ARTHRITIS & RHEUMATISMVol. 43, No. 9, September 2000, pp 2100–2109© 2000, American College of Rheumatology

MODULATION OF COLLAGENASE 3 IN HUMAN OSTEOARTHRITICCARTILAGE BY ACTIVATION OF EXTRACELLULAR

TRANSFORMING GROWTH FACTOR b

Role of Furin Convertase

FLORINA MOLDOVAN, JEAN-PIERRE PELLETIER, FRANÇOIS MINEAU, MARTINE DUPUIS,JEAN-MARIE CLOUTIER, and JOHANNE MARTEL-PELLETIER

Objective. Treatment of normal cartilage withtransforming growth factor b (TGFb) can increase thesynthesis of collagenase 3 by chondrocytes and mimicthe in situ distribution of this enzyme in osteoarthritic(OA) cartilage, which occurs predominantly in the deepzone. In this study, we examined the elements of theTGFb system that are potentially relevant to this effect.

Methods. TGFb1 and TGFb2 levels in culturedcartilage explants were determined by enzyme-linkedimmunosorbent assay (ELISA). OA cartilage explantswere treated with small latent TGFb1 complex in thepresence of various inhibitors, and collagenase 3 levelswere determined by ELISA. The inhibitors were againstserine proteases, plasmin, cathepsins, furin, and aneutralizing antibody against the mannose-6 phosphate/insulin-like growth factor 2 receptor (M6P/IGF-2R).Small latent TGFb1, TGFb receptor types I, II, and III(TGFbRI, RII, and RIII), M6P/IGF-2R, and furin wereimmunolocalized in cartilage.

Results. Our data showed that latent TGFb1 isthe major isoform that is synthesized; levels of 17.2 61.7 pg/mg and 1.1 6 0.3 pg/mg tissue wet weight(mean 6 SEM) were found for total TGFb1 andTGFb2, respectively, in OA cartilage. A general serineprotease inhibitor abrogated activation of both endoge-

nous and exogenous small latent TGFb1. Plasmin andfurin inhibitors and anti–M6P/IGF-2R reduced the lev-els of exogenous small latent TGFb1 complex–inducedcollagenase 3 by 33%, 95%, and 76%, respectively, butthe cathepsin inhibitor had no effect. Immunolocaliza-tion of the small latent TGFb1 complex as well as ofTGFbRI and RII revealed a statistically significantincrease in the chondrocyte score in only the deep zoneof OA cartilage. The M6P/IGF-2R level was significantlyhigher in OA cartilage in both the superficial and deepzones. Furin was found in normal cartilage exclusivelyin the superficial zone, whereas in OA cartilage, a levelsimilar to that in normal cartilage was found in thesuperficial zone, but a significantly higher cell score(mean 6 SEM 23.6 6 4.7%) was registered in the deepzone.

Conclusion. The mechanisms of TGFb activation/activity with regard to collagenase 3 modulation incartilage appear to be controlled by furin convertasewith or without M6P/IGF-2R. These factors and thesmall latent TGFb complex are increased in the deepzone of OA cartilage, corresponding to the preferentialsite of collagenase 3 production.

Osteoarthritis (OA) ultimately results in the de-generation of articular cartilage. The first macroscopicsigns of joint destruction in this disease consist ofcartilage swelling, fibrillation, and disruption of thecollagen framework. Damage to the collagen network isa critical event in the pathogenesis of OA becausecollagenase macromolecules play a key mechanical rolein cartilage tissue; when this framework fails, the tissuedegenerates irreversibly.

The mechanisms leading to the development of

Supported by the Canadian Arthritis Society.Florina Moldovan, MD, PhD, Jean-Pierre Pelletier, MD,

Francois Mineau, MSc, Martine Dupuis, MSc, Jean-Marie Cloutier,MD, Johanne Martel-Pelletier, PhD: Centre hospitalier de l’Universitede Montreal, Hopital Notre-Dame, Montreal, Quebec, Canada.

Address reprint requests to Johanne Martel-Pelletier, PhD,Unite de recherche en Arthrose, Centre hospitalier de l’Universite deMontreal, Hopital Notre-Dame, 1560 rue Sherbrooke Est, Montreal,Quebec H2L 4M1, Canada.

Submitted for publication February 9, 2000; accepted inrevised form April 21, 2000.

2100

cartilage erosion in OA are not completely understood.It seems clear, however, that proteolytic degradation ofthe extracellular matrix is of particular importance (1,2),and evidence points to the increased production ofmetalloproteases (MMPs) in this process (2,3). In theMMP family, the collagenases are the enzymes respon-sible for native collagen degradation. Recent progress inthe identification of the MMP involved in collagenmatrix degradation has shown that, in addition to colla-genase 1 (MMP-1), collagenase 3 (MMP-13) is also ofparticular importance in this process (4–10).

Collagenase 3 is detected in increased amounts inOA cartilage, yet preferentially cleaves type II collagenand is 5–10 times more active than collagenase 1 on thissubstrate (4,5). Another interesting finding is that thesecollagenases show a differential topographic distributionwithin OA cartilage. Collagenase 3 is found predomi-nantly in the lower intermediate and deep layers (deepzone) of the cartilage (6,9), in contrast with collagenase1, which is detected mostly in the superficial layers(9,11). On the whole, data from the literature suggestthat collagenase 3 is involved in the remodeling phase ofthe tissue. Of interest, investigators at our laboratoryhave shown that this selective increase in production ofcollagenase 3 could be mimicked by stimulation ofnormal cartilage with transforming growth factor b(TGFb) (6).

TGFb belongs to a superfamily of structurallyrelated regulatory proteins, which includes 3 mamma-lian TGFb isoforms (b1, b2, and b3) (12). The activeform of TGFb binds with high affinity to cell-surfacereceptors. Three distinct glycosylated receptors of TGFb(TGFbR), known as types I, II, and III (or betaglycan),which have molecular weights of ;53, 83–110, and250–310 kd, respectively, have been identified. All 3receptors bind all 3 ligands with high affinity, but typesRI and RII bind TGFb2 with less affinity than TGFb1and TGFb3. Studies suggest that TGFbRI and RII aredirectly responsible for signaling (13), while TGFbRIIIappears to be involved in presenting TGFb to the otherreceptors.

