Fibroblast growth factor expression during skeletal fracture healing in mice



Fibroblast growth factors (FGFs) are important signaling molecules that regulate many stages of endochondral bone development. During the healing of a skeletal fracture, several features of endochondral bone development are reactivated. To better understand the role of FGFs in skeletal fracture healing, we quantitatively evaluated the temporal expression patterns of Fgfs, Fgf receptors (Fgfrs), and molecular markers of bone development over a 14-day period following long bone fracture in a mouse model. These studies identify distinct groups of FGFs that are differentially expressed and suggest active stage-specific roles for FGF signaling during the fracture repair process. Developmental Dynamics 238:766–774, 2009. © 2009 Wiley-Liss, Inc.


Skeletal fracture repair recapitulates several stages of endochondral bone development. During endochondral bone development, undifferentiated mesenchymal cells differentiate into chondrocytes. These chondrocytes then undergo the processes of proliferation, hypertrophy, ossification, death, vasculogenesis, and finally, replacement with bone-forming osteoblasts (Provot and Schipani,2005). After a fracture, signals that initiate the repair process originate from the bone marrow, injured bone matrix and inflammatory cells that infiltrate the site of injury. The initial injury leads to mesenchymal stem cell recruitment and differentiation, cartilaginous callus formation, cartilage resorption, primary spongiosa formation, and finally, secondary bone formation and remodeling (Gerstenfeld et al.,2003,2006; Shapiro,2008).

Genes that function during development are often re-expressed and re-used during physiological challenges. The Fibroblast Growth Factor (Fgf) and Fgf receptor (Fgfr) gene families encode essential signaling molecules that function throughout all stages of endochondral bone development. For example, deficiencies in FGF2 have been linked to decreased bone formation and bone mass (Montero et al.,2000), while FGF9 and FGF18 signaling regulate hypertrophic chondrocyte differentiation, skeletal vascularization, and osteoblast/osteoclast recruitment to the growth plate (Liu et al.,2002,2007; Hung et al.,2007). The FGFRs also function in osteogenic differentiation and maturation. FGFR1 signaling in osteochondroprogenitor cells regulates osteoblast maturation (Jacob et al.,2006). FGFR2 is important in skeletal growth and bone density (Yu et al.,2003), and mutations in FGFR2 are linked to a variety of craniosynostosis syndromes in humans (Morriss-Kay and Wilkie,2005; Marie et al.,2008). Skeletal overgrowth has been seen in association with FGFR3 deficiency in both mouse and human (Colvin et al.,1996; Toydemir et al.,2006). Disruption of FGF signaling during bone development results in a range of phenotypes from relatively mild shortening of limbs to severe dismorphology, including truncated and missing limbs.

Although the importance of the FGF family during skeletal development is well established, the role of FGFs during fracture healing is less clear. Multiple studies have used a rat femoral fracture model to study the expression of Fgfrs during fracture healing. Fgfr1 expression, localized to subperiosteal cells, was rapidly up-regulated 1 day following fracture and peaked after 14 days (Nakajima et al.,2001a; Rundle et al.,2002). Fgfr2 was also expressed in the periosteum and in the chondrocytes of the fracture callus, but its expression was not observed to be regulated during the fracture repair process (Rundle et al.,2002). The onset of Fgfr3 expression initiated 10 days after fracture in prehypertrophic and hypertrophic chondrocytes of the fracture callus (Rundle et al.,2002; Nakajima et al.,2003).

FGF ligands have not been systematically examined during the fracture repair process. Immunohistochemical studies showed that FGF1 (aFGF) was expressed 3 days after fracture and localized to the perichondrium (Bourque et al.,1993). Nakajima et al. examined but did not detect Fgf2 expression during fracture healing (Nakajima et al.,2001a); however, Pacicca et al. reported increased Fgf2 expression during distraction osteogenesis (Pacicca et al.,2003). Haque et al. showed increased expression of FGF1, FGF2, and FGF18 in fibroblasts and chondrocytes in the callus region in a rabbit model of distraction osteogenesis (Haque et al.,2007). FGF2 has been tested as a potential therapeutic molecule for skeletal repair. Injection of FGF2 at the fracture site increased the size of the cartilaginous callus, although it did not improve the ossification or remodeling aspects of the repair process (Nakajima et al.,2001b,2007). These studies highlight a potential role for FGF signaling during fracture repair; however, no study has comprehensively evaluated the coordinated expression of Fgf ligands and receptors throughout the fracture repair process.

