Inhibition of Endogenous TGF-β During Experimental Osteoarthritis Prevents Osteophyte Formation and Impairs Cartilage Repair

Osteoarthritis (OA)² is a joint disease that is characterized by degradation of articular cartilage and the formation of new bone at the joint margins, so-called osteophytes. Articular cartilage proteoglycan (PTG) loss and fibrillation of the articular surface are early events in the OA process. At later stages clefts are formed in the cartilage, while end-stage OA shows erosion of the articular cartilage.

TGF-β is a multifunctional cytokine involved in crucial biological processes such as development, extracellular matrix synthesis, cell proliferation/differentiation, and tissue repair. So far, three mammalian isoforms have been described: TGF-β1, -β2, and -β3. TGF-β is produced in a latent form, associated with the latency-associated peptide. TGF-β can be activated after dissociation of latency-associated peptide from the mature protein (1). TGF-β signaling starts with binding of TGF-β to TGF-β-RII, a constitutively active serine/threonine kinase (2). After transphosphorylation of TGF-β-RI by TGF-β-RII the signal is further propagated involving Smad proteins (3, 4).

TGF-β has been suggested as an important factor in the pathogenesis of OA. One indication is the significant levels of active TGF-β in the synovial fluid of OA patients (5, 6). Indeed, OA changes of the cartilage have been described after exposure of knee joints to TGF-β, supporting a role for TGF-β in the pathogenesis of OA (7, 8). Local administration of TGF-β in the knee joint also induced inflammation and fibrosis (7, 8, 9). Another remarkable finding was the formation of osteophytes after multiple intra-articular (i.a.) injections of TGF-β protein or adenoviral overexpression of TGF-β1 in the knee joint (7, 8, 10). Osteophytes are newly formed bony structures located at the joint margins, and their occurrence is strongly associated with OA. Osteophytes originate from activated periosteum leading to new cartilaginous outgrowths that eventually ossify into osteophytes via the process of endochondral ossification. In developing osteophytes, mRNAs for TGF-β and TGF-β protein expression are strongly up-regulated (11, 12, 13). These data suggest that TGF-β induces osteophyte formation.

Although TGF-β seems implicated in pathology, TGF-β has also been suggested as a beneficial factor in cartilage repair. We have previously shown that injection of TGF-β in naive murine knee joints results in an increase in PTG synthesis and PTG content of articular cartilage (14). Moreover, i.a. injection of TGF-β during experimental arthritis resulted in protection from PTG loss (15). In addition, effects of IL-1, such as inhibition of cartilage PTG synthesis and release of cartilage PTG content, could be blocked by local application of TGF-β (16, 17). This demonstrates that TGF-β is able to counteract the deleterious effects of IL-1, a cytokine considered to be a key mediator during erosive joint diseases. Taken together, these experiments suggest a protective function for TGF-β in articular cartilage.

As can be concluded from the discussion above, TGF-β appears to have a dualistic role in OA: protection against cartilage damage but induction of osteophyte formation. Experiments conducted to investigate the role of TGF-β in joint diseases are mainly based on TGF-β supplementation. In this study, we aimed to determine the role of endogenous TGF-β during experimental OA. We selectively blocked endogenous TGF-β via systemic treatment with soluble TGF-β-RII (solRII), the cloned scavenging extracellular domain of the TGF-β-RII, an approach that to our knowledge has never been previously used in OA. solRII has a very high affinity for TGF-β1 and -β3, the two most abundant isoforms of TGF-β in the knee joint. Moreover, due to the small size of solRII it can penetrate the articular cartilage and affect chondrocytes. We show for the first time that inhibition of endogenous TGF-β during experimental OA dramatically decreases osteophyte size but enhances PTG loss. Our study implies a crucial role for endogenous TGF-β in osteophyte formation and cartilage maintenance.

