We have previously identified in articular cartilage an abundant pool of the heparin-binding growth factor, fibroblast growth factor 2 (FGF-2), which is bound to the pericellular matrix heparan sulfate proteoglycan, perlecan. This pool of FGF-2 activates chondrocytes upon tissue loading and is released following mechanical injury. In vitro, FGF-2 suppresses interleukin-1–driven aggrecanase activity in human cartilage explants, suggesting a chondroprotective role in vivo. We undertook this study to investigate the in vivo role of FGF-2 in murine cartilage.
Basal characteristics of the articular cartilage of Fgf2−/− and Fgf2+/+ mice were determined by histomorphometry, nanoindentation, and quantitative reverse transcriptase–polymerase chain reaction. The articular cartilage was graded histologically in aged mice as well as in mice in which osteoarthritis (OA) had been induced by surgical destabilization of the medial meniscus. RNA was extracted from the joints of Fgf2−/− and Fgf2+/+ mice following surgery and quantitatively assessed for key regulatory molecules. The effect of subcutaneous administration of recombinant FGF-2 on OA progression was assessed in Fgf2−/− mice.
Fgf2−/− mice were morphologically indistinguishable from wild-type (WT) animals up to age 12 weeks; the cartilage thickness and proteoglycan staining were equivalent, as was the mechanical integrity of the matrix. However, Fgf2−/− mice exhibited accelerated spontaneous and surgically induced OA. Surgically induced OA in Fgf2−/− mice was suppressed to levels in WT mice by subcutaneous administration of recombinant FGF-2. Increased disease in Fgf2−/− mice was associated with increased expression of messenger RNA of Adamts5, the key murine aggrecanase.
These data identify FGF-2 as a novel endogenous chondroprotective agent in articular cartilage.
The structural integrity of articular cartilage is determined principally by homeostasis of the 2 major macromolecules of the extracellular matrix, type II collagen and the chondroitin sulfate–rich proteoglycan aggrecan. In healthy tissue, there is a balance between anabolic (synthetic) and catabolic (degradative) processes that allows matrix turnover. Excessive catabolic activity results in matrix breakdown, a hallmark of osteoarthritis (OA). One key early event in matrix breakdown is loss of aggrecan, which is caused by aggrecanase enzymes, members of the ADAMTS family. In humans, ADAMTS-4 and ADAMTS-5 are thought to be the major aggrecanases in cartilage (1, 2). In the mouse, deletion of ADAMTS-5, but not ADAMTS-4, was shown to protect against the development of OA and inflammatory arthritis, suggesting that ADAMTS-5 is the main murine aggrecanase (3, 4).
We have identified fibroblast growth factor 2 (FGF-2) as a potential regulatory molecule in articular cartilage. It is bound to the heparan sulfate chains of the proteoglycan perlecan in the pericellular matrix of human and porcine cartilage, where it acts as a mechanotransducer (5, 6). Upon loading, FGF-2 is made available to cell surface tyrosine kinase receptors and activates intracellular signaling pathways including ERK, one of the MAPKs (7). FGF-2 is also released from the pericellular pool upon physical injury to the tissue (8).
In order to determine the role of FGF-2 in articular cartilage, we recently investigated the influence of FGF-2 on the breakdown of aggrecan in human articular cartilage explants. We found that FGF-2 suppressed interleukin-1 (IL-1)– or tumor necrosis factor–stimulated aggrecanase activity in explants of normal knee cartilage in a dose-dependent manner (9). Because of these findings, we investigated whether FGF-2 was chondroprotective in vivo. Fgf2−/− mice are viable, fertile, and morphologically indistinguishable from their wild-type (WT) littermates under normal conditions (10). We examined the knee articular cartilage of naive Fgf2−/− and Fgf2+/+ mice, and then we compared both age-related cartilage degeneration in Fgf2−/− and Fgf2+/+ mice and cartilage degeneration following surgical destabilization of the medial meniscus (DMM), a well-established model of OA.
