Activin A is an anticatabolic autocrine cytokine in articular cartilage whose production is controlled by fibroblast growth factor 2 and NF-κB




Proteomic analysis has previously shown that activin A, a member of the transforming growth factor β family, is produced by human articular cartilage. This study was undertaken to investigate whether activin A affects cartilage matrix catabolism and how its production is regulated.


The effect of exogenous activin A on interleukin-1–induced aggrecanase-generated neoepitope production was assessed by Western blotting, using medium from human cartilage explants. Levels of activin A production were determined by enzyme-linked immunosorbent assay. For genes of interest, messenger RNA (mRNA) induction in cartilage explants or primary chondrocyte monolayers was assessed by reverse transcriptase–polymerase chain reaction. Activin A activity in cartilage explant medium was measured by incubating it with human dermal fibroblasts and determining the increase in phospho-Smad2 by Western blotting.


Activin A (1–10 ng/ml) suppressed aggrecanase-mediated cleavage of aggrecan in human articular cartilage. Activin A mRNA and protein secretion were induced by dissection and culture of human or porcine articular cartilage. This activin A was biologically active. Its production was due to an active cellular process and was enhanced in osteoarthritic (OA) tissue. Activin A production on dissection was reduced by 80% by the fibroblast growth factor (FGF) receptor inhibitor PD173074 and by 70% by the IKK inhibitor BMS345541.


Activin A is potentially an anticatabolic molecule in articular cartilage. Its expression is induced by wounding in an FGF-2– and NF-κB–dependent manner. OA cartilage produced more activin A than did normal cartilage in vitro.

Proteomic analysis of the proteins secreted by human articular cartilage in vitro has revealed activin A as a potential regulatory molecule of the tissue (1). Activin A is a member of the transforming growth factor β (TGFβ) family (2, 3) and is a homodimer of proteolytically processed inhibin βA chains. Inhibin β-chains also heterodimerize with the inhibin α-chain to form inhibins. Activins bind and activate activin receptors on the cell surface. Inhibin binds to the receptors but does not activate them.

There are 4 genetically different mammalian inhibin β-chains: A, B, C, and E. These potentially form either homodimers or heterodimers, giving rise to activins A, B, AB, and so on. Proteomic analysis of cartilage has not revealed any inhibin α-chains, suggesting that activin, rather than inhibin, is the likely product (1). Both pro–inhibin βA (35-kd) and processed inhibin βA (14-kd) forms have been found, but no inhibin βB or βC chains have been identified. Activins, besides binding to cell surface receptors, also bind to follistatin or follistatin-related protein, which act as inhibitors by preventing binding to the receptors.

Activin was originally discovered as a molecule in ovarian follicular fluid that caused pituitary cells in culture to secrete follicle-stimulating hormone, while inhibin was found to be an inhibitor of this process (2, 3). Inhibin βA–knockout mice have severe craniofacial abnormalities, suggesting that activin A or inhibin is important for intramembranous development of bone (4). Activin A is produced at sites of inflammation and is induced in cultured cells by inflammatory stimuli, such as interleukin-1 (IL-1) or bacterial lipopolysaccharide, and by growth factors (5). It is strongly expressed in injured skin (6), in inflamed gastrointestinal mucosa (7, 8), and in the lung in conditions associated with pulmonary fibrosis (9). It is also found in synovial fluid in inflammatory arthritis (10).

Overexpression of follistatin in the basal epidermis of the skin was found to delay wound healing in mice, while overexpression of activin A increased the amount of granulation tissue formed and subsequent scarring, suggesting that activins play a physiologic role in the healing process (11). The function of activin A in inflammation is uncertain. Its relationship to TGFβ makes it likely to be a regulatory molecule, but its diverse effects on gene expression in different cell types are consistent with both proinflammatory and antiinflammatory roles (3).

The activin receptors are related to those of TGFβ (12). Activins and inhibins bind to the activin type II or IIB receptors. The type II receptors are constitutively active serine/threonine protein kinases which, when engaged, bind, phosphorylate, and activate the type I receptor, activin receptor–like kinase 4 (ALK-4), which is also a serine/threonine protein kinase. ALK-4 then phosphorylates the transcription factors Smad2 and Smad3. Smad2 and Smad3 are also activated by the TGFβ receptor. TGFβ is a pleiotropic cytokine. It is immunomodulatory, chemotactic for leukocytes, stimulates deposition of collagen, and may inhibit expression of inflammatory response proteins (13). It can reduce the IL-1–induced degradation of cartilage proteoglycan (14–16).