The expression of TGF-β1, -β2, and -β3 was studied during papain-induced OA. The papain model is characterized by PTG depletion of articular cartilage, which is followed by attempted replenishment of the articular cartilage with PTGs at approximately day 10 (18). The right knee joints of mice were injected with 1 U papain solution; the left knee joints served as internal controls. Knee joints were dissected at days 7 and 14. Immunohistochemistry was performed on cryosections with specific Ab against TGF-β1, -β2, and -β3 (R&D Systems, Abingdon, U.K.). As a negative control, the primary Ab was replaced with chicken IgYs or goat IgGs. Biotin-labeled secondary Ab were used (Vector Laboratories, Burlingame, CA) followed by a biotin-streptavidin detection system (Vectra elite kit; Vector Laboratories). Bound complexes were visualized via reaction with 3′,3′-diaminobenzidine (Sigma-Aldrich, St. Louis, MO) and H₂O₂. Sections were briefly counterstained with hematoxylin and mounted with permount.

Because TGF-β1 and -β3 were detected by immunohistochemistry, TGF-β1 and -β3 levels in knee joints were quantified via ELISA. Therefore, 1 U papain was injected in the right knee joints of mice (n = 6); uninjected left knee joints served as internal controls. After 3 days, patellae were isolated and placed in RPMI medium/0.1% BSA. After 2 or 10 h of culturing, the medium was analyzed for the presence of TGF-β1 and -β3 via ELISA. In short, ELISA plates were coated with solRII (R&D Systems). TGF-β1 or -β3 standards (R&D Systems) or samples were applied and subsequently incubated with anti-TGF-β1 or -β3 Ab. The appropriate secondary biotinylated Ab were used, followed by incubation with streptavidin-polyperoxidase conjugate. Bound complexes were detected by reaction with orthophenylenediamine (Sigma-Aldrich) and H₂O₂. Absorbance was read at 492 nm using an ELISA plate reader (Multiscan MCC/340; Titertek, Huntsville, AL).

Construction of the yeast expression vector and screening of solRII-expressing clones are described elsewhere (19). In short, the complete extracellular domain of human TGF-β type II receptor was cloned in the pPic-9 expression plasmid (Invitrogen, San Diego, CA). During the cloning procedure the cleavage signal of the Saccharomyces cerevisiae α-factor secretion signal peptide and a sequence coding for six consecutive histidine residues (6xHis tag) were introduced via PCR.

For the production of solRII the Pichia pastoris expression system (Invitrogen) was used. P. pastoris was cultured in the Bioflow 3000 tabletop fermenter (New Brunswick Scientific, Edison, NJ). In essence, the fermentation process was conducted as described elsewhere (20). Yeast culture supernatant was filtered through a 0.2-μm filter (Schleicher & Schull, Dassel, Germany). solRII was purified from the supernatant by means of a 6xHis tag and a nickel-nitrilotriacetic acid column (Qiagen, Leusden, The Netherlands). Bound protein was eluted from the column with 300 mM imidazole in wash buffer. Imidazole was removed via a HiPrep 26/10 desalting column (Amersham Pharmacia Biotech, Little Chalfont, U.K.). The protein was subsequently further purified and concentrated using m.w. cut-off 100,000 and 10,000 filters (Millipore, Etten-Leur, The Netherlands) to a final concentration of ∼60 mg/ml. Although yeast cells do not contain LPS, the endotoxin level of purified protein was analyzed via the endosafe kinetic turbidimetric assay (Charles River Endosafe, Kent, U.K.). However, yeast supernatant contained an unknown interfering substance that confounded test results (data not shown). Therefore, an endotoxin inhibitor, polymyxin B (PMB; Sigma-Aldrich), was added before implantation as a precautionary measure.

SDS-PAGE analysis of recombinant solRII was performed on a 12% gel under denaturating conditions. Proteins were visualized via a standard silverstaining procedure.

For Western blotting analysis, solRII protein was blotted on a polyvinylidene difluoride membrane (Amersham Pharmacia Biotech). solRII protein was detected by an Ab against the extracellular part of the TGF-β-RII (R&D Systems) in combination with the appropriate secondary biotin-labeled Ab (Vector Laboratories). Bound complexes were detected by the ECL Plus detection reagents (Amersham Pharmacia Biotech).

