Impaired Fracture Healing in the Absence of TNF-α Signaling: The Role of TNF-α in Endochondral Cartilage Resorption


  • The authors have no conflict of interest


TNF-α is a major inflammatory factor that is induced in response to injury, and it contributes to the normal regulatory processes of bone resorption. The role of TNF-α during fracture healing was examined in wild-type and TNF-α receptor (p55−/−/p75−/−)-deficient mice. The results show that TNF-α plays an important regulatory role in postnatal endochondral bone formation.

Introduction: TNF-α is a major inflammatory factor that is induced as part of the innate immune response to injury, and it contributes to the normal regulatory processes of bone resorption.

Methods: The role of TNF-α was examined in a model of simple closed fracture repair in wild-type and TNF-α receptor (p55−/−/p75−/−)-deficient mice. Histomorphometric measurements of the cartilage and bone and apoptotic cell counts in hypertrophic cartilage were carried out at multiple time points over 28 days of fracture healing (n = 5 animals per time point). The expression of multiple mRNAs for various cellular functions including extracellular matrix formation, bone resorption, and apoptosis were assessed (triplicate polls of mRNAs).

Results and Conclusions: In the absence of TNF-α signaling, chondrogenic differentiation was delayed by 2–4 days but subsequently proceeded at an elevated rate. Endochondral tissue resorption was delayed 2–3 weeks in the TNF-α receptor (p55−/−/p75−/−)-deficient mice compared with the wild-type animals. Functional studies of the mechanisms underlying the delay in endochondral resorption indicated that TNF-α mediated both chondrocyte apoptosis and the expression of proresorptive cytokines that control endochondral tissue remodeling by osteoclasts. While the TNF-α receptor ablated animals show no overt developmental alterations of their skeletons, the results illustrate the primary roles that TNF-α function contributes to in promoting postnatal fracture repair as well as suggest that processes of skeletal tissue development and postnatal repair are controlled in part by differing mechanisms. In summary, these results show that TNF-α participates at several functional levels, including the recruitment of mesenchymal stem, apoptosis of hypertrophic chondrocytes, and the recruitment of osteoclasts function during the postnatal endochondral repair of fracture healing.


Fracture healing, like all other wound repair responses, is initiated through the induction of an innate immune response.(1–3) The processes of postnatal fracture healing, unlike the repair of many other connective tissues, does not lead to scar tissue formation but to the full regeneration of the damaged bone.(4) Thus, many of the molecular mechanisms known to regulate skeletal tissue formation during embryological development are recapitulated during postnatal fracture healing.(5) The repair of injured bone, however, requires the coordinated participation of hematopoietic and immune cell types within the marrow space of the injured tissue in conjunction with the vascular and skeletal cell precursors that are recruited from the periosteum and surrounding soft tissues.(4–6)

While TNF-α has been shown to be essential in the mediation of both the innate and acquired immune responses,(7–9) it has also been shown to be involved in the normal mediation of bone repair.(10,11) TNF-α belongs to a very large cytokine superfamily that is currently known to be composed of at least 27 receptors and 18 ligands. The oldest known and most extensively studied members of this cytokine superfamily are TNF and lymphotoxin (LT).(9) This group of TNF family members is made up of TNF-α, TNF-β (LT-α), and LT-β. Both TNF ligands exist as homotrimers, whereas the LT-β exists only as a heterotrimer of LT-α1 and LT-β2. There are three receptors in this family: TNFR1 (p55), TNFR2 (p75), and LT-β receptor. TNF-α and TNF-β homotrimers bind both TNFR1 and TNFR2, but the LT-β/TNF-β heterotrimers only bind to the LT-β receptor. These molecules are induced in response to a multitude of inflammatory stimuli and either activate apoptosis or facilitate cell survival, dependent on the cell type and intracellular pathway that is activated.(12,13) While TNF-α elicits apoptosis through receptor-activated death domain containing second signal proteins, cell survival is mediated through the activation of NFκB or AP1 transcription factors. Thus, the dichotomy of cellular responses induced by TNF-α resides in the specific receptors and their interactions with specific downstream signal transduction pathways.(13)

