A combined series of Fgf9 and Fgf18 mutant alleles identifies unique and redundant roles in skeletal development

Abstract

Fibroblast growth factor (FGF) signaling is a critical regulator of skeletal development. Fgf9 and Fgf18 are the only FGF ligands with identified functions in embryonic bone growth. Mice lacking Fgf9 or Fgf18 have distinct skeletal phenotypes; however, the extent of overlapping or redundant functions for these ligands and the stage-specific contributions of FGF signaling to chondrogenesis and osteogenesis are not known. To identify separate versus shared roles for FGF9 and FGF18, we generated a combined series of Fgf9 and Fgf18 null alleles. Analysis of embryos lacking alleles of Fgf9 and Fgf18 shows that both encoded ligands function redundantly to control all stages of skeletogenesis; however, they have variable potencies along the proximodistal limb axis, suggesting gradients of activity during formation of the appendicular skeleton. Congenital absence of both Fgf9 and Fgf18 results in a striking osteochondrodysplasia and revealed functions for FGF signaling in early proximal limb chondrogenesis. Additional defects were also noted in craniofacial bones, vertebrae, and ribs. Loss of alleles of Fgf9 and Fgf18 also affect the expression of genes encoding other key intrinsic skeletal regulators, including IHH, PTHLH (PTHrP), and RUNX2, revealing potential direct, indirect, and compensatory mechanisms to coordinate chondrogenesis and osteogenesis.

1. Introduction

Skeletal development in vertebrate organisms occurs via intramembranous ossification with mesenchymal cells differentiating directly to osteogenic cells, and endochondral ossification, with an intermediary cartilaginous template prefiguring the future bone (Erlebacher et al., 1995; Karsenty and Wagner, 2002; Long and Ornitz, 2013). Following mesenchymal cell condensation, chondrogenic differentiation produces immature, type II collagen-expressing chondrocytes, while cells at the periphery differentiate into the surrounding perichondrium, expressing type I collagen. Immature chondrocytes initially separate into a population of round, reserve chondrocytes with low mitotic indices and a population of ovoid, columnar, CyclinD1-positive proliferating chondrocytes with high mitotic indices that will subsequently mature to hypertrophic chondrocytes. Centrally located cells are the first to undergo proliferative and hypertrophic differentiation. During hypertrophic maturation, chondrocytes expand dramatically to drive elongation of the cartilage anlagen and secrete a type X collagen-rich matrix. Trabecular bone and bone marrow replace distal hypertrophic chondrocytes to form the primary ossification center in the bone shaft (Gibson, 1998; Gerber and Ferrara, 2000). Cortical bone is formed separately by osteoblasts derived from osteoprogenitor cells in the perichondrium (Caplan and Pechak, 1987). Chondrogenesis and osteogenesis are tightly regulated by multiple signaling pathways and molecules.

The involvement of FGF pathways in skeletogenesis was realized with the discovery that activating mutations in FGF receptors (FGFRs) caused human skeletal dysplasia and craniosynostosis syndromes, including the most common form of genetic dwarfism, achondroplasia (Rousseau et al., 1994; Shiang et al., 1994; Bellus et al., 1995; Tavormina et al., 1995; Wilkie, 1997; Cohen, 2000; Wilkie, 2000; Britto et al., 2001; Morriss-Kay and Wilkie, 2005; Johnson and Wilkie, 2011). A distinctive feature in achondroplasia is rhizomelia, where the proximal limb is more severely affected than distal elements. FGF-related craniosynostotic patients also exhibit limb abnormalities of variable severity. Normal FGF signaling requires binding of FGF ligands and heparan sulfate chains to FGFRs at the cell surface, inducing receptor dimerization and activation of downstream signaling (Eswarakumar et al., 2005; Ornitz and Itoh, 2015). Thus far, physiologic roles in endochondral bone growth have been demonstrated for only three FGF ligands, FGFs 2, 9 and 18, although others, such as Fgfs 7, 8, and 17, are also expressed in some skeletal elements (Mason et al., 1994; Finch et al., 1995; Xu et al., 1999). The difficulty in identifying additional FGFs has been attributed to potential redundancy between ligands; however, direct evidence of this phenomenon has been lacking in skeletogenesis.

Fgf2 is expressed in differentiated chondrocytes, periosteum, and osteoblasts (Sullivan and Klagsbrun, 1985; Hurley et al., 1994, 1999; Sabbieti et al., 1999, 2005). Though a function for Fgf2 has not been identified in chondrogenesis, Fgf2 null mice have reduced trabecular bone mass, indicating a role in osteogenesis (Montero et al., 2000). Fgf9 and Fgf18 share partially overlapping expression patterns in limb bud mesenchyme, in mesenchyme surrounding condensations, and in the perichondrium (Maruoka et al., 1998; Colvin et al., 1999; Garofalo et al., 1999; Ohuchi et al., 2000; Liu et al., 2002; Hung et al., 2007). Previous studies characterizing Fgf9 and Fgf18 single knockout skeletons identified similar roles for these genes (Liu et al., 2002, 2007; Ohbayashi et al., 2002; Hung et al., 2007). Specifically, the encoded ligands individually promote increased rates of cell division in proliferating chondrocytes and positively regulate initiation of chondrocyte hypertrophy at mid-gestational stages. Each of these ligands possesses unique roles as well. FGF9 activity was restricted to stylopod elements (the most proximal limb compartments such as hindlimb femur) while FGF18 activity was present throughout the developing limb. Additionally, FGF18 functions were biphasic in skeletogenesis: FGF18 negatively regulates chondrocyte proliferation and hypertrophic differentiation at later gestational stages (embryonic day 16.5; E16.5), but positively regulates these processes at earlier time points (E14.5). Opposing functions were not observed for FGF9 at different developmental stages. Biphasic activity of FGFR3 has also been observed in vivo: at E14.5–E15.5, FGFR3 promotes chondrocyte proliferation, while at late embryonic and postnatal stages FGFR3 inhibits chondrocyte proliferation and differentiation (Iwata et al., 2000).

