The BMP Antagonists Follistatin and Gremlin in Normal and Early Osteoarthritic Cartilage: an Immunohistochemical Study G

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

International Cartilage Repair Society

The BMP antagonists follistatin and gremlin in normal and early osteoarthritic cartilage: an immunohistochemical study G. Tardif Ph.D., J.-P. Pelletier M.D., C. Boileau Ph.D. and J. Martel-Pelletier Ph.D.* Osteoarthritis Research Unit, University of Montreal Hospital Centre, Notre-Dame Hospital, Montreal, Quebec, Canada

Summary

Objective: Bone morphogenic protein (BMP) activities are controlled in part by antagonists. In human osteoarthritic (OA) cartilage, the BMP antagonists follistatin and gremlin are increased but differentially regulated. Using the OA dog model, we determined if these BMP antagonists were produced at different stages during the disease process by comparing their in situ temporal and spatial distribution. Methods: Dogs were sacrificed at 4, 8, 10 and 12 weeks after surgery; normal dogs served as control. Cartilage was removed, differentiating fibrillated and non-fibrillated areas. Immunohistochemistry and morphometric analyses were performed for follistatin, gremlin, BMP-2/4 and IL-1b. Growth factor-induced gremlin expression was assessed in dog chondrocytes. Results: Follistatin and gremlin production were very low in normal cartilage. Gremlin was significantly up-regulated in both non-fibrillated and fibrillated areas at 4 weeks, and only slightly increased with disease progression. Follistatin showed a time-dependent increased level in the non-fibrillated areas with significance reached at 8e12 weeks; in the fibrillated areas significant high levels were seen as early as 4 weeks. In the whole cartilage, follistatin and IL-1b temporal production showed similar patterns; this was also true for gremlin and BMP-2/4, though BMP- 2/4 production was already high in the normal dogs. Interestingly, data revealed that basic fibroblast growth factor (bFGF) could be another factor increasing gremlin expression early in the disease process. Comparison between superficial and deep zones revealed similar patterns for follistatin and IL-1b in the superficial zone only; gremlin and BMP-2/4 had similar patterns in both zones. Conclusion: Data show, for the first time, different spatial and temporal production of gremlin and follistatin in cartilage during OA progression. These findings may reflect different roles for each antagonist in this disease. ª 2008 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved. Key words: BMP antagonists, Follistatin, Gremlin, Osteoarthritis, Animal model.

e Introduction and OA adult human cartilage10 12. BMP-13 (cartilage derived morphogenic protein (CDMP)-2, growth and differen- Osteoarthritis (OA) is a degenerative disease of the articular tiation factor (GDF)-6) and BMP-14 (CDMP-1, GDF-5) joints, characterized by cartilage degradation. The breakdown were also found in both normal and OA adult cartilage11. of the cartilage matrix leads to the development of fibrillation Despite increased BMP levels in the OA cartilage, overall and fissures, the appearance of gross ulcerations, and full- anabolic activities are decreased with disease progres- thickness loss. Alterations in the collagen network and proteo- 1,2 sion, suggesting the presence of some factor(s) interfer- glycans are observed early during the disease process . ing with their activity. Among the factors up-regulated in human OA are the BMP activities are controlled in part by specific antago- bone morphogenic proteins (BMPs) and some of their an- 13,14 e nists , which prevent BMP interaction with their specific tagonists3 6. BMPs belong to the transforming growth fac- 7,8 receptors. Many BMP antagonists have been identified and tor (TGF)-ß super-family of secreted signaling molecules . they differ in their specificity and affinity for the BMPs. For They play important roles during embryogenesis, and pro- example, gremlin binds BMP-2, -4 and -7 with similar affin- mote the formation of cartilage and bone. They also con- ity15,16, whereas follistatin also binds these BMPs, but with tribute to the tissue homeostasis of adult cartilage by a higher affinity for BMP-717,18. promoting the production of some cartilage macromole- The levels at which BMP antagonists are expressed in cules. Several BMPs are present in the adult human car- 9 pathological tissues may lead to decreased BMP anabolic tilage and some are found elevated during OA. For activity, consequently affecting tissue repair and remodel- example, BMP-2/4 was shown to be increased in OA car- 6 ing. We previously reported the involvement of three BMP tilage and present in osteophytes . BMP-6 and BMP-7 antagonists, follistatin, gremlin and chordin, in human appear to be present at similar levels in both normal OA3,4. From these previous studies, we hypothesized that follistatin and gremlin might be produced in the cartilage *Address correspondence and reprint requests to: Dr Johanne at different stages during the OA process. Martel-Pelletier, Ph.D., Osteoarthritis Research Unit, University of It is not possible to monitor the early stages of primary hu- Montreal Hospital Centre, Notre-Dame Hospital, 1560 Sherbrooke Street East, Montreal, Quebec, Canada H2L 4M1. Tel: 1-514- man OA, not only because of the slow course of the degen- 890-8000x26658; Fax: 1-514-412-7582; E-mail: jm@ erative process, but also because of ethical considerations martelpelletier.ca that preclude taking biopsies at different stages of the dis- Received 23 April 2008; revision accepted 29 June 2008. ease. However, the use of animal models allows for such

