To determine whether the basic fibroblast growth factor (bFGF) mediates signal transduction in articular cartilage in response to mechanical loading.
To determine whether the basic fibroblast growth factor (bFGF) mediates signal transduction in articular cartilage in response to mechanical loading.
Articular cartilage from porcine metacarpophalangeal or knee joints was cyclically loaded (62.5–250N) for 2 minutes in the absence or presence of a bFGF receptor inhibitor, SB 402451 (250 nM). Activation of the extracellularly regulated kinase MAP kinase ERK was measured by Western blot analysis. Changes in protein synthesis were assessed by measuring the incorporation of 35S-Met/Cys into proteins secreted by cartilage explants or by isolated chondrocytes.
Rapid activation of the ERK MAP kinase occurred when articular cartilage was loaded. This was dependent upon release of the bFGF because it was restricted by the FGF receptor inhibitor. Loaded explants were shown to release bFGF. Loading or bFGF stimulation of explants induced synthesis and secretion of tissue inhibitor of metalloproteinases 1 (TIMP-1), which was inhibited by SB 402451.
Cyclical loading of articular cartilage causes bFGF-dependent activation of ERK and synthesis of TIMP-1.
Mechanical factors are important for the maintenance of normal articular cartilage, as well as in its degeneration. For many years, it has been accepted that mechanical injury predisposes cartilage to the development of osteoarthritis (OA). Historically, this is illustrated by syndromes such as “coal miner's back” and “foundry worker's elbow.” OA of the knee is more common in individuals whose occupations involve heavy labor (1), and it is well established that sporting injuries predispose the individual to premature disease. In recent years, interest has extended to the effects of noninjurious mechanical stimuli. There is mounting evidence that weight bearing may be important for the maintenance of cartilage integrity by providing an anabolic stimulus in a manner analogous to bone. In young horses, exercise is crucial for development and maintenance of cartilage (2). Furthermore, thinning of knee articular cartilage occurs within 6 months of loss of weight bearing in patients with spinal cord injury (3).
Biosynthetic responses of articular cartilage to mechanical loading have been detected in vitro. In general, increased sulfate incorporation, an indicator of proteoglycan synthesis, is detected in cartilage explants that have been cyclically loaded (4, 5), whereas decreased synthesis occurs following static loading (5–7). The latter effect may be due to a reduction in fluid flow, with resultant decreases in oxygen uptake and ATP production (8). These findings have been confirmed at the messenger RNA (mRNA) level for aggrecan (7, 9, 10). Other genes that have been studied in similar systems include matrix metalloproteinases (MMPs), whose expression increases with cyclic mechanical stimulation at both the protein (11) and mRNA levels (12). Other investigators have failed to demonstrate such changes in MMP expression, but have reported a reduction in type II collagen mRNA upon cyclical loading (13). Cell death, by apoptosis and necrosis, is increased in cartilage explants subjected to high static loads, especially in the superficial zone (14–16), although loading the tissue still attached to bone causes less cell death (17).
Much speculation has surrounded the transduction mechanism of mechanical stimuli in the cells in articular cartilage. Integrins have long been considered candidate mechanotransducers. They are abundant in articular cartilage, where their roles include tethering extracellular matrix proteins and participating in focal adhesion formation (18). Although their short intracellular domains are thought not to have signaling activity, they are able to bind signaling molecules such as the adaptor protein Shc (19) and to activate tyrosine kinase receptors in the absence of receptor ligation (20). Integrin-dependent responses to mechanical stimuli have been reported in a variety of cell types, including bone (21) and isolated chondrocytes (22). Other postulated mechanisms of mechanotransduction include activation of potassium channels (23) and induction of the signaling molecules SMAD-6 and SMAD-7 (24) in endothelial cells in response to shear stress. There is currently no direct evidence that any of these mechanisms is involved in mechanical signal transduction in articular cartilage.
We have previously shown that the extracellularly regulated kinases (ERKs 1 and 2), also known as p44 and p42 MAPKs, are rapidly activated in a sustained manner upon cutting cartilage (25). This activation was due to the release of basic fibroblast growth factor (bFGF) from an extracellular heparan sulfate–bound pool. The existence of an extracellular pool led us to investigate the possibility that mechanical loading of cartilage might, by displacing components of the extracellular matrix, cause an FGF-dependent activation of ERK and consequent changes in gene expression.
Porcine articular cartilage was from either the metacarpophalangeal joints or the knee joints of 3–6-month-old pigs obtained from a local slaughterhouse within 16 hours of slaughter.