TNF-α has been extensively studied in bone and cartilage metabolism, and for a number of years it has been implicated in the mediation of osteoclastogenesis.(10,11) The levels of TNF-α production have been shown to be elevated in surgical or natural menopausal states, thus implicating its involvement in controlling coupled bone turnover. Recently, a novel member of the TNF receptor superfamily, RANKL, has been shown to be a key regulator of osteoclastogenesis in conjunction with macrophage-colony stimulating factor (M-CSF). Bone mass has been shown to be regulated through the balanced production of RANKL, which controls osteoclast formation, and its soluble receptor antagonist osteoprotegerin (OPG).(14–16) A number of studies have also shown that TNF-α can potentially induce osteoclast formation either by directly inducing osteoclast formation or through the stimulation of RANKL expression.

In previous studies we demonstrated that the ratio of expression of OPG, RANKL, and M-CSF were tightly coupled during fracture healing and appeared to be involved in the regulation of both endochondral resorption and primary bone remodeling. In the same study, we examined the expression of TNF-α and its receptors and demonstrated that TNF-α and both its receptors were induced at very high levels immediately after injury but showed markedly reduced levels of expression during the period of cartilage formation. However, once cartilage resorption and primary bone remodeling were initiated, TNF-α expression rapidly increased, and during this period, TNF-α protein was localized in mesenchymal cells, osteoblasts, and at very high levels in hypertrophic chondrocytes.(3) In other studies, we have also shown that TNF-α induced hypertrophic chondrocytes to undergo apoptosis.(17) The role of TNF-α in homeostatic bone remodeling and its role in inflammatory response to injury have been examined in numerous laboratories. The cumulative data suggest that TNF-α might carry out important functional processes during bone repair. The focus of the current studies was to define the molecular nature of the functional processes that TNF-α mediates during fracture healing.


Production of simple transverse fractures

Eight- to 10-week-old male transgenic and strain-matched control mice were used for this study. Transgenic mice null for both TNF-α receptors(18) (TNF-α p55−/−p75−/− homozygous mice) and strain-matched controls were from Jackson Laboratories (Bar Harbor, ME, USA). Closed, transverse, mid-diaphyseal fractures of the tibias were generated by controlled blunt trauma using a modification of the technique developed for rats.(19)

Callus cell and articular chondrocyte cultures

Cell populations enriched in growth chondrocytes were prepared from murine fracture calluses at 10 days after injury. Calluses were microdissected, and the areas circumscribed by the callus were isolated under a microdissection scope. The bone was sagitally cut in half, and the underlying cortical bone tissue was removed. Cells were dissociated from the soft calluses by trypsin collagenase treatment for 5 h using the digestion procedures developed for the preparation of avian chondrocytes.(20) Femoral head articular chondrocytes cultures were prepared as described previous.(21) Both types of cells were plated at 2 × 106 cells and grown in DMEM with 10% fetal bovine serum. All measurements of mRNA expression profiles were performed using cultured cells from at least two to three separate preparations. Averaged values from the multiple experiments are presented. TNF-α additions were carried out in primary cultured chondrocytes at 2–4 days after plating. Recombinant mouse TNF-α (PharMingen, San Diego, CA, USA) was added at 0.1 and 1.0 ng/ml. Cells were harvested 24 h after the TNF-α treatments.


Tissue samples were fixed in ice-cold 4% paraformaldehyde for 3 days, followed by decalcification in 14% EDTA for up to two weeks. Pins were then removed from the bones before embedding and sectioning. Fixed and decalcified tissues were dehydrated in graded ethanol up to 100%, transferred to xylene, and embedded in paraffin. Five-micron thin paraffin sections were placed on poly L-lysine coated slides, dried overnight, and either used immediately or stored at 4°C. Sections were stained with Safranin O-Fast Green.(22) Tartrate Resistant Acid Phosphatase (TRACP) enzyme was detected by using an azo-dye coupling method with the slight modification of fast red violet LB salt (Sigma F-3381, Sigma Chemicals, St. Louis, MO) replacing the fast red TR salt.