263 264 G. Tardif et al.: BMP antagonists in early osteoarthritis time-oriented studies. One of the well-characterized types (3) substitution of the primary antibody with an autologous pre-immune se- of surgically induced OA models is the sectioning of the an- rum. These controls showed only background staining. Each section was examined under a light microscope (Leitz Diaplan; Wild Leitz, St. Laurent, QC, terior cruciate ligament (ACL) in dogs, as the pathological Canada) and photographed with a Retiga OEM Fast camera (QImaging, Sur- structural changes seen in this model reproduce those of rey, BC, Canada). human OA. These changes include cartilage destruction, synovial inflammation, subchondral bone alterations, and the presence of osteophytes19e22. In this model, the earliest MORPHOMETRIC ANALYSIS time point reported for site-specific gene expression is 2 Antigen levels were quantified as previously described4 by determining weeks post-surgery, with no evidence of OA-related the number of chondrocytes that stained positive. The total number of cells changes in gross histologic (cell morphology, density and and the number of cells that stained positive for the specific antigen were first counted for the whole cartilage and then separately for each zone of the car- distribution and matrix architecture and composition) or bio- tilage, at 40 magnification. Each slide was examined from 3 fields of the chemical (proteoglycan and collagen levels) assess- superficial (superficial and upper intermediate layers) and 3 fields of the ments23. However, noticeable changes are detected by 4 deep (lower intermediate and deep layers) zones. The results were ex- weeks and major cartilage lesions occur within 12 weeks pressed as the percentage of cells that stained positive for the antigen, the 20,24,25 maximum being 100%. Each slide was subjected to a double-blind evalua- post-surgery . In addition, the OA dog model also tion; the final score was determined by consensus of the 2 readers. shows parameters indicative of the involvement of pro- inflammatory and catabolic factors similar to those seen in 22,26 human OA at the clinical stage of the disease . There- CELL CULTURE fore, because this model displays the characteristics of Chondrocytes were released from full-thickness strips of whole normal human OA, it is suitable for the study of OA progress. dog cartilage by sequential enzymatic digestion at 37C28. The cells were The aim of this study was to investigate the production of seeded at high density (105/cm2) and cultured in Dulbecco’s modified the two BMP antagonists, follistatin and gremlin, using the Eagle’s medium (DMEM; Gibco BRL, Burlington, ON, Canada) supple- experimental OA dog model, to examine their temporal mented with 10% heat-inactivated fetal calf serum (FCS; Gibco BRL) and an antibiotic mixture (100 units/ml penicillin base and 100 mg/ml streptomycin and spatial changes in the cartilage, and to correlate their base; Gibco BRL) at 37 C in a humidified atmosphere of 5% CO2/95% air. levels with those of anabolic (BMP-2/4) and catabolic The effects of growth factors on gremlin expression were assessed by (IL-1b) factors involved in OA cartilage. To this end, immu- pre-incubating first-passage cultured chondrocytes in DMEM/0.5% FCS for nohistochemistry combined with morphometric analysis 24 h followed by another 24 h incubation with fresh DMEM/0.5% FCS containing bFGF (100 ng/ml; R&D Systems), GDF-9 (100 ng/ml; Peprotech was used as it is the most appropriate methodology to Inc., Rocky Hill, NJ, USA) or BMP-2 (100 ng/ml; R&D Systems). Total ribo- investigate the production of these factors at specific loca- nucleic acid (RNA) was extracted and processed for real-time polymerase tions and to follow their presence throughout the evolution chain reaction (PCR). of the disease.

RNA EXTRACTION AND REAL-TIME PCR

Materials and methods Total RNA was extracted and quantified4. Reverse transcription was primed with random hexamers. Real-time PCR was performed as previously ANIMALS described4 in the Rotor-Gene RG-3000A (Corbett Research, Mortlake, Twenty-four adult crossbred dogs (2e3 years of age, 20e25 kg each) Australia) with the SYBR GreenER PCR Master Mix (Invitrogen, Carlsbad, were used. OA was induced in 17 dogs by surgically sectioning the ACL CA, USA) used according to the manufacturer’s specifications. The 18S of the right knee through a stab wound25. Prior to surgery, the animals was used as the housekeeping gene. The data were given as a threshold DðDCt Þ were anesthetized intravenously with pentobarbital sodium (25 mg/kg) and cycle (Ct). Fold changes in gene expression were calculated as 2 , intubated. Following surgery, the dogs were kept in animal care facilities where DCt ¼ Ct stimulated Ct18S, and D(DCt) ¼ DCt stimulated DCt control. The primer efficiencies for the genes were the same as those for the 18S for 1 week, and were then sent to a housing farm. Seven non-operated (nor- 0 mal) animals were used as controls. Dogs were housed in a large pen where gene. The dog-specific primer sequences were 5 -GAATCAGGGTT CGATTCCG (sense), 50-CCAAGATCCAACTACGAGC (antisense) for 18S they could exercise ad libitum. All dogs were sacrificed by intravenous injec- 0 0 tion. Operated dogs were sacrificed 4 weeks (n ¼ 4), 8 weeks (n ¼ 4), 10 and 5 -ATGTGACCGAGCGCAAGTAC (sense), 5 -ATATGCATCGACA weeks (n ¼ 5), or 12 weeks (n ¼ 4) after surgery. This study was approved CTGCTTG (antisense) for gremlin. by the Institutional Ethics Committee.