Besides FGF signaling, additional pathways serve as important regulators of endochondral ossification, including Indian hedgehog (IHH), Parathyroid hormone-like peptide (PTHLH), and Runt-related transcription factor 2 (RUNX2) (de Crombrugghe et al., 2001; Karsenty and Wagner, 2002). IHH induces CyclinD1 expression and chondrocyte proliferation, and promotes differentiation of immature chondrocytes to proliferating chondrocytes (Karp et al., 2000; Long et al., 2001). Ihh^−/− mice have shortened limbs due partly to a severe chondrocyte proliferation defect and delayed initiation of chondrocyte maturation (St-Jacques et al., 1999). IHH also induces Pthlh expression, which, with its receptor, Parathyroid hormone 1 receptor (PTH1R), maintains chondrocytes in a proliferative state and inhibits chondrocyte maturation. Disruption of the PTHLH pathway in mice results in short-limbed dwarfism and extensive premature chondrocyte hypertrophy (Karaplis et al., 1994; Lanske et al., 1996). RUNX2 is an essential transcription factor that is required for osteogenesis (Komori et al., 1997; Otto et al., 1997) and functions to promote hypertrophic chondrocyte differentiation (Enomoto et al., 2000; Takeda et al., 2001; Ueta et al., 2001; Stricker et al., 2002). Runx2^−/− embryos exhibit delayed chondrocyte maturation throughout the developing skeleton, and proximal limb elements lacked hypertrophic chondrocytes (Inada et al., 1999; Kim et al., 1999). RUNX2 also influences chondrocyte proliferation and differentiation through direct transcriptional activation of Ihh (Yoshida et al., 2004).

Our past studies showed similar expression sites and suggested overlapping roles for Fgf9 and Fgf18 in skeletal development. Here, we demonstrate profound functional redundancy of FGF9 and FGF18. By generating a series of Fgf9 and Fgf18 mutant alleles, we further characterize their roles in different phases of skeletogenesis and identify differential limb compartment-specific activity levels. In the developing stylopod, FGF9 and FGF18 are required for initial differentiation of chondrocytes from an immature state to CyclinD1-positive, columnar proliferating cells. Together, they promote Ihh and Runx2 expression in E12.5 cartilaginous condensations and are required to maintain expression of these key signaling molecules during midgestation (E14.5). These findings demonstrate important regulatory functions for FGF9 and FGF18 in developing bone.

2. Results

2.1. Loss of Fgf alleles causes a severe osteochondrodysplasia

Viable compound heterozygotes (Fgf9^+/−; Fgf18^+/−) were generated by breeding mice harboring one functional allele of Fgf9 (Fgf9^+/−) (Colvin et al., 2001) with mice harboring one functional allele of Fgf18 (Fgf18^+/−) (Liu et al., 2002). The Fgf9^+/−; Fgf18^+/− mice were subsequently intercrossed to obtain a series of Fgf9 and Fgf18 mutant alleles. All possible genotypes were recovered in Mendelian ratios; however, mice homozygous for Fgf9 and/or Fgf18 null alleles exhibited perinatal lethality, as reported previously (Colvin et al., 2001; Liu et al., 2002; Ohbayashi et al., 2002).

Skeletal preparations at E18.5 were stained with Alcian blue and Alizarin red for gross assessment of the combined roles of Fgf9 and Fgf18 (Figs. 1 and 2). As previously reported, Fgf18-deficient skeletons had delayed mineralization of endochondral and intramembranous bones. Axial skeletal abnormalities included wavy ribs and increased spinal curvatures, which we observed in ~30% of Fgf18^−/− mice (Fig. 1B). Limb skeletal defects seen in all individuals included an angulated radius, a curved ulna, and a bowed tibia. A small unossified fibula was present in ~30% of Fgf18^−/− mice (Fig. 2B). Cleft palate was present in >95% of these animals (not shown). Fgf9-deficient skeletons (Fig. 2D) had 100% penetrant proximal limb anomalies including an enlarged deltoid tuberosity in the humerus and disproportionate shortening of the femur with delayed mineralization (Liu et al., 2002, 2007; Ohbayashi et al., 2002; Hung et al., 2007).

Fig. 1.

Morphological analysis of the skeletal phenotypes in the combined series of Fgf9 and Fgf18 mutant alleles. Skeletal preparations at E18.5 with alcian blue-stained cartilage and alizarin red-stained mineralized bone arranged in order of overall phenotypic severity. (A) Control (Fgf9^+/+; Fgf18^+/− for clarity abbreviated as 9^+/+; 18^+/−). (B-E) Increasing loss of Fgf9 and Fgf18 alleles as indicated. (F) Fgf9/18 double knockout (9^−/− ; 18^−/−) showing a severe osteochondrodysplasia.

Fig. 2.

Evaluation of Fgf mutants’ appendicular skeletons at E18.5 ordered by increasing severity of hindlimb rhizomelia. (A) Control limb skeletons. (B) Fgf18^−/− forelimb with kinked radius (arrow), increased curvature of the distal ulna (arrowhead), bowed tibia (arrow), and shortened fibula (arrowhead). (C) Removal of one allele of Fgf9 in an Fgf18^−/− background worsens the phenotype seen in (B). Additional features in (C) include a small rudimentary fibula (arrowhead) and misshapen autopod elements. (D) Fgf9^−/− forelimb with enlarged deltoid tuberosity (arrow) and hindlimb with shortened femur; note reduced area of alizarin red staining. (E) Removal of one allele of Fgf18 in an Fgf9^−/− background worsens the hindlimb rhizomelia (arrow). Further reduced femur length and area of alizarin red staining in (E) compared to (D); enlarged deltoid tuberosity (arrow). (F) Fgf9^−/− ; Fgf18^−/− rhizomelic forelimb and hindlimb. In the forelimb, boomerang-shaped humerus (arrow), severely angulated radius (arrowhead), v-shaped scapula (asterisk). In the hindlimb, small misshapen pelvic bones, cartilaginous femur rudiment (arrowhead), curved tibia (arrow), unossified fibular rudiment. Hindlimb autopod with severely abnormal rotation. (G) Comparison of femur to tibia length ratios between control and Fgf mutant hindlimbs shows statistically significant reduction in length ratio for Fgf9^−/−, Fgf9^−/−; Fgf18^+/−, Fgf9^−/−; Fgf18^−/− hindlimbs (p<0.05, ANOVA). FL, forelimb and HL, hindlimb.

