDS-polyacrylamide gel electrophoresis under reducing and non-reducing conditions followed by immunoblotting with an anti-FGFR antibody. Under reducing conditions, XFGFR1-WT and XFGFR1-C249Y both migrated as a monomeric form of approximately 95 kDa (Fig. 5). Under non-reducing conditions, XFGFR1-C249Y gave an additional species of approximately 200 kDa, consistent with the size of a disulfide-linked homodimer. XFGFR2-WT, XFGFR2-C268F, and XFGFR2-C268F/C332Y were analyzed in a similar manner. Under reducing conditions XFGFR2-WT, XFGFR2-C268F, and XFGFR2-C268F/C332Y migrated as monmeric form of approximately 110 -125 kDa. Under nonreducing conditions XFGFR2-C268F gave an additional species of ϳ260 -280 kDa consistent with the formation of a disulfide linked dimer, whereas XFGFR2-WT and XFGFR2-C268F/ C332Y yielded only monomeric forms of the receptor. Antiphosphotyrosine immunoblotting and in vitro kinase experiments confirmed the anti-FGFR immunoblotting experiments demonstrating that XFGFR1-C249Y and XFGFR2-C268F form covalent dimers through intermolecular disulfide bonds (data not shown).
Mesoderm Induction by XFGFR1 and XFGFR2 Mutants-Xenopus embryos were injected with the indicated amounts of wild-type or mutant XFGFR1 and XFGFR2 RNAs at the two-FIG. 2. Immunoblotting of lysates from oocytes expressing various mutant forms of XFGFR1 and XFGFR2. A, equivalent amounts of wild-type or mutant XFGFR1 (R1) RNA injected oocyte lysates were resolved by SDS-8.5% PAGE, transferred to nitrocellulose, probed with an antiphosphotyrosine monoclonal antibody and bound antibodies visualized with enhance chemiluminescence procedures. B, the immunoblot in A was stripped and reprobed with an anti-XFGFR1 monoclonal antibody to demonstrate equivalence of expression of receptor mutants. C, equivalent amounts of wild-type or mutant XFGFR2 (R2) RNA injected oocyte lysates were resolved by SDS-8.5% PAGE, and subjected to antiphosphotyrosine immunoblotting as in A. D, the immunoblot in C was stripped and reprobed with a rabbit anti-FGFR antibody to demonstrate equivalence of receptor expression in injected oocytes. For A and C the molecular mass markers (in kilodaltons) are shown to the left. The arrows to the right in A and C denote tyrosinephosphorylated polypeptides that are more prominent in oocytes expressing constitutively activated receptors than in oocytes expressing wild-type receptors.
FIG. 3. Autophosphorylation of wild-type and mutant forms of XFGFR1 and XFGFR2. Lysates were prepared from control oocytes or oocytes injected with the indicated RNAs encoding wild-type or mutant forms of XFGFR1 or XFGFR2. Equivalent amounts of receptor protein as shown for Fig. 2 were immunoprecipitated from oocyte lysates with anti-FGFR antibodies, washed, and subjected to in vitro kinase reactions as described previously (32). Products of the in vitro kinase reaction were separated by electrophoresis on SDS-8.5% PAGE and subjected to autoradiography. cell stage. Animal caps were dissected from injected embryos at stage 8 -9. When sibling control embryos reached neurula stages, animal caps were scored for elongation, an indicator of mesoderm induction (42). Caps from uninjected embryos rarely exhibited elongation, whereas uninjected embryos treated with FGF exhibited elongation indicative of mesoderm induction in greater than 90% of cases (see Table I) (42). Animal caps from embryos injected with 10 -20 pg of XFGFR1-WT RNA per embryo did not exhibit elongation above that of untreated controls. Injection of high doses (50 -100 pg/embryo) of XFGFR1-WT did result in elongation in 10 -20% of cases. In contrast, animal caps from embryos injected with 5-10 pg/ embryo of XFGFR1-C289R RNA exhibited