1. Top of page
  2. Abstract
  6. Acknowledgements


To compare gene expression in normal and osteoarthritic (OA) human chondrocytes using microarray technology. Of the novel genes identified, we selected follistatin, a bone morphogenetic protein (BMP) antagonist, and investigated its expression/regulation as well as that of 3 other antagonists, gremlin, chordin, and noggin, in normal and OA chondrocytes and synovial fibroblasts.


Basal and induced gene expression were determined using real-time polymerase chain reaction. Gene regulation was monitored following treatment with inflammatory, antiinflammatory, growth, and developmental factors. Follistatin protein production was measured using a specific enzyme-linked immunosorbent assay, and localization of follistatin and gremlin in cartilage was determined by immunohistochemical analysis.


All BMP antagonists except noggin were expressed in chondrocytes and synovial fibroblasts. Follistatin and gremlin were significantly up-regulated in OA chondrocytes but not in OA synovial fibroblasts. Chordin was weakly expressed in normal and OA cells. Production of follistatin protein paralleled the gene expression pattern. Follistatin and gremlin were expressed preferentially by the chondrocytes at the superficial layers of cartilage. Tumor necrosis factor α and interferon-γ significantly stimulated follistatin expression but down-regulated expression of gremlin. Interleukin-1β (IL-1β) had no effect on follistatin but reduced gremlin expression. Conversely, BMP-2 and BMP-4 significantly stimulated expression of gremlin but down-regulated that of follistatin. IL-13, dexamethasone, transforming growth factor β1, basic fibroblast growth factor, platelet-derived growth factor type BB, and endothelial cell growth factor down-regulated the expression of both antagonists.


This study is the first to show the possible involvement of follistatin and gremlin in OA pathophysiology. The increased activin/BMP-binding activities of these antagonists could affect tissue remodeling. The data suggest that follistatin and gremlin might appear at different stages during the OA process, making them interesting targets for the treatment of this disease.

Osteoarthritis (OA) is a progressive, debilitating disease of the joints characterized by the erosion of articular cartilage. Although much is known about the expression and regulation of genes associated with OA (e.g., metalloprotease, tissue inhibitor of metalloproteases, inflammatory cytokines, and extracellular matrix proteins such as collagens and proteoglycans), it is likely that the expression of many other genes is also affected during this pathologic process, and more is yet to be learned in order to gain a comprehensive understanding of this disease.

Microarray technology was developed to analyze the expression of thousands of genes in a very short time. Complemented by techniques such as real-time reverse transcription–polymerase chain reaction (RT-PCR) or Northern blotting, it is a powerful tool for analyzing genetic expression on a large scale to obtain novel information about a given pathophysiology. This technology has already been applied to study a number of systems ranging from chondrogenesis to cancer and, more recently, to the field of arthritis (1–4).

Using microarray technology, we identified follistatin, a bone morphogenetic protein (BMP) and activin-binding protein, as being significantly up-regulated in OA chondrocytes. Activin and BMPs belong to the transforming growth factor β (TGFβ) superfamily of secreted signaling molecules (5). BMPs were first identified for their ability to induce bone formation. They have a wide range of biologic activities in embryogenesis and in the maintenance and repair of bone, cartilage, and other tissues in adults (5–10).

To date, very little is known about the role of the BMPs and their antagonists in OA pathophysiology. Nakase et al have shown that BMP-2/4 is present in chondrocytes in adult OA cartilage and osteophytes, as well as in neonatal growing articular cartilage, but is scarce in normal adult articular cartilage (11). It has been reported that the expression of BMP-2 is stimulated by interleukin-1β (IL-1β) and tumor necrosis factor α (TNFα) in normal and OA human chondrocytes (12), and that BMP-2 and BMP-6 are expressed in arthritic synovium and up-regulated by inflammatory cytokines (13). However, in another recent study, Bobacz et al (14) reported that BMP-6 is expressed in both normal and adult OA human cartilage but with no significant difference of expression. BMP-2 was found to stimulate in vivo proteoglycan synthesis in normal murine joints (15) but could not counteract the catabolic effect of IL-1β (16), something that BMP-7 was capable of doing in human cultured articular chondrocytes (17). Contradictory results were reported for BMP-7. Chubinskaya et al (18) observed that BMP-7 could be up-regulated in OA cartilage, while Bobacz et al (19) did not demonstrate a differential expression between normal and OA chondrocytes. In addition, it was recently shown that each BMP has a different pattern of distribution in joint articular tissue (20).

