FGF9 regulates early hypertrophic chondrocyte differentiation and skeletal vascularization in the developing stylopod

Abstract

Gain-of-function mutations in fibroblast growth factor (FGF) receptors result in chondrodysplasia and craniosynostosis syndromes, highlighting the critical role for FGF signaling in skeletal development. Although the FGFRs involved in skeletal development have been well characterized, only a single FGF ligand, FGF18, has been identified that regulates skeletal development during embryogenesis. Here we identify Fgf9 as a second FGF ligand that is critical for skeletal development. We show that Fgf9 is expressed in the proximity of developing skeletal elements and that Fgf9-deficient mice exhibit rhizomelia (a disproportionate shortening of proximal skeletal elements), which is a prominent feature of patients with FGFR3-induced chondrodysplasia syndromes. Although Fgf9 is expressed in the apical ectodermal ridge in the limb bud, we demonstrate that the Fgf9^−/− limb phenotype results from loss of FGF9 functions after formation of the mesenchymal condensation. In developing stylopod elements, FGF9 promotes chondrocyte hypertrophy at early stages and regulates vascularization of the growth plate and osteogenesis at later stages of skeletal development.

Introduction

Fibroblast growth factor (FGF) signaling is essential for vertebrate limb development, from the initial formation and outgrowth of the limb bud to patterning and growth of the skeletal elements (Mariani and Martin, 2003; Min et al., 1998; Niswander, 2003; Ohuchi et al., 1997). The early phases of limb induction require the expression of Fgf10 in lateral plate mesoderm, which activates FGFR2b in the overlying epithelium. Subsequently, Fgf8 is expressed in the apical ectodermal ridge (AER) along with Fgfs 4, 9 and 17 (Colvin et al., 1999; Martin, 1998; Maruoka et al., 1998). Signals emanating from this specialized ectoderm are required for proximal-to-distal limb outgrowth, as well as maintenance of mesenchymal gene expression in the limb bud. Previous studies have shown that conditional inactivation of Fgf8 (Fgf8^cko) in the AER at early stages results in hypoplasia or agenesis of proximal limb elements as well as less severe abnormalities in the intermediate and distal elements (Lewandoski et al., 2000; Moon and Capecchi, 2000). Conditional inactivation of both Fgfs 4 and 8 (Fgf4/8^cko) in the AER ultimately results in limb agenesis; however, small limb buds do form initially. The defects in the proximal skeletal elements of the Fgf8^cko and Fgf4/8^cko animals are thought to result from the dramatic increase in apoptosis observed in the proximal limb bud mesenchyme (Boulet et al., 2004; Sun et al., 2002).

After limb bud initiation and outgrowth, development of the appendicular skeleton occurs via endochondral ossification, with cartilage anlagen prefiguring the future bone (Erlebacher et al., 1995; Karsenty and Wagner, 2002). Following formation of mesenchymal condensations, cells within these condensations differentiate into proliferating chondrocytes which elaborate an extracellular matrix rich in type II collagen, while cells at the periphery differentiate into the surrounding perichondrium. As the cartilaginous template elongates, chondrocytes within the anlagen mature and hypertrophy, secreting a matrix rich in type X collagen. Concomitantly, skeletal vascularization is stimulated by angiogenic growth factors such as vascular endothelial growth factors (VEGF) (Gerber et al., 1999; Zelzer and Olsen, 2005) produced in the perichondrium and hypertrophic cartilage. These changes allow perichondrial angiogenesis to occur and, subsequently, invasion of blood vessels, osteoblasts, and osteoclasts into the avascular cartilaginous template to form the primary ossification center. In a separate, but related, process, osteoprogenitor cells in the perichondrium differentiate into osteoblasts to form cortical bone. These events: chondrogenesis, vascularization and osteogenesis, are coordinated by multiple signaling molecules including IHH, PTHrP, VEGF and FGF to regulate skeletogenesis (de Crombrugghe et al., 2001; Karsenty and Wagner, 2002; Ornitz and Marie, 2002; Zelzer and Olsen, 2005).

Multiple human skeletal dysplasia syndromes result from activating mutations in FGF receptors, highlighting the critical role for FGF signaling in bone development. For example, chondrodysplasias such as achondroplasia, hypochondroplasia and thanatophoric dysplasia are due to mutations in FGFR3 (Bellus et al., 1995; Rousseau et al., 1994; Shiang et al., 1994; Tavormina et al., 1995). The limb dysmorphology of these patients is characterized by rhizomelic shortening, with proximal elements affected more severely than intermediate and distal elements. Various craniosynostosis syndromes are caused by activating mutations in FGFR1 – 3 and result in premature fusion of the cranial sutures (Britto et al., 2001; Cohen, 2000; Morriss-Kay and Wilkie, 2005; Wilkie, 1997, 2000). Afflicted patients also exhibit variable abnormalities in limb development in addition to their calvarial deformities. Despite the identification and functional characterization of the FGFRs involved in skeletogenesis, to date only two of the 22 known FGFs, FGF2 and FGF18, have been shown to have functional roles in skeletal development.

Fgf18 is expressed in the perichondrium and joint spaces and regulates early stages of cartilage development by promoting chondrocyte proliferation and differentiation while, at later stages, FGF18 functions to inhibit chondrogenesis by signaling to FGFR3. Additionally, FGF18 promotes skeletal vascularization by inducing Vegf expression and positively regulates osteogenesis by signaling to FGFR 1 and 2 in the developing perichondrium/periosteum (Liu et al., 2002, 2007; Ohbayashi et al., 2002). Besides Fgf18, expression of Fgfs 7, 8 and 17 has also been observed in the perichondrium; however, gene targeting experiments have not elucidated roles for these ligands in skeletal development (Finch et al., 1995; Mason et al., 1994; Xu et al., 1999). Fgf2 expression has been demonstrated in proliferating and prehypertrophic chondrocytes, periosteal cells and osteoblasts (Hurley et al., 1994, 1999; Sabbieti et al., 1999, 2005; Sullivan and Klagsbrun, 1985). Mice lacking Fgf2 have defects in osteogenesis, which result in aberrant trabecular bone formation (Montero et al., 2000). The identification of additional endogenous FGF ligand(s) for FGFRs expressed in bone has been problematic, likely due to functional redundancy between ligands or, alternatively, activity restricted to specific skeletal elements or to limited developmental stages.

