JOURNAL OF BONE AND MINERAL RESEARCHVolume 24, Number 12, 2009Published online on June 29, 2009; doi: 10.1359/JBMR.090603� 2009 American Society for Bone and Mineral Research

Fra-1/AP-1 Impairs Inflammatory Responses and Chondrogenesisin Fracture Healing

Toru Yamaguchi,1,2 Yasunari Takada,2,3,4 Kenta Maruyama,3 Kouji Shimoda,5 Yoshinori Arai,6 Nobuhito Nango,7

Naoto Kosaki,1 Hironari Takaishi,1 Yoshiaki Toyama,1 and Koichi Matsuo3,4

ABSTRACT: Inflammation inevitably follows injury of various tissues, including bone. Transgenic over-expression of Fra-1, a component of the transcription factor activator protein-1 (AP-1), in various tissuesprogressively and globally enhances bone formation, but little is known about the possible effects of Fra-1/AP-1 on fracture healing. We created a transverse fracture of the mouse tibial diaphysis and examinedfracture healing radiologically, histologically, and immunologically. Strikingly, fracture union was delayedeven though the bone formation rate in callus was higher in Fra-1 transgenic (Tg) mice. In these mice,chondrogenesis around the fracture site was impaired, resulting in accumulation of fibrous tissue, whichinterferes with the formation of a bony bridge across the callus. Curiously, immediately after fracture,induction of the inflammatory mediators TNF-a, interleukin (IL)-6, and Cox-2 was significantly suppressed inFra-1 Tg mice followed, by the reduced expression of Sox-9 and BMP-2. Because serum prostaglandin E2

(PGE2) levels were dramatically low in these mice, we administered PGE2 to the fracture site using a slow-release carrier. The accumulation of fibrous tissue in Fra-1 Tg mice was significantly reduced by PGE2

administration, and chondrogenesis near the fracture site was partially restored. These data suggest that theFra-1-containing transcription factor AP-1 inhibits fracture-induced endochondral ossification and bonybridge formation presumably through suppression of inflammation-induced chondrogenesis.J Bone Miner Res 2009;24:2056–2065. Published online on June 29, 2009; doi: 10.1359/JBMR.090603

Key words: rodent, cytokines, gene regulation, inflammation

Address correspondence to: Koichi Matsuo, MD, PhD, School of Medicine, Keio University,35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan, E-mail: matsuo@sc.itc.keio.ac.jp

INTRODUCTION

HEALING OF BONE fractures regenerates broken boneand restores its original shape.(1,2) Fracture healing is

a sequential process that includes inflammation, angio-genesis, mesenchymal progenitor recruitment, cartilageformation, and callus remodeling.(2) After fracture, notonly the bone itself but also the periosteum, blood vessels,surrounding soft tissue, and bone marrow are injured.Disruption of periosteal and endosteal blood vessels andvessels inside bone results in hematoma formation, hyp-oxia, and angiogenesis. Inflammatory cells such as neu-trophils and macrophages infiltrate the injured area, pro-ducing inflammatory mediators such as TNF-a, interleukin(IL)-6, and prostaglandin E2 (PGE2) and leading to pain,swelling, and fever. Mesenchymal progenitors, which makeup the fibrous tissue around the fracture gap, undergoendochondral bone formation, initially differentiating in-to chondrocytes to form cartilage, which is replaced bycalcified bone and bone marrow. At the periosteum adja-cent to the fracture gap, the mesenchymal progenitors

differentiate into osteoblasts for intramembranous ossifi-cation. The molecular events underlying endochondral andintramembranous ossification after fracture resemble thoseoccurring during development,(3,4) except for the occur-rence of inflammatory processes seen in fracture healing.(5)

Fos family proteins, such as c-Fos, Fra-1, Fra-2, andFosB, heterodimerize with Jun proteins to form the di-meric transcription factor activator protein-1 (AP-1).(6)

The prototypical Fos family member c-Fos regulates in-flammatory mediators in mice.(7–9) When mice lacking c-Fosare injected with lipopolysaccharide (LPS), a cell wall com-ponent of gram-negative bacteria, they experience pro-nounced endotoxic shock characterized by hypothermia andtachycardia, which can be partially neutralized by anti-TNF-a antibody.(8) Fra-1 (Fos-related antigen, encoded by Fosl1)is similar to c-Fos in terms of its heterodimerization andDNA-binding properties. Fra-1 Tg mice, which overexpressFra-1 under the control of the H-2k promoter in numeroustissues, including bone, brain, heart, kidney, liver, lung, andspleen, develop osteosclerosis because of cell autonomous,systemic enhancement of osteoblastic bone formation,(10)

and mice lacking Fra-1 develop osteopenia.(11) Enhancedosteoblastogenesis is also observed in mice overexpressing

1Department of Orthopedic Surgery, School of Medicine, Keio University, Tokyo, Japan; 2These authors contributed equally to thiswork; 3Department of Microbiology and Immunology, School of Medicine, Keio University, Tokyo, Japan; 4Center for IntegratedMedical Research, School of Medicine, Keio University, Tokyo, Japan; 5Laboratory Animal Center, School of Medicine, Keio Uni-versity, Tokyo, Japan; 6Nihon University School of Dentistry, Tokyo, Japan; 7Ratoc System Engineering Co., Ltd., Bunkyo-ku, Tokyo,Japan.

The authors state that they have no conflicts of interest.

2056

a truncated form of FosB, DFosB.(12) Because the nano-structure of bone overproduced in Fra-1 Tg mice is normalexcept for a transient delay in mineralization,(13) one mightnaively expect that Fra-1–induced bone formation wouldaccelerate fracture healing or counteract osteoporosis.

In this study, we analyzed fracture healing in Fra-1 Tgmice using a tibial diaphyseal fracture model. Our dataindicate that inflammatory responses and the initiation ofendochondral callus formation are impaired in Fra-1 Tgmice at the fracture site.