Because activin A shares a postreceptor mechanism with TGFβ, we investigated the effect of activin A on articular cartilage. It physiologically antagonized aggrecanolysis stimulated by IL-1. Osteoarthritic (OA) cartilage explants appeared to secrete more activin A than normal tissue, so we therefore explored how this potentially autocrine anticatabolic cytokine might act and how its production in cartilage is controlled.



Porcine articular cartilage was dissected from the metacarpophalangeal (MCP) joints of freshly slaughtered 3–6-month-old pigs obtained from a local abattoir. Normal and OA human femoral head cartilage and OA knee cartilage were obtained from patients at Charing Cross Hospital (London, UK) who were undergoing joint replacement surgery for either fractured femoral neck or OA. Normal knee cartilage was obtained from patients at the Royal National Orthopaedic Hospital (RNOH; Stanmore, Middlesex, UK) who were undergoing surgery for tumors involving the lower leg (not involving the knee). All human samples were used with consent from the local ethics committees (Riverside Research Ethics Committee and Joint RNOH/Institute of Orthopaedics and Musculoskeletal Science Research Ethics Committee), and written informed consent was obtained from all patients. Human dermal fibroblasts were obtained from infant foreskin and were used at passages 7–16.


Fetal calf serum and Dulbecco's modified Eagle's medium (DMEM) (both from BioWhittaker, Verviers, Belgium) were supplemented with 25 mM HEPES, 1.25 units/ml penicillin, 100 μg/ml streptomycin, and 2 μg/ml amphotericin B. The DuoSet enzyme-linked immunosorbent assay (ELISA) kit for activin A was obtained from R&D Systems (Minneapolis, MN). Recombinant human activin A, TGFβ, follistatin, and neutralizing mouse monoclonal antibodies (mAb) to activin A and TGFβ were also from R&D Systems. Rabbit polyclonal antibody to phosphorylated Smad2 was obtained from Cell Signaling Technology (Beverly, MA), and murine mAb to phosphorylated ERK was from Sigma (Poole, UK). Rabbit polyclonal antibody to ERK was from Santa Cruz Biotechnology (Santa Cruz, CA). Chondroitinase ABC and keratanase were obtained from Seikagaku Kogyo (Tokyo, Japan). Mouse mAb to aggrecan neoepitope ARGSV was provided by Bruce Caterson (University of Cardiff, Cardiff, UK). Sheep tissue inhibitor of metalloproteinases 1 (TIMP-1) antibody was provided by Hideaki Nagase (Kennedy Institute, London, UK), and mAb to TIMP-3 was obtained from Fuji Chemical (Takaoka, Japan). Horseradish peroxidase (HRP)–linked anti-mouse IgG or anti-rabbit IgG was from Dako (Glostrup, Denmark). The electrochemiluminescence Western blot detection system was from Amersham Biosciences (Little Chalfont, UK).

The fibroblast growth factor receptor (FGFR) kinase inhibitor PD173074, also called SB402451, was provided by Stephen Skaper (GlaxoSmithKline, London, UK). U0126 was from Promega (Madison, WI), and MG-132 was from Biomol (Plymouth Meeting, PA). SB202190, LY294002, BMS345541, PP1, PP2, Gö6983, and Gö6976 were from Calbiochem (La Jolla, CA).

Preparation and culture of cartilage explants.

Cartilage was dissected into DMEM (supplemented as indicated above) and washed briefly to remove synovial fluid. Cartilage was cut into pieces ∼3–4 mm3. It was then cultured at 0.5 gm/ml, 1 gm/ml, or as indicated, in DMEM for up to 20 hours.

Assay for aggrecanase-generated neoepitope.

Three pieces of cartilage were cultured in 1 well of a 48-well plate in 250 μl of medium for 24 hours in the absence or presence of IL-1 (10 ng/ml). For each assay, 3 culture wells were used (i.e., 9 explants). The medium samples were deglycosylated at 37°C overnight with chondroitinase ABC and keratanase in 50 mM Tris HCl (pH 7.5), 50 mM sodium acetate, and 10 mM EDTA. Deglycosylated proteoglycan from the medium was precipitated with ice-cold acetone, air-dried, and resuspended in sodium dodecyl sulfate (SDS) sample buffer. Samples were loaded onto an 8% Tris–glycine gel and separated by SDS–polyacrylamide gel electrophoresis (PAGE) under reducing conditions. Protein was electrophoretically transferred to polyvinylidene difluoride (PVDF) membrane, and aggrecan fragments were detected with the BC-3 antibody to the ARGSV neoepitope. After reaction with HRP-linked anti-mouse IgG antibody, proteins were detected by electrochemiluminescence.