STATISTICAL ANALYSIS SPECIMEN SELECTION AND PREPARATION Data are expressed as the mean S.E.M. Statistical significance was as- Full-thickness sections of articular cartilage from tibial plateaus of operated sessed by 2-tailed Student’s t test, and P values < 0.05 were considered dogs were removed from the lesional (fibrillated) and non-lesional (non-fibril- significant. lated) areas as previously described27. Specimens from normal dogs obtained from the same anatomical sites presented no macroscopic or microscopic lesions. Cartilage specimens were fixed in TissuFix #2 (Chaptec, Montreal, QC, Canada), embedded in paraffin, and 5 mm sections prepared. Results

MORPHOLOGICAL CHANGES

IMMUNOHISTOCHEMISTRY The monitoring of specific factors during the OA process was done with cartilage obtained from non-operated (nor- The specimens were processed as previously described27. The primary antibodies were a mouse monoclonal anti-human follistatin (50 mg/ml; R&D mal) or operated dogs sacrificed at 4, 8, 10 and 12 weeks Systems, Minneapolis, MN, USA), a goat polyclonal anti-human gremlin post-surgery. Specimens were taken from the non-lesional (4 mg/ml; Santa Cruz, Santa Cruz, CA, USA), a rabbit polyclonal anti-human (non-fibrillated) and lesional (fibrillated) areas to localize BMP-2/4 (4 mg/ml; Santa Cruz), a rabbit antiserum against IL-1b (2.1 mg/ml; the temporal and spatial production of the factors. Cartilage Rockland, Gilbertsville, PA, USA), and a rabbit polyclonal anti-bFGF (1/400 dilution; Chemicon, Temecula, CA, USA). Preliminary experiments showed histology of the normal dogs (Fig. 1) showed normal struc- that each of the antibodies recognized the specific proteins. ture, cellularity, and matrix staining. OA changes over time Staining specificity was determined by: (1) immunoadsorption for 1 h at were similar to those described previously24,26,29: few le- 37 C with the corresponding protein or peptide: 20 molar excess recombi- sions were detected at 4 weeks post-surgery, more lesions nant follistatin (R&D Systems), gremlin immunizing peptide (Santa Cruz), BMP-2/4 peptide (Santa Cruz), recombinant IL-1b protein (R&D Systems), appeared between the 4th and 8th weeks and progressed recombinant bFGF (R&D Systems); (2) omission of the primary antibody; over time, being most severe at 10 and 12 weeks. OA Osteoarthritis and Cartilage Vol. 17, No. 2 265

Normal

Superficial zone

Deep zone

OA (8 weeks) Non-Fibrillated Fibrillated

Superficial zone

Deep zone

Fig. 1. Representative histological sections of dog tibial plateaus from normal (non-operated) and OA cartilage 8 weeks post-surgery from non- fibrillated and fibrillated areas. Cartilage from normal dogs shows normal structure, cellularity, and matrix staining. The non-fibrillated area of the OA cartilage presents surface irregularity (dotted arrows), hypercellularity and cloning (arrows) and slight discoloration of the matrix at the superficial zone. The fibrillated area of the OA cartilage is characterized by fibrillation (arrowheads), hypocellularity, and matrix discoloration extending to the lower layers, as well as loss of matrix integrity and cartilage thickness. Magnification: 100.

non-lesional areas showed surface irregularities, hypercel- (P < 0.02) than in the normal cartilage and only slightly in- lularity and cloning, and slight discoloration of the matrix creased over time. at the superficial layer. The lesional areas showed fibrilla- The spatial production was monitored by evaluating pro- tions, hypocellularity and matrix discoloration at the superfi- tein levels in the superficial (superficial and upper interme- cial and intermediate layers, reflecting loss of matrix diate layers) and deep (lower intermediate and deep layers) integrity and cartilage thickness. zones of the cartilage. As illustrated in Figs. 3 and 4, differ- ences were found in the distribution of the antagonists between the superficial and deep zones. IMMUNOHISTOMORPHOMETRIC ANALYSIS

Immunohistomorphometry was chosen to determine the Superficial zone (Fig. 3) relative amounts of follistatin, gremlin, BMP-2/4 and IL-1b over time during OA progression and their locations in the Each factor studied was present in the superficial zones cartilage. of the normal dogs, but at different levels: follistatin and There was a time-dependent increase of follistatin pro- gremlin at low levels of about 4%, IL-1b and BMP-2/4 at duction in the non-fibrillated OA cartilage (Fig. 2). This is 13% and 22%, respectively. There was a progressive pro- in contrast to the fibrillated areas, where follistatin levels duction of follistatin over time in the non-fibrillated areas were already high at week 4 and remained relatively stable of OA cartilage that reached statistical difference from over time. week 8. In contrast, follistatin levels in the fibrillated areas Gremlin levels (Fig. 2) were already statistically higher at were already significantly higher at week 4 compared to 4 weeks in both the non-fibrillated (P < 0.008) and fibrillated the normal (P < 0.001) and the non-fibrillated areas at 12 (P < 0.0001) areas compared to the normal cartilage. weeks (P < 0.02). The levels in the fibrillated areas did Levels in the non-fibrillated and fibrillated areas of OA car- not vary from 4 to 10 weeks (21e24%), and a slight de- tilage only slightly increased over time. Furthermore, grem- crease was found at 12 weeks. lin levels were not significantly different in the non-fibrillated Gremlin levels were significantly higher at 4 weeks in and fibrillated areas for each given week. the non-fibrillated areas (P < 0.03) and the levels slightly BMP-2/4 levels (Fig. 2) were high in the normal dogs, and increased with disease progression, but not significantly slightly increased over time in the non-fibrillated cartilage to compared to week 4. In the fibrillated areas, levels reach statistical significance at 10 and 12 weeks post-sur- were already statistically higher at 4 weeks (P < 0.001) gery. Levels in the fibrillated cartilage at 4 weeks post-sur- and remained elevated and stable until the end of the gery were significantly higher than in the normal and non- study at week 12. fibrillated cartilage for the same week and remained stable BMP-2/4 levels were relatively high in the normal carti- over time. lage, and did not vary significantly in the OA non-fibrillated The IL-1b production profile followed a pattern similar to cartilage, although a small increase was found at 10 and 12 that of follistatin (Fig. 2); a significant increase occurred weeks. The levels at these times were similar to those seen over time in the non-fibrillated cartilage, while levels at at 4 weeks in the fibrillated areas. A small but significant in- week 4 in the fibrillated cartilage were significantly higher crease was noted at week 10 in the fibrillated areas. 266 G. Tardif et al.: BMP antagonists in early osteoarthritis