elongation responses in 74 -86% of animal caps in the absence of exogenously added FGF, and were similar to FGF treated control animal caps. To determine whether activation of XFGFR1 by the C289R mutation is specific to arginine, we also tested a C289K mutation in this assay. The XFGFR1-C289K mutant (50 -100 pg of RNA/embryo) did not induce elongation above that of XFGFR1-WT. Similarly, expression of the kinase-deficient XFGFR1-C289R/K420A (50 -100 pg of RNA/embryo) mutant which contains the activating transmembrane domain mutation did not induce elongation in animal cap explants. The expression of the tyrosine kinase domain mutant, XFGFR1-K562E (5-10 pg of RNA/embryo), was also found to result in ligand independent elongation of 79 -94% of animal caps. In addition, two extracellular domain mutations of XFGFR1 were tested in this assay. The XFGFR1-C249Y mutant induced elongation of animal caps when embryos were injected with 25-50 pg of RNA/embryo (42-89%), whereas XFGFR1-P160R did not induce elongation when embryos were injected with 50 -100 pg of RNA. Injection of XFGFR2-WT RNA at either 50 or 100 pg/embryo rarely induced elongation (Table I) (37). Expression of XFGFR2-C268F, which contains a mutation that creates an unpaired cysteine residue in the IgIII domain of XFGFR2, also resulted in elongation of 78 -85% of animal caps when embryos were injected with 25-50 pg of RNA (Table I). These results are similar to that reported for the XFGFR2-C332Y mutant (37). Finally, the XFGFR2-C268F/C332Y mutant which exhibited higher tyrosine kinase levels than XFGFR2-WT in oocytes, failed to induce elongation in animal cap explants.
FGF-mediated mesoderm induction has been shown previously to be blocked by expression of a dominant negative Raf (C4-Raf) (44) or a dominant negative Ras (N17-Ras) (45). Coinjection of C4-Raf with XFGFR1-C289R, K562E, C249Y, and XFGFR2-C268F inhibited FGF-independent mesoderm formation as measured by animal cap elongation (Table I). We have previously reported that both C4-Raf and N17-Ras inhibited mesoderm induction by the constitutively activated XFGFR1-C332Y (37). Here we demonstrate that ligand-independent mesoderm induction by XFGFR1-C289R can be inhibited by N17-Ras as well as by the dominant negative Raf (Table I). Since all constitutively activated receptors are inhibited by C4-Raf and selected mutants are also inhibited by N17-Ras, it is likely that the dominant negative Ras will inhibit the actions of all constitutively activated receptors.
To further evaluate ligand-independent mesoderm induction by these mutant FGFRs, we have isolated RNA from animal caps incubated to control stages 22-30 and measured the induction of muscle specific ␣-actin mRNA by RNA gel blot analysis (37) (Fig. 6). Injection of 5-10 pg of XFGFR1 mutants C289R and K562E or 25-50 pg of XFGFR1-C249Y was sufficient to induce ␣-actin mRNA expression to a level similar to that induced by uninjected animal caps treated with FGF. Injection of similar amounts of XFGFR1-WT did not induce ␣-actin expression. However, injection of 2-10-fold greater amounts of two-loop XFGFR1-WT (50 -100 pg) did occasionally induce low levels of ␣-actin expression. In addition, injection of 25-50 pg of XFGFR2-C268F RNA also induced ␣-actin mRNA expression, whereas XFGFR2-WT and XFGFR2-C268F/C332Y did not induce expression when injected at 50 -100 pg/embryo. Furthermore, ␣-actin mRNA induction by XFGFR1 mutants C289R, K562E, C249Y, and the XFGFR2-C268F mutant was inhibited by co-injection of C4-Raf RNA. These results confirm the ligand-independent mesoderm inducing capacity of receptor mutants XFGFR1-C289R, K562E, and C249Y, and XFGFR2-C268F as determined by both morphological and molecular criteria.
