Single-cell RNA sequencing data analysis

We analyzed public-available single-cell RNA-seq data from ImmPort (study SDY998) featuring articular synovial T cells, B cells, monocytes, and fibroblasts from 3 OA and 18 RA patients, sequenced using the CEL-Seq2 method [22]. Following UMI-based gene quantification, log2(CPM) transformation, and quality control, 5265 cells and 32391 genes were used for further analysis.

miRNA sequencing data analysis

We analyzed public-available RNA-seq of extracellular vesicles, isolated from SF of RA patients with high (n = 7)- or low (n = 5)-grade inflammation classified by leukocyte counts from Foers et al. [28]. Differentially expressed miRNAs were defined with adjusted P-value < 0.05 and log2FC  > 1.2 or  < − 1.2.

miRNA target prediction

miRNA targeting HTR2A was predicted by RNAhybrid 2.2 [29] and TargetScanHuman 8.0 [30]. Overlapping results of RNAhybrid, conserved sites in results of TargetScan and miRNAs lower expressed in high-inflammation RA, we identified differentially expressed miRNAs targeting HTR2A in RA.

Patient material

Synovial fluid and tissue were obtained from RA and OA patients with informed consent and Renji Hospital Human Research Ethics Committee approval (identification no. 2013-126). Cells and supernatant of SF were frozen at -80℃ until required. Half of each synovial tissue piece was diced, digested with 100U/ml liberase TL and 100U/ml DNase I (Roche, Switzerland) at 37 °C for 15 min, and then passed through filter. ST cells were preserved in liquid nitrogen for future experiments. The other half of each synovial tissue piece was fixed for Immunohistochemistry (IHC) staining.

Immunohistochemistry staining and analysis

Paraffin-embedded tissue sections (5 μm) were prepared using RM2125 (Leica, Germany). After heating at 55 °C for 30 min, sections were deparaffinized with xylene, rehydrated with ethanol, and antigen retrieval at 95 °C for 20 min. Endogenous peroxide was removed with 3% H2O2 for 15 min. Slides were blocked and then stained with primary antibodies, anti-human HTR2A (Invitrogen, USA), IL-6 and CXCL12 (Proteintech, USA) overnight at 4 °C. Subsequent steps included secondary antibody staining, HRP treatment, DAB staining, and hematoxylin counterstaining.

Slides were digitally scanned (Servicebio, China) and quantified by QuPath-0.4.2 software. Fibroblast identification criteria were Nucleus Perimeter  > 19 and Nucleus Circularity  < 0.8; marker positivity was determined according to Cytoplasm DAB OD max value,  > 0.6 for HTR2A,  > 0.4 for IL-6,  > 0.7 for MMP13,  > 0.55 for MMP14 and  > 0.7 for CXCL12.

Exosome isolation and identification

Exosome isolation was performed using differential ultracentrifugation method as described [31]. The concentration and size distribution of exosomes were identified by Nanoparticle Tracking Analysis (NTA) method via NanoFCM, China. Morphology identification of exosomes was performed with Transmission Electron Microscopy (TEM). Exosome-specific markers (CD9, CD81, CD63) were verified by western blot.

Western blot

EVs or cells were resuspended in lysis buffer. Protein concentration was detected by BCA Kit (Beyotime BioTech, China). Purified proteins were separated in 10%SDS-PAGE and transferred to PVDF membrane at 350 mA for 50 min. After washing and blocking, the membrane was incubated overnight with human primary antibodies CD9, CD81, CD63, and Calnexin (1000 × dilution, Beyotime BioTech, China), HTR2A (1000 × dilution, Abclonal, China), GAPDH (2000 × dilution, Abclonal, China). Secondary antibodies were applied for 120 min, and images were acquired using a chemiluminescence detection system (Beyotime BioTech, China).