Fgf9 is expressed in the AER of the developing limb bud and in migrating myoblasts (Colvin et al., 1999). Fgf9 transcripts have also been observed surrounding and within precursor skeletal elements at E12.5 and E14.5 (Garofalo et al., 1999). Here we show Fgf9 expression localized to the perichondrium/periosteum, trabecular bone, muscle and loose mesenchyme surrounding developing bone. Mice lacking Fgf9 (Fgf9^−/−) exhibit rhizomelia, a prominent limb patterning defect observed in patients afflicted with achondroplasia. In Fgf9-deficient mice, the hindlimb femora are affected more severely than the forelimb humeri. The Fgf9^−/− limb phenotype can be interpreted as a consequence of loss of Fgf9 expression in the AER resulting in a skeletal patterning defect that primarily affects mesenchymal condensations, or loss of Fgf9 expression at later stages of skeletogenesis directly affecting chondrogenesis, osteogenesis and/or vascularization of proximal skeletal elements. Here we demonstrate that Fgf9^−/− mice have normal limb bud development and mesenchymal condensations, but in stylopod elements, decreased chondrocyte proliferation, delayed initiation of chondrocyte hypertrophy and abnormal osteogenesis secondary to defects in skeletal vascularization.

Materials and methods

Skeletal preparations and bone morphometric analysis

Skeletons were prepared as described previously (Colvin et al., 1996) using Alizarin red S and Alcian blue (Sigma) staining to visualize mineralized bone and cartilage, respectively. Lengths of individual bones were measured at E18.5 using Canvas X (ACD Systems) software.

Generation of mice

Fgf9^−/− and Fgfr3^−/− mice were generated as described (Colvin et al., 1996; Colvin et al., 2001). Conditional alleles for FGFR1 and FGFR2 have been previously described (Pirvola et al., 2002; Yu et al., 2003). Mice homozygous for conditional alleles of both FGFR1 and FGFR2 (Fgfr1^flox/flox;Fgfr2^flox/flox) were crossed with mice carrying one copy of Dermo-1-cre (Yu et al., 2003) and heterozygous for null alleles of FGFR1 and FGFR2 (Fgfr1^Δ/+;Fgfr2^Δ/+;cre/+) to obtain litters containing wild-type, Fgfr1^cko and Fgfr2^cko embryos. Note that Fgfr1 and Fgfr2 single CKO embryos are also missing one functional allele of FGFR2 and FGFR1, respectively.

Histological analysis, immunohistochemistry and in situ hybridization

Tissues were fixed in 4% paraformaldehyde. Histomorphometry was carried out on hematoxylin and eosin (H&E)-stained 5 µm paraffin sections using Axio Vision 3.0 software (Zeiss). Hypertrophic zone heights were measured along the midline of the distal femoral and proximal tibial growth plates at E17.5 and E18.5. At least 3 sections were measured for each embryo. Mineralized bone was visualized by von Kossa staining with methyl green counterstaining. For PECAM (CD31) immunohistochemistry, embryos were fixed in 4% paraformaldehyde followed by 20% sucrose infiltration. Tissues were embedded in OCT (Tissue-Tek®) and 10-µm cryostat sections were made. Sections were fixed in 0.2% glutaraldehyde prior to peroxidase blocking with 3% H₂O₂ in methanol. The primary antibody was anti-PECAM (BD PharMingen). Secondary antibody was biotinylated anti-rat IgG (BD PharMingen). Immunoreactivity detection was performed using a streptavidin–biotin–peroxidase complex (sABC-HRP, DakoCytomation K0377) and DAB (Zymed). Sections were counterstained with hematoxylin. Whole-mount in situ hybridization was performed according to standard protocol. Radioactive in situ hybridization was performed as described (Naski et al., 1998).

Analysis of cell proliferation and apoptosis

Anti-bromodeoxyuridine (BrdU) immunohistochemistry was carried out as described (Naski et al., 1998). BrdU-positive nuclei of reserve and proliferating chondrocytes were counted and the area of the reserve and proliferating chondrocyte zones was measured using ImageJ 1.36b software (NIH). The number of BrdU-positive nuclei per 0.01 mm² area was calculated for each embryo examined. At least three sections were analyzed through a 10× objective for each embryo. TUNEL assay was performed using the In Situ Cell Death Detection Kit, POD (Roche) per manufacturer’s instructions.

Embryonic limb explant cultures

Forelimb cartilage was dissected from E14.5 VEGF-LacZ mouse embryos (Miquerol et al., 1999) and placed at the air-fluid interface on Transwell filters (Corning) containing DMEM with 10% fetal calf serum (GibcoBRL), antibiotic antimycotic solution (Sigma) and 2 µg/ml heparin. Media were supplemented with 250 ng/ml recombinant FGF10 (PeproTech), recombinant FGF9 (Pepro-Tech) or BSA. Explants were cultured for 2 days at 37 °C/5% CO₂ under humidified conditions.

Limb explant histochemistry and immunohistochemistry

Whole-mount β-galactosidase staining was performed as described (Liu et al., 2007). After post-fixing in 4%paraformaldehyde, the explants were infiltrated with 20% sucrose and embedded in OCT at −20 °C for sectioning. Frozen sections (8 µm) were mounted in toluene-based media. The contralateral limb explant was used for PECAM whole-mount immunohistochemistry as described (Lavine et al., 2006). For histological analysis, 5-µm paraffin sections of stained explants were counterstained with hematoxylin.

Results

Loss of Fgf9 affects the proximal skeletal elements of the developing limb

Comparison of Alizarin red- and Alcian blue-stained skeletons of E18.5 and P0 mice showed that the skeletons of Fgf9^−/− mice were approximately 10–15% smaller than littermate controls. Although the general patterning of the Fgf9^−/− skeletal elements resembled that of control animals, the Fgf9^−/− hind-limb proximal elements were disproportionately shorter, while the intermediate and distal elements of the limbs appeared unaffected (Figs. 1A, B). Fgf9^−/− mice also lacked the third tro-chanter (Fig. 1C), normally located on the lateral aspect of the femur (Bateman, 1954). Additionally, closer examination of the Fgf9^−/− humerus revealed an enlarged deltoid tuberosity (Fig. 1A). Thus, bone growth and patterning defects were observed in the stylopod elements of Fgf9^−/− mice. In comparison with controls, the Alizarin red-stained region of the Fgf9^−/− femur appeared smaller, suggesting that bony mineralization was decreased.