Three bones each from eight groups (7dKO, 7dWT, 10dKO, 10dWT, 14dKO, 14dWT, 21dKO, 21dWT) were used for histomorphometric analysis. Four serial 5-μm sections from the center of the callus from each bone were stained with Safranin-O fast red green. Each of these sections was photographed with an Olympus BX51 light microscope attached to a digital camera. Placing the callus on a horizontal plane in the center of the field, each photograph was taken at 1.25× and downloaded onto Image-Pro Plus Version for Windows. An area of interest (AOI) was created by loading a uniform box (5 × 7.86365 mm) onto the photograph and centering the callus within the box. The bone was then outlined within the uniform AOI, excluding any muscle, soft tissue, or periosteum. Using a color match program, the total area of the cartilage (red) and bone (green) was quantified using a filter range of 573.921 μm −5.73921e + 03 μm. Specimen means were calculated for the individual bones and then used to create group means, SDs, and SEs for each of the groups.

The number of TRACP+ cells was determined in adjacent serial sections to those that were processed for the quantification of cartilage and bone. For these determinations, three serial 5-μm sections from the center of the callus were stained for TRACP activity. Microscopic images were captured as described above and two microphotographic images (10.5 × 7.8 mm) were taken of contiguous sections at 10× and downloaded onto Image-Pro Plus Version for Windows. The total number of fields counted per time point per group were N = 24. Using a color match program, the osteoclasts from each photograph were quantified using a spot filter using predefined pixel values defining the lower and upper dimensions for the spots that were counted. The numbers from the two adjacent pictures were combined and specimen means were calculated and used to create group means, SDs, and SEs.

TUNEL assay

TUNEL assays were carried out using an ApopTag kit from Intergen Corp. (Purchase, NY, USA) using the modification in the detection method of Xu et al.(23) Images were digitally captured, and quantitative measurements were made from calluses of three animals counting (n = 3) fields for each set of counts taken from a total area of 0.311 mm2. Analysis was performed with Image-Pro Plus software (Media Cybernetics, Silver Spring, MD, USA). The dead cell number was determined to be those cells that were stained for the TUNEL reaction and determined as a percentage of the total number of nuclei that were visualized.

RNA analysis

Animal tissues were collected in duplicate sets of pooled samples (n = 10 for days 1 and 3; n = 5 for days 7, 14, 21, and 28). Bones were retrieved on the indicated days, and the mid-diaphyseal region of the bone that circumscribed the fracture site was excised, and the tissue powdered under liquid nitrogen with a mortar and pestle. Total RNA was extracted from the powder as previously described. RNA quantities were determined by OD260 nm, and sample integrity was monitored by visualization of ribosomal RNAs with ethidium bromide after denaturing RNA gel electrophoresis.(24)

RNase protection analysis

mRNA expression during fracture healing was quantitatively assessed by ribonuclease protection analysis (RPA). Probes for the 5′ splice variants of COL2A1 and murine aggrecan (large proteoglycan) were constructed in our laboratories by selective reverse transcriptase-polymerase chain reaction (RT-PCR) cloning. RT-PCR reaction products were cloned to create the specific probes for aggrecan and type II collagen α 1 splicing variants. Primer sequences were 5′-tgacaacttctttgccaccg-3′, 5′-ctcagcatcctgggcacatta-3′ for aggrecan, and 5′-atgagggagcggtagagac-3′, 5′-tatacctctgcccattctgc-3′ for type II collagen. The cRNA probe for type II collagen α1 chain was designed to discriminate the two splicing variants, IIA and IIB. The probe for TRACP was obtained through a materials transfer agreement with Amgen Inc. (Torre Pines, CA, USA). Remaining linearized probes for the various mouse genes that were assayed in these studies were purchased from PharMingen Corp. (San Diego, CA, USA). Single-stranded32P-labeled cRNA were generated using RNA transcription kits purchased from PharMingen Corp. RNase protection analysis and quantitative analysis were as previously described.(3) All mRNA expression data were normalized to the L32 ribosomal protein mRNA expression and are expressed as a relative percentage.