MATERIALS AND METHODS

Mice

Fra-1 Tg mice were established as previously described.(10)

Mice were initially bred on a C57BL/6J 3 CBA mixedbackground and backcrossed onto C57BL/6J for 7–12generations. Fra-1 Tg mice were produced by naturalbreeding or in vitro fertilization using Fra-1 Tg males andC57BL/6J females. Littermates served as controls. All ex-periments were performed in accordance with the animaluse and care guidelines of Keio University.

Transverse fracture model

We performed a transverse osteotomy at the tibialmidshaft as reported previously(14) with minor modifica-tions. Six-week-old mice (n = 13 for each genotype) wereanesthetized by intraperitoneal injection of 0.01 ml/g bodyweight 2.5% avertin.(15) A longitudinal incision was madealong the front of the lower thigh. A transverse fracturewas generated in the tibial diaphysis 6 mm below the kneejoint using a 22-mm-diameter diamond disk of 0.4-mmthickness (Shofu, Kyoto, Japan) driven by a dental engine(Toyo Associates, Tokyo, Japan). To stabilize the fracture,either a 0.29-mm-diameter stainless steel rod (SN-2370;Terumo, Tokyo, Japan) for soft X-ray imaging or a 0.31-mm-diameter stainless steel pipe (Teshima, Gunma, Japan)for in vivo mCT imaging was inserted as an intramedullarynail from the lateral condyle of the tibia at the knee joint.The skin was sutured and mice were warmed on a heatedwater pad to enhance recovery from anesthesia. Mice wereallowed free unrestricted weight bearing in cages.

Closed fracture model

A closed fracture was generated using a guillotine-likedevice driven by a dropped weight(16) without an intra-medullary rod (n = 4 per data point). Six-week-old mice wereanesthetized with avertin as described above, and the tibiawas placed on the support stage to make a fracture in thesupine position at the tibial diaphysis. The distance betweenthe blunt guillotine blade and the stage was set at 1.5 mm. The500-g weight was dropped from a height of 35 cm. The forceof the descending weight was transmitted to the impact disc,driving the guillotine ramming system and fracturing thetibial diaphysis without inducing significant soft tissue trauma.

Soft X-ray analysis

Radiographic images were obtained on days 0, 14, 17,21, and 28 for wildtype and Fra-1 Tg mice. Mice were

anesthetized with avertin and radiographs were taken witha soft X-ray apparatus (CMB-2; Softex, Tokyo, Japan)using industrial X-ray film (IX; Fuji Film, Tokyo, Japan).At least four mice were imaged per data point.

mCT analysis

Mice were anesthetized with 2% isoflurane and analyzedusing an in vivo mCT scanner (R_mCT; Rigaku, Tokyo,Japan) operated at 90 kV and 200 mA with a copper filter of0.1 mm using 512 projections over 17 s (n = 4 for eachgroup). In vivo mCT imaging was performed on all 16mice 2, 7, 14, 21, and 28 days after the fracture. Thevoxel resolution was 20 mm. CT images were reconstructedand observed using i-view-R software (J. Morita, Kyoto,Japan). Independently, using the same mCT scanner, iso-lated bone samples at 14 days after fracture were imagedwith extending the exposure time to 120 s, and the gap areawas quantified.

Histological analysis

Tibias were fixed in 4% freshly prepared para-formaldehyde for 7 days, decalcified in 0.5 M EDTA for2 wk, and embedded in paraffin. Sections 4 mm thick werestained with van Gieson (n = 5 for Wt, n = 3 for Tg), Alcianblue-Nuclear fast red (n = 5 for Wt, n = 8 for Tg), or forTRACP activity (n = 5 for Wt, n = 8 for Tg) using a kit(Sigma, St. Louis, MO, USA) with 20 mM tartrate. Thearea of Alcian blue–positive chondrocytes and the num-bers of TRACP+ osteoclasts were measured in a longitu-dinal section of each mouse. Four windows of 625 (cross) 3

1250 mm (longitudinal) were placed above and below thefracture line as well as to the left and right just outside thecortical bone and were analyzed under a microscope(Axiovert 135; Carl Zeiss, Jena, Germany). The data fromthe four windows were averaged for each animal. Osteocalcin-positive osteoblasts were stained with rabbit anti-osteocalcinantibody (Alexis, San Diego, CA, USA) using Alexa Fuor647-labeled anti-rabbit antibody (Invitrogen, Carlsbad, CA,USA) (n = 5 for Wt, n = 8 for Tg). Each 625 3 1250-mmwindow (see above) was further divided into 0- to 250- and250- to 1250-mm areas from the fracture line and was an-alyzed using a laser scanning confocal microscope (LSM510 META; Carl Zeiss). Quantification was performedusing Image J software (NIH). Fra-1–positive cells werestained with rabbit anti-Fra-1 antibody (Santa Cruz Bio-technology, Santa Cruz, CA, USA) using Alexa Fuor 647–labeled anti-rabbit antibody (Invitrogen). The slide wasmounted with Vectashield containing DAPI (Vector,Burlingame, CA, USA).

Dynamic histomorphometry

Mice were injected with 100 mg/kg xylenol orange(Sigma) at 4 days before death and 16 mg/kg calcein (WakoPure Chemicals, Tokyo, Japan) at 1 day before death.Bones were fixed in 4% paraformaldehyde and embeddedin methylmethacrylate (MMA) as described.(17) Five-micrometer sections were coverslipped unstained for analysis.Histomorphometry was carried out using an Osteomeasuresystem (OsteoMetrics, Atlanta, GA, USA).

FRA-1 SUPPRESSES FRACTURE UNION 2057

Quantitative RT-PCR

Mice were killed, and the fractured legs after removal ofskin were frozen in liquid nitrogen and cut axially 1 mmabove and below the fracture plane (n = 3 for Wt, and n = 4for Tg). Muscles were included for analysis at day 1 andexcluded for analysis at day 7. Total RNAs were extractedusing ISOGEN (Nippon Gene, Toyama, Japan). cDNAwas synthesized from 1–2 mg total RNA using randomhexamer (Qiagen, Valenca, CA, USA) with PrimeScript(Takara Bio, Tokyo, Japan). Real-time PCR was per-formed on an ABI Prism 7000 machine (Applied Bio-systems, Foster City, CA, USA). mRNA levels werequantified using a standard curve generated with seriallydiluted plasmids containing the PCR amplicon and nor-malized to Gapdh expression. Experiments were repeatedthree times with three independent cDNA syntheses.