A fracture method, established for the mouse, creates a transverse fracture in a prenailed tibia, allowing for maintained axial alignment and union without displacement (Hiltunen et al.,1993; Sakano et al.,1999). This method, which provides some stability to the fracture site (similar to that provided by casting) without rigid external fixation, provides a clinically relevant tool for the study of fracture healing. To better understand the role of FGFs in bone healing, we quantitatively evaluated the temporal expression patterns of Fgfs, Fgfrs, and molecular markers of bone repair at four time points over 14 days following long bone fracture using this model. These results implicate several distinct roles for FGF signaling at multiple stages of the fracture repair process.


Tibial Fracture Model

Tibial fractures, internally stabilized by insertion of an intramedullary wire, were allowed to heal over a 14-day period. Radiographic imaging (Fig. 1A) and histology (Fig. 1B and Supp. Fig. 1. S1, which is available online) of longitudinal sections of the bone/fracture were used to stage the healing process. One day postfracture (1 DPF), radiography (Fig. 1A2,A3) and histology (hematoxylin and eosin [H&E] staining, Supp. Fig. S1B) showed skeletal nonunion of trabecular bone (picro-sirius red positive, Fig. 1B2) with small pockets of polysaccharide-rich cells (alcian blue positive, Fig. 1B2; Blumer et al.,2004), indicating normal osteoblast/osteoclast mediated remodeling. After 4 DPF, radiography confirmed skeletal nonunion (Fig. 1A4,A5), while histology showed increased granulomatous tissue in the marrow space and inflammatory cells, signifying the beginning of a fracture callus (H&E staining, Supp. Fig. S1C), as well as alcian blue positive osteoprogenitor cells in the expanded periosteum near the fracture site (Fig. 1B3). After nine days (9 DPF), radiography showed bony bridging spanning the fractured ends of the tibia, indicating some ossification (Fig. 1A6,A7), and histology showed a large fracture callus with areas of active chondrogenesis, spongy (trabecular) bone formation, and woven bone trabeculae rimmed by osteoblasts (Supp. Fig. S1D). The newly forming trabecular bone, along with the bone collar that surrounds the blue stained chondrocyte-rich cartilage callus, was identified as bone matrix (Fig. 1B4). After 14 days (14 DPF), radiography showed complete fracture union with mature cortical bone throughout the site of the previous fracture (Fig. 1A8,A9). Histology demonstrated remnants of cartilage surrounded by areas of trabecular bone (Fig. 1B5 and Supp. Fig. S1E).

Figure 1.

Radiography, alcian blue/picro-sirius red staining, and Sox9 in situ hybridization of mouse tibia over the first 14 days after fractures. A: Radiographs of pinned tibia, including magnification of the fracture site. A1: Unfractured tibia. A2,A3: fractured tibia. A4,A5: 4 days postfracture (DPF); A6,A7: 9 DPF; A8,A9: 14 PDF. Dashed lines indicate area used for RNA harvest as well as magnified X-ray. B: Alcian blue/picro-sirius red staining of the mouse tibia over the 14 days after fracture. B1: Unfractured tibia. B2: 1 DPF; B3: 4 DPF; B4: 9 DPF; B5: 14 DPF. C:Sox9 expression at 4 DPF(C1), 9 DPF (C2,C3), and 14 DPF (C4) in the mouse tibia. Black and white arrows indicate approximate midline of the fracture; blue arrows indicate bony bridging in A6. Black boxes in B correspond to areas shown in C. Scale bars = 500 μm in B,C.