Western blotting for determination of phospho-ERK and phospho-Smad2 content.

Cartilage tissue was snap-frozen and kept at −70°C until used. Cellular protein was extracted with radioimmunoprecipitation assay (RIPA) buffer (20 mM Tris HCl [pH 7.4], 150 mM NaCl, 5 mM EDTA, 1% [volume/volume] Triton X-100, 0.1% [weight/volume] SDS, 1% [w/v] sodium deoxycholate, 0.5% [v/v] Igepal CA-630, 1 mM phenylmethylsulfonyl fluoride, 10 μM E64, 1 μg/ml pepstatin A, and 10 μg/ml aprotinin) at 4°C for 2 hours with agitation. Extracts were clarified by centrifugation (at 13,000 revolutions per minute for 20 minutes) and mixed with reducing SDS sample buffer. Samples were boiled, subjected to SDS-PAGE on a 10% gel, transferred to PVDF membrane, and immunostained, and proteins were visualized by electrochemiluminescence, all as described above. For cells, fibroblasts were lysed in ice-cold RIPA buffer and left on ice for 10 minutes. Lysates were scraped and aspirated from the culture surface, clarified by centrifugation, and mixed with reducing SDS sample buffer. Samples were boiled, subjected to SDS-PAGE on a 12.5% gel, and immunostained for phospho-Smad2, which was visualized by electrochemiluminescence as described above.

RNA isolation and reverse transcriptase–polymerase chain reaction (RT-PCR).

Approximately 200 mg of cartilage was homogenized with a Polytron rotor-stator homogenizer in 3 ml of TRI Reagent (Helena Biosciences, Gateshead, UK), then extracted with chloroform. The aqueous phase was mixed with an equal volume of 70% (v/v) ethanol, and RNA was extracted with the QIAamp RNA Blood Mini kit (Qiagen, Chatsworth, CA). Total RNA (1 μg) was reverse-transcribed using Superscript II (Invitrogen, San Diego, CA). PCR was performed using Ready-to-Go PCR beads (Amersham Biosciences) with the appropriate primers. Amplified DNA was analyzed on a 1% (w/v) agarose gel and then visualized by ethidium bromide staining.

RT-PCR primers (MWG Biotech, Ebersberg, Germany) and conditions were as follows: for ADAMTS-4, sense 5′-ACCACTTTGACACAGCCATTCTG-3′ and antisense 5′-ACCCCCACAGGTCCGAGAGCAG-3′ (annealing at 63°C, 27 cycles); for ADAMTS-5, sense 5′-TGTGCTGTGATTGAAGACGAT-3′ and antisense 5′-GACTGCAGGAGCGGTGAGTGG-3′ (annealing at 60°C, 24 cycles); for inhibin βA (17), sense 5′-GACATCCGGACTGCCTGCGAGCAG-3′ and antisense 5′-GTAGCCGGACGGAGCGATTAGCCAGTC-3′ (annealing at 65°C, 30 cycles); and for GAPDH, sense 5′-CATGGAGAAGGCTGGGGCTC-3′ and antisense 5′-ATGAGGTCCACCACCCTGTT-3′ (annealing at 60°C, 23 cycles).


Physiologic antagonism of the aggrecanase-mediated cleavage of aggrecan in articular cartilage by activin A.

Explants of normal human femoral condylar cartilage were rested for 24 hours after dissection into culture medium. They were then stimulated with IL-1 or were left resting for a further 24 hours, after which the culture medium was analyzed by Western blotting for released aggrecan fragments containing the neoepitope ARGSV. This sequence forms the new N-terminal sequence generated by aggrecanase cleavage of the major cartilage proteoglycan aggrecan between E373 and A374. The likely potential aggrecanases in cartilage are ADAMTS-4 and ADAMTS-5 (18–20).

IL-1 stimulation caused the expected release of ARGSV-bearing fragments from the cartilage (Figure 1). Inclusion of recombinant human activin A in the culture, which was added 1 hour before the addition of IL-1, suppressed the release of the proteoglycan fragment in a dose-dependent manner. The results shown were obtained with cartilage from a 16-year-old girl (Figure 1A) and a 51-year-old woman (Figure 1B). Suppression occurred with 1–10 ng/ml activin A. The steady-state levels of both ADAMTS-4 and ADAMTS-5 mRNA were increased by IL-1 stimulation of isolated human chondrocytes, but these were unaffected by activin A at 10 ng/ml (Figure 1C). Some suppression of ADAMTS-5 mRNA was seen at the high concentration of 100 ng/ml. We have been unable to detect either ADAMTS-4 or ADAMTS-5 proteins by Western blotting with a number of different antibodies.