TGFb is synthesized as a 390–amino acid precur-sor (pre–pro-TGFb), with the active part of the mole-cule contained in the 112 amino acids at the C-terminus(12). The pre–pro-TGFb is cleaved, which releases thepro-TGFb form (small latent complex). This pro-TGFbis also cleaved at residue 278 by furin convertase, whichis dimerized at the N-terminal and termed the b-latent–associated protein (LAP-TGFb). This LAP-TGFb bindsnoncovalently to the active TGFb, forming anothersmall latent complex. The active form of TGFb is a

single disulfide-linked dimer of 2 identical chains of 25kd. Often covalently linked to the LAP-TGFb is anotherprotein called latent TGFb binding protein, or LTBP.This LTBP is produced by the cells, but is from a genedistinct from that of the pre–pro-TGFb. In many cellsystems, either the large latent complex (LTBP–LAP-TGFb) or the small latent complexes (pro-TGFb andLAP-TGFb) are secreted from cells, but neither com-plex possesses biologic activity. Active TGFb may alsobe released. (For a schematic diagram of TGFb synthe-sis and maturation, see refs. 14–16).

In vitro, TGFb can be activated by exposure toacid, urea, or heat (.80°C) in the presence of a proteo-lytic enzyme such as plasmin, or by binding proteins suchas thrombospondin under certain conditions (17–19).Although the in vivo mechanisms of extracellular TGFbactivation are not fully characterized, storage of latentforms and the mechanism for TGFb activation appear tobe specific for each cell and tissue type. These critical invivo events seem to be highly regulated and may resultfrom cooperative action requiring more than one step.In human cartilage, the in vivo activation of the matrixlatent TGFb still remains to be elucidated.

Since TGFb is an important regulator of cartilagedevelopment and metabolism and is involved in collage-nase 3 production in OA cartilage, we first investigatedthe level and localization of small latent TGFb andTGFb receptors in this tissue. Furthermore, we exploredthe possible mechanisms of in situ activation of smalllatent TGFb in the extracellular matrix and its role inthe up-regulation of collagenase 3 production. Weshowed that small latent TGFb1, as well as TGFbRI andTGFbRII, are significantly elevated in the lower inter-mediate and deep layers (deep zone) of OA cartilage,and that the mannose-6 phosphate/insulin-like growthfactor 2 receptor (M6P/IGF-2R) as well as furin conver-tase appear to be implicated in the activation process ofsmall latent TGFb in the matrix of human OA cartilage.

PATIENTS AND METHODS

Specimen selection. Cartilage specimens (tibial pla-teaus) were obtained from OA patients (mean 6 SEM age67 6 2 years; 6 male, 14 female) undergoing knee replacement.All patients were diagnosed as having OA based on theAmerican College of Rheumatology criteria (20). Normalcartilage specimens were obtained at autopsy within 12 hoursof death, from individuals with no history of joint disease(mean 6 SEM age 48 6 8 years; 6 male, 4 female).

For all samples, histologic examinations were per-formed and sections were stained with Safranin O and fastgreen, as previously described (21,22). For normal controls,only cartilage that exhibited normal macroscopic and normal

ACTIVATION OF TGFb IN OA CARTILAGE 2101

histologic morphology was used. The severity of OA cartilagelesions was evaluated using the Mankin histologic/histochemical scale (23). On histopathologic examination, al-terations to the cell surface, hypercellularity with cell cloning,and decreased Safranin O staining were given Mankin scoresof 3–8.

TGFb determination. To determine which of theTGFb isoforms were present in normal and OA cartilage, thelevels of total and active (before acidification) TGFb1 andTGFb2 were measured quantitatively using specific enzyme-linked immunosorbent assays (ELISA; R&D Systems, Minne-apolis, MN). Cartilage explants (;150 mg) were washed andincubated in the serum-free BGJ culture medium supple-mented with glutamine (Sigma-Aldrich, Oakville, Ontario,Canada) and antibiotics (100 units/ml penicillin, 100 mg/mlstreptomycin; Gibco BRL–Life Technologies, Burlington, On-tario, Canada) for 5 days at 37°C. Culture medium aliquotswere used either directly or after acidification, according to themanufacturer’s instructions. The minimum detectable dose ofTGFb1 or TGFb2 was ,7 pg/ml.

Immunohistochemistry. For immunohistochemistry,cartilage specimens were processed as previously described(6,21,24), using the avidin–biotin–peroxidase complex method(Vectastain ABC kit; Dako Diagnostics Canada, Mississauga,Ontario, Canada). Briefly, specimens were fixed in 4% para-formaldehyde and embedded in paraffin. Sections (5m) ofparaffin-embedded specimens were deparaffinized, dehy-drated, and then either preincubated with chondroitinase ABC(0.25 units/ml in phosphate buffered saline; Sigma-Aldrich) for60 minutes at 37°C (when using anti–small latent TGFb1,anti-TGFbRI, anti-TGFbRII, anti-TGFbRIII, and anti–M6P/IGF-2R antibodies) or followed by heating at 65°C for 20minutes in 10 mM sodium citrate buffer, pH 6.0 (when usingantifurin antibody), and overlaid with antibodies (final concen-trations 5 mg/ml for anti–small latent TGFb1, anti-TGFbRI,anti-TGFbRII, and anti-TGFbRIII, 10 mg/ml for anti–M6P/IGF-2R, and 1:500 dilution for antifurin) for 18 hours at 4°C ina humidified chamber.

The antibody used for small latent TGFb1 was amonoclonal antibody (R&D Systems). All TGFbR antibodiesused were from Santa Cruz Biotechnology (Santa Cruz, CA)and were rabbit polyclonal antibodies for TGFbRI and RIIand a goat polyclonal antibody for TGFbRIII. The M6P/IGF-2R antibody (25) was generously provided by Dr. P. Lobel(Robert Wood Johnson Medical School, Piscataway, NJ). Theantifurin antibody was donated by Chiron Corporation (Em-eryville, CA) and is a polyclonal rabbit antibody.