Skeletons of compound heterozygous mice (Fgf9^+/−; Fgf18^+/−) were similar to controls, although a mild developmental delay in chondrocyte maturation was observed at E14.5 (Fig. S1). However, when one allele of Fgf9 was removed in the context of Fgf18 nullizygosity (Fgf9^+/−; Fgf18^−/−), Fgf18^−/− skeletal phenotypes became 100% penetrant with regard to long bone, rib, and vertebral column abnormalities (Figs. 1B, C and 2B, C), and Alizarin red staining was absent in fibular remnants and autopod phalanges. In the reciprocal experiment, removal of one copy of Fgf18 from Fgf9-deficient mice (Fgf9^−/−; Fgf18^+/−) worsened the rhizomelic hindlimb phenotype and reduced the area of Alizarin red staining (Figs. 1D, E and 2D, E).

Congenital absence of Fgf9 and Fgf18 (Fgf9^−/−; Fgf18^−/−) caused an extreme osteochondrodysplasia that was 100% penetrant (Fig. 1F). Multiple craniofacial defects were apparent, including striking calvarial agenesis with only small, ossified, frontal bones present, maxillary and mandibular hypoplasia, and a wide cleft palate (not shown). This phenotype is consistent with both Fgf9 and Fgf18 expression in cranial mesenchyme (Kim et al., 1998; Ohbayashi et al., 2002; Reinhold and Naski, 2007), and supports a previously unidentified role for Fgf9 in intramembranous as well as endochondral ossification (Fig. 1F) (Fakhry et al., 2005). Patterning and growth of the axial skeleton was also abnormal, including rib, sternal, and vertebral anomalies. Rhizomelic shortening was observed throughout the appendicular skeleton, and hindlimbs were affected more severely than forelimbs (Fig. 2F). Multiple distinctive anomalies included: bifurcated scapula blade, boomerang-shaped humerus, kinked radius, curved ulna, incomplete pelvis, hypoplastic and malformed femur rudiment, tiny fibular remnant, markedly bowed tibia, absent patella, and mal-positioned hindlimb autopods. Long bones (femur, fibula, phalanges) lacked Alizarin red staining at E18.5, indicating increased severity of mineralization defects compared to other skeletal elements. These data confirm partial functional redundancy between Fgf9 and Fgf18.

Morphometric analyses were performed on the hindlimb skeletons from the combined series of Fgf9 and Fgf18 mutant alleles at E18.5 and P0 to quantify the degree of stylopodial shortening (Fig. 2G) by normalizing femur lengths to tibia lengths. In Fgf18^−/− and Fgf9^+/−; Fgf18^−/− animals, the femur/tibia (F/T) ratios did not differ significantly from controls. As previously reported (Hung et al., 2007), the F/T ratio in Fgf9^−/− mice demonstrated a significantly smaller proximal element (control F/T ratio=0.97±0.02; Fgf9^−/− F/T ratio=0.85±0.03; p<0.05, ANOVA). Removal of one copy of Fgf18 in an Fgf9-deficient animal further reduced the F/T ratio (Fgf9^−/−; Fgf18^+/− F/T ratio=0.64±0.03; p<0.01, ANOVA), while loss of all four Fgf alleles distorted the proportions of the limb elements the most dramatically (Fgf9^−/−; Fgf18^−/− F/T ratio=0.34±0.04; p<0.01, ANOVA). Thus, the skeletal phenotypes worsen in a predictable and reproducible manner with increasing loss of Fgf alleles, demonstrating a dosage-dependent effect of these encoded ligands on bone development.

2.2. Regional FGF activity levels in the developing limb

Since our skeletal data suggest significant overlap in functions for FGF9 and FGF18, we reexamined the distribution of ligand activities by comparing histological sections from proximal and distal limbs of the combined series of Fgf9 and Fgf18 mutant alleles. Normally, the femoral primary ossification center has formed by E16.5 (Fig. 3A). Fgf18^−/− femurs (Fig. 3B) have delayed trabecular bone formation (Liu et al., 2002, 2007; Ohbayashi et al., 2002). Removal of one copy of Fgf9 in the Fgf18^−/− background further delayed vascular invasion of the mid-shaft region at this stage (Fig. 3B and C). By E18.5, the marrow cavity was formed, but the trabecular bone length appeared reduced compared to control. Loss of Fgf9 retarded formation of the trabecular bone even more severely (Fig. 3A and D). Deleting one copy of Fgf18 in the Fgf9^−/− background further hindered progression of chondrocyte maturation and osteogenesis. Only a narrow central hypertrophic chondrocyte zone was visible in the E16.5 diaphysis, which remained avascularized until E18.5 (Fig. 3E). These findings demonstrate that FGF9 activity impacts proximal limb bone development to a greater extent than FGF18, since loss of Fgf9 alleles has a more detrimental effect than loss of Fgf18 alleles. In mice lacking four Fgf alleles (Fig. 3F), femoral chondrocytes were predominantly small, round, and disorganized-appearing. Ovoid, columnar proliferating chondrocytes and ballooned, hypertrophic cells were markedly absent. Fgf9^−/−; Fgf18^−/− cartilage histology was remarkably similar between E16.5 and E18.5, indicating stalled chondrogenesis and suggesting that loss of FGF signaling causes a profound and early defect in chondrocyte differentiation.

Fig. 3.

