HepG2 cells were purchased from American Type Culture Collection and additional STR profiling was performed (Fig. S1). T-MSCs previously obtained from patients (Ewha University Medical Center Institutional Review Board; approval no. EUMC 2018-01-011-002) at Ewha Womans University Seoul Hospital (Seoul, Korea) were maintained as previously described (20). Patients provided informed written consent for the use of their tissue.

To analyze cell phenotype, T-MSCs were washed with FACS buffer (0.5 FBS and 0.1% NaN₃ in PBS), blocked with 0.5 µg/ml purified rat anti-mouse CD16/CD32 (BD Pharmingen) at 4°C for 5 min and stained with FITC anti-CD11b (cat. no. ICRF44, mouse IgG₁, κ), FITC anti-CD45 (2D1, mouse IgG₁, κ), FITC anti-CD73 (AD2, mouse IgG₁, κ), FITC anti-CD90 (5E10, mouse IgG₁, κ), FITC anti-CD105 (43A3, mouse IgG₁, κ) and FITC mouse IgG₁, κ isotype (MOPC-21) antibodies (all BioLegend, Inc.) at 0.5 µg/ml for 20 min at room temperature. After staining, the cells were fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich; Merck KGaA) in phosphate-buffered saline (PBS) to final 0.5% PFA. Stained cells were acquired using a Novocyte flow cytometer (ACEA Bioscience, Inc.). Acquired cells were analyzed by FlowJo software (v10, TreeStar, Inc.).

For adipogenic differentiation, T-MSCs were seeded at a density of 1×10⁴ cells/well in a 12-well plate and cultured with adipogenic medium (Invitrogen; Thermo Fisher Scientific, Inc.) in a humid atmosphere with 5% CO₂ at 37°C for 3 weeks. Medium was replaced every 3-4 days. After 3 weeks, adipogenic cultures of T-MSCs were rinsed with PBS and fixed in 4% PFA for 5 min at room temperature. The wells were dried completely and stained with Oil red O (Sigma-Aldrich; Merck KGaA) for 10 min at room temperature. The Oil red O solution was removed and wells were immediately washed with distilled water four times. Wharton's jelly-derived (WJ-)MSCs were purchased from PromoCell GmbH. Ethics approval was received for the use of primary cells (Ewha Institutional Biosafety Committee; approval no. IBC-past-096). HepG2 cells were cultured in Minimum Essential Medium Eagle (MEM; Welgene, Inc.) with 10 FBS (Welgene, Inc.) and 1% penicillin/streptomycin solution (P/S; Capricorn Scientific GmbH) in a humid atmosphere with 5% CO₂ at 37°C. T-MSCs and WJ-MSCs were cultured in low-glucose Dulbecco's modified Eagle medium (DMEM; Welgene, Inc.) with 10 FBS (Welgene, Inc.) and 1% P/S in 100-mm cell culture plates. For preparation of T-MSC CM (T-CM), T-MSCs at 80% confluence were washed four times with PBS (Welgene, Inc.) and medium was replaced with serum-free DMEM. The medium was collected after 48 h, centrifuged at 190 x g for 5 min at room temperature, passed through a 0.2-µm filter (MilliporeSigma) and concentrated 20-fold using a 3-kDa Amicon Ultra centrifugal filter unit (EMD Millipore) with high-speed centrifugation (Sorvall LYNX4000; Thermo Fisher Scientific, Inc.) at 5,000 x g for 1 h at 4°C. T-CM for animal experiments was frozen, whereas T-CM and CM from WJ-MSCs for exosome isolation and reverse transcription-quantitative (RT-q)PCR were used immediately.

For EV isolation, Minimal Information for Studies of EVs (MISEV) 2018 guidelines proposed by the International Society for EVs (ISEV) were referred to for separation and characterization (11).

To isolate EVs from T-MSCs, 1/5 volume of ExoQuick-TC reagent (System Biosciences) was added to 10 ml T-CM and mixed by vigorous inverting. Following incubation at 4°C overnight, the mixture was centrifuged at 1,500 x g for 30 min at 4°C. The supernatant was removed, and final centrifugation at 1,500 x g was performed for 5 min at room temperature. The visible EV-containing pellet was resuspended in 100-500 µl PBS for Nanosight particle tracking analysis (Nanosight NS300; Malvern Instruments, Ltd.) and for protein concentration analysis via bicinchoninic acid (BCA) protein assay (Thermo Fisher Scientific, Inc.). Following BCA assay, 6X SDS-PAGE loading buffer was added for sample preparation for immunoblotting and stored at −80°C until use.

