Inhibition of fibroblast activation protein ameliorates cartilage matrix degradation and osteoarthritis progression

Fibroblast activation protein (Fap) is a serine protease that degrades denatured type I collagen, α2-antiplasmin and FGF21. Fap is highly expressed in bone marrow stromal cells and functions as an osteogenic suppressor and can be inhibited by the bone growth factor Osteolectin (Oln). Fap is also expressed in synovial fibroblasts and positively correlated with the severity of rheumatoid arthritis (RA). However, whether Fap plays a critical role in osteoarthritis (OA) remains poorly understood. Here, we found that Fap is significantly elevated in osteoarthritic synovium, while the genetic deletion or pharmacological inhibition of Fap significantly ameliorated posttraumatic OA in mice. Mechanistically, we found that Fap degrades denatured type II collagen (Col II) and Mmp13-cleaved native Col II. Intra-articular injection of rFap significantly accelerated Col II degradation and OA progression. In contrast, Oln is expressed in the superficial layer of articular cartilage and is significantly downregulated in OA. Genetic deletion of Oln significantly exacerbated OA progression, which was partially rescued by Fap deletion or inhibition. Intra-articular injection of rOln significantly ameliorated OA progression. Taken together, these findings identify Fap as a critical pathogenic factor in OA that could be targeted by both synthetic and endogenous inhibitors to ameliorate articular cartilage degradation.


INTRODUCTION
Osteoarthritis (OA) is one of the most common orthopedic diseases worldwide, 1,2 with an approximately 26.6% prevalence rate among people over 45 years old. 3 The most prominent pathological changes of OA include articular cartilage degeneration, osteophyte formation, low-grade inflammation and subchondral bone remodeling. [4][5][6][7][8] After initial mechanical erosion during aging, articular cartilage and synovium express a panel of proteolytic enzymes and proinflammatory factors that accelerate cartilage matrix degradation and OA progression. [9][10][11] Early-stage OA patients can be treated by microfracture, osteochondral allograft transplantation, or biomaterial implantation, 12,13 while late-stage OA patients can only be treated by replacement plastic surgeries. 14 Pharmacological treatments such as nonsteroidal antiinflammatory drugs and intra-articular glucocorticoid injection are recommended by most guidelines to relieve inflammation and pain. [15][16][17] In contrast, the effectiveness of other drugs, such as glucosamine, chondroitin and hyaluronic acid, remains controversial. [15][16][17] Therefore, drugs with higher efficacy for OA treatment are under intensive investigation. 18 Aggrecan (Acan) and type II collagen (Col II) are the two major components of the cartilage matrix. 19 Whereas Acan degradation is mainly mediated by a disintegrin and metalloprotease with thrombospondin motifs (ADAMTS) family proteases, 20 Col II degradation is mainly catalyzed by matrix metalloproteinases (MMPs). 21 Collagenases, such as MMP1/8/13, are responsible for unwinding and cleaving the triple-helical Col II fiber into large fragments, [22][23][24][25] after which gelatinases, such as MMP2/9, further digest them into smaller peptides. 26,27 Although MMPs were initially considered potential drug targets for OA treatment, none of the nonselective MMP inhibitors passed clinical trials due to potential side effects, including joint stiffness, inflammation and pain. 28,29 In contrast, selective MMP inhibitors demonstrated low solubility and permeability, 23 precluding their clinical applications. In addition to ADAMTS and MMPs, serine proteases, such as thrombin, matriptase and HtrA1, were also implicated in OA progression. [30][31][32][33] However, no serine protease inhibitors have been approved to treat OA.