To determine staining specificity, 3 controls were used:1) adsorbed immune serum (1 hour at 37°C) with 10-fold molarexcess of the recombinant human latent TGFb1 (rHuTGFb1;R&D Systems), the blocking peptides for TGFbRI and RIII(Santa Cruz), full-length TGFbRII (Santa Cruz), and recom-binant human furin (Alexis Biochemicals, San Diego, CA); 2)omission of the primary antibody; and 3) substitution of theprimary antibody with IgG (Nordic Immunology, Tilburg, TheNetherlands) following the same experimental protocol.

The evaluation of chondrocytes staining positive wasperformed using our previously published methods (6,9,24).For each specimen, 6 microscopic fields were examined (mag-nification 3 40): 3 fields at the superficial and upper interme-diate layers (superficial zone), and 3 fields at the lowerintermediate and deep layers (deep zone). Results were ex-

pressed as the percentage of chondrocytes staining positive(cell score). For scoring, slides were counterstained withnuclear fast red stain (Digene Diagnostics, Silver Springs,MD), which gives a light pink–stained nucleus. For photo-graphs, however, the slides were not counterstained.

Effect of inhibitors on small latent TGFb1–inducedcollagenase 3. To determine a potential nonspecific effect ofthe protease inhibitors and anti–M6P/IGF-2R used in thisstudy, preliminary experiments were performed with OA chon-drocytes in culture. The study was further carried out oncartilage explants that were obtained by aseptically dissectingfull-thickness strips of cartilage across the articular cartilage.The cartilage (;150 mg) was cut into pieces, rinsed severaltimes in Dulbecco’s modified Eagle’s medium (DMEM; GibcoBRL–Life Technologies), and incubated for 72 hours at 37°Cin a humidified atmosphere of 5% CO2/95% air in a mediumcontaining DMEM, 0.5% heat-inactivated fetal calf serum(HyClone Laboratories, Logan, UT), and antibiotics (100units/ml penicillin, 100 mg/ml streptomycin) in the presence orabsence of 160 ng/ml of small latent rHuTGFb1 complex(R&D Systems) and in the presence of enzyme inhibitors E-64(10 mM; an inhibitor of cathepsins B and L; Roche Diagnostics,Laval, Quebec, Canada), a2-antiplasmin (4 mg/ml; an inhibitorof plasmin activity; Calbiochem-Novabiochem, San Diego,CA), decanoyl-Arg-Val-Lys-Arg-chloromethylketone (100mM; an irreversible inhibitor of furin activity; Alexis Biochemi-cals), Pefabloc SC (2 mM 4-[2-aminoethyl]-benzenesulfonylfluoride; a specific serine protease inhibitor; Roche Diagnos-tics), and a neutralizing antibody against M6P/IGF-2R (10mg/ml; generously provided by Dr. P. Lobel). At the end of theincubation period, the milieu was used for collagenase 3determination, which was carried out with a specific ELISA(Amersham Pharmacia Biotech, Baie d’Urfe, Quebec, Cana-

Figure 1. Representative Western immunoblot of small latent trans-forming growth factor b1 (TGF-b1) complex secretion by humanosteoarthritic (OA) chondrocytes. Cells were incubated in serum-freeculture medium for 5 days at 37°C. Culture medium was concentrated10-fold, and 20 mg protein was electrophoresed under nonreducingconditions on sodium dodecyl sulfate–10% polyacrylamide gels. rh 5recombinant human.

2102 MOLDOVAN ET AL

da). This ELISA enables the measurement of both pro- andactive collagenase 3, with a sensitivity of 32 pg/ml.

Western immunoblotting. Experiments were also con-ducted to discriminate which of the small latent complex(monomer or dimer) is released by chondrocytes. OA chon-drocytes were incubated in the serum-free BGJ culture me-dium supplemented with glutamine and antibiotics for 5 daysat 37°C. The culture medium was concentrated 10-fold byultrafiltrating the sample solution through an anisotropicmembrane (Centriplus; Millipore, Nepean, Ontario, Canada).Twenty micrograms of protein was electrophoresed undernonreducing conditions on sodium dodecyl sulfate–10% poly-acrylamide gels, and transferred electrophoretically onto anitrocellulose membrane (Hybond-C Extra; Amersham Phar-macia Biotech). The membranes were immersed overnight inSuperBlock blocking buffer (Pierce, Rockford, IL), and anti-bodies were diluted in the same buffer. Following an overnightincubation with 2 mg/ml of the primary antibody anti–latent

Figure 2. Detection of small latent transforming growth factor b1(TGF-b1) complex in representative sections of human normal andosteoarthritic cartilage. Immunostaining revealed that chondrocytes ofnormal cartilage (A) stained specifically for latent TGF-b1 almostexclusively at the upper superficial layers, whereas osteoarthriticspecimens (B) showed specific staining throughout the cartilage tis-sues. C, Osteoarthritic cartilage incubated with the adsorbed recom-binant human latent TGF-b1 showed only background staining. D,Histograms of the cell score for small latent TGF-b1 complex in thesuperficial and deep zones of normal and osteoarthritic cartilage. Barsshow the mean and SEM scores. P values indicate the differencebetween osteoarthritic and normal cartilage and between the superfi-cial and deep zones for normal cartilage. (Original magnification 3 63in A–C.)

Figure 3. Detection of small latent transforming growth factor b(TGF-b) receptor types I (RI), II (RII), and III (RIII) in representa-tive sections of human normal and osteoarthritic cartilage. Immuno-staining showed that RI and RII were detected preferentially in thesuperficial zone of normal cartilage (A and C, respectively), while inosteoarthritic cartilage, both receptors were detected throughout theentire thickness of the cartilage (B and D, respectively). E, Histogramsof the cell score for the 3 receptor types in normal and osteoarthriticspecimens from the superficial and deep zones. Bars show the meanand SEM scores. P values indicate the difference between osteoar-thritic and normal cartilage and between the superficial and deepzones for normal cartilage. (Original magnification 3 63 in A–D.)