Fig. 1.

Proximal limb development in Fgf9^−/− mice. (A, B) Skeletal preparations of control and Fgf9^−/− forelimbs (A) and hindlimbs (B) at P0. Note enlarged deltoid tuberosity (arrows) in the Fgf9^−/− humerus. The Fgf9^−/− femur is significantly shorter than control with a smaller region of Alizarin red staining. The third trochanter is missing on the lateral aspect of the femur (arrows in panel B). (C) Higher magnification of femora from (B) to show absence of third trochanter (arrow) in Fgf9^−/− mice. (D) Comparison of femur to tibia length ratios between control and Fgf9^−/− hindlimbs shows statistically significant reduction in length ratio for Fgf9^−/− limb. *Student’s T test, n=3, p=0.013.

Morphometric analyses were performed on limb skeletal preparations of Fgf9^−/− and littermate controls at E18.5. In the forelimb, the mutant humerus length was not significantly shorter (92% of control, n=3, p=0.06). In the hindlimb, the mutant femur length was significantly decreased (82% of control, n=3, p=0.008). Because the entire skeleton of Fgf9^−/− mice was 10–15% smaller than that of control animals, we compared ratios of limb element lengths to determine whether the proximal elements of the limbs were disproportionately reduced compared to the intermediate limb compartment. In the forelimb, the ratio of humerus to radius or ulna length did not reveal a significant difference between Fgf9^−/− and control mice (data not shown). However, the femur to tibia ratio showed a significantly smaller hindlimb stylopod in Fgf9^−/− mice compared to control (control femur/tibia (F/T) ratio=0.96±0.03; Fgf9^−/− F/T ratio=0.85 ± 0.04; p=0.013, n=3) (Fig. 1D).

Fgf9 expression during limb development

Previous studies have localized Fgf9 expression to the AER of the developing limb bud at E10.5 and to skeletal myoblasts at E10.5–E12.5 (Colvin et al., 1999). Fgf9 expression has also been observed at later stages during skeletal development in regions corresponding to mesenchymal condensations in the limb at E12.5 and in mesenchyme surrounding rib condensations at E14.5 (Garofalo et al., 1999). Our examination of spatial and temporal patterns of Fgf9 expression in the limb showed that, at E12.5, Fgf9 was expressed in the mesenchyme surrounding the cartilaginous condensations; however, its expression was excluded from the condensations (Figs. 2A, B). At later stages of limb development, Fgf9 was expressed in the perichondrium/periosteum, as well as in surrounding skeletal muscle and overlying skin (Figs. 2C–F). Low levels of Fgf9 mRNA were also detectable in the primary spongiosa at this stage. This expression pattern suggests that Fgf9 could regulate multiple stages of skeletogenesis since, like Fgf18 (Liu et al., 2002; Ohbayashi et al., 2002), this ligand is juxtaposed to Fgfr3-expressing proliferating chondrocytes, Fgfr1-expressing hypertrophic chondrocytes and Fgfr1 and Fgfr2-expressing cells in the perichondrium and periosteum (Naski and Ornitz, 1998; Orr-Urtreger et al., 1991; Peters et al., 1992, 1993). Additionally, Fgf9 expression in the skeletal musculature surrounding bone could serve as an additional source of ligand that may also impact skeletal development.

Fig. 2.

Fgf9 expression patterns in the developing limb. (A) Longitudinal section through the hindlimb of an E12.5 embryo stained with hematoxylin and eosin (H&E). (B) Fgf9 in situ hybridization of a nearby section to that shown in panel A. Note Fgf9 transcripts are excluded from the condensations (f—femur, t—tibia, a—autopod). (C) H&E staining of longitudinal section through femur and proximal tibia at E16.5. (D) Nearby section to panel C showing Fgf9 expression in the perichondrium/periosteum (po), trabecular bone (tb), surrounding muscle (m) and skin (s) but not in proliferating (p) or hypertrophic (h) chondrocytes. (E) H&E staining of a longitudinal section through the ulna (top) and radius (bottom) at E15.5. (F) Nearby section to panel E showing Fgf9 expression in the perichondrium/periosteum (po) and skin (s).

Loss of Fgf9 from the AER does not cause limb patterning defects

Transgenic mice with reduced AER-FGF signaling due to a conditional loss of Fgf8 (Fgf8^cko) or a combination of Fgfs 4 and 8 (Fgf4/8^cko) from the AER have shortened or absent proxifmal limb elements (Boulet et al., 2004; Lewandoski et al., 2000; Moon and Capecchi, 2000; Sun et al., 2002). These mice also had smaller limb bud sizes and decreased expression of mesenchymal limb bud genes. Although the Fgf9^−/− mice have stylopod element defects similar to those observed in the Fgf8^cko and Fgf4/8^cko mice, the Fgf9^−/− limb buds appeared normal in size between E9.5 and E11.5 (Figs. 3A–F and data not shown). To determine whether any molecular changes were detectable in Fgf9^−/− limb buds, the expression patterns of Fgf8 and Shh were examined by whole-mount in situ hybridization. Similar levels of Fgf8 expression were observed in Fgf9^−/− and control limb buds (Figs. 3A, B). Shh (important for anterior–posterior (AP) patterning and AER maintenance) expression in the posterior limb mesenchyme requires AER-FGF signaling (Laufer et al., 1994; Niswander et al., 1994; Zuniga et al., 1999). Shh expression levels appeared slightly decreased and diffuse in Fgf9^−/− limb buds compared to controls (Figs. 3C, D). In the Fgf4/8^cko embryos, increased apoptosis in proximal limb mesenchyme was observed, which was hypothesized to result in a smaller proximal mesenchymal condensation and consequently smaller proximal skeletal elements. At E10.5 and E11.5, we observed similar levels of apoptosis in the proximal limb bud and in the AER of control and Fgf9^−/− mice (Figs. 3E, F and data not shown). Together, these data suggest that AER-FGF function is preserved in Fgf9^−/− mice and that FGFs 4, 8 and 17 can functionally compensate for any loss of FGF9 signaling.