Fibroblast activation protein (Fap) is a serine protease with both dipeptidyl peptidase and endopeptidase activities. 34,35 It is known to degrade denatured type I collagen (Col I), alpha2-antiplasmin and FGF21 in vivo [35][36][37][38][39] as well as several other substrates in vitro. 40,41 Fap was upregulated in activated fibroblasts and critically involved in tumor microenvironment formation, inflammation and wound healing. [42][43][44][45][46][47] In the musculoskeletal system, Fap is highly expressed in bone marrow stromal cells (BMSCs) and osteoblasts, which function as osteogenic suppressors. 48 Genetic deletion or pharmacological inhibition of Fap promotes bone formation and inhibits bone resorption, suggesting that it is a potential anti-osteoporosis drug target. 48 Interestingly, Fap was also detected in the synovium of OA patients, although at a lower level compared to rheumatoid arthritis (RA) patients. 49,50 Whether Fap plays an important role in regulating OA progression remains poorly understood.
Osteolectin (Oln), also known as Clec11a or Scgf, was initially identified as a growth factor that promotes hematopoietic colony formation in vitro. 51,52 Our previous in vivo studies showed that Oln is highly expressed in osteochondral lineage cells and functions as an anabolic factor that promotes bone formation by activating the Wnt pathway. 53, 54 We also demonstrated that Oln interacts with Fap and inhibits its protease activity, 48 suggesting a novel mechanism by which Oln promotes bone formation and osteogenic differentiation. 48 In the present study, we found that synovium-derived Fap exacerbates OA progression. Oln forms a protective barrier on the cartilage surface and shows the opposite effects. A mechanistic study revealed that Fap promotes the degradation of Col II in osteoarthritic cartilage, which could be ameliorated by intra-articular injection of a small molecule inhibitor of Fap or recombinant Oln.

Fap is significantly upregulated in OA synovium
To test whether Fap is involved in OA pathogenesis, we first analyzed its expression in the synovium of patients diagnosed with OA or acute joint injury (control). Immunofluorescence staining showed that human FAP was mainly detected in the OA synovium ( Fig. 1a and Supplementary Fig. 1a). Quantitative realtime PCR (qPCR) confirmed that FAP mRNA was significantly upregulated in OA synovium compared to control synovium (Fig.  1b). Western blot analysis showed an even greater increase in FAP protein levels in the OA synovium (Fig. 1c). Consistent with the human data, Fap was only marginally detected in the synovium of sham-operated mice, whereas it was significantly upregulated in OA synovium after surgical destabilization of the medial meniscus (DMM) (Fig. 1d and Supplementary Fig. 1b). To test whether the inflammatory microenvironment induced by OA 10 regulates Fap expression, we stimulated primary human synovial fibroblasts with IL-1β ex vivo. We found that FAP mRNA was significantly increased in a time-dependent manner, suggesting that proinflammatory factors promote Fap expression in the synovium (Fig. S1c).
Genetic deletion or pharmacological inhibition of Fap ameliorates OA progression To test whether Fap regulates OA progression, we performed unilateral knee joint DMM surgery in Fap-deficient (Fap KO) and wild-type control mice. Three days after the surgery, we administered weekly intra-articular injections of a Fap-selective small molecule inhibitor (FAPi, 40 μg·kg −1 body weight) 55 or vehicle control (PBS) in the DMM-treated knee joints for 8 weeks. Mice were then sacrificed, and knee joint sections of the medial compartment were stained with safranin O and fast green.
Histomorphometry analyses in the contralateral knee joint without DMM surgery showed no significant differences between Fap KO and control mice ( Fig. S2a-f), suggesting that Fap is dispensable for knee joint development and maintenance under a steady state. In contrast, the knee joint with DMM surgery showed significant cartilage erosion, subchondral bone thickening and synovitis in control mice ( Fig. 2a-f). Notably, these OA symptoms were significantly ameliorated in Fap KO mice ( Fig. 2a-f). Intraarticular administration of FAPi in control mice also significantly ameliorated cartilage erosion, subchondral bone thickening and synovitis compared to the vehicle group ( Fig. 2a-f). Importantly, FAPi administration showed no therapeutic effects in Fap KO mice, indicating that the inhibitor functioned in a Fap-dependent manner ( Fig. 2a-f). Immunofluorescent staining of Col2a1 in the articular cartilage showed similar results (Fig. 2g, h). We also used micro-CT to analyze osteophyte formation and found that the intra-articular administration of FAPi significantly ameliorated osteophyte formation in control but not Fap KO mice (Fig. S3a,  b). There was also a trend toward decreased osteophyte formation in Fap KO mice compared to control mice after vehicle treatment ( Supplementary Fig. 3a, b).
