Proteomic Analysis of Articular Cartilage Shows Increased Type II Collagen Synthesis in Osteoarthritis and Expression of Inhibin βA (Activin A), a Regulatory Molecule for Chondrocytes*

We show that proteomic analysis can be applied to study cartilage pathophysiology. Proteins secreted by articular cartilage were analyzed by two-dimensional SDS-PAGE and mass spectrometry. Cartilage explants were cultured in medium containing [35S]methionine/cysteine to radiolabel newly synthesized proteins. To resolve the cartilage proteins by two-dimensional electrophoresis, it was necessary to remove the proteoglycan aggrecan by precipitation with cetylpyridinium chloride. 50–100 radiolabeled protein spots were detected on two-dimensional gels of human cartilage cultures. Of 170 silver-stained proteins identified, 19 were radiolabeled, representing newly synthesized gene products. Most of these were known cartilage constituents. Several nonradiolabeled cartilage proteins were also detected. The secreted protein pattern of explants from 12 osteoarthritic joints (knee, hip, and shoulder) and 14 nonosteoarthritic adult joints were compared. The synthesis of type II collagen was strongly up-regulated in osteoarthritic cartilage. Normal adult cartilage synthesized little or no type II collagen in contrast to infant and juvenile cartilage. Potential regulatory molecules novel to cartilage were identified; pro-inhibin βA and processed inhibin βA (which dimerizes to activin A) were produced by all the osteoarthritic samples and half of the normals. Connective tissue growth factor and cytokine-like protein C17 (previously only identified as an mRNA) were also found. Activin induced the tissue inhibitor for metalloproteinases-1 in human chondrocytes. Its expression was induced in isolated chondrocytes by growth factors or interleukin-1. We conclude that type II collagen synthesis in articular cartilage is down-regulated at skeletal maturity and reactivated in osteoarthritis in attempted repair and that activin A may be an anabolic factor in cartilage.

Osteoarthritis (OA) 1 is a common joint disease characterized by degeneration of articular cartilage. Since cartilage has very limited capacity for repair, the loss is effectively irreversible. Prevalence studies show that most people over the age of 65 have some evidence of the disease (1, 2). Little is known about the molecular mechanism of cartilage destruction in OA, particularly the early events. It is thought that there is an imbalance between anabolism and catabolism of the extracellular matrix, there being an increase in catabolism. It has been suggested that this increased breakdown of matrix is due to the production of degradative enzymes such as the matrix metalloproteinases (MMPs) and members of the disintegrin and metalloproteinase (ADAM) family (3,4). The increase in proteinase expression may be due to inflammatory cytokines such as interleukin-1 (Il-1) and tumor necrosis factor (4,5). However, it is unclear whether these degradative processes are a primary event or a secondary reaction.
Articular cartilage consists mainly of extracellular matrix, the principal organic components of which are type II collagen fibers and aggregates of the large proteoglycan aggrecan. The only cells in cartilage, the chondrocytes, contribute less than 5% to the total volume (6) and are responsible for the synthesis and degradation of matrix components. Very little is known about the normal endogenous control mechanisms of matrix turnover in articular cartilage. To study the regulation of the synthesis of proteins in cartilage and to understand better the molecular basis of the osteoarthritic process, we have developed a method for the proteomic analysis of explanted tissue in which secreted proteins are separated using two-dimensional electrophoresis and identified by mass spectrometry (MS). We analyzed secreted proteins because the tissue is mostly composed of extracellular material and used metabolic radiolabeling to detect newly synthesized molecules. Limiting the study to secreted proteins means that only a few hundred need to be separated, which is feasible by medium format two-dimensional gel electrophoresis. Since cartilage cannot be directly studied in the joint, we used explanted human articular cartilage in culture.