Histological analyses of proximal and distal hindlimb skeletal development in Fgf mutants. (A–F) H&E-stained sections of E16.5 (upper panels) and E18.5 femurs (lower panels). (G–L) E18.5 metatarsal bones. (A,G) wild-type bones with well-established primary ossification centers. (B,H) Fgf18^−/− ossification centers are formed and femoral hypertrophic zones are enlarged compared to control (A,G). (C,I) Fgf9^+/− ; Fgf18^−/− skeletal elements with delayed vascular invasion and enlarged hypertrophic chondrocyte zone in proximal bone (C) and even greater delay in distal hindlimb bone (I) compared to Fgf18^−/− bones (B,H). (D,J) Fgf9^−/− skeletal elements with delayed vascular invasion, enlarged hypertrophic chondrocyte zone, and shortened proximal bone (D) but a normal appearing ossification center in the distal bone (J). (E,K) Fgf9^−/− ; Fgf18^+/− skeletal elements with only a small central zone of chondrocyte hypertrophy in the proximal element at E16.5 and delayed vascularization until E18.5 (E). In Fgf9^−/− ; Fgf18^+/− distal hindlimb bone (K), trabecular bone formation is slightly retarded compared to control (G) but advanced compared to Fgf9^−/− ; Fgf18^+/− proximal bone (E). (F,L) Fgf9^−/− ; Fgf18^−/− skeletal elements show inhibition of chondrocyte differentiation with small round chondrocytes throughout at E16.5 and E18.5 in tiny proximal bone cartilaginous rudiments (F). In distal bone (L), hypertrophic chondrocytes occupy the center of the bone demonstrating advanced development compared to the proximal bone (F). green line: trabecular bone region.

Developmental delays were also observed distally; however, differentiation defects were less severe than those found more proximally. At E18.5, the primary ossification center was established in control metatarsal elements (Fig. 3G). Consistent with previous reports (Hung et al., 2007), loss of Fgf9 did not affect distal limb development (Fig. 3J). In Fgf9^−/−; Fgf18^+/− mice (Fig. 3K), the trabecular bone region appeared slightly shortened compared to Fgf9^−/− mice (Fig. 3J). Fgf18^−/− metatarsal ossification center formation was more severely delayed (Fig. 3H), and additional inactivation of one allele of Fgf9 (Fgf9^+/−; Fgf18^−/−) significantly enhanced the phenotype (Fig. 3I), since the cartilage elements had a large single central zone of hypertrophic chondrocytes with small foci of vascularization at E18.5. These data show that FGF9 activity exists in the distal skeleton but in the presence of wild type levels of FGF18 does not contribute much to distal skeletal growth. The Fgf9^−/−; Fgf18^−/− metatarsals (Fig. 3L) contained a single zone of hypertrophic chondrocytes and remained avascularized at this stage. Although all Fgf9^−/−; Fgf18^−/− limb compartments exhibited delayed bone development, chondrocyte maturation of autopod elements was significantly more advanced than in the stylopod. These data support our previously proposed gradient model that FGF9 activity exists throughout the appendicular skeleton with its highest levels of activity localized proximally, while FGF18 activity levels are more uniform in the various segments (Hung et al., 2007).

2.3. FGF ligands promote chondrocyte maturation in the developing limb

Type II collagen (Col2a1) expression was used to identify reserve and proliferating chondrocytes and type X collagen (Col10a1) expression was used to identify hypertrophic chondrocytes in proximal limb elements. Col2a1 expression was present in all genotypes by E16.5, indicating that chondrogenic differentiation had occurred by this stage (Fig. 4A, E, I, M, Q, U). However, analysis of Col10a1 expression in the combined series of mutant alleles demonstrated genotype-specific deficiencies in chondrocyte maturation.

Fig. 4.

Molecular and histochemical analyses of chondrogenesis and osteogenesis in the combined series of mutant alleles. In situ hybridization on E16.5 femoral longitudinal sections for (A,E,I,M,Q,U) Col2a1, (B,F,J,N,R,V) Col10a1, and (C,G,K,O,S,W) Col1a1 expression. (D,H,L,P,T,X) von Kossa staining (vKos) of (A–D) wild-type, (E–H) Fgf18^−/− , (I–L) Fgf9^+/-; Fgf18^−/− , (M–P) Fgf9^−/− , (Q–T) Fgf9^−/− ; Fgf18^+/-, (U-X) Fgf9^−/− ; Fgf18^−/− femur (arrows).

Although Fgf9 and Fgf18 single knockout mice had similarly delayed initiation of chondrocyte maturation at E14.5 and expanded hypertrophic zones at E16.5–E17.5 (Liu et al., 2002, 2007; Hung et al., 2007), compound Fgf mutants showed varying degrees of genotype-dependent chondrocyte maturation defects. For example, in the Fgf9^+/−; Fgf18^−/− femur, in situ hybridization of Col10a1 demonstrated two distinct zones that were enlarged and located closer together than in controls, indicating delayed development (Fig. 4J). In the Fgf9^−/−; Fgf18^+/− animals, a single zone of Col10a1 expression was visible in a much shorter cartilage anlagen, suggesting increased severity of chondrocyte maturation defects (Fig. 4R). Col10a1 expression was not detectable throughout the Fgf9^−/−; Fgf18^−/− femur (Fig. 4V), indicating that inactivation of both FGFs inhibits development of femoral chondrocytes, at least until E18.5.

Mineralization of hypertrophic chondrocyte matrix and osteogenesis were also impaired by loss of Fgf alleles. For example, in the Fgf9^+/−; Fgf18^−/− femur, reduced mineralization, assayed by von Kossa staining, was evident surrounding terminally differentiated hypertrophic chondrocytes in the bone collar and in the primary ossification center (compare Fig. 4D and L). In the Fgf9^−/−; Fgf18^+/− animals, rare mineralized deposits were located around a few femoral hypertrophic chondrocytes, even though a centralized hypertrophic zone had developed (Fig. 4T). In Fgf9^−/−; Fgf18^−/− stylopod elements, mineralization was absent (Fig. 4X). Osteogenic cell differentiation was also evaluated using type I collagen (Col1a1) as a mature osteoblast marker. Control bones showed high levels of Col1a1 expression in the periosteum, bone collar, and trabecular bone regions at E16.5 (Fig. 4C). In Fgf9^+/−; Fgf18^−/− mice, Col1a1 levels were reduced in the bone collar and not yet detectable within the bone shaft (Fig. 4K). In Fgf9^−/−; Fgf18^+/− (Fig. 4S) and Fgf9^−/−; Fgf18^−/− (Fig. 4W) mice, low levels of Col1a1 expression were present in perichondrial cells, indicating a marked delay or absence of mature osteoblast development. These data suggest that loss of FGF ligands causes severe impairments in chondrocyte maturation and osteogenesis. The bony defects may result from earlier problems in chondrogenesis and/or indicate a separate role for FGF ligands in osteogenic differentiation. To distinguish between these possibilities, we examined the expression of Runx2, a marker of osteoprogenitor cells and osteoblasts, and a mature osteoblast marker, bone sialoprotein (Bsp) (Chen et al., 1992), in intramembranous bone, which forms directly from mesenchymal condensations without a cartilaginous intermediate (Ornitz and Marie, 2002). At E14.5, Runx2 (Fig. S2A and E) and Bsp (Fig. S2B and F) transcript levels were decreased in frontal bones of Fgf9^−/−; Fgf18^−/− skulls compared to control, indicating a likely deficit in osteoprogenitor cell populations and a paucity of functional osteoblasts. At E16.5, very little mineralization of the frontal bone was present compared to control (Fig. S2C, D, G, H). These data demonstrate that FGF9 and FGF18 also function directly to positively regulate osteogenesis.