To test whether FAPi administration could ameliorate articular symptoms after the onset of OA, we first performed DMM surgery in wild-type mice and waited for 4 weeks. We then administered weekly intra-articular injections of FAPi or vehicle control for another 8 weeks. Compared to sham-operated mice, vehicletreated mice showed profound cartilage erosion, subchondral bone thickening and synovitis after DMM (Fig. 3a-f), which were significantly ameliorated by FAPi treatment (Fig. 3a-f). Immunofluorescent staining of Col2a1 in the articular cartilage showed similar results (Fig. 3g, h). Taken together, these data showed that  56 Consistent with this, Fap was mainly detected in the synovium, but not articular cartilage, of normal and OA knee joints in adult mice (Fig. 1d). To test whether Fap regulates chondrocyte function in a cell-autonomous manner, we cultured primary articular chondrocytes from Fap KO or wild-type newborn mice and stimulated the cells with IL-1β to mimic OA pathogenesis. [57][58][59] Consistent with a previous study, 60 qPCR analysis in control chondrocytes showed significantly decreased Acan and Col2a1 mRNA levels and significantly increased Mmp3 mRNA levels after IL-1β stimulation (Fig. S4a-c). Compared to control chondrocytes, Fap KO chondrocytes showed similar Acan, Col2a1 and Mmp3 expression before and after IL-1β stimulation (Fig. S4a-c). Consistent with this, western blot analysis showed significantly decreased Acan and Col2a1 protein levels and significantly increased Mmp3 protein levels after IL-1β stimulation (Fig. S4d, e), with no significant differences between Fap KO and control chondrocytes ( Supplementary Fig. 4d, e).  Quantification of the OARSI score (d), subchondral bone thickness (e) and synovitis score (f). OARSI scores were calculated based on the erosion of the medial tibial plateau cartilage (a, bottom). Subchondral bone thickness was measured as the mean distance of five evenly distributed measuring points between the lower edge of the articular cartilage and the roof of the cancellous bone. Synovitis scores were calculated by summing the enlargement of the synovial lining cell layer, density of the resident cells and inflammatory infiltration. g, h Immunostaining of Col II in the knee joints (g) with quantifications (h) (n = 6 mice per genotype in each treatment group). DAPI staining indicates the nucleus. Scale bars: 100 μm. The statistical significance was assessed using two-way ANOVAs with Sidak's multiple comparison tests. Data are presented as the mean ± SD (*P < 0.05, **P < 0.01, ***P < 0.001) Next, we treated primary wild-type chondrocytes with 10 μg/ml FAPi for different periods ex vivo but still found no significant changes in Acan, Col2a1 or Mmp3 mRNA levels ( Supplementary  Fig. 5a-c). Different doses of FAPi were also administered with or without IL-1β stimulation during ex vivo culture, which did not significantly affect the expression of Acan, Col2a1 or Mmp3 in primary wild-type chondrocytes ( Fig. S5d-f). Taken together, these data suggest that Fap does not autonomously regulate chondrocyte function.