Fig. 3.

Loss of Fgf9 from the AER does not affect limb bud development. (A, B) Fgf8 expression in the AER of E10.5 control (A) and Fgf9^−/− (B) hindlimb buds visualized by whole-mount in situ hybridization (WISH). (C, D) Shh expression in the AER of E10.5 control (C) and Fgf9^−/− (D) hindlimb buds visualized by WISH. (A–D) Dorsal view, anterior to the left, posterior to the right. (E, F) TUNEL staining on transverse sections through E11.5 embryos at the level of the hindlimb bud showing similar levels of apoptosis in control (E) and Fgf9^−/− (F) hindlimbs; dorsal at the bottom, proximal to the left. (G, H) Control (G) and Fgf9^−/− (H) H&E-stained E12.5 longitudinal hindlimb sections showing similar condensation sizes for proximal and intermediate limb elements. z, zeugopod; s, stylopod. (I, J) Sox9 in situ hybridization of nearby sections to panels G and H. (K, L) ColII in situ hybridization of nearby sections to panels G and H. (I–L) Dark-field images.

Defects in AER-FGF signaling resulted in reduced proximal condensation sizes during early limb development in Fgf8^cko and Fgf4/8^cko mice. However, in Fgf9^−/− mice, histological sections revealed that mesenchymal condensations corresponding to proximal, intermediate and distal skeletal elements appeared normal in size and shape at E12.5 (Figs. 3G, H). Although Shh expression was slightly decreased in the Fgf9^−/− limb bud, Fgf9^−/− autopod elements were indistinguishable from controls, suggesting that normal SHH signaling was present. Additionally, Sox9, a marker of prechondrogenic condensations, was expressed at control levels in Fgf9^−/− limb condensations (Figs. 3I, J). Levels of type II collagen (ColII) expression in Fgf9^−/− mice appeared similar to controls (Figs. 3K, L), indicating that mesenchymal cells within the condensations were undergoing chondrogenic differentiation. These data demonstrate that mesenchymal aggregation and condensation, as well as early stages of chondrogenic differentiation, are not significantly altered during the initial stages of limb development in Fgf9^−/− mice and suggest that loss of Fgf9 from the AER does not cause the observed rhizomelic phenotype.

Delayed skeletogenesis in Fgf9^−/− proximal limb elements

Histological analyses at later stages of limb development revealed delayed endochondral ossification of fore- and hindlimb stylopod elements in mice lacking Fgf9 (Figs. 4A–D, G–J). Normally, at E15.5, the primary ossification center in the humerus has formed; however, at this stage, the diaphysis of the Fgf9^−/− humerus contained predominantly hypertrophic chondrocytes with a narrow zone of vascularization (Figs. 4A, B). In the hindlimb at E15.5, blood vessels were visible in the control femoral diaphysis, but the Fgf9^−/− femur remained avascular with relatively few hypertrophic chondrocytes (Figs. 4C, D). By E16.5 the primary ossification center had formed in the control femur, while the shaft of the Fgf9^−/− bone contained an expanded zone of hypertrophic chondrocytes with a small focus of vascularization and absent trabecular bone formation (Figs. 4G, H). By E18.5 the ossification center in the Fgf9^−/− femur was well-developed with abundant blood vessels, although the trabecular bone region was reduced in size (Figs. 4I, J). Delayed bone formation was not apparent in the intermediate and distal elements of the Fgf9^−/− limbs (Figs. 4E, F and data not shown).

Fig. 4.

Histological analysis of Fgf9^−/− hindlimbs. (A–F) H&E sections of E15.5 humeri (A, B), femora (C, D) and tibiae (E, F). (A, C, E) control, (B, D, F) Fgf9^−/−. Note reduced area of vascular invasion in Fgf9^−/− humerus. In Fgf9^−/− femur, the hypertrophic chondrocyte zone is shortened and no vascular invasion of the cartilage is visible. Development of the Fgf9^−/− tibia is similar to the control indicating no developmental delay. (G, H) H&E staining of E16.5 control (G) and Fgf9^−/− (H) femora. Note that the primary ossification center has developed in the control bone while only the initial stages of vascular invasion are apparent in Fgf9^−/− bone. (I, J) H&E sections of E18.5 control (I) and Fgf9^−/− (J) femora. Ossification centers are now well-developed in control and Fgf9^−/− bones, but the Fgf9^−/− femur is shorter and broader with decreased areas of trabecular bone formation and an enlarged hypertrophic chondrocyte zone. Trabecular bone regions are indicated with green brackets in panels G–J. (K–N) H&E sections of E14.5–E14.75 control (K), Fgfr1^cko (L), Fgfr2^cko (M) and Fgfr3^−/− (N) femora showing a smaller hypertrophic chondrocyte zone in tissue missing Fgfr1 but not Fgfr2 or Fgfr3.

In vitro studies have shown that FGF9 can activate the c splice forms of FGFRs 1, 2 and 3, FGFR4, and the b splice form of FGFR3 (Ornitz et al., 1996; Zhang et al., 2006). In developing bone, Fgfrs 1 and 2 are expressed in the mesenchymal condensation, immature and reserve zone chondrocytes, the perichondrium/periosteum and various stages of osteoblast development (Jacob et al., 2006; Ornitz and Marie, 2002; Yu et al., 2003). Fgfr1 is also expressed in hypertrophic chondrocytes and Fgfr3 is expressed in proliferating chondrocytes (Ornitz and Marie, 2002) and mature osteoblasts (Xiao et al., 2004). Since Fgf9 is widely expressed in the perichondrium/periosteum, trabecular bone, surrounding muscle and soft tissue, it can potentially interact with all three Fgfrs. To ascertain which FGFR(s) could serve as the physiological receptor for FGF9, we compared the histological phenotype from limbs of Fgfr3^−/− mice and from conditional knock out mice lacking Fgfr1 (Fgfr1^cko) or Fgfr2 (Fgfr2^cko) in osteochondroblast lineages with those of Fgf9^−/− mice. These studies revealed that Fgfr1^cko mice exhibited a consistent delay in hypertrophic cartilage differentiation at mid-embryonic stages (Fig. 4L), similar to that observed in the Fgf9^−/− mice, while Fgfr2^cko mice (Fig. 4M) or Fgfr3^−/− mice (Fig. 4N) did not exhibit histomorphological alterations at this stage.