Figure 1.

Activin A inhibition of aggrecan cleavage in human articular cartilage stimulated with interleukin-1α (IL-1α). A and B, Normal human femoral condylar cartilage was dissected from fresh surgical specimens from a 16-year-old girl (A) and a 51-year-old woman (B). Explants were washed and rested overnight in serum-free Dulbecco's modified Eagle's medium (DMEM), preincubated for 30 minutes in the presence or absence of activin A, and then left untreated or treated with 10 ng/ml IL-1α for 24 hours. Aggrecan fragments released into conditioned medium were detected by Western blotting with the antibody to the ARGSV neoepitope as described in Materials and Methods. C, Confluent human chondrocytes (from a 43-year-old man) in primary culture were serum-starved, preincubated for 30 minutes in the presence or absence of activin A, and then left untreated or treated with 10 ng/ml IL-1α for 4 hours. RNA was extracted and subjected to reverse transcriptase–polymerase chain reaction for ADAMTS-4, ADAMTS-5, and GAPDH. D, Human cartilage explants (from a 51-year-old woman) were rested overnight in serum-free DMEM, preincubated for 30 minutes in the presence or absence of activin A, and then left untreated or treated with 10 ng/ml IL-1α for 24 hours. Secreted tissue inhibitor of metalloproteinases 1 (TIMP-1) and TIMP-3 were detected by Western blotting. In experiments for TIMP-3, heparin was added to the medium before addition of activin A, to trap the protein.

Aggrecanases are inhibited by TIMP-3 (21). We examined the effect of activin A on the expression of TIMP-3 and, because of the relationship of activin A to TGFβ, on the expression of TIMP-1 (22). Activin A increased the amount of TIMP-1 in the cartilage explant medium, while IL-1 decreased it (Figure 1D). However, activin A did not increase the amount of TIMP-1 in the presence of IL-1. At the mRNA level, the amount of TIMP-1 was unaffected by IL-1 stimulation, but the modest 2-fold increase caused by activin A (similar to that shown for protein in Figure 1D) was prevented (results not shown). IL-1 stimulation caused a reduction in the amount of TIMP-3 in the medium, and this was also unaffected by the presence of activin A (Figure 1D). At the mRNA level, neither IL-1 nor activin A had any consistent marked effect on the levels of TIMP-3 expression (results not shown).

Activin A induction in cartilage by dissection and culture. The fact that activin A inhibits the aggrecanase cleavage induced by IL-1 suggests that it could be an autocrine anticatabolic molecule. It was therefore important to establish how the production of activin A was controlled. Proteomic analysis showed that both large and small forms of inhibin βA chain were produced by cartilage. These corresponded to the precursor pro–inhibin βA, and the mature processed inhibin βA, which homodimerizes to form activin A. We used ELISA to measure the levels of production of activin A.

Articular cartilage was dissected from porcine MCP joints into DMEM and cultured for 16 hours, and the activin A content in the medium was determined (Figure 2A). To establish the extent to which activin A was newly synthesized or was preformed in the tissue, explants were cultured with the protein synthesis inhibitor cycloheximide (CHX) or were rapidly frozen and thawed 3 times upon dissection. These explants released little detectable activin A, which suggested that the cytokine from the live explants was synthesized in the culture. It is not possible to measure the synthesis of activin A protein in vivo; however, inhibin βA mRNA was barely detectable in cartilage ex vivo, but was strongly induced after dissection and culture (Figure 2B). Higher levels of mRNA were present at 4 hours after injury than at 20 hours after injury. Thus, activin A expression is strongly induced by the physical injury of dissection. Interestingly, explants of human OA hip and knee cartilage secreted more of the cytokine than did tissue from joints with an apparently normal surface (Figure 2C).

Figure 2.