ACTIVATION OF TGFb IN OA CARTILAGE 2103

TGFb1 (R&D Systems), the blots were washed several timesin TTBS (20 mM Tris, pH 7.4, 137 mM NaCl, and 0.1% Tween20), and then incubated for 1.5 hours at 22°C with the secondantibody (anti-mouse antibody–horseradish peroxidase conju-gates; 1:50,000 dilution) and washed again. Detection wascarried out using SuperSignal substrate (Pierce).

Statistical analysis. Data are expressed as the mean 6SEM. Statistical significance was assessed by Student’s 2-tailedt-test or Student’s paired t-test where appropriate. P values lessthan 0.05 were considered significant.

RESULTS

We first determined the TGFb isoforms pro-duced by human normal (n 5 4) and OA (n 5 7) explantcartilage. Our data showed that b1 was the majorisoform produced in human chondrocytes, and that thiswas found mostly in latent form. Total TGFb1 levels of17.2 6 1.7 and 5.1 6 0.2 pg/mg tissue wet weight(mean 6 SEM) were recorded in OA and normalcartilage explants, respectively. Total TGFb2 levels were1.1 6 0.3 pg/mg tissue wet weight in OA cartilage and atthe detection limit (.0.2 pg/mg tissue wet weight) innormal cartilage. Active TGFb1 was found in 4 of 7 OAcartilage samples, at a mean level of 0.5 6 0.1 pg/mgtissue wet weight, and in none of the normal samples.

The determination of latent TGFb complex re-leased by OA chondrocytes (n 5 4) was carried out usingWestern immunoblot under nonreducing conditions.The results (Figure 1) revealed 2 bands: a major bandwith an apparent molecular weight of ;50–60 kd, cor-responding to the monomeric form of the latent TGFbcomplex, and a second band with an apparent molecularweight of ;110 kd, corresponding to the dimeric form.

We further investigated the production and topo-graphic distribution of small latent TGFb1 in normal(n 5 5) and OA (n 5 6) human articular cartilage. Thiswas performed using immunohistochemical and mor-phologic analyses. As expected, the results revealed thatboth normal and OA cartilage samples produced smalllatent TGFb1. In normal cartilage (Figures 2A and D),a strong immunoreactivity toward latent TGFb1 wasfound mostly in the superficial layer and in the upperintermediate layers (superficial zone) of cartilage. How-ever, in the lower intermediate and deep layers (deepzone), only a punctate staining pattern was noted, withbarely detectable immunoreactivity. OA cartilage (Fig-ures 2B and D) showed increased immunoreactivitythroughout the entire thickness of the tissue, although inthe superficial zone, a higher level of immunoreactivechondrocytes (30.6 6 9.9%) was found compared withthat in normal cartilage (21.0 6 5.0%); the difference

did not reach statistical significance. However, in thedeep zone, latent TGFb demonstrated a significantincrease in cell score in OA cartilage (20.7 6 6.4%)when compared with that in normal cartilage (0.7 60.5%) (P , 0.02). In control studies in which immuneserum–adsorbed latent rHuTGFb1 (Figure 2C) or non-immune serum was used, only background staining wasobserved.

One of the critical steps in controlling the bio-logic activity of a factor is the binding of the ligand to itsspecific receptors. The level and distribution of TGFbreceptors are important elements when determiningTGFb activity. In normal (n 5 5) and OA (n 5 6)cartilage, we examined the levels and topographic dis-tribution of each of the TGFb receptor types. Theresults (Figure 3) showed that human articular cartilageproduced all 3 TGFb receptors in a similar topographiclocalization. However, when normal and OA specimenswere compared, important differences in the levels ofthese receptors were noted. In normal and OA cartilage,the number of immunoreactive chondrocytes forTGFbRI and for TGFbRII (Figures 3A–E) was higherthan that recorded for TGFbRIII (Figure 3E) in boththe superficial and deep cartilage zones. Moreover, foreach receptor type, the chondrocyte score in normalcartilage was significantly higher (4–7 times) in thesuperficial zone than in the deep zone. In OA cartilage,values similar to those found in normal cartilage were

Figure 4. The effect of protease inhibitors or an anti–mannose-6phosphate/insulin-like growth factor 2 receptor (anti-M6P R) oncollagenase 3 production stimulated by small latent transforminggrowth factor b1 (TGF-b1) complex in human osteoarthritic cartilageexplants. Cartilage explants (n 5 4–7) were incubated in the absence(control) or presence of 160 ng/ml of small latent TGF-b1 complex andin the presence of enzyme inhibitors and a neutralizing anti-M6P R. Atthe end of incubation, production of collagenase 3 was determined inthe culture medium by a specific enzyme-linked immunosorbent assay.Bars show the mean and SEM percentage of collagenase 3 over controllevels (which were assigned the value of 100%; indicated with a dashedline). P values indicate the difference compared with autologouscontrol.

2104 MOLDOVAN ET AL

observed in the superficial zone, but values were signif-icantly increased in the deep zone for TGFbRI (P ,0.04) and RII (P , 0.007) (Figure 3E). The level ofTGFbRIII was also higher in the deep zone of OAcartilage compared with that in normal cartilage, but thedifference did not reach statistical significance.

The in vivo activation of TGFb appears to de-pend on a variety of elements and events, including thecell type, differentiation, and microenvironment, toname just a few. This indicates that the exact physiologicprocess involved must be determined within a particular

context. Therefore, we assessed the effect of inhibitingdifferent pathways that have been implicated in theactivation of extracellular small latent TGFb1 withregard to collagenase 3 production. Preliminary experi-ments performed with unstimulated OA chondrocytes(n 5 5) showed that neither the inhibitors, except forPefabloc SC, nor anti–M6P/IGF-2R had an effect on thebasal collagenase 3 level. Pefabloc SC showed an abro-gation of collagenase 3 production (results not shown).Studies were also carried out on 2 OA cartilage explantstreated with furin inhibitor and a2-antiplasmin. Valueswere similar to basal levels (controls) (results notshown). Results with small latent TGFb1–treated carti-lage (n 5 4–7) (Figure 4) showed that thiol proteases donot appear to be implicated in this process, since noinhibition was found when a specific inhibitor of thecathepsin family, E-64, was tested. Treatment with a2-

Figure 5. Detection of mannose-6 phosphate/insulin-like growth fac-tor 2 receptor (M6P/IGF-IIR) in representative sections of humannormal and osteoarthritic cartilage. Immunostaining of normal speci-mens (A) showed chondrocytes that stained specifically for M6P/IGF-IIR preferentially in the superficial zone of cartilage; very few chon-drocytes stained positive in the deep zone. Osteoarthritic specimens(B) showed specific staining throughout the cartilage tissues. C,Histograms of the cell score for M6P/IGF-IIR in normal and osteoar-thritic cartilage from the superficial and deep zones. Bars show themean and SEM scores. P values indicate the difference betweenosteoarthritic and normal cartilage and between the superficial anddeep zones for normal cartilage. (Original magnification 3 63 in A andB.)