Transcriptional analyses of normal and diseased cartilage have enabled the parallel analysis of a large complement of genes in the osteoarthritic process (7,8). However, mRNA levels do not necessarily correlate with protein expression and reveal nothing about processing or post-translational modification. A proteomic approach, in which proteins are identified and quantified directly, is therefore a valuable complement to such transcriptomic studies. Our proteomic analysis shows a marked increase of type II collagen synthesis in osteoarthritic cartilage and has revealed that articular cartilage makes two potentially regulatory molecules, activin A and connective tissue growth factor (CTGF).

EXPERIMENTAL PROCEDURES
Materials-[ 35 S]methionine/[ 35 S]cysteine and recombinant plateletderived growth factor (PDGF) were purchased from Amersham Biosciences. Recombinant activin A and epidermal growth factor (EGF) were from R&D Systems (Abingdon, UK), and recombinant basic fibroblast growth factor (FGF-2) was obtained from PeproTech (London, UK). Recombinant human IL-1␣ was prepared in-house. The activin A enzyme-linked immunosorbent assay kit came from R&D Systems. Pronase E was from BDH Laboratory Supplies. DMEM and fetal calf serum were obtained from BioWhittaker (Verviers, Belgium). All other reagents were the best available grade from Sigma.
Sources of Cartilage-Porcine articular cartilage was dissected from the metacarpophalangeal joints of 3-6-month-old pigs within 24 h of slaughter. Human articular cartilage was obtained from Charing Cross Hospital, Hammersmith, London and the Royal National Orthopaedic Hospital, Stanmore, London with approval of the appropriate local ethical committee. Informed consent was given in all cases. OA cartilage came from joint replacement operations, whereas control samples were from femoral heads removed following trauma or from individuals undergoing amputations and resections for reasons other than joint disease. Some normal cartilage specimens were from fresh postmortem autopsies performed at Algemeen Ziekenhuis Sint-Jan, Bruges, Belgium or at the Department of Rheumatology, University of Ghent, Belgium. The age range was 22-86 and 53-83 years for normal and osteoarthritic patients, respectively. Details of the OA and control samples studied are summarized in Table S1 (see supplementary material).
Preparation of Cartilage Explant Conditioned Medium-Cartilage explants were dissected into serum-free DMEM (1 ml/g of cartilage) supplemented with 25 mM HEPES, penicillin (100 units/ml), streptomycin (100 g/ml), and amphotericin (2 g/ml). The explants were washed once and left overnight in serum-free DMEM. The following day, the tissue was incubated in methionine/cysteine-free DMEM (30 min) and then in the same medium containing [ 35 S]methionine/cysteine (200 Ci/ml/g of cartilage) for 5 h.
Precipitation of Proteoglycans with Cetyl Pyridinium Chloride (CPC)-Glycosaminoglycan content was estimated as chondroitin sulfate using the dimethyl methylene blue assay (9). Whale and shark chondroitin sulfate (Sigma) was used as a standard. A 5% aqueous (w/v) cetylpyridinium chloride solution was prepared, and 3 mg of CPC/mg of glycosaminoglycan (GAG) were added to each sample (unless otherwise stated). After 30 min at room temperature, the proteoglycan/CPC-containing precipitate was centrifuged, and the supernatant was removed. Samples for two-dimensional electrophoresis were dialyzed (100-kDa cut off) against water at 4°C overnight and then lyophilized. The pellets were washed with 0.4 M sodium acetate in 90% ethanol and 90% ethanol to remove CPC and sodium acetate, respectively. The residual proteoglycan-rich pellet was mixed with 4 volumes of sample buffer, boiled for 5 min, and then run on a 12.5% SDS-PAGE gel.