The blocking capacity of the recombinant solRII was determined via a TGF-β competition ELISA using solRII as a capture agent. To assure the stability of the solRII for the duration of the implantation period, the used osmotic pumps were retrieved at the end of the experiment and flushed with 200 μl PBS. The diluted samples were also tested via ELISA. Samples of recombinant solRII were mixed with TGF-β1 in a molar ratio of 1000:1–8000:1. The TGF-β1 ELISA was performed as described in Quantification of TGF-β1 and TGF-β3 isoforms.

Male C57BL/6 mice aged 12 wk were used. Animals were kept in filter-top cages with a wood chip bedding. They were fed a standard diet and tap water ad libitum.

Alzet osmotic pumps (models 1007D and 2002; ALZA, Palo Alto, CA) were filled with the solution containing solRII (60 mg/ml)/PMB (2.63 mg/ml) or PMB alone and as a control empty pumps were used. One day after i.p. implantation of the osmotic pumps, 1 U papain solution was injected into the right knee joint. We aimed to induce a relatively mild OA to allow for diminished and increased PTG loss of the cartilage due to the treatment with solRII. Left knee joints served as internal controls for the systemic treatment with solRII/PMB and PMB. To administer solRII for 3 wk, first an osmotic pump model 2002 was implanted i.p., which after 14 days was replaced by a model 1007D pump. After 7, 14, or 21 days mice were sacrificed and whole knee joints were dissected and fixed for 4 days in 4% formalin. Subsequently, knee joints were decalcified in 5% formic acid and processed for paraffin embedding. Tissue sections (7 μm) were stained with Safranin O (Saf-O)/fast red.

[³⁵S]Sulfate incorporation was studied by i.p. injection of radiolabeled sulfate (75 μCi) 4 h before dissection of the knee joints. After histological processing, 7-μm tissue sections were dipped in K₅ emulsion (Ilford, Basildon, Essex, U.K.) and exposed for 3–5 wk. Then the slides were developed and stained with H&E. All findings are the result of two independent experiments.

The surface area of the osteophytes was measured using the Qwin image analysis system (Leica Imaging Systems, Cambridge, U.K.), using a JVC 3-CCD color video camera. The size of the osteophytes was determined by manual selection of the surface area of the osteophyte in five semiserial sections per knee, spaced ∼50 μm apart. Measurements of eight knees per group were averaged.

The PTG content of articular cartilage was also measured using an image analysis system as described elsewhere (21). The noncalcified layer of the patellar and femoral cartilage and cartilage of both the lateral and medial sides of the tibia-femur plateau was selected by hand. Loss of red staining (loss of PTGs) was calculated by measuring the amount of blue light passing through the tissue section. Of each knee joint, three sections per cartilage compartment were measured and the average was taken. Of each compartment the most severe depletion was considered as maximal PTG loss. The noncalcified cartilage of the control left knees were also measured and averaged and the PTG content was stated as being 100%. All measurements were expressed on a scale of 0–100% PTG content as measured by Saf-O staining.

We also noted a thinning of the articular cartilage, which resulted in a decrease in the thickness of the cartilage. This loss of noncalcified articular cartilage of the patella and femur was measured in three sections, spaced ∼50 μm apart by manual selection of the noncalcified cartilage. Then the surface area was determined and values were averaged.