Absence of TNF-α leads to a delay in endochondral tissue formation

The effects of TNF-α receptor deficiency were first assessed at the histological level (Fig. 1). These results showed a clear pattern of altered fracture repair in the TNF p55−/−-p75−/− receptor-deficient mice compared with the wild-type mice, throughout the 28-day time course of repair. At 7 days after fracture, most of the chondrocytes in the TNF receptor-deficient mice were immature. These mice also lacked formation of new trabecular bone adjacent to the fracture site along the periosteal surface. In contrast, robust bone formation along the periosteal surface and mature cartilage were present in the day 7 wild-type mice. By day 14, the majority of the chondrocytes in callus tissue in the wild-type animals were hypertrophic, and the callus tissue was being resorbed with considerable new bone laid down on areas of mineralized cartilage. At the same time points, the receptor-deficient mice showed far fewer hypertrophic chondrocytes, fewer areas of cartilage undergoing resorption, and large amounts of loose unorganized mesenchymal tissue adjacent to the cartilage tissue (Fig. 1B). By 21 days, all of the cartilage tissue had been resorbed and replaced with trabecular bone and a well-formed marrow was present in the callus tissue of the wild-type mice, yet a clear delay was seen in resorption of the callus tissues within the TNF-α receptor-deficient mice. Residue amounts of cartilage tissue were still evident at this time, but very little mature trabecular bone was observed in the TNF receptor null animals. The transgenic animals displayed poorly formed marrow space within the callus tissues and a persistence of large amounts of undifferentiated mesenchymal cells with a gross enlargement of the calluses. In contrast, bones from the wild-type animals showed remodeling toward their original size (Figs. 1A and 1B). Histomorphometric assessment of the callus tissues validated the qualitative appearance of these tissues and further demonstrated that there was an increased quantity of cartilage tissue in the TNF-α receptor null calluses (Fig. 1C) throughout the period in which primary bone formation was proceeding in the wild-type mice.

Figure FIG. 1..

Histological analysis of fracture healing in wild-type and TNF-α receptor (TNFR)-deficient (p55−/−/p75−/−) mice. (A) Panoramic view of the calluses at 7, 14, and 21 days after fracture in wild-type (WT) and TNFR-deficient (p55−/−p75−/−) mice (sagittal section, safranin-O and fast green, magnification, ×40). *Site of fracture. (B) Representative 7-, 14-, and 21-day calluses showing delay in endochondral progression in the absence of TNF-α signaling (safranin O and fast green, magnification, ×200). Arrows denote areas of cartilage. (C) Histomorphometric assessment of the percentage cartilage.

To determine if TNF-α signaling affected cartilage cell differentiation, a set of extracellular matrix gene products that are associated with chondrogenic differentiation (collagen types II, X, and large proteoglycan/aggrecan) were assessed over the time course of fracture healing (Figs. 2A and B). A clear 3- to 4-day delay and an initial reduced level of expression was seen in the induction of these genes, suggesting that the differentiation of mesenchymal cells to chondrocytes was delayed in the absence of TNF-α signaling. However, once chondrogenic differentiation was initiated, enhanced expression of these genes was observed, and differentiation appeared to proceed at a normal rate based on the maintenance of an unchanged ratio of type II/X collagen mRNA expression (Fig. 2B). The uniform delay in chondrogenesis is also evident by comparing the expression of aggrecan with collagen type II expression (Fig. 2A, bottom panel). A further assessment of the rate of chondrogenic differentiation was made by examining expression of the two splicing variants of type II collagen that are representative of more embryonic or immature stages of chondrogenic development (COL2A1a) and the isoform that is representative of the more mature chondrocytes (COL2A1b).(25) The relative ratio of the mature cartilage form (COL2A1b) to the that of the immature form (COL2A1a) was then determined, and while both isotypes were equally delayed in their expression, no apparent differences in the ratio of expression of the immature to mature isotypes were seen at any of the time points after fracture (Fig. 2A, bottom panel).

Figure FIG. 2..

Expression of cartilage extracellular matrix proteins during fracture healing in wild-type and TNFR-deficient mice. (A) Upper 2 RPAs: Representative RPA analysis of collagen type II (COL2A1), collagen type X (COL10A1), and the ribosomal protein L32 mRNA expression. Lower RPA: Representative RPA analysis of collagen type II, 5′ splice variants IIA and IIB, aggrecan large proteoglycan (AG), and L32 mRNA expression. The times after fracture are noted. (B) Graphical analysis of the relative levels of expression for collagen types II and X mRNA over the time course of fracture healing.