MATERIALS AND METHODS
The generation of Fgf2-knockout mice has been described elsewhere (10). A fragment of the Fgf2 gene was replaced with a 3.2-kb hypoxanthine guanine phosphoribosyltransferase (Hprt) minigene, effectively removing the first 59 amino acids, which are important for heparin and receptor binding and for mitogenic activity. Heterozygous (Fgf2+/−) breeding pairs were maintained on a 50% 129Sv/50% black Swiss background and were provided by The Jackson Laboratory (Bar Harbor, ME). Genotyping was performed using reverse transcriptase–polymerase chain reaction (RT-PCR) with the following primer pairs: 5′-CGA-GAA-GAG-CGA-CCC-ACA-C-3′ (forward) and 5′-CCA-GTT-CGG-GGA-CCC-TAT-T-3′ (reverse) for the WT allele, and 5′-AGG-AGG-CAA-GTG-GAA-AAC-GAA-3′ (forward) and 5′-CCC-AGA-AAG-CGA-AGG-AAC-AAA-3′ (reverse) for the target (deleted) allele. Full-cousin mice were used in all experiments, and, where possible, experiments were performed with littermate controls.
Surgical induction of OA.
Surgical induction of OA was performed in 10–12-week-old mice by microsurgical release of the anterior horn of the medial meniscus from its tibial attachment (3). Sham surgery consisted of medial capsulotomy only. All invasive procedures were approved by the UK Home Office and followed institutional ethical and procedural guidelines.
For FGF-2 and perlecan immunostaining, hips from 6-week-old mice were snap-frozen and sectioned (6-μm sections) using a CV1900 cryostat (Leica Microsystems, Milton Keynes, UK) equipped with a tungsten blade. Tissue was fixed for 15 minutes with methanol, then immunostained and viewed using Ultraview confocal microscopy (60× oil immersion lens; PerkinElmer, Waltham, MA) as described previously (6). For tissues used for scoring of OA, dissected knee specimens were fixed in 10% formalin, decalcified in dilute formic acid, and embedded in paraffin. Coronal sections (8-μm thick) were cut and stained with Safranin O. Sections were evaluated at 80-μm intervals.
Severity of cartilage destruction was assessed histologically using a modification of a previously described 7-point scale (0 = normal; 1 = surface fibrillations; 2 = loss of superficial cartilage and surface delamination, with shallow fissures but no frank ulceration; 3 = ulceration of noncalcified cartilage only; 4 = vertical clefts extending into subchondral bone; 5 = ulceration extending into calcified cartilage but not into subchondral bone, <80% cartilage loss; and 6 = ulceration extending into subchondral bone and/or >80% cartilage loss) (11). Histologic scoring was performed by 2 blinded observers. All articular surfaces within each section (medial femur, medial tibia, lateral femur, lateral tibia) were graded separately, and the scores were added to give a score for the section (range 0–24). The histologic summed score (maximum 72) was calculated from the sum of the 3 highest section scores to provide a composite indicator of both the severity and the extent of cartilage damage. Intra- and interobserver agreement were very good, as reflected by intraclass correlation coefficients (ICCs) of >0.8.
Cartilage histomorphometry was performed on uncompressed digital images using MCID Analysis 7.0 software (Imaging Research, St. Catherines, Ontario, Canada). For neoepitope immunostaining, embedded sections were deparaffinized and treated for 30 minutes with 0.1% hyaluronidase in phosphate buffered saline (PBS). Sections were blocked with normal goat serum and then incubated for 1 hour with anti-NITEGE antibodies (1:850 dilution; kindly provided by Dr. John Mort, Shriner's Hospital, Montreal, Quebec, Canada). Antigen was visualized using streptavidin-conjugated peroxidase (ABC kit; Vector, Burlingame, CA) according to the manufacturer's recommendations.
Safranin O–stained paraffin sections were digitally photographed in uncompressed TIFF format with specialized software (Spot v4.5; Diagnostic Instruments, Sterling Heights, MI) using a Spot Insight Color Mosaic camera (Diagnostic Instruments) attached to a Leitz Wetzlar Dialux 22EB light microscope (Leitz, Wetzlar, Germany). Histomorphometric analysis was performed with MCID Analysis 7.0 software. The derivation of the morphometric indices measured is as described below.
The general region of interest (e.g., the area around the tibial plateau or femoral condyle) was outlined. A preset function based on chromatic thresholding was then used to precisely select only the cartilaginous component, following which an algorithm was applied that measured multiple vertical distances across the selected region of cartilage, before returning the maximum and mean values. These values were taken as the maximum and mean “thickness” measurements, respectively.