Fibroblast, T cell and monocyte isolation

Single-cell suspension of synovial tissue was purified into fibroblasts after 4 passages in one month. Single-cell suspension of SF was stained with FITC-TCRb and percp5.5-CD14 (Biolegend, USA) at dark 4 °C for 10–15 min. After washing and centrifugation, cell pellets were resuspended with propidium iodide/PBS buffer (Biolegend, USA) and sorted into T cells and monocytes by BD FACSAria™ III. Single-cell suspension of each subset was pooled from 6 OA patients.

miRNA extraction and quantification

miRNA extraction used the miRcute Serum/Plasma miRNA Isolation Kit (TIANGEN, China) as instructed. Briefly, 900 ul lysis buffer was mixed with each 200 ul EV-biofluid or each cell pellet thoroughly. 1 pmol cel-miR-39 (RIBOBIO, China) per 200 ul EV-biofluid or 10^5 cells was added as an exogenous control. The mixture was placed at room temperature for 5 min to separate nucleic acids and protein. RNA was separated into aqueous phase by 200 ul of chloroform and then transferred into twice the volume of absolute ethanol. After passing the obtained liquid through the Column miRelute, 700 ul of MRD buffer and 500 ul of RW buffer were added into the Column miRelute for RNA purification. Lastly, RNA was eluted in 20 ul of RNase-Free ddH2O.

MiRNAs were reverse transcribed and the expression levels were measured by quantitative real-time polymerase chain reaction (qPCR) on a QuantStudio 7 Flex instrument (Applied Biosystems). MiRNA PCR was performed with the Bulge-Loop miRNA qRT-PCR Starter Kit, Bulge-Loop hsa-miR-23b-3p Primer Set, Bulge-Loop hsa-miR-23b-5p Primer Set, Bulge-Loop hsa-miR-214-3p Primer Set, Bulge-Loop hsa-miR-3120-5p Primer Set, Bulge-Loop hsa-miR-615-3p Primer Set and Bulge-Loop cel-miR-39 Primer Set (RIOBO, China) according to manufacturer’s instructions. cDNA synthesis reaction contained 1 μL of total RNA, 0.5 μL of Bulge-LoopTM miRNA RT Primer (5 μM), 1 μL of RTase Mix, 1 μL 5 × Reverse Transcription Buffer, 1.5 μL of RNase-free H2O. The reaction mixture was incubated at 42 °C for 60 min, followed by 70 °C for 10 min. The synthesized cDNA was stored at  − 20 °C until further use. qPCR reactions were performed in a total volume of 10 μL, containing 5 μL of 2 × SYBR Green Mix, 1 μL of template DNA, 0.4 μL of Bulge-LoopTM miRNA Forward Primer (5 μM) and 0.4 μL of Bulge-LoopTM Reverese Primer (5 μM) and 3.2 μL of RNase-free H2O. The qPCR cycling conditions consisted of an initial denaturation step at 95 °C for 10 min, followed by 40 cycles of denaturation at 95 °C for 2 s, annealing at 60 °C for 20 s, and extension at 70 °C for 10 s. A final extension step was performed at 72 °C for 5 min. Assays set up in 96-well PCR plates (Integrated Sciences, Australia) included 3 duplicates of each sample.

For miRNAs extracted from SF-EVs, data was normalized by the initial synovial fluid volume which was used for EV isolation, and was calibrated by the cel-miR-39 to eliminate the bias caused by different synovial fluid volume, RNA isolation efficiencies and PCR efficiencies among samples. For miRNAs extracted from cells, data was normalized by exogenous control. Gene expression was relatively quantified with 2−ΔΔCt method.

miRNA transfection

RA fibroblast cell line (PriMed-Icell-003, icell bioscience, China) at 60–80% confluence was transfected. Lipofectamine® RNAiMAX Reagent (Thermo Fisher Scientific, USA) was diluted in Opti-MEM® Medium (Gibco, USA) and mixed with miRNA mimics (RIBOBIO, China) in Opti-MEM® Medium at a 1:1 ratio. After 5 min at room temperature, the miRNA-lipid complex was added into 6 wells plate to incubate cells at 37℃ for 2–3 days. miRNAs mimics include hsa-miR-23b-3p, hsa-miR-23b-5p, hsa-miR-214-3p, hsa-miR-3120-5p, hsa-miR-615-3p. Cel-miR-39 acts as a negative control to check the off-target effect. Also, blank control was designed to assess any non-specific effects or variations in gene expression resulting from factors such as transfection reagents, culture conditions, or sample processing techniques.

Statistics

Analyses were performed using R 4.2.0 and GraphPad Prism 8.0. Data was analyzed by Student t-test, Mann–Whitney test, Chi-square test, Fisher’s exact test and Spearman’s correlation when applicable. Differences were considered significant if p < 0.05 (indicated as * for p < 0.05, ** for p < 0.01, *** for p < 0.001 and **** for p < 0.0001 when measurement data, and * for p < 0.05 when enumeration data).