ELISA

Mice were anesthetized, and blood was collected (n = 3for each genotype for transverse fractures, n = 4 for closedfractures). Serum was prepared by removing blood clot bycentrifugation. Serum PGE2 levels were analyzed in du-plicate using an ELISA kit (Parameter Prostaglandin E2

Immunoassay; R&D Systems, Minneapolis, MN, USA).

Luciferase assay

Transfection was performed with 0.25 mg of the 0.9-kbCox2 promoter-luciferase pGL3 construct (2891 ; +9)(18)

and 0.05 mg of pB-Actin-RL (Promega, Madison, WI, USA)using FuGENE 6 (Roche, Mannheim, Germany) into bonemarrow–derived macrophage-colony stimulating factor (M-CSF)–dependent macrophages, which were prepared asdescribed previously.(19) Twenty-four hours later, cells werestimulated with various concentrations of TNF-a (R&DSystems) for 72 h. Cell lysates were prepared and analyzedusing a Dual Luciferase Kit (Promega) with a MicroLumatPlus microwell plate luminometer (LB96V; Berthold Tech-nologies, Bad Wildbad, Germany) in triplicate.

Skeletal preparation

Embryos at 14.5 days postcoitum (d.p.c.) were fixed in100% ethanol for 4 days and transferred to acetone (n = 6for each genotype). Staining with Alizarin red S (Sigma)and Alcian blue 8GX (Sigma) was performed as describedpreviously.(20)

Administration of PGE2 pellets

PGE2 and control pellets for long-term continuous re-lease (0.1 mg/d for 21 days) were purchased from InnovativeResearch of America (Rockville, MD, USA). Each pelletwas inserted between the tibia and fibula adjacent to thefracture site during osteotomy (n = 24 for each genotype).

Statistical analysis

All data were expressed as the means ± SE. The statis-tical significance of differences between values was evalu-ated by Student’s t-test.

RESULTS

Time course of fracture healing in Fra-1 Tg mice

A transverse fracture of the tibial diaphysis was createdin 6-wk-old wildtype and Fra-1 Tg mice, and the fracturewas immobilized by insertion of an intramedullary rod.Figure 1A depicts the representative radiographs showingthe fracture healing in this model. In wildtype mice, a cal-cified callus was apparent at 14 days, and bone union wasconfirmed at 21 days. In contrast, callus formation over thefracture site was much less prominent and fracture unionwas still not observed by day 28 in Fra-1 Tg mice (Fig. 1A).Impaired fracture healing in Fra-1 Tg mice was less evidentin healing of the fibula, which was often accidentally frac-tured during surgical manipulation of the tibia. As was thecase for the open fracture, a closed fracture model alsoshowed impaired fracture union in Fra-1 Tg mice (Fig. 1B).These data indicate that Fra-1 Tg mice cannot efficientlyinitiate callus formation over the fracture site.

Endochondral ossification during fracture healing

We next analyzed histological sections of the transversefracture in wildtype and Fra-1 Tg mice at 7, 14, 21, and 28days after surgery. Staining of collagen-containing boneand osteoid with van Gieson picro-fuchsin showed thatmembranous ossification was induced efficiently around theperiosteum near the fracture site in both wildtype and Fra-1Tg mice by 14 days after surgery (Fig. 2A). However, Al-cian blue staining showed that cartilage formation wasmuch less prominent at the fracture site in Fra-1 Tg than inwildtype mice (Fig. 2B). The space around the fracture sitecontaining chondrocytes in wildtype mice was often devoidof chondrocytes and was occupied by fibroblastic cells inFra-1 Tg mice (Fig. 2C), indicating an impairment of frac-ture-induced chondrogenic differentiation in Fra-1 Tg mice.

Twenty-eight days after fracture, the fracture gap wasconnected by massive accumulation of bony callus inwildtype mice, and most of the cartilaginous callus hadbeen resorbed (Figs. 2A and 2B). In contrast, in Fra-1 Tgmice, the fracture gap was still not connected by callus atday 28 after fracture in the presence of chondrocytes withirregular morphology. These observations indicate thatfracture unions that depend on endochondral ossificationare impaired in Fra-1 Tg mice.

Fra-1 expression and bone cells aroundthe fracture site

We analyzed the expression of Fra-1 at the fracture siteby immunofluorescence microscopy. In wildtype mice, Fra-1expression in the callus tissue was relatively low and wasbarely detectable in chondrocytes (Fig. 3A). In contrast, inFra-1 Tg mice, Fra-1 was highly expressed in various celltypes around the fracture site including fibroblastic pro-genitor cells (Figs. 3B and 3C). Quantification of the callusareas occupied by chondrocytes in histological sectionsshowed that the cartilaginous area peaked at day 14 inwildtype mice, whereas chondrogenesis was impaired inFra-1 Tg mice (Fig. 4A). In the same area, the number ofTRACP+ osteoclasts/chondroclasts also increased over the

2058 YAMAGUCHI ET AL.

28-day measurement period in wildtype mice, with a peakat day 14, whereas it was reduced in Fra-1 Tg mice (Fig. 4B).In the central region of the callus at the fracture line(0–250 mm), the number of osteocalcin-positive osteo-blasts was increased during 14–28 days in wildtype mice butfailed to increase in Fra-1 Tg mice, consistent with an

impairment of endochondral ossification near the fracturesite (Fig. 4C). However, more peripheral to the fractureline (250–1000 mm), the number of osteoblasts wascomparable between the two genotypes. Moreover, thebone formation rate was significantly increased in the bonycalluses of Fra-1 Tg mice compared with those of thewildtype controls, indicating that enhanced bone forma-tion, which is the osteoblast cell-autonomous phenotype inFra-1 Tg mice,(10) is consistently observable during fracturehealing.