Figure 6. Detection of furin in representative sections of humannormal and osteoarthritic cartilage. Immunostaining of normal speci-mens (A) showed chondrocytes that stained specifically for furinalmost exclusively at the upper superficial layers of cartilage. Osteo-arthritic specimens (B) showed specific staining throughout the carti-lage tissues. In C, osteoarthritic cartilage incubated with IgG showedonly background staining. D, Histograms of the cell score for furin innormal and osteoarthritic cartilage from the superficial and deepzones. Bars show the mean and SEM scores. P values indicate thedifference between osteoarthritic and normal cartilage and betweenthe superficial and deep zones for normal cartilage. (Original magni-fication 3 63 in A–C.)

ACTIVATION OF TGFb IN OA CARTILAGE 2105

antiplasmin resulted in only partial inhibition (33%) oflatent TGFb–induced collagenase 3 synthesis. Interest-ingly, when M6P/IGF-2R was blocked, the level of latentTGFb–induced collagenase 3 was inhibited by 76%. Thespecific serine protease inhibitor, Pefabloc SC, blockedboth endogenous and exogenous latent TGFb activa-tion, since collagenase 3 production was abrogated.Moreover, inhibition of furin activity induced a 95%reduction in exogenous latent TGFb–induced collage-nase 3.

Since these prior data showed that M6P/IGF-2Rand furin appear to be important in exogenous latentTGFb activation with respect to collagenase 3 produc-tion, we further examined the topographic localizationand production of these 2 factors in cartilage. The resultsrevealed that both M6P/IGF-2R (Figure 5) and furin(Figure 6) were produced by normal chondrocytes, andtheir topographic distribution was very similar to that ofsmall latent TGFb1. In normal cartilage, these 2 pro-teins were detected almost exclusively in the superficialzone (Figures 5A and C and Figures 6A and C). AnM6P/IGF-2R cell score of 18.5 6 3.1% was recorded inthe superficial zone, compared with 2.0 6 1.3% in thedeep zone of normal cartilage (n 5 5) (P , 0.0007)(Figures 5A and C). In OA cartilage (n 5 6), M6P/IGF-2R cell scores in the superficial (52.8 6 8.6%) anddeep (52.2 6 4.6%) zones of cartilage were significantlyincreased compared with the scores in normal cartilage(P , 0.003 and P , 0.0001, respectively) (Figures 5Band C). Furin cell scores of 16.2 6 5.3% and 0.8 6 0.5%were found in normal cartilage in the superficial anddeep zones, respectively (Figures 6A and D). In OAcartilage, furin cell scores of 19.9 6 3.9% and 23.6 64.7% were found in the superficial and deep zones,respectively (Figures 6B and D). Compared with normalcartilage, a significant enhancement (P , 0.002) of furinwas found in only the deep zone of OA cartilage (Figure6D).

DISCUSSION

This study aimed at investigating which factorsfrom the TGFb system might be involved in the in situup-regulation of collagenase 3 production in OA carti-lage. The study brings new and interesting findings withregard to the topographic localization and production ofTGFb elements. It also provides new insight into themechanisms involved in the activation of matrical latentTGFb with respect to collagenase 3 production in OAcartilage, and the role of M6P/IGF-2R and furin in thisprocess. We demonstrated that TGFb is secreted mostly

in latent form in OA cartilage. Moreover, our data,which showed a preferential increase in the levels ofboth small latent TGFb and TGFb receptor types I andII in the lower intermediate and deep layers (deep zone)of OA cartilage, support the concept that an elevation ofTGFb activity in this zone of the pathologic tissue maybe responsible for the increase in collagenase 3 levels.Furthermore, our study revealed that the mechanisms ofTGFb activation appear to be controlled by furin con-vertase and M6P/IGF-2R, since these factors were alsosignificantly increased in the deep zone of OA cartilage,and that inhibiting their activity abrogates or markedlyreduces the production of collagenase 3 induced bysmall latent TGFb.

We showed that both TGFb1 and TGFb2 iso-forms are produced by chondrocytes, and that b1 is themajor isoform found in human cartilage. The TGFb2level is very low and represents only ;6% of the TGFb1isoform. Moreover, ;97% of the TGFb1 released fromhuman chondrocytes is in a latent form. These data areconsistent with the previous findings of Villiger and Lotz(26), in which both isoforms were found in normalcartilage, TGFb1 was predominant, and only a very lowlevel of the active form was detected.

In general, TGFb appears to be released fromcells via a constitutive secretory pathway in several formsof latent complexes (14–16). In this study, we showedthat OA chondrocytes released the 2 forms of smalllatent TGFb1 complex, the monomeric and dimericforms, but with a higher proportion in the monomericform. However, the exact role within the extracellularmatrix of each of these complexes remains to be eluci-dated.

