|Home | About | Journals | Submit | Contact Us | Français|
M. H., D. M. O. and H. K. developed the project and wrote the manuscript. M. H., S. H. and H. K. contributed to the purification of FGF9 proteins, mitogenic assays, analytical gel filtration chromatography, analytical heparin affinity chromatography, surface plasmon resonance analysis, skeletal preparation, histological analyses and in situ hybridization of sections, implantation of FGF9 beads in mouse forelimb buds, IP/Western analysis. H. M., A. K. and H. K. contributed to the identification of the Eks mutation. N. O., N. F. and M. T. contributed to the MD simulation. T. N. and S. I. contributed to the implantation of FGF9 beads in mouse fetal skulls. R. A., M. S. and S. Y. contributed to the analytical ultracentrifugation. W. S. and A. K. contributed to the retroviral misexpression. Y. M-K. contributed to the in situ hybridization of sections.
The spontaneous dominant mouse mutant, Elbow-knee-synostosis (Eks), exhibits elbow and knee joint synosotsis, and premature fusion of cranial sutures. Here we identify a missense mutation in the Fgf9 gene that is responsible for the Eks mutation. Through investigation of the pathogenic mechanisms of joint and suture synostosis in Eks mice, we identify a key molecular mechanism that regulates FGF9 signaling in developing tissues. We show that the Eks mutation prevents homodimerization of the FGF9 protein and that monomeric FGF9 binds to heparin with a lower affinity than dimeric FGF9. These biochemical defects result in increased diffusion of the mutant FGF9 protein (FGF9Eks) through developing tissues, leading to ectopic FGF9 signaling and repression of joint and suture development. We propose a mechanism in which the range of FGF9 signaling in developing tissues is limited by its ability to homodimerize and its affinity for extracellular matrix heparan sulfate proteoglycans.
The fibroblast growth factors (FGFs) are widely expressed in developing and adult tissues and have diverse functions in organogenesis, tissue repair, nervous system control, metabolism and physiological homeostasis1. In humans and mice, the 22 FGF ligands are expressed in a spatiotemporally regulated manner and mediate signals through 7 different isoforms of FGF receptors (FGFRs)1. The pharmacologic potential of FGF ligands has been highlighted by identification of gain-of-function mutations in genes encoding Fgfrs 1–3 in patients with chondrodysplasia and craniosynostosis syndromes2, 3. These human diseases identify essential roles for FGF signaling not only in development but also for homeostasis of bones and joints.
Based on these clinical, genetic and biochemical studies in humans and mice, the coordinated development of bones and joints appears to rely on critical levels of FGFR signaling. This suggests that spatiotemporal constraints on FGF signaling are a prerequisite for appropriate functions in vivo and are indeed modulated at several distinct levels. First, there is spatiotemporal restricted expression of FGF ligands. Among the 22 FGF ligands, FGF2, FGF4, FGF7, FGF8, FGF9, FGF10, FGF17 and FGF18 are expressed in the limb bud and developing skeleton4–6. Of these, loss-of-function mutations have demonstrated the involvement of FGF2, FGF9 and FGF18 in chondrogenesis and/or osteogenesis7–10. Induction of chondrodysplastic phenotypes by overexpression of FGF9 in mice also demonstrates its ability to affect chondrogenesis11. Other elements implicated in FGF signaling are the heparan sulfate proteoglycans (HSPGs). Genetic studies in mice and Drosophila melanogaster suggest that HSPGs regulate the distribution and receptor binding of FGF ligands12, 13. Finally, structural analyses of FGF9 suggest that it may form homodimers that could affect its ability to signal14, 15. Since FGF9 homodimerization occludes several critical receptor binding sites, an autoinhibitory mechanism may function to modulate FGF9-dependent signal transduction. However, a functional demonstration of this proposed mechanism is lacking.
We have previously reported that a dominant mouse mutant, Elbow-knee-synostosis (Eks), exhibits radiohumeral and tibiofemoral synostosis, craniosynostosis (Supplementary Fig. 1), and lung hypoplasia16. In this study, we identify a novel missense mutation, which replaces Asn143 with Thr in the Fgf9 gene in Eks mutant mice. We designate this mutant allele as Fgf9Eks and show that this mutation prevents homodimerization of FGF9, consequently decreasing the affinity of FGF9 for heparin. As a result, FGF9Eks is more diffusible in developing tissues leading to ectopic FGF9 signaling in the prospective joints and sutures where it functions to repress development. Molecular dynamics (MD) calculations suggest that the reduction in FGF9 affinity for heparin is due to the predominance of the monomeric form rather than to changes in its intrinsic affinity for heparin. We thus propose a mechanism in which the range of FGF9 signaling in developing tissues is limited through regulation of its affinity for HS, which is at least in part controlled by the FGF9 monomer/dimer equilibrium. These observations could have far-reaching implications for the pharmacologic manipulation of FGF signaling under a variety of circumstances and in a wide range of tissues.
