Positive Feedback Regulation Between USP15 and ERK2 Promotes Cartilage Repair in Experimental Osteoarthritis

Background.The transforming growth factor-β (TGF-β) signaling pathway plays an essential role in maintaining homeostasis in joints affected by osteoarthritis (OA). Determine the specic relationship between non-SMAD and classical SMAD signaling pathways. Methods.First,we found from human cartilage tissue specimens positive expressions of TGF-β1 and USP15between normal andosteoarthritiscartilage tissueby immunohistochemical.Then, overexpressing and knocking out USP15 or ERK2 followed bythe latest gene editing technology Crispr/Cas9 supported the studyinvitro.And we established rat knee OAmodelsthrough anterior cruciate ligament transection in combination with partial medial meniscectomy (ACLT+pMMx)after 8 weeksand conducted the corresponding Adeno-associated virus (AAV) intraarticular injections. Next, experimental data including immunohistochemical staining,HE/Safranin-O fast green staining,histological evidence analysis,relevant Osteoarthritis Research Society International (OARSI) scoresand expressions of molecules in cartilage anabolismindicated that the relationship betweenERK2 and USP15 throughTGF-β signaling pathway in vivo.Finally, we observed the specic mechanismof USP15 on ERK2 throughco-immunoprecipitation,deubiquitination assay, immunouorescence and nuclear plasma separation. Results.Here, we detected that ERK2 of non-SMAD signaling can maintain the cartilage phenotype by activating the TGF-β/Smad signaling pathway throughincreasing USP15 in vitro and in vivo. Most importantly, USP15 is required during this process and can form a complex with ERK2 to regulate ubiquitination of ERK2.Specically,USP15, instead of USP15 C269S, deubiquitinates and activates ERK2, followed by translocation of phosphorylated ERK2 from the cytoplasm to the nucleus. Correspondingly, the level of phosphorylated ERK2 was increased by injecting AAV-mediated overexpressing USP15 into the rat OA model. Conclusion.Taken together, our study revealed a positive feedback regulation between USP15 and ERK2, which plays a critical role in the interaction of SMAD and non-SMAD signaling pathways to promote cartilage repair in experimental OA. in the OA models followed by real-time quantitative PCR, n=3 for AAV-mediated USP15 overexpression groups in the OA models followed by real-time quantitative PCR). The data were presented as the mean ± SD. (*P<0.05, **P<0.01, ***P<0.001). All experiments were performed at least three times. groups in the sham surgery models followed by real-time quantitative PCR). The data were presented as the mean ± SD. (*P<0.05, **P<0.01, ***P<0.001). All experiments were performed at least three times.


Introduction
Osteoarthritis (OA), the most common chronic arthritis, is a complex and multifactorial disorder characterized by excessive degradation of articular cartilage that results in pain and disability (1,2) . Disruption in the balance between anabolic and catabolic signaling pathways causes extracellular matrix damage to articular cartilage (3,4) . Although many signaling mechanisms contribute to protective of cartilage, the stimulating molecular activities and pathways involved in inhibiting the development of OA remains imperfectly understood.
The transforming growth factor-β (TGF-β) signaling pathway plays an essential role in maintaining tissue homeostasis through anabolic signaling for cartilage growth and repair of cartilage (5) . TGF-β signaling is initiated by speci c type I and II serine/threonine kinase receptors. The speci c receptor-regulated (R)-SMAD proteins, SMAD2 and SMAD3, which form a complex with SMAD4, a co-SMAD, are phosphorylated via the activated TGF-β type I receptor (TbRI). The SMAD complex is then e ciently transferred into the nucleus to regulate the transcription of target genes (6) . Furthermore, the TGF-β signaling pathway can be modi ed by ubiquitination to regulate downstream molecules (7) . Deubiquitinating enzymes (DUBs) can regulate target proteins by removing polyubiquitin chains. Recently, the ability of DUB USP15 to deubiquitinate and stabilize R-SMADs and TGF-β receptor I has gained great importance in our understanding of TGF-β signaling activity (8,9) , which attracted our attention in our current research.