Whole cartilage 35 Follistatin 35 Gremlin 30 30 25 25 p< 0.0001

20 p< 0.0001 p< 0.0001 p< 0.0001 20 p< 0.0001 p< 0.0001 p< 0.0001 p< 0.0001 p< 0.0001

15 p< 0.0001 15 p< 0.001 p< 0.008 p< 0.0001 10 p< 0.0001 10 p< 0.03 5 5

% of positive chondrocytes 0 0 N 4 8 10 12 4 8 10 12 N 4 8 10 12 4 8 10 12 NF F NF F

BMP-2/4 IL-1

35 p< 0.02 35

p< 0.003 30 p< 0.01

30 p< 0.003 p< 0.007 p< 0.0001 p< 0.05 25 p< 0.01 25 p< 0.004 p< 0.001 p< 0.0007 p< 0.0008 20 20 p< 0.02 15 15 10 10 5 5 0 0

% of positive chondrocytes N 4 8 10 12 4 8 10 12 N 4 8 10 12 4 8 10 12 NF F NF F

Fig. 2. Temporal production levels of follistatin, gremlin, BMP-2/4, and IL-1b in the full thickness of cartilage from normal and operated dogs at 4, 8, 10 and 12 weeks in non-ﬁbrillated (NF) and ﬁbrillated (F) areas. Protein production was determined by immunomorphometric analysis as described in Materials and methods. P values were calculated by 2-tailed Student’s t test.

IL-1b levels increased over time in the non-fibrillated areas at week 4 (P < 0.0007). These levels decreased at areas of the cartilage; statistical significance was reached 8e10 weeks but reverted at 12 weeks. Of note, these for weeks 8e12 compared to the normal cartilage. Levels values were not significantly different within the weeks. IL- decreased slightly at week 12, suggesting attenuation in 1b levels were low in the deep layers of normal cartilage the inflammatory process. IL-1b levels in the fibrillated (3%) and increased in a time-dependent fashion in both areas were significantly higher at week 4 (P < 0.02), re- non-fibrillated and fibrillated areas. Levels were maximal mained relatively constant at weeks 8 and 10, and as at week 10, and as in the superficial zone, decreased seen in the non-fibrillated cartilage, decreased slightly at slightly at week 12 in the non-fibrillated areas. week 12. In summary, gremlin and BMP-2/4 levels in the deep In summary, follistatin and IL-1b levels in the superficial zone followed a similar production pattern especially in zone showed similar profiles in both fibrillated and non- the non-fibrillated areas, IL-1b increased with disease pro- fibrillated areas; gremlin and BMP-2/4 also showed similar gression, and follistatin levels remained very low. profiles, the only difference being a high BMP-2/4 level in the normal cartilage. FACTORS INVOLVED IN THE UP-REGULATION OF GREMLIN

Deep zone (Fig. 4) Our previous studies have shown that BMP-2/4 was the only factor tested that could up-regulate gremlin expres- The levels of the factors studied were lower in the deep sion in human chondrocytes4. However, in the normal (Fig. 4) than in the superficial zones (Fig. 3). There were dog cartilage, the high levels of BMP-2/4 contrasted with very few cells showing immunoreactivity toward follistatin the low levels of gremlin and its statistical up-regulation in the deep zone for both non-fibrillated and fibrillated areas. early during the disease process. In order to identify which Slightly higher levels were found in the fibrillated areas com- factor could have contributed to the increased gremlin pro- pared to the non-fibrillated, which reached significance only duction in the OA cartilage, we determined the effects of after week 8. Gremlin production in the superficial zone in- some factors either related to the OA process or known creased with disease progression in both the non-fibrillated up-regulators of gremlin expression, BMP-2, bFGF and and fibrillated areas. GDF-9. BMP-2/4 production in the deep zone differed from that Chondrocytes (n ¼ 5) were treated for 24 h with these seen in the superficial zones: normal cartilage levels were factors and the expression level of gremlin was determined lower (6%), production increased over time in the non-fibril- by real-time PCR. The results showed that GDF-9 did not lated areas from week 4 (12%) to week 12 (26%), and significantly affect gremlin expression but that BMP-2 and levels were already significantly higher in the fibrillated bFGF markedly induced gremlin expression by 1.6 0.1 Osteoarthritis and Cartilage Vol. 17, No. 2 267

Superficial zone Follistatin Gremlin 40 40 35 35 30 30 p< 0.0001 p< 0.0001 p< 0.0001

25 25 p< 0.0001 p< 0.0001 p< 0.0001 p< 0.0001 p< 0.0001 20 20 p< 0.0001 p< 0.0002 p< 0.0001 15 p< 0.003 15 p< 0.0001 p< 0.003