Whereas TGFb has been shown to be expressedand synthesized by articular cartilage chondrocytes, itsexact localization and mechanisms for physiologic acti-vation in this tissue, and particularly during pathologicevents, are poorly understood. The findings of an in-creased cell score in OA cartilage for the small latentTGFb1 complex as well as for the 3 TGF receptor typesin the deep zone further support the hypothesis of apreponderance of TGFb system activity at this location.One of the important factors controlling TGFb biologicactivity is the specific cell receptors. To determine if thepresence and level of the TGFb receptors in OA carti-lage corresponded to preferential collagenase 3 produc-tion, we investigated their topographic distribution. Weshowed that OA cartilage demonstrated a higher num-ber of chondrocytes producing the 3 TGFb receptortypes than did normal tissue, which results from theirenhancement only in the deep zone of the cartilage (a

2106 MOLDOVAN ET AL

significant increase was reached for receptor types I andII). Our data on TGFbRII appear to be in contrast withthe findings of Boumediene et al (27), who showed adecrease in the expression of messenger RNA for thisreceptor in an experimental OA rabbit model. However,the latter study was performed using specimens from anOA animal model that presented only very mild lesions.Histopathologic analysis of the OA specimens used inthis study demonstrated moderate-to-severe lesions,which could possibly provide an explanation for thesedifferences. Moreover, it is suggested in the literaturethat it is the ratio of TGFb receptor types I and II thatis of the utmost importance in the modulation of theTGFb response. Likewise, the hypothesis that TGFbacts as a switch (28,29) could very well be applied to thereceptors, depending on the stage of the disease; theproduction of a receptor type could vary depending onthe cell environment of the tissue.

Although the elevated TGFb receptor levels inthe deep zone of OA cartilage could result in theincreased activity of TGFb and explain the collagenase 3up-regulation, this seems to be an insufficient explana-tion, since TGFb is secreted mostly in latent form. Inthat regard, the critical step appears to be the activationof the latent molecule, which is believed to be handledby a tightly regulated and localized mechanism. In thisstudy, we examined some of the possible physiologicfactors present in the cartilage that may be involved inlatent TGFb activation. It is proposed that extracellularlatent TGFb activation occurs at the cell surface in orderto be proximate to the TGFb receptors, thus facilitatingsignaling and averting adventitious events. This processwould involve factors bound to or in proximity to theTGFb membrane receptor. Many enzymes have beensuggested as factors, including cathepsin B and plasmin.

Although cathepsin B levels have been found tobe elevated in human OA cartilage (30), this enzyme didnot seem to play a role in latent TGFb–induced colla-genase 3 production, since no effect was recorded whenOA cartilage was incubated with a specific cathepsininhibitor. Moreover, although this study did not com-pletely rule out the effect of plasmin, it did not appear tobe the sole mechanism, since a2-antiplasmin blockedonly one-third of the latent TGFb–induced collagenase3 production. In support of this finding, previous studieshave shown that plasminogen knockout mice displaynone of the pathologic features of TGFb knockout mice(31,32). In addition, it was recently reported that theproduction of active TGFb by alveolar macrophages wassimilar in urokinase plasminogen activator–deficientmice (uPA2/2; this enzyme is responsible for the cleav-

age of plasminogen to plasmin) and in their normal(uPA1/1) counterparts (33). Moreover, using immuno-histochemistry analysis, investigators at our laboratorypreviously showed that uPA was found exclusively in thesuperficial layers of OA cartilage, in contrast with colla-genase 3, which was found preferentially in the deepzone (34). One could therefore suggest that in cartilage,plasmin-mediated latent TGFb activation would be lesscontrolled compared with the mechanism proposedherein (see below). Activation with plasmin would befaster and independent of protein synthesis, which mightbe required in the tissue response to damage but is notnecessarily required during the tissue-remodeling phase,such as occurs with collagenase 3 production.

Our data, however, strongly support the conceptof the involvement of a combined effect of M6P/IGF-2Rand furin convertase in matrical small latent TGFbcomplex activation. Indeed, incubating OA cartilagewith a neutralizing antibody against M6P/IGF-2R mark-edly reduced latent TGFb–induced collagenase 3 pro-duction. Likewise, incubating OA cartilage explants withan inhibitor against furin abrogated small latent TGFb1complex–induced collagenase 3 production. With regardto the specificity of the furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone, it has been reportedthat it can also inhibit another protease from this family,PACE-4 (35). Although our data cannot discriminatebetween the inhibition of these 2 enzymes, PACE-4expression is not found in all tissues, and its presence hasnever been demonstrated in articular cartilage. Ourimmunohistochemical data support the involvement offurin. Indeed, the results show a statistically significantenhancement of furin in only the deep zone of OAcartilage when compared with that in normal cartilage,corresponding to the preferential location of collagenase3 in this disease tissue. Furthermore, the reportedfindings of TGFb induction of furin production in othercell systems (36) reinforce our concept.

Furin is an enzyme belonging to the mammaliansubtilisin-like family of convertases, and appears to bethe prime candidate for the intracellular enzymaticcleavage of pro-TGFb (36,37). The mechanism appearsto occur by a cleavage of the precursor at specific pairsof basic amino acids: Arg275-His-Arg-Arg278 sequencebefore the Ala279 (36). Also present in the small latentcomplex is another putative site of furin cleavage, theArg243-Arg244. In humans, the enzyme furin is localizedmainly in the trans-Golgi network, but recent studieshave demonstrated that it also can translocate betweenthe cell surface and the trans-Golgi via endosomes(38,39). It is therefore possible that the extracellular

ACTIVATION OF TGFb IN OA CARTILAGE 2107

small latent complex of TGFb is cleaved by furin at thecell surface and/or that, as proposed by Jirtle et al (40),these complexes are endocytosed and cleaved intracell-ularly by furin in the trans-Golgi, thus releasing activeTGFb. This could occur by the binding of small latentTGFb complexes to M6P/IGF-2R. The role of M6P/IGF-2R in facilitating the latent TGFb activation hasbeen found for many cell types (41–44). This transmem-brane glycoprotein, in addition to targeting newly syn-thesized lysosomal enzymes, can bind extracellular li-gands, resulting in their subsequent endocytosis andtransport. This receptor’s role in facilitating the activa-tion of TGFb would be by the binding of 2 N-linkedglycosylation sites on the small latent TGFb complexthat contain M6P residues (45). The localization of furinat the cell surface and/or in the trans-Golgi mightexplain why antifurin activity abolished the small latentTGFb complex activation, and why neutralizing M6P/IGF-2R reduced activation by only 76%.