BMP antagonists comprise a family of structurally unrelated proteins. They regulate the activities and functions of the different BMPs by forming a complex with them and preventing their proper binding to the receptors (7, 8). Representatives of the BMP antagonist family include follistatin, gremlin, chordin, and noggin. Because each antagonist differs in its specificity and affinity for a specific BMP, each plays a different role, depending on cell and tissue type, in the spatial and temporal regulation of BMP activity.

Follistatin binds BMP-2, BMP-4, and BMP-7, with a higher affinity for BMP-7, although the affinity for BMPs is lower than that for activin (21, 22). Follistatin can inhibit BMP signaling in a manner different from that of the other antagonists, by binding to BMP receptors and BMPs, forming a trimeric complex (22). Gremlin binds BMP-2, BMP-4, and BMP-7 (23, 24). It is highly expressed in nondividing and terminally differentiated cells; it has been shown to regulate limb bud development, inhibit chondrogenesis and cell replication, and induce apoptosis in vitro (25). Chordin binds BMP-2, BMP-4, and BMP-7, with a higher affinity for BMP-2 and BMP-4; it does not bind activin or TGFβ1 (26). Overexpression of chordin in the embryonic chick limb system delays chondrocyte maturation and supports a role for chordin as a negative regulator of endochondral ossification (27). Noggin binds BMP-2, BMP-4, and BMP-7, with a higher affinity for BMP-2 and BMP-4 (28). It decreases osteoblast formation (29, 30), inhibits membranous ossification, and prevents chondrogenesis and limb development (31, 32).

This study is the first to evaluate the expression/production of 4 BMP antagonists in normal and OA human chondrocytes and synovial fibroblasts, as well as the regulation of 2 of them by different factors involved in the OA process. We show that the BMP antagonists are expressed and regulated differently in normal and OA chondrocytes. Such a change in the levels of BMP antagonists in OA tissues will likely impede the biologic action of the BMPs, suggesting a role in the progression of the disease.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Factors and chemicals.

Recombinant human IL-1β and IL-6 were purchased from Genzyme (Cambridge, MA); IL-4, IL-10, IL-13, epithelial growth factor (EGF), TNFα, interferon-γ (IFNγ), TGFβ1, basic fibroblast growth factor (bFGF), platelet-derived growth factor type BB (PDGF-BB), and activin A were from R&D Systems (Minneapolis, MN); prostaglandin E2 (PGE2) was from Cayman Chemical (Ann Arbor, MI); and dexamethasone and retinoic acid were from Sigma-Aldrich (Oakville, Ontario, Canada). Recombinant human BMP-2 and BMP-4 were a generous gift from the Genetics Institute (Cambridge, MA). Each factor was used at a concentration of 10 ng/ml, except for IL-1β (100 pg/ml), TNFα (5 ng/ml), IFNγ (10 units/ml), PGE2 (1 μg/ml), dexamethasone (10−7M), and retinoic acid (10−6M); these concentrations were chosen based on previous observations.

Specimen selection.

Normal human cartilage (femoral condyles and tibial plateaus) and synovial membranes were obtained from individuals within 12 hours of death (n = 15; mean ± SEM age 64 ± 6 years). These individuals had no history of joint disease and died of causes unrelated to arthritic diseases, including cardiorespiratory arrest, cerebral hemorrhage, pulmonary embolism, and pulmonary edema. The tissue was examined macroscopically and microscopically to ensure that only normal tissue was used. Human OA cartilage and synovial membranes were obtained from patients undergoing total knee arthroplasty (n = 19; mean ± SEM age 72 ± 8 years). All patients were evaluated by a certified rheumatologist (J-PP) and were diagnosed as having OA (33). These specimens represented moderate to severe OA as defined according to macroscopic criteria (34). At the time of surgery, the patients had symptomatic disease requiring medical treatment in the form of acetaminophen, nonsteroidal antiinflammatory drugs, or selective cyclooxygenase 2 inhibitors. None had received intraarticular steroid injections within 3 months prior to surgery. The institutional Ethics Committee Board of the Hôpital Notre-Dame approved the use of the human articular tissue.

Cell cultures.

Chondrocytes and synovial fibroblasts were released from full-thickness strips of articular cartilage and synovial membrane, respectively, by sequential enzymatic digestion at 37°C, as previously described (35). The chondrocytes were extracted from the whole OA cartilage. The cells were seeded at high density (105/cm2) in tissue culture flasks and cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco BRL, Burlington, Ontario, Canada) supplemented with 10% heat-inactivated fetal calf serum (FCS; Gibco BRL) and an antibiotic mixture (100 units/ml penicillin base and 100 μg/ml streptomycin base; Gibco BRL) at 37°C in a humidified atmosphere of 5% CO2/95% air. Cells were maintained in culture for 5–7 days. Primary chondrocyte cultures were used for the microarray assays, and firstpassage cultured chondrocytes were used for all other experiments. Synovial fibroblasts were used at the second or third passage.