Two-dimensional Gel Electrophoresis-The lyophilized residues were dissolved in 9.5 M urea, 1% (w/v) dithiothreitol, 2% CHAPS, and 0.5% carrier ampholyte buffer (Amersham Biosciences) supplemented with proteinase inhibitors and loaded into 13-cm-long linear pH 3-10 immobilized pH gradient dry strip (Amersham Biosciences) by in-gel rehydration. Samples (50 -100 g) were isoelectrically focused using a Multiphor II flatbed electrophoresis system (Amersham Biosciences) at 300 V for 1 min, then ramped to 3500 V for 1.5 h, and then kept at 3500 V for 3.5 h. Prior to the second dimension separation, disulfide bonds were reduced by incubating the immobilized pH gradient strips for 15 min with 65 mM dithiothreitol in equilibration buffer (2% SDS, 6 M urea, 30% v/v glycerol, and l50 mM Tris, pH 8.8). Free SH-groups were alkylated by treatment with 260 mM iodoacetamide in equilibration buffer for 15 min. Following equilibration, the strips were transferred to a 12.5% polyacrylamide gel (Laemmli (10) without stacking gel) and run at 8 mA. Gels were fixed and silver-stained using a mass spectrometrycompatible protocol (11). After staining, gels were soaked for 2 h in 3% glycerol and dried prior to visualization of metabolically radiolabeled proteins by autoradiography.
Quantification of Protein Expression-The autoradiograms were scanned using a Bio-Rad 710 imaging densitometer, and the autoradiographic patterns were analyzed using Phoretix 2D software (version 6.01; Nonlinear Dynamic Ltd., Newcastle, UK). The area and pixel intensity of each spot were measured, enabling calculation of individual spot volumes, which were expressed as a percentage of the integrated spot volume for the entire gel.
Data-dependent MS/MS acquisitions were performed on precursors with charge states of 2, 3, or 4 over a survey mass range of 540 -1200. Known trypsin autolysis products and keratin-derived precursor ions were automatically excluded. Proteins were identified by correlation of uninterpreted tandem mass spectra to entries in SwissProt/TREMBL, using ProteinLynx Global Server (14). One missed cleavage per peptide was allowed, and the fragment ion tolerance was set to 100 ppm. Carbamidomethylation of cysteine was assumed, but other potential modifications were not considered in the first pass search. All matching spectra were reviewed manually, and in cases in which the score reported by ProteinLynx global server was less than 100, additional searches were performed against the NCBI non-redundant data base using MASCOT, which utilizes a robust probalistic scoring algorithm (15).
Isolation of Chondrocytes-Chondrocytes were isolated from the cartilage by digestion with Pronase E (1 mg/ml/g of cartilage) for 30 min at 37°C followed by collagenase (1 mg/ml/g of cartilage) for 5 h at 37°C. The digest was strained and then centrifuged at 500 ϫ g for 8 min. Pellets were washed twice and then resuspended in DMEM containing 10% fetal calf serum supplemented with 25 mM HEPES, penicillin (100 units/ml), streptomycin (100 g/ml), and amphotericin (2 g/ml). Cells were counted and plated on 12-well plates (diameter of 22.6 mm) at a density of 2.5 million cells/well (100% confluent).
Encapsulation of Chondrocytes in Alginate-Chondrocytes from the femoral condyle of a patient with OA were isolated by incubating cartilage with Pronase E (1 mg/ml) for 30 min followed by collagenase (1 mg/ml) overnight. The digest was strained and centrifuged at 500 ϫ g for 5 min. The pellet was washed twice with DMEM supplemented with 10% (v/v) fetal calf serum. Cells were counted and resuspended in 1.2% (w/v) (Keltone® LV, ISP Alginates, Tadworth, UK) in 0.15 M NaCl at a density of 4 ϫ 10 6 cells/ml, which was passed dropwise through a 25-gauge needle into 102 mM CaCl 2 . After 10 min of polymerization, beads were washed twice in 0.15 M NaCl and finally in DMEM supplemented with 10% (v/v) fetal calf serum and HEPES. The cells were cultured for 5 weeks in the same medium in a humid atmosphere of 5% CO 2 in air at 37°C. The medium was replaced twice weekly.