Reduced inflammatory responses in Fra-1 Tg mice

Because inflammation precedes chondrogenic callusformation during fracture healing, we compared bonefracture–induced inflammation between wildtype and Fra-1 Tg mice. Twenty-four hours after osteotomy, we analyzedexpression of the pro-inflammatory cytokines Tnfa and Il6(encoding TNF-a and IL-6, respectively) at the fracture

FIG. 1. Time course of fracture healing in Fra-1 Tg mice. (A)Open fracture. Representative radiographic images of tibias at theindicated numbers of days after transverse fracture generation inwildtype and Fra-1 Tg mice (n = 13 for each genotype). Note thatthe intramedullary rod was removed by day 28. (B) Closed fracturein wildtype and Fra-1 Tg mice (n = 4 for each genotype). Fracturelines are indicated by arrowheads.

FIG. 2. Histological analysis of fracture healing in Fra-1 Tg mice.(A) Bone (collagen) was stained red in paraffin sections of tibialfractures (van Gieson). Wt, n = 5; Tg, n = 3. Arrowheads indicatecallus formed by membranous ossification. Scale bar, 500 mm. (B)Cartilage was stained blue in adjacent sections (Alcian blue). Wt,n = 5; Tg, n = 8. Scale bar, 500 mm. (C) Higher magnification ofthe area indicated in the box in B. Scale bar, 200 mm.

FRA-1 SUPPRESSES FRACTURE UNION 2059

site. Tnfa and Il6 expression was efficiently induced bybone fracture in wildtype but not in Fra-1 Tg mice (Figs.5A and 5B). We also determined the levels of Cox2 (cy-clooxygenase 2 [Cox-2]), which is responsible for synthesisof PGE2, an inflammatory mediator functioning in fracturehealing.(21,22) Cox2 expression around the fracture site wasinduced in both wildtype and Fra-1 Tg mice but was sig-nificantly reduced in Fra-1 Tg mice (Fig. 5C). At 24 h aftersurgery, the mRNA levels of NF-kB components (Nfkb1,Nfkb2, Rela), Hif1a, Rankl, Sox9, Col2a (type II collagen),Bmp2 (bone morphogenetic protein-2), and Tgfb1 (TGF-b1) were comparable between wildtype and Fra-1 Tg mice(Figs. 5D–5G, 5I, and 5K–5N). Curiously, expression ofangiogenic Vegfa (vascular endothelial growth factor A)was reduced (Fig. 5H) and anti-osteoclastogenic Opg (os-teoprotegerin) was increased in Fra-1 Tg mice in responseto surgery (Fig. 5J). At present, it is unclear whether al-terations in Vegfa or Opg expression participate in theregulation of chondrogenesis around the fracture site.Because it is known that PGE2 is needed for fracturehealing,(21,22) we further analyzed the serum levels ofPGE2. Two days after surgery, serum PGE2 levels werelower in Fra-1 Tg mice than in wildtype mice (Fig. 5O). Inthe closed fracture model using a guillotine-like device,(16)

we also found that serum PGE2 levels were dramatically

lower in Fra-1 Tg mice even before fracture and remainedlow for 8 days after fracture (Fig. 5P).

To examine Cox2 promoter activity in the presence ofFra-1, we transiently transfected a Cox2-luciferase constructinto macrophages prepared from wildtype and Fra-1 Tg miceand added increasing concentrations of TNF-a to culturesto stimulate the promoter activity. The basal Cox2 promoteractivity was higher in wildtype than in Fra-1 Tg macro-phages, and TNF-a treatment further augmented the activityin wildtype but not in Fra-1 Tg macrophages, suggestingthat Fra-1 represses Cox2 promoter activity (Fig. 5Q).

Reduced expression of chondrogenic factors inFra-1 Tg mice

Fourteen days after fracture, expression of inflammatorymediators such as Il6 and Cox2 remained reduced in Fra-1

FIG. 3. Expression of Fra-1 at the fracture site. (A) Alcian bluestaining and immunofluorescence detection of Fra-1 expression inthe callus 14 days after fracture in wildtype mice (n = 5). Scale bar,100 mm. (B) Analysis of Fra-1 Tg mice (n = 3) as in (A). (C) Highermagnification of the area indicated in the box in B. Scale bar, 10 mm.

FIG. 4. Bone parameters in the tibial callus. (A) Quantificationof chondrocyte (Ch) area Wt, n = 3; Tg, n = 3. (B) Number ofTRACP+ osteoclasts Wt, n = 3; Tg, n = 3. (C) Number of osteo-calcin-positive osteoblasts. Wt, n = 3; Tg, n = 3. (D) Double fluo-rochrome labeling. Xylenol orange (red) and calcein (green) wereinjected intraperitoneally on days 10 and 13, respectively, and thetissues were collected on day 14 after surgery. Scale bar, 100 mm.Bone formation rates (BFR) were measured. Wt, n = 4; Tg, n = 4.*p < 0.05, **p < 0.01.

2060 YAMAGUCHI ET AL.

Tg mice, although expression of Tnfa was comparable be-tween wildtype and Fra-1 Tg mice (Figs. 6A–6C). At thistime point, there were no significant differences in theexpression of Nfkb1, Nfkb2, Rela, Hif1a, Vegfa, Rankl, andOpg between the two genotypes (Figs. 6D–6J). However,induction of both chondrogenic factor Sox9 and its targetCol2a1 was significantly lower in tissues containing thefracture callus in Fra-1 Tg mice compared with controls(Figs. 6K and 6L). We next studied expression of Bmp2around the fracture site. Bmp2 stimulates both Sox9 ex-pression and fracture healing.(23–25) Whereas wildtypecontrol mice showed elevated Bmp2 expression near thefracture, Bmp2 expression was not significantly elevated inFra-1 Tg mice and remained as low as that seen in the

wildtype, unfractured controls (Fig. 6M). Tgfb1 expression,on the other hand, was not significantly induced in thisfracture model in either wildtype or Fra-1 Tg mice (Fig.6N). These data suggest that the impairment of chondro-genic callus formation is caused by reduced BMP-2 in-duction in Fra-1 Tg mice.