The effect of cytokines and growth factors on gene transcription and protein production was assessed by preincubating confluent cells in DMEM/0.5% FCS for 24 hours followed by an incubation with fresh DMEM/0.5% FCS containing the factors being studied. The cells were incubated for 24 hours when processed for RT-PCR and for 48 hours when assayed for protein production.


Normal and OA chondrocytes were grown as described above. Three comparison pairs (1 normal and 1 OA specimen) were prepared for each assay. RNA extraction and concomitant complementary DNA (cDNA) synthesis and 32P labeling were performed with the Atlas Pure Total RNA Labeling System (Clontech, Palo Alto, CA). Briefly, total RNA was extracted with saturated phenol and treated with DNase I. Poly(A+) RNA was enriched by annealing with biotinylated oligo(dT) and binding to streptavidin magnetic beads. The purified RNA was reverse transcribed and labeled with α32P-dATP. The labeled probes were hybridized overnight to the Atlas Human 1.2 Array II using ExpressHyb (Clontech) hybridization solution, and the arrays were exposed to x-ray films for 2 days. The results were analyzed with AtlasImage 2.0 software (Clontech).

RNA extraction, RT, and real-time PCR.

Total cellular RNA from normal and OA chondrocytes and synovial fibroblasts was extracted with TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's specifications and treated with the DNA-free DNase Treatment and Removal Kit (Ambion, Austin, TX) to ensure complete removal of chromosomal DNA. The RNA was quantitated using the RiboGreen RNA Quantitation Kit (Molecular Probes, Eugene, OR). The RT reactions were primed with random hexamers, as described previously (36), with 3 μg of total RNA in a 100-μl final reaction volume.

Real-time quantitation of follistatin, gremlin, chordin, noggin, and GAPDH messenger RNA (mRNA) was performed in the GeneAmp 5700 Sequence Detection System (Applied Biosystems, Foster City, CA) with 2× QuantiTect SYBR Green PCR Master Mix (Qiagen, Valencia, CA), used according to the manufacturer's specifications.

Briefly, 45 ng of the cDNA obtained from the RT reactions was amplified in a total volume of 50 μl consisting of 1× Master Mix, uracil N-glycosylase (UNG; 0.5 units; Epicentre Technologies, Madison, WI), and the gene-specific primers, which were added at a final concentration of 200 nM. The primer sequences were as follows: for follistatin, 5′-AGAGCCTGCTTCCTCTGAG and 5′-AGCTGTAGTCCTGGTCTTC; for gremlin, 5′-ATGTGACGGAGCGCAAATAC and 5′-TGGATATGCAACGACACTGC; for chordin, 5′-CGCATCAGTGGACACATTG and 5′-TTCTGCAGCAGCATATGAGC; for noggin, 5′-CAAGAAGCAGCGCCTAAG and 5′-GTACTGGATGGGAATCCAG; and for GAPDH, 5′-CAGAACATCATCCCTGCCTCT and 5′-GCTTGACAAAGTGGTCGTTGAG.

The tubes were first incubated for 2 minutes at 50°C (UNG reaction), then at 95°C for 15 minutes (UNG inactivation and polymerase activation), followed by 40 cycles, each of which consisted of denaturation (94°C for 15 seconds), annealing (60°C for 30 seconds), extension (72°C for 30 seconds), and data acquisition (77°C for 15 seconds). The data were collected and processed with GeneAmp 5700 SDS software (Applied Biosystems) and given as a threshold cycle (Ct), corresponding to the PCR cycle at which an increase in reporter fluorescence above a baseline signal can first be detected. Plasmid DNAs containing the target gene sequences were used to generate the standard curves. When comparing the basal expression levels in normal and OA cartilage, the Ct was converted to the number of molecules, and the values for each sample were calculated as the ratio of the number of molecules of the target gene/number of molecules of GAPDH. Change in gene expression following stimulation by the various factors was calculated with the following formula: fold change = 2math image, where ΔCt = Ct stimulated − Ct GAPDH, and Δ(ΔCt) = ΔCt stimulated − ΔCt control. The primer efficiencies for the test genes were the same as that for the GAPDH gene.