Treatment with solRII during experimental OA increases PTG loss compared with untreated animals. We investigated whether solRII had an effect on the mRNA expression of matrix metalloproteinase (MMP)-3, -13, and -14, a disintegrin and metalloproteinase with thrombospondin motif (ADAMTS)-4, and -5, and tissue inhibitors of MMPs (TIMP)-1 and -3. Therefore, papain (1 U) was injected in the right knee joints of mice (n = 6 per group). Uninjected left knee joints served as internal controls. After 3 days, patellae were isolated and placed in 200 μl RPMI 1640 (Dutch modification) medium (Life Technologies, Breda, The Netherlands) for 24 h as described previously (22, 23). Patellae were treated with solRII (10 μg/ml) or IL-1 (10 ng/ml) as a positive control or were not treated. Next, RNA was isolated and treated with reverse transcriptase (Life Technologies). The RT-PCR protocol started with a 1-min denaturation at 92°C followed by annealing at 60°C for 1 min (58°C for TIMP-3) and a 1-min extension at 72°C. PCR products were subjected to electrophoresis in a 1% low-melting agarose gel (Seaplaque; SanverTech, Boechout, Belgium) containing ethidium bromide. For each primer set, the cycle number at which the PCR product was first detected on the agarose gel was identified. It was determined that in this phase the PCR still performs linearly. The PCR products were excised from the gel. PCR products were quantified using PicoGreen (Molecular Probes, Eugene, OR), a fluorescent dye that selectively stains dsDNA, according to the manufacturer’s recommendations. GAPDH levels were used to correct for the amount of template added to the PCR mix. For control and papain-injected knee joints the mRNA levels in the nontreated group were stated as 100%. Each PCR was performed in triplicate and results were averaged.

Results were analyzed via a Student t test. Results were considered significant at p < 0.05.

The expression of TGF-β isoforms was studied in papain-injected joints and compared with normal knee joints, using immunohistochemical methods on cryosections of whole knee joints.

In the periosteum of normal knee joints, hardly any expression of TGF-β1 and TGF-β3 was observed. After injection of papain the expression of both TGF-β isoforms was enhanced in the periosteum, where the expression of TGF-β3 seemed more intense than TGF-β1 expression (Fig. 1, A–D). The expression was present at locations known to develop osteophytes. TGF-β2 expression was not detected in normal periosteum or in periosteum after papain injection. Thus, TGF-β1 and -β3 seem to be up-regulated after papain injection at locations in the periosteum linked to osteophyte formation.

Besides expression in the periosteum, chondrocytes in the articular cartilage expressed TGF-β1 and -β3 isoforms. The expression was mostly confined to the noncalcified layer of the cartilage. As in the periosteum, this expression was up-regulated after papain injection compared with noninjected left knee joints (Fig. 1, E–H). In addition, in the cartilage the staining for TGF-β3 seemed more intense than TGF-β1 staining. Again, no staining for TGF-β2 was observed. Negative controls, such as substitution of the primary Ab with chicken IgYs or goat IgGs, did not result in a color reaction (Fig. 1, I and J).

The results show that TGF-β1 and TGF-β3 isoforms are expressed in periosteum and articular cartilage and are up-regulated after papain injection. Next, we tried to quantify TGF-β1 and -β3 levels in knee joints and we studied the effect of papain injection.

The effect of papain injection on TGF-β1 and TGF-β3 protein levels in knee joints was compared with noninjected knee joints via ELISA. TGF-β1 was detected in 2-h patellae wash-outs from control left knee joints (Table I). Part of this released TGF-β is probably the result of the patella isolation procedure itself. Nevertheless, injection of papain increased TGF-β1 protein expression significantly from 10 pg/ml in control patellae to 30 pg/ml (p < 0.05). TGF-β3 protein was also detected in wash-outs from control patellae (200 pg/ml, Table I). Injection of papain increased the expression of TGF-β3 2-fold in 2-h wash-outs (p < 0.001). Approximately 20 times more TGF-β3 than TGF-β1 was detected, which supports the observation from the immunohistochemistry experiment that murine knee joints seem to contain more TGF-β3 than TGF-β1. The up-regulation of TGF-β1 and -β3 protein was also found in 10-h wash-outs with similar results (data not shown). We proceeded to study the effect of TGF-β inhibition during experimental OA on these joint structures. We made use of solRII, the extracellular part of the TGF-β-RII, as a specific TGF-β inhibitor. solRII has a high affinity for TGF-β1 and -β3 and only low affinity for TGF-β2.

Before using solRII to inhibit TGF-β in vivo we characterized the recombinant solRII as produced by P. pastoris. Therefore, purified solRII was analyzed via SDS-PAGE and Western blot analysis. Recombinant solRII showed a pattern of protein bands ranging from ∼25 to 40 kDa (data not shown). The different bands are the result of heterogeneous glycosylation (19). The Ab used did not cross-react with BSA. Our recombinant solRII was estimated to be >95% pure.

