Human BMSCs were kindly provided by the Stem Cell Bank, Chinese Academy of Sciences, harvested from healthy people and cultured in Dulbecco’s minimum essential medium (DMEM; BasalMedia, China). Mandibular condylar chondrocytes (MCCs) were extracted from the condyles of the temporomandibular joints (TMJs) of New Zealand rabbits and cultured in DMEM. All media were supplemented with 10% fetal bovine serum (FBS, Yeasen, China) and 1% penicillin-streptomycin (BasalMedia, China), and all cells were cultured in a humidified incubator (Thermo Fisher Scientific, United States) with 5% CO₂ at 37°C.

First, sEVs were isolated from the medium supernatant of BMSCs without FBS or penicillin-streptomycin, which was harvested after 48 h of culture. After centrifugation (4,000 rpm, 10 min), the supernatant was sequentially minifilter through polyvinylidene fluoride (PVDF) membrane filters at 450 nm (Merck Millipore, Germany) and then 200 nm (PALL, United States). Then, the flow-through was ultrafiltered by 100-kD centrifugal filter devices (Merck Millipore, Germany). A 15-ml device is usually used, and a 0.5-ml device can be used to further concentrate the sEVs, which can then be harvested by reverse centrifugation. Finally, sEVs isolated from approximately 80-ml medium supernatant were suspended in 1 × phosphate buffer saline (PBS) buffer or culture medium through the use of 0.5-ml centrifugal filter devices (Merck Millipore, Germany) for further experiments.

For particle size and concentration determination, nanoparticle tracking analysis (NTA) was performed with a NanoSight NS300 (Malvern, United Kingdom) equipped with fast video capture and NTA analytical software. Nanoparticles were illuminated by the laser, and their movements under Brownian motion were captured for 60 s. Videos were analyzed by the software to provide the nanoparticle concentration and size distribution profiles.

The BMSC-sEV samples were added onto a piece of copper grid for 1 min. Then, a drop of 2% uranyl acetate was added onto the copper grid for 1 min. After drying for 10 min, the cells were examined under a transmission electron microscope (Thermo Fisher Scientific, United States).

BMSC-sEVs and cells were lysed with radioimmunoprecipitation assay (RIPA) lysis buffer (Yeasen, China) containing protease inhibitors. The lysates were boiled with Protein SDS-PAGE Loading Buffer (GenScript, China), electrophoresed through 4–20% polyacrylamide gels (GenScript, China) and transferred onto 0.45-μm PVDF membranes (Absin, China). The membranes were blocked using 5% skim milk (Yeasen, China) in PBS. Antibodies against the following proteins were used for immunoblotting (IB) analysis: CD81 (SBI, Japan), CD63 (SBI, Japan), Ras-related protein 5 (Rab5; Biovision, United States), ALG2-interacting protein (Alix; Thermo Fisher Scientific, United States), glucose-regulated protein 94 (GRP94; Thermo Fisher Scientific, United States), aggrecan (ACAN; Bioss, China), SRY-related high-mobility group box 9 (SOX9; Bioss, China), matrix metalloproteinase 13 (MMP13; Bioss, China), RUNX family transcription factor 2 (RUNX2; Bioss, China), Collagen I (Bioss, China), proliferating cell nuclear antigen (PCNA; Proteintech, United States), cartilage-forming factors-type II collagen (Col-II; Novus, United States), Autotaxin (Abcam, China), Yes-associated protein (YAP; Absinthe, China), phosphorylated Yes-associated protein (p-YAP; LifeSpan BioSciences, United States), RhoA (Absinthe, China), large tumor suppressor kinase 1 (LATS1; Absinthe, China), large tumor suppressor kinase 2 (LATS2; Absinthe, China), and glyceraldehyde 3-phosphate dehydrogenase (GAPDH; Proteintech, United States). Horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG and anti-mouse IgG (Proteintech, United States) were used as secondary antibodies. Detection was performed using Chemiluminescent HRP Substrate (Merck Millipore, Germany), and signals were captured and observed using Molecular Imager^® ChemiDoc^TM XRS + (Bio-Rad, United States).

By using chemical method modeling-collagenase injection, a TMJOA animal model was established. The posterior pole of the TMJ condyle can be found behind the outer canthus of 12- to 18-week-old New Zealand rabbits. In its mouth opening position, with left forefinger tip pressing on the area of joint space, the syringe needle held in the right hand was thrust into the joint space inward, forward, and downward, parallel to the infraorbital margin. If the needle entered approximately 0.5 cm and was withdrawn without blood, approximately 0.25 ml of 4 mg/ml collagenase II (Yeasen, China) was injected into the TMJ, and this group was named OA group (the control group was injected with approximately 0.25 ml normal saline). 4 weeks later, the TMJOA rabbit model was identified by morphological evaluation and histological and molecular biological examination. All protocol and procedures employed in vivo were ethically reviewed and approved by the Institutional Animal Care and Use Committee at Zhejiang Laboratory Animal Center (Approval No. ZJCLA-IACUC-20050012) for the rational care and use of laboratory animals.

