The approval for performing the project was provided by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University. Cartilage tissue specimens were collected from 21 OA patients undergoing total knee arthroplasty and 17 healthy traumatic amputees at the First Affiliated Hospital of Zhengzhou University, and the written informed consent from these subjects was obtained before enrolling.

Human chondrocyte cell line CHON-001 cells (CRL-2846, ATCC, Manassas, VA, United States) were cultured in DMEM (30–2002, ATCC) at 37°C, 5% CO₂ with 0.1 mg/mL G-418, and 10% fetal bovine serum (FBS; PAN Biotech, Aidenbach, Germany). In addition, IL-1β (Sigma-Aldrich, St. Louis, MO, United States) at different doses (0, 5, 10, and 15 ng/ml) was used to stimulate CHON-001 cells for 24 h at 37°C to mimic the OA chondrocyte model in vitro as previously described (Cheleschi et al., 2018; Wang et al., 2019). 10 ng/ml IL-1β treatment was selected for further investigation.

After multiple treatments, the CHON-001 cell culture medium was harvested and the levels of IL-6 and tumor necrosis factor α (TNF-α) were subsequently measured, according to the operation manual of specific ELISA kits (R&D Systems, Minneapolis, MN, United States).

For the Counting Kit (CCK-8) assay, un-treated or treated CHON-001 cells (approximately 2000 cells/well) were plated into 96-well plates at 37°C conditions containing 5% CO₂. After 24 h of culture, 10 μL CCK-8 reagent (Dojindo, Kumamoto, Japan) was mixed for 4 h. At indicated time points, the determination of absorbance was conducted using a microplate reader at 450 nm. For the 5-ethynyl-2′-deoxyuridine (EdU) assay, 2×10⁵ cells were introduced into 24-well plates, followed by a mixture with 50 μM EdU (Solarbio, Beijing, China) for 2 h. After being fixed in a 4% formaldehyde solution, cells were reacted with the Apollo reaction cocktail and DAPI at 37°C in the dark, followed by visualization under a fluorescence microscope.

In general, CHON-001 cells were harvested and 1×10⁶ cell suspension was prepared. After being washed and digested, treated cells were re-suspended in 100 μl binding buffer and incubated with 5 μl Annexin V-FITC and 10 μl PI (Bender Med System, Vienna, Austria) at room temperature away from light. 15 min later, cell apoptosis was sorted and analyzed with the help of a flow cytometer.

Using a proteinase inhibitor mixture in RIPA buffer (Beyotime, Shanghai, China), lysis of tissue samples and CHON-001 cells was prepared. Subsequently, an equal amount of protein was separated by 10% SDS-PAGE and transferred onto nitrocellulose membranes (Millipore, Molsheim, France), which were then blocked with 5% non-fat milk. After hybridization with the primary antibodies: Bax (ab53154), Cleaved-caspase-3 (1:1000, ab2302), MMP13 (1:1000, ab39012), Aggrecan (1:1000, ab36861), ADAMTS5 (1:1000, ab45047), and GAPDH (1:1000, ab9485) all night at 4°C, the membranes were incubated with secondary antibody (1:10000, ab205718) for 1 h at 37°C. At last, the protein blots were subjected to the enhanced chemiluminescence (ECL) reagent (Millipore) for visualization.

Generally, RNA extraction in clinical specimens or CHON-001 cells was accomplished using TRIzol (Invitrogen, Paisley, Scotland, United Kingdom), and their 260/280 value was distributed from 1.8 to 2.0. Furthermore, all RNAs were treated with 3 U/μg of RNase R (Epicentre, Madison, WI, United States). For circRNA and mRNA, reverse transcription was carried out according to the Bio-Rad iScript kit (Bio-Rad Laboratories, Hercules, CA, United States), followed by RT-qPCR using iQSYBR Green SuperMix (Bio-Rad). For miRNA detection, All-in-One™ miRNA First-Strand cDNA Synthesis kit and the All-in-One™ miRNA qPCR kit (GeneCopoiea, Rockville, MD, United States) were applied, referring to the operation manual. Samples were normalized to GAPDH and U6, and the 2^–ΔΔCt method analyzed the relative expression. The primers were listed in Table 1.

Small interfering RNAs specific to circ_0037658 (si-circ_0037658), the inhibitors and mimic of miR-665 (miR-665 and anti-miR-665), and their negative controls (si-NC, miR-NC, and anti-miR-NC) were obtained from Ribobio (Guangzhou, China). Meanwhile, overexpression vectors pCD5-circ_0037658 (circ_0037658) and ADAMTS5, and their controls (pCD5-ciR and pcDNA), were purchased from Geneseed (Guangzhou, China). Subsequently, CHON-001 cells at 60% confluence were transfected with the above transfection agents using Lipofectamine 3000 reagent (Invitrogen) for 48 h. After that, the transfection efficiency of the harvested CHON-001 cells was detected using RT-qPCR or western blot, followed by treatment with IL-1β.