Col II is a novel substrate of Fap both in vitro and in vivo Fap participates in extracellular matrix remodeling in many diseases, such as pancreatic cancer, myocardial infarction, and thin-cap fibroatheroma, 61-63 suggesting that it might play similar roles during OA progression. Since Fap is a serine protease that degrades denatured or Mmp1-cleaved Col I, 35,63-65 we hypothesized that Fap could also degrade Col II, the key component of the cartilage matrix. Similar to native Col I, native Col II was resistant to digestion by recombinant Fap (rFap) (Supplementary Fig. 6a). Thus, we heat-inactivated Col II at 95°C for 10 min to depolymerize it into monomers and then incubated it with rFap at 37°C for 24 h. Remarkably, we observed a dose-and time-dependent degradation of denatured Col II by rFap (Fig. 4a, b). Preincubation with FAPi for 30 min dose-dependently inhibited the degradation of Col II by rFap (Fig. 4c), further validating the efficacy of FAPi. Our recent study demonstrated that Oln functions as an endogenous Fap inhibitor to promote bone formation. 48 To test whether Oln inhibits the degradation of denatured Col II by rFap, we overexpressed mouse Fap, alone or together with mouse Oln, in HEK293T cells and immunoprecipitated Fap (Fig. 4d). Consistent with our previous study, 48 Oln physically interacts with Fap, which does not affect the amount of Fap that was expressed or precipitated ( Fig. 4d). Immunoprecipitated Fap also degraded denatured Col II, which could be partially inhibited by Oln coexpression (Fig. 4e).
Next, we tested whether Fap exhibits gelatinase activity similar to MMP2/9, which further digests MMP13-cleaved Col II into smaller fragments. 26,27 Recombinant human MMP13 (rMMP13, 10 μg·mL −1 ) was able to digest native Col II into 6 major fragments in a time-dependent manner (Fig. 4f). When we terminated the protease activity of MMP13 using EDTA after 12 h of incubation and then added rFap (10 μg·mL −1 ) for another 12 h, the 55, 40 and 30 kD fragments could be further digested by rFap (Fig. 4f, g). To test whether Fap degrades Col II in a pathological setting, we measured the Fap level in synovial fluids from OA patients by ELISA and found that its average concentration was approximately 150 ng·mL −1 (Fig. S6b). A previous study reported that the concentration of MMP13 in OA synovial fluids is approximately 200 ng·mL −1 . 66, 67 We then cultured primary chondrocytes from newborn mice and administered rFap (200 ng·mL −1 ), rMMP13 (200 ng·mL −1 ), or rFap plus rMMP13 for 24 h (Fig. 4h). qPCR analysis showed no significant changes in Col2a1 mRNA levels in any treatment group ( Supplementary Fig. 6c). Western blot analysis showed unaltered Col2a1 protein levels in total cell lysates after rFap or rMMP13 treatment ( Fig. 4i) but significantly decreased Col2a1 protein levels in the rFap plus rMMP13 group (Fig. 4i). Together, these data indicate that Fap works in concert with MMP13 to degrade Col II both in vitro and ex vivo. Step 1 Step 2 Relative grayscale 0.  To test whether rFap exacerbates OA progression and Col II degradation in vivo, we administered weekly intra-articular injections of rFap after the onset of OA (4 weeks after DMM surgery) for 8 weeks. Compared to vehicle controls, rFap-treated mice showed significantly increased articular cartilage erosion, subchondral bone thickening and synovitis (Fig. 5a-f). Micro-CT analysis showed increased osteophyte formation after rFap injection (Fig. S7a, b). Immunofluorescent staining of Col2a1 showed a significant decrease in the Col2a1 + area in the articular cartilage of rFap-treated mice (Fig. 5g, h). Together with the fact that the Col II level was significantly elevated in the articular cartilage by genetic or pharmacological inhibition of Fap after DMM (Figs. 2g, h and 3g, h), we concluded that Col II is a novel substrate of Fap in vivo.
To test whether Fap degrades Acan, another key component of the cartilage matrix, we incubated rFap with recombinant human Acan (G1-IGD-G2 domain) at 37°C for 24 h. rFap could not digest either native or denatured Acan ( Supplementary Fig. 6d, e), indicating that Fap specifically degrades Col II in the cartilage matrix.
Oln is significantly downregulated in OA cartilage Since Oln functions as an endogenous inhibitor of Fap, 48 we tested whether Oln also regulates OA progression. Our previous study showed that Oln is highly expressed in growth plate chondrocytes, osteoblasts and BMSCs. 53 Immunofluorescent staining showed that Oln is also expressed in the superficial layer of normal human articular cartilage but absent in the lesioned cartilage from OA patients ( Fig. 6a and Supplementary Fig. 8a). qPCR analysis showed that OLN levels were significantly decreased in lesioned OA cartilage compared to adjacent normal cartilage (Fig. 6b), which was confirmed by western blotting (Fig. 6c). In contrast to Fap, Oln was not detected in either normal or OA patient synovium (Supplementary Fig. 8b).