Delayed initiation of chondrocyte hypertrophy in Fgf9^−/− mice

The narrowed zone of hypertrophic chondrocytes observed in the Fgf9^−/− femur at E15.5 (Figs. 4C, D) suggested that chondrocyte maturation might be defective in proximal skeletal elements. To examine chondrogenesis in the mutant limbs, expression studies were performed for ColII, a marker of proliferating chondrocytes, and type X collagen (ColX), a specific marker of hypertrophic chondrocytes. At E14.5, ColII expression was present throughout the control femur, except in the central region of the cartilage, where chondrocytes had undergone hypertrophy. Cells with downregulated ColII expression showed high levels of ColX mRNA (Figs. 5A, C, E). In the Fgf9^−/− femur at the same stage, ColII expression was present throughout the entire cartilage anlagen while ColX expression was completely absent, indicating that chondrocytes had not yet matured to the hypertrophic stage (Figs. 5B, D, F). A decrease in the pool of proliferating chondrocytes available to undergo hypertrophic differentiation could explain the observed delay in chondrocyte maturation. BrdU incorporation studies to assess cell proliferation rates at E13.5 and E14.5 showed a significant reduction in proliferation of Fgf9^−/− femoral chondrocytes at E14.5 compared to control (Table 1). At E13.5, no significant difference in proliferation was detected.

Fig. 5.

Expression of chondrocyte differentiation markers at E14.5 and E17.5. H&E staining of E14.5 (A, B) and E17.5 (G, H) femora. Type II collagen (ColII) expression at E14.5 (C, D) and E17.5 (I, J). Type X collagen (ColX) expression at E14.5 (E, F) and E17.5 (K, L). Hypertrophic zone regions are indicated with green brackets in panels A, G, H. A, C, E, G, I, K are sections from control littermates. Panels B, D, F, H, J, L are sections from Fgf9^−/− mice.

Table 1.

Cell proliferation in E13.5 and E14.5 control and Fgf9^−/− femora

Age	Genotype	na	BrdU+cells/0.01 mm²	P value
E13.5	+/+, +/−	7	34.4±1.7	ns
	−/−	4	35.2±1.6
E14.5	+/+, +/−	6	28.0±2.2	<0.04
	−/−	5	24.5±2.5

ns, not significant.

^an, number of animals examined.

At late stages of embryogenesis (E17.5–18.5), ColII and ColX expression levels were similar between the control and Fgf9^−/− distal femoral growth plates (Figs. 5I–L); however, the hypertrophic zone (HZ) height was significantly enlarged in the Fgf9^−/− femora (Figs. 5G, H, Table 2). In contrast, the hypertrophic zone height in the proximal tibia did not differ significantly between Fgf9^−/− and control mice (Table 2). The expanded hypertrophic chondrocyte zone in the Fgf9^−/− femur could result from increased proliferation, in effect enlarging the pool of precursor chondrocytes available to undergo hypertrophy, a reduction in the rate of loss of terminally differentiated hypertrophic chondrocytes, or a delay in the initiation of chondrocyte loss due to delayed formation of the primary ossification center.

Table 2.

Bone morphometric data

Growth plate	Genotype	na	HZb height (mm)	P value
Distal femur	+/+, +/−	3	0.35±0.02	0.001
	−/−	3	0.49±0.01
Proximal tibia	+/+, +/−	3	0.46±0.03	0.18
	−/−	3	0.50±0.03

^an, number of animals examined at E17.5–E18.5.

^bHypertrophic chondrocyte zone mean height (mm).

Ihh and PTHrP signaling are delayed during early chondrogenesis in Fgf9^−/− mice

Consistent with these findings, the patterns of expression for Ihh, Ptc and Pth1r expression suggested delayed chondrocyte differentiation in the Fgf9^−/− femur at E14.5 (Fig. 6). For example, Ihh is normally expressed in prehypertrophic chondrocytes of developing bone, which flank the centrally located hypertrophic chondrocytes of the control femur in two distinct zones at E14.5. In the Fgf9^−/− cartilage, Ihh is expressed in the prehypertrophic chondrocytes, but these cells are only found in the center of the anlagen (Figs. 6C, D). Similarly, Ptc is normally expressed in proliferating and prehypertrophic chondrocytes and the perichondrium, while its expression is downregulated in hypertrophic chondrocytes. In the control femur at this stage, Ptc expression is evident in the perichondrium as well as in the populations of proliferating and prehypertrophic chondrocytes that flank the centrally located hypertrophic chondrocytes. However, in the Fgf9^−/− femur, expression of Ptc appears diffusely throughout the cartilage anlagen and in the perichondrium, with no central area of downregulation due to the absence of mature hypertrophic chondrocytes (Figs. 6E, F). Expression of Pth1r is barely detectable in the Fgf9^−/− femur at this stage (Figs. 6G, H), while it is easily detected in the control prehypertrophic and hypertrophic chondrocyte populations. By E18.5 (Figs. 6K–P), patterns and levels of Ihh, Ptc and Pth1r expression in Fgf9^−/− mice were similar to controls, indicating that chondrogenesis at later stages of bone development was not significantly affected by loss of FGF9.

Fig. 6.

Expression of Ihh, Ptc and Pth 1 r at E14.5 and E18.5. H&E staining of E14.5 (A, B) and E18.5 (I, J) femora. (C–H, K–P) Dark-field images. Ihh expression at E14.5 (C, D) and E18.5 (K, L). Ptc expression at E14.5 (E, F) and E18.5 (M, N). Pth 1 r expression at E14.5 (G, H) and E18.5 (O, P). Panels A, C, E, G, I, K, M, O are sections from control littermates. Panels B, D, F, H, J, L, N, P are sections from Fgf9^−/− mice.