We proceeded to determine the blocking activity of freshly prepared solRII with an ELISA. To fully neutralize TGF-β, a 4000-fold excess of solRII was required (data not shown). solRII in excess of 2000- and 1000-fold prevented detection of TGF-β in the order of 84 and 68%, respectively.

To apply the TGF-β inhibitor in vivo, we first checked the stability of the solRII in osmotic pumps, implanted i.p. in mice. Therefore, after 7 and 14 days the remainder of the protein was flushed from the retrieved osmotic pumps. Comparable concentrations of freshly and retrieved solRII were equally effective in blocking TGF-β, indicating no loss of blocking activity of the retrieved solRII protein (data not shown). This implies that, during the complete period, active solRII was administered.

To determine the role of endogenous TGF-β in osteophyte formation we treated mice with solRII during experimental OA. Papain injection in the knee joint was used to cause the formation of osteophytes at specific locations. Osteophytes developed on the femur close to the medial collateral ligament and on the tibia between the articular cartilage and where the growth plate meets the joint space. Early osteophytes consisted of large round chondrocytes, which stained with Saf-O and had not yet undergone the process of endochondral ossification (Fig. 2, A and E). Autoradiographic analysis showed that these chondrocytes had incorporated radioactive sulfate in high amounts, suggesting a very active cell metabolism (Fig. 2, B and F). Treatment with recombinant solRII reduced the average size of the osteophytes located on the femur significantly (p < 0.02) from 9616 μm² in the solvent-treated animals to 2746 μm² in the solRII-treated mice (Fig. 3). On the tibia, the average size of the osteophytes was reduced from 9849 μm² in solvent-treated mice to 2653 μm² in solRII-treated animals (p = 0.007). Autoradiographic analysis showed a less-active cell metabolism in the solRII group than in the solvent-treated group. These results strongly implicate endogenous TGF-β as an important factor in osteophyte development.

The role of endogenous TGF-β on articular cartilage PTG content after papain injection was studied using recombinant solRII. Inhibition of TGF-β activity for 7 days with solRII decreased PTG staining in five of six cartilage compartments compared with treatment with solvent (Fig. 4, C–F, and Fig. 5). The effect of TGF-β inhibition was most striking in patella and femoral cartilage. In these cartilage compartments a significant reduction (p < 0.001) in relative PTG content of 43 and 38%, respectively, was observed (Fig. 4, C and D, and Fig. 5). Also, the lateral femur and lateral tibia, as well as the medial femur, showed a significant decrease in PTG content (p < 0.05). No significant effects after 7 days of solRII treatment were observed in medial tibia cartilage. However, after 14 days of treatment with solRII, the cartilage PTG content of the medial tibia was significantly less than in the solvent-treated animals (p < 0.05) (Fig. 4, G and H). The other cartilage compartments did not significantly differ between treatments at day 14 (data not shown). After 21 days of treatment with solRII, the overall PTG content of the articular cartilage of the tibia/femur plateau was significantly lower than in the control-treated mice (p < 0.001) (data not shown).

Treatment with solvent, compared with animals receiving an empty pump, had no significant effects on PTG content in any of the cartilage surfaces on days 7, 14, or 21 (data not shown). The control left knee joints of mice receiving solRII for 7, 14, or 21 days were indistinguishable from left knee joints of solvent-treated animals (Fig. 4, A and B), indicating that systemic treatment with solRII itself did not induce pathological effects in the knee joint.

Besides a decrease in PTG content, solRII treatment also had an effect on the thickness of uncalcified cartilage (Fig. 4, C and D). Measurements of patellar cartilage revealed a significant decrease in surface area between solvent- and solRII-treated animals, 22,610 and 16,351 μm², respectively, and p < 0.001 (Table II). The femoral uncalcified cartilage also showed a significant decrease (p < 0.05) after solRII treatment (22,463 vs 26,077 μm² in solvent-treated animals). These findings further imply a role for endogenous TGF-β in cartilage maintenance.