Antibodies to ERKs 1 and 2 (9122) and phosphorylated ERKs 1 and 2 (9101S) were from New England Biolabs (Beverly, MA). Neutralizing and immunoblotting antibodies to bFGF and recombinant platelet-derived growth factor (PDGF) were obtained from Upstate Biotechnology (Milton Keynes, UK). Secondary antibodies were purchased from Dako (Glostrup, Denmark). Recombinant bFGF was from PeproTech (Rocky Hill, NJ). Recombinant insulin-like growth factor 1 (IGF-1), epidermal growth factor (EGF), and vascular endothelial growth factor (VEGF) were from R&D Systems (Minneapolis, MN). Fibroblast growth factor receptor 1 (FGFR-1) inhibitor (SB 402451, also known as PD 173074) (26) was a generous gift of Stephen Skaper (Glaxo SmithKline, London, UK). Enhanced chemiluminescence (ECL) reagents and radioisotopes (35S-Met/Cys) were from Amersham Pharmacia Biotech (Little Chalfont, UK). Pronase E was from BDH Chemicals (Poole, UK). Collagenase 1A was from Sigma (Poole, UK). Dulbecco's modified Eagle's medium (DMEM) and fetal calf serum (FCS) were obtained from PAA Laboratories (Linz, Austria). All other reagents were the best available grade from Sigma.
Cartilage explants (0.125 gm) were cultured in DMEM (500 μl) with 25 mmoles of HEPES (25 mM), penicillin (100 units/ml), streptomycin (100 μg/ml), and amphotericin (2 μg/ml). Isolated chondrocytes were incubated in DMEM supplemented with 10% FCS.
Cartilage was incubated with Pronase E (1 mg/ml/gm of cartilage) for 30 minutes at 37°C, followed by collagenase (1 mg/ml/gm of cartilage) for 5 hours at 37°C. The digest was passed through a 70-μm cell strainer, then centrifuged at 1,400 revolutions per minute for 5 minutes. Pellets were washed, then resuspended in DMEM containing 10% FCS. Cells were counted and plated on 24-well plates (1.5 cm diameter) at a density of 1 million cells per well (100% confluent).
Metacarpal bones with intact articular cartilage were dissected from porcine feet within 16 hours of slaughter. Two metacarpal bones, which were fused, were embedded in cement and placed in a humidified chamber at 37°C for 1 hour to equilibrate (Figure 1). The cartilage was loaded using an Instron 5565 materials-testing machine (Instron, High Wycombe, UK) with a custom-made loading device carrying a ball-bearing probe, which fit the concave shape of the articular surface. For each metacarpal bone, one surface was loaded and the other was used as a control. Cartilage was dynamically loaded (1 second on, 1 second off for 2 minutes), with peak loads ranging from 62.5 to 250N using an Instron Merlin computer control system. Pressure-sensitive paper (Fuji Photo Film Company, Kanagawa, Japan) confirmed good coverage (∼100 mm2) by the probe, and stress was estimated to be between 0.6 and 2.5 mPa. Following loading, metacarpals were left at 37°C for 20 minutes, or for the times specified, then dissected rapidly into lysis buffer (1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate [SDS], 50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 10 mM NaF, 25 mM β-glycerophosphate, 10 mM tetrasodium pyrophosphate, 1 mM sodium vanadate, supplemented with fresh protease inhibitors) on ice. After 1 hour, lysates were removed and examined for the presence of phosphorylated and total ERKs by Western blot analysis.
Articular cartilage was dissected from the tibial plateaus of 3–6-month-old pigs within 16 hours of slaughter. Explants (0.6 cm × 0.6 cm × 0.2 cm) were cultured in serum-free medium for 48 hours with either vehicle or SB 402451 (250 nM). Explants were dynamically loaded in an unconfined manner using an Instron 5565 materials-testing machine with a custom-made loading device with a plane-ended indenter (10-mm diameter) (Figure 2). Loads ranged from 62.5 to 250N (1.7–6.9 MPa). Cartilage was immersed in medium and loaded for the times specified. All experiments were performed in a humidified chamber at 37°C. After loading, explants were either left unloaded for 18 minutes before lysis or were cultured for longer periods for metabolic labeling experiments. Medium conditioned by loading was retained for analysis.