FGF9 promotes osteogenesis in the stylopod

The reduced area of Alizarin red-staining and delayed formation of the primary ossification centers suggested abnormal osteogenesis in Fgf9^−/− mice. Consistent with our previous findings, von Kossa staining showed abundant mineralization in control bones but significantly decreased mineralization of both trabecular and cortical regions in Fgf9^−/− bones at E16.5 (Figs. 7A, B and data not shown). Defective osteoblastogenesis is one potential cause for delayed bone formation in Fgf9^−/− mice and may be due to a deficiency in osteoblast progenitor cells or a defect in osteoblast proliferation or maturation. To characterize osteoblast development, the expression of osteoblast differentiation markers, Runx2 (Cbfa1), Type I collagen (ColI), and Osteocalcin were examined. Runx2 is expressed in osteoprogenitor cells in the perichondrium and in hypertrophic chondrocytes. Similar levels of expression of Runx2 were detected in control and Fgf9^−/− perichondrium at E14.5, indicating that similar numbers of osteoprogenitor cells were present (Figs. 7E, F). However, ColI and Osteocalcin levels were markedly decreased in the perichondrium/periosteum and trabecular bone regions of Fgf9^−/− mice from E16.5 to E18.5, which suggested that there were fewer mature osteoblasts in the absence of Fgf9 (Figs. 7G, H and data not shown). In vitro osteoblast culture data suggest that FGF9 signaling may directly stimulate osteogenic differentiation and matrix mineralization in a stage-dependent manner (Fakhry et al., 2005; Jacob et al., 2006; Valverde-Franco et al., 2004). Thus, FGF9 may function to regulate osteoblast proliferation or differentiation by signaling directly to FGFRs expressed in osteoblast lineages. Alternatively, FGF9 could control the influx of osteoblasts into the trabecular bone region.

Fig. 7.

Osteoblast and osteoclast markers at E14.5 and E16.5. (A, B) von Kossa-stained E16.5 femora. (C, D) H&E-stained E14.5 femora. (E, F) Runx2/Cbfa1 expression at E14.5. (G, H) ColI expression at E16.5. (I, J) MMP9 expression at E16.5. (K, L) TRAP expression at E16.5. (E–L) Dark-field images. Panels A, C, E, G, I, K are sections from control littermates. Panels B, D, F, H, J, L are sections from Fgf9^−/− mice.

Deficiency of osteoclasts in Fgf9^−/− bone

To analyze the osteoclast populations in control and Fgf9^−/− femurs, in situ hybridization studies were performed with the osteoclast-specific markers, matrix metalloproteinase 9 (MMP9) and tartrate-resistant acid phosphatase (TRAP). At E16.5, MMP9 expression was localized to the cartilage–bone interface as well as to the trabecular region in control bones, whereas MMP9 expression was markedly reduced in the Fgf9^−/− femur, with transcripts detectable only in the vascular invasion front of the developing bone (Figs. 7I, J). TRAP expression at E16.5 and TRAP staining at E17.5–E18.5 were similarly decreased in the Fgf9^−/− femur compared to control (Figs. 7K, L and data not shown). These data indicate that loss of Fgf9 results in a reduction in osteoclast cell populations in the perichondrium and primary spongiosa of developing bone.

Delayed vascular invasion in Fgf9^−/− femurs

Vascularization of the developing skeletal elements is a key step in endochondral ossification, allowing the influx of osteoblast and osteoclast progenitors into the initially avascular cartilaginous matrix surrounding hypertrophic chondrocytes (Colnot et al., 2005; Zelzer and Olsen, 2005). Angiogenesis of developing bone was ascertained by PECAM (CD31) immunohistochemistry. At E16.5, blood vessels were present in the perichondrium/periosteum and in the marrow cavity of control femur, while in Fgf9^−/− mice, blood vessels were clearly visible in the perichondrium surrounding the femur cartilage but had not yet invaded the future marrow cavity (Figs. 8A, B). These data suggest that Fgf9 is important for promoting skeletal vascularization and that the delayed osteogenesis in the proximal limbs of Fgf9^−/− mice may result from defects in angiogenesis.

Fig. 8.

Vascularization and expression of Vegf and Vegf receptors in the developing femur. PECAM (CD31) immunohistochemistry (brown signal) counterstained with hematoxylin showing abundant vascularization in control femur (A) compared with Fgf9^−/− femur (B) at E16.5. (C, D) H&E staining of E14.5 femora. (E, F) Dark-field images showing Vegf expression at E14.5. Note Vegf expression in the center of control cartilage anlagen (arrow in E) and in the perichondrium. In Fgf9^−/− femur, Vegf transcripts are only detectable in the perichondrium (F). (G, H) H&E staining of E16.5 femora. (I–L) Dark-field images showing Vegfr1 (I, J) and Vegfr2 (K, L) expression at E16.5. Panels A, C, E, G, I, K are sections from control littermates. Panels B, D, F, H, J, L are sections from Fgf9^−/− mice.

Decreased VEGF signaling in Fgf9^−/− mice

Vascular endothelial growth factor (VEGF) is essential for blood vessel formation and chondrocyte maturation in the developing skeleton (Maes et al., 2002; Ornitz, 2005; Zelzer et al., 2002; Zelzer and Olsen, 2005). Vegf is expressed in the perichondrium and subsequently in hypertrophic chondrocytes and functions to promote vascularization of the cartilage anlagen. In Fgf9^−/− mice, delayed vascular invasion may be due to decreased VEGF signaling. To test this hypothesis, we examined Vegf expression in control and Fgf9^−/− stylopod elements (Figs. 8C–F). At E14.5, similar levels of Vegf expression were observed in the perichondrium. However, in control bone, Vegf was also expressed centrally in hypertrophic chondrocytes, whereas in Fgf9^−/− mice, there was little Vegf expression in this region. Vegf receptor (Vegfr) expression levels correlate with the intensity of VEGF signaling (Barleon et al., 1997; Gerber et al., 1999; Shen et al., 1998; Zelzer et al., 2002). At E16.5, both Vegfr1 and Vegfr2 expression levels were decreased in the perichondrium and diaphysis of Fgf9^−/− femora compared to control, consistent with decreased Vegf expression in the cartilage at E14.5 and reduced vascular invasion of the Fgf9^−/− hypertrophic zone (Figs. 8I–L).