Normal developmental chondrogenesis inFra-1 Tg mice

Embryonic chondrogenesis occurs in the absence of in-flammation. To evaluate the effect of Fra-1 overexpressionduring embryonic chondrogenesis, wildtype and Fra-1 Tgembryos were stained with Alcian blue. At 14.5 d.p.c.,there were no gross differences in chondrogenesis between

FIG. 5. Expression of pro-inflammatory mediators andregulators of osteochondraldifferentiation 24 h after frac-ture. Expression levels of (A)Tnfa, (B) Il6, (C) Cox2, (D)Nfkb1, (E) Nfkb2, (F) Rela,(G) Hif1a, (H) Vegfa, (I)Rankl, (J) Opg, (K) Sox9, (L)Col2a, (M) Bmp2, and (N)Tgfb1 at the tibial diaphysis24 h after osteotomy (Fracture+) or in controls (Fracture 2).n = 3 for Wt; n = 4 for Fra-1 Tgmice. The mRNA levels (copynumber) were normalized tothose of Gapdh. (O) SerumPGE2 levels. Two days afteropen tibial fracture. n = 3 forWt; n = 3 for Fra-1 Tg mice perpoint. (P) Time course afterclosed tibial fracture generatedwith a guillotine-like device.n = 4 per data point. (Q) Cox2promoter activity in Wt andFra-1 Tg macrophages stimu-lated with the indicated TNF-aconcentrations for 72 h beforeharvest. Cultures were assayedin triplicate wells. *p < 0.05,**p < 0.01.

FRA-1 SUPPRESSES FRACTURE UNION 2061

embryos of wildtype and Fra-1 Tg mice (Fig. 6O). Growthplates of the proximal tibias at postnatal day 3 were alsoindistinguishable between wildtype and Fra-1 Tg mice (Fig.6P). Because the Fra-1 transgene expression is activatedembryonically(10) and developmental chondrogenesis wasapparently intact, the impaired chondrogenic callus for-mation observed in Fra-1 Tg mice was likely limited toinflammation-induced chondrocyte differentiation.

Improvement of fracture healing by PGE2

We administered PGE2 to the fracture site using pelletsthat allow long-term continuous release (0.1 mg/d for 21days) and examined whether such treatment amelioratedthe impaired healing in Fra-1 Tg mice. Fracture healing was

monitored using in vivo mCT imaging (Fig. 7A). We his-tologically examined the gap between calluses above andbelow the fracture line after PGE2 treatment in wildtypeand Fra-1 Tg mice. Alcian blue staining showed that the gapin Fra-1 Tg mice treated with PGE2 contains cartilage (Fig.7B). Furthermore, the gap area was significantly reduced inFra-1 Tg mice treated with PGE2 (Fig. 7C). Therefore,PGE2 administration restored cartilage formation and sig-nificantly reduced the fracture gap in Fra-1 Tg mice.

DISCUSSION

Here we report that endochondral bone repair is impairedin mice broadly overexpressing the transcription factor

FIG. 6. Expression of pro-inflammatory mediators andregulators of osteochondraldifferentiation 7 days afterfracture. Fracture-induced ex-pression. Expression levels of(A) Tnfa, (B) Il6, (C) Cox2,(D) Nfkb1, (E) Nfkb2, (F)Rela, (G) Hif1a, (H) Vegfa, (I)Rankl, (J) Opg, (K) Sox9, (L)Col2a, (M) Bmp2, and (N)Tgfb1 in tibial diaphysis werequantified by qRT-PCR forwildtype (Wt, n = 4) and Fra-1Tg mice (n = 5) 7 days afterosteotomy (Fracture +) or incontrols (Fracture 2). ThemRNA levels (copy number)were normalized to those ofGapdh. *p < 0.05, **p < 0.01.(O) Whole mount preparationsof 14.5 d.p.c. wildtype and Fra-1 Tg embryos stained for car-tilage (Alcian blue). Note thatembryos at this developmentalstage are negative for Alizarinred staining (calcified bones).(P) Longitudinal section oftibial growth plates of 3-day-old mice stained with Alcianblue.

2062 YAMAGUCHI ET AL.

Fra-1/AP-1. In Fra-1 Tg mice, fusion of calluses over thefracture line is disrupted by fibrous tissues, which otherwiseshould differentiate into chondrocytes to initiate endo-chondral bone repair. Curiously, the lack of chondrogenesisover the fracture line is associated with reduced in-flammatory responses around the fracture site. An intrinsicdefect in chondrocytes is unlikely because both wholemount staining of embryos and histological analysis ofgrowth plates in postnatal Fra-1 Tg mice showed no grossabnormality. Furthermore, the morphology of growth platesin adult Fra-1 Tg mice was normal.(10) Collectively, theseresults suggest that the chondrogenic potential of mesen-chymal progenitors during development is likely intact inFra-1 Tg mice. Therefore, inflammation-induced chondro-genesis may be specifically impaired in Fra-1 Tg mice.

After fracture, mesenchymal progenitors derived fromthe bone marrow or the periosteum near the fracture are

believed to differentiate into chondrocytes or osteoblasts.Accumulation of Fra-1–positive fibroblastic cells was ob-served at the site of fracture in Fra-1 Tg mice, suggestingthat recruitment of chondroprogenitors per se is intact.When mesenchymal progenitors fail to initiate endochon-dral bone repair, these cells persistently occupy the spacearound the fracture line and interfere with bony fusion ofcalluses across the fracture line.