Purified EVs were diluted to 1:1,000 in PBS. A total of 5 µl diluted EVs was dropped on Formvar-carbon-coated EM grids. The grids were stained with 2% uranyl acetate for 2 min at room temperature and removed using filter paper. Finally, the grids were viewed using a H-7650 TEM (Hitachi, Ltd.) at 80 kV. Digital images were captured at a magnification of 70,000-200,000.

Diluted EVs were dropped on Poly L-lysine cover glass and prefixed with 0.25% glutaraldehyde for 30 min at room temperature. After washing in PBS, samples were maintained in 1% osmium tetroxide for 30 min for final fixation at room temperature. Then, samples were washed and dehydrated by serial dilution with ethanol and critical point drying. Finally, samples were mounted onto stubs, sputter-coated with gold by Quorum Technologies and examined with a Sigma-300 microscope (Zeiss GmbH) at a magnification of ×70,000.

Male BALB/c nude mice (n=6, age, 5 weeks; weight, 18±2 g) were purchased from OrientBio. All animals were housed at 21-23°C and 51-54% humidity in a pathogen-free environment on a 12/12-h light/dark cycle and allowed free access to food and water. All animals were monitored every day for health and behavior. The method of euthanasia was carbon dioxide inhalation followed by cervical dislocation (20% vol/min for a cage size of 8×13×5 inches) and animal death was confirmed by cardiac and respiratory arrest. The experimental procedures were approved by the Animal Care and Use Committee of the College of Medicine, Ewha Womans University (Seoul, Korea; approval no. EUM18-0408). To construct a hepatoma xenograft model, 3×10⁶ HepG2 cells were suspended in 100 µl low-glucose DMEM and injected subcutaneously into the right back of each animal (n=3) (21). To assess the effect of T-CM, HepG2 cells were suspended in 100 µl CM from 5×10⁵ T-MSCs and injected at the same position (n=3). Five days after the injection, mice were euthanized by carbon dioxide and cervical dislocation and the tumor was isolated, chopped and seeded onto a 100-mm cell culture plate. Images were captured following 2 and 9 days of culturing in a humid atmosphere with 5% CO₂ at 37°C using an inverted light microscope (Olympus Corporation) at ×100 magnification, and cell clusters were calculated manually by counting cluster of >50 cells.

Equal concentrations of EV (5 µg/lane) from T-MSCs were loaded onto 5% stacking/10% separating sodium dodecyl sulfate (SDS)-PAGE, separated by electrophoresis, transferred to polyvinylidene difluoride (PVDF) membranes, blocked with 5% skimmed milk in TBST (50 mM Tris HCl, pH 7.6, 150 mM NaCl, 0.1% Tween-20) for 1 h at room temperature and incubated with primary antibodies overnight at 4°C. Gels were stained with Coomassie blue solution (0.1 Coomassie brilliant blue R-250, 40.0 methanol and 10.0% acetic acid in water) overnight at room temperature followed by incubation with de-staining solution (40 methanol and 10% acetic acid in water) for 2 h at room temperature. All primary antibodies were prepared by diluting in 3.00 BSA (Bovogen Biologicals Pty, Ltd.) and 0.02% sodium azide (Sigma-Aldrich; Merck KGaA) in TBST. Anti-CD63 mouse monoclonal antibody (cat. no. ab193349; 1:1,000; IgG₁κ) was purchased from Abcam; anti-CD81 (cat. no. sc-166029; 1:200; IgG_2bκ) and anti-β-actin mouse monoclonal antibody (cat. no. sc-47778; 1:3,000; IgG₁κ) were purchased from Santa Cruz Biotechnology, Inc. PVDF membranes were washed three times for 10 min each in TBST and incubated with horseradish peroxidase-conjugated goat anti-mouse IgG (H+L) antibody (cat. no. #1706516; BioRad Laboratories, Inc.), diluted in TBST (1:4,000), for 1 h at room temperature. Following incubation, the membranes were washed three times for 10 min each in TBST and developed using an EZ-Western Lumi Femto Western blot kit (Doo Gene Bio Co., Ltd.). Images were obtained using ImageQuant LAS 3000 (FUJIFILM Wako Pure Chemical Corporation). The pixel densities of protein bands were analyzed using UN-SCAN-IT-gel 6.1 software (Silk Scientific, Inc.).