After 30 minutes, 250 μl of ice-cold dissociative lysis buffer was added and the samples were left mixing at 4°C. Lysates were removed after 60 minutes and clarified by centrifugation. Samples were mixed with 4× sample buffer (0.4M Tris HCl, pH 6.88, 40% glycerol, 0.4% SDS, 1% β-mercaptoethanol, 0.01% bromphenol blue), boiled for 5 minutes, run on 12.5% SDS-polyacrylamide gel, and transferred to polyvinylidene difluoride membranes. Membranes were blocked (30 minutes) in 5% dried milk, then treated with primary antibody (1 hour), washed 3 times (phosphate buffered saline/0.05% Tween 20), and incubated (1 hour) with secondary horseradish peroxidase–conjugated antibody. The signal was developed by ECL and visualized by autoradiography.
Chondrocytes were washed in serum-free medium for 1 hour, then stimulated overnight. The following day, they were washed in Met/Cys-free medium and incubated with 15 μCi/well 35S-Met/Cys. After 6 hours, medium was removed, centrifuged (13,000 rpm for 5 minutes), mixed with sample buffer, run on 12.5% SDS-polyacrylamide gel, and silver-stained. Newly synthesized proteins were visualized by autoradiography. Silver-stained bands that corresponded to labeled ones were excised and identified by mass spectrometry (25).
Explants were loaded (cyclically for 4 hours at 100N), stimulated with interleukin-1 (IL-1; 120 ng/ml) or bFGF (200 ng/ml), or cut. The latter involved cutting the explant (0.6 × 0.6 × 0.2 cm) into pieces ∼0.1 × 0.1 × 0.1 cm with a sharp scalpel. After overnight incubation, explants were washed with Met/Cys-free medium, then pulsed with 25 μCi/well 35S-Met/Cys. After 6 hours, medium was removed and the level of glycosaminoglycan (GAG) in each sample was measured by dimethylmethylene blue assay. To remove proteoglycan selectively, 3 mg of cetylpyridinium chloride/mg of GAG was added to each sample. Samples were mixed and then centrifuged (13,000 rpm for 5 minutes) to precipitate GAG chains. Supernatants were mixed with sample buffer and processed as above.
In-gel digestion of silver-stainable bands with trypsin was performed according to published methods (27–29) and modified for use with a robotic digestion system (30) (Investigator ProGest; Genomic Solutions, Huntington, UK). Aliquots (0.5 μl) of the digest solution were deposited onto the matrix film, allowed to dry, and desalted by brief washing with a few microliters of 0.1% trifluoroacetic acid. Matrix-assisted laser desorption ionization mass spectra were recorded with a TofSpec 2E spectrometer (Micromass, Manchester, UK), and the resulting peptide mass fingerprints were searched against a local copy of the National Center for Biotechnology Information (NCBI) nonredundant database (http://www.ncbi.nlm.nih.gov) using the Protein Probe search engine (Micromass).
Desalted samples were dissolved in 1–2 μl of 50% methanol/0.1% aqueous formic acid and mounted in the source of a quadrupole time-of-flight spectrometer (Micromass). Amino acid sequences were determined semimanually from the daughter ion spectra using the Pepseq peptide sequencing program (Micromass). The deduced sequence was searched against the nonredundant database at the National Center for Biotechnology Information using BLAST (31).
Initial experiments were carried out by loading articular cartilage attached to the subchondral bone in order to approximate physiologic conditions as closely as possible (Figure 1). Cyclical loading of cartilage for 2 minutes (followed by 18 minutes resting at 37°C) resulted in activation of ERKs 1 and 2, as shown by the increase in the phosphorylated forms (Figure 3). Activation was caused by all loads tested, and showed a graded response with the magnitude of load. The extent of activation was equivalent to that induced by cutting cartilage (Figure 3), where cartilage was dissected from the metacarpal and incubated in serum-free medium for 20 minutes before lysis. Unloaded metacarpals were treated under identical conditions, and dissected into lysis buffer after 20 minutes. No activation of ERK was seen in these explants, or in cartilage dissected directly from the joint into lysis buffer with no incubation. Western blot analysis for ERK protein showed that similar amounts of the kinase were loaded in each lane (Figure 3).
To determine the time course of ERK activation upon loading, metacarpals were cyclically loaded for 2 minutes, then either the articular cartilage was dissected and extracted immediately or the metacarpals were left for increasing amounts of time before dissection and extraction. Activation of ERK was rapid, since it was seen immediately following loading (2 minutes), and was maintained for up to 15 minutes (Figure 4). Western blot analysis for ERK protein showed that similar amounts of the kinase were added in each lane (Figure 4). Similar experiments, in which cartilage was extracted up to 3 hours following loading, showed that activation was sustained (results not shown).