Fgf9 induces Vegf expression and skeletal vascularization in limb explant cultures

Analyses of Fgf9^−/− limbs suggest that FGF9 promotes vascular development by stimulating VEGF signaling in proximal skeletal elements. To test whether FGF9 was sufficient to induce Vegf expression, E14.5 forelimbs from mice heterozygous for a β-galactosidase-tagged allele of Vegf were cultured with bovine serum albumin (BSA), FGF10 or FGF9. Vegf expression levels were not affected by treatment with BSA or FGF10, which signals to epithelial FGFR2b; however, Vegf expression levels were upregulated in limbs treated with FGF9 (Figs. 9A–C). Cryosections revealed increased Vegf expression in the thickened perichondrium and hypertrophic chondrocytes of FGF9-treated explants (Figs. 9D–F). These data demonstrate that FGF9 is sufficient to induce VEGF in skeletal tissue. Since VEGF is an important stimulator of skeletal angiogenesis, these data support the hypothesis that FGF9 regulates vascularization of developing bone by positively regulating Vegf expression. PECAM immunohistochemistry was performed on contralateral cartilage explants to determine whether treatment with FGF9 could promote vascularization. Explants cultured with FGF10 did not show an increase in vascularization compared with the BSA control. In contrast, explants cultured with FGF9 showed a marked increase in perichondrial vascularization (Figs. 9G–L).

Fig. 9.

FGF9 induces Vegf expression and vascular development in limb explant cultures. E14.5 forelimb cartilage explants from heterozygous Vegf-LacZ embryos were cultured with BSA (A, D, G, J), FGF10 (B, E, H, K) or FGF9 (C, F, I, L). (A–C) staining for β-galactosidase enzymatic activity as a measure of Vegf expression. (D–F) Frozen sections of boxed areas in panels A–C showing increased β-galactosidase activity in the thickened perichondrium (arrow in panel F) and hypertrophic chondrocytes of the explant cultured with FGF9 (asterisk in panel F). (G–I) PECAM immunohistochemistry (brown signal) on contralateral explants to those shown in panels A–C. Note marked increase in perichondrial vascularization of the explant cultured with FGF9 (I). (J–K) Paraffin sections of panels G–H counterstained with hematoxylin to show FGF9-induced perichondrial vascularization and thickening (arrow in panel L).

Discussion

Here we present evidence that FGF9 participates in multiple steps of endochondral ossification to regulate skeletal development in the proximal limb. Mice lacking FGF9 have rhizomelic limb shortening, initiated at the earliest stages of skeletal growth. In Fgf9^−/− mice, this limb deformity is a consequence of loss of Fgf9 expression in the muscle, perichondrium/ periosteum, trabecular bone and loose mesenchyme surrounding the developing bone rather than loss of Fgf9 expression in the AER, since development of limb buds and mesenchymal condensations was normal. Analysis of Fgf9^−/− limbs revealed that, in stylopod elements, loss of FGF9 function resulted in decreased chondrocyte proliferation, delayed initiation of chondrocyte hypertrophy and aberrant formation of mineralized bone secondary to defects in skeletal angiogenesis.

FGF9 signaling regulates early stages of chondrogenesis

In the proximal limb elements of Fgf9^−/− mice, the hypertrophic zone size was decreased at midgestation, suggesting that FGF9 signaling is required for normal chondrocyte maturation. A similar histological phenotype has been observed throughout the skeleton in mice lacking FGF18. In vitro studies have shown that both FGF9 and FGF18 are able to activate all FGFRs expressed in bone (Ornitz et al., 1996; Zhang et al., 2006). To determine which FGFR was transducing the FGF9 signal, growth plate histology was examined from Fgfr3^−/− mice and from mice lacking FGFR1 or FGFR2 in skeletal lineages. Only the Fgfr1^cko mice showed a similar decrease in hyper-trophic zone size at this stage, indicating that FGF9 likely signals through FGFR1 to initiate hypertrophic chondrocyte differentiation during very early stages of chondrogenesis.

The smaller zone of chondrocyte hypertrophy at early stages in Fgf9^−/− mice may be due to decreased chondrocyte proliferation at E14.5, which would result in fewer chondrocytes available to undergo differentiation. Reduced proliferation was also observed in Fgf18^−/− mice (Liu et al., 2007). At earlier stages of chondrogenesis, effects on proliferation in Fgf9^−/− proximal limbs may be partially masked by redundancy with FGF18 (Davidson et al., 2005; Iwata et al., 2000; Liu et al., 2002, 2007; Ohbayashi et al., 2002). In addition to direct regulation of chondrocyte growth and differentiation by FGF signaling, other pathways known to be important in chondro-genesis, such as the IHH and PTHrP signaling pathways, may also modulate chondrocyte proliferation in response to changes in FGF signaling. During early chondrogenesis, IHH regulates chondrocyte maturation through induction of PTHrP (St-Jacques et al., 1999) and also promotes chondrocyte proliferation through a PTHrP-independent pathway (Karp et al., 2000). In the stylopod elements of Fgf9^−/− mice at E14.5, Ihh, Ptc and Pth1r expression levels were decreased, suggesting that FGF9 may also regulate chondrogenesis indirectly through these pathways. Reduced levels of Ihh were also observed at early stages in Fgf18^−/− limbs (Liu et al., 2007). Previous studies showed that FGF18 and FGFR3 inhibited Ihh expression in prehypertrophic chondrocytes at late embryonic and postnatal stages (Liu et al., 2002, 2007; Naski et al., 1998). Our observations of the concomitant decreases in both Ihh expression and chondrocyte proliferation suggest that FGF9 and FGF18 may also regulate chondrocyte proliferation and differentiation indirectly by regulating IHH signaling. In support of this model, Iwata et al. (2000) observed increased chondrocyte proliferation and increased Ptc expression at E15.5 in mice harboring an Fgfr3 gain of function mutation.