We also analyzed the localization of Oln in the mouse knee joint and found that it is expressed in the superficial layer of articular cartilage and meniscus (Fig. 6d). Consistent with the human data, mouse Oln was dramatically downregulated after DMM surgery (Fig. 6d). Mouse Oln was also marginally detected in the boundary between the synovium and meniscus, which was not affected by DMM ( Supplementary Fig. 8c). No Oln signal could be detected in IgG control or Oln-deficient (Oln KO) mice ( Supplementary Fig. 8d). Interestingly, Oln mRNA was significantly decreased in primary chondrocytes after IL-1β stimulation ex vivo ( Supplementary Fig.  8e), suggesting that proinflammatory factors inhibit Oln expression during OA progression.
Genetic deletion of Oln exacerbates OA progression To test whether Oln regulates OA progression, we performed DMM surgery in Oln KO, Oln/Fap KO (double KO) and wild-type control mice. Histomorphometry analysis showed no significant differences among Oln KO, double KO and control mice in the contralateral knee joints without DMM surgery (Fig. S9a-f), suggesting that neither Oln nor Fap are required for articular cartilage development or maintenance under steady state. In the knee joints with DMM surgery, Oln KO mice showed significantly increased cartilage erosion, subchondral bone thickening, and synovitis compared to control mice (Fig. 7a-f). Importantly, these defects were significantly ameliorated in double KO mice (Fig. 7a-f). Col2a1 staining of the articular cartilage showed similar results (Fig. 7g, h).
Consistent with this finding, weekly intra-articular administration of FAPi also significantly ameliorated OA progression in Oln KO mice (Fig. S10a-h). Oln deficiency did not affect Acan and Col2a1 levels in primary chondrocytes after IL-1β stimulation, while there was a modest but significant increase in Mmp3 levels compared to those in wild-type controls (Fig. S11a-e). Taken together, these findings indicate that the genetic deletion of Oln exacerbates OA progression, which could be partially rescued by Fap deletion or inhibition.

Intra-articular injection of recombinant Oln attenuates OA progression
To test whether recombinant Oln (rOln) showed similar therapeutic effects as those of FAPi, we administered weekly intra-articular injections of recombinant rOln (240 μg·kg −1 body weight) or vehicle control (0.05 mg·mL −1 hyaluronic acid) in wild-type mice Quantification of the OARSI score (d), subchondral bone thickness (e) and synovitis score (f). g, h Immunostaining of Col II in the knee joints (g) with quantification (h) (n = 6 mice per genotype). DAPI staining indicates the nucleus. Scale bars: 100 μm. The statistical significance was assessed using one-way ANOVAs with Tukey's multiple comparison tests. Data are presented as the mean ± SD (*P < 0.05, **P < 0.01, ***P < 0.001) Fap inhibition ameliorates OA A Fan et al.
3 days after DMM surgery for a total of 8 weeks. Histomorphometry analysis showed significantly decreased cartilage erosion, subchondral bone thickening, and synovitis compared to vehicle controls ( Fig. S12a-f), which was confirmed by Col2a1 staining of articular cartilage (Fig. S12g, h). To test whether rOln could ameliorate articular symptoms after the onset of OA, we administered weekly intra-articular injections of rOln or vehicle control in wild-type mice 4 weeks after DMM surgery for a total of 8 weeks. The vehicle-treated DMM group showed profound cartilage erosion, subchondral bone thickening and synovitis compared to the sham-operated group, and these effects were significantly ameliorated in the rOln-treated group (Fig. 8a-f). Similar results were obtained by immunostaining of Col2a1 in articular cartilage (Fig. 8g, h). Taken together, these data showed that intra-articular rOln administration significantly attenuated OA progression.