Extracellular signal-regulated kinase (ERK), a member of the MAPK family, has been shown to be responsible for maintaining matrix anabolism and homeostasis of chondrocytes (10) . As described previously, ERK, part of the non-SMAD pathway of TGF-β signaling, modulates the TGF-β1/Smad pathway through enhancing Smad2/3 transcriptional activity (11,12) . Of note, crosstalk mechanisms between ERK and SMAD pathways have been reported in many biological responses (13)(14)(15)(16) . But the speci c associations between ERK and SMAD pathways in OA have not yet been fully elucidated.
In the present study, we report that a positive feedback regulation between USP15 and ERK2 related to the SMAD and non-SMAD pathway could serve as an important quality control mechanism for prevention of cartilage matrix degradation in OA tissues.

Materials And Methods
Collection of human cartilage Human articular cartilage was obtained from the femoral heads of female patients undergoing total hip replacement surgery at the Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, following procedures approved by the Institutional Review Board. Cartilage from female patients with a fracture to the neck of the femur (n = 5, ages 65-95 years) was compared with cartilage from female patients with OA (n = 5, ages 36-87 years). OA was diagnosed according to the The American College of Rheumatology criteria for the classi cation (17) . The patients with fractures had no known history of joint disease, and their cartilage was free of lesions. These human cartilage tissues were used for subsequent histomorphological analysis by Safranin-O fast green staining and immunohistochemistry. All patients signed the informed consent.

Animals
Two-month-old male Norway rats (280-300 g; n = 42) were purchased from Shanghai SLAC Laboratory Animal Co. (Shanghai, China). All rats were randomly assigned to two models: sham surgery models (n = 14) and OA models (n = 28). All animal studies were conducted according to the guidelines of animal care and were approved by the Animal Care Committee of Shanghai Jiao Tong University.
In vivo rat OA model Twenty four 2-month-old male Norway rats were anaesthetized by intraperitoneal injection with 3% pentobarbital sodium until cessation of all sensory re exes. The right knees of rats were routinely disinfected with povidone iodine before surgery. The right knee joint was exposed by incision from the medial side of the patella, and the medial collateral ligament was removed to open the articular cavity. The rats then underwent anterior cruciate ligament transection in combination with partial medial meniscectomy (ACLT + pMMx) on the right knee as described previously (18,19) . Finally, the incisions were sutured and ACLT + pMMx surgeries were induced after complete hemostasis was achieved. In the control group, the sham operations were performed on the model rats. After 8 weeks, OA models were successfully induced in rats that received the ACLT + pMMx surgeries.

Preparation of AAV vectors
Adeno-associated virus (AAV) serotype 2 vectors are AAV plasmids that can encode target genes and enhanced green uorescent protein (EGFP) cDNA under control of the cytomegalovirus (CMV) enhancer.
Cell culture, transient transfections, and lentiviral infections Primary chondrocytes were isolated from rat cartilage tissue as previously described (23) . Brie y, rats were euthanized by cervical dislocation and the hyaline cartilage of the femoral and tibial surfaces was removed with a scalpel. The pared hyaline cartilage was digested after cutting into pieces with hyaluronidase and collagenase II treatment. The isolated cells were cultured with Dulbecco's Modi ed Eagle Medium (DMEM) containing 10% fetal bovine serum (FBS) at 37 °C under a humidi ed atmosphere of 5% CO 2 . To amplify the cultured cells, they were subcultured using 0.25% trypsin/EDTA. The isolated chondrocytes were identi ed by immunostaining for collagen II and Aggrecan.