While the histological and mRNA profile assessments of the transgenic animals provided an excellent descriptive analysis of consequences of the deficiency in TNF-α signaling during fracture healing, these data do not define the underlying molecular mechanism that produced the delayed fracture healing. The functional mechanisms by which TNF-α affected chondrocytes within the callus tissue was investigated in cultured cells isolated from callus tissues taken 10 days after fracture, thereby allowing the direct assessment of cellular response to exogenously added TNF-α. Initial experiments were carried out to establish that the cells that were isolated from the callus tissues would express a cartilage phenotype in vitro. A histological examination of the isolated tissue showed that the cartilage component of the callus tissues could be cleanly isolated from 10-day postfracture specimens. Enzymatic digestion of the callus tissue released cells from the callus tissue that had a typical polygonal appearance of cultured chondrocytes (Fig. 3, top panel).(17) Subsequent molecular analysis of the phenotype of these cells, as determined by mRNA profiling, demonstrated that no mature osteogenic cells were present in the cultures based on the absence of the expression of osteocalcin. However, after 4 days in culture, the expression of type X collagen was lost, and type II collagen and aggrecan expression progressively diminished with increasing expression of the COL2A1a isoform, suggesting that the chondrocytes were reverting to a more embryonic phenotype (Fig. 3, top panel, left and middle). Therefore, all subsequent experiments were performed on cells cultured for only 4–6 days. After carrying out these initial control experiments, callus cultures were prepared from wild-type and TNF-α receptor-deficient mice. We then assessed if TNF-α treatment would directly affect the differentiated state of these cells by examining the expression of their extracellular matrix gene expression. These results demonstrated that TNF-α treatment did not have a direct effect on the expression of the differentiated extracellular matrix phenotype of these cells (Fig. 3, bottom panel, right).

Figure FIG. 3..

Establishment of callus chondrocyte cell cultures and the effect of TNF-α on extracellular matrix (ECM) protein mRNA expression. (A) Upper panel: Morphological characterization of calluses and isolated chondrocytes. Left panel shows a sagittal section through the healing fracture site (hematoxylin and eosin, magnification, ×25). Middle panel shows tissue morphology of the callus dissected from underlying bone used for enzymatic dissociation (hematoxylin and eosin, magnification, ×40). Right panel shows cell morphology of chondrocytes after 4 days of growth in culture (magnification, ×40; inset, ×100). Lower panel: (left and middle) RPA analysis of cartilage ECM mRNA expression over 21 days of growth in culture. Additional ECM mRNAs, osteocalcin (OC), bone sialoprotein (BSP), collagen type I α1 (COL1A1), and osteopontin (OPN) are indicated. (right) Effect of TNF-α treatment on ECM mRNA expression in 8-day cultures from wild-type and receptor null mice. Concentrations of TNF-α additions are noted.

TNF-α signaling mediates chondrocyte apoptosis

In our previous studies, TNF-α had been shown to induce apoptosis in cultured avian embryonic chondrocytes.(17) A series of studies was next carried out to examine the role of TNF-α in the mediation of chondrocyte apoptosis. The expression of a cassette of mRNAs that regulate the apoptotic cascade were examined within the callus tissues over the period of chondrogenic differentiation, hypertrophy, and resorption in vivo (Figs. 4A and 4B). The expression of several of these mRNAs including Fas, Caspase 8, and TRAIL showed delayed decreased or absent expression across the time course of fracture repair (Fig. 4A). We then examined if TNF-α would increase the expression of these same genes in the cultured callus chondrocytes and compared while these cells to those isolated from the mutant animals (Fig. 4B). Not only are the reduced levels of these genes in vivo identical to those inducible, in vitro, in response to TNF-α, they also show a lack response within the mutant cells. Finally because the callus tissues from which we prepared the cell cultures are a mix of different tissues, we also tested the effect of TNF-α treatment in pure populations of femoral head articular chondrocytes and compared its effects to that seen in cultured callus cells (Fig. 4C). As can be seen in this experiment, almost all of the pro-apoptotic mRNAs that are assayed in this cassette of genes are induced within the pure populations of chondrocytes.

Figure FIG. 4..