In summary, the study results are consistent withthe notion of a preferential increase in TGFb biologicactivity/activation in the deep zone of OA cartilage,which is responsible for the up-regulation of collagenase3 production in this pathologic tissue zone. The in-creased levels of latent TGFb in this zone, as well as ofTGFb receptors and furin convertase, are indicative ofthis process. Based on our data, a possible sequence ofevents that may account for the effect of TGFb activa-tion on collagenase 3 production in the deep zone of OAcartilage could be summarized as follows. Small latentTGFb complexes are synthesized and released from thechondrocytes, preferentially in the deep zone of thecartilage. These complexes bind to M6P/IGF-2R and areproteolytically processed by furin at the cell membrane,or are endocytosed and cleaved intracellularly by furin,thus releasing active TGFb.

ACKNOWLEDGMENTS

We thank Dr. P. Lobel for the M6P/IGF-2R neutral-izing antibody, Dr. C. Dubois (Sherbrooke University, Sher-brooke, Quebec, Canada) for her productive exchanges, F. C.Jolicoeur for his excellent technical assistance, and S. Mc-Carthy for her assistance in manuscript preparation.

REFERENCES

1. Pelletier JP, Martel-Pelletier J, Howell DS. Etiopathogenesis ofosteoarthritis. In: Koopman WJ, editor. Arthritis and allied con-ditions: a textbook of rheumatology. 13th ed. Baltimore: Williams& Wilkins; 1997. p. 1969–84.

2. Dean DD, Martel-Pelletier J, Pelletier JP, Howell DS, WoessnerJF Jr. Evidence for metalloproteinase and metalloproteinase

inhibitor imbalance in human osteoarthritic cartilage. J Clin Invest1989;84:678–85.

3. Martel-Pelletier J, McCollum R, Fujimoto N, Obata K, CloutierJM, Pelletier JP. Excess of metalloproteases over tissue inhibitorof metalloprotease may contribute to cartilage degradation inosteoarthritis and rheumatoid arthritis. Lab Invest 1994;70:807–15.

4. Reboul P, Pelletier JP, Tardif G, Cloutier JM, Martel-Pelletier J.The new collagenase, collagenase-3, is expressed and synthesizedby human chondrocytes but not by synoviocytes: a role in osteo-arthritis. J Clin Invest 1996;97:2011–9.

5. Mitchell PG, Magna HA, Reeves LM, Lopresti-Morrow LL,Yocum SA, Rosner PJ, et al. Cloning, expression, and type IIcollagenolytic activity of matrix metalloproteinase-13 from humanosteoarthritic cartilage. J Clin Invest 1996;97:761–8.

6. Moldovan F, Pelletier J-P, Hambor J, Cloutier J-M, Martel-Pelletier J. Collagenase-3 (matrix metalloprotease 13) is preferen-tially localized in the deep layer of human arthritic cartilage in situ:in vitro mimicking effect by transforming growth factor b. ArthritisRheum 1997;40:1653–61.

7. Shlopov BV, Lie W-R, Mainardi CL, Cole AA, Chubinskaya S,Hasty KA. Osteoarthritic lesions: involvement of three differentcollagenases. Arthritis Rheum 1997;40:2065–74.

8. Martel-Pelletier J, Pelletier JP. The recently discovered collage-nase-3: a key role in osteoarthritis. In: Hamanishi C, Tanaka H,editors. Advances in osteoarthritis. Tokyo: Springer-Verlag; 1998.p. 121–33.

9. Fernandes JC, Martel-Pelletier J, Lascau-Coman V, Moldovan F,Jovanovic D, Raynauld JP, et al. Collagenase-1 and collagenase-3synthesis in early experimental osteoarthritic canine cartilage: animmunohistochemical study. J Rheumatol 1998;8:1585–94.

10. Tardif G, Pelletier J-P, Dupuis M, Geng C, Cloutier J-M, Martel-Pelletier J. Collagenase 3 production by human osteoarthriticchondrocytes in response to growth factors and cytokines is afunction of the physiologic state of the cells. Arthritis Rheum1999;42:1147–58.

11. Nguyen Q, Mort JS, Roughley PJ. Preferential mRNA expressionof prostromelysin relative to procollagenase and in situ localiza-tion in human articular cartilage. J Clin Invest 1992;89:1189–97.

12. Massague J, Heino J, Laiho M. Mechanisms in TGF-beta action.Ciba Found Symp 1991;157:51–65.

13. Hall FL, Benya PD, Padilla SR, Carbonaro-Hall D, Williams R,Buckley S, et al. Transforming growth factor-beta type-II receptorsignalling: intrinsic/associated casein kinase activity, receptor in-teractions and functional effects of blocking antibodies. Biochem J1996;316:303–10.

14. Feige JJ, Quirin N, Souchelnitskiy S. TGFb, un peptide biologiquesous controle: formes latentes et mecanismes d’activation.Medecine/Sciences 1996;12:929–39.

15. Centrella M, Horowitz MC, Wozney JM, McCarthy TL. Trans-forming growth factor-b gene family members and bone. EndocrRev 1994;15:27–39.

16. Lyons RM, Moses HL. Transforming growth factors and theregulation of cell proliferation. Eur J Biochem 1990;187:467–73.

17. Taipale J, Miyazono K, Heldin CH, Keski-Oja J. Latent transform-ing growth factor-beta 1 associates to fibroblast extracellularmatrix via latent TGF-beta binding protein. J Cell Biol 1994;124:171–81.

18. Brown PD, Wakefield LM, Levinson AD, Sporn MB. Physico-chemical activation of recombinant latent transforming growthfactor-beta’s 1, 2, and 3. Growth Factors 1990;3:35–43.

19. Lyons RM, Keski-Oja J, Moses HL. Proteolytic activation of latenttransforming growth factor-beta from fibroblast-conditioned me-dium. J Cell Biol 1988;106:1659–65.

20. Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, etal. Development of criteria for the classification and reporting ofosteoarthritis: classification of osteoarthritis of the knee. ArthritisRheum 1986;29:1039–49.

2108 MOLDOVAN ET AL

21. Pelletier JP, Faure MP, Di Battista JA, Wilhelm S, Visco D,Martel-Pelletier J. Coordinate synthesis of stromelysin,interleukin-1, and oncogene proteins in experimental osteoarthri-tis: an immunohistochemical study. Am J Pathol 1993;142:95–105.