Transient transfections and lentiviral infections were performed as previously described (24) . HEK293T cells were transiently transfected with pLenti-CMV-mUSP15-FLAG-GFP-Puro, pLenti-CMV-mUSP15C269S- Quantitative real-time RT-PCR Total RNA was isolated from the ATDC5 cells or articular cartilage in rats by using Trizol (TaKaRa, Japan) according to the manufacturer's instructions. RNA was reverse-transcribed to cDNA with Reverse Transcription Kit (TaKaRa, Japan). Relative mRNA amount was quanti ed by Light Cycler 480 system (Roche, Switzerland) with SYBR Green I Kit (TaKaRa, Japan). The mouse primer sequences of the Col2a1, Col10a1, Sox9, Runx2 and GAPDH used were as follows: Col2a1: 5'-TACTGGAGTGACTGGTCCTAAG-3' (Reverse). Relative amounts of mRNA were standardized and calculated as previously described (25,26) .

Co-immunoprecipitation and deubiquitination assay
Monolayer cultured cells were added to ice cold RIPA buffer and centrifuged at 10,000 g for 10 min at 4 °C. The supernatant was transferred to a fresh 1.5 mL conical centrifuge tube on ice with 5 µL of indicated antibodies and 20 µL of resuspended volume of Protein A/G PLUS-Agarose (Santa Cruz Biotechnology), and then incubated overnight at 4 °C on a rocker platform. Co-immunoprecipitates were collected and the supernatant discarded. The pellet was washed 4 × with 1.0 mL RIPA buffer. After the nal wash, samples were boiled for 10 min and analyzed by Western blotting. To detect ERK2 deubiquitination, transient transfections and lentiviral stable infections were performed before coimmunoprecipitation.

Results
USP15 is negatively correlated with cartilage destruction in human cartilage of OA First of all, we performed relevant histomorphological analysis of each human sample by Safranin-O staining in our study. The results showed that compared with normal human cartilage tissue, the cartilage surface of human OA cartilage tissue was rough, uneven and severely damaged. Subsequently, the Safranin-O color of the hyaline cartilage became lighter, the thickness decreased, and its OARSI score statistically signi cant increased (Fig. 1A, C).
From the statistics, OA human cartilage revealed weaker Aggrecan and stronger MMP13 immunostaining than was observed in the normal human cartilage (Fig. 1B, C). Relative to the normal human cartilage, immunohistochemical staining results showed that levels of TGF-β1 and USP15 decreased in human cartilage of OA (Fig. 1B, C). These results inspired us to further study the role of USP15 in the SMAD pathway of OA in the following experiments.

USP15 is conducive to cartilage repair in vitro and in vivo
We assessed the mechanism of USP15 regulating the TGF-β signaling pathway. In ADTC5 cells, ITS induction solution was added to continuously induce cartilage phenotype, and PCR detection of cartilage phenotype related marker molecules was performed at 7, 14 and 21 days respectively. It was con rmed that the expression levels of Col2a1 and Sox9 would peak at 14 days. Col10a1 peaked after 21 days of induction, and Runx2 decreased gradually after 7 days of induction. According to the results of this experiment and relevant literature reports (29,30) , we set 14 days as the optimal cartilage induction time ( Fig. 2A). All the subsequent experiments required ATDC5 cells that had been induced. Overexpression of USP15, rather than USP15 mutant USP15 C269S (enzymatically inactive USP15), led to an increase of phosphorylated SMAD2 (p-SMAD2) (Fig. 2B). In contrast, endogenous USP15 knocked out by Crispr/Cas9 in untreated and TGF-β1-treated ATDC5 cells decreased the level of p-SMAD2 instead of total expression of SMAD2 and SMAD4 (Fig. 2C). Compared with the role of Crispr USP15a, Crispr USP15b has a more obvious in uence on decreasing p-SMAD2, so we selected the sequence of Crispr USP15b for subsequent experiments. Immuno uorescence staining in ATDC5 cells demonstrated that USP15 could increase molecules involved in cartilage anabolic metabolism Col2a1 and Aggrecan concentrations, and the Col10a1 of inhibiting the formation of cartilage decreased (Fig. 2D,E).