Role of TNF-α signaling in the induction of chondrocyte apoptosis. (A) Upper panel: RPA analysis of the expression of various mRNAs associated with the induction of apoptosis (caspase 8 [CP8], fas, tumor necrosis factor-related apoptosis-inducing ligand [TRAIL]), days after fracture are noted. Lower panel: Graphical analysis of selected mRNA expression over the time course of fracture healing in vivo. (B) Effect of TNF-α treatment on apoptosis-associated mRNA expression in callus chondrocytes after 8 days of culture. Right panel: RPA analysis of the expression of various mRNAs associated with the induction of apoptosis (CP8, fas, TRAIL), days after fracture are noted. Left panel: Graphical analysis of mRNA expression in callus cultures as a function of TNF-α addition. Concentrations of TNF-α addition are noted. (C) Comparison of the effect of TNF-α treatment on apoptosis-associated mRNA expression in callus chondrocytes to femoral head articular chondrocytes after 8 days of culture. Left panel: RPA profiles from callus cultures. Right panel: RPA profiles from femoral articular head chondrocytes. Concentrations of TNF-α addition are noted.

While these genes provide an assessment of whether apoptotic cascade has been activated, a direct determination of apoptosis of the chondrocytes within callus tissues in vivo was made by using the TUNEL assay (Figs. 5A and 5B). These results directly display that there was a higher percentage of apoptotic cells within the wild-type callus tissue at 7 and 10 days compared with the receptor null mice. At later time points, the TNF-α receptor null group exhibited enhanced apoptosis and the levels of apoptosis appeared to catch up with that of the wild-type animals.

Figure FIG. 5..

Assessment of apoptosis in vivo. TUNEL assay of callus tissues at 7, 10, and 14 days after fracture in wild-type and TNFR (p55−/−p75−/−) mice. (A) Comparison of micrographs of representative histological sections reacted for TUNEL and counterstained with H&E. Nuclei showing DNA nicking are labeled dark brown to black. (B) Graphical analysis of the percentage of cells with darkly stained nuclei. Error bars show SE. (magnification, ×100).

TNF-α signaling regulates endochondral tissue resorption

Because of TNF-α's known role in stimulating bone resorption, our final experiments assessed if the deficiency of TNF-α signaling would also alter the activity of the osteoclasts/chondroclasts involved in the resorption of the mineralized cartilage. Histological analysis of osteoclast/chondroclast cells within the fracture callus was carried out throughout the repair period. At early times at day 7, there was an initial increase in the number of osteoclasts seen along the periosteal surfaces where intramembranous bone in which growth was occurring. There was a subsequent penetration of the osteoclasts within the callus tissue by the osteoclasts, with almost no osteoclasts seen in the calluses of the mutant animals before 14 days, followed by a very rapid increase in osteoclast numbers seen by 21 days (Fig. 6B). These data were quantitatively further confirmed by examining the mRNA expression of TRACP (Fig. 6B). This analysis showed a clear shift of several days in the period of time when maximal TRACP mRNA expression was seen. We then examined whether TNF-α changed the expression profile of the proresorptive regulators that promoted osteoclast activity similar within the callus tissues in vivo. An examination of the mRNA expression of the primary known regulators of osteoclast activity (M-CSF, RANKL, and OPG) in the intact callus tissue showed that both M-CSF and RANKL were initially reduced and ultimately delayed, while OPG levels were much higher throughout the time course of fracture healing in the receptor-deficient animals (Figs. 7A and 7B). In vitro studies showed that TNF-α directly stimulated the expression M-CSF in the cultured callus chondrocytes while concurrently decreasing the expression of OPG. In contrast, in the transgenic cells, M-CSF showed a lower basal level of expression and failed to be induced with TNF-α treatment while OPG showed a higher basal level of expression in the untreated cells and remained elevated in the presence of TNF-α treatment. It is interesting to note that while we observed altered RANKL expression in vivo, we failed to detect quantifiable levels of RANKL in the callus cultures, even when we treated the cells with TNF-α (Fig. 7B), suggesting that a different type of cell synthesizes RANKL.

Figure FIG. 6..