We have previously shown that ERK is rapidly activated upon cutting articular cartilage due to the release of bFGF from an extracellular pool (25). To determine whether the ERK activation caused by loading was mediated by bFGF, we performed experiments in either the absence or the presence of the FGFR inhibitor SB 402451. This is a small lipophilic molecule that binds to the intracellular ATP binding site of the receptor tyrosine kinase and inhibits activation. We have estimated the 50% inhibition concentration of SB 402451 to be ∼10 nM on bFGF-stimulated chondrocytes, and 1–10 μM on PDGF-, IGF-, and EGF-stimulated cells (data not shown). To allow the inhibitor to penetrate the tissue, it was necessary to switch to a system in which cartilage explants were used (Figure 2). In this system, higher concentrations (>100 nM) of SB 402451 were required to inhibit bFGF-stimulated ERK activation (Figure 5a). For subsequent explant experiments, 250 nM SB 402451 was used to fully inhibit the bFGF effect. Explants were dissected and preincubated for 48 hours with serum-free medium containing either the vehicle or 250 nM SB 402451. Medium with or without the inhibitor was changed daily. Unconfined cyclical loading of cartilage explants for 2 minutes activated ERK at all loads tested (Figure 5). This was inhibited by the presence of SB 402451. As expected, ERK activation in response to cutting was also inhibited, while activation by IL-1, which does not activate ERK through a tyrosine kinase receptor, was unaffected.
ERK is activated in primary cultures of chondrocytes in monolayer when they are treated with bFGF, and this is prevented by the presence of neutralizing antibodies to the growth factor (Figure 6). ERK was also activated when the cells were treated with medium in which explants had been finely chopped. This activation was inhibited by the same antibodies, indicating the release of bFGF when cartilage was cut (Figure 6). To see whether bFGF was released upon loading the explants, the medium surrounding the loaded cartilage was removed and tested on chondrocytes. The medium activated ERK, and the active material was neutralized by antibodies to bFGF, but not by an isotype-matched antibody control. No ERK activation occurred with medium conditioned by rested, nonloaded cartilage explants (results not shown). To confirm the release of bFGF following loading, medium conditioned by cut, loaded, or rested (in serum-free medium for 30 minutes) cartilage was run on an SDS-polyacrylamide gel and immunoblotted for bFGF. Recombinant bFGF (5 ng) gave a strong signal at 17 kd. Similar bands were detected in medium from cut and loaded cartilage, but not in medium from rested cartilage (Figure 6b). To exclude release of intracellular bFGF as a result of cell injury, medium conditioned by rested or loaded cartilage was assayed for lactate dehydrogenase (LDH), an intracellular enzyme. No LDH was detected in medium conditioned by rested or loaded cartilage (results not shown).
Stimulation of chondrocytes in monolayer culture with bFGF induced a number of secreted proteins as assessed by the incorporation of radiolabeled Met/Cys (Figure 7a). These included TIMP-1, MMP-1, and gp38, which were previously identified by mass spectrometry (25). Interestingly, stimulation with other growth factors, such as IGF-1, EGF, PDGF, and VEGF, did not change the pattern of protein expression despite the fact that, with the exception of VEGF, they are all able to activate ERK in isolated chondrocytes (Figure 7b). IL-1, an inflammatory cytokine that is known to have profound effects on articular cartilage, also induced changes in gene expression. Some of these changes were similar to those induced by bFGF. The pattern of secreted proteins induced by bFGF was similar when cartilage explants, rather than isolated chondrocytes, were stimulated (results not shown), and TIMP-1 was the most prominent band. In addition to ERK, bFGF also stimulates the p38 MAP kinase pathway. To determine whether induction of TIMP-1 and MMPs by bFGF was through the ERK or the p38 MAP kinase pathway, chondrocytes were stimulated with bFGF in the presence of either the MEK inhibitor, U0126 (20 μM), or the p38 inhibitor, SB 202190 (1 μM). Induction of TIMP-1 was unaffected by SB 202190, but was completely abrogated by U0126, indicating that the changes in protein synthesis in response to bFGF are dependent upon the ERK MAP kinase pathway (results not shown).
To demonstrate an FGF-dependent change in gene expression following cyclic loading, we metabolically labeled cartilage explants that had been loaded for 4 hours at 100N in medium containing either the vehicle or SB 402451 (250 nM). Cartilage explants were also labeled following treatment with IL-1 or recombinant bFGF, or following cutting. As expected, there was an induction of TIMP-1 upon either IL-1 or bFGF stimulation (Figure 8). TIMP-1 induction was also observed after the explant was mechanically loaded or cut. Induction of TIMP-1 by bFGF, cutting, or loading was inhibited when the experiments were performed in the presence of SB 402451. This was consistent with bFGF mediating changes in gene expression, as well as protein kinase activation, caused by mechanical stimuli. TIMP-1 induction following IL-1 stimulation was, as expected, not affected by the presence of the inhibitor.