To further determine whether USP15 has an inhibitory effect on chondrocyte destruction, we injected rat OA models with AAV-mediated USP15 overexpression in situ. Immunostaining results demonstrated there were higher levels of USP15, p-SMAD2, Col2a1 and Aggrecan in the AAV-mediated USP15 overexpression groups than the control groups of the OA models (Fig. 2F, I). Histological analysis was performed in the HE/Safranin-O fast green staining (Fig. 2G-I). Compared to the control group, the stained Safranin-O color of the cartilage in the USP15 overexpression group was deeper and the articular plane was repaired (Fig. 2H). The thickness of hyaline cartilage was measured and OARSI was scored for each group according to the staining results. It can be concluded that USP15 overexpression can signi cantly increase the thickness of hyaline cartilage, and its OARSI score is much lower than that of the control group (Fig. 2I).To further verify the protective role of USP15 in cartilage, we detected some promoting cartilage anabolic molecules by real-time quantitative PCR in rat cartilage tissue, and the results showed that the overexpression of USP15 increased the expression of Col2a1, Aggrecan and Sox9 (Fig. 2J).
ERK2 can maintain the cartilage phenotype by regulating the TGF-β signaling pathway through USP15 To determine whether ERK2 can regulate USP15 to in uence the TGF-β signaling pathway, we transfected lentivirals to overexpress and knock out ERK2 in ATDC5 cells. When ERK2 was overexpressed, the expression of USP15 increased in untreated and TGF-β1-treated ATDC5 cells. Subsequently, p-SMAD2 was evidently activated in the TGF-β signaling pathway (Fig. 3A). These results also indicated that the levels of USP15 and p-SMAD2 were suppressed with knocked out of ERK2 (Fig. 3C). After treating ATDC5 cells with the extracellular-signal-regulated kinase (ERK)-pathway inhibitor PD98059, the level of p-SMAD2 decreased despite overexpression of ERK2 (Fig. 3B). Furthermore, total expression of SMAD2 and SMAD4 were unchanged after undergoing these treatments (Fig. 3A-C). Finally, immuno uorescence staining results in ATDC5 cell revealed that ERK2 can also maintain the cartilage phenotype and inhibit the destruction factor of cartilage formation by statistic analysis of data (Fig. 3D,E).
Notably, it was reported that ERK2 is crucial for the activity of R-Smads (31) . We also observed that AAVmediated ERK2 knockdown reduced the levels of USP15, p-SMAD2, Col2a1 and Aggrecan in rat sham surgery model knee joints (Fig. 3F). Under the in uence of ERK2 knockdown, the HE/Safranin-O staining thickness of the cartilage in OA was signi cantly reduced and the articular surface became rough (Fig. 3G-I). In the sham surgery models, we then performed real-time quantitative PCR in the control groups and the AAV-mediated ERK2 knockdown groups of rat cartilage to detect some of the molecules that promote anabolic metabolism of cartilage, and the results showed that the knockdown of ERK2 reduced the expression of Col2a1, Aggrecan and Sox9 (Fig. 3J). In vitro and in vivo experiments showed that ERK2 had a certain inhibitory effect on cartilage degradation, and it could enhance TGF-β signaling pathway by upregulation of the expression of USP15, thus affecting cartilage repair.
ERK2 requires USP15 to in uence the TGF-β signaling pathway We transfected lentivirus ERK2 overexpression and USP15 knockout at the same time in ATDC5 cells in vitro. The levels of p-SMAD2 declined with decreasing USP15 despite overexpression of ERK2 (Fig. 4A).
Immuno uorescence staining results further revealed that compare to ERK2, USP15 has more in uence on maintenance of cartilage phenotype such as increasing the expressing of Col2a1 and Aggrecan except for the expressing of Col10a1 (Fig. 4B,C). Taken together, these results con rm that USP15 is required for ERK2 to in uence the TGF-β signaling pathway.