In situ hybridization for Sox9 was used to verify areas of osteochondrogenic activity. In pinned, but unfractured bone, as well as at 1 DPF, some faint stain was detected in the bone marrow cavity (data not shown). This is most likely due to the injury caused by the pin insertion. At 4 DPF, Sox9 appeared to be active in the expanded periosteum near the fracture site, indicating activated osteochondroprogenitor cells (Fig. 1C1). At 9 DPF, Sox9 activity was seen throughout the fracture callus, particularly in the periosteum, chondrocytes, and the osteoblasts surrounding the newly formed trabecular bone (Fig. 1C2,C3). By 14 DPF, Sox9 expression was no longer detected in the periosteum, but remained in the small chondrocyte pockets of the fracture callus (Fig. 1C4). These data are consistent with previously published Sox9 expression in a mouse tibial fracture model (Sakano et al.,1999).

Quantitative Gene Expression

Because FGF signaling is active during many stages of endochondral skeletal development, we sought to determine whether Fgf and Fgfr expression were differentially regulated during the fracture healing process and how other signaling pathways known to interact with FGF signaling pathways correlated with the expression of different members of the Fgf family. Based on the histological and radiographic time course, RNA was prepared from the fracture site at one, four, nine and fourteen days after acute fracture. An initial semiquantitative screen was performed on Fgfs known to be expressed in developing endochondral bone, Fgfs expected to be expressed in differentiating cells that populate the fracture site, and Fgfs known to be important in wound healing. Fgfs1, 2, 5, 6, 9, 16, 17, and 18 all showed expression in at least one time point. Fgfs4, 8, and 20 were not detected during fracture healing, but were expressed in control mouse embryonic day 10.5 RNA (data not shown).

Quantitative RT-PCR was used to monitor transcript levels of Fgfs and other genes at the four selected time points after fracture. Three different groups of genes with similar expression patterns emerged from this analysis (Figs. 2–4). The first group included early response genes, with peak expression at 1 DPF. The second group, peaking at 9 DPF, included genes involved in chondrogenesis. The third group included genes active in ossification and skeletal remodeling and showed highest levels of expression at 14 DPF. The expression levels of all genes were normalized to expression of the housekeeping gene, Hprt, and were compared with levels in unfractured bone.

Figure 2.

Quantitative reverse transcriptase-polymerase chain reaction (RT-PCR (qPCR) of genes responding early to tibial pinning and fracture, measured relative to unmanipulated animals. A–F:Fgf2 (A), Fgf5 (B), Fgf6 (C), Fgf9 (D), Bmp2 (E), and Vegfa (F) at 0, 1, 4, 9, and 14 days after manipulation. Closed diamonds (♦) represent fracture data, open diamonds (◊) represent pinned only data. All values were normalized to the housekeeping gene Hprt and represent mRNA levels as fold difference relative to unmanipulated animals. Error bars indicate standard error of the mean from n = 3 per time point. P values: ★ < 0.05, ★★ < 0.01, ★★★ < 0.001, and ☆ < 0.05, ☆☆ < 0.01, and ☆☆☆ < 0.001 for fractured and pinned data, respectively, as calculated using the Student's t-test.

Figure 3.

Quantitative reverse transcriptase-polymerase chain reaction (qPCR) of genes with expression peaking at 9 days after tibial pinning and fracture measured relative to unfractured animals. A–G: Fgf16 (A), Fgf18 (B), Fgfr3 (C), Ptch1 (D), Sox9 (E), Col2a1 (F), and Col10a1 (GA) at 0, 1, 4, 9, and 14 days after manipulation. Symbols and statistics are as in Figure 2.

Figure 4.

Quantitative reverse transcriptase-polymerase chain reaction (qPCR) of genes with maximal expression at 14 days after tibial fracture measured relative to unfractured animals. A–H:Fgf1 (A), Fgf17 (B), Fgfr1 (C), Fgfr2 (D), Cd68 (E), Bsp (F), Runx2 (G), and Osx (H) at 0, 1, 4, 9, and 14 days postfracture (DPF). Symbols and statistics are as in Figure 2.

Damage to the bone marrow cavity and endosteum, due to the insertion of the intramedullary pin, was expected to induce some change in gene expression. Consequently, we evaluated pinned animals at 1 and 9 days after pinning and compared them with unmanipulated animals. We also compared the data of pinned groups with that of fracture groups at corresponding time points to determine if elevated levels of gene expression were due to the response evoked by induced fracture or in response to pin insertion alone.