The results strongly suggest that ERK activation induced by loading articular cartilage is mediated by bFGF. This activation is rapid, occurring within 2 minutes of the stimulus. Activation was seen over a wide range of loads, and the magnitude of the response was graded accordingly. These loads are considered to be within the normal physiologic range generated during weight bearing. For example, peak compressive stresses in humans while walking range from 5 to 8 MPa (32, 33).
Many stimuli, including other growth factors and IL-1, activate the ERK pathway. That ERK activation upon loading is mediated by bFGF is inferred from the specificity of the SB 402451 inhibitor, the neutralization of the active material released by loading cartilage with an antibody to bFGF, and the identification of bFGF in the medium by immunoblotting. Furthermore, it is less likely that gene induction upon loading is due to other growth factors, because stimulation of chondrocytes with several of these did not change the pattern of secreted proteins. This is perhaps surprising because the majority of growth factors tested activated ERK, indicating cell surface receptor expression. Basic FGF activates both the p38 and the ERK pathways in chondrocytes, and it was possible that changes in gene expression were dependent upon the p38 rather than the ERK MAP kinase pathway. This was not the case, because induction of TIMP-1 and MMP-1 by bFGF was inhibited in the presence of the MEK inhibitor U0126, but not the P38 inhibitor SB 202190.
The rapid activation of ERK upon loading suggests that release of bFGF is from a preexisting pool. It was previously found that articular cartilage has an extracellular heparan sulfate–bound pool of bFGF, which could be released by treating the tissue with the heparan sulfate–degrading enzyme heparitinase (25). Cutting cartilage released bFGF from this store. Basic FGF may also be found in the cytoplasm of cells and could be released upon cell injury. However, the low levels (5–10%) of total intracellular LDH released upon finely cutting the tissue suggested that much of the bFGF released was from an extracellular source. The cyclical mechanical loading used in the present experiments caused no detectable LDH release and seems unlikely to have caused cell breakage. Although it seems more likely that the bFGF is from an extracellular source, it is conceivable that loading causes an active secretion of bFGF from an intracellular source. The mechanism whereby extracellular bFGF, which is presumably tightly bound to heparan sulfate, activates cells is unclear. It might be released by the activation of endogenous proteases (34), heparanases (35), or be made available at the cell surface as the tissue deforms.
Basic FGF has been implicated as a mediator of mechanical signaling in other systems. Cultured vascular endothelial cells released bFGF when subjected to shear stress, and this led to their proliferation and differentiation (36). It was suggested that the release of bFGF from the cells was integrin-dependent because blocking αvβ3 (a fibronectin receptor) inhibited release. However, we have found no evidence that the release of bFGF upon cutting cartilage is integrin-dependent. We have been unable to detect activation of ERK in isolated chondrocytes or explants with integrin ligands, such as fibronectin, or specific peptides containing the RGD motif (Vincent TL, et al: unpublished observations). We have also been unable to inhibit release of bFGF following cutting with inhibitory RGD peptides (Vincent TL, et al: unpublished observations).
Could bFGF mediate the biologic responses to loading that others have observed in vitro? These include increased proteoglycan synthesis (4) and induction of genes involved in matrix remodeling, such as MMPs (11). Increases in proteoglycan synthesis have been detected in subconfluent rabbit costal chondrocytes (37) and in adult bovine cartilage explants following bFGF stimulation (38). We did not find significant changes in proteoglycan synthesis in either isolated chondrocytes or explants following stimulation with high doses of bFGF. However, we found, both here and previously (25), that bFGF induces MMP-1 expression in chondrocytes.
The question remains whether bFGF is responsible for beneficial anabolic effects, or is catabolic and, if so, whether it plays a part in cartilage degeneration. It may be that at low mechanical stress bFGF predominantly stimulates anabolic pathways (e.g., up-regulation of TIMP-1), which maintain cartilage integrity. At high stresses, such as following injury, excessive activation of the pathway may lead to changes in gene expression that favor catabolic events and lead to degeneration of the tissue. This bidirectional response of bFGF is consistent with the findings of Sah et al (38), who observed that the catabolic or anabolic nature of the response varied with the concentration of the growth factor. It is likely that mechanical injury activates additional pathways, and the context of the bFGF signal may be important in determining whether there is an anabolic or catabolic outcome. Either way, understanding the mechanisms of mechanical sensing in articular cartilage is likely to provide insight into the physiology and pathology of the tissue.