To further evaluate the role of USP15 in the TGF-β signaling pathway, we used injections of AAV-mediated USP15 overexpression and ERK2 knockdown simultaneously in the OA models rat articular cavities. According to the results of immunohistochemistry, when overexpressing USP15 and knocking down ERK2 at the same time, the expression of p-SMAD2, Col2a1 and Aggrecan in the cartilage of the knee joint of rats was relatively increased, indicating that the TGF-β signaling pathway was strengthened (Fig. 4D). In this rescue case, the HE/Safranin-O staining of the cartilage increased, and we can see that the increasing thickness of the hyaline cartilage, the smoother articular surface, and the decreasing loss of chondrocytes (Fig. 4E, F). We then performed a quantitative analysis of hyaline cartilage thickness and OARSI score to further con rm our observation (Fig. 4G). Next the real-time quantitative PCR results showed that when overexpression of USP15 was combined with the knocking down of ERK2, the expression of Col2a1, Aggrecan and Sox9 quantity increased to the inhibition of the destroyed cartilage (Fig. 4H). In the in vivo experiments of rats, when the overexpression of USP15 was combined with the knocking down of ERK2 in the OA model, its effect was nally to inhibit the injury of the knee cartilage of rats. Therefore, the TGF-β signaling pathway was mainly in uenced by the trend of USP15, and in vitro and in vivo experiments showed that USP15 played a full and necessary role in ERK2's transduction of TGF-β signaling pathway to the downstream cartilage phenotype.

USP15 can form a complex with ERK2
We investigated the underlying relationship between USP15 and ERK2 by co-immunoprecipitation assays. Indeed, endogenous USP15 can form a complex with endogenous ERK2 in rat articular chondrocytes (Fig. 5A). The binding interaction was further found between ectopically overexpressed Flag-tagged USP15 and HA-tagged ERK2 by co-immunoprecipitation in 293T cells (Fig. 5B).
To observe the USP15 binding region involved in ERK2 interaction, we used three truncated proteins from USP15 to co-immunoprecipitate with ERK2 (Fig. 5C). The results demonstrated that the N terminus domain of USP15 was bound to ERK2 (Fig. 4C), which was further exhibited by immuno uorescence staining results revealing the colocalization of exogenous ERK2 and USP15 or its mutant in ATDC5 cells (Fig. 5D).
Since USP15 can interact with ERK2, we hypothesized that USP15 deubiquitinates ERK2. First, we coexpressed HA-tagged ERK2 along with His-tagged ubiquitination and found that ERK2 could be ubiquitinated in ATDC5 cells (Fig. 6A). Next, HA-tagged ERK2 and His-tagged Ubiquitination expression vectors were co-expressed with Flag-tagged USP15 wild-type or USP15C269S in ATDC5 cells (Fig. 6B). A loss of ubiquitination of ERK2 was detected by overexpression of USP15 wild-type but not the catalytically inactive USP15 mutant. Conversely, USP15 depletion markedly enhanced the incorporation of ubiquitin into ERK2 (Fig. 6C).
We therefore investigated whether USP15 affects the stability of ERK2. Surprisingly, the total amount of ERK2 was unchanged by treatment with the proteasome inhibitor MG132 (Fig. 6D). When treated with cycloheximide, overexpressed Flag-tagged USP15 could not protect the ERK2 half-life (Fig. 5E). Additionally, knocking out USP15 expression could not in uence the ERK2 concentration (Fig. 6D). However, these data indicated that MG132 induced ERK2 activation and active USP15 could increase phosphorylated ERK2 (p-ERK2) stability and half-life (Fig. 6D,E).