There are other mouse models showing delayed ornonunion bone fractures. For example, unlike Fra-1 Tgmice, mice lacking placental growth factor (PlGF) exhibita massive accumulation of cartilage in the calluses.(2) Cu-riously, these mice show reduced inflammatory reaction totrauma in the absence of PlGF. Moreover, fibrous tissuewas observed at the fracture line, discontinuing the bonybridge in two thirds of these mice.

In Fra-1 Tg mice, production of inflammatory mediatorssuch as TNF-a, IL-6, and PGE2 was reduced, presumablyin both chondrogenic progenitors and macrophages, duringfracture healing (Fig. 8). Accumulating lines of evidencesuggest that Fos proteins such as c-Fos and Fra-1 suppressthe production of inflammatory mediators in response toToll-like receptor signaling.(8,26) Several laboratories haveanalyzed bone fracture healing in the absence of pro-inflammatory cytokine signaling. For example, mice lackingTNF receptors (p552/2/p752/2) have been shown to ex-hibit delayed endochondral tissue formation.(5,27) How-ever, direct regulation of chondrogenesis by inflammatorycytokines may be unlikely, because inflammatory cytokinessuch as TNF-a and IL-1b suppress expression of chon-drogenic Sox9.(28) Interestingly, we observed that BMP-2mRNA was induced at fracture sites in wildtype but not inFra-1 Tg mice. Consistent with this observation, it has beenreported that inflammatory cytokines induce BMP-2 ex-pression in chondrocytes and endothelial cells(29–31) andthat BMP-2 is absolutely needed to initiate normal frac-ture healing.(25) Therefore, suppression of inflammatorycytokine expression may explain the reduced BMP-2 ex-pression in cells located around the fracture site in the Fra-1 Tg mice and the resulting impaired endochondral frac-ture healing (Fig. 8).

The rate-limiting step in prostaglandin synthesis is cat-alyzed by Cox-2, which is inducible by pro-inflammatorycytokines and other stresses. We showed that both Cox2expression and serum PGE2 levels were significantly lowerin Fra-1 Tg mice than in wildtype controls. It has beenreported that mice lacking Cox2 show delayed fracturerepair,(21,22) that mice lacking the PGE2 receptor EP4 showfracture nonunion,(32) that an EP2 receptor-selective ago-nist enhances fracture healing in rat models,(33) and thatCox-2/EP4 agonists enhanced chondrogenesis and boneformation in aged mice.(34) Therefore, we examinedwhether PGE2 administration would enhance fractureunion in Fra-1 Tg mice. We found that the fracture gap wasreduced over the fracture line by PGE2 administration, andchondrogenesis was partially restored near the fracture sitein Fra-1 Tg mice, suggesting that PGE2 is necessary toinitiate chondrogenic callus formation. Although we ob-served reduced numbers of osteoclasts in the callus, thisreduction correlated with the reduced cartilaginous area

FIG. 7. Effects of PGE2 administration on fracture healing inFra-1 Tg mice. (A) In vivo mCT images of callus formation andunion over 28 days after fracture (n = 4 for each group). PGE2 slowreleasing and control pellets (*) were placed between the tibia andfibula near the fracture site during osteotomy. (B) RepresentativemCT images and Alcian blue staining of the gap area (white ar-rowheads) in isolated bone samples from PGE2-treated Wt and Tgmice 14 days after fracture. (C) The gap areas between callusesabove and below the fracture line were quantified based on themCT images on isolated bone specimens 14 days after fracture (n =4 per each data point). *p < 0.05. Scale bar, 500 mm.

FRA-1 SUPPRESSES FRACTURE UNION 2063

(Figs. 3A and 3B). Therefore, we think that the reducedosteoclast numbers may be secondary to the reduced car-tilaginous area.

We did not analyze hypoxia in this study, but bothhypoxia and inflammation are early physiological re-sponses to bone fracture. Disruption of the vascular supplyto the fracture site and resultant hypoxia accounts for theincreased numbers of chondrocytes and formation ofa cartilaginous callus. Indeed, hypoxia induces transcrip-tion of Vegf and Sox9 through HIF-1a activation, andVEGF enhances fracture repair.(35–37) Similarly, HIF-1a

activation accelerates bone regeneration in response todistraction osteogenesis.(38) Therefore, the link betweenangiogenesis and reduced inflammation in Fra-1 Tg mice,especially the regulation of VEGF, should be analyzed inthe future.

This study showed a critical role of endochondral ossi-fication in bridging calluses across the fracture line. Evenwhen membranous ossification is highly active in Fra-1 Tgmice as shown by an increase in the bone formation rate inthe bony callus, lack of efficient chondrogenesis results inimpaired formation of bony callus over the fracture.Importantly, our results suggest that induction of chon-drogenesis during the earliest stage of fracture healing iscritically associated with inflammatory responses regulatedby AP-1. In conclusion, the osteogenic transcription factorFra-1 suppresses production of inflammatory mediatorsduring early stages of fracture repair. These inflammatorymediators may be critical to initiate endochondral ossifi-cation after bone fracture. Clinically, development of newapproaches to improve fracture healing and to treat

osteoporosis by increasing bone mass is needed.(39) Ourresults show a previously undescribed role for Fra-1/AP-1in regulating fracture-induced endochondral ossification.

ACKNOWLEDGMENTS

The authors thank Yuko Takagaki for advice on in vivoexperiments, Mayako Asakawa, Sayaka Fukuse, andEmiko Kobayashi for technical help, and Erwin F. Wagnerfor provision of the Fra-1 Tg mice. This work was fundedby Grants-in-Aid for Exploratory Research (17659479)and for Scientific Research (B) (19390399) from JSPS.

REFERENCES

1. Gerstenfeld LC, Einhorn TA 2006 Fracture healing: The bi-ology of bone repair and regeneration. In: Favus MJ (ed.)Primer on the Metabolic Bone Diseases and Disorders ofMineral Metabolism, 6th ed. American Society for Bone andMineral Research, Washington, DC, USA, pp. 42–48.