Effect of Dominant Negative FGFR1 on Signaling by Activated FGFRs-FGF mediated mesoderm induction has been shown previously to be blocked by the expression of a dominant negative FGFR (XFD) (5). To gain some insight into the mechanism by which these constitutively activated receptors transmit the mesoderm inducing signal, we have co-injected XFD RNA with RNA encoding activated FGFRs and assayed for animal cap elongation. Embryos at the two-cell stage were injected with XFGFR1-C289R, XFGFR1-K562E, or XFGFR2-C268F RNA without or with XFD RNA at a 1:10 or 1:50 ratio and allowed to develop to stage 8 -9 before dissection. Animal caps were incubated to neurula stages at which time they were scored for elongation. Animal caps from embryos injected with XFGFR1-C289R, XFGFR1-K562E, and XFGFR2-C268F all ex- hibited elongation responses typical for these activated receptors (Fig. 7). Co-injection of XFD RNA resulted in a dose-dependent inhibition of animal cap elongation induced by XFGFR1-C289R and XFGFR2-C268F expression. In contrast, XFD expression had little effect on animal cap elongation induced XFGFR1-K562E. These results indicate that there may be fundamental differences in the mechanisms by which these individual mutations result in the ligand-independent activation of FGFRs. DISCUSSION Within the last two years over 30 point mutations resulting in single amino acid substitutions have been identified in the human FGFR1, FGFR2, and FGFR3 genes (9,22). These substitutions occur in several structural domains of these receptors including the interloop region between IgII and IgIII, several sites in IgIII, the transmembrane domain, and the tyrosine kinase domain (9). The diversity of these mutations do not indicate a common mechanism for FGFR dysfunction. How-  6. Induction of ␣-actin mRNA expression by FGFR mutants. Embryos at the two-cell stage were injected with the indicated wild-type or mutant FGFR RNAs in the specified amounts. For experiments involving the dominant negative Raf mutant, C4-Raf RNA was injected at a 10-fold excess over the indicated amount of receptor RNA. Animal caps (6 each) were dissected at stage 8 -9 and harvested at stage 22-30. Total RNA was isolated and ␣-actin mRNA expression was analyzed by RNA gel blot hybridization. The ␣-actin mRNA band is indicated by the arrow. The upper two bands are cytoskeletal actin transcripts.

FIG. 7. Inhibition of constitutively activated FGFRs by XFD.
Embryos were injected at the two-cell stage with 10 pg of either XFGFR1-C289R or XFGFR1-K562E or 50 pg of XFGFR2-C268F without or with either a 10-or 50-fold excess of XFD RNA. Embryos were dissected at stage 8 -9 and incubated in 0.5 ϫ MMR containing 1 mg/ml BSA and 50 g/ml gentamicin. Embryos were scored for evidence of convergent extension (elongation), an indicator of mesoderm induction, at neurula stages. The results are representative of at least three independent experiments. ever, all of these mutations have been implicated in the etiology of several dysmorphic syndromes of craniofacial and limb development in humans (22). We have previously reported that microinjection of Xenopus embryos with RNA encoding a FGFR2 protein with a C332Y point mutation found in Crouzon syndrome results in constitutive activation of FGFR2 (37). Therefore, we hypothesized that many of the other mutations identified in human FGFR1, FGFR2, and FGFR3 that are associated with dysmorphic syndromes of bone growth and development could also result in ligand-independent activation. To test this hypothesis, we made missense point mutations in the extracellular, transmembrane, and tyrosine kinase domains of XFGFR1 analogous to mutations identified in human FGFR1, FGFR2, and FGFR3 (22). Due to the high degree of sequence identity and structural similarity between members of the FGFR family, we anticipate that mutations made at analogous positions within the FGFR family will have similar functional consequences. Furthermore, we have made additional mutations in XFGFR2 that are analogous to mutations in human FGFR2 to extend our previous observations (37).
Achondroplasia, the most common form of dwarfism, has been shown to be associated with point mutations in the transmembrane domain of human FGFR3 (23,24). Greater than 90% of these mutations result in a G380R substitution (23,24). To test the hypothesis that this mutation will activate FGFRs, we made the analogous mutation in XFGFR1. This mutation results in increased tyrosine kinase activity of the mutant receptor when expressed in Xenopus oocytes, as well as ligandindependent mesoderm induction in animal caps when expressed in embryos. A similar mutation made in XFGFR2 also resulted in increased tyrosine kinase activity. 2 The constitutive activation of XFGFR1 by the C289R mutation is similar to the constitutive activation of the neu oncogene by a V664E mutation in its transmembrane domain (46). It has been postulated that the V664E mutation in neu results in a stabilization of a dimeric conformation of the receptor with concomitant activation tyrosine kinase activity (47). Our data indicate that a similar mechanism may be responsible for the activation of FGFRs bearing the transmembrane domain mutation described here. Activation of tyrosine kinase receptors at this position in the transmembrane domain appears to be specific to Arg since substitution with Lys at this position did not result in ligand-independent activation of the receptor. In addition, it has been reported recently that substitution of the transmembrane domain of neu with the mutant human FGFR3 (G380R) transmembrane domain results in ligand-independent activation of neu (48).