The TIFF image was first converted to a 16-bit grey-scale image in the red channel. The cartilaginous component was selected as described above. The averaged optical density (OD) was then read—this corresponded to the OD calculated on a grey scale obtained in the red component of the light crossing the section. The OD was a function of the intensity of Safranin O staining. Although this technique is not fully quantitative, it has been validated previously and provides an approximate measure of proteoglycan content (12). Each OD reading was normalized to the OD of the epiphyseal plate in the same tissue section (which acted as a staining control), and the resulting value was expressed as a ratio. This normalized OD reading was used as an indicator of the degree of proteoglycan staining. The repeatability of measurements made with the MCID software was excellent, with a vanishingly low variance. The intraobserver and interobserver ICCs were also >0.9, indicating excellent rater agreement.
RNA was extracted from whole joints following removal of the skin and muscle bulk. Tissue was snap-frozen and then extracted using TRIzol reagent (Invitrogen, Paisley, UK) and purified using a spin column kit (Qiagen, Hilden, Germany). RNA was reverse-transcribed to complementary DNA and quantified with the TaqMan real-time PCR system (Applied Biosystems, Foster City, CA), using prevalidated primers/probe mixes obtained from the same company, at a concentration of 900 nM for each primer and 200 nM for the probe. Real-time PCR was performed using a RotorGene 6000 thermocycler (Corbett Research, Mortlake, New South Wales, Australia). Data capture and primary analysis were carried out with RotorGene 6000 software (version 1.7) from the same company. The thermocycling conditions were as follows: denaturation at 95°C for 10 minutes, followed by 45 cycles of a denaturation step at 95°C for 2 seconds, followed by an annealing/extension step at 60°C for 30 seconds. All samples were measured in triplicate and compared with expression levels of the housekeeping gene GAPDH.
RefSeq accession codes were as follows: for Fgf2, NM_008006.1; for Adamts4, NM_172845.1; for Adamts5, NM_011782.1; for Mmp13, NM_008607.1; for Timp1, NM_011593.2; and for Timp3, NM_011595.2.
Pharmacologic rescue of Fgf2−/− mice.
Recombinant FGF-2 (R&D Systems, Minneapolis, MN) was administered subcutaneously at a dosage of 1 μg every other day, beginning 10 days before surgical OA induction and continuing until the time when mice were killed (2 or 4 weeks after surgery). Control animals were injected with an equal volume of PBS.
Nonparametric comparisons were made using the Mann-Whitney 2-tailed U test unless specified otherwise. Multiple comparisons were made using one-way analysis of variance with post hoc analysis as indicated. P values less than 0.05 were considered significant.
Colocalization of FGF-2 with perlecan in the pericellular matrix of murine articular cartilage.
We previously demonstrated colocalization of FGF-2 and perlecan in the pericellular matrix of porcine and human articular cartilage, and verified this binding through in vitro studies using surface plasmon resonance (6). Pericellular staining of FGF-2 was confirmed in articular cartilage of Fgf2+/+ mice, where it colocalized with perlecan as determined using confocal microscopy. Staining for FGF-2 was absent in Fgf2−/− animals, although some nonspecific staining was apparent on the superficial articular surface. Pericellular distribution of perlecan was maintained in Fgf2−/− mice (Figure 1).
Increased OA in aged Fgf2−/− mice.
We first examined the appearance and characteristics of the articular cartilage of 3-month-old mice (the age at which OA was to be induced). No differences in cartilage thickness or intensity of Safranin O staining were observed between Fgf2−/− and Fgf2+/+ mice by histomorphometric analysis (Figures 2a–d). We also tested the integrity of the matrix by nanoindentation, using atomic force microscopy. No difference in mechanical properties of the matrix was detected between either genotype (data not shown). There was also no difference in gene expression of the principal matrix proteins aggrecan and type II collagen between Fgf2−/− and Fgf2+/+ mice as judged by messenger RNA (mRNA) level (Figure 2e).