A, Secretion of activin A by freshly dissected cartilage. Porcine cartilage was dissected and cultured for 16 hours in 1 gm/ml serum-free Dulbecco's modified Eagle's medium (DMEM). Some explants were cultured with 10 μg/ml cycloheximide, some were frozen and thawed 3 times before culture, and some were left untreated. Activin A content in medium was determined by enzyme-linked immunosorbent assay (ELISA). Bars show the mean ± SD of triplicate cultures. ∗∗∗ = P < 0.0001 versus untreated cultures, by unpaired t-test. B, Induction of inhibin βA mRNA by dissection. Porcine articular cartilage was dissected directly into ice-cold TRI Reagent and homogenized prior to RNA extraction (0 hours), or was cultured in serum-free DMEM for 4 hours or 20 hours before homogenization and extraction of RNA. Reverse transcriptase–polymerase chain reaction for determination of inhibin βA and GAPDH levels was carried out as described in Materials and Methods. C, Levels of activin A production in normal and osteoarthritic (OA) cartilage. Human articular cartilage was dissected from normal-appearing femoral heads of 4 patients (3 women and 1 man; mean age 83 years [range 78–86 years]) who were undergoing surgery for femoral neck fracture (normal) or from the joints of 8 patients (5 women and 3 men; mean age 70 years [range 54–84 years]) who were undergoing joint replacement for OA. Cartilage explants were cultured for 16 hours in serum-free DMEM, and explants were then incubated for a further 5 hours in 1 gm/ml fresh serum-free DMEM. Medium was collected, and activin A content was determined by ELISA. Horizontal bars show the mean. ∗∗ = P = 0.0043 versus normal cartilage, by unpaired t-test.

Production of biologically active activin A by cartilage.

It was important to show that the cartilage protein was biologically active. It might be inactive, or its activity might be masked by either follistatin or inhibin. Phosphorylation of Smad2 in human skin fibroblasts was used as an indicator of activity of the activin A in cartilage conditioned medium. Fibroblasts were used as a readily available human target cell. They were serum-starved, then treated for 45 minutes with the stimulus, harvested, and subjected to Western blotting for determination of phospho-Smad2 content (Figure 3). Blots were also stained for ERK to check for even loading. The strongly induced phosphorylated band caused by adding TGFβ (Figure 3, lane 2) or activin A (lane 3) was identified as phospho-Smad2. The faint retarded band was unidentified.

Figure 3.

Phosphorylation of Smad2 in human fibroblasts due to the presence of activin A in medium conditioned with live cartilage explants. Conditioned medium from both human and porcine cartilage explant cultures was collected and diluted 1:4 with fresh Dulbecco's modified Eagle's medium (DMEM). These diluted medium samples were preincubated for 2 hours alone or with either a neutralizing antibody (ab) to activin A (5 μg/ml) or recombinant follistatin (2.5 μg/ml). Confluent human dermal fibroblasts were kept overnight in serum-free DMEM and then stimulated for 45 minutes with the test samples in serum-free DMEM. Cells were washed and lysed. The lysates were subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis and immunoblotted for phosphorylated Smad2. Samples were also blotted for ERK protein to show even loading of the lysates. Samples tested contained either no stimulus, 10 ng/ml transforming growth factor β (TGFβ), 50 ng/ml activin A, medium conditioned with human cartilage (Hu. Med.), or medium conditioned with porcine cartilage (Porc. Med.), with or without antibody to activin A or recombinant follistatin as indicated.

The effect of activin A was prevented by preincubating it with either a neutralizing antibody (Figure 3, lane 4) or with follistatin (lane 5). Culture medium conditioned with either human cartilage (lane 6) or porcine cartilage (lane 9) increased the level of phosphorylation of Smad2 as compared with that seen in unstimulated cells (lane 1). Both human and porcine conditioned medium induced additional fainter retarded bands, which were unidentified. Adding the activin A neutralizing antibody nullified the effect of both the human (lane 7 versus lane 6) and porcine (lane 10 versus lane 9) cartilage conditioned medium. Follistatin was also added to cartilage conditioned medium (lanes 8 and 11), and it suppressed Smad2 phosphorylation to the same extent as did antibody to activin A. These results show that the activin A produced by cartilage was active and accounted for essentially all Smad2 phosphorylation caused by the conditioned medium.

Factors regulating the production of activin A by articular cartilage.

Having shown that activin A is synthesized by injured cartilage, is biologically active, and accounts for the TGFβ-like activity produced by tissue in culture, we investigated factors regulating its synthesis. We have previously shown that wounding or dissecting articular cartilage causes the release of FGF-2 from a pericellular pool where it is sequestered by perlecan (23, 24). The release causes prolonged chondrocyte activation, as judged by stimulation of ERK (23). Dissection also rapidly activates inflammatory signaling pathways. There is strong activation of JNK and p38 MAPK and degradation of IκB (25). The signaling upon injury is sufficient to induce the expression of inflammatory response genes. The mechanism of activation of inflammatory signaling is unknown. FGF-2 release explains the prolonged ERK activation, but FGF-2 does not activate JNK or NF-κB in cartilage. It is likely that activin A production was dependent on the chondrocyte activation caused by dissection.