2.4. Mesenchymal condensation and early chondrogenic differentiation are initiated normally in Fgf mutants

Fgf9 is expressed in the apical ectodermal ridge and in mesenchyme surrounding skeletogenic condensations. Thus, some of the observed phenotypes could derive from earlier developmental events. Previous studies on Fgf9-deficient hindlimbs suggested that Fgf9 did not regulate formation of the mesenchymal condensation (Hung et al., 2007). However, the substantially increased severity of limb phenotypes in the Fgf9^−/−; Fgf18^+/− and Fgf9^−/−; Fgf18^−/− mice indicated that Fgf18 may have partially compensated for loss of Fgf9 since both encoded ligands are expressed in the limb bud (Ohuchi et al., 2000). To examine whether these genes were involved in limb patterning and early chondrogenesis, Sox9 expression studies were performed at E11.5 and E12.5 in control and Fgf9^−/−; Fgf18^−/− hindlimbs (Fig. 5A-D). As shown, Fgf9^−/−; Fgf18^−/− proximal condensations were not disproportionate at these stages, indicating qualitatively normal formation of chondrogenic condensations. At E12.5, levels of Col2a1 in Fgf9^−/−; Fgf18^−/− hindlimbs were also similar to controls, indicating that mesenchymal cells within the condensations were undergoing chondrogenic differentiation (Fig. 5E and F). Additionally, proliferation rates of E12.5 Fgf9^−/−; Fgf18^−/− femoral chondrocytes, assessed by BrdU incorporation studies, were similar to controls (control % BrdU-positive cells=39.3±5.1; Fgf9^−/−; Fgf18^−/− % BrdU-positive cells=38.3±5.0; p=0.67.) These data demonstrate that mesenchymal aggregation and condensation, as well as early stages of chondrogenic differentiation, occurred normally during the initial stages of skeletal development in mice lacking Fgf9 and Fgf18.

Fig. 5.

Formation of mesenchymal condensations and initiation of chondrogenesis is normal in mice lacking Fgf9 and Fgf18. (A,B) Sox9 expression in (A) control and (B) Fgf9^−/− ; Fgf18^−/− hindlimbs at E11.5 visualized by whole-mount in situ hybridization. Note similar size of stylopod condensations (arrows). (C,D) Sox9 in situ hybridization in (C) control and (D) Fgf9^−/− ; Fgf18^−/− E12.5 hindlimb longitudinal sections showing similar condensation sizes for proximal and intermediate limb elements. (E,F) Col2a1 expression of nearby sections to panels C and D. fe, femur and t, tibia.

2.5. FGF signaling promotes differentiation of immature chondrocytes to proliferating chondrocytes

In the Fgf9^−/−; Fgf18^−/− mutants, our histological analyses demonstrated that proximal limb chondrocytes maintained morphological characteristics of immature chondrocytes since they remained small, round, and were not arranged into organized linear columns typical of differentiated, proliferating chondrocytes, even by E18.5 (Fig. 3F). These data indicated that columnar proliferating chondrocytes may be deficient or defective in the Fgf9^−/−; Fgf18^−/− mice. Consistent with this hypothesis, BrdU incorporation studies at E14.5 (Fig. 6A, B, E) showed that cellular proliferation rates in the Fgf9^−/−; Fgf18^−/− distal femur were not significantly different compared to that of reserve zone chondrocytes in the control distal femoral growth plate, but were reduced by 13% compared to control proliferating zone chondrocytes (ANOVA, p<0.001). To determine a molecular basis for the proliferation defect, we analyzed CyclinD1 (CCND1) levels in control and Fgf9^−/−; Fgf18^−/− developing cartilages. This cell cycle regulator promotes G1-S cell cycle transition, is required for chondrocyte proliferation, and is specifically localized to columnar proliferating chondrocytes (and perichondrium) (Beier et al., 2001; Long et al., 2001). At E14.5, nuclear localization of CyclinD1 was observed in the proliferating zone chondrocytes of the control femur, but was markedly absent in the Fgf9^−/−; Fgf18^−/− proximal long bone cartilage, though abundant CyclinD1-staining was visible in the mutant perichondrium (Fig. 6C and D). At E16.5, CyclinD1 levels remained reduced in the Fgf9^−/−; Fgf18^−/− femur (Fig. S3B). Apoptosis assays by TUNEL staining at E14.5 and E16.5 showed that cell death was not increased in mutant limb elements compared to control (data not shown). These data suggest that FGF9 and FGF18 promote differentiation of immature chondrocytes to CyclinD1-positive proliferating chondrocytes and cellular proliferation during endochondral bone growth.

Fig. 6.

Regulation of chondrocyte proliferation and differentiation of immature chondrocytes to CyclinD1-positive columnar cells by FGF signaling. (A,B) Alcian blue-staining and immunohistochemical detection of BrdU-labeled epiphyseal growth plate chondrocytes in (A) control distal femur and (B) Fgf9^−/− ; Fgf18^−/− femur at E14.5. (C,D) CyclinD1 immunohistochemistry to detect columnar proliferating chondrocytes in (C) control distal femoral epiphyseal growth plate and (D) Fgf9^−/− ; Fgf18^−/− femur at E14.5. Note the absence of CyclinD1-positive cells in the center of the Fgf9^−/− ; Fgf18^−/− femoral chondrocyte anlagen, but an abundance of CyclinD1-stained cells in mutant perichondrium. (E) Cell proliferation data at E14.5 in control and Fgf9^−/− ; Fgf18^−/− distal femoral growth plate, * p<0.001, ANOVA. reserve chondrocyte zone (RCZ), proliferating chondrocyte zone (PCZ), distal femur (DF).