The Eks mutation was mapped between the polymorphic markers D14Mit62 and D14Mit5 on mouse chromosome 14 (ref. 16). Among 169 genes located in this interval, Fgf9 seemed a likely candidate for the Eks mutation since FGF9 is a ligand for FGFR2c and FGFR3c17 and is expressed in the developing limbs, cranial sutures and developing lung8, 18, 19. Sequence analysis of Fgf9 cDNA from homozygous Eks mice revealed an A to C substitution at position 428, which resulted in the replacement of Asn143 with Thr (Fig. 1a). Intriguingly, the Asn143 residue in FGF9 is highly conserved among most FGF proteins (Fig. 1b) and is predicted to be a critical amino acid residue for homodimerization and receptor binding14, 15.
We used a genetic approach to determine whether the N143T substitution in Fgf9 was responsible for the Eks phenotype. We observed a Mendelian pattern of inheritance of the mutation among 976 offspring of Eks heterozygous (Fgf9Eks/+) matings, with heterozygous mice exhibiting mild skeletal phenotypes and homozygous Fgf9Eks/Eks littermates exhibiting severe skeletal phenotypes. The Eks phenotype and the mutation in Fgf9 co-segregated in all cases. The absence of recombination between Eks and Fgf9 among nearly 2000 meiotic events provides strong evidence that the Eks mutation is allelic with Fgf9.
Eks is a dominant mutation and joint synostosis and premature fusion of sutures have also been reported in mice expressing gain-of-function alleles of Fgfr1 (ref. 20) and Fgfr2c (Fgfr2C342Y)21. Furthermore, homozygous Fgf9 null mutants (Fgf9−/−) exhibit rhizomelia but do not show joint or suture synostosis8. Thus we hypothesized that FGF9Eks encodes a gain-of-function mutation. To test this possibility, we first investigated whether there were phenotypic similarities between Eks mutants and gain-of-function mutants for Fgfr1 and Fgfr2C342Y. Since initiation of elbow joint development was primarily impaired in Fgfr1 gain-of-function transgenic mice20, we examined radiohumeral joint development in Fgf9Eks/Eks mice (Fig. 2a–l). Gdf5 (ref. 22) and Col2a1 (ref. 23) demarcate the prospective elbow joint and cartilaginous condensation, respectively. Gdf5 expression in the prospective joint space was observed as early as E11.5 in Fgf9+/+ control mice (Fig. 2i), but was completely absent in Fgf9Eks/Eks mice (Fig. 2j). Analysis of the prospective cartilage revealed a gap of Col2a1 expression at the prospective elbow joint in E11.5 wild type embryos (Fig. 2k). The gap of Col2a1 expression was absent in Fgf9Eks/Eks mice (Fig. 2l). In summary, ectopic chondrocyte differentiation in the prospective elbow joint of Fgf9Eks/Eks mice appears very similar to that seen in transgenic mice that ectopically express an activated Fgfr1 kinase domain in the presumptive joint field20.
Premature fusion of coronal sutures in Fgfr2C342Y mice results from excess osteogenic differentiation within the coronal suture mesenchyme21. To determine whether Fgf9Eks/Eks mice had similar histological features, we examined mineralization and the expression of the early osteoblast differentiation markers Spp124 and Runx225 in the coronal suture. At E15.5, both wild type and Fgf9Eks/Eks mice showed similar coronal suture histology (Fig. 2m–t). However, at E16.5 von Kossa staining revealed significantly more overlap of the frontal and parietal bones in Fgf9Eks/Eks mice compared to Fgf9+/+ mice (Fig. 2w, x). Spp1 expression domains, which demarcate preosteoblasts and osteoblasts, showed wide separation of the frontal and parietal bones in Fgf9+/+ mice (Fig. 2y), however there was significant overlap in the Fgf9Eks/Eks mice (Fig. 2z). Runx2 is highly expressed in immature osteoblasts at the leading edge of the frontal and parietal bones (Fig. 2a’). In E16.5 Fgf9Eks/Eks mice, the intensity of Runx2 expression in the coronal suture was lower than in Fgf9+/+ mice (Fig. 2b’), suggesting premature differentiation of the osteoblasts at the presumptive suture. These phenotypes reflect abnormal osteogenic differentiation within the coronal suture mesenchyme and indicate that the defect in suture development occurs before E16.5. Taken together, these observations suggest that the FGF9Eks mutation mediates excess FGFR signals within the prospective joints and sutures, preventing joint formation and promoting the fusion of cranial sutures.