Stimulation of Chondrocytes in Alginate-After 5 weeks of alginate culture, 30 beads (ϳ0.75 ϫ 10 6 cells in 30 beads) were reseeded into a 24-well plate and cultured in 500 l of DMEM supplemented with HEPES overnight. The following day, the beads were washed twice with the same medium and stimulated with either activin A (R&D Systems, Abingdon, UK) (100 ng/ml) or TGF␤ (PeproTech EC Ltd., London, UK) (10 ng/ml) for 48 h.
Western Blotting for TIMP-1-The harvested conditioned medium was precipitated with trichloroacetic acid and subjected to SDS-PAGE with 12% (w/v) acrylamide gel. The proteins separated in the gel were electrotransferred onto polyvinylidene difluoride membrane and reacted with sheep anti-human TIMP-1 (from Professor Hideaki Nagase, Kennedy Institute of Rheumatology) followed by horseradish peroxidase-conjugated rabbit anti-(sheep IgG) IgG (DAKO A/S, Glostrup, Denmark). Immunoreactive TIMP-1 was visualized with enhanced chemiluminescence (ECL, Amersham Biosciences).
Reverse Transcriptase-PCR-Chondrocytes were serum-depleted for 5 h and then stimulated for 24 h. RNA was isolated from the cells using RNeasy mini-columns (Qiagen Ltd., Crawley, UK) and reverse-transcribed into DNA using Superscript II (Invitrogen). PCR amplification was performed using PuRe Taq ready-to-go PCR beads (Amersham Biosciences). The primers used for inhibin ␤A subunit were 5Ј-CCTC-CCAAAGGATGTACCCAAC-3Ј (sense strand) and 5Ј-GTGATGATCTC-CGAGGTCTGCT-3Ј (antisense strand). The primers were derived from the human sequence of activin ␤A-chain (accession number NM_002192).

Two-dimensional Electrophoresis of Proteins Secreted by Articular Cartilage, the Need for Proteoglycan
Removal-In initial two-dimensional electrophoresis experiments, it was found that proteins secreted by cartilage explants did not focus in the first dimension. This was probably due to the presence of highly anionic proteoglycans, particularly aggrecan, interfering with isoelectric focusing. Aggrecan was therefore removed by precipitation with the cationic detergent CPC (16).
The GAG content of medium conditioned by culturing cartilage (as estimated with the dimethyl methylene blue assay) varied but was normally around 200 g/ml for porcine and 100 g/ml for human material. After the addition of CPC, precipitates were centrifuged, and both the supernatants and the proteoglycan-rich pellets were analyzed by one-dimensional electrophoresis (Fig. 1a). The addition of 1 mg of CPC/mg of GAG removed about 80% of the GAG from porcine cartilage conditioned medium (Fig. 1b), but some smaller proteins (molecular mass less than 30 kDa) were lost from the supernatant and were found in the precipitate (Fig. 1a). When 2-4 mg of CPC/mg of GAG were added, more than 95% of the GAG was precipitated, and little silver-stainable protein was present in the pellet, although cartilage link protein was identified by HPLC MS/MS (Fig. 1a, arrow). Precipitation was carried out at room temperature since nonspecific co-precipitation of several proteins was observed at 4°C (data not shown). For routine use, it was decided to add 3 mg of CPC/mg of GAG, and the medium was left for 30 min at room temperature. After centrifugation, the medium was dialyzed, and the proteins secreted by porcine cartilage were then well resolved by twodimensional electrophoresis (see supplementary material, Fig.  S1 and Table S2).
Two-dimensional Electrophoresis of Proteins Secreted from Human Cartilage-Human cartilage was dissected from fresh surgical samples, washed overnight with serum-free DMEM to remove extraneous proteins, and then metabolically radiolabeled with [ 35 S]methionine/cysteine for 5 h. CPC precipitation was carried out as described above. Altogether, 12 osteoarthritic samples and 17 controls with macroscopically normal cartilage were analyzed (Supplementary Table S1).