2.6. Loss of FGFs attenuates IHH and RUNX2 signaling in the developing limb

Coordination of chondrocyte proliferation and differentiation in the developing growth plate is tightly regulated by several key pathways. Expression studies were carried out to determine whether Fgf9 and Fgf18 interact with IHH-PTHLH and RUNX2 signaling to control chondrogenesis. At E12.5, Ihh mRNA transcripts appear throughout the limb cartilage primordia (Fig. 7A and B). By E14.5, Ihh expression is localized to two distinct zones of prehypertrophic chondrocytes that flank a central domain of matured hypertrophic chondrocytes (Fig. 7I). In E12.5 Fgf9^−/−; Fgf18^−/− hindlimbs, Ihh expression was greatly diminished throughout the proximal and intermediate limb elements, with the femur and fibula affected more than the tibia (Fig. 7E and F). By E14.5, Ihh expression was not detected in the mutant femur, where prehypertrophic chondrocytes failed to develop. However, Ihh expression was still present in the mutant tibial prehypertrophic chondrocytes (Fig. 7M). Expression of the IHH receptor, Patched homolog 1 (Ptch1), serves as a measure of IHH signaling strength and was detected in perichondrium and in immature proliferating chondrocytes of E14.5 control femur (Fig. 7J). In contrast, Ptch1 expression was not detectable in the mutant femoral cartilage at this stage, confirming loss of Hedgehog signaling in the proximal limb (Fig. 7N). IHH interacts with the PTHLH pathway to promote chondrocyte proliferation and to delay hypertrophic differentiation in the developing skeletal elements (Karp et al., 2000). Parathyroid hormone 1 receptor (Pth1r) expression was absent in Fgf9^−/−; Fgf18^−/− femur sections at E14.5 (Fig. 7O), although it was readily detected in the prehypertrophic and hypertrophic chondrocyte populations of mutant tibia (Fig. 7O) and control femur (Fig. 7K). Similarly, Pthlh expression was reduced in the femoral periarticular regions of Fgf9^−/−; Fgf18^−/− hindlimb sections at E14.5 (Fig. S4B), although it was readily detected in the mutant tibial periarticular region.

Fig. 7.

Ihh, Pth1r and Runx2 expression is decreased in Fgf9^−/− ; Fgf18^−/− mutants. (A,E) Ihh and (C,G) Runx2 expression in E12.5 hindlimbs visualized by whole-mount in situ hybridization. (B,F,I,M) Ihh and (D,H,L,P) Runx2 expression in longitudinal sections of E12.5 (B,F,D,H) and E14.5 (I,M,L,P) hindlimbs visualized by in situ hybridization. (J,N) Ptch1, (K,O) Pth1r expression in longitudinal sections of E14.5 hindlimbs. Panels A–D, I–L are from control littermates. Panels E–H, M–P are from Fgf9^−/− ; Fgf18^−/− mice. femur (fe), tibia (t), fibula (fi), femoral elements (arrows).

In Ihh null mice, chondrocytes undergo rapid and premature hypertrophy due to loss of a functional IHH-PTHLH signaling loop. Although Hedgehog signaling is absent, Fgf9^−/−; Fgf18^−/− femoral chondrocytes fail to mature by E18.5, indicating a possible inability to initiate the hypertrophic differentiation program. This suggests that FGF signaling may modulate another gene or pathway that promotes chondrocyte maturation upstream of IHH.

Runx2 is one such gene that plays a critical role in hypertrophic differentiation. Runx2^−/− mice have short limbs that lack Ihh and Pthlh expression as well as hypertrophic chondrocytes in proximal limb skeletal elements (Inada et al., 1999; Kim et al., 1999), similar to the Fgf9^−/−; Fgf18^−/− phenotype. Runx2 is normally detected in centrally located immature chondrocytes and in the perichondrium of the femoral element at E12.5 (Fig. 7C and D); by E14.5 Runx2 transcripts are found in prehypertrophic chondrocytes and maintained in perichondrial cells (Fig. 7L). In the Fgf9^−/−; Fgf18^−/− femur at E12.5 (Fig. 7H), Runx2 expression levels were reduced, but transcripts were appropriately localized to immature chondrocytes and perichondrium. However, by E14.5 Runx2 was absent from the perichondrium and hardly detectable in the immature chondrocytes of the Fgf9^−/−; Fgf18^−/− femur, although it was present in the mutant tibia where hypertrophic chondrocytes did develop (Fig. 7P). Thus, the observed reduced/absent Ihh and Pthlh expression is likely a consequence of loss of RUNX2 in the Fgf9^−/−; Fgf18^−/− mutant.

3. Discussion

Through analysis of the combined series of Fgf9 and Fgf18 mutant alleles, we show that FGF9 and FGF18 participate in several stages of endochondral ossification, including regulation of the onset of columnar chondrocyte differentiation and CyclinD1-dependent cellular proliferation that is not observed with either individual ligand. In addition, we show that these FGFs positively regulate the IHH-PTHLH pathway and Runx2 expression to control chondrogenesis between E12.5 and E14.5.

3.1. Overlapping functions and graded activity levels of FGF9/18 in the developing limb

Analysis of compound Fgf mutants delineates proximal-distal differences between FGF ligands in the appendicular skeleton. FGF9 activity is most pronounced in the developing stylopod, since loss of both Fgf9 alleles causes rhizomelia. However, FGF18 also shares function in this region with FGF9 since removal of one Fgf18 allele worsens the Fgf9 null phenotype. Similarly, the Fgf18 null skeletal phenotype is exacerbated throughout the limb, including the autopod, when one allele of Fgf9 is deleted demonstrating that the domain of FGF9 function is more widespread than previously recognized due to masking of its activity by redundancy with FGF18. The extensive osteochondrodysplasia observed in mice lacking both Fgf9 and Fgf18, compared to mice lacking either gene, is definitive evidence that these encoded ligands have fundamental, partially redundant roles in skeletogenesis.