Enzyme-linked immunosorbent assay (ELISA).

Follistatin levels in the culture medium were quantitated by a specific ELISA (R&D Systems). The values are calculated as picograms of follistatin released by 3 × 105 cells during a 48-hour incubation period.


Normal and OA cartilage was processed for immunohistochemical analysis as previously described (37). The specimens were fixed in 4% paraformaldehyde and embedded in paraffin. Sections (5μ) of paraffin-embedded specimens were deparaffinized in toluene, hydrated in a graded series of ethanol, and heated at 65°C for 20 minutes in 10 mM citrate buffer (pH 6.0). The specimens were then preincubated with chondroitinase ABC (0.25 units/ml in PBS [pH 8.0]) for 60 minutes at 37°C and with Triton X-100 (0.3%) for 30 minutes at room temperature for gremlin or with Triton X-100 (0.3%) for 30 minutes at room temperature for follistatin. Slides were then washed in PBS followed by 0.3% hydrogen peroxide/methanol for 30 minutes. They were further incubated for 60 minutes with 2% normal serum (Vector, Burlingame, CA) and overlaid with the primary antibodies for 18 hours at 4°C in a humidified chamber. The antibodies were a mouse monoclonal anti-human follistatin (R&D Systems) used at 50 μg/ml, and a goat polyclonal anti-human gremlin (IgG fraction; Santa Cruz Biotechnology, Santa Cruz, CA) used at 2 μg/ml.

Each slide was washed 3 times in PBS (pH 7.4) and stained using the avidin–biotin complex method (Vectastain ABC kit; Vector Laboratories, Ontario, Quebec, Canada). The color was developed with 3,3′-diaminobenzidine (Vector Laboratories) containing hydrogen peroxide. The slides were counterstained with eosin. Controls were used to determine the specificity of staining by substituting the primary antibodies with preimmune mouse and goat IgG (Chemicon, Temecula, CA), which was used at the same concentration as that of the primary antibody. Controls showed only background staining.

The total number of chondrocytes and the total number of positively stained chondrocytes were counted as previously described (37). Briefly, for each cartilage specimen, the total number of chondrocytes and the total number of positively stained chondrocytes were counted separately, at 40× magnification, from 3 fields of the superficial (superficial and upper intermediate layers) zone and 3 fields of the deep (lower intermediate and deep layers) zone. The total number of chondrocytes and the number of positively stained chondrocytes were quantitated separately for each zone of the cartilage. The final results were expressed as the percentage of positively stained chondrocytes. Each slide was subjected to an evaluation by 2 observers, with >95% agreement.

Statistical analysis.

Data are expressed as the mean ± SEM. Statistical significance was assessed by analysis of variance and with the 2-tailed Student's t-test, when applicable. P values less than 0.05 were considered significant.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Microarray analysis.

We used the Atlas Human 1.2 Array II to compare the expression of 1,196 genes in 3 pairs of normal and OA human chondrocytes. This array was chosen because it includes a variety of genes: cell surface antigens, basic transcription factors, transcription activators and repressors, cell–cell adhesion receptors, immune system proteins, oncogenes and tumor suppressors, calcium-binding proteins, transporter and carrier proteins, general metabolism and enzymes, ion channels, extracellular matrix proteins, trafficking and targeting proteins, growth factors, cytokines and chemokines, hormones and communication proteins, kinases and phosphatases, proteases and protease inhibitors, receptors, motility proteins, and housekeeping genes.

The genes that were up-regulated or down-regulated by ≥2-fold in at least 2 of the 3 comparison pairs are listed in Table 1. Of note, expression of the follistatin and calcitonin gene-related peptide 1 receptor genes was up-regulated in all 3 comparison pairs, while expression of the S100 calcium-binding protein A1 and receptor activity-modifying protein 2 genes was down-regulated in all 3 pairs. The cation-dependent mannose 6-phosphate receptor and dynactin 150-kd genes were up-regulated by at least 2-fold in 2 comparison pairs and by 1.6-fold in the third pair. The ATP-regulated potassium channel ROM-K and the guanine nucleotide exchange factor MSS4 genes were down-regulated 2-fold in 2 comparison pairs and were also down-regulated in the third pair but, because of weak signals, could not be quantitated accurately.