Early Response Genes

Fgf2 is typically expressed in injured tissue. As expected, Fgf2 was significantly (P < 0.05, 0.01, respectively) up-regulated at 1 DPF in both the pin control and in fractured bone (Fig. 2A). Consistent with previous observations (Bourque et al.,1993), Fgf1 also increased significantly (P < 0.05, 0.001, respectively) in both the pin control and in the fractured bone groups at this time point (Fig. 4A). Of interest, Fgf5 showed dramatic up-regulation at 1 DPF (84- ± 0.2-fold) preferentially in fractured bone compared with the pin only control (38- ± 10.4-fold; Fig. 2B).

Fgfr2 expression increased significantly (P < 0.01, 0.05, respectively) in both the pin control and in fractured bone at 1 DPF (Fig. 4D). Fgfr1 was also significantly (P < 0.05) up-regulated 1 day after fracture, consistent with prior reports (Fig. 4C; Nakajima et al.,2001a). Cd68, a gene expressed in macrophages (Greaves and Gordon,2002), Bmp2, which has been shown to be required for the initiation of fracture repair (Tsuji et al.,2006), and Vegfa, a modulator of angiogenesis (Geiger et al.,2005), also showed a significant (P < 0.001, 0.001, 0.01, respectively) early response to pin placement and fracture (Figs. 2E,F, 4E). Vegfa expression has also been linked to osteogenesis (Gerber et al.,1999; Mayr-Wohlfart et al.,2002; Zelzer et al.,2002) along with Bmp2 (Tsuji et al.,2006; Kang and He,2008), which may explain the delayed but steady increase in expression seen during later stages of the fracture healing process. Fgf6 and Fgf9 (Fig. 2C,D) demonstrated similar trends to Bmp2 and Vegfa in response to fracture but were not statistically significant. At 1 DPF, no change in expression was observed for markers of chondrogenesis or osteogenesis.

The early response of Fgf2 is likely due to the effect of the pin insertion as no difference was found between Fgf2 up-regulation in the pinned control and the fracture specimens at 1 DPF. However, the sustained increase in Fgf2 expression at 9 DPF is more likely due to the fracture itself, as there is no difference between the pin control and the unpinned data at 9 DPF. The early response of Fgf5, however, is most likely due to the induced fracture rather than pin insertion (2.2 ± 0.3 at 1 DPF relative to pinned control, P < 0.05). This indicates that Fgf5 may be important in the early stages of fracture healing. Interestingly, evolutionary analysis of the Fgf family grouped Fgfs1, 2, and 5 together (Itoh and Ornitz,2008). Thus, these three Fgfs may act redundantly in the early response to skeletal injury.

Delayed Response Genes and Chondrogenic Factors

The 4 to 9 DPF time frame was characterized by an increase in expression of genes known to be important in chondrogenesis. Sox9 (Fig. 3E), an early marker of osteochondroprogenitor cells and chondrocyte differentiation, has been shown to be required for cartilage formation (Wright et al.,1995) and is increased during this time. Bsp, Runx2, and Osx, genes expressed in active periosteum and in osteoblasts (Chen et al.,1991; Komori,2002; Nakashima and de Crombrugghe,2003), all increased during this time (Fig. 4F–H). Consistent with endochondral bone formation, active chondrogenesis and differentiation of hypertrophic chondrocytes, Col2a1 and Col10a1 were significantly (P < 0.05, 0.01, respectively) increased at 9 DPF (Fig. 2F,G). Cd68 also continued to increase during this period, indicating continued macrophage infiltration. Ptch1, an indicator of active hedgehog signaling (Chen and Struhl,1996), was also significantly (P < 0.001) up-regulated at 9 DPF (Fig. 3D).

Several members of the Fgf and Fgf receptor family were also expressed at 9 DPF. Consistent with previous findings (Rundle et al.,2002; Nakajima et al.,2003), Fgfr3 was significantly up-regulated (P < 0.05) (Fig. 3C). During endochondral bone formation, Fgfr3 is expressed in proliferating chondrocytes (Peters et al.,1993) and in mature osteoblasts and osteocytes (Xiao et al.,2004). Fgf16 and Fgf18 were also significantly up-regulated (P < 0.001, 0.01, respectively) at 9 DPF (Fig. 3A,B). Fgf18 has been shown to be expressed in the perichondrium and periosteum (Liu et al.,2002), and both ligands have been shown to signal to Fgfr3 (Zhang et al.,2006).