USP15 changes the location of ERK2 and stimulates the p-ERK2 level As previously mentioned, USP15 failed to affect the expression of ERK2. However, we evaluated the subcellular distribution of endogenous ERK2 and p-ERK2 by treatments with overexpressed Flag-tagged USP15 instead of USP15C269S. According to the results of isolating the nuclear and cytoplasmic proteins or immuno uorescence microscopy, overexpressing USP15 instead of its mutant could increase the cytoplasmic signal of endogenous ERK2 and decrease the nuclear amount of endogenous ERK2 (Fig. 7A, B). Importantly, treatment with overexpressed Flag-tagged USP15 activated nuclear p-ERK2 validated by immuno uorescence measurements (Fig. 7C). Furthermore, through the experimental study of immunohistochemistry, we also veri ed once again that the overexpressed USP15 can increase the phosphorylation of ERK2 in the OA models comparing the control groups and the AAV-mediated USP15 overexpression groups (Fig. 7D, E).

Discussion
OA is the most common type of arthritis with in ammatory disease in the synovial joints (32,33) . It is characterized by gradual degeneration of articular cartilage and changes in the structure and function of the entire joint (34,35) . In recent years, it has been reported that the canonical SMAD signaling of the TGF-β signaling pathway contribute signi cantly to cartilage maintenance and repair. To our knowledge, the intricate mechanisms between SMAD and non-SMAD pathways in OA tissues were not yet fully characterized (36)(37)(38) . In our study, we further investigated speci c regulation between USP15 and ERK2 related to the SMAD and non-SMAD pathway to provide an innovative direction to explore the occurrence and treatment of OA.
In order to investigate the expression of TGF-β signaling pathway related molecules in human cartilage, we rst collected normal and osteoarthritic human cartilage following hip replacement. Then, we found that the destruction of the hip joint cartilage with OA was quite serious. The surface of the cartilage was rough with cracks. The thickness of hyaline cartilage was signi cantly reduced compared with the normal sample and the OARSI score was also signi cantly increased, which were similar to previous reports (39) . Secondly, as indicated by the immunohistochemical results, the more severe the cartilage surface destruction, the lower the expression levels of Aggrecan, TGF-β1 and USP15, and on the contrary, the higher the expression levels of MMP13. This result suggests that we need to pay more attention to the signaling activities of TGF-β signaling pathway related USP15 in our study.
From the beginning of the vitro experiment, we found that endogenous USP15 could bind to ERK1 in rat chondrocytes through co-immunoprecipitation (Supplement Fig. 1A). Therefore, we started to verify the relationship between them. But the western blot results showed that the expression level of USP15 was unchanged when ERK1 knocked down in ATDC5 cells (Supplement Fig. 1B). Then, we observed that the amount of USP15 was constant by knocking out or overexpressing ERK1 in ATDC5 cells (Supplement Fig. 1C, D). De nitely, we chose to study the relationship between USP15 and ERK2 deeply in the followup experiments.
It was reported that USP15 could be regarded as a responding marker of the TGF-β signaling pathway (40) . Consistent with the in vivo measurement indications of AAV-mediated USP15 overexpression, we observed that USP15 stimulated the TGF-β signaling pathway such as p-SMAD2, p-ERK2 and caused chondrocyte remodeling with increasing Col2a1, Aggrecan, and Sox9 levels. Histological assessments further revealed that USP15 was bene cial to developments of cartilage repair. Our ndings with ATDC5 cells further demonstrated that only activated USP15, and not USP15C269S, is important for enhancing SMAD signaling responses.
ERK2 of non-SMAD pathways is known to regulate the TGF-β signaling pathway through SMADs (41) . As expected, in vitro data suggested that ERK2 overexpression can increase the level of p-ERK2 and activate p-SMAD2 by enhancing USP15 concentrations. Even though ERK2 was overexpressed, the amount of p-SMAD2 was still decreased with p-ERK2 inhibited. This result demostrated that overexpressed ERK2 also can stimulate p-SMAD2 directly through increasing p-ERK2. Similarly, knocking down ERK2 in the sham surgery models could weaken the TGF-β signaling pathway followed by cartilage damage. Hence, elevating ERK2 levels suppressed cartilage damage through the TGF-β signaling pathway. To address whether USP15 is required between ERK2 and p-SMAD2, we performed rescue experiments in both rat models and ATDC5 cells. Taken as a whole, although ERK2 was knocked down, p-SMAD2 expression was increased due to upregulation of USP15 expression. Eventually, the expression levels of molecules such as Col2a1,Aggrecan and Sox9 that promoting cartilage repair elevated.