Thanatophoric dysplasia type II is a severe and lethal form of dwarfism that resembles homozygous achondroplasia in phenotype (25). Thanatophoric dysplasia type II is associated with a missense mutation (K650E) in the tyrosine kinase domain of FGFR3 (25). The analogous mutation in the two Ig loop form of XFGFR1 (K562E) results in increased tyrosine kinase activity when expressed in Xenopus oocytes and ligand-independent mesoderm induction when expressed in animal caps. Due to the high degree of sequence similarity between XFGFR1 and human FGFR3 in the tyrosine kinase domain, the K650E mutation in FGFR3 is likely to be an activating mutation. The K562E mutation we have introduced into XFGFR1 is adjacent to a pair of tyrosine residues (Tyr-559 and Tyr-560) that are highly conserved among many receptor tyrosine kinases that become phosphorylated upon receptor activation (8). In fact, both XFGFR1 and FGFR3 share the identical sequence motif DYYKK (residues 558 -562 and 646 -650, respectively) within the kinase domain which is nearly identical to the sequence DYYRK (residues 1161-1165) found in the insulin receptor kinase domain (49). A mutation in the insulin receptor that introduces an R1164Q substitution near the autophosphorylation site has been identified in patients with non-insulin-dependent diabetes and has also been shown to be an activating mutation (49). Analysis of the crystal structure of the insulin receptor kinase domain demonstrates that Tyr-1162, in the unphosphorylated state, acts as a kinase inhibitor by binding to the active site of the enzyme (49,50). This conformation appears to make the ATP-binding site inaccessible, further inhibiting kinase activity (49,50). The identification of missense mutations in both the insulin receptor (49) and FGFRs (25) that alter the two basic residues adjacent to the major autophosphorylation site, resulting in constitutive activation, indicates that binding of the inhibitor tyrosine hydroxyl group to the active site may be affected by these mutations. It is possible that this mutation may alter the structure of the tyrosine kinase domain allowing it to assume an activated conformation. Crystallographic studies of the insulin receptor tyrosine kinase suggests that this indeed may be the case (49).
Each of the four FGFRs have distinct patterns of expression during embryonic development indicating that each may have specialized functions in this process (9). FGFR1, FGFR2, and FGFR3 are expressed in bone primordia (16,17). However, during endochondral ossification, only FGFR3 is expressed in resting chondrocytes, suggesting a role for FGFR3 in this process (17). Disruption of the fgfr3 gene in the mouse results in progressive bone dysplasia with excessive endochondral bone growth (21). Contrasting the fgfr3 Ϫ/Ϫ phenotype with the achondroplasia and thanatophoric dysplasia phenotypes suggests that these dysmorphic syndromes are the result of mutations that activate FGFR3 (21,(23)(24)(25). Evidence presented here further supports the conclusion that mutations in FGFR3 associated with abnormalities of skeletal development result in constitutive receptor activation.