The predicted involvement of the Asn143 residue in homodimerization and receptor activation14, 15 suggests that changes in these processes may account for the apparent gain-of-function activity of the FGF9Eks mutant. Homodimerization of FGF9 has been proposed to occlude receptor binding sites and consequently mediate an autoinhibitory mechanism for FGF9 signaling. We thus hypothesized that the Eks mutation might impair the autoinhibition. To test this possibility, we compared the degree of homodimerization of wild type (FGF9WT) and FGF9Eks proteins by analytical ultracentrifugation. FGF9WT and FGF9Eks were expressed in E. coli and purified by column chromatography (Supplementary Methods and Supplementary Fig. 2).
The molecular mass and association constant of FGF9WT and FGF9Eks were determined by sedimentation equilibrium centrifugation using the purified recombinant proteins (Fig. 3a, b). An estimated average molecular mass of FGF9WT and FGF9Eks was 39,264 and 32,929 Da, respectively, whereas the calculated monomeric molecular mass was 20,090 and 20,077 Da, respectively. These data suggest that FGF9WT primarily exists as a dimer, whereas FGF9Eks exists in a monomer/dimer equilibrium. The calculated association constants of FGF9WT and FGF9Eks were 10.4 µM−1 and 0.119 µM−1, respectively. We further measured the sedimentation coefficient of FGF9Eks by sedimentation velocity centrifugation. The overlay plots of c(s)-sedimentation coefficient distributions show that FGF9WT has a unimodal peak at 3.0 S for a single ideal species, whereas FGF9Eks has bimodal peaks (2.2 S and 3.1 S) for two ideal species (Fig. 3c, d). This observation suggests that FGF9WT is present primarily as a dimer, whereas FGF9Eks exists primarily as a monomer. These interpretations are consistent with the retarded elution of FGF9Eks relative to FGF9WT on a gel filtration column (Fig. 3e). Therefore, FGF9Eks is defective in homodimer formation.
To examine whether the Eks mutaion altered the binding of FGF9 to FGFRs by impairing the autoinhibitory mechanisms, we compared the ability of FGF9WT and FGF9Eks to activate FGFRs by assaying the mitogenic activity of both proteins on BaF3 cells expressing individual FGF receptors17. FGFR-expressing BaF3 cell lines were treated with increasing concentrations of purified recombinant FGF9 in the presence of 1 µg/ml heparin. Compared to FGF9WT, FGF9Eks showed less activity on cells expressing any of the FGFRs except FGFR3c, where FGF9Eks showed equivalent activity (Fig. 4a–g). To test the ability of heparin to enhance FGF9 activity, the BaF3 cell lines were treated with increasing concentrations of heparin in the presence of 0.2 nM FGF9WT or FGF9Eks. FGF9Eks also showed a decreased heparin-dependent mitotic response on all FGFRs except FGFR3c, where FGF9Eks showed higher activity in the presence of high concentrations of heparin (Fig. 4h–n). Since FGF9Eks does not mediate excess signaling via FGFRs, other properties of the mutant protein must be responsible for the phenotype of the Eks mice.
The decreased heparin-dependent mitogenic activity of FGF9Eks suggested that its affinity for heparin may be reduced. Heparin is functionally very similar to HS, which is present in most tissues in the form of HSPGs. HSPGs function to modulate FGFR activation directly, by mediating FGF-FGFR interactions, and indirectly, by binding FGF ligands and regulating their diffusion through the extracellular matrix and thus their access to distant FGFRs1, 12, 13, 26. Since the gain-of-function property of FGF9Eks may not involve direct interaction with the FGFRs, we hypothesized that its decreased affinity for heparin might allow increased diffusion and thus bioavailability in tissues. To address this possibility, we first measured FGF9/heparin affinity by heparin affinity chromatography (Fig. 5a). FGF9WT was eluted from heparin-conjugated agarose with 1.50 M NaCl as a single peak. In contrast, most FGF9Eks was eluted at 1.38 M NaCl and a small fraction eluted at 1.10 M NaCl.
We next measured the kinetic constants for the FGF9Eks/heparin interaction using surface plasmon resonance analysis (Fig. 5b, c and Supplementary Table 1). The resulting sensorgrams were used for kinetic parameter determination by globally fitting the experimental data to a 1:1 interaction model. The association rate constant (ka) of FGF9Eks was slightly greater than that of FGF9WT, whereas the dissociation rate constant (kd) of FGF9Eks was 18 fold greater than that of FGF9WT. The dissociation constants (KD) for FGF9WT and FGF9Eks were 0.71±0.02 nM and 5.24±0.03 nM respectively, representing an 86% decrease in affinity of the FGF9Eks protein for heparin.