10 p< 0.004 10 5 5 0 0 N 410128 48N 10 12 410128 48 10 12 % of positive chondrocytes NF F NF F

BMP-2/4 IL-1

40 40 p< 0.02 35 35 p< 0.002 30 p< 0.0003 p< 0.002 30 p< 0.002 p< 0.002 25 25 p< 0.03 p< 0.03 20 20 15 15 10 10 5 5 0 0 N 410128 410128 N 410128 410128 % of positive chondrocytes NF F NF F

Fig. 3. Temporal production levels of follistatin, gremlin, BMP-2/4, and IL-1b in the superficial zone of cartilage from normal and operated dogs at 4, 8, 10 and 12 weeks in non-fibrillated (NF) and fibrillated (F) areas. Protein production was determined by immunomorphometric analysis as described in Materials and methods. P values were calculated by 2-tailed Student’s t test.

fold (P < 0.0001) and 3.2 0.9 fold (P < 0.03), respectively. play critical mediatory roles. Indeed, it is hypothesized that Hence, we looked for bFGF production in normal and OA during the course of the disease insufficient or poor quality dog cartilage (8 weeks) to verify whether the presence of of new matrix macromolecules occurs as a consequence of this factor could correlate with the up-regulation of gremlin deficient stimulation. Since growth factors of the BMP family production with time. Results showed (Fig. 5) that although are important regulators of matrix production, decreased ac- similar levels of bFGF (20e22%) were produced in both tivity of these factors through increased production of BMP normal and OA cartilage and positive-staining cells were antagonists could lead to the development of an altered ma- mainly localized in the superficial zones of both, a very trix. We and others have shown that BMP antagonists are important difference was the presence of marked positive produced in adult cartilage, and that some are found at staining throughout the extracellular matrix in the OA higher levels in human OA cartilage3e5. We have also cartilage only. shown that in human OA cartilage, follistatin and gremlin are produced in different areas3,4: follistatin is found primarily in the superficial layers of cartilage, while gremlin Discussion is mostly detected in the upper intermediate layers. As follistatin is up-regulated by the inflammatory molecules Animal models of OA allow evaluation of temporal and tumor necrosis factor (TNF)-a and interferon (IFN)-g4,we spatial changes occurring in joint tissues during the pro- hypothesized that this antagonist would be linked particu- gression of the disease. Although no animal model will rep- larly to the inflammatory process of the disease, while resent exactly the events leading to human OA, they gremlin, because of its up-regulation by BMPs, would be provide a good basis for understanding the role of specific associated preferentially with the remodeling phase. molecules in the evolution of the disease. We chose the Results from this study confirm that the production of both ACL dog model as it mimics changes seen in human OA follistatin and gremlin is up-regulated during OA, and that at the molecular level but also because pro-inflammatory follistatin is located mostly in the superficial zone of the car- factors play a role in bringing about the structural changes, tilage while gremlin is produced in both the superficial and as they do in human OA. As our study aimed at investigat- deeper layers. These data, in addition to the significant ing the production of factors in cartilage in a temporal and up-regulation of gremlin at 4 weeks post-surgery, earlier spatial manner, immunohistochemisty/morphometric analy- than follistatin, confirm our hypothesis that gremlin is in- sis was used as it provides data not only on the relative volved during the early phases. abundance of a protein, but also its specific location The very early stage of OA has been characterized by throughout the evolution of the disease. a hypertrophic biochemical repair reaction and increased While the etiology of articular cartilage degeneration seen synthesis of extracellular matrix. BMP-2 and -4 are growth in OA is not fully understood, cytokines and growth factors factors that are expressed at this stage. Nakase et al.6 268 G. Tardif et al.: BMP antagonists in early osteoarthritis

Deep zone Follistatin Gremlin

20 20 p< 0.0005 p< 0.003 p< 0.002 15 15 p< 0.008 p< 0.007 p< 0.02

10 p< 0.01 10

5 p< 0.004 5 p< 0.009

0 0 N 410128 410128 N 410128 48 10 12 % of positive chondrocytes NF F NF F

BMP-2/4 IL-1

40 20 p< 0.0001 p< 0.003

35 p< 0.004 p< 0.001 p< 0.002

30 p< 0.02

p< 0.0007 15 25 p< 0.02 20 p< 0.02 10 p< 0.01 15 10 5 5 0 0 N 410128 410128 N 410128 48 10 12 % of positive chondrocytes NF F NF F

Fig. 4. Temporal production levels of follistatin, gremlin, BMP-2/4, and IL-1b in the deep zone of cartilage from normal and operated dogs at 4, 8, 10 and 12 weeks in non-fibrillated (NF) and fibrillated (F) areas. Protein production was determined by immunomorphometric analysis as described in Materials and methods. P values were calculated by 2-tailed Student’s t test. have reported that BMP-2 is scarce in normal adult human than follistatin, are associated with the remodeling phase of cartilage, is found in moderately damaged OA cartilage in the cartilage matrix. the upper and middle layers, and in severely damaged car- Moreover, the fact that follistatin is involved later during tilage extends to the deep layers. Interestingly, our previous the OA process as well as its similar topographical distribu- work showed a similar production pattern for gremlin in hu- tion with IL-1b at the superficial zone, suggests that the two man OA cartilage4. In this study, similar expression patterns molecules may be regulated by similar pro-inflammatory of gremlin and BMP-2/4 were also observed: they are pro- stimuli, strongly supporting a link between follistatin produc- duced in both superficial and deep cartilage zones, in con- tion and the inflammatory process of this disease. Also trast to follistatin which is produced mainly in the superficial supportive of this link is that follistatin preferentially binds zone, strongly suggesting that BMP-2/4 and gremlin, rather activin, a molecule expressed during some inflammatory