DISCUSSION
OA is a degenerative orthopedic disease that severely affects the quality of life of elderly people worldwide. Apart from surgical treatments, very few medicines have been developed to effectively treat OA symptoms by either promoting cartilage regeneration or preventing its degradation. 68 In the present study, we found that synovial Fap levels positively correlated with OA progression and that genetic deletion or pharmacological inhibition of Fap significantly ameliorated OA symptoms. Furthermore, we showed that Fap degrades Col II both in vitro and in vivo, and this effect could be inhibited by both FAPi and its endogenous inhibitor Oln. Interestingly, Oln is expressed in the superficial layer of the articular cartilage and meniscus to form a protective barrier, which negatively correlates with OA progression. Genetic deletion of Oln promotes OA progression in a Fap-dependent manner, while intra-articular administration of rOln ameliorates OA. Together, this study identified Fap as a potential drug target to treat OA.
Fap is structurally and functionally similar to Dpp4, 69 which is a dipeptidyl peptidase that shares many substrates with Fap. 70 DPP4 inhibitors, such as sitagliptin, are orally available drugs that have been approved by the FDA to treat type 2 diabetes. 71 Although the FAPi we used in this study is only 9-fold selective over Dpp4, 55 it showed no therapeutic effects in Fap KO mice (Fig. 2). This finding suggests that Fap, but not Dpp4, is the downstream target of FAPi. Fap has both membrane-bound and soluble forms. 39 Since synoviocytes do not infiltrate the cartilage matrix in OA, 9 we reason that the soluble form of Fap secreted from OA synovium contributes to Col II degradation, which was confirmed by our ELISA analysis (Supplementary Fig. 6b). In contrast to OA, RA is characterized by severe inflammation and synoviocyte infiltration. 9 Given that Fap + synovial fibroblasts are known to play key roles in RA 72 and that genetic deletion of Fap significantly ameliorates articular cartilage damage in a mouse model of RA, 50 it is intriguing to test whether Fap inhibitors could also be used to treat RA in the future.
Previous studies showed that Fap degrades denatured Col I 35,63 as well as MMP1-cleaved native Col I. 64,65 Here, we provide evidence that Fap also degrades denatured Col II and MMP13cleaved native Col II in vitro and primary chondrocyte-derived Col II ex vivo, thereby adding a new member to the substrate list of Fap. While MMP13 can cleave native Col II into large fragments, Fap helps to digest them into even smaller peptides (Fig. 4f). It is worth noting that Fap is a prolyl-specific serine protease and that 17% of COL2A1 amino acids are composed of proline. Thus, MMP13 and Fap work together for the efficient degradation of native Col II, the major component of the articular cartilage. Col II is also an in vivo substrate of Fap during OA progression. The Col II level was significantly increased after genetic deletion or inhibition of Fap (Figs. 2h and 3h) and significantly decreased after intra-articular administration of rFap (Fig. 5h). Unfortunately, we were unable to test whether FAPi attenuates OA progression in Col2a1-deficient mice, which died perinatally due to chondrodysplasia. 73,74 Therefore, we cannot rule out the possibility that other substrates might also mediate the therapeutic effects of Fap inhibition.
MMPs play critical roles in cartilage degradation during OA progression, 75 while tissue inhibitors of metalloproteinases (TIMPs) protect against OA. 76 Whether similar antagonistic mechanisms exist to maintain the homeostasis of the articular cartilage remains poorly understood. Our previous study showed that Oln interacts with Fap and functions as an endogenous inhibitor to promote bone mineralization. 48 Here, we found that Oln also localizes to the articular surface, which forms a protective barrier to prevent Fap-dependent degradation of Col II. During OA progression, cartilage erosion first occurs at the articular surface due to mechanical wear, 77 which leads to Oln disruption and further deterioration of the cartilage matrix. Notably, Oln KO mice did not exhibit spontaneous OA, possibly because the Fap level is low in the knee joint under steady state (Fig. 1). Unlike FAPi, rOln cannot directly inhibit the serine protease activity of Fap in vitro (data not shown). However, Oln overexpression partially inhibited immunoprecipitated Fap (Fig. 4e), suggesting that they could form an inhibitory complex in vivo together with other protein adaptors. This study falls short of dissecting the detailed molecular mechanism by which Fap is inhibited by Oln, which could be achieved by purification of the inhibitory complex coupled with crystallography/cryo-EM analysis. We also noted that rOln showed limited therapeutic effects in vivo when dissolved in PBS (data not shown). However, HA seemed to be a better vehicle, probably by immobilizing rOln on the articular surface. Future studies are needed to optimize the vehicle and route of administration for FAPi and rOln and to test whether they could ameliorate OA progression in larger animal models and clinical trials.