Other inherited skeletal disorders including Apert (30), Crouzon (29), Jackson-Weiss (31), and Pfeiffer (28) syndromes have been associated with mutations in the extracellular domains of either human FGFR1 or FGFR2. Results presented in this report and previously (37) indicate that missense point mutations that result the loss of one of the two conserved cysteine residues that stabilize the structure of the IgIII domain, as seen in cases of Crouzon (29), Jackson-Weiss (31), and Pfeiffer (28) syndromes, result in covalent dimerization through the formation of intermolecular disulfide bonds. Our data further indicate that the loss of either Cys residue that forms the disulfide bond in IgIII of FGFR2 is sufficient to result in covalent dimerization and constitutive activation. Our results also indicate that loss of one of the paired Cys residues that form IgIII in XFGFR1 also results in covalent dimerization, increased tyrosine kinase activity, and ligand-independent mesoderm induction in animal caps expressing the mutant receptor. All of these Cys mutations result in the loss of ligand binding by the mutant receptors. These results support our initial hypothesis that due to the high degree of sequence identity and structural similarity between members of the FGFR family, a mutation that constitutively activates one member of the family will also activate other members of the receptor family. Accordingly, these results indicate that mutations analogous to those described here will likely activate all members of the FGFR family.
As part of our studies on the effects of mutating Cys-268 or Cys-332 on XFGFR2 function, we also constructed a mutant with both Cys residues mutated (XFGFR2-C268F/C332Y). Interestingly, this mutant demonstrated elevated levels of tyro-sine kinase activity above that of XFGFR2-WT, but similar to that seen with the XFGFR2-C268F mutant when expressed in Xenopus oocytes. In Xenopus oocytes, activation of XFGFR2-WT tyrosine kinase in the absence of ligand is presumably due to a high density of receptors in the plasma membrane (37). Further increases in tyrosine kinase activation of XFGFR2-C268F are due to stabilization of the dimeric conformation through intermolecular disulfide bonds. However, unlike the XFGFR2-C268F mutant, XFGFR2-C268F/C332Y is unable to form covalent dimers as would be predicted by the absence of a free Cys residue. XFGFR2-C268F/C332Y is also unable to bind radiolabeled ligand indicating that these mutations disrupt the structure of IgIII. It is possible that the loss of both Cys residues in the XFGFR2-C268F/C332Y mutant results in a partial unfolding of the Ig III domain resulting in a less compact and more flexible conformation. A less compact and more flexible conformation could increase the probability of receptor interaction on the oocyte plasma membrane and may lead to activation of the receptor tyrosine kinase in a greater number of mutant receptors compared to wild-type receptors. Although XFGFR2-C268F/C332Y did display elevated levels of tyrosine phosphorylation above that of the wild-type receptor, this receptor mutant was unable to induce mesoderm in animal caps in a ligand-independent manner. These results indicate that high level expression of some mutant receptors in the oocyte plasma membrane may result in increased tyrosine kinase activity, without resulting in a measurable biological response such as mesoderm induction or mitogenesis. Additional mutational analysis of FGFR2 IgIII is required to further address this issue.
At present, the only mutation identified in human FGFR1 is a missense mutation resulting in a P252R substitution in the linker region between the IgII and IgIII domains that is associated with Pfeiffer syndrome (27). We have made the analogous mutation in the two Ig loop form of XFGFR1 and expressed the protein in Xenopus oocytes and developing embryos. Expression of this mutant receptor in oocytes resulted in tyrosine kinase activity comparable to that of the wild-type receptor and less than that of other mutant receptors used in these studies. Expression of XFGFR1-P160R in animal caps failed to induce mesoderm formation in animal caps even when RNA was injected at 2-10-fold higher amounts than for the other mutant receptors. The observation that this mutant seemed to bind slightly more radiolabeled FGF than FGFR1-WT in cross-linking experiments led us to investigate whether this mutant receptor might respond to lower concentrations of FGF-1 than wild-type receptor in an animal cap assay. Injection of embryos with equal amounts of either XFGFR1-WT or XFGFR1-P160R, dissection of animal caps, and treatment with various doses of FGF-1 did not reveal any differences in response between the two receptors. 2 These results have several possible interpretations. First, it is possible that in cases where these mutations have been identified, other mutations may exist that have gone undetected. However, this seems unlikely since an analogous mutation has been identified in FGFR2 in individuals with Apert syndrome (30). Second, since this region of the receptor has been implicated to play a role in ligand binding, it is possible that the ligand binding specificity or affinity has been altered and we have not detected these subtle changes with our assay system. Third, it is possible that this mutation may have an effect on the three Ig-loop form of FGFR1 but not on the two Ig-loop form. The effects of this mutation on receptor function as well as its role in skeletal disorders requires further investigation.