Spontaneous degeneration of the articular cartilage was assessed histologically at 3, 6, and 12 months. The articular cartilage in all 4 quadrants of the joint, from at least 8 sections across the joint, was scored (see Materials and Methods). To measure the severity as well as the extent of chondral damage, the scores from the 4 quadrants were added together, and the highest 3 scores from an individual joint were summed (summed score). While there was no apparent spontaneous degradation of the cartilage at age 3 months, at age 6 months both Fgf2−/− and Fgf2+/+ mice showed incipient articular cartilage degradation, which was significantly greater in the Fgf2−/− mice (Figures 3a and b). By 9 months further significant increase in disease was apparent in Fgf2−/− mice (Figure 3a). When the summed scores of the medial and lateral compartments were considered separately, the Fgf2−/− mouse cartilage showed severe degeneration in both compartments, while the Fgf2+/+ mouse cartilage showed severe degeneration only in the medial compartment (Figures 3b and c). Histologic analysis also showed that at 9 months, the medial compartment erosion in Fgf2−/− mice, as compared with that in WT mice, extended more deeply into the subchondral bone, at times beyond the growth plate (Figure 3b).
Increased OA in Fgf2−/− mice following surgical destabilization of the knee.
We next examined the susceptibility of 12-week-old Fgf2+/+ and Fgf2−/− mice to surgically induced OA, using DMM as described previously (3). DMM results in a progressive OA-like disease, in which the articular cartilage degenerates with little or no synovitis. We compared the summed scores from Fgf2+/+ and Fgf2−/− mice at 2, 4, and 8 weeks following surgery. Compared with Fgf2+/+ littermate controls, acceleration of disease was seen in Fgf2−/− mice at all time points examined following DMM surgery (Figure 4a). Representative histologic sections are shown from sham-operated Fgf2+/+ mice and from DMM-operated Fgf2+/+ and Fgf2−/− mice (Figure 4b). Sham-operated knees (in which the joint capsule was opened but the meniscotibial ligament was spared) from mice of either genotype showed no cartilage damage (data not shown). Slowly progressive cartilage damage was seen, however, in the contralateral, unoperated knees of Fgf2−/− mice 4 and 8 weeks after DMM surgery. This damage was significantly greater than that in Fgf2+/+ mice at 8 weeks (Figure 4c). This was likely due to increased joint loading through the unoperated side, because of progressive arthritis in the operated knee. These findings indicate that the absence of FGF-2 in articular cartilage hastens matrix breakdown and the development of OA, under conditions of both increased physiologic (i.e., with age, and on the unoperated contralateral side) and pathologic (i.e., the operated side) joint loading.
Subcutaneous delivery of FGF-2 reverses the accelerated OA phenotype of Fgf2−/− mice following DMM surgery.
Even though we had determined by nanoindentation and histomorphometry that the cartilage of Fgf2−/− mice was similar to that of WT mice, it was possible that susceptibility to OA in Fgf2−/− mice was due to an intrinsic weakness in the tissue that had arisen during development. In order to address this, we treated Fgf2−/− mice with subcutaneous recombinant human FGF-2. Others have achieved therapeutic dosing of FGF-2 by subcutaneous injection (13) following detailed kinetic studies in the mouse (14). Although no data are available on the extent to which FGF-2 is able to penetrate cartilage via subcutaneous administration, the presence of detectable FGF-2 in the plasma suggests that access to the joint is feasible. Mice were pretreated with either subcutaneous FGF-2 or PBS from 10 days prior to DMM surgery to 2 or 4 weeks after surgery. Fgf2−/− mice treated with FGF-2, but not those treated with PBS, were protected from accelerated OA 2 weeks following surgery and showed histologic scores comparable with those of Fgf2+/+ mice treated with PBS (Figure 4d). A trend toward protection was also seen in mice treated for 4 weeks following surgery, but this did not reach significance (data not shown). The ability to reverse the susceptibility of the strain by FGF-2 treatment suggested that FGF-2 had an active chondroprotective role in mature tissue.
Increased ADAMTS-5 expression and activity in Fgf2−/− mouse cartilage in vivo.