Co-immunocoprecipitation assays in our current studies showed that USP15 could bind to ERK2. Because USP15 is an indispensable deubiquitinating enzyme for the TGF-β signaling pathway, we speculated that USP15 could deubiquitinate ERK2. To verify this deubiquitination modi cation, we determined that ERK2 could be deubiquitinated. In agreement with our prediction, we analyzed the speci c deubiquitination effects of USP15 towards ERK2. If USP15 was knocked out, ERK2 would undergo more ubiquitination modi cations. The most common role in ubiquitination is protein degradation (42)(43)(44) Surprisingly, even though USP15 could deubiquitinate ERK2, the degradation of endogenous ERK2 was not changed by USP15. Next, we identi ed that the total ERK2 levels were unchanged when treated with the proteasome inhibitor MG132. Interestingly, overexpressed USP15 could stabilize p-ERK2 with or without MG132 treatment. From the half-life experiment with cycloheximide, we observed that overexpressed Flag-tagged USP15 could prolong the half-life of p-ERK2. To some extent, the degradation of p-ERK2 was prevented by USP15. Otherwise, USP15 instead of USP15C269S, can cause an increase in the distribution of ERK2 in the cytoplasm and a decrease in the distribution of ERK2 in the nucleus. Then, USP15 activates ERK2, and p-ERK2 from the cytoplasm is translocated to the nucleus. In vivo, the level of p-ERK2 increased by AAV-mediated upregulated USP15 in rat OA model knee joints. Finally, a working model for positive feedback regulation between USP15 and ERK2 that enhanced the intracellular responses to the TGF-β signaling pathway, which may present potential therapeutic value for OA treatment was described (Fig. 8).
The limitations of this study included that there were only rat OA models instead of human clinical samples to explore deeper mechanisms in the further study. What's more, primary chondrocytes isolated from rat cartilage tissue were replaced by ATDC5 cells in the part of vitro study because of the di cult lentiviral infections in them.
In conclusion, the present study suggests that positive feedback regulation between USP15 and ERK2 plays a critical role in the interactions of SMAD and non-SMAD signaling pathways. However, the cascade of interconnected responses provides new insights into the pathogenesis of OA. Additionally, the positive feedback loop should be considered a powerful therapeutic target for the clinical treatment for OA.

Declarations
Ethics approval and consent to participate: The human and animal study were reviewed and approved by Animal CareCommittee of Shanghai Jiao Tong University.     in each group were calculated, and the data were presented as the mean ± SD. (*P<0.05, **P<0.01, ***P<0.001). (D-F) The expression levels of p-SMAD2, Col2a1 and Aggrecanin the two groups were detected by immunohistochemistry, and the cartilage tissue morphology in the two groups were detected by HE/Safranin-O fast green staining. (n=4 for control groups in the OA models, n=4 for AAV-mediated USP15 overexpression and ERK2 knock-down groups in the OA models). D: (Scale bars=50um). E: (Scale bars=50um). F:(Scale bars=200um). (G) The positive cells number of immunohistochemistry of p-SMAD2, Col2a1 and Aggrecan in each group were calculated, and the hyaline cartilage thickness and histological analysis of OARSI scores were quanti ed. The data were presented as the mean ± SD.
(*P<0.05, **P<0.01, ***P<0.001). (H). Relative gene expressions of Col2a1, Aggrecan and Sox9 associated with cartilage anabolic metabolism in two groups were detected by real-time quantitative PCR. (n=3 for control groups in the OA models followed by real-time quantitative PCR, n=3 for AAV-mediated USP15 overexpression and ERK2 knock-down groups in the OA models followed by real-time quantitative PCR).