Table 1. Normal and osteoarthritic human chondrocyte differential gene expression as determined by the microarray technique*
Fold increase, rangeFold decrease, rangeGeneGenBank accession no.
  • *

    RNA from normal (n = 3) and osteoarthritic (n = 3) chondrocytes was extracted and processed for the Atlas Human 1.2 Array II assay, as described in Materials and Methods. The results were analyzed with AtlasImage 2.0 software. The genes listed were differentially regulated by 2-fold or more in at least 2 of the 3 comparison pairs. Genes that were differentially regulated by 2-fold or more in all 3 comparison pairs are shown in boldface.

2.0–5.0Calcitonin gene-related peptide 1 receptorL76380
2.3–2.8Cation-dependent mannose 6-phosphate receptorM16985
2.2–2.5Dynactin 150-kd isoformX98801
6.7–14.2Skeletal muscle LIM-protein 3L42176
4.3–5.0Na- and Cl-dependent taurine transporterZ18956
3.0–3.5Phospholipid transfer proteinL26232
6.5–9.3Lipopolysaccharide-binding proteinM35533
2.2–3.0Aldehyde dehydrogenase ALDH7U10868
2.5–3.5Cytochrome P450 1B1U03688
2.8–6.0Cartilage-derived morphogenetic protein 1U13660
3.3–5.2Myristoylated alanine-rich protein kinase C substrateM68956
3.5–6.2Proenkephalin AJ00123
3.5–10.0Connective tissue growth factor–like proteinAF074604
2.0–2.33-β-hydroxy-Δ5-steroid dehydrogenaseM27137
2.0–4.8S100 calcium-binding protein A1X58079
2.1–2.3Receptor activity-modifying protein 2AJ001015
2.0–2.1ATP-regulated potassium channel ROM-KU12541
2.4–2.7Guanine nucleotide exchange factor MSS4S78873
2.0–3.1Cartilage oligomeric matrix proteinL32137
2.0–2.3β-hexosaminidase α chainM13520
15.4–15.5Apolipoprotein DJ02611
3.2–3.8Membrane-associated phospholipase A2 group 2AM22430
2.0–5.3G protein–coupled receptor kinase 4L03718
3.7–6.0Endosome-associated proteinL40157
4.0–4.5Cellular retinoic acid–binding protein IS74445

BMP antagonist gene expression.

The microarray data revealed a significant up-regulation of the follistatin gene in OA chondrocytes. We pursued this study by determining the expression of 3 other members of the BMP antagonist family, gremlin, chordin, and noggin, which were chosen for their differential binding and affinity to BMPs. Basal expression levels of the antagonists were measured by real-time PCR in human chondrocytes and synovial fibroblasts, comparing normal with OA.

As shown in Table 2, the expression of follistatin and gremlin was significantly up-regulated in OA chondrocytes. Chordin was also expressed in chondrocytes, with similar values observed for normal and OA cells. Noggin, however, was not detectable in either normal or OA cells. BMP antagonists were also expressed in synovial fibroblasts but at a lower level than in chondrocytes. The expression of follistatin, gremlin, and chordin was down-regulated, but not significantly, in OA cells. Noggin, as for chondrocytes, was not detectable in synovial fibroblasts.

Table 2. BMP antagonist gene expression in normal and OA human chondrocytes and synovial fibroblasts*
GeneChondrocytesSynovial fibroblasts
Normal (n = 5)OA (n = 9)PNormal (n = 5)OA (n = 7)P
  • *

    Values are the mean ± SEM. Total RNA was extracted from the cells and processed for real-time polymerase chain reaction. The threshold cycles obtained were converted to the number of molecules, and the values for each sample were calculated as the ratio of the number of molecules of the target gene/number of molecules of GAPDH. Statistical significance was assessed by the Student's 2-tailed t-test. BMP = bone morphogenetic protein; OA = osteoarthritis; NS = not significant; ND = not detectable.

Follistatin2.36 ± 1.0710.84 ± 2.490.030.43 ± 0.090.22 ± 0.07NS
Gremlin6.55 ± 1.5640.25 ± 9.980.030.062 ± 0.020.039 ± 0.01NS
Chordin0.096 ± 0.0210.093 ± 0.01NS0.55 ± 0.17 × 10−30.45 ± 0.09 × 10−3NS

Follistatin production.

We compared the protein production of follistatin, using a specific ELISA, in normal and OA human chondrocytes and synovial fibroblasts. As shown in Table 3, both chondrocytes and synovial fibroblasts produced follistatin. As observed with the gene expression results, OA chondrocytes produced significantly more follistatin than did the normal cells (6.4-fold; P < 0.02). OA synovial fibroblasts produced less follistatin compared with the normal cells, but the difference did not reach statistical significance. In contrast with the gene expression levels, normal synovial fibroblasts synthesized more follistatin than did chondrocytes. This could be related to the fact that the half-life and/or stability of the mRNA could differ for the 2 cell types.