Role of TNF-α signaling in the resorption of calcified cartilage. (A) Histological analysis of TRACP+ multinucleated cells in fracture calluses of wild-type and TNFR-deficient mice. Representative histological sections of calluses at 7, 14, and 21 days after fracture are presented. TRACP+ cells are stained bright red (magnification, ×100). (B) Quantification of TRACP+ cells per unit area of callus tissue. Error bases denote SD of the measurements. (C) Upper panel: RPA of TRACP mRNA expression. Days after fracture are noted. Lower panel: Graphical analysis of TRACP mRNA expression.

Figure FIG. 7..

Role of TNF-α signaling in the induction of proresorptive factors by chondrocytes. (A) Upper panel: RPA of the expression of mRNAs for the primary regulators of osteoclast differentiation and activity. Days after fracture are noted. Lower panel: Graphical analysis of the expression of these mRNAs over the time course of fracture healing in vivo. (B) Effect of TNF-α treatment on the expression of RANKL, M-CSF, and OPG in callus chondrocytes after 8 days in culture. Left panel: RPA examining the expression of RANKL, M-CSF, and OPG in culture. Concentrations of TNF-α addition are noted. Right panel: Graphical analysis of mRNA expression in callus cultures as a function of TNF-α addition.


There is a growing understanding that inflammatory and immune cell cytokines carry out central functions in the initiation and induction of downstream responses to many different kinds of injuries in the absence sepsis.(26,27) The goals of this study were to determine the functional role of TNF-α in closed fracture healing induced by blunt trauma. Our observations from this study show that in the absence TNF-α signaling the temporal progression of fracture healing was considerably delayed with effects observed at three discrete levels: (1) there is a delay during the initial healing phase in the either the recruitment of mesenchymal cells or the initiation of these cells to undergo skeletogenic cell differentiation; (2) there is a delay in chondrocyte apoptosis during the endonchondral period; and (3) there is also a delay in the resorption of the mineralized cartilage during the endonchondral period. These effects indicate that TNF-α plays a role in early steps of mesenchymal cell differentiation, carries out a primary role in the facilitation of chondrocyte apoptosis, and plays a role in the regulating the production of factors that control the resorption of mineralized cartilage. Concerning the effects of TNF-α on either mesenchymal cell recruitment or differentiation, one of the primary molecular responses to TNF-α signaling is the activation of NF-κB, which has been shown to be essential in early limb bud formation.(28)

While it is interesting to speculate on the underlying molecular mechanism by which the lack of TNF-α alters initial skeletal tissue formation, the current studies focus specifically on the molecular mechanisms by which TNF-α signaling contributes to the resorption of calcified cartilage during fracture healing. Crucial to the progression of endochondral skeletal development are the physical removal of the mineralized cartilage anlage and the remodeling of the mineralized cartilage tissue and formation of primary trabecular bone. Chondrocyte apoptosis has been shown to be an important component in the sequential progression of endochondral bone formation.(29,30) Past studies have shown that TNF-α and Fas ligand, another TNF-α superfamily member, are two of the primary molecules that control the programmed cell death of chondrocytes during developmental and fracture-induced endochondral bone formation.(31) In previous studies, we demonstrated that TNF-α induces apoptosis in vitro in populations of both hypertrophic and permanent cartilage chondrocytes isolated from avian tissues.(17) When combined with the data from our current experiments, these data suggest that the apoptotic actions of TNF-α may be dually regulated through the induction of Fas. TNF-α also seems to upregulate TRAIL, which is one of the other known pro-apoptotic mediators. This indicates that TNF-α is upstream of both TRAIL and Fas-L mediation of apoptosis. From the standpoint of coordinating apoptotic removal of the cells and tissue resorption, apoptotic induction could then be further enhanced through the presentation of Fas ligand, which would be produced in high levels on the arrival of the hematopoietic and resorptive cells into the endochondral tissue or through the TNF-α-induced expression of TRAIL, which acts in an autocrine manner to self-induce apoptosis. Such redundancy could then explain the eventual induction of apoptosis in the mutant mice, because low levels of Fas are present in the absence of TNF-α in the callus cultures, yet our data also suggests that TNF-α synergizes the actions of Fas-L through the induction of the Fas receptor.