Human samples
This study was carried out in compliance with the Helsinki Declaration. Synovium, articular cartilage and synovial fluids were collected from the knee joints of OA patients (Kellgren-Lawrence grade 78 3 or 4) during arthroplasty procedures. Cartilage samples were examined under the stereoscope. The smooth regions with no obvious lesions were separated as normal cartilage, while the rough regions were separated as lesioned cartilage. Control synovial tissues were obtained from patients with no radiographic cartilage changes undergoing exploratory arthroscopy. Detailed patient information is summarized in Table S1.

Mice
Fap KO mice (#024288) 56 were obtained from the Jackson Laboratory and maintained by crossing with C57BL/6J wild-type mice (#000664). Parts of exons 4 and 5 were replaced by a LacZneo cassette in this strain, which led to the genetic ablation of Fap. The generation of Oln KO mice was previously described. 53 Wildtype C57BL/6J mice were used for in vivo administration of rOln and FAPi. Male mice at approximately 10 weeks of age were used to perform the DMM surgery. qPCR Total RNA was extracted from human samples or mouse primary chondrocytes using TRIzol reagent (Invitrogen) and then reverse transcribed into cDNA using a 5X All-In-One MasterMix kit (ABM). qPCR was performed using iTaq Universal SYBR Green Supermix (Bio-Rad) on a CFX96 real-time system (Bio-Rad). The PCR primers used included hACTB (NM_001101. Cryosection and immunostaining Human synovium, cartilage or mouse knee joint samples were fixed in 4% paraformaldehyde overnight at 4°C. Mouse knee joints were decalcified in 10% EDTA (pH 7.4) for 5-7 days. All samples were dehydrated in 30% sucrose for 1 day at room temperature and sectioned at 10 μm using the CryoJane tape-transfer system (Leica). Sections were incubated in blocking buffer (PBS with 5% horse serum) for 1 h at room temperature and then stained DMM model Mice were anesthetized with 1.5% pentobarbital (80 mg·kg −1 body weight). Ketoprofen (5 mg·kg −1 body weight) was administered subcutaneously for pain relief. Knee joints were shaved and sterilized with Betadine and 70% ethanol, and a 4 mm longitudinal incision was made from the inferior pole of the patellar to the proximal tibial plateau. The joint capsule was opened medial to the patellar tendon with microiris scissors. The fat pad over the intercondylar area was bluntly dissected. For the DMM operation, the medial meniscotibial ligament was incised with microiris scissors, and the mobility of the anterior horn of the medial meniscus was tested. For the sham operation, no other procedure was performed. The joint capsule and the skin were then closed separately with a 7-0 cutting PGA suture line (Jinhuan Medical).

Intra-articular injection
Mice were randomly chosen for each intervention group. Mice were anesthetized with isoflurane, and knee joints were sterilized with 70% ethanol. The needle of the insulin syringe was sagittally inserted into the intercondylar area of the mouse knee, where 10 μl of FAPi (MCE, HY-101801), rFap (R&D Systems, 8647-SE-010) or rOln was injected. For the prevention model, intra-articular injections of FAPi (40 μg·kg −1 body weight, in PBS), rOln (240 μg·kg −1 body weight, in 0.05 mg·mL −1 HA) or the corresponding vehicle were given 3 days after the surgery. For the treatment model, intra-articular injections of FAPi, rFap (40 μg·kg −1 body weight, in PBS), rOln or the corresponding vehicle were given 4 weeks after DMM surgery.