Table 3. Follistatin production in human normal and OA chondrocytes and synovial fibroblasts*
 Follistatin productionP
  • *

    Values are the mean ± SEM picograms of follistatin released by 3 × 105 cells during a 48-hour incubation period. The statistical significance was assessed by Student's 2-tailed t-test. OA = osteoarthritis; NS = not significant.

 Normal (n = 5)627 ± 3970.02
 OA (n = 10)4,012 ± 880
Synovial fibroblasts  
 Normal (n = 6)8,576 ± 3,370NS
 OA (n = 7)5,926 ± 1,267

Cartilage immunolocalization of follistatin and gremlin.

To determine the topographic distribution of follistatin and gremlin in cartilage, normal (n = 3) and OA (n = 5) human cartilage specimens were processed for immunohistochemical analysis using specific antibodies. OA specimens included samples showing different degrees of cartilage damage. Positive chondrocyte staining for follistatin in normal cartilage was demonstrated preferentially at the superficial zone (16.8 ± 2.7%), with only a punctuated staining pattern observed at the deep zone (2.1 ± 0.9%) (Figure 1). In the OA specimens, higher levels of positively stained cells were found, preferentially at the superficial zone as compared with the deep zone (29.7 ± 5.4% and 5.8 ± 2.3%, respectively). For both normal and OA cartilage a statistically significant difference (P < 0.007) was found when superficial and deep zones were compared. The follistatin localization pattern described above was similar regardless of the level of cartilage damage.

thumbnail image

Figure 1. Follistatin and gremlin immunostaining in representative sections of human and osteoarthritic (OA) cartilage. Follistatin was observed in both normal and OA specimens, preferentially at the superficial zone. Gremlin was not produced in normal cartilage but was observed in OA specimens, preferentially at the superficial zone. Unlike follistatin, gremlin generally was not found at the very upper superficial layers in OA cartilage. Arrows indicate positively stained chondrocytes. (Original magnification × 60.)

Download figure to PowerPoint

Gremlin was not detected in normal cartilage. OA chondrocytes produced this BMP antagonist and, as seen with follistatin, did so preferentially at the superficial zone (25.2 ± 2.5%) rather than at the deep zone (6.9 ± 1.6%) (Figure 1). A P value of <0.0003 was recorded when positively stained cells were compared in the superficial and deep zones of OA cartilage. Interestingly, and in contrast with follistatin, gremlin was not generally found at the very superficial layer of the cartilage, but rather below the first few chondrocyte layers (upper intermediate layers).

Follistatin and gremlin regulation.

In order to gain more insight into the regulation of the antagonists and to identify factors contributing to their elevated levels in OA chondrocytes, cells were treated with different factors that might affect cartilage physiology/pathophysiology. These included inflammatory (IL-1β, TNFα, IFNγ, IL-6, PGE2), antiinflammatory (IL-4, IL-10, IL-13, dexamethasone), and growth/developmental factors (TGFβ1, BMP-2, BMP-4, activin A, bFGF, PDFG-BB, EGF, retinoic acid). Cells (n = 5) were treated for 24 hours with these factors, and the expression levels of follistatin and gremlin were determined by real-time PCR.

As illustrated in Figure 2, the expression of follistatin and gremlin was differentially affected by some factors. The inflammatory factors TNFα and IFNγ up-regulated follistatin but down-regulated gremlin, and the antiinflammatory factor IL-4 down-regulated gremlin. In contrast, growth factors BMP-2 and BMP-4 significantly up-regulated gremlin but down-regulated follistatin. IL-1β and retinoic acid did not affect follistatin expression but down-regulated gremlin, while activin A down-regulated follistatin but did not affect gremlin. PGE2 and retinoic acid up-regulated follistatin, and the latter down-regulated gremlin. The other factors tested either did not affect (IL-6, IL-10) or down-regulated (IL-13, dexamethasone, TGFβ1, bFGF, PDGF-BB, EGF) expression of both follistatin and gremlin.