Late Response Genes and Bone Remodeling Factors

Decreasing expression of many of the chondrogenic markers at 14 DPF and the continual increase in expression of ossification markers indicates that remodeling occurs at this late stage. Bsp and Runx2, both previously shown to be expressed in osteoblasts lining trabecular surfaces and in osteocytes in newly formed woven bone (Amir et al.,2007), and Osx, recently shown to be expressed in mesenchymal progenitor cells in the periosteum, as well as immature chondrocytes and osteoblasts within the fracture callus (Kaback et al.,2008), all peaked at 14 DPF (P < 0.001), consistent with active ossification. Elevated Cd68, at this stage, is consistent with ongoing macrophage or osteoclast-mediated bone remodeling.

Two FGF receptors (Fgfr1 and Fgfr2) were maximally expressed at this time point. Consistent with previous studies, Fgfr1 expression continued to increase after fracture. Contrary to previous studies (Rundle et al.,2002), we found Fgfr2 expression significantly (P < 0.05) up-regulated after fracture and peaking at later stages (P < 0.001). FGF1 and FGF17, both of which can signal to FGFR1 and FGFR2 (Zhang et al.,2006), also showed maximum expression at 14 DPF (P < 0.001). Fgf17 (Fig. 4B) expression has been observed in the perichondrium (Xu et al.,1999). Additionally, Fgfr3, Fgf9, Fgf16, and Fgf18 were expressed at significantly (P < 0.001, 0.01, 0.01, 0.05, respectively) elevated levels, consistent with resolving chondrogenesis and ongoing osteogenesis.


In this study, we defined a time course for Fgf ligand and receptor expression during fracture healing (summarized in Fig. 5), using a well-established mouse model for fracture repair. Fgfs not previously known to be expressed during the fracture repair process were identified. Importantly, we have identified concordance between the expression of particular ligands and their known receptors during different stages of fracture repair. These studies suggest that different components of the fracture repair process could be regulated through pharmacological or genetic manipulation of specific FGF–FGF receptor signaling pathways. Further studies using genetically engineered mouse models could help to identify the interactions and pathways involving the FGF family and identify gene targets for pharmacological or therapeutic interventions.

Figure 5.

Graphical representation of relative gene expression over the 14-day postfracture time period.


Fracture Model

Female Black Swiss Webster mice (Taconic) were used in this study according to the guidelines set forth by The American Association for Laboratory Animal Science. The mice were 8 weeks of age at the time of fracture and weighed ∼23 g. Mice were anesthetized with an IP injection of KAX (ketamine, acepromazine, xylazine in phosphate buffered saline [PBS]) before any experimental procedure. A closed tibial fracture was made using a method adapted from Bourque et al. (1992). A ∼1-cm incision was made medial and parallel to the patellar tendon. Using a 23-gauge needle, a hole was made in the proximal end of the tibia, medial to the patellar tendon. A 0.4-mm-diameter stainless steel pin was introduced into the intramedullary canal through the needle, running the full length of the tibia. Pin placement was verified using X-ray tomography. The fracture was created using a blunt-ended guillotine-type apparatus and verified by X-ray. The proximal end of the pin was bent and cut so as not to interfere with knee movement, and the wound was closed with two to three interrupted, nylon sutures. The result was a highly reproducible transverse fracture of the tibia with little soft tissue damage. Six mice were used for each time point in this study: unmanipulated (Time 0), 1 DPF, 4 DPF, 9 DPF, and 14 DPF. For each group, 3 mice were used for histology and 3 were used for quantitative PCR. Eight mice were used in a pinned but not fractured control group and analyzed at 1 and 9 days post-pin.