In previous studies, we had shown that the period of active chondrogenesis during fracture healing was associated with a high ratio of OPG relative to M-CSF and RANKL, yet during endochondral resorption the levels of OPG fell while the expression of M-CSF and RANKL increased.(3) Such results suggest that the remodeling of mineralized cartilage is controlled like bone resorption through the balance (ratio) of RANKL to its antagonist OPG in conjunction with M-CSF. While our previous studies did not address the cellular origin of these factors, the current studies show that endochondral chondrocytes most likely regulate the resorption of the mineralized cartilage tissue through the direct regulation of at least two of these cytokines, M-CSF and OPG. Our data would further suggest that the primary protection of the chondrogenic tissues from resorption during the early periods of cartilage tissue formation after fracture is derived from the high levels of OPG produced by these cells. In this context, it is interesting to note the 20–30% higher basal level of this gene's expression in the chondrogenic cells in the absence of TNF-α signaling and the downregulation of its expression by exogenously added TNF-α. These findings may have considerable implications to the destructive role that TNF-α plays in pathologies that effect articular cartilage,(32,33) because high levels of OPG have also been observed in this type of cartilage tissue.(34) We did not detect the induction of basal expression of RANKL by the cultured callus chondrocytes, suggesting that this cytokine is probably presented to the osteoclasts by hematopoietic cell types when vessels grow into the areas of the mineralized cartilage. While these data suggest that the processes that mediate endochondral resorption and bone remodeling phases are dependent on the same molecules (M-CSF, OPG, and RANKL), the data suggest at the same time that these processes are regulated locally in the calcified cartilage and bone tissues in a different manner. The conclusion that the mechanisms that regulate calcified cartilage resorption are different from that of bone are further supported by the studies of RANKL (TRANCE)-deficient mice or mice in which RANKL expression was rescued by the engineering of RANKL expression in lymphocytes. In these studies, RANKL expression by lymphocytes was able to promote osteoclast development and rescue the osteopetrosis in both the marrow space and woven bone replacement in the cortical shafts. It did not, however, correct the chondrodysplasia of the epiphyseal and metaphyseal area. These observations led the authors to conclude that RANKL was under differing mechanisms of local tissue control in cartilage as opposed to bone.(35)

One of the most intriguing aspects of this study was that the multiple molecular mechanisms affected by the removal of TNF-α signaling were coordinated solely through its actions on chondrocytes. Previous studies have shown that chondrocytes appear to centrally control other processes in skeletal tissue formation including angiogenic processes that accompany tissue resorption(36) and the production of the morphogenetic signals that selectively drive osteogenesis.(37) The emergent picture from both past and current studies would suggest that the mechanisms controlling cartilage differentiation, growth, maturation, and apoptosis are the key rate controlling steps in osteogenesis, and alterations in any number of functions that chondrocytes carry out affect the subsequent progression of bone formation.(37–39) In this context, our technical approach of examining cassettes of genes that define particular molecular processes such as apoptosis or the mechanisms that regulate resorption show that when one aspect of a molecular process is altered, there is a cascading and coordinated effect on the other molecular mechanisms that regulate that process. Furthermore, these studies show that multiple genes of a specific molecular process will be coordinately regulated together in the same manner.

While the current accepted paradigm is that endochondral progression that occurs during postnatal fracture repair recapitulates the processes that occur during embryological skeletal development,(5) the current data suggests that not all aspects of the regulatory mechanisms that control these separate processes are the same. If the period of delay observed in fracture repair were translated to embryological skeletal development, it would produce an observable phenotype. However, TNF receptor-deficient animals show no overt anomalies in the embryological or postnatal aspects of the development of their skeletons. It is therefore important to note that the postnatal tissue environment is very different from that of embryological development. Unlike embryological bone development in which the marrow space is concurrently being formed and the immune system is naïve, fracture repair in mature animals occurs in close physical association with an immune competent marrow element in a skeletal tissue that is weight bearing. Finally, an inflammatory process that has been produced as a consequence of trauma initiates fracture healing. Thus, we conclude that fracture repair has numerous complexities that are unique from embryological skeletal tissue formation and will not be regulated by exactly the same regulatory mechanisms.


This work has been supported with a grant from NIAMS AR 47045 (LCG). Institutional support was provided by the Department of Orthopedic Surgery, Boston University School of Medicine (T-JC).