ADAMTS-5 is the major aggrecan-degrading enzyme in murine cartilage; mice deficient in ADAMTS-5 are protected from surgically induced OA as well as from inflammatory arthritis (3, 4). We extracted mRNA from joints of mice 2 weeks following DMM or sham surgery. Adamts5 mRNA was increased in joints of both Fgf2+/+ and Fgf2−/− mice 2 weeks after DMM surgery compared with that in sham-operated and contralateral, unoperated control joints (Figure 5a). This increase was significantly greater in Fgf2−/− mice than in Fgf2+/+ mice. Two other proteinase genes, Adamts4 and Mmp13 (a collagenase), were also induced in knees following DMM surgery, but there was no further increase in the Fgf2−/− mouse joints, indicating that the expression of these proteinases during the chondral response to joint destabilization was not influenced by endogenous FGF-2 (Figure 5a). We also looked at regulation of the intrinsic tissue inhibitors of metalloproteinases (TIMPs) 1 and 3. Timp1, but not Timp3, was induced following DMM surgery, but no difference in expression was detected between Fgf2+/+ and Fgf2−/− mouse knees (Figure 5a). In WT mice, up-regulation of Fgf2 mRNA was detected in joints following DMM surgery (Figure 5a).
To confirm that ADAMTS-5–mediated aggrecanolysis was occurring in the articular cartilage following DMM surgery, immunohistochemical evaluation was performed using an antibody against the aggrecanase-generated neoepitope (NITEGE). This antibody cross-reacts with the cleaved epitope of murine aggrecan (NVTEGE) that is retained in cartilage following ADAMTS-mediated cleavage (15). Staining of the NVTEGE epitope was apparent in the superficial layer of articular cartilage in both Fgf2+/+ and Fgf2−/− mouse cartilage 2 weeks following DMM surgery, but was significantly reduced in tissue from sham-operated animals. No staining was seen when a nonimmune antibody was used (Figure 5b). Table 1 shows the threshold cycle values for the ADAMTS enzymes, matrix metalloproteinase 13 (MMP-13) and TIMP-3. Interestingly, despite a clear role for ADAMTS-5 in the course of murine OA, the abundance of ADAMTS-5 (and ADAMTS-4) mRNA in the joint was low compared with that of MMP-13 and TIMP-3 mRNA.
Threshold cycle (Ct) values are listed for Adamts4, Adamts5, Mmp13, Timp3, and GAPDH from joints subjected to destabilization of the medial meniscus (DMM) or sham operation (Sham) as well as from unoperated (NO) joints. Each value is an average of 3–5 samples from experiments performed in triplicate. The Ct value denotes the cycle number at which it is possible to measure the amplified polymerase chain reaction product. The Ct value is inversely proportional to the copy number of the gene of interest template; a low Ct value represents a higher initial copy number of the gene of interest in the starting material.
This is the first description of an endogenous matrix-bound growth factor acting as a chondroprotective agent in vivo; mice deficient in FGF-2 developed accelerated OA with age or following surgical destabilization of the knee. Our earlier observation that FGF-2 inhibits IL-1–driven aggrecanolysis in human explants (9) suggests that the chondroprotective effect of FGF-2 might be due, at least in part, to suppression of aggrecanolysis mediated by ADAMTS-5. Indeed, Adamts5 mRNA was superinduced in Fgf2−/− mice compared with Fgf2+/+ mice 2 weeks following DMM surgery. We were unable to visualize ADAMTS-5 protein due to lack of suitable antibodies and the likely low abundance of the enzyme, although we were able to demonstrate ADAMTS-mediated aggrecanolysis in the cartilage of both Fgf2−/− and Fgf2+/+ mice.
Regulation of ADAMTS-5 gene expression has not been extensively studied. It was initially described as a constitutively expressed aggrecanase in bovine (16, 17) and human (18) chondrocytes, although we and others have demonstrated regulation by inflammatory cytokines in murine (4), bovine (19), and human (9) tissue. The putative promoter region of the gene has predicted binding sites for a number of transcription factors, including NF-κB, which are likely to be involved in gene expression following activation of inflammatory signaling (20). It is not clear how FGF-2 influences ADAMTS-5 gene expression in cytokine-treated cartilage explants. The mechanism by which this occurs in vivo is likely to be yet more complicated because of the expression and effects of FGF-2 in tissues other than articular cartilage.