2. Maes C, Coenegrachts L, Stockmans I, Daci E, Luttun A,Petryk A, Gopalakrishnan R, Moermans K, Smets N, VerfaillieCM, Carmeliet P, Bouillon R, Carmeliet G 2006 Placentalgrowth factor mediates mesenchymal cell development, carti-lage turnover, and bone remodeling during fracture repair. JClin Invest 116:1230–1242.

3. Vortkamp A, Pathi S, Peretti GM, Caruso EM, Zaleske DJ,Tabin CJ 1998 Recapitulation of signals regulating embryonicbone formation during postnatal growth and in fracture repair.Mech Dev 71:65–76.

4. Ferguson C, Alpern E, Miclau T, Helms JA 1999 Does adultfracture repair recapitulate embryonic skeletal formation?Mech Dev 87:57–66.

5. Gerstenfeld LC, Cho TJ, Kon T, Aizawa T, Tsay A, Fitch J,Barnes GL, Graves DT, Einhorn TA 2003 Impaired fracturehealing in the absence of TNF-a signaling: The role of TNF-ain endochondral cartilage resorption. J Bone Miner Res18:1584–1592.

6. Wagner EF, Eferl R 2005 Fos/AP-1 proteins in bone and theimmune system. Immunol Rev 208:126–140.

7. Matsuo K, Ray N 2004 Osteoclasts, mononuclear phagocytes,and c-Fos: New insight into osteoimmunology. Keio J Med53:78–84.

8. Ray N, Kuwahara M, Takada Y, Maruyama K, Kawaguchi T,Tsubone H, Ishikawa H, Matsuo K 2006 c-Fos suppressessystemic inflammatory response to endotoxin. Int Immunol18:671–677.

9. Maruyama K, Sano G, Ray N, Takada Y, Matsuo K 2007 c-Fos-deficient mice are susceptible to Salmonella enterica se-rovar Typhimurium infection. Infect Immun 75:1520–1523.

10. Jochum W, David JP, Elliott C, Wutz A, Plenk H Jr, MatsuoK, Wagner EF 2000 Increased bone formation and osteo-sclerosis in mice overexpressing the transcription factor Fra-1.Nat Med 6:980–984.

11. Eferl R, Hoebertz A, Schilling AF, Rath M, Karreth F, KennerL, Amling M, Wagner EF 2004 The Fos-related antigen Fra-1is an activator of bone matrix formation. EMBO J 23:2789–2799.

12. Sabatakos G, Sims NA, Chen J, Aoki K, Kelz MB, Amling M,Bouali Y, Mukhopadhyay K, Ford K, Nestler EJ, Baron R2000 Overexpression of DFosB transcription factor(s) in-creases bone formation and inhibits adipogenesis. Nat Med6:985–990.

13. Roschger P, Matsuo K, Misof BM, Tesch W, Jochum W,Wagner EF, Fratzl P, Klaushofer K 2004 Normal mineraliza-tion and nanostructure of sclerotic bone in mice over-expressing Fra-1. Bone 34:776–782.

14. Shimoaka T, Kamekura S, Chikuda H, Hoshi K, Chung UI,Akune T, Maruyama Z, Komori T, Matsumoto M, Ogawa W,

FIG. 8. A model for the impaired fracture union in Fra-1 Tgmice. On fracture, chondrogenic progenitors are recruited to thefracture site. Fra-1/AP-1 suppresses transcription of inflammatorymediators such as Tnfa (TNF-a) and Cox2 in macrophages andosteochondroprogenitors. Reduced expression of Cox2 results inimpaired production of PGE2. The periosteal cells at the fracturesite produce a reduced amount of BMP-2 in the presence of in-sufficient amounts of inflammatory mediators. BMP-2 otherwiseinduces chondrogenesis to initiate endochondral bone repair. InFra-1 Tg mice, chondrogenesis is impaired and fibrous tissues oc-cupy the periosteal space around the fracture site, interfering withendochondral fracture union.

2064 YAMAGUCHI ET AL.

Terauchi Y, Kadowaki T, Nakamura K, Kawaguchi H 2004Impairment of bone healing by insulin receptor substrate-1deficiency. J Biol Chem 279:15314–15322.

15. Nagy A, Gertsenstein M, Vintersten K, Behringer R 2003Buffers and Solutions (Appendix 1) Manupulating the MouseEmbryo: A Laboratory Manual, 3rd ed. Cold Spring HarborLaboratory Press, Cold Spring Harbor, NY, USA.

16. Bonnarens F, Einhorn TA 1984 Production of a standard closedfracture in laboratory animal bone. J Orthop Res 2:97–101.

17. Sims NA, Clement-Lacroix P, Da Ponte F, Bouali Y, Binart N,Moriggl R, Goffin V, Coschigano K, Gaillard-Kelly M, KopchickJ, Baron R, Kelly PA 2000 Bone homeostasis in growth hormonereceptor-null mice is restored by IGF-I but independent of Stat5.J Clin Invest 106:1095–1103.

18. Saunders MA, Sansores-Garcia L, Gilroy DW, Wu KK 2001Selective suppression of CCAAT/enhancer-binding protein bbinding and cyclooxygenase-2 promoter activity by sodiumsalicylate in quiescent human fibroblasts. J Biol Chem 276:18897–18904.

19. Zhao C, Irie N, Takada Y, Shimoda K, Miyamoto T, NishiwakiT, Suda T, Matsuo K 2006 Bidirectional ephrinB2-EphB4signaling controls bone homeostasis. Cell Metab 4:111–121.

20. Komori T, Yagi H, Nomura S, Yamaguchi A, Sasaki K,Deguchi K, Shimizu Y, Bronson RT, Gao YH, Inada M, SatoM, Okamoto R, Kitamura Y, Yoshiki S, Kishimoto T 1997Targeted disruption of Cbfa1 results in a complete lack ofbone formation owing to maturational arrest of osteoblasts.Cell 89:755–764.

21. Simon AM, Manigrasso MB, O’Connor JP 2002 Cyclo-oxygenase 2 function is essential for bone fracture healing.J Bone Miner Res 17:963–976.