thumbnail image

Figure 2. Follistatin and gremlin regulation by different factors. Osteoarthritic (OA) chondrocytes (n = 5) were incubated in the absence (control; CTL) or presence of various factors. Total RNA was extracted and processed for real-time polymerase chain reaction, as described in Materials and Methods. The fold change in gene expression following stimulation by the various factors was calculated as follows: fold change = 2math image, where ΔCt = Ct stimulated − Ct GAPDH, and Δ(ΔCt) = ΔCt stimulated − ΔCt control. The level expressed by the nonstimulated (control) cells was given an arbitrary value of 1. Values are the mean and SEM fold change. Statistical significance was assessed by analysis of variance. IL-1β = interleukin-1β; TNFα = tumor necrosis factor α; IFNγ = interferon-γ; PGE2 = prostaglandin E2; DEX = dexamethasone; TGFβ1 = transforming growth factor β1; BMP-2 = bone morphogenetic protein 2; ACT.A = activin A; bFGF = basic fibroblast growth factor; PDGF-BB = platelet-derived growth factor type BB; EGF = endothelial growth factor; Ret.A. = retinoic acid; ∗ = P < 0.05 and ∗∗ = P < 0.001 versus control.

Download figure to PowerPoint

These data show that follistatin and gremlin are differentially regulated in OA chondrocytes. The strongest stimulation of follistatin occurred with IFNγ and TNFα, and that of gremlin occurred with BMP-2 and BMP-4. The other growth factors were generally effective at inhibiting both BMP antagonists, and inflammatory factors were more inhibitory to gremlin than to follistatin.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Microarrays are powerful screening tools that have allowed us to identify several novel genes with potential involvement in OA pathophysiology. Among the novel genes found to be modulated in OA cartilage, we selected the BMP antagonist follistatin, as well as 3 other BMP antagonists, for further characterization and regulation in normal and OA human articular tissue cells. This study is the first to show that the expression and synthesis of the BMP antagonists follistatin and gremlin are up-regulated in cultured OA chondrocytes and cartilage, suggesting their potential involvement in human OA.

The BMPs and their antagonists constitute a complex, dynamic system regulated at different levels in order to respond to specific environmental stimuli. In the adult joint, an equilibrium between the levels of BMPs and their antagonists exists to maintain the normal turnover of the tissues. This equilibrium is affected in response to inflammation or injury, when a higher level of BMP activity is required. The levels at which the BMP antagonists are expressed in OA cartilage may disturb activin/BMP activities, leading to decreased anabolic activities, affecting tissue repair and remodeling. Because BMPs are now targeted as therapeutic molecules to stimulate healing/reengineering of cartilage, BMP antagonists also should be taken into consideration when using this system.

Our results show differential expression of the antagonists not only within a given cell type (chondrocytes) but also between different cell types (chondrocytes, synovial fibroblasts). Differential expression of the antagonists is not unexpected because the role and expression/regulation of BMPs and their antagonists are different in cartilage and synovial membrane. The fact that normal and OA synovial fibroblasts do not differ significantly in their levels of follistatin and gremlin may reflect the complex interplay between stimulatory and inhibitory factors present in the synovial membrane.

In OA pathology some cytokines and other inflammatory factors (although not the primary cause of this disease) are definitely closely related to its progression. We tested several cytokines to verify whether 1 or more of these factors influenced the expression of the antagonists and contributed to the higher levels of follistatin and gremlin observed in OA cartilage. Moreover, as a number of tissue repair agents are present in these diseased tissues, we also tested the effect of some growth factors. Our results demonstrated a complex pattern of regulation, indicating that the increased expression of follistatin and gremlin in cartilage may not necessarily result from only 1 or the same stimulus. Indeed, we did not find a single factor that simultaneously triggered the increase of both. Follistatin was increased by IFNγ and TNFα and gremlin was not; gremlin expression was increased by BMP-2 and BMP-4, but follistatin was not. The overall expression of follistatin and gremlin is increased in OA cartilage, and, because of their different binding affinities, these BMP antagonists could likely play different roles and/or appear at different stages during the OA process, and could also be responsible for sequential autocrine feedback in the control of BMP/activin activities. Very few factors stimulate expression of follistatin and gremlin compared with the number of factors that are inhibitory. This supports the hypothesis that their expression is timed with specific stages in the progression of OA.

Although the factor or factors that initiate OA are still unknown, the very early stage of the disease is characterized by a hypertrophic biochemical repair reaction and an enhanced synthesis of extracellular matrix. At this stage, chondrocytes express genes involved in the repair of cartilage, such as BMP-2. The new matrix thus formed might not be quantitatively and qualitatively adequate to regenerate new functional cartilage and could result in a gradual loss of this tissue. The cartilage fragments released may trigger inflammatory reactions from the synovial membrane, or even within the cartilage, which will promote the expression of catabolic factors, including matrix metalloproteases, and contribute to the maintenance of OA.