Taken together, inhibition of endogenous TGF-β activity resulted in a decreased PTG content and decreased cartilage thickness, indicating a pivotal protective role for endogenous TGF-β in cartilage.

Inhibition of endogenous TGF-β during experimental OA reduces articular cartilage PTG content. To gain a better understanding of how TGF-β inhibition leads to enhanced PTG loss, we investigated the effect of solRII on the expression of cartilage-degrading proteinases MMP-3, -13, and -14, and ADAMTS-4 and -5, or inhibitors of degradation TIMP-1 and -3.

Treatment with solRII had no significant effect on the MMP-3 expression in patellae of noninjected knee joints. However, after papain injection, solRII treatment resulted in an ∼4-fold up-regulation of MMP-3 mRNA (Table III). Although no effect of solRII was seen on MMP-13 expression in control knee joints, treatment with solRII after papain injection resulted in an ∼4-fold up-regulation of MMP-13 mRNA (Table III). We could not identify significant effects of solRII treatment on MMP-14, ADAMTS-4 and -5, or TIMP-1 and -3 mRNA expression in patellae from normal and papain-injected joints (data not shown).

We used stimulation with IL-1 as a validation of the method used to determine the up-regulation of MMP mRNA levels. As expected, treatment with IL-1 up-regulated MMP-3 and -13 mRNA expression significantly in both noninjected and papain-injected knee joints (data not shown). These results indicate the ability of isolated patella tissue to respond to stimuli as IL-1.

From these data we conclude that the observed enhanced PTG loss in articular cartilage after solRII treatment is probably the result of up-regulated MMP-3 and -13 expression.

Our experiments identified for the first time the role of endogenous TGF-β during experimental OA. Via administration of the solRII we inhibited endogenous TGF-β. This enabled us to study the impact of neutralization TGF-β on osteophyte formation and articular cartilage degradation during experimental OA. Blocking of endogenous TGF-β in the knee joint resulted in an almost complete inhibition of osteophyte formation. Furthermore, systemic delivery of solRII not only decreased articular cartilage PTG content but also resulted in cartilage loss, probably via the up-regulation of MMP expression.

We chose the extracellular part of TGF-β-RII as a TGF-β antagonist because it has a very high affinity for TGF-β1 and -β3 (24, 25), which are both abundantly present in the joint (12, 26, 27). Moreover, because of its small size of 25 kDa, solRII can penetrate cartilage and directly affect chondrocytes (28). Chondrocytes are difficult cells to target due to the dense network of collagen fibrils and PTGs that make up articular cartilage and render it virtually impenetrable for large molecules such as Abs.

The solRII we produced via the P. pastoris expression system is a biologically active protein. A 4000-fold excess of recombinant solRII to TGF-β was able to completely neutralize TGF-β activity. This is well in range with other studies in which a 1,600- to 20,000-fold excess of solRII was needed to fully inhibit TGF-β activity (25, 29). The solRII obtained from the osmotic pumps retrieved from the mice after implantation showed no loss of blocking activity compared with freshly produced solRII. This indicates that during the complete duration of the experiment we were able to inhibit TGF-β activity.

One of the main characteristics of OA is the formation of new cartilage and bone on the joint edges, so-called osteophytes. Osteophytes are the principal cause of pain in OA patients and thus an unwanted feature of the disease. Osteophytes are believed to originate from the periosteum. Periosteal cells have the potential to undergo chondrogenesis and osteogenesis in vivo and in vitro (30, 31, 32). In developing osteophytes, not only is TGF-β mRNA strongly expressed but also TGF-β-R expression can be found (12, 13). Chondrogenesis of periosteal cells can be enhanced by TGF-β in a dose-dependent manner (33, 34, 35). We previously showed that already 1 day after i.a. injection of TGF-β increased PTG synthesis was observed in periosteal cells at sites that later show massive osteophyte development (14). However, these studies did not prove a role of endogenous TGF-β during osteophyte formation. In this study we show expression of TGF-β1 and -β3 in periosteum after papain injection. Blocking of TGF-β1 and -β3 activity by use of the solRII resulted in a dramatic reduction in osteophyte formation. So, here we show for the first time that endogenous TGF-β is essential in osteophyte formation during experimental OA.