The above studies indicate the FGF9Eks mutation concurrently affects monomer/dimer equilibrium and affinity for heparin/HS. We thus went on to address whether the Asn143Thr mutation directly affects the affinity of FGF9 for heparin versus directly affecting homodimerization and secondarily affecting heparin affinity. However, direct biochemical measurements of the affinity of the two species for heparin are not possible because monomeric and dimeric forms of FGF9 are in equilibrium. We therefore analyzed the configuration of heparin-binding domains in monomeric and dimeric FGF9 using MD simulations and calculated the binding free energy between FGF9 and heparin using the molecular mechanics Poisson-Boltzmann/surface area (MM-PBSA) method. It is well-known that the binding free energies calculated by this method show good qualitative but not quantitative agreement with experimental observations27.
To model the heparin binding affinity of FGF9, we performed MD simulations of 2:2 FGF9WT-heparin and 2:2 FGF9Eks-heparin complexes based on a 2:2:2 FGF2-FGFR1-heparin crystal structure (Protein Data Bank (PDB) ID: 1FQ9)28. The conformations of two heparin oligosaccharides in each complex were influenced by strong electrostatic repulsions, resulting in the exclusion of one heparin oligosaccharide from the complex (data not shown). This analysis suggested that 2:2 FGF9-heparin complexes would be unstable. In contrast, MD simulations of 2:1 FGF9WT-heparin, 2:1 FGF9Eks-heparin, 1:1 FGF9WT-heparin and 1:1 FGF9Eks-heparin complexes suggested that these complexes are stable (Fig. 5d–g). MD simulations of dimeric FGF9-heparin complexes did not show a big difference in heparin-binding free energies for 2:1 FGF9WT-heparin (dimeric FGF9WT-heparin) and 2:1 FGF9Eks-heparin (dimeric FGF9Eks-heparin) complexes (Fig. 5d, e). This is due to the strong interaction between the negatively charged heparin oligosaccharide chain and the array of basic amino acid residues located in the heparin binding site near the groove of the dimer interface in both dimeric complexes. In addition, flexibility of the heparin oligosaccharide chain would help to maintain electrostatic interactions. Similarly, there was little difference in the heparin-binding free energies in 1:1 FGF9WT-heparin (monomeric FGF9WT-heparin) and 1:1 FGF9Eks-heparin (monomeric FGF9Eks-heparin) complexes (Fig. 5f, g). This is also due to heparin oligosaccharide chain flexibility, the strong negative charge of the heparin oligosaccharide and the presence of several basic amino acid residues in the heparin binding site. Therefore, the Eks mutation does not appear to influence the heparin binding affinity of either the dimeric or monomeric FGF9-heparin complex. Because the heparin binding free energies for dimeric FGF9 (Fig. 5d, e) were smaller than those for monomeric FGF9-heparin for both FGF9WT and FGF9Eks (Fig. 5f, g), the reduced binding affinity to heparin of the FGF9Eks protein is most likely due to the shift in the monomer/dimer equilibrium towards the monomer. In summary, the Eks mutation primarily affects homodimerization of FGF9 and secondarily heparin affinity.
Heparin/FGF2 interactions have previously been shown to regulate the diffusibility of FGF2 (ref. 26, 29). We hypothesized that the diffusibility of FGF9Eks in tissues would be increased because of its lower affinity for heparin, leading to ectopic localization outside of the normal signaling domain and consequently ectopic activation of FGFRs. However, this model can only be considered if the following two prerequisites are met; first, Fgf9 and Fgfrs are expressed in the proximity of the prospective elbow and knee joints and coronal sutures and, second, the increased diffusibility of FGF9Eks is dominant over its decreased ability to activate FGFRs.
We first examined the expression of Fgf9 and Fgfr1, −2 and −3 in the forelimb buds in E10.5 and E11.5 mice. Fgf9 was expressed in migrating myoblasts, both in Fgf9+/+ and Fgf9Eks/Eks mice (Supplementary Fig. 3a, b, i, j). At E10.5, Fgfr1, −2 and −3 were expressed diffusely in the limb bud mesenchyme, overlapping the expression domain of Col2a1 in both Fgf9+/+ and Fgf9Eks/Eks tissues (Supplementary Fig. 3c–h). At E11.5, Fgfr1 expression was excluded from the cartilaginous condensation, whereas Fgfr2 and Fgfr3 expression was observed mainly in this location (Supplementary Fig. 3k–p). Therefore, mesenchymal cells in the prospective elbow joint express Fgfrs.
Previous reports showed that Fgf9 is expressed in the developing frontal and parietal bones, particularly strongly at the rims of the bones18. Fgfr1, −2 and −3 are expressed within and around the developing frontal and parietal bone domains30. Thus, the first prerequisite was validated.