Fig. 5. Representative sections of bFGF immunostaining from normal and OA dog cartilage 8 weeks post-surgery. bFGF was found in chondrocytes from both normal and OA cartilage, mainly at the superficial zone. However, diffuse staining in the extracellular matrix was observed only in the OA cartilage throughout the cartilage layers. Magnification: 100. Osteoarthritis and Cartilage Vol. 17, No. 2 269 arthropathies30, over BMPs. Of note, and according to our also binds factors other than BMPs, such as the polypeptide previous work with human OA chondrocytes4, IL-1b is not YWHAH (tyrosine 3-monooxygenase/tryptophan 5-mono- the likely factor responsible for the increased follistatin pro- oxygenase activation protein eta polypeptide)39, a protein duction, but was used in this study as it is indicative of the that mediates signal transduction by binding to phosphoser- inflammatory process occurring in this tissue. ine-containing proteins, and the Slit proteins, molecules in- One unexpected finding was that the BMP-2/4 levels volved in the regulation of monocyte chemotaxis40.In were relatively high in the normal dog cartilage and were addition, gremlin was shown to activate the transcription not up-regulated in the superficial zone of the cartilage, of p21Cip141, involved in the cell cycle. Finally, gremlin whereas gremlin levels were low in the normal cartilage was also reported to be a proangiogenic factor42. Although and significantly increased at 4 weeks in both non-fibrillated these factors have not yet been related to the OA process, it and fibrillated areas. These data suggest that factors other will be interesting to investigate whether gremlin’s other ac- than BMP-2/4 are able to stimulate gremlin in dog OA car- tivities/properties might be involved in the physiological or tilage. We looked for factors that might trigger the up-regu- pathological cartilage processes. lation of gremlin in dog chondrocytes. GDF-9 has been In summary, despite the constant presence of BMPs in shown to increase gremlin expression in mouse granulosa OA cartilage, these molecules might not be optimally active cells31, but our data showed that this factor does not signif- due to the increased levels of specific antagonists resulting icantly alter gremlin expression in dog chondrocytes. In con- in decreased production of matrix macromolecules such as trast, bFGF as well as BMP-2 markedly up-regulated collagen and proteoglycan, and a shift toward degradation gremlin expression in dog chondrocytes; furthermore, rather than matrix synthesis and repair. The potential inter- bFGF was found at an increased level in the OA cartilage ference with BMP activity of different BMP antagonists dur- matrix. Although bFGF is one of many factors present in ing OA progression should be taken into consideration in OA cartilage, it may modulate gremlin expression in vivo the design of cartilage repair therapies involving BMPs. as it is produced by the chondrocytes, stored in the extracellular matrix and is released from the matrix upon cartilage damage/injury. In addition, bFGF was suggested to Conflict of interest contribute to the progression of OA as it interferes with the anabolic activities of insulin-like growth factor (IGF)-1 This research was supported by a grant from the CIHR/ and BMP-7 in chondrocytes32 and stimulates matrix metal- IMHA. loprotease (MMP)-13 expression33, a metalloprotease ca- pable of degrading many components of the extracellular e matrix28,34 36. Our data show that bFGF is present in chon- Acknowledgments drocytes as well as in the OA cartilage matrix throughout the cartilage layers. The release of this growth factor may We thank Martin Boily for his technical assistance in immu- then contribute to the up-regulation of gremlin seen in OA nohistochemistry, David Hum for assistance in the cell cul- dog cartilage. ture and real-time PCR experiments, and Virginia Wallis for In our previous study of gremlin regulation in OA human the manuscript preparation. This research was supported chondrocytes4, we found that among all the cytokines and by a grant from the CIHR/IMHA. growth factors tested, only BMP-2 and -4 increased gremlin while IL-1b decreased its expression. Data from this study are in agreement with this observation, as they show an in- References b verse correlation between gremlin and IL-1 production: 1. Hollander AP, Pidoux I, Reiner A, Rorabeck B, Bourne R, Poole RA. gremlin production in the whole cartilage decreases at 10 Damage to type II in aging and osteoarthritis starts at the articular sur- weeks, which is also the time when IL-1b increases, and face, originates around chondrocytes, and extends into the cartilage its production is again increased at 12 weeks, when IL-1b with progressive degeneration. J Clin Invest 1995;96:2859e69. 2. Dodge GR, Poole RA. Immunohistochemical detection and immuno- is decreased. chemical analysis of type II collagen degradation in human normal, IL-1b has been reported to increase BMP-2/4 expression rheumatoid, and osteoarthritic articular cartilages and in explants of 37 in normal and OA human chondrocytes . This cytokine may bovine articular cartilage cultured with interleukin-1. J Clin Invest also be a regulating factor for BMP-2/4 levels in the dog 1989;83:647e61. b 3. Tardif G, Pelletier JP, Hum D, Boileau C, Duval N, Martel-Pelletier J. model. Indeed, although IL-1 basal levels were relatively Differential regulation of the bone morphogenic protein antagonist high in the whole cartilage, production slowly increased chordin in human normal and osteoarthritic chondrocytes. Ann Rheum with disease progression in the non-fibrillated areas and Dis 2006;65:261e4. was already significantly higher at week 4 in the fibrillated 4. Tardif G, Hum D, Pelletier JP, Boileau C, Ranger P, Martel-Pelletier J. Differential gene expression and regulation of the bone morphoge- areas. This pattern was somewhat similar for BMP-2/4. netic protein antagonists follistatin and gremlin in normal and osteoar- However, although BMP-2/4 levels increased over time, thritic human chondrocytes and synovial fibroblasts. Arthritis Rheum this was accompanied by increased levels of gremlin and fol- 2004;50:2521e30. listatin, thus reducing the anabolic activities of the BMPs. 5. Nakayama N, Han CY, Cam L, Lee JI, Pretorius J, Fisher S, et al.A novel chordin-like BMP inhibitor, CHL2, expressed preferentially in The up-regulation of gremlin and follistatin in OA cartilage chondrocytes of developing cartilage and osteoarthritic joint cartilage. is likely the consequence of the actions of different factors, Development 2004;131:229e40. resulting in differential production at different stages of the 6. Nakase T, Miyaji T, Tomita T, Kaneko M, Kuriyama K, Myoui A, et al. disease. Gremlin is associated with the remodeling/early Localization of bone morphogenetic protein-2 in human osteoarthritic cartilage and osteophyte. Osteoarthritis Cartilage 2003;11:278e84. phase and follistatin with the inflammatory process. This 7. Canalis E, Economides AN, Gazzerro E. Bone morphogenetic proteins, suggests the possibility that these proteins have functions their antagonists, and the skeleton. Endocr Rev 2003;24:218e35. other than being extracellular BMP antagonists in the OA 8. Ducy P, Karsenty G. The family of bone morphogenetic proteins. Kidney process. Indeed, different functions were recently discov- Int 2000;57:2207e14. 9. Lories RJ, Luyten FP. Bone morphogenetic protein signaling in joint ho- ered for gremlin. In relation to the BMP system, gremlin meostasis and disease. Cytokine Growth Factor Rev 2005;16:287e98. was recently found to interact with the intracellular BMP-4 10. Bobacz K, Gruber R, Soleiman A, Erlacher L, Smolen JS, Graninger WB. 38 precursor and prevent the secretion of mature BMP-4 .It Expression of bone morphogenetic protein 6 in healthy and 270 G. Tardif et al.: BMP antagonists in early osteoarthritis