METAFECTENE PRO (Biontex Laboratories GmbH) was used according to manufacturer's protocol to transfect cells. Briefly, mixed solutions containing 1 µg AccuTarget™ fluorescein-labeled miRNA negative control, inhibitor #1 (cat. no. SMC-4101, Bioneer, bioneer.co.kr/20-smc-4101.html) or has-miR-199a-3p inhibitor (Bioneer) were added to 50 µl serum-free medium. Then, 3 µl METAFECTENE PRO was added to 50 µl serum-free medium at room temperature for 20 min. The sequence of has-miR-199a-3p inhibitor was 5′-ACA GUA GUC UGC ACA UUG GUU A-3. ′ Following incubation, the mixture was carefully added to cells. After 5 h, cells were collected for RT-qPCR and wound healing assay was performed.

CM from five different donor originated T-MSCs and DMEM (negative control) were sent to Macrogen, Inc. for miRNA sequencing by SMARTer smRNA-Seq method (22) using TruSeq Small RNA Library Prep kit (RS-200-0012, Illumina, Inc.).Ribosomal RNA removed reads were aligned to reference genome (miRBase v22.1) and non-coding RNA database (RNAcentral 14.0) to classify known miRNA and other type of RNA. Novel miRNA prediction is performed by miRDeep2 (v2.0.0.8). To reduce systemic bias, size factors from the count data were estimated and Trimmed Mean of M-values normalization with edgeR R library was applied. Statistical analysis was performed using Fold Change, exact-Test using edgeR per comparison pair. Hierarchical clustering analysis was also performed using complete linkage and Euclidean distance as a measure of similarity to display the expression patterns of differentially expressed miRNAs that satisfied the |fold-change|≥2 and P<0.05 criteria. All data analysis and visualization of differentially expressed genes was performed using R 3.3.1 (r-project.org). miRNA target gene prediction was performed by TargetScanHuman 7.2 (targetscan.org/vert_72/). In order to analyze signaling pathways associated with miRNA from exosomes, miRNA-target genes were analyzed by Kyoto Encylcopedia of Genes and Genomes (KEGG) analysis. Genes derived from mirDIP were further analyzed by Database for Annotation, Visualization and Integrated Discovery v6.8 to identify enriched biological pathways.

To isolate miRNA, T-CM and CM from WJ-MSCs was added to an appropriate volume of Exo2D™ for RNA (ExosomePlus) according to the manufacturer's protocol, and inverted several times. The mixture was incubated at 4°C for 30 min and centrifuged at 3,000 x g for 30 min at 4°C. The supernatant was removed and resuspended in 100 µl PBS and the RNA concentration was measured by BioPhotometer D30 (Eppendorf) and adjusted to 1 µg/µl. miRNA was converted into complementary DNA (cDNA) using MystiCq™ microRNA cDNA Synthesis Mix (Sigma-Aldrich; Merck KGaA) according to the manufacturer's protocol. MystiCq Universal PCR Primer (Sigma-Aldrich; Merck KGaA) was used as a reverse primer for RT-qPCR. MystiCq microRNA qPCR Control Primer RNU6-1 (cat. no. MIRCP00001) and MystiCq microRNA qPCR Assay Primer hsa-miR-199a-3p (cat. no. MIRAP00244; both Sigma-Aldrich; Merck KGaA) were used as forward primers. Total RNA was extracted using an RNeasy Plus Mini kit (Qiagen GmbH) to isolate RNA of transfected HepG2 cells. Total RNA (1 µg) was transcribed into cDNA using RT reagent (ElpisBiotech, Inc.), and RT-qPCR was performed. Primer sequences were as follows: CD151 forward, 5′-ATG GGT GAG TTC AAC GAG AAG A-3′ and reverse, 5′-GCA GGC TGA TGT AGT CAC TCT -3′; integrin α3 (ITGA3) forward, 5′-TGT GGC TTG GAG TGA CTG TG-3′ and reverse, 5′-TCA TTG CCT CGC ACG TAG C-3′; ITGA6 forward, 5′-ATG CAC GCG GAT CGA GTT T-3′ and reverse, 5′-TTC CTG CTT CGT ATT AAC ATG CT-3′ and human GAPDH (192 bp) forward, 5′-GGT AAA GTG GAT ATT GTT GCC ATC AAT G-3′and reverse, 5′-GGA GGG ATC TCG CTC CTG GAA GAT GGT G-3′. The mixture was prepared in each well of a Fast Optical 96-well reaction plate (Applied Biosystems; Thermo Fisher Scientific, Inc.) using 0.5 reverse/forward primer each, 10.0 1X SensiFAST SYBR Hi-ROX mix (Bioline; Meridian Bioscience, Inc.), 8.0 deionized water and 1.0 µl cDNA. Amplification was performed in triplicate with 40 cycles of 15 sec denaturation at 95°C, 1 min annealing at 62°C and 30 sec extension at 72°C using StepOnePlus Real-Time RCR System (Applied Biosystems; Thermo Fischer Scientific, Inc.). The relative fold expression and changes were calculated using the 2^−∆∆Cq method (23).