22. Zhang X, Schwarz EM, Young DA, Puzas JE, Rosier RN,O’Keefe RJ 2002 Cyclooxygenase-2 regulates mesenchymalcell differentiation into the osteoblast lineage and is criticallyinvolved in bone repair. J Clin Invest 109:1405–1415.

23. Uusitalo H, Hiltunen A, Ahonen M, Gao TJ, Lefebvre V,Harley V, Kahari VM, Vuorio E 2001 Accelerated up-regulationof L-Sox5, Sox6, and Sox9 by BMP-2 gene transfer duringmurine fracture healing. J Bone Miner Res 16:1837–1845.

24. Kugimiya F, Kawaguchi H, Kamekura S, Chikuda H, Ohba S,Yano F, Ogata N, Katagiri T, Harada Y, Azuma Y, NakamuraK, Chung UI 2005 Involvement of endogenous bone mor-phogenetic protein (BMP) 2 and BMP6 in bone formation.J Biol Chem 280:35704–35712.

25. Tsuji K, Bandyopadhyay A, Harfe BD, Cox K, Kakar S,Gerstenfeld L, Einhorn T, Tabin CJ, Rosen V 2006 BMP2activity, although dispensable for bone formation, is requiredfor the initiation of fracture healing. Nat Genet 38:1424–1429.

26. Morishita H, Saito F, Kayama H, Atarashi K, Kuwata H,Yamamoto M, Takeda K 2009 Fra-1 negatively regulateslipopolysaccharide-mediated inflammatory responses. IntImmunol 21:457–465.

27. Lehmann W, Edgar CM, Wang K, Cho TJ, Barnes GL, KakarS, Graves DT, Rueger JM, Gerstenfeld LC, Einhorn TA 2005Tumor necrosis factor alpha (TNF-a) coordinately regulatesthe expression of specific matrix metalloproteinases (MMPS)and angiogenic factors during fracture healing. Bone 36:300–310.

28. Murakami S, Lefebvre V, de Crombrugghe B 2000 Potentinhibition of the master chondrogenic factor Sox9 gene by

interleukin-1 and tumor necrosis factor-a. J Biol Chem275:3687–3692.

29. Feng JQ, Xing L, Zhang JH, Zhao M, Horn D, Chan J, BoyceBF, Harris SE, Mundy GR, Chen D 2003 NF-kB specificallyactivates BMP-2 gene expression in growth plate chondrocytesin vivo and in a chondrocyte cell line in vitro. J Biol Chem278:29130–29135.

30. Csiszar A, Smith KE, Koller A, Kaley G, Edwards JG,Ungvari Z 2005 Regulation of bone morphogenetic protein-2 expression in endothelial cells: Role of nuclear factor-kBactivation by tumor necrosis factor-a, H2O2, and high in-travascular pressure. Circulation 111:2364–2372.

31. Fukui N, Ikeda Y, Ohnuki T, Hikita A, Tanaka S, Yamane S,Suzuki R, Sandell LJ, Ochi T 2006 Pro-inflammatory cytokinetumor necrosis factor-a induces bone morphogenetic protein-2in chondrocytes via mRNA stabilization and transcriptionalup-regulation. J Biol Chem 281:27229–27241.

32. Li M, Healy DR, Li Y, Simmons HA, Crawford DT, Ke HZ,Pan LC, Brown TA, Thompson DD 2005 Osteopenia andimpaired fracture healing in aged EP4 receptor knockoutmice. Bone 37:46–54.

33. Li M, Ke HZ, Qi H, Healy DR, Li Y, Crawford DT, ParalkarVM, Owen TA, Cameron KO, Lefker BA, Brown TA,Thompson DD 2003 A novel, non-prostanoid EP2 receptor-selective prostaglandin E2 agonist stimulates local bone for-mation and enhances fracture healing. J Bone Miner Res18:2033–2042.

34. Naik AA, Xie C, Zuscik MJ, Kingsley P, Schwarz EM, AwadH, Guldberg R, Drissi H, Puzas JE, Boyce B, Zhang X,O’Keefe RJ 2009 Reduced COX-2 expression in aged mice isassociated with impaired fracture healing. J Bone Miner Res24:251–264.

35. Wang Y, Wan C, Deng L, Liu X, Cao X, Gilbert SR, BouxseinML, Faugere MC, Guldberg RE, Gerstenfeld LC, Haase VH,Johnson RS, Schipani E, Clemens TL 2007 The hypoxia-inducible factor a pathway couples angiogenesis to osteogenesisduring skeletal development. J Clin Invest 117:1616–1626.

36. Robins JC, Akeno N, Mukherjee A, Dalal RR, Aronow BJ,Koopman P, Clemens TL 2005 Hypoxia induces chondrocyte-specific gene expression in mesenchymal cells in associationwith transcriptional activation of Sox9. Bone 37:313–322.

37. Street J, Bao M, deGuzman L, Bunting S, Peale FV Jr, FerraraN, Steinmetz H, Hoeffel J, Cleland JL, Daugherty A, vanBruggen N, Redmond HP, Carano RA, Filvaroff EH 2002Vascular endothelial growth factor stimulates bone repair bypromoting angiogenesis and bone turnover. Proc Natl AcadSci USA 99:9656–9661.

38. Wan C, Gilbert SR, Wang Y, Cao X, Shen X, Ramaswamy G,Jacobsen KA, Alaql ZS, Eberhardt AW, Gerstenfeld LC,Einhorn TA, Deng L, Clemens TL 2008 Activation of thehypoxia-inducible factor-1a pathway accelerates bone re-generation. Proc Natl Acad Sci USA 105:686–691.

39. Khosla S, Westendorf JJ, Oursler MJ 2008 Building bone toreverse osteoporosis and repair fractures. J Clin Invest 118:421–428.

Received in original form September 29, 2008; revised form May21, 2009; accepted June 25, 2009.

FRA-1 SUPPRESSES FRACTURE UNION 2065