Under sterile conditions, the condyles of 12- to 18-week-old New Zealand rabbits were excised, and surrounding soft tissues were removed. Cartilage tissues on the condyles were separated and cut into 1-mm³ size; and after rinsing with PBS, 0.25% trypsin was added and treated for 30 min. Then, high-glucose DMEM containing 10% FBS was added to terminate dissociation. After discarding supernatant, 2 ml 0.3% collagenase (Yeasen, China) and 4 ml high-glucose DMEM containing 10% FBS were added and placed in 37°C. After 12 h, the MCCs were collected and cultured in six-well plates. When MCCs grew to 80–90% confluency, 10 ng/ml recombinant human interleukin-1 beta (IL-1β; Novoprotein, China) was added into the MCC medium and cultured in 37°C for 24 h, and then identification of TMJOA cell model, such as expression of OA-related factors, was launched.

In vivo, normal and TMJOA model rabbits were randomly assigned into control groups and experimental groups, separately, and there were 6–8 TMJs of rabbits (i.e., 3–4 rabbits) in each group (one rabbit died in the OA group, which means that there were six TMJs of rabbits in that OA group), half male and half female. In the first in vivo experiment, the animals were divided into four groups: control group (rabbits in the former control group treated with PBS), OA group (rabbits in the former OA group treated with PBS), sEV group (rabbits in the former control group treated with sEVs), and OA^sEV group (rabbits in the former OA group treated with sEVs). This experiment was launched three times, for 4-week, 6–week, and 8-week observation, separately. In the second in vivo experiment, the animals were divided into four groups: OA group (rabbits in the former OA group treated with PBS), OA + sEV group (rabbits in the former OA group treated with sEV), OA + sEV^–/–ATX group (rabbits in the former OA group treated with sEV^–/–ATX), and OA + sEV^Ziritaxestat group (rabbits in the former OA group treated with sEV^Ziritaxestat). This experiment was launched three times, for 4 weeks, 6 weeks, and 8 weeks of observation, separately. In each in vivo experiment, approximately 200 μl (4–8) × 10⁸ BMSC-sEVs extracted from about 5.5 × 10⁷ BMSCs were injected into each TMJ in the experimental groups, once per side per capita. In vitro, approximately (2–4) × 10⁸ BMSC-sEVs aforementioned were used in the experimental groups (the control group was added with normal saline).

Condyle specimens were scanned and analyzed by the high-resolution micro-computed tomography (micro-CT) scanner U-CT-XUHR (Milabs, Netherlands) and Imalytics Preclinical 2.1 software. Scanning parameters were 0.24 μA current, 50 kV voltage, 15-ms exposure time, and 30 deg/s angle speed. Subchondral cancellous bone was defined as the cancellous bone region 0.5 mm beneath the calcified cartilage–bone junction. The cylinder with volume of (π × 1.5² × 1) mm³ was selected for each specimen to calculate the parameters, including bone volume fraction (BVF, BV/TV), bone surface/bone volume ratio (BS/BV), bone mineral density (BMD), trabecular thickness (Tb.Th), and trabecular spacing (Tb.Sp).

Hematoxylin–eosin (HE) staining and specific staining for condylar cartilage (such as safranine O/fast green and Alcian blue) were adopted to identify the successful establishment of the TMJOA animal model.

Immunohistochemistry (IHC) for PCNA, Col-I, Col-II, ACAN, SOX9, MMP13, and RUNX2 was adopted to analyze the changes in the cartilage tissue of the condyle in the normal and TMJOA states with or without BMSC-sEV treatment.

The histologic observations of normal and TMJOA animal tissues were scored using the International Cartilage Regeneration & Joint Preservation Society (ICRS) Visual Histological Assessment Scale (Supplementary Table 1) and Wakitani Histological Grading Scale (Supplementary Table 2). The ICRS Visual Histological Assessment Scale contains six features (surface, matrix, cell distribution, cell population viability, subchondral bone, and cartilage mineralization), scored from 0 to 3, and the Wakitani Histological Grading Scale contains five categories (cell morphology, matrix staining, surface regularity, thickness of cartilage, and integration of donor with host adjacent cartilage), scored from 4 to 0, total maximum 14. Three independent blinded technicians from the Department of Pathology in our hospital were recruited to score, respectively.

Stem cell-derived small extracellular vesicles were dyed with Exo-Glow Exosome Labeling Kit (SBI, Japan). Single-stranded RNAs in the sEVs were fluorescently labeled in red by acridine orange (AO). F-actin was stained green with Actin-stain 488 fluorescent phalloidin (Cytoskeleton Inc., United States). Finally, the cell nucleus was stained blue with 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific, United States). Fluorescence images were acquired on a TCS SP2 laser scanning confocal microscope (Leica Microsystems, Germany).