FGF9 promotes skeletal vascularization and osteogenesis

In Fgf9^−/− mice, abnormalities in Alizarin red and von Kossa staining were consistent with decreased mineralization of the femora, suggesting that FGF9 positively regulates osteogenesis in the developing stylopod. Defective bone formation was further confirmed by the observation that osteoblast and osteoclast markers were downregulated in Fgf9^−/− mice. These findings may be a result of earlier defects in chondrogenesis, or alternatively FGF9 may directly regulate osteogenesis, as has been demonstrated by in vitro culture studies (Fakhry et al., 2005). Osteoblasts are derived from mesenchymal progenitor cells that reside in the perichondrium, while osteoclasts originate from hematopoietic lineages. Although it is possible that loss of FGF9 signaling results in independent effects on these two distinct cell lineages, a simpler explanation for the reduced numbers of osteoblasts and osteoclasts in the developing bone is that skeletal angiogenesis, which is required for both of these cell populations to gain entry into the avascular cartilaginous tissue, is delayed in the absence of FGF9. In support of this model, PECAM immunohistochemistry showed delayed vascularization of Fgf9^−/− stylopod elements. A delay in the formation of the primary spongiosa could also explain the increased hypertrophic zone size observed at late embryonic stages since accumulation of hypertrophic chondrocytes has been reported in several mouse models with impaired skeletal angiogenesis and is thought to result from delayed removal of terminally differentiated hypertrophic chondrocytes (Gerber et al., 1999; Haigh et al., 2000; Vu et al., 1998).

Blood vessel formation is a complex process which is orchestrated by multiple signaling pathways. Various angiogenic factors, including FGF, VEGF, angiopoietins and IHH, are involved in vascularization of the developing skeleton (Colnot et al., 2005; Liu et al., 2007; Zelzer and Olsen, 2005). Several explanations could account for the delayed vascular invasion of the Fgf9^−/− stylopod growth plates. FGF signaling promotes growth of vascular endothelium (Auguste et al., 2003; Kanda et al., 2004; Seghezzi et al., 1998) and could act directly on endothelial cells to facilitate their invasion of the growth plate. However, conditional inactivation of Fgfr1 and Fgfr2 in endothelial cells did not affect skeletal development (KJL and DMO, unpublished observation), suggesting that FGFs do not signal directly to the endothelial cells in the developing growth plate. FGF signaling can also promote Vegf expression, either directly or indirectly through other signaling pathways (Kanda et al., 2004; Lavine et al., 2006; Seghezzi et al., 1998). Consistent with this possibility, Vegf expression was reduced in the absence of FGF9. Additionally, in forelimb cartilage explant cultures, FGF9 was sufficient to induce Vegf expression and to stimulate cell proliferation and vascular growth within the perichondrium (a prerequisite for invasion of the growth plate). Mice lacking Fgfr3 showed reduced expression of Vegf in their expanded zone of hypertrophic chondrocytes and impaired blood vessel formation at the chondro-osseous junction (Amizuka et al., 2004); however, Fgfr3^−/− mice did not show a delay in the formation of the primary ossification center (Colvin et al., 1996). These data suggest that, during skeletal development, FGF9 signals to FGFR1 and/or FGFR2 in the perichondrium/periosteum to potentiate VEGF signaling and to promote perichondrial angiogenesis. Additionally, FGF9 may signal to FGFR3 in proliferating chondrocytes and/or to FGFR1 in hypertrophic chondrocytes to upregulate VEGF signaling within the growth plate. This source of VEGF would stimulate vascularization of the hypertrophic cartilage and the chondro-osseous junction once the primary ossification center has been established (Zelzer et al., 2002; Zelzer and Olsen, 2005).

Gradient model for FGF9 and FGF18 in the developing limb

Three FGF ligands, FGFs 2, 9 and 18, have identified physiological roles in skeletogenesis. FGF2 signaling affects trabecular bone architecture but not the onset of ossification or vascularization (Montero et al., 2000). Fgf9 and Fgf18 display some overlap in their expression patterns and mice lacking either FGF9 or FGF18 have delayed chondrogenesis, skeletal vascularization and mineralization. However, loss of FGF9 causes a specific defect in the formation of the stylopod elements despite the fact that Fgf9 is expressed throughout the developing limb, while loss of FGF18 causes a more uniform delay, affecting proximal, intermediate and distal compartments of the developing limb (Liu et al., 2007). These differences may occur because there is an increased dosage requirement of FGFs for development of proximal versus distal limb elements or because different FGF ligands exhibit different levels of activity depending on their location. Based on our data, a model can be proposed in which functional gradients of FGF9 and FGF18 activity modulate various aspects of skeletal development along a proximal-distal axis. This model predicts, for example, that a delayed vascularization phenotype observed in the stylopod in Fgf9^−/− embryos may become more severe in the stylopod and extend more distally if one copy of Fgf18 is removed. Alternatively, the Fgf18^−/− phenotype may become more severe if one copy of Fgf9 is removed. Preliminary observations of these genetic crosses suggest that this type of functional redundancy does occur (Fig. 10).

Fig. 10.

Functional gradient model of FGF9 and FGF18 in the developing limb. Levels of FGF9 function vary throughout the limb, with highest levels in the stylopod and diminishing levels in more distal segments. In contrast, FGF18 function remains relatively constant throughout the stylopod, zeugopod and autopod compartments.

Another interesting feature of the Fgf9^−/− phenotype is the disproportionate shortening of the proximal skeletal elements (rhizomelia). Rhizomelia is a prominent deformity seen in patients with achondroplasia who harbor gain of function mutations in Fgfr3. Interestingly, this proximal limb defect has not been observed in mice with activating mutations in Fgfr3 (Chen et al., 1999, 2001; Iwata et al., 2001; Li et al., 1999; Naski et al., 1998; Segev et al., 2000; Wang et al., 1999) but has been observed postnatally in transgenic mice overexpressing Fgf9 in cartilage (Garofalo et al., 2003). Furthermore, in Fgfr3^−/− mice, the limb overgrowth phenotype is only seen beginning at late embryonic stages, whereas the rhizomelia observed in Fgf9^−/− mice occurs at earlier stages. Because signaling through FGFR3 may have a biphasic response during development, it is possible that loss of FGF9 would result in increased proximal limb growth at postnatal stages. If this were the case, specifically inhibiting FGF9 postnatally may provide a therapeutic avenue for the treatment of achondroplasia.