FGF-2 added to porcine explants adapted to culture induced activin A mRNA (Figure 4A). We therefore first assessed the role of FGF-2 in activin A production caused by dissection and culture by using the compound PD173074, an inhibitor of the FGFR tyrosine kinase. This agent is selective for FGFR but also exerts some activity against the platelet-derived growth factor receptor (23, 26). Figure 4B shows the results of an experiment conducted with porcine cartilage. Medium was removed from freshly dissected tissue at the indicated times and replaced with fresh serum-free DMEM. Activin A content in medium samples harvested at different time points was determined. Production of activin A peaked between 8 and 16 hours and declined thereafter. When cartilage was dissected into medium containing PD173074, activin A synthesis was inhibited ∼60% (Figure 4B). A similar experiment was carried out with human femoral head cartilage, and activin A production was inhibited to a similar degree (Figure 4C).

Figure 4.

Dependence of activin A expression on fibroblast growth factor 2 (FGF-2). A, FGF-2 induction of inhibin βA mRNA in rested cartilage explants. Explants of porcine cartilage were rested for 24 hours in serum-free Dulbecco's modified Eagle's medium (DMEM) and then left untreated or treated overnight with 20 ng/ml FGF-2. RNA was extracted and subjected to reverse transcriptase–polymerase chain reaction for inhibin βA and GAPDH. Results of 1 of 2 experiments are shown. B and C, Inhibition of activin A production in freshly dissected porcine (B) and human (C) articular cartilage by FGF receptor inhibitor (PD173074). Articular cartilage was dissected into 1 gm/ml of serum-free DMEM containing vehicle (darkly shaded bars) or 250 nM FGF receptor inhibitor (PD173074) (lightly shaded bars). Medium was collected and replaced with fresh DMEM in the presence or absence of the inhibitor at the times shown, and the activin A content was measured. Experiments were performed in triplicate; bars show the results of 1 representative experiment.

Activation of signaling caused by dissection of cartilage is very rapid, and we have found that to achieve maximal inhibition of signaling with pharmacologic agents, it is better to not only dissect the cartilage into a solution containing the inhibitor, but also to first inject the inhibitor into the joint so that it distributes through the cartilage before the tissue is injured. To test the effect of the FGFR inhibitor on the activation of chondrocytes caused when cartilage is dissected, we injected it into porcine MCP joints. After 2 hours the joints were opened, and the articular cartilage was dissected into DMEM containing the inhibitor. The explants were either lysed immediately or after 30 minutes and were Western blotted for determination of phospho-ERK and total ERK protein content (Figure 5A).

Figure 5.

Efficient blocking of signaling and activin A expression induced by dissection and culture of articular cartilage, by intraarticular injection of inhibitors. Porcine metacarpophalangeal joints were initially injected with 3 ml of serum-free Dulbecco's modified Eagle's medium (DMEM) containing vehicle (veh) (0.1% DMSO), 250 nM fibroblast growth factor receptor inhibitor PD173074 (PD), 20 μM MEK-1 inhibitor U0126 (U), 10 μM of IKK-2 inhibitor BMS345541 (BMS), a combination of PD173074 and BMS345541, or 10 μg/ml cycloheximide (cyclo) or were left untreated (none). Following equilibration at 37°C for 2 hours, cartilage was dissected. A, Cartilage was dissected directly into ice-cold lysis buffer or was cultured for 30 minutes in serum-free DMEM in the presence or absence of inhibitor prior to lysis. Phospho-ERK content in cartilage lysates was determined by Western blotting. The membrane was stripped and reprobed for determination of total ERK content. B, Cartilage was dissected into 1 gm/ml of DMEM in the presence or absence of inhibitor and cultured for 16 hours. Activin A content in the medium was determined by enzyme-linked immunosorbent assay. Bars show the mean ± SD of 8 different cultures. ∗∗∗ = P < 0.0001 versus untreated cartilage (none), by unpaired t-test. C, Cartilage was dissected directly into ice-cold TRI Reagent or was dissected and cultured for 4 hours in DMEM in the presence or absence of inhibitor prior to homogenization of cartilage explants, RNA extraction, and reverse transcriptase–polymerase chain reaction for determination of inhibin βA and GAPDH levels.

The explants left for 30 minutes after dissection showed, as expected, strong phosphorylation of ERK when compared with those lysed immediately. This activation was unaffected by the presence of vehicle, but was nearly completely prevented by the presence of PD173074 (Figure 5A). A similarly effective inhibition of ERK activation was seen when U0126, an inhibitor of MEK-1, was used (Figure 5A). Figure 5B shows that when PD173074 was injected before dissection, activin A production was inhibited over 80%. Taken together, the results of these pharmacologic experiments strongly suggested that activin A production was highly dependent on FGF-2.