RNA Isolation

Animals were euthanized in a CO2 chamber. The fractured (left) leg and contralateral (right) leg were dissected out and muscle and tendon removed without disrupting the periosteum or soft tissue callus. The pin was removed and the tissue was snap frozen by submersion in liquid nitrogen. Frozen tissues were pulverized in a dry ice cooled stainless steel flask with a ball bearing in a Micro Dismembrator (Sartorius) at 2,000 RPM for 20 sec. RNA was stabilized with TRIzol reagent (Invitrogen) and isolated using a chloroform extraction. RNA purification was completed using the RNeasy kit (Qiagen), and DNA contamination was eliminated using an on-column RNase-Free DNase Set (Qiagen) according to the manufacturer's instructions.

cDNA Synthesis

Superscript III DNA First Strand Synthesis System (Invitrogen) for cDNA synthesis was used according to the manufacturer's instructions. Five hundred ng of RNA was used as starting material for the reactions; cDNA was eluted in 30 μl H2O and diluted to a final volume of 130 μl.

Quantitative Real Time PCR

Gene expression was quantified using Taqman Gene Expression Assays (Applied Biosystems) (Table 1) and an ABS7500 Fast Thermocycler with the following program on the Fast setting: 95°C for 20 sec followed by 45 cycles of 95°C for 3 sec and 30 sec at 60°C. Target genes were normalized to Hprt. All samples were represented as relative expression normalized to the amount of mRNA in the control group. Error bars represent the standard error of the mean. Student's t-test was used to determine P values.

Table 1. Gene Expression Assay Primers
GeneRefSeq (accession no.)ABI catalog no.
Fgfr1NM_001079908.1, NM_010206.2Mm00438923_m1
Fgfr2NM_201601.2, NM_010207.2Mm00438941_m1
VegfaNM_001025250.3, NM_001025257.3, NM_009505.4Mm00437304_m1

Histology and In Situ Hybridization

Histological reagents and chemicals were purchased from Sigma Chemical Company (St. Louis, MO) or from Fisher Scientific (Springfield, NJ). Tissues to be used for histology were dissected in DEPC-PBS and fixed in formalin overnight. Tissues were washed in DEPC-PBS, decalcified in 20% ethylenediaminetetraacetic acid (EDTA) for 28 days, washed in DPEC-PBS and stored in 70% ethanol until embedded in paraffin. Five-micrometer sections were cut and placed on positively charged slides. Sections were stained with hematoxylin and eosin (H&E).

For the alcian blue and picro-sirius red stain (, briefly, sections were deparaffinized and hydrated in dH2O, stained in Gill's 2× hematoxylin for 2 min, rinsed in dH2O, PBS and again in dH2O, stained in alcian blue for 30 min and washed in dH2O. Slides were then stained in picro-sirius red for 1 hr, followed by two washes in acidified water. Slides were then dehydrated and mounted.

For in situ hybridization, slides were deparaffinized, rehydrated in graded ethanol solutions, and washed in PBS. Slides were then fixed in 4% paraformaldehyde (PFA) for 25 min and washed in PBS. Tissues were digested with proteinase K (5 mg/10 ml PBS) for 10 min at 37°C, washed in PBS, then post-fixed in 4% PFA for 10 min, and again washed in PBS. Slides were bathed in 2× standard saline citrate (SSC) for 5 min and 0.1 M triethanolamine (TEA) for 5 min. A mixture of 0.25% acetic acid in 0.1 M TEA (25 min) was used for acetylation, and the slides were then washed again in 2× SSC. Slides were dehydrated in graded ethanol before hybridization (1:100 probe in high salt buffer) overnight at 52°C in a humid chamber (50% formamide and 50% 2× SSC).

Slides were washed in a mixture of 50% formamide, 1× SSC, and 0.1% Tween-20 at 52°C for 15 min then again for 30 min and finally rinsed in MABT at room temperature. Slides were then blocked with 2% Boehringer blocking reagent and 20% sheep serum in MABT in a humid chamber for 1.5 hr at room temperature. Secondary antibody was added to this solution (1:2,000 antibody to blocking solution) for overnight incubation at room temperature.

The following day, slides were repeatedly rinsed in MABT (5 times for 30 min each) on a gentle rocker before pH equilibration in NTMT (2 times at 10 min each). NBT/BCIP (0.66%) in NTMT was applied to slides for color reaction. After appropriate color development, slides were submerged in PBS to halt the reaction. Zymed GVA Mount was used to mount the slides.


We thank T. Imamura for help with skeletal fractures and C. Press, C. Smith, M. Seppa, and A. Snyder-Warwick for critically reading the manuscript.