Several hundred protein spots were usually observable on the silver-stained gels, of which 170 were excised, digested in-gel, and identified by HPLC MS/MS (see supplementary material, Fig. S2 and Table S3). Many of these were not chondrocyte-derived but were plasma or other proteins originating from synovial fluid and blood cells (e.g. hemoglobin and carbonic anhydrase). The autoradiographic patterns representing newly synthesized chondrocyte proteins were simpler, with between 50 and 100 well focused protein spots (depending on the sample loading) being observable. A typical autoradiographic pattern from cultured osteoarthritic cartilage is shown FIG. 1. The effect of CPC concentration upon precipitation of proteins (a) and GAG (b) from cartilage explant medium. Porcine cartilage explants were cultured in DMEM (1 g/ml) for 5 h. The medium was removed, and its GAG content was estimated with the dimethyl methylene blue assay. 800-l samples were treated for 30 min at room temperature with 1-4 mg of CPC/mg of GAG. The precipitate was spun down, and the supernatant was removed. a, supernatants and pellets were analyzed by SDS-PAGE and silver-stained. Link protein detected in the pellet is indicated with an arrow. b, the GAG remaining in the supernatant after CPC precipitation was measured.
in Fig. 2. 19 radiolabeled proteins were identified by HPLC-MS/MS (Fig. 2). These are listed in Table I, together with some nonlabeled cartilage proteins and several radiolabeled proteins, which were absent from the two-dimensional pattern but were detected by one-dimensional electrophoresis. Many are present on the two-dimensional gel (Fig. 2) as multiple gel spots because of glycosylation or other modifications. The re-producibility of the two-dimensional patterns is shown by comparison of seven autoradiograms from normal and OA joints representing a wide age range (Fig. 3).
Lumican and clusterin were abundantly present in multiple isoelectric forms and hindered detection of other proteins migrating to the same region of the gel. For example, MMP-2 and MMP-3 were sometimes obscured by lumican. MMP-1 was not observed on the two-dimensional gels but may be hidden in the lumican cluster since it was identified by MS after one-dimensional electrophoresis (Table I). Relative expression levels of these MMPs in OA and control samples were therefore difficult to assess from the two-dimensional gel patterns.
Expression of YKL-40 (also known as cartilage glycoprotein 39 and chitinase-3-like protein 1) is reportedly increased in OA cartilage (see Refs. 21, 24, and 25). However, we found its production variable; it was detected in 12/14 adult controls and in 4/12 OA samples. In contrast, the related protein YKL-39 was a consistent feature of the two-dimensional patterns in both normal and OA cartilage.
Two interesting potentially regulatory molecules, inhibin ␤A and CTGF, were identified. These proteins have not previously been described in articular cartilage (see below). Fig. 2, two high molecular mass metabolically labeled spots (apparent molecular mass Ͼ116 kDa) were identified as type II collagen. The larger species was presumed to be the pro-␣-chain, and the smaller was presumed to be the processed ␣-chain. The type II collagen C-terminal propeptide was detected in 1-3 isoelectric forms of about 35 kDa (Figs. 2 and 3). Procollagen C-proteinase enhancer protein was also detected, but the C-proteinase itself was not found. The OA explants (Figs. 2 and 3a) were clearly synthesizing and processing type II collagen.