It was likely that other signaling pathways activated by injury also contributed to the change in expression of the cytokine. One candidate is the pathway leading to the activation of NF-κB. Performing the dissection in the presence of the IKK inhibitor BMS345541 reduced the production of activin A by about 70% (Figure 5B). Another way of inhibiting activation of NF-κB is to use proteasome inhibitors, which prevent the degradation of phosphorylated IκB. MG-132, a widely used inhibitor, also reduced activin A production (Table 1). Combining the IKK inhibitor with the FGFR inhibitor did not have a significantly greater effect than either agent alone (Figure 5B).

Table 1. Summary of the effects of signaling pathway inhibitors on the production of activin A caused by dissection and culture*
Inhibitor (concentration tested)/target% inhibition, mean (range)
  • *

    Intact porcine metacarpophalangeal joints were injected with 3 ml of serum-free Dulbecco's modified Eagle's medium (DMEM) plus 10 μg/ml cycloheximide, the given inhibitor, or the vehicle (serum-free DMEM plus 0.1% DMSO) or were left untreated and equilibrated at 37°C for 2 hours. Joints were then opened and cartilage was dissected into 1 gm/ml of DMEM in the presence or absence of inhibitor and cultured for 16 hours. Activin A content in medium was determined by enzyme-linked immunosorbent assay. Each experiment was performed with quadruplicate cultures for each condition. Mean activin A production (ng/ml) was measured in the presence and absence of each inhibitor, and the percentage inhibition was calculated. The range shows minimum and maximum percentage inhibition where >1 experiment was performed. FGFR = fibroblast growth factor receptor; PI 3-kinase = phosphatidylinositol 3-kinase; PKC = protein kinase C.

BMS345541 (10 μM)/IKK71 (63–78)
MG-132 (10 μM)/proteasome49 (28–70)
SB202190 (20 μM)/p38 MAPK17 (2–33)
PD173074 (250 nM)/FGFR82 (79–84)
U0126 (10 μM)/MEK-151 (34–66)
LY294002 (10 μM)/PI 3-kinase12
Gö6983/Gö6976 (100 nM)/PKC0
PP2 (10 μM)/Src family kinases60 (52–66)
PP1 (10 μM)/Src family kinases51

To achieve maximal inhibition, we used CHX in the experiment, which reduced activin A production by more than 90% (Figure 5B). Since NF-κB is antiapoptotic, we were concerned that the reduction in activin A production caused by inhibitors of NF-κB activation over a period of 20 hours was due to cell death. However, inhibition of inhibin βA mRNA was seen at an early time point (4 hours) after dissection with each inhibitor alone and in combination (Figure 5C). The p38 MAPK inhibitor SB202190 was also tested but had no significant effect on activin A production under these conditions (Table 1). We were unable to block JNK signaling pharmacologically in cartilage, so its contribution could not be determined. These results suggest that the burst of activin A production after dissection and culture, which peaks between 8 and 16 hours, is dependent on both FGF-2 and NF-κB.

Stimulation of FGFR activates several signaling pathways. In addition to the ERK MAPK pathway (which is probably Ras-dependent), protein kinase C (PKC), phosphatidylinositol 3-kinase (PI 3-kinase), and Src family kinases are all likely to be activated. The effects of inhibitors of these pathways were tested (Table 1). LY294002, a PI 3-kinase inhibitor, and PKC inhibitors had no effect on activin A production. However, both the MEK-1 inhibitor U0126 and the Src family kinase inhibitors PP1 and PP2 caused inhibition (Figure 5B and Table 1). It is therefore likely that the ERK pathway and Src family kinases are important for the effect of FGF-2.


The results of the present study showed that a function of activin A in cartilage may be to oppose cartilage-catabolizing stimuli such as IL-1. This means it is potentially an autocrine anticatabolic cytokine and might limit cartilage degradation in inflammation. The mechanism of action is unclear, since activin A did not regulate mRNA for the known IL-1–inducible aggrecanases ADAMTS-4 and ADAMTS-5. It also did not alter the amounts of TIMP-3 mRNA or protein in the medium. Activin A modestly increased TIMP-1 production, but this was not seen when IL-1 was present (TIMP-1 is only a weak inhibitor of aggrecanases). Investigation of the mechanism of action of activin A was hampered by our inability to detect the aggrecanases in cartilage cultures by Western blotting.

It was surprising that activin A accounted for virtually all of the active material that activated Smad2 in the culture medium from human OA and porcine cartilage. However, we did not acidify the medium in order to release TGFβ from its latency peptide, to which it may be bound. A study published in 1991 showed that calf articular cartilage produced TGFβ and sequestered it in the extracellular matrix (27), and several other studies have demonstrated the expression of TGFβ mRNA and activity by cultured chondrocytes (28–31). In light of the findings of the present study, the relative importance of TGFβ and activin A as potentially chondroprotective proteins endogenous to articular cartilage needs to be investigated.