It is possible that FGF-2 was in part inhibiting aggrecanase activity through induction of TIMP-3, one of the 4 TIMP family members, which is a high-affinity inhibitor of ADAMTS-4 and ADAMTS-5 (21). However, we found no evidence that FGF-2 regulated the expression of TIMP-3 mRNA. Indeed, Timp3 mRNA was not differentially expressed in Fgf2−/− and Fgf2+/+ mice following DMM surgery or in vitro following stimulation of human cartilage with exogenous FGF-2 or IL-1 (9).
It is not known what signals and which cells are responsible for aggrecanase expression following surgical destabilization of the knee. Although there is little or no synovitis seen in this model when cartilage degeneration is occurring (after 4 weeks), it is possible that cytokine production within the synovium or joint cavity drives proteinase production. It is also possible that proteinase expression is directly induced in the chondrocyte by altered joint biomechanics. Destabilization of the meniscus is likely to have 2 main consequences following the initial cutting injury. The first is increased load transmitted through the weight-bearing region of the joint. The second is joint instability resulting in increased shear stress over the articular surfaces. It is not known whether chondrocytes are able to sense and respond to such changes, but our previous finding that simple cutting of cartilage is sufficient to activate JNK and p38 MAPK pathways and to induce inflammatory response genes such as IL-1 in the chondrocyte (22) lends support to such a theory.
The role of FGF-2 in cartilage has been most studied in the growth plate, where it has an inhibitory effect on chondrocyte proliferation. Gain-of-function mutations in human FGF receptor 3 (FGFR-3) (23–27) and overexpression of FGF-2 (28) or intravenous treatment of mice with recombinant FGF-2 (29) all result in a reduction in the proliferating zone of the growth plate and in subsequent shortening of the long bones (achondroplastic dwarfism). The absence of a developmental phenotype in Fgf2−/− mice is thought to be due to compensation by other FGF family members such as FGF-18 which are present in the growth plate (30). It is possible that Fgf2−/− mice exhibited a mild subclinical chondrodysplasia that rendered the cartilage more susceptible to OA in adult life. However, subcutaneous delivery of FGF-2 was able to slow arthritis in Fgf2−/− mice to levels seen in Fgf2+/+ animals, suggesting that accelerated OA in Fgf2−/− mice was not due to an intrinsic matrix weakness that had arisen during development, but rather that it was due to the loss of FGF-2–mediated suppression of matrix breakdown in postnatal tissue.
The disparate functions of FGF-2 in pre- and postnatal cartilage may be due to differences in receptor expression; FGFR-3 expression is high in the growth plate (31), but FGFR-1 and FGFR-2 are the predominant receptors in healthy mature articular cartilage (Vincent T, Chia S-L: unpublished observations), and it is probably through these that the anticatabolic effects of FGF-2 are occurring. Our results suggest that other FGFs do not compensate for the loss of FGF-2 in adult cartilage. This may be because of the high relative abundance of FGF-2 in articular cartilage compared with other FGFs, and also because of the mechanical control of its bioavailability, which ensures that FGF-2–mediated signaling occurs during weight bearing as well as following chondral injury (7, 8).
The role of FGF-2 in tissue injury responses in other organs has also emerged in recent years; Fgf2−/− mice have delayed healing following excisional skin wounding (32), and FGF-2 up-regulates neurogenesis and protects neurons from degeneration in the adult hippocampus after traumatic brain injury (33). These observations identify FGF-2 as a tissue remodeling cytokine and highlight its importance in tissue homeostasis and repair.
Finally, could loss of the chondroprotective role of FGF-2 contribute to the development of human OA? We have observed strong pericellular staining for perlecan and FGF-2 in OA cartilage, suggesting that their expression is maintained in disease. It is not clear at present whether mechanical signals are unable to activate FGF-2–mediated chondroprotection in damaged tissue or whether anticatabolic processes are simply overwhelmed by degradative ones in established OA. Nonetheless, understanding the molecular basis for these observations should enable us to harness the properties of such anticatabolic cytokines in the future management of OA.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Vincent had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Chia, Sawaji, Inglis, Saklatvala, Vincent.
Acquisition of data. Chia, Burleigh, McLean, Inglis.
Analysis and interpretation of data. Chia, Sawaji, Saklatvala, Vincent.