To address the second prerequisite for our model, we examined the inhibitory effects of FGF9WT and FGF9Eks on joint development by ectopically expressing them in the chicken limb bud using a replication component retroviral vector (RCAS) transduction. RCAS-Fgf9WT, RCAS-Fgf9Eks, or empty RCAS virus was used to infect the prospective hindlimb bud region in the lateral plate mesoderm. FGF9WT and FGF9Eks were expressed throughout the hindlimb buds (Fig. 6a, b). Ectopic expression of both Fgf9WT and Fgf9Eks caused knee joint fusion (Fig. 6c, d, e) whereas no abnormalities were induced by the empty vector (Fig. 6f, g). Therefore FGF9Eks retains inhibitory effects on joint development as well as FGF9WT. It is notable that skeletal defects induced by the expression of FGF9WT were widespread whereas those mediated by FGF9Eks were limited to the prospective joint regions. This is consistent with our finding that FGF9Eks preferentially activates FGFR3c (Fig. 4), which is expressed in the bone anlagen (Supplementary Fig. 3), whereas FGF9WT is expected to activate all of the mesodermally expressed FGFRs.
To examine the inhibitory effects of FGF9WT and FGF9Eks on suture development, FGF9-soaked-AffiGel-Blue beads were implanted in the coronal suture of normal mouse fetal skulls around the initiation stage of the suture defect, at E15, by ex utero surgery. We first confirmed that approximately equal amounts of FGF9WT and FGF9Eks were loaded in each AffiGel-Blue bead and that the diffusion rate of FGF9WT and FGF9Eks from the beads was almost identical (Supplementary Fig. 4). The expression of Spp1, an early osteoblast differentiation marker upregulated by FGF9, was examined 24 hours after in utero bead placement. Grafts of FGF9WT and FGF9Eks beads also promoted ectopic Spp1 expression at the leading edges of the frontal and parietal bones (Fig. 6h–k). This FGF9-induced ectopic expression of Spp1 resembled that observed in the Fgf9Eks/Eks coronal suture (Fig. 2z). Therefore, ectopic expression of either FGF9Eks or FGF9WT within the suture inhibits suture development.
To examine whether the diffusibility of FGF9Eks in mesenchymal tissue is increased in comparison with FGF9WT, we measured the diffusibility of FGF9WT and FGF9Eks in the skull following bead implantation (Fig. 7a–e). Because FGF9 upregulates Spp1 expression, we could measure the area of high Spp1 expression as an indication of the distance over which FGF9 exerts its effects. Implantation of FGF9Eks beads resulted in a larger area of Spp1 expression (High Spp1 expression area / Frontal and parietal bone anlagen area = 82.9 ± 7.3%) (± s.e.m.) compared with beads loaded with FGF9WT (50.7 ± 4.4%; P=0.0035) suggesting that the mutant protein diffused more effectively through the developing tissue.
Next, we investigated the diffusibility of FGF9Eks in forelimb buds. FGF9WT- or FGF9Eks-soaked-AffiGel-Blue beads were grafted into the dorsal and central forelimb bud region of Fgf9−/− embryos around the initiation stage of the joint defect, at E10.5. FGF9 protein released from the beads into mesenchymal tissue 2 hours after implantation was detected by immunohistochemistry using FGF9 antibodies on sections (Fig. 7f–h). This analysis showed that FGF9Eks permeated through the limb bud mesenchyme to a greater extent (Relative disffusion area = 138 ± 12%) (± s.e.m.) than FGF9WT (100 ± 14%; P=0.043) supporting the hypothesis that FGF9Eks is more diffusible than FGF9WT.
To examine whether diffusion of endogenous FGF9Eks is increased in comparison to FGF9WT, we determined the level of activation of FGFRs in the prospective elbow joint of Fgf9Eks/Eks mice. As a readout for FGFR signaling, we examined the expression of Etv4 and Etv5, both of which are known to be transcriptionally activated by FGF signaling, in the forelimb buds31. In wild type E10.5 limbs, we did not observe the intensive expression of Etv4 or Etv5 within the region undergoing cartilaginous condensation demarcated by Col2a1 expression (Fig. 7m, o). However, in E10.5 Fgf9Eks/Eks limbs, we found ectopic expression of both Etv4 and Etv5 in the cartilaginous condensation (Fig. 7n, p). At E11.5, Etv4 and Etv5 were expressed in the myoblasts and cells surrounding the cartilaginous condensation in wild type mice (Fig. 7u, w), whereas in Fgf9Eks/Eks mice, we observed clear expression of Etv4 and Etv5 in the cartilaginous condensation including the prospective elbow joint position (Fig. 7v, x). These results demonstrate ectopic FGF signaling in the prospective elbow joint in Fgf9Eks/Eks mice. Since the ectopic expression domain of Etv4 and Etv5 in Fgf9Eks/Eks mice was not consistent with the Fgf9 expression domain (Supplementary Fig. 3a, b, i, j), this outcome is likely due to increased diffusion of FGF9Eks protein over a larger area than FGF9WT.