Activin A was clearly induced by wounding the cartilage and could provide a homeostatic mechanism to limit damage to the extracellular matrix. Production of the cytokine was dependent partly on FGF-2 and partly on inflammatory signaling, since it was markedly reduced by the FGFR inhibitor, the IKK inhibitor, and the proteasome inhibitor. In the context of inflammatory signaling, FGF-2 therefore appears to enhance the induction of the gene. Since FGF-2 is a normal pericellular constituent and activates chondrocytes upon physiologic loading (24, 32), the question arises as to whether FGF-2 is a sufficient stimulus, and whether activin A is produced in the absence of inflammatory stress and exerts a chondroprotective action normally in the tissue.

Activin A mRNA was detectable in the cartilage ex vivo, and when we used FGF-2 to treat explants that had been adapted to culture, we saw an increase in mRNA. This suggests that FGF-2 is a sufficient stimulus for activin A expression. However, we were not able to detect a significant increase in secreted activin A from these explants. This was puzzling, but the explants were spontaneously secreting some activin A all the time, and the amount that is continuously removed by the cells is unknown. In previous studies using isolated chondrocytes (1) or fibroblasts (5), FGF-2 was shown to be a sufficient stimulus for activin A production.

The in vitro experiments undertaken in the present study should be interpreted with caution, but it is likely that activin A is produced at a low level normally, and not solely after wounding or injury. However, further experiments are needed to determine whether activin A functions in the tissue under normal, as well as stressful, conditions. It is difficult in these experiments to separate the effects of dissection from the effects of tissue culture. OA cartilage explants secreted more of the cytokine than did normal tissue; to what extent this reflects increased production in vivo, perhaps as an attempt to reduce aggrecanolysis, or represents a change in responsiveness of the tissue to the insult of dissection and culture is not known.

We have only examined the expression and response of one component of the activin/inhibin system to cartilage injury. Inhibin α-chains and follistatin may be produced, and their expression might also be altered by, injury. Interestingly, follistatin mRNA expression is increased in OA cartilage (33), and if this is reflected at the protein level, there should be a reduction in activin activity.

It is possible to draw some tentative conclusions regarding the intracellular signaling pathways controlling activin A production in cartilage. Activin A production was dependent on NF-κB, as evidenced by the reduced production caused by both the IKK and proteasome inhibitors. With regard to other inflammatory pathways, blocking p38 MAPK had no effect, and we were unable to block JNK specifically in cartilage with the available agents. FGF-2 strongly activates ERK in cartilage, and the inhibitor of activation of MEK-1, which blocks the activation of ERK, had an effect comparable with that of the inhibitor of FGFR itself. FGF-2 would be expected to activate PKC and PI 3-kinase, but these did not appear to be necessary for activin A induction. The Src family kinase inhibitors reduced activin A production; therefore, FGF-2 controls activin A expression via ERK and possibly Src.

It should be stressed that although wounding the cartilage rapidly activates inflammatory signaling and causes instantaneous release of FGF-2, the mechanism of these phenomena is unknown. FGF-2 alone does not activate inflammatory signaling (NF-κB and JNK), and we have not been able to identify any soluble material released from the injured tissue as a stimulus of the inflammatory activation.

The possibility that articular cartilage acts via endogenous anticatabolic mechanisms has implications for the pathogenesis and treatment of OA. The disease is generally thought to result from matrix destruction due to production of proteinases caused by a cryptic stimulus. However, if there are endogenous anticatabolic mechanisms, and activin A may represent one, then failure of these could lead to an imbalance between catabolism and anabolism, resulting in excessive aggrecan degradation. By the same logic, therapy could be directed at boosting endogenous protective mechanisms.


Dr. Watt 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 design. Alexander, Watt, Sawaji, Hermansson, Saklatvala.

Acquisition of data. Alexander, Watt, Sawaji.

Analysis and interpretation of data. Alexander, Watt, Sawaji, Hermansson, Saklatvala.

Manuscript preparation. Alexander, Watt, Saklatvala.

Statistical analysis. Alexander, Watt.


We thank Ms Judith Hynes for assistance and Bruce Caterson of Cardiff University for the aggrecan neoepitope antibodies. We are grateful to Steve Crane and Adrienne Flanagan at RNOH in Stanmore and our surgical colleagues there and at Charing Cross Hospital for providing human tissues.