Synthesis of Collagen Type II Is Increased in Cartilage Explants from Young or OA Subjects-In the osteoarthritic sample shown in
Inspection of the two-dimensional patterns from 12 osteoarthritic and 17 normal samples listed in Supplementary Table  S1 showed that the collagen II ␣-chain and the C-propeptide were made by most of the OA samples but not by the normal adult controls (Figs. 2 and 3, a, e, and f). Because the pro-␣chain and ␣-chain spots often focused poorly, the normalized FIG. 2. Autoradiograph of two-dimensional electrophoresis of OA cartilage explant medium (OA11). 2 g of cartilage were removed and washed overnight in serum-free DMEM. The following day, the cartilage was pulsed with [ 35 S]methionine/cysteine (200 Ci/ml) for 5 h. The explant medium was removed and treated with 3 mg of CPC/mg of GAG for 30 min at room temperature. The supernatant was removed and dialyzed into water at 4°C and lyophilized. 60 g of proteins were used for isoelectric focusing on a pH 3-10 gradient. Focused proteins were further separated by SDS-PAGE on a 12.5% gel. Metabolically labeled protein spots were identified by mass spectrometry. Coll, collagen; ASPIC, acidic secreted protein in cartilage; COMP, cartilage oligomeric protein.  spot volume of the C-propeptide was used as a measure of type II collagen synthesis (Fig. 4). The C-propeptide was a prominent feature in 9 out of 12 OA samples (0.4 -6.3% of total spot volume). It was undetectable in one and only weakly present in two others (about 0.1% of total spot volume). In samples from control adult cartilage, the C-propeptide either was undetectable or comprised less than 0.3% of total spot volume. In contrast, cartilage from two very young subjects (7 weeks and 6 years) synthesized significant amounts of collagen type II ␣-chain and C-propeptide (Fig. 3, b and c). The latter was also just detectable in a further sample from a 13 year old (Fig. 3d). Newly synthesized type II collagen and its C-terminal propeptide were also prominent in the medium of cultured explants of porcine cartilage (see supplementary material, Fig.  S1 and Table S2). The porcine cartilage was from animals that were 3-6 months old at slaughter (i.e. skeletally immature). Taken together, these results suggest that in healthy cartilage type II collagen, synthesis declines with skeletal maturity but may be reactivated in OA.
Secretion of the ␤A-chain of Activin/Inhibin by OA Cartilage Explants-A diagonal line of spots with molecular masses around 45 kDa was identified as pro-inhibin ␤A by HPLC MS/MS (Figs. 2 and 3). A single spot at 14 kDa corresponded to mature inhibin ␤A. The increasing mass and acidity of the 45-kDa spot chain is consistent with complex or hybrid type N-glycosylation, and inspection of the sequence reveals a single Asn-X-Thr N-linked glycosylation sequence (Asn-165). By contrast, the fully processed form, which lacks this consensus glycosylation site, is observed as a single spot in a position consistent with its calculated mass and pI (12,976/7.1).
Inhibin ␤A-chains homodimerize by disulfide bonding to form activin A or heterodimerize with inhibin ␣-chains to form inhibin A. Since no inhibin ␣-chains or other types of ␤-chain were detected, presumably only activin A (a ␤A-␤A homodimer) is present.
Proinhibin ␤A was secreted by all the OA explants, and the mature protein was present in 10/12 samples. The production of proinhibin ␤A and inhibin ␤A in adult control samples was variable and only detected in half of the samples. Production was detectable in the samples from the three youngest subjects (7 weeks, 6 years, and 13 years) (Fig. 3, b, c, and d).
Activin A Is Induced by Growth Factors and Cytokines-Activin is produced by cultured fibroblasts and keratinocytes in response to stimulation by growth factors or the inflammatory cytokines IL-1 and tumor necrosis factor-␣ (17). Serum-starved monolayers of human chondrocytes were therefore treated for 24 h with IL-1, activin A, TGF␤1, FGF-2, PDGF, or EGF to identify potential mediators of pro-inhibin ␤A induction in cartilage. All of these agents, including activin A itself, induced activin A mRNA (Fig. 5a). The concentration of activin A protein in the culture medium was also increased by these stimuli; IL-1 and TGF␤ caused the highest increase, consistent with their effect on the mRNA (Fig. 5b). The sample containing activin A was not assayed for protein.