osteoarthritic human articular chondrocytes and stimulation of matrix experimental dog osteoarthritis by licofelone is associated with a re- synthesis in vitro. Arthritis Rheum 2003;48:2501e8. duction in the synthesis of MMP-13 and cathepsin K. Bone 2004;34: 11. Bobacz K, Gruber R, Soleiman A, Graninger WB, Luyten FP, Erlacher L. 527e38. Cartilage-derived morphogenetic protein-1 and -2 are endogenously 28. Reboul P, Pelletier JP, Tardif G, Cloutier JM, Martel-Pelletier J. The new expressed in healthy and osteoarthritic human articular chondrocytes collagenase, collagenase-3, is expressed and synthesized by human and stimulate matrix synthesis. Osteoarthritis Cartilage 2002;10: chondrocytes but not by synoviocytes: A role in osteoarthritis. J Clin 394e401. Invest 1996;97:2011e9. 12. Chubinskaya S, Merrihew C, Cs-Szabo G, Mollenhauer J, McCartney J, 29. Pelletier JP, Mineau F, Faure MP, Martel-Pelletier J. Imbalance between Rueger DC, et al. Human articular chondrocytes express osteogenic the mechanisms of activation and inhibition of metalloproteases in the protein-1. J Histochem Cytochem 2000;48:239e50. early lesions of experimental osteoarthritis. Arthritis Rheum 1990;33: 13. Yanagita M. BMP antagonists: their roles in development and involvement 1466e76. in pathophysiology. Cytokine Growth Factor Rev 2005;16:309e17. 30. Gribi R, Tanaka T, Harper-Summers R, Yu J. Expression of activin A in 14. Balemans W, Van Hul W. Extracellular regulation of BMP signaling in inflammatory arthropathies. Mol Cell Endocrinol 2001;180:163e7. vertebrates: a cocktail of modulators. Dev Biol 2002;250:231e50. 31. Pangas SA, Jorgez CJ, Matzuk MM. Growth differentiation factor 9 reg- 15. Topol LZ, Bardot B, Zhang Q, Resau J, Huillard E, Marx M, et al. Biosyn- ulates expression of the bone morphogenetic protein antagonist grem- thesis, post-translation modification, and functional characterization of lin. J Biol Chem 2004;279:32281e6. Drm/Gremlin. J Biol Chem 2000;275:8785e93. 32. Loeser RF, Chubinskaya S, Pacione C, Im HJ. Basic fibroblast growth 16. Hsu DR, Economides AN, Wang X, Eimon PM, Harland RM. The Xen- factor inhibits the anabolic activity of insulin-like growth factor 1 and opus dorsalizing factor Gremlin identifies a novel family of secreted osteogenic protein 1 in adult human articular chondrocytes. Arthritis proteins that antagonize BMP activities. Mol Cell 1998;1:673e83. Rheum 2005;52:3910e7. 17. Iemura S, Yamamoto TS, Takagi C, Uchiyama H, Natsume T, 33. Im HJ, Muddasani P, Natarajan V, Schmid TM, Block JA, Davis F, et al. Shimasaki S, et al. Direct binding of follistatin to a complex of bone- Basic fibroblast growth factor stimulates matrix metalloproteinase-13 morphogenetic protein and its receptor inhibits ventral and epidermal via the molecular cross-talk between the mitogen-activated protein cell fates in early Xenopus embryo. Proc Natl Acad Sci U S A 1998; kinases and protein kinase Cdelta pathways in human adult articular 95:9337e42. chondrocytes. J Biol Chem 2007;282:11110e21. 18. Yamashita H, ten Dijke P, Huylebroeck D, Sampath TK, Andries M, 34. Monfort J, Tardif G, Roughley P, Reboul P, Boileau C, Bishop PN, et al. Smith JC, et al. Osteogenic protein-1 binds to activin type II receptors Identification of opticin, a member of the small leucine-rich repeat pro- and induces certain activin-like effects. J Cell Biol 1995;130:217e26. teoglycan family, in human articular tissues: a novel target for MMP- 19. Dedrick DK, Goulet RW, O’Connor BL, Brandt KD. Preliminary report: in- 13 in osteoarthritis. Osteoarthritis Cartilage 2008;16:749e55. creased porosity of the subchondral plate in an accelerated canine 35. Monfort J, Tardif G, Reboul P, Mineau F, Roughley P, Pelletier JP, et al. model of osteoarthritis. Osteoarthritis Cartilage 1997;5:71e4. Degradation of small leucine-rich repeat proteoglycans by matrix met- 20. Brandt KD. Insights into the natural history of osteoarthritis provided by alloprotease-13: identification of a new biglycan cleavage site. Arthri- the cruciate-deficient dog. An animal model of osteoarthritis [Review]. tis Res Ther 2006;8:R26. Ann N Y Acad Sci 1994;732:199e205. 