From these results, we propose a mechanism of elbow joint synostosis in Fgf9Eks/Eks mice in which FGF9Eks produced by myoblasts diffuses beyond its normal range and ectopically activates FGFRs in the prospective elbow joint, preventing joint formation (Fig. 7y). Similarly, FGF9Eks produced at the rims of the frontal and parietal bones diffuses beyond its normal area and ectopically activates FGFRs in the coronal suture mesenchyme, promoting the fusion of coronal sutures (Fig. 7z).
A prediction of our model is that the severity of synostotic phenotypes will correlate with a shift in the equilibrium of FGF9 from dimer toward monomer. By MD simulations, we estimated that FGF9WT homodimer, FGF9WT/Eks heterodimer and FGF9Eks homodimer have 10, 8, 6 inter-monomer hydrogen bonds, respectively (Fig. 8a–c and Supplementary Table 2). Based on these results, the calculated binding free energy of the FGF9WT/Eks heterodimer was between those of the FGF9WT and FGF9Eks homodimers (Fig. 8a–c). FGF9WT/Eks heterodimers are suggested to be more stable than FGF9Eks homodimers. If our model is correct, FGF9WT should interfere with FGF9Eks action by forming FGF9WT/Eks heterodimers. In other words, skeletal phenotypes due to the Eks mutation should be alleviated by the expression of FGF9WT.
We first sought experimental evidence that the FGF9WT/Eks heterodimer was substantially more stable than the FGF9Eks homodimer. We addressed this issue by immunoprecipitation (IP)/Western analysis after tagging FGF9WT and FGF9Eks with FLAG- or Myc-peptides. FLAG-FGF9WT or -FGF9Eks were over-expressed together with Myc-FGF9WT or -FGF9Eks in COS7 cells and the culture supernatants were subjected to the IP/Western analysis (Fig. 8d). We readily detected FLAG-FGF9WT/Myc-FGF9WT dimers. FLAG-FGF9Eks/Myc-FGF9WT dimers were detected at lower level, however, FLAG-FGF9Eks/Myc-FGF9Eks dimers did not form under these conditions. These data suggest that FGF9WT and FGF9Eks can form heterodimers that are more stable than FGF9Eks homodimers.
We finally examined whether the elbow synostosis caused by FGF9Eks could be alleviated by the expression of FGF9WT. We thus compared the severity of elbow synostosis in Fgf9Eks/+ and compound heterozygous mutants (Fgf9Eks/−) relative to that of Fgf9+/− and Fgf9Eks/Eks mice. Elbow joint formation was not affected in Fgf9+/− mice (Fig. 8e), whereas elbow synostosis in Fgf9Eks/Eks mice involved both cartilaginous and bony components (Fig. 8h). In contrast, the synostosis in Fgf9Eks/+ mice was limited to the cartilaginous component (Fig. 8f), while the involvement of the bony component in Fgf9Eks/− mice was similar to that of Fgf9Eks/Eks mice (Fig. 8g, h). Therefore, elbow synostosis in Fgf9Eks/− is more severe than in Fgf9Eks/+. These findings strongly support our model that the monomer/dimer status of FGF9 influences its affinity for heparin/HS and, consequently, its distribution in developing tissues.
In the present study, we identified a missense mutation in the Fgf9 gene that is responsible for the Eks mutant phenotype, which includes elbow and knee joint synostosis, and craniosynostosis. We further demonstrate that the Asn143Thr mutation in FGF9 favors formation of the monomeric form of FGF9, which binds to heparin with a lower affinity than dimeric FGF9. The decreased affinity for heparin/HS leads to increased diffusion of the mutant protein in developing tissues, resulting in ectopic FGF9 signaling. We propose that regulation of the monomer/dimer equilibrium of FGF9, and potentially of other FGFs, and its affinity for heparin/HS is a mechanism that functions to shape FGF9 concentration gradients in developing tissues. We further propose that these biochemical properties of FGF9 restrict its signaling activity to limited skeletal domains. Data presented here and in previous studies indicate that low FGF signaling in the presumptive joint space is necessary for the formation of the joint space and maintenance of an open suture20, 21. Common usage of receptor binding and homodimerization sites of FGF9 could be at least in part instrumental for local modulation of FGF9 signaling activity.
Homodimerization is suggested to be a common feature of the FGF9/16/20 subfamily32 and of FGF2 (ref. 33, 34). It is not known to what extent homodimerization affects the activity of other FGFs. The discovery that a mutation in Fgf9 can affect homodimerization, affinity for heparin/HS, and biological activity suggests that pharmacological agents that affect FGF homodimerization could be useful tools to modulate its activity.