During the course of these studies we noted that 10 -20% of animal caps from embryos injected with 50 -100 pg of XFGFR1-WT (two Ig-loop form) RNA were induced to form mesoderm. This is much less induction than was observed for XFGFR1-C289R (74 -86%) or XFGFR1-K562E (80 -94%) which were injected at 10-fold lower amounts (5-10 pg) of RNA. Indeed, when 10 -20 pg of XFGFR1-WT was injected into embryos, no mesoderm induction was observed supporting our conclusion that the mutant receptors are constitutively activated. However, the 10 -20% induction of mesoderm observed in animal caps from embryos injected with 50 -100 pg of XFGFR1-WT RNA is more than was observed for XFGFR2-WT (three Ig-loop form). Animal caps from embryos injected with 50 -100 pg of XFGFR2-WT RNA gave at most 3% induction of mesoderm. It is possible that the two Ig-loop form of the receptor has higher basal tyrosine kinase activity or is more sensitive to endogenous ligand. A recent report indicates that the two Ig-loop form of FGFR1 has at least a 10-fold higher affinity for both FGF-1 and heparin, supporting the later possibility (51). It has been hypothesized that the lower affinity for ligand of the three Ig-loop form of FGFR1 is due to a conformation of the receptor in which the Ig-loop I interacts with Ig-loops II and III resulting in a partial inhibition of FGF binding (51).
FGF-mediated mesoderm induction in animal caps has been previously shown to be inhibited by a dominant negative Ras (45), dominant negative Raf (44), and a dominant negative FGFR (XFD) (5). Expression of a dominant negative Raf or a dominant negative Ras with the constitutively activated FGFRs resulted in an inhibition of ligand-independent mesoderm induction by these receptors. These results are consistent with the ability of dominant negative forms of Raf or Ras to disrupt the FGFR signaling pathway. We also employed the dominant negative FGFR construct XFD in these studies. Results from these experiments demonstrate that different mutations may activate FGFRs by different mechanisms. Co-injection of XFD with either XFGFR1-C289R or XFGFR2-C268F at a ratio of 50:1 resulted a 2.5-3.0-fold reduction in the number of animal caps elongated. In contrast, experiments performed in a similar manner with XFD and XFGFR1-K562E demonstrate that XFD was unable to inhibit ligand-independent animal cap elongation by this FGFR mutant. These results have several implications. First, constitutive activation of XF-GFR1-C289R and XFGFR2-C268F appears to require dimerization, that can be inhibited, presumably by the formation of non-functional heterodimers with XFD. Second, the failure of XFD to inhibit ligand-independent mesoderm induction by XF-GFR1-K562E indicates that the mechanism by which this mutant is activated may be different than than for XFGFR1-C289R or XFGFR2-C268F. As discussed previously, the K562E mutation may alter the conformation of the kinase domain such that Tyr-559 binds less tightly to the active site of the kinase thus relieving autoinhibition. In this more permissive conformation, it may be possible for two kinase domains to interact and trans-phosphorylate one another without the necessity to juxtapose the extracellular and/or transmembrane domains. Alternatively, in the more permissive conformation, which may also unmask the ATP-binding site, cis-autophosphorylation may be favored, thus abrogating the need for dimerization to obtain activation of the kinase. The lack of inhibition of XFGFR1-K562E by XFD supports both of these possibilities. The possibility that different mutations may activate FGFRs by different mechanisms suggests that there may be some correlation to mode or extent of FGFR activation and the severity of the phenotype observed. It is interesting to note that thanatophoric dysplasia II has the most severe phenotype of the skeletal dysplasias associated with FGFR mutations (25), and that the thanatophoric dysplasia type II mutation K562E in XFGFR1 appears to be the most activating mutation in an animal cap assay.
The results in this report demonstrate that many of the mutations identified in human FGFRs that are associated with skeletal disorders result in the ligand-independent activation of these receptors. Our results also indicate that there may be subtle functional differences between FGFRs bearing different mutations that may affect the severity of the disorder. Furthermore, the identification of gain-of-function mutations for FGFRs will facilitate an analysis of the relative contributions of each receptor to developmental processes.