3.2. FGF ligand-receptor relationships in early chondrogenesis

Loss of Fgfr1 and Fgfr2 in pre-condensation limb mesenchyme results in a severely dysmorphic and underdeveloped appendicular skeleton (Yu and Ornitz, 2008). This phenotype is reminiscent of the extreme short-limbed dwarfism seen in Fgf9^−/−; Fgf18^−/− mice. However, in Fgfr1/2 conditional knockout mice (targeted with Col2a1^Cre or Dermo1^Cre), the skeletal elements are uniformly reduced in length versus disproportionately shortened in the Fgf9^−/−; Fgf18^−/− animals (Yu et al., 2003; Jacob et al., 2006). Additionally, the Fgfr1/2 conditional knockout femur is ossified while in the Fgf9^−/−; Fgf18^−/− proximal limb only a tiny cartilaginous rudiment exists. These phenotypic differences suggest that overlap between FGFR1/2 and FGF9/18 signaling is partial. Possible explanations include a relatively later time of Cre-mediated inactivation of Fgfr1/2 compared to when FGF9 and FGF18 first begin to signal in limb mesenchyme, or, alternatively, that FGF9 and FGF18 signal to an additional FGF receptor, such as FGFR3.

During early developmental stages, Fgfr2 is expressed at high levels in limb mesenchymal condensations, and Fgfr3 is expressed in differentiating chondrocytes (Peters et al., 1992; Delezoide et al., 1998). Given the proximity of Fgf9 in mesenchyme surrounding the condensations and Fgf18 in nascent perichondrium (Liu et al., 2002; Ohbayashi et al., 2002; Hung et al., 2007), it is likely that FGF9 and FGF18 could signal locally to FGFR2 and FGFR3.

Mice lacking Fgf9 or Fgf18 have reduced chondrocyte proliferation and delayed initiation of chondrocyte maturation at E14.5. This phenotype is consistent with FGF9/18 signaling to FGFR3 since activation of this receptor has been shown to positively regulate early chondrogenesis (Iwata et al., 2000, 2001; Davidson et al., 2005; Havens et al., 2008). However, loss of both Fgf9 and Fgf18 causes an earlier and much more severe chondrocyte differentiation defect than that seen in Fgfr3 null mice. Thus, in addition to FGFR3, FGF9/18 may also signal to FGFR2 in skeletogenic condensations to regulate the initial steps of endochondral ossification.

By E16.5, experimental evidence suggests that FGF18 signals to FGFR3 to inhibit chondrocyte proliferation and hypertrophic maturation, since both Fgf18^−/− and Fgfr3^−/− growth plate phenotypes include expanded proliferating and hypertrophic zones as well as increased Ihh expression (Colvin et al., 1996; Deng et al., 1996; Liu et al., 2002, 2007; Ohbayashi et al., 2002). By contrast, although Fgf9^−/− proximal limb growth plates have expanded hypertrophic zones, the proliferating zone height and Ihh expression levels are not increased, suggesting that FGFR3 may not be the primary target of FGF9 beyond E16.5. Rather, FGFR1 may be a primary receptor for FGF9, since inactivation of Fgfr1 in limb mesenchyme delays the initiation of chondrocyte hypertrophy (Hung et al., 2007), leading to an expanded hypertrophic zone without significant changes in IHH signaling (Jacob et al., 2006).

We conclude that FGF9/18 likely signals to FGFR2/3 by E12.5 to positively regulate chondrogenesis, while by E16.5, FGF18/FGFR3 signaling inhibits chondrocyte proliferation and differentiation and FGF9/FGFR1 signaling promotes hypertrophic differentiation. Refinement of FGF/FGFR relationships involved in early chondrogenesis will require precise cell-specific knockouts of FGFRs.

3.3. Fgf9 and Fgf18 regulate early chondrogenesis

In the Fgf9^−/−; Fgf18^−/− proximal hindlimb, CyclinD1-positive columnar proliferating chondrocytes fail to form by E18.5. Thus, simultaneous loss of Fgf9 and Fgf18 results in a much earlier disruption in chondrogenesis than previously identified in the Fgf9 or Fgf18 single knockout mice, occurring just after initial mesenchymal cell differentiation to Col2a1-expressing chondrocytes. These findings suggest a critical role for FGF signaling in differentiation of immature chondrocytes to proliferating chondrocytes and in regulation of CyclinD1 expression either directly or indirectly. An indirect mechanism by which FGFs may influence chondrocyte differentiation and their proliferative capacity is by modulating IHH-PTHLH signaling. IHH regulates chondrocyte hypertrophy through induction of Pthlh (St-Jacques et al., 1999) and promotes differentiation of immature chondrocytes to rapidly dividing, CyclinD1-positive proliferating chondrocytes through a PTHLH-independent pathway (Karp et al., 2000; Long et al., 2001; Kobayashi et al., 2005). In support of this model, reduced levels of Ihh were observed in the Fgf9 and Fgf18 single knockout mice at mid-gestational stages (Hung et al., 2007; Liu et al., 2007). Additionally, absence of both encoded ligands further decreased Ihh levels at an even earlier stage of limb chondrogenesis (E12.5). Hedgehog signaling was undetectable in the Fgf9^−/−; Fgf18^−/− proximal limb cartilage primordium by E14.5, when differentiated chondrocytes are normally well-established. In mice harboring an Fgfr3 gain of function mutation, Iwata et al. (2000) observed increased chondrocyte proliferation and increased Ptch1 expression at E15.5, providing additional supportive evidence that FGFs may control early chondrogenesis through the IHH-PTHLH pathway.