Activin A Induces TIMP-1 Protein in Chondrocytes-Since activin A is a member of the TGF␤ family, and TGF␤ is known to induce expression of TIMP-1 in chondrocytes (18), we stimulated alginate-encapsulated human articular chondrocytes with activin A or TGF␤ for 48 h. Immunoblotting showed that both stimuli increased levels of TIMP-1 protein in the culture medium (Fig. 6a), and in the case of activin A, induction was concentration-dependent (Fig. 6b). DISCUSSION This study is, to our knowledge, the first application of twodimensional electrophoresis and MS to analyze proteins made by articular cartilage. Removal of sulfated proteoglycans by treatment with CPC was essential for successful two-dimensional electrophoresis. Between 50 and 100 radiolabeled protein spots were visible, although many of these were below the silver stain detection limit. Inevitably, since our analytical strategy entailed a deliberate removal of proteoglycans, we did not detect aggrecan or link protein (which was, however, found in the proteoglycan-rich pellet). It is likely that small proteoglycans such as biglycan and decorin were also precipitated with the pellet since they were not observed on the two-dimensional gels. Given that aggrecan and link protein were precipitated by CPC and that fibronectin and MMP-1 were not resolved on two-dimensional electrophoresis, we accounted for a total of 27 cartilage proteins secreted by cartilage.
Two-dimensional electrophoresis has inherent limitations, particularly for the analysis of hydrophobic, high molecular mass, or extremely acidic and basic proteins. Generally, 2-3 g of cartilage are needed to obtain two-dimensional gels from which radiolabeled proteins can be identified by mass spectrometry. In the case of OA samples, it was usually necessary to use all available material so that diseased tissue may sometimes be mixed with a proportion of undamaged cartilage. Thus differences between the protein expression patterns in osteoarthritic and normal cartilage may be masked by the heterogeneity of the diseased material.
Gene expression in normal and osteoarthritic cartilage has been studied using microarrays and by sequencing cDNA libraries. These approaches depend on obtaining 10 -20 g of mRNA/sample and thus require comparable amounts of cartilage to the present study. Gene expression profiling cannot predict a priori the quantities of protein made, post-translational modifications, or protein-protein interactions.
We have previously shown that dissection of cartilage results in activation of the extracellularly regulated kinase (ERK) pathway due to release of basic FGF, which is sequestered in the extracellular matrix (19). Dissection and explantation also transiently activate the c-Jun N-terminal kinase (20). Although the explants were rested and washed overnight, the patterns of proteins synthesized by the cartilage may differ from in vivo. Cell activation on explantation could also obscure differences between osteoarthritic and normal tissue and could induce expression of MMPs, inhibin ␤A/activin A, and TIMP-1. Caution is thus needed when comparing the behavior of OA and normal samples. Basic FGF was not found on the twodimensional gels of the cartilage culture medium, but it is extremely basic, which would not favor its isoelectric focusing, and it is present in low concentration.
Our observation that type II collagen synthesis and processing were increased in osteoarthritic cartilage is consistent with microarray data (7), in situ hybridization studies (21), and early work using [ 3 H]proline incorporation (22). In contrast, Kumar et al. (23), who sequenced cDNA libraries derived from OA and normal cartilage, did not observe differences in type II collagen mRNA expression. However, since mRNA from several subjects was pooled, individual variation could have biased the results.
Although we consistently observed collagen type II C-terminal propeptide, procollagen C-proteinase (BMP-1) was absent from the gels. Procollagenase enhancer protein, which binds the type II collagen C-terminal propeptide and potentiates the action of procollagen C-proteinase (24), was observed at the protein level but has not been detected in published transcriptomic studies (7,23). Collagen type VI ␣-chain was readily detectable, but apart from a 30-kDa unlabeled fragment of type XI, other collagens were not found.