Histomorphometry analysis Dissected mouse knee joints were fixed in 4% paraformaldehyde overnight at 4°C and decalcified in 10% EDTA for 14 days. Paraffin-embedded knee joints were sectioned at 6-μm thickness and stained with hematoxylin, 0.5% Safranin O solution and 0.2% Fast Green solution. Grading was performed by calculating OARSI scores (medial tibial plateau) 79 and synovitis scores (summation of lining layer enlargement, density of resident cells, and inflammatory infiltration). 80 Subchondral bone thickness was measured as the mean distance of five evenly distributed measuring points between the lower edge of the articular cartilage and the roof of the cancellous bone. 81,82 The average score of 6 knee sections was calculated for statistical analysis.
Micro-CT analysis Dissected mouse knee joints were fixed in 4% paraformaldehyde overnight at 4°C and replaced with PBS before micro-CT analysis. Knee joints were scanned at an isotropic voxel size of 7 μm, with a peak tube voltage of 70 kV and current of 0.114 mA (mCT 35; Scanco Medical AG, Bassersdorf, Switzerland). A three-dimensional Gaussian filter (s = 0.8) with a limited, finite filter support of one was used to suppress noise in the images, and a threshold of 220-1 000 was used to segment mineralized bone from the air and soft tissues. The region of interest was selected for calcified osteophyte tissues, and the bone volume was calculated to measure the osteophyte size.

Immunoprecipitation
Fap-HA, Oln-Flag, and control GFP constructs were transfected alone or together into the HEK293T cell line (ATCC) using Lipofectamine 3000 (Thermo Fisher Scientific). Three days after transfection, cells were washed with ice-cold PBS and lysed in 500 μL ice-cold RIPA buffer (100 mmol·L −1 Tris-HCl, 150 mmol·L −1 NaCl, 1% Triton-X-100, 1X protease inhibitor cocktails). After centrifugation at 12 000 r·min −1 for 5 min, the supernatants were incubated with 10 μL anti-HA affinity gel (Sigma) overnight at 4°C and washed three times with RIPA buffer and three times with PBS. Immunoprecipitated proteins were incubated with denatured Col II at 37°C for 24 h and then subjected to 6% SDS-PAGE and colloidal blue staining as described above. Immunoprecipitation efficiency was detected by western blotting with anti-HA (Sigma, H6908) and anti-Flag (Sigma, F1804) antibodies.
ELISA Secreted FAP levels in human synovial fluids were measured using a Human FAP ELISA Kit (Abcam, ab193701) according to the manufacturer's instructions. Briefly, human synovial fluids were diluted 1:100 using PBS buffer, and 100 μL of diluted synovial fluid was added to a precoated 96-well ELISA plate at 4°C for 16 h. The plate was then washed 4 times with 1× wash buffer, and 100 μL of biotinylated human FAP detection antibody (diluted in 1× assay diluent B) was added and incubated at room temperature for 1 h. After washing with 1× wash buffer 4 times, 100 μL of HRP Streptavidin solution (diluted in 1× assay diluent B) was added and incubated at room temperature for 45 min. After washing with 1× wash buffer 4 times, 100 μL of TMB one-step substrate reagent was added to each well and incubated at room temperature in the dark for 30 min. Finally, 50 μL of stop solution was added to each well, and the optical density was measured at 450 nm.

Statistics
The statistical significance of differences between two groups was assessed using two-tailed Student's t tests. The statistical significance of differences among more than two groups was assessed using one-way ANOVAs with Tukey's multiple comparison tests or two-way ANOVAs with Sidak's multiple comparison tests. All data are presented as the mean ± SD. P values less than 0.05 were considered significant (*P < 0.05, **P < 0.01, ***P < 0.001).

DATA AVAILABILITY
All data are available from the corresponding authors upon reasonable request.