Our data show that the growth factors BMP-2 and BMP-4 are favorite candidates for triggering the increase in gremlin expression. In OA, gremlin may first appear at the hypertrophic stage, when the increased level of BMP-2 stimulates its expression. IL-1β, which plays a pivotal role in cartilage degradation and is involved early in the disease, may also indirectly up-regulate gremlin by inducing BMP-2 in cartilage (12). Interestingly, Fukui et al (12) reported that, in moderately damaged OA cartilage, BMP-2 was expressed mostly in the middle and deep layers; in severely damaged OA cartilage, it was not seen at the fibrillated surface, a pattern that parallels the expression of gremlin-positive chondrocytes. Thus, the absence of gremlin near the surface of OA cartilage may result from the absence of BMP-2 as well as its inhibition by inflammatory factors. The overall gremlin expression levels could be influenced by both the presence of BMP-2 and the inhibitory effects of inflammatory factors. Thus, the IL-1β catabolic effect combined with the inhibition of the BMP anabolic activity would eventually result in severe cartilage degradation and an increased level of synovitis and inflammatory factor production. The presence of the inflammatory factor TNFα produced later by the diseased articular cells could also stimulate BMP-2 (12), which could in turn induce gremlin. Gremlin would then be present in both the early and late stages of OA.

Follistatin, in contrast to gremlin, may have a stronger link to the inflammatory aspect of OA and may appear later during the OA process, possibly induced by the presence of TNFα and IFNγ, 2 cytokines involved in inflammation and in severe OA. Follistatin binds activin A, a molecule expressed in inflammatory arthropathies (38) that is capable of inducing cell proliferation of rheumatoid arthritis (RA) synovial fibroblasts (39) and is released during inflammatory episodes at about the same time as the release of TNFα (40). Furthermore, Thornton et al (41) showed that a mouse follistatin-like gene was highly expressed along the margin of contact between the inflammatory synovial pannus and eroding bone in collagen-induced arthritis.

We showed that TNFα and IFNγ significantly increased follistatin levels, and that follistatin was mainly localized at the superficial zone of OA cartilage; this is the zone where TNFα had been previously located (42). Although the involvement of IFNγ in OA could be debated, it was reported that both OA and RA synovial lymphocytes have a similar Th1 profile, and that OA lymphocytes could produce IFNγ, although not as much as lymphocytes from RA (43). During the inflammatory episodes, follistatin would be preferentially increased over gremlin, the expression of which is decreased by most cytokines. Activin A and the other growth factors are not likely to trigger the increase in follistatin, because these molecules inhibit rather than stimulate its expression.

It has been suggested that BMPs could induce apoptosis (44–46). Because the major role of the BMP antagonists is to modulate BMP activity, changes in the levels of the antagonists may indirectly affect the level of apoptosis in the tissue. Gremlin, for example, was found to regulate programmed cell death in the developing avian limb (25). Whether or not follistatin and gremlin protect OA cartilage from apoptosis remains to be evaluated.

The finding that BMP antagonists are expressed in chondrocytes and are subjected to differential regulation opens the door to a new field of investigation, and experiments are under way to understand the role of each antagonist in OA pathophysiology. Normal chondrocytes might express low levels of BMPs and their antagonists, which are responsible for the normal turnover of the tissues. In OA, there is a reexpression of genes normally involved in embryogenesis and development in an attempt to repair the cartilage: the gene coding for BMPs, matrix metalloproteinase 13, wingless, and frizzled, for example, respond to the healing signal (20, 47–49) and are found at a higher level in human OA cartilage. Therefore, the follistatin and gremlin genes, normally expressed in embryogenesis, should be added to the list of genes reactivated during OA. The balance of BMP antagonist levels, as well as inflammatory or BMP factors, during the OA process may play a critical role in influencing the progression of the disease, making these antagonists interesting targets for the treatment of this pathologic condition. The patients with OA who were selected for this study had long-established OA disease, and future experiments with OA animal models will help in establishing expression levels of the BMP antagonists in the early stages of the disease.

In summary, new players in the field of OA pathophysiology have been identified. The study of the biologic functions of follistatin and gremlin is a challenging new area of research and may open new directions for the treatment of OA.


  1. Top of page
  2. Abstract
  6. Acknowledgements

We thank Martin Boily for his expert technical assistance in immunohistochemistry and the secretarial staff for manuscript preparation. We also thank the Genetics Institute (Cambridge, MA) for generously providing BMP-2 and BMP-4.


  1. Top of page
  2. Abstract
  6. Acknowledgements