Using Starbase software, the prediction of the binding sequences was achieved. Sequence fragment of circ_0037658 and ADAMTS53′ untranslated region (3′UTR) containing miR-665-matched binding sites or miss-matched target sites was designed and inserted into the pGLO reporter vector (Promega, Fitchburg, WI, United States) by Hanbio (Shanghai, China), generating WT/MUT-circ_0037658 and WT/MUT-ADAMTS5 3′UTR reporter vectors. Co-transfection of CHON-001 cells was conducted with the generated vectors and miR-665 or miR-NC for 48 h, followed by the analysis of luciferase activities in cell lysates was performed using a dual-luciferase reporter assay kit (Promega).

In this assay, a commercial Magna RNA immunoprecipitation kit (Millipore) was applied to verify the interaction between miR-665 and circ_0037658 or ADAMTS5 in CHON-001 cells. In short, CHON-001 cells at 80% confluency were incubated with complete RIP lysis buffer. Then, cell lysates were incubated with anti-Argonaute2 (Ago2, ab186733, 1:50, Abcam) or immunoglobulin G (IgG, ab172730, 1:100, Abcam) for 4 h at 4°C before treating magnetic protein A/G beads for 2 h. After being digested with proteinase K, the RNAs on the beads were retrieved and analyzed using RT-qPCR.

The collected data were exhibited as mean ± standard deviation (SD) and processed according to GraphPad Prism8 software. Data differences were compared according to Student’s t-test or one-way analysis of variance (ANOVA) with Tukey’s tests. The expression association was assessed using Pearson correlation analysis. p-value < 0.05 was deemed as a statistically significant difference.

An in vitro model of OA was established in human articular chondrocyte (CHON-001 cells) treated with IL-1β. Data displayed that the secretions of pro-inflammatory cytokines in cells were highly induced in a dose-dependent manner upon IL-1β (Figure 1A). Subsequently, cell proliferation ability gradually declined after treatment of 0–15 ng/ml IL-1β for 24 h (Figures 1B,C). On the contrary, the cell apoptosis rate was significantly elevated in IL-1β-administered chondrocytes (Figure 1D). Consistently, the western blot assay presented a significant enhancement in Bax and Cleaved-caspase-3 of treated cells (Figure 1E). Furthermore, previous studies have confirmed that MMP13 upregulation (a key cartilage-degrading enzyme) and Aggrecan downregulation (a main proteoglycan in the articular cartilage) are critical events in the early stage of OA (Burrage et al., 2006; Troeberg and Nagase, 2012; Haller et al., 2015). As expected, the western blot assay verified that the treatment of IL-1β led to a significant increase in MMP13 and a substantial decrease in Aggrecan in CHON-001 cells (Figure 1F). Together, IL-1β treatment might induce cell injury in OA, and 10 ng/ml IL-1β treatment for 24 h was chosen to mimic chondrocyte injury for further research.

According to the data shown in Figure 2A, the circ_0037658 level was significantly increased in 21 OA cartilage tissues versus 17 healthy cartilage tissues using the RT-qPCR assay. Consistently, our data verified that circ_0037658 was highly expressed in IL-1β-stimulated CHON-001 cells in a dose-dependent manner (Figure 2B). Whereafter, the circ_0037658 was resistant to RNase R in CHON-001 cells, and linear GAPDH was reduced (Figure 2C). Subsequently, the transfection efficiency of si-circ_0037658 was verified in Figure 2D. After that, the results of ELISA presented that IL-1β exposure elicited a significant enhancement in IL-6 and TNF-α levels in chondrocytes, which was partly relieved by the introduction of si-circ_0037658 (Figure 2E). Moreover, CCK-8 assay and EdU assay exhibited that IL-1β treatment might significantly inhibit cell proliferative ability, while the silence of circ_0037658 abrogated these effects (Figures 2F,G). In parallel, circ_0037658 downregulation strikingly attenuated the promotion of IL-1β on cell apoptosis (Figures 2H,I), as evidenced by decreased Bax and Cleaved-caspase-3 (Figure 2J). In terms of ECM degradation, IL-1β-mediated increase in MMP13 and decrease in Aggrecan were ameliorated by circ_0037658 deficiency in chondrocytes (Figure 2K). Collectively, circ_0037658’s absence might partly reverse IL-1β-induced chondrocyte injury.