22. Martel-Pelletier J, Pelletier J-P, Cloutier J-M, Howell DS,Ghandur-Mnaymneh L, Woessner JF Jr. Neutral proteases capa-ble of proteoglycan digesting activity in osteoarthritic and normalhuman articular cartilage. Arthritis Rheum 1984;27:305–12.

23. Mankin HJ, Dorfman H, Lippiello L, Zarins A. Biochemical andmetabolic abnormalities in articular cartilage from osteoarthritichuman hips. II. Correlation of morphology with biochemical andmetabolic data. J Bone Joint Surg Am 1971;53:523–37.

24. Saha N, Moldovan F, Tardif G, Pelletier J-P, Cloutier J-M,Martel-Pelletier J. Interleukin-1b–converting enzyme/caspase-1 inhuman osteoarthritic tissues: localization and role in the matura-tion of interleukin-1b and interleukin-18. Arthritis Rheum 1999;42:1577–87.

25. Sleat DE, Lobel P. Ligand binding specificities of the two mannose6-phosphate receptors. J Biol Chem 1997;272:731–8.

26. Villiger PM, Lotz M. Differential expression of TGF beta isoformsby human articular chondrocytes in response to growth factors.J Cell Physiol 1992;151:318–25.

27. Boumediene K, Conrozier T, Mathieu P, Richard M, Marcelli C,Vignon E, et al. Decrease of cartilage transforming growth factor-beta receptor II expression in the rabbit experimental osteoarthri-tis: potential role in cartilage breakdown. Osteoarthritis Cartilage1998;6:146–9.

28. Sporn MB, Roberts AB. Peptide growth factors are multifunc-tional. Nature 1988;332:217–9.

29. Sporn MB, Roberts AB. The multifunctional nature of peptidegrowth factors. In: Sporn MB, Roberts AB, editors. Handbook ofexperimental pharmacology. Heidelberg: Springer-Verlag; 1990. p.3–15.

30. Martel-Pelletier J, Cloutier JM, Pelletier JP. Cathepsin B andcysteine protease inhibitors in human OA: Effect of intra-articularsteroid injections. J Orthop Res 1990;8:336–44.

31. Shull MM, Ormsby I, Kier AB, Pawlowski S, Diebold RJ, Yin M,et al. Targeted disruption of the mouse transforming growthfactor-beta 1 gene results in multifocal inflammatory disease.Nature 1992;359:693–9.

32. Kulkarni AB, Huh CG, Becker D, Geiser A, Lyght M, FlandersKC, et al. Transforming growth factor beta 1 null mutation in micecauses excessive inflammatory response and early death. Proc NatlAcad Sci U S A 1993;90:770–4.

33. Matrat M, Lardot C, Huaux F, Broeckaert F, Lison D. Role ofurokinase in the activation of macrophage-associated TGF-beta insilica-induced lung fibrosis. J Toxicol Environ Health 1998;55:359–71.

34. Martel-Pelletier J, Faure MP, McCollum R, Mineau F, CloutierJM, Pelletier JP. Plasmin, plasminogen activators and inhibitor inhuman osteoarthritic cartilage. J Rheumatol 1991;18:1863–71.

35. Denault JB, D’Orleans-Juste P, Masaki T, Leduc R. Inhibition ofconvertase-related processing of proendothelin-1. J CardiovascPharmacol 1995;26 Suppl 3:S47–50.

36. Blanchette F, Day R, Dong W, Laprise MH, Dubois CM. TGF-beta1 regulates gene expression of its own converting enzymefurin. J Clin Invest 1997;99:1974–83.

37. Dubois CM, Laprise MH, Blanchette F, Gentry LE, Leduc R.Processing of transforming growth factor beta 1 precursor byhuman furin convertase. J Biol Chem 1995;270:10618–24.

38. Mallet WG, Maxfield FR. Chimeric forms of furin and TGN38 aretransported with the plasma membrane in the trans-Golgi networkvia distinct endosomal pathways. J Cell Biol 1999;146:345–59.

39. Schafer W, Stroh A, Berghofer S, Seiler J, Vey M, Kruse ML, et al.Two independent targeting signals in the cytoplasmic domaindetermine trans-Golgi network localization and endosomal traf-ficking of the proprotein convertase furin. EMBO J 1995;14:2424–35.

40. Jirtle RL, Carr BI, Scott CD. Modulation of insulin-like growthfactor-II/mannose 6-phosphate receptors and transforming growthfactor-beta 1 during liver regeneration. J Biol Chem 1991;266:22444–50.

41. Dennis PA, Rifkin DB. Cellular activation of latent transforminggrowth factor beta requires binding to the cation-independentmannose 6-phosphate/insulin-like growth factor type II receptor.Proc Natl Acad Sci U S A 1991;88:580–4.

42. De Bleser PJ, Jannes P, van Buul-Offers SC, Hoogerbrugge CM,van Schravendijk CF, Niki T, et al. Insulinlike growth factor-II/mannose 6-phosphate receptor is expressed on CCl4-exposed ratfat-storing cells and facilitates activation of latent transforminggrowth factor-beta in cocultures with sinusoidal endothelial cells.Hepatology 1995;21:1429–37.

43. Ghahary A, Tredget EE, Mi L, Yang L. Cellular response to latentTGF-beta1 is facilitated by insulin-like growth factor-II/mannose-6-phosphate receptors on MS-9 cells. Exp Cell Res 1999;251:111–20.

44. Kovacina KS, Steele-Perkins G, Purchio AF, Lioubin M, Miya-zono K, Heldin CH, et al. Interactions of recombinant and platelettransforming growth factor-beta 1 precursor with the insulin-likegrowth factor II/mannose 6-phosphate receptor. Biochem BiophysRes Commun 1989;160:393–403.

45. Purchio AF, Cooper JA, Brunner AM, Lioubin MN, GentryLE, Kovacina KS, et al. Identification of mannose 6-phosphatein two asparagine-linked sugar chains of recombinant trans-forming growth factor-beta 1 precursor. J Biol Chem 1988;263:14211–5.

ACTIVATION OF TGFb IN OA CARTILAGE 2109