Cell proliferation of the MCCs was determined by the Cell Counting Kit-8 (CCK-8; Yeasen, China). MCCs were seeded in a 96-well plate in triplicate. They were treated with BMSC-sEVs for 24 h. During the next few days, the proliferation of the MCCs was detected at a wavelength of 450 nm, as measured by SpectraMax i3 (Molecular Devices, United States).

MCCs were cultured at 37°C until they were in the logarithmic phase and then treated with DMEM without FBS containing BMSC-sEVs for 24 h. The medium with the cells was then placed in the top chamber of a Transwell (Corning, United States) inserted into a 24-well plate. DMEM with 10% FBS was added to the bottom chamber, and the cells were cultured for approximately 24 h. Then, the cells that had migrated into the bottom chamber were fixed with 4% paraformaldehyde, dyed with 0.1% crystal violet, and observed by microscopy (Olympus, Japan).

The cell cycle distribution of the MCCs was determined by a Cell Cycle and Apoptosis Analysis Kit (Yeasen, China). Cell sediments were washed with PBS and resuspended in binding buffer. Propidium iodide (PI) was added to the solution and incubated for 15 min at room temperature in the dark. Then, the samples were analyzed by flow cytometry (FCM; CytoFLEX LX, Beckman, United States).

Total RNA of the BMSCs and MCCs was extracted by TRIeasy^TM Total RNA Extraction Reagent (Yeasen, China) and reverse transcribed into cDNA using Hifair^® II 1st Strand cDNA Synthesis SuperMix for qPCR (Yeasen, China) and a Mastercycler instrument (Eppendorf, Germany). All qPCR programs were performed using Hieff UNICON^® Power qPCR SYBR Green Master Mix (Yeasen, China) and the CFX384^{^TM} Real-Time System (Bio-Rad, United States). mRNA expression was quantified by quantitative PCR (qPCR) using the 2^–ΔΔCT relative quantitation method, and GAPDH served as the internal control. The primers are listed in Table 1.

To knock down the expression of autotaxin, three gene-specific short interfering RNAs (siRNAs) (Table 2) were synthesized (GenePharma, China) and used. The cells were transfected with these siRNAs using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific, United States). After 48 h, the knockdown efficacy was confirmed by qPCR and immunoblotting to quantitate the expression of the genes at both the transcription and translation levels in the cells and in their sEVs. To inhibit the expression of autotaxin, a specific inhibitor, ziritaxestat (GLPG1690) (Selleck, United States), was used.

Statistical analyses were performed with SPSS 19.0 (IBM Corp., United States), and the figures were made with GraphPad Prism 8 software (GraphPad Software, United States). Both parametric and non-parametric inferential statistics were used depending on whether the data were normally distributed tested by Kolmogorov–Smirnov test. Two-tailed t-test and Mann–Whitney U test were used to identify differences between groups. One-way analysis of variance (ANOVA) and Kruskal–Wallis tests were carried out for multiple group comparisons. Three replicates were set for each treatment. The data are presented as the mean ± standard error of the mean (SEM). P value less than 0.05 was considered statistically significant.

Human BMSCs were harvested from the healthy human iliac bone marrow. Viewed by optical microscopy, their cell form was fibroblast-like, with a long barracuda, fish, vortex, or reticular arrangement (Figure 1A). By FCM, the BMSC markers CD105, CD29, and CD44 were identified (whose counts were 96.1, 98.7, and 99.4%, respectively), and the results complied with the BMSC identification criteria (Figures 1B–D).

By centrifugation, microfiltration (0.45 and 0.20 μm) and ultrafiltration (100 kD), BMSC-sEVs were obtained (Figure 1E). Through transmission electron microscopy (TEM), many “cup-shaped” and “tea saucer-shaped” vesicles with membrane structures could be seen, with diameters of approximately 30–120 nm. The size and morphology of the BMSC-sEV samples observed under the microscope conformed to the relevant definitions and characteristics of sEVs (Figure 1F). The NTA results showed that the particle size distribution of the BMSC-sEV samples obtained in this experiment was concentrated at approximately 106 nm. Each 1 ml BMSC-sEV sample contained approximately 1.05 × 10⁷ particles, derived from approximately 5.5 × 10⁷ BMSCs (Figure 1G). The NTA results suggested that the BMSC-sEV samples obtained in this experiment conformed to the relevant definitions and characteristics of sEVs, consistent with the TEM results mentioned above. In addition, the results showed that the concentration and purity of the sEV samples from the BMSC sources were satisfactory.

sEVs can also be identified by positive and negative markers. Positive markers of sEVs, such as CD63, CD81, Rab5, and Alix, were clearly detected in the BMSC-sEVs and their donor cells, whereas the negative marker GRP94 was only detected in the BMSCs (Figure 1H).

These results show that the sEV separation and purification technology used in this study is reliable and effective; the BMSC-sEV samples are in line with the recognized definition and identification standards of sEVs, with satisfactory concentration and purity.