In contrast to control cartilage from adult donors, normal cartilage from young subjects (and skeletally immature pigs) secreted collagen type II C-propeptide and ␣-chain at similar levels to OA samples. Thus type II collagen synthesis may decline with skeletal maturity but becomes reactivated in OA.
Inhibin ␤A-chain (a member of the TGF␤ superfamily (25)) was secreted by most cultured OA samples, but its production by normal cartilage was variable. Its expression may be induced by the cell activation caused by explantation and the release of basic FGF (19,20). We found that it could be induced in human chondrocytes by IL-1 or several growth factors including basic FGF. It increased chondrocyte expression of TIMP-1. In view of its relationship to TGF␤, it is likely that activin A is anabolic for cartilage.
Inhibin and activin were originally discovered in ovarian follicular fluid. Activin stimulated but inhibin inhibited release of follicle-stimulating hormone from the pituitary cells. Activin was subsequently shown to be a powerful mesodermal inducer of embryonic ectoderm in Xenopus (26). Activin binds to het-erodimeric receptor complexes (ActRI and ActRII) similarly to TGF␤ receptors, the activation of which leads to phosphorylation of the transcription factors Smad 2 and Smad 3, which transduce the signal from the cytoplasm to the nucleus via interaction with Smad 4 (27,28). It is interesting that Smad 3 null mice displayed OA-like degenerative changes in their cartilage (29). Activin also binds to the naturally occurring glycoprotein inhibitor follistatin (30), which we did not find.
Activin is produced at sites of inflammation (31)(32)(33) and has been shown to play a role in wound healing in mice (34). It may also be involved in bone formation since inhibin ␤A-chain knockout mice have craniofacial abnormalities (35). It has not previously been described in articular cartilage but induces a modest enhancement of type II collagen gene expression and proteoglycan synthesis in chondrocytes (36). In situ hybridization and immunocytochemistry will be needed to determine whether activin is expressed by osteoarthritic lesions.
A further potential regulatory molecule, CTGF, was present as a low abundance 20-kDa spot in osteoarthritic samples. It was also produced by the young porcine cartilage. CTGF has previously been detected by screening cDNA libraries from OA and normal articular cartilage (23). Full-length CTGF is ϳ36 kDa, but cleaved forms between 10 and 20 kDa have been observed in serum-stimulated mouse fibroblast cultures and physiological fluids (37). It promotes proliferation and differentiation of chondrocytes in culture (38) and has been detected in mouse fracture callus (39). Since CTGF is induced by TGF␤ (40), it may conceivably act downstream of activin.
Cytokine-like protein C17 mRNA was originally reported in CD34 ϩ bone marrow stem cells (41), but the protein has not previously been detected. It is predicted to fold into four ␣-helices, a characteristic feature of hematopoietic cytokines and interleukins. The abundance of both cytokine-like C17 and CTGF was very low, and further study is required to see whether their expression differs in normal and OA cartilage.
In conclusion, we have shown that proteomic technologies can be applied to articular cartilage and can reveal potential disease-specific alterations in protein expression. Refinement of these techniques will enable definition of phenotypic changes during the progression of OA and could also be used to evaluate the phenotype of chondrocytes during differentiation from stem cells for use in tissue engineering.
Acknowledgment-We are grateful to Sandy Welson for technical assistance in the early stages of this study.
FIG. 6. Induction of TIMP-1 protein in human chondrocytes in alginate by stimulation with activin A and TGF␤ (a). Chondrocytes in alginate were either unstimulated (lane 1) or stimulated with activin A (100 ng/ml) (lane 2) or TGF␤ (10 ng/ml) (lane 3) for 48 h. The conditioned culture medium was subjected to SDS-PAGE on a 12% gel, blotted onto a polyvinylidene difluoride membrane, and probed with an antibody against TIMP-1. In b, chondrocytes in alginate were stimulated with increasing concentration of activin A and processed as above. Lane 1 is unstimulated, and lanes 2-4 is 10, 50, 100 ng/ml activin A, respectively.