To identify the mutation responsible for the Eks mutant phenotype, we surveyed the cDNA sequence of Fgf9 from normal (+/+), heterozygous (Eks/+) and homozygous (Eks/Eks) mice through reverse transcription-PCR (RT-PCR) and direct sequencing of RT-PCR products.
For genotyping of the Eks allele, genomic DNA spanning the Eks mutation was amplified by PCR using specific primers (Supplementary Table 3). PCR products were digested with the diagnostic BsrI restriction enzyme. Wild type mice show two bands of 147 bp and 42 bp, whereas the Eks allele shows one band of 189 bp (Supplementary Fig. 5).
Bones and cartilage of E17.5 fetuses were stained with Alizarin red and Alcian blue as previously described35. For histological preparations, tissues were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned at 5 µm, and stained with hematoxylin and eosin and von Kossa.
In situ hybridization of paraffin sections was performed as previously described36, using radiolabeled antisense RNA for Gdf5 (MGI: 95688), Col2a1 (MGI: 88452), Spp1 (MGI: 98389), Runx2 (MGI: 99829), Fgf9 (MGI: 104723), Fgfr1 (MGI: 95522), Fgfr2 (MGI: 95523), Fgfr3 (MGI: 95524), Etv4 (MGI: 99423) and Etv5 (MGI: 1096867). In situ hybridization after beads implantation in fetal skulls was performed as previously described24.
FGF9WT and FGF9Eks expression and purification were performed as described in Supplementary Methods.
All analytical ultracentrifuge experiments were performed on a Beckman Coulter XL-I analytical ultracentrifuge. The samples were diluted in 25 mM ammonium acetate buffer (pH 5.5) containing 120 mM NaCl. The partial specific volumes were estimated as 0.7317 mL/g (FGF9WT) or 0.7322 mL/g (FGF9Eks) by SEDNTERP software. All experiments were performed at 20°C and the absorbance wavelength was 280 nm. Sedimentation equilibrium experiments were carried out with six channel centerpieces, with loading concentrations of 0.8, 0.4 and 0.2 mg/ml. Data were obtained at 12, 14 and 16 k rpm for FGF9WT or at 14, 16 and 18 k rpm for FGF9Eks. A total equilibration time of 16 hours was used for each speed with scans taken at 12, 14 and 16 hours. The sedimentation equilibrium data were analyzed using the Beckman XL-A/XL-I Data Analysis software. All nine data sets (three speeds, three concentrations) were fitted together by "Self Association Model" calculation. Sedimentation velocity experiments were carried out with double sector centerpieces. Protein concentrations were 0.4, 0.3 or 0.2 mg/ml. The absorbance data were scanned 100 times every five minutes at 40 k rpm. The measurements data were analyzed by SEDFIT software.
Purified FGF9WT and FGF9Eks (100 µl of 2 mg/ml) were loaded onto a Superdex75 10/300 GL column (GE healthcare) equilibrated with a 25 mM ammonium acetate buffer (pH 5.5) containing 120 mM NaCl. The samples were eluted with the same buffer.
The ability of FGF9WT and FGF9Eks to transduce signals via FGFRs was analyzed by a mitogenic assay using BaF3 cells expressing specific FGFRs as described previously17. 5,000 cells were plated per well in a 96-well assay plate in media containing varying concentrations of FGF9 and heparin (Wako). FGF9WT or FGF9Eks were added to each well for a total volume of 200 µl per well. The cells were then incubated at 37°C for 36 hours. 1 µCi of [3H]thymidine was added to each well in 20 µl of media. The cells were harvested after 4 hours by filtration through glass fiber paper and the incorporated [3H]thymidine was counted on a Wallac MicroBeta TriLux scintillation counter (PerkinElmer).
Three mg of purified FGF9WT and FGF9Eks were loaded onto a 1 ml HiTrap heparin HP column (GE healthcare) equilibrated with 25mM ammonium acetate buffer (pH5.5) containing 120 mM NaCl. The bound FGF9WT or FGF9Eks were eluted with a linear gradient of NaCl (120 mM to 2.0 M) in the same buffer.