To further understand the potential mechanism of circ_0037658 regulating OA development, we further searched for circ_0037658-interacting miRNAs using the online software Starbase. Data suggested that five potential target miRNAs of circ_0037658 from Starbase were found. Then, all these mRNAs in CHON-001 cells were subjected to RT-qPCR analysis responding to circ_0037658 downregulation. Among these five miRNAs, miR-665 displayed the highest fold change (Supplementary Figure S1). Therefore, we chose miR-665 for further research. Furthermore, some binding sites between circ_0037658 and miR-665 were exhibited in Figure 3A. Meanwhile, miR-665 expression was significantly improved in miR-665 mimic-transfected cells (Figure 3B). Then, data presented only combined miR-665 mimic and WT-circ_0037658 transfection reduced luciferase activity in chondrocytes (Figure 3C). Synchronously, circ_0037658 and miR-665 could be abundantly precipitated together by the Ago2 antibody compared with the IgG antibody using RIP assay (Figure 3D), further verifying the combination between circ_0037658 and miR-665. Of interest, the RT-qPCR assay presented a significant decrease of miR-665 in 21 OA cartilage tissues compared with 17 healthy cartilage tissues (Figure 3E). Moreover, miR-665 content was gradually reduced in IL-1β-treated CHON-001 cells (Figure 3F). According to Pearson correlation analysis, circ_0037658 was negatively associated with miR-665 in OA cartilage tissue samples (Figure 3G). Moreover, the RT-qPCR assay displayed that the circ_0037658 level was significantly increased in IL-1β-triggered CHON-001 cells after the transfection of circ_0037658 (Figure 3H), suggesting that the pCD5-circ_0037658 overexpressing vector was successfully transfected into IL-1β-administered CHON-001 cells. Then, the introduction of si-circ_0037658 enhanced the miR-665 level in IL-1β-induced CHON-001 cells, and pCD5-circ_0037658 exhibited the opposite results (Figure 3I). Overall, circ_0037658 interacted with miR-665.

Then, we further explored the influence of circ_0037658 and miR-665 on IL-1β-induced cell damage. After diverse transfection, we found that the co-transfection of anti-miR-665 might significantly counteract circ_0037658 downregulation-induced miR-665 level promotion in IL-1β-treated CHON-001 cells (Figure 4A). Functionally, circ_0037658 silencing might significantly block the secretion of these pro-inflammatory cytokines in the treated CHON-001 cells, while miR-665 downregulation could effectively abolish these effects (Figure 4B). In contrast, the miR-665 inhibition partly overturned the positive impact of circ_0037658 silencing on cell proliferative ability in IL-1β-exposed CHON-001 cells (Figures 4C,D). In addition, circ_0037658 silencing repressed CHON-001 cell apoptosis rate, but anti-miR-665 ameliorated these effects in treated cells (Figure 4E), accompanied by higher Bax and Cleaved-caspase-3 levels (Figure 4F). Furthermore, the si-circ_0037658-mediated MMP13 reduction and Aggrecan increase were attenuated by miR-665 inhibitor in IL-1β-treated CHON-001 cells, meaning that miR-665 knockdown might abolish the repression effect of circ_0037658 deletion on ECM degradation of CHON-001 cells (Figure 4G). All in all, miR-665’s absence partly weakened the suppressive role of si-circ_0037658 on IL-1β-triggered CHON-001 cell injury.

A predicted miR-665-binding site was found in the 3′UTR of ADAMTS5 (Figure 5A). Then, miR-665 overexpression significantly decreased the luciferase activity of the WT-ADAMTS5 3′UTR reporter vector but not that of the MUT-ADAMTS5 3′UTR reporter vector in CHON-001 cells (Figure 5B). Consistently, miR-665 and ADAMTS5 levels were enriched in the Ago2 immunoprecipitates using the RIP assay (Figure 5C). Interestingly, inversely associated with miR-665 expression (Figure 5E), ADAMTS5 was raised in OA cartilage tissues (Figures 5D,F). As we expected, the ADAMTS5 protein level was gradually enhanced in CHON-001 cells after the treatment of 0–15 ng/ml IL-1β (Figure 5G). Furthermore, the overexpression or knockdown efficiency of miR-665 in CHON-001 cells was measured using RT-qPCR assay (Figure 5H). Then, the western blot assay displayed that miR-665 upregulation significantly reduced ADAMTS5 expression in CHON-001 cells, whereas the miR-665 inhibitor presented an opposite result (Figure 5I). Overall, ADAMTS5 acts as a direct downstream target gene of miR-665.

Then, to elucidate their correlation cell model of OA, we conducted rescue assays. As displayed in Figure 6A, miR-665 overexpression significantly declined ADAMTS5 protein level in CHON-001 cells, while these effects were partly counteracted after co-transfection with pcDNA-ADAMTS5. Subsequently, ELISA exhibited that ADAMTS5 upregulation might overturn miR-665 mimic-mediated IL-6 and TNF-α levels inhibition in IL-1β-induced CHON-001 cells (Figure 6B). Meanwhile, the overexpression of miR-665 significantly elevated cell proliferative ability, whereas pcDNA-ADAMTS5 partly reversed this impact in CHON-001 cells (Figures 6C,D). Simultaneously, the upregulation of ADAMTS5 in IL-1β-stimulated CHON-001 cells effectively ameliorated the repression of miR-665 overexpression on cell apoptosis rate (Figure 6E), which was further confirmed by Bax and Cleaved-caspase-3 expression (Figure 6F). For ECM degradation, the re-introduction of pcDNA-ADAMTS5 might partly relieve the miR-665-triggered decline in MMP13 and increase in Aggrecan in CHON-001 cells (Figure 6G). Then, western blot indicated that the deficiency of circ_0037658 might dampen ADAMTS5 protein expression in CHON-001 cells, and anti-miR-665 might significantly attenuate the impact (Figure 6H), suggesting the circ_0037658/miR-665/ADAMTS5 axis. In summary, miR-665 might abate IL-1β-triggered cell damage by targeting ADAMTS5.