One of the first signs of OA is the loss of PTGs from the articular cartilage, which ultimately leads to cartilage loss. Several lines of evidence suggest an important role for TGF-β in cartilage maintenance. We and others (14, 15, 36) showed that TGF-β can increase PTG synthesis in normal and OA chondrocytes in vivo and in vitro. Furthermore, local administration of TGF-β can suppress IL-1-induced matrix degradation (17, 37), probably via down-regulating MMPs (38, 39) and/or increasing the level of TIMPs (38, 40). TGF-β also inhibits terminal differentiation of chondrocytes (hypertrophy), which has been shown to occur during OA. Serra et al. (41) have shown that blocking of TGF-β signaling in transgenic mice expressing a truncated, kinase-defective TGF-β-RII in skeletal tissue results in hypertrophic chondrocytes leading to OA changes in these mice. Recently it was shown that, in Smad-3-deficient mice, articular chondrocytes undergo abnormal terminal hypertrophic differentiation, which leads to OA (42). These studies demonstrate that loss of TGF-β responsiveness in the articular cartilage can lead to OA changes.

Because TGF-β has an important regulatory function in articular cartilage, we studied effects of solRII treatment on articular chondrocytes. First we determined the expression of TGF-β isoforms during papain-induced OA. The results show that TGF-β1 and -β3 protein expression occurs in chondrocytes of normal articular cartilage, which is up-regulated after i.a. papain injection. In contrast, we were unable to detect TGF-β2 in normal knee joints or during papain-induced OA. So, TGF-β1 and -β3 are likely to play a role in cartilage matrix during experimental OA. Next, we show that inhibition of endogenous TGF-β in articular cartilage leads to a decreased PTG content in nearly all articular cartilage compartments. The loss of PTGs could be due to impaired synthesis of PTGs, enhanced breakdown, or poor retention of PTGs in the cartilage. These results are in line with our earlier studies that show that exogenous TGF-β increases PTG synthesis and content (14, 15). Our present results illustrate that the solRII is able to penetrate the cartilage and affect chondrocyte metabolism, leading to reduced PTG content. In conclusion, endogenous TGF-β is a necessary agent to protect articular cartilage from PTG loss.

Besides effects on PTG content, TGF-β deprivation also appeared to increase cartilage loss. We found that the thickness of cartilage was decreased after blocking TGF-β. Increased cartilage loss might be related to a decreased resistance to mechanical loading. Because inhibition of endogenous TGF-β decreased PTG content, and because PTGs are important molecules responsible for the resistance of articular cartilage (43, 44, 45), loss of PTGs may lead to loss of cartilage. Therefore, endogenous TGF-β appears important in the prevention of cartilage loss.

We found that inhibition of endogenous TGF-β leads to enhanced cartilage PTG loss. This can be the result of either up-regulation of cartilage-degrading proteases or down-regulation of the natural inhibitors of these enzymes. To clarify the mechanistics of TGF-β inhibition on articular cartilage, the mRNA levels of several MMPs, ADAMTS, and TIMPs in the knee joint were determined after exposure to solRII. Our results suggest that inhibition of endogenous TGF-β via solRII enhances cartilage PTG loss via up-regulation of MMP-3 and MMP-13. We observed only a small stimulation on ADAMTS-4 and -5 mRNA expression after solRII treatment. The expression levels of TIMP-1 and -3 were unchanged after solRII and IL-1 treatment. Thus, endogenous TGF-β suppresses the expression of several MMPs, such as MMP-3 and -13.

In summary, inhibition of endogenous TGF-β leads to an increased loss of PTGs from the cartilage and enhances cartilage loss during experimental OA, most likely through the up-regulation of MMPs. This indicates a protective role for endogenous TGF-β on cartilage. Moreover, with the aid of the soluble form of the TGF-β-RII as a TGF-β antagonist, we identified endogenous TGF-β as the main contributor to osteophyte development during experimental OA.