However, without Hedgehog signaling, chondrocytes undergo accelerated and extensive hypertrophic differentiation by E18.5 subsequent to an initial delay in cellular maturation (St-Jacques et al., 1999). In Fgf9^−/−; Fgf18^−/− mice, although expression of Ihh and Ptch1 was essentially absent by E14.5, hypertrophic chondrocytes were not observed in the proximal hindlimb. One explanation for this failed maturation could be that IHH signaling is present but below the level of in situ hybridization detection. Another possibility is that FGF9 and FGF18 may regulate a molecule/pathway upstream of IHH, with abrogation of this regulation leading to a severe differentiation defect and subsequent loss of Hedgehog signaling.

Runx2^−/− mice are phenotypically similar to Fgf9^−/−; Fgf18^−/− mice since they exhibit absence of both Ihh expression and hypertrophic chondrocytes in the stylopod elements. RUNX2 is essential for chondrocyte proliferation and maturation in the proximal limb. RUNX2 functions upstream of IHH since Ihh is not expressed in Runx2^−/− stylopod elements, and influences chondrocyte proliferation and differentiation through direct transcriptional activation of Ihh (Inada et al., 1999; Kim et al., 1999; Takeda et al., 2001; Yoshida et al., 2004; Chen et al., 2014). In the current study, during early stages of chondrogenesis, FGF9 and FGF18 appear to be required for Runx2 and Ihh expression and therefore likely function through these factors to control differentiation of immature to proliferating and then hypertrophic chondrocytes. The epistatic relationship between Runx2 and Fgf18 is complex, though, because at later stages Runx2 reportedly upregulates Fgf18 in the perichondrium, leading to inhibition of chondrocyte proliferation and hypertrophy (Hinoi et al., 2006). These findings are in accordance with results of our previous work showing biphasic functions for FGF18 during chondrocyte development (reviewed in Ornitz and Marie (2015)). Thus, a novel role of FGF9 and FGF18 is to provide a permissive cue for immature proliferative chondrocytes to undergo differentiation through interactions with key chondrogenic pathways.

4. Conclusions

Together, these data demonstrate that FGF9 and FGF18 are critical ligands that modulate many aspects of embryonic bone formation. The extreme phenotype caused by simultaneous loss of FGF9 and FGF18 demonstrates extensive redundancy between these ligands in early chondrogenesis and osteogenesis. These studies show that FGFs function by acting upstream of the IHH-PTHLH axis and RUNX2 to promote differentiation of immature to columnar proliferating chondrocytes and to enhance cellular proliferation. The mechanisms by which FGF ligands control skeletogenesis are still being elucidated; experimental evidence thus far indicates FGF9/18 have distinctive gradients of activity across the proximodistal limb axis, with varied, even oppositional, functions depending on developmental context and timing.

5. Materials and methods

5.1. Generation of mice

Fgf9^+/− mice (Colvin et al., 2001) were crossed with Fgf18^+/− mice (Liu et al., 2002) to generate Fgf9^+/−; Fgf18^+/− animals. These were subsequently intercrossed to generate a combined series of Fgf9 and Fgf18 mutant alleles. Controls used in this study were genotypes Fgf9^+/+; Fgf18^+/+, Fgf9^+/−; Fgf18^+/+, or Fgf9^+/+; Fgf18^+/−.

5.2. Skeletal preparations and bone morphometric analysis

Skeletons were prepared as described previously (Colvin et al., 1996). Carcasses were skinned and eviscerated, then soaked in acetone for 12–24 h, stained with Alizarin red S and alcian blue (Sigma) (72 h), cleared in 1% KOH/20% glycerol, and stored in glycerol. Individual bone lengths were measured at E18.5 or P0 using Canvas 12 (ACD Systems) software. Three mice from each genotype group were analyzed.

5.3. Histological analysis

Tissues were fixed in 4% paraformaldehyde/PBS and embedded in paraffin. Sections (5 μm) were stained with hematoxylin and eosin (H&E). AxioVision 4.0 software (Zeiss) was used for computer imaging. Mineralized bone was visualized by von Kossa staining with methyl green counterstaining.

5.4. Immunohistochemistry

For CyclinD1 immunostaining, paraffin sections were dewaxed, rehydrated, antigen unmasked in citrate buffer, pH 6.0 for 20 min at 95 °C, incubated in hydrogen peroxide/PBS, and blocked in 5% sheep serum. Prediluted CyclinD1 antibody (ThermoScientific Cat. RM-9104-R7) and biotinylated goat anti-rabbit IgG were used. Amplification was achieved by using Vectastain ABC-peroxidase reagent and DAB for visualization (Vector).

5.5. In situ hybridization

Whole-mount in situ hybridization was performed as previously described (Wilkinson, 1992). Section in situ hybridization was performed using digoxigenen-labeled probes or ³³P-labeled probes as described (Naski et al., 1998) except that hybridization buffer for digoxigenen-labeled probes was as follows: 50% formamide, 5×SSC, 50 μg/mL yeast tRNA, 100 μg/mL heparin, 0.2% Tween-20, 0.5% CHAPS, 5 mM EDTA pH8.0. In situ probes used were Sox9 (Bi et al., 1999), Col2a1 (Kohno et al., 1984), Col10a1 (Jacenko et al., 1993), Col1a1 (Rossert et al., 1995), Ihh (Bitgood and McMahon, 1995), Ptch1 (Goodrich et al., 1996), Pthlh and Pth1r (Lanske et al., 1996), Runx2 (Ducy et al., 1997), and Bsp (Chen et al., 1991).

5.6. Proliferation and apoptosis analysis

For embryos, time-mated females were injected intraperitoneally with BrdU (100 μg/g body weight) 1 h prior to sacrifice. For Alcian blue staining, paraffin sections were dewaxed, rehydrated, and incubated in 3% acetic acid for 3 min followed by 2.5% Alcian blue solution, pH2.5 with acetic acid, for 45 min. Subsequently, BrdU immunohistochemistry (mouse IgG, BD Pharmingen, Cat. 555627 was performed as described (Naski et al., 1998). For statistical analysis, nine sections were analyzed per genotype (three sections from three different specimens per genotype), viewed through a 20× objective. The percentage of BrdU-positive nuclei versus total nuclei was calculated as the proliferation index. Data shown as mean±SD. TUNEL assay was performed using the In Situ Cell Death Detection Kit, POD (Roche) per manufacturer’s instructions.