36. Tardif G, Pelletier JP, Dupuis M, Geng C, Cloutier JM, Martel-Pelletier J. 21. Dedrick DK, Goldstein SA, Brandt KD, O’Connor BL, Goulet RW, Collagenase 3 production by human osteoarthritic chondrocytes in re- Albrecht M. A longitudinal study of subchondral plate and trabecular sponse to growth factors and cytokines is a function of the physiolog- bone in cruciate-deficient dogs with osteoarthritis followed up for 54 ical state of the cells. Arthritis Rheum 1999;42:1147e58. months. Arthritis Rheum 1993;36:1460e7. 37. Fukui N, Zhu Y, Maloney WJ, Clohisy J, Sandell LJ. Stimulation of BMP- 22. Pelletier JP, Martel-Pelletier J, Ghandur-Mnaymneh L, Howell DS, 2 expression by pro-inflammatory cytokines IL-1 and TNF-alpha in Woessner JF Jr. Role of synovial membrane inflammation in cartilage normal and osteoarthritic chondrocytes. J Bone Joint Surg Am matrix breakdown in the Pond-Nuki dog model of osteoarthritis. Arthri- 2003;85-A(Suppl 3):59e66. tis Rheum 1985;28:554e61. 38. Sun J, Zhuang FF, Mullersman JE, Chen H, Robertson EJ, 23. Stoker AM, Cook JL, Kuroki K, Fox DB. Site-specific analysis of gene Warburton D, et al. BMP4 activation and secretion are negatively reg- expression in early osteoarthritis using the Pond-Nuki model in ulated by an intracellular gremlin-BMP4 interaction. J Biol Chem 2006; dogs. J Orthop Surg 2006;1:8. 281:29349e56. 24. Fernandes JC, Martel-Pelletier J, Lascau-Coman V, Moldovan F, 39. Namkoong H, Shin SM, Kim HK, Ha SA, Cho GW, Hur SY, et al. The Jovanovic D, Raynauld JP, et al. Collagenase-1 and collagenase-3 bone morphogenetic protein antagonist gremlin 1 is overexpressed synthesis in early experimental osteoarthritic canine cartilage. An im- in human cancers and interacts with YWHAH protein. BMC Cancer munohistochemical study. J Rheumatol 1998;25:1585e94. 2006;6:74. 25. Pelletier JP, Martel-Pelletier J, Altman RD, Ghandur-Mnaymneh L, 40. Chen B, Blair DG, Plisov S, Vasiliev G, Perantoni AO, Chen Q, et al. Howell DS, Woessner JF Jr. Collagenolytic activity and collagen ma- Cutting edge: bone morphogenetic protein antagonists Drm/Gremlin trix breakdown of the articular cartilage in the Pond-Nuki dog model of and Dan interact with Slits and act as negative regulators of monocyte osteoarthritis. Arthritis Rheum 1983;26:866e74. chemotaxis. J Immunol 2004;173:5914e7. 26. Pelletier JP, Faure MP, Di Battista JA, Wilhelm S, Visco D, Martel- 41. Chen B, Athanasiou M, Gu Q, Blair DG. Drm/Gremlin transcriptionally Pelletier J. Coordinate synthesis of stromelysin, interleukin-1, and activates p21(Cip1) via a novel mechanism and inhibits neoplastic oncogene proteins in experimental osteoarthritis. An immunohisto- transformation. Biochem Biophys Res Commun 2002;295:1135e41. chemical study. Am J Pathol 1993;142:95e105. 42. Stabile H, Mitola S, Moroni E, Belleri M, Nicoli S, Coltrini D, et al. The 27. Pelletier JP, Boileau C, Brunet J, Boily M, Lajeunesse D, Reboul P, et al. bone morphogenic protein antagonist Drm/gremlin is a novel pro- The inhibition of subchondral bone resorption in the early phase of angiogenic factor. Blood 2006;109:1834e40.