Surface plasmon resonance analysis for measurements of the interactions of FGF9WT-heparin and FGF9Eks-heparin were performed using a BIAcore 3000 instrument (Biacore AB). In order to immobilize heparin (Wako) on the streptavidin-conjugated sensor chip SA, 100 µg/ml biotinylated heparin in HBS-EP buffer was injected at a flow rate of 10 µl/min and was immobilized to 63 response units (RU). All measurements were carried out at room temperature, and refractive index errors due to bulk solvent effects were corrected by subtracting responses on the non-coated sensor chip for the FGF9WT and FGF9Eks concentrations used. To obtain kinetic data, different concentrations of analytes (FGF9WT and FGF9Eks) in HBS-EP were injected over the heparin sensor chip at a flow rate of 20 µl/min. At the end of each sample injection (120 s), HBS-EP buffer was passed over the sensor surface to monitor the dissociation phase. Following 120 s of dissociation, the sensor surface was regenerated by injection of 5 µl of 1 M NaCl in HBS-EP. Five different analyte concentrations were used to determine the kinetic parameters for each interaction. Kinetic parameters were obtained by global fitting of the sensorgrams to a 1:1 model using BIAevaluation software.
Starting structures of monomeric FGF9WT, dimeric FGF9WT, monomeric FGF9WT-heparin, dimeric FGF9WT-heparin, and those of FGF9Eks for MD simulations were taken from the PDB (PDB ID: 1IHK)14. The structures of monomeric FGF9Eks, dimeric FGF9Eks and heterodimeric FGF9WT/Eks were constructed based on FGF9WT using molecular modeling software MOE (Chemical Computing Group, Inc.). A hexasasaccharide (UAP-SGN-IDU-SGN-IDU-SGN) is used as a heparin oligosaccharide. UAP is 1,4-dideoxy-5-dehydro-O2-sulfo-glucuronic acid, SGN is O6-disulfo-glucosamin and IDU is 1,4-dideoxy-O2-sulfo-glucuronic acid. For monomeric FGF9WT-heparin and dimeric FGF9WT-heparin simulations, heparin oligosaccharide was bound to FGF9WT structures obtained from MD simulations using the molecular docking program GOLD (version 3.0)37. In the docking protocol, the standard default setting of GA parameters was used and GoldScore was used as the scoring function. The structures of monomeric FGF9Eks-heparin and dimeric FGF9Eks-heparin were built in the same manner as FGF9WT-heparin complexes. All the starting structures for MD simulations were surrounded by TIP3P water molecules38 spherically. After energy minimizations, all MD simulations were carried out for 10 ns at 300 K using modified Amber 8.0 (ref. 39) for MDGRAPE3 system40. The Amber ff03 force field41 was adopted, and the simulation time step was set at 1 s. The binding free energies were calculated by the molecular mechanics Poisson-Boltzmann/surface area (MM-PBSA) method42 utilizing the last 2 ns MD trajectories.
Mouse Fgf9WT and Fgf9Eks cDNAs were cloned into the RCASBP(A) vector43. The virus solutions were injected into the hind limb bud of chicken embryos at Hamburger-Hamilton (HH) stage 17. The expression of mouse Fgf9 transcripts and skeletal changes were examined 2 and 5 days after injection, respectively.
AffiGel-Blue beads (BioRad) soaked in 100 µg/ml FGF9WT or FGF9Eks were implanted onto E15 mouse skulls by ex utero surgery as previously described24 Operated heads were collected 24 hours later and Spp1 transcripts were detected by whole-mount in situ hybridization. The area of Spp1 expression was measured using NIH image software.
AffiGel-Blue beads that had been soaked in 500 µg/ml FGF9WT or FGF9Eks were implanted into the dorsal and central region of E10.5 Fgf9−/− forelimb buds. Limb buds were subsequently cultured for 2 hours on Transwell filters (Costar, Coaning) in serum-free medium [BGJb, 2 mg/ml BSA, penicillin (50 units/ml), streptomycin (50 µg/ml)], in a humid, 37°C, and 5% CO2 environment. Explants were fixed in 4% paraformaldehyde and embedded in paraffin. Sections through the equator of the bead were analyzed for exogenous FGF9 using goat anti-human FGF9 antibody (R&D Systems) and a cell and tissue staining kit HRP-AEC system (R&D Systems). The signal area and intensity were analyzed using NIH image software.
cDNA fragments encoding the full length mouse FGF9WT and FGF9Eks proteins were cloned into the p3xFLAG-CMV-14 vector (Sigma-Aldrich) and into the pCMV-Tag3B vector (Stratagene) to allow expression of FGF9 proteins with either C-terminal Myc or 3xFLAG tags. These vectors were transfected into COS7 cells, and 48 h later, culture supernatants were incubated with anti-FLAG M2 affinity beads (Sigma-Aldrich) for 1h at RT and washed three times with PBS, and then subjected to Western blots with anti-FLAG M2 antibody (Sigma-Aldrich) or anti-Myc antibody (Upstate).
This study was supported in part by the RIKEN Structural Genomics/Proteomics Initiative (RSGI) and the National Project on Protein Structural and Functional Analysis, Ministry of Education, Culture, Sports, Science and Technology of Japan (S. Y.) and NIH grant HD049808 (D. M. O.).