A scratch wound healing assay was performed to compare the effect of T-CM or exosomes on the migration capability of HepG2. HepG2 cells (2×10⁵) were seeded onto a 12-well cell culture plate with MEM containing 10% FBS and 1% P/S and incubated to 90-100% confluence at 37°C in a 5% CO₂ incubator. When the confluence of cells was reached, a scratch wound was made in the cell monolayers using a cut cell scraper. The transfected cells were allowed to grow for an additional 24 and 48 h in the presence of CM or EVs (10 µg/ml) with serum-free MEM. Images were taken at 0, 24 and 48 h using an inverted light microscope (Olympus Corporation; ×40 magnification). The migration distance was quantified using ImageJ software v1.53 g (imagej.nih.gov). The percentage of area closure was calculated as follows: Final wound width/initial wound width ×100.

Data are presented as the mean ± SEM (n>3). Statistical significance was determined by one-way ANOVA with multiple comparison by Sidak test applied to the wound healing assay with T-CM or miRNA inhibitors. Paired t-test was used for wound area closure after T-CM treatment. Student t-test (unpaired) was used for the comparison of miR expression between WJ-MSCs and T-MSCs. All analyses were performed using Prism 9.2 (GraphPad Software, Inc.). P<0.05 was considered to indicate a statistically significant difference.

Investigation of the effect of subcutaneous injection of CM of T-MSCs at the right back of mice showed that T-CM with HepG2 cells formed a smaller tumor mass than injection of DMEM with HepG2 cells after 9 days (Fig. 1A). At 5 days after the injection, no mice had died and the plated cells from tumor masses were separated, chopped and seeded for culture. Formation of cell clusters was assessed at 2 and 9 days. The cell clusters in the HepG2 + T-CM group tumor mass were less prolific than those from the control group (Fig. 1B). Comparison of the number of clusters on culture plates (Fig. 1C) indicated that T-CM exerted an inhibitory effect on HepG2 cell tumor growth and expansion in vitro. The in vitro scratch assay compared the effect of T-CM on the migration of HepG2 cells were obtained after 24 and 48 h (Fig. 1D); the distance between the scratch of the control group (DMEM) was closer and the area became smaller as time passed. By contrast, in the T-CM group, wound area remained almost same as that at 0 h. These results suggested that T-CM had an inhibitory effect on the migration of HepG2 cells (Fig. 1E and F).

The morphology of T-MSC small EVs isolated from CM and observed by TEM and SEM showed a round or oval shape with diameter <200 nm (Fig. 2A). Immunoblot analysis indicated positive expression of CD63 and CD81 (Fig. 2B). The particle size distribution and concentration of isolated exosomes were analyzed by Nanosight particle tracking analysis (Fig. S2). The mean diameter of particles was 192.3±21.6 nm and the concentration was 1.28×10⁸±2.38×10⁷ particles/ml.

To search the highly expressed miRNAs in T-MSC EVs, five different origins of T-MSCs established in the lab were selected. These five T-MSCs showed typical surface markers of MSCs (Fig. S3A). For the comparison of differentiation potential, adipogenic differentiation of these five different T-MSCs was induced (Fig. S3B). After confirming these cells possessed chracteristics of MSCs [plastic adherence, phenotype marker expression (CD73⁺, CD90⁺, CD105⁺, CD45⁻ and CD11b⁻) and adipocyte differentiation], miRNA sequening was performed. Mycoplasma contamination was not detected in small RNA composition report from Macrogen. The highly expressed miRNAs are listed in Table I. A heatmap (Fig. S4A) was constructed to demonstrate the results of hierarchical clustering analysis (Euclidean method, complete Linkage), which clusters the similarity of mature miRNAs and samples by expression level (normalized value) from a significant list. Fig. S4B shows the number of up- and downregulated mature miRNAs based on fold change compared with those of negative control. The upregulated miRNAs with similar expression levels were grouped together using the normalized value of each sample (Fig. S4C). Enriched signaling pathways among top 20 miRNA-target genes were analyzed by KEGG analysis; genes belonging to cancer pathway were highly enriched (Fig. S5). Exosomal RNA was extracted from T-MSC and WJ-MSC CM. To validate miRNA sequencing, hsa-miR-199a-3p, hsa-miR-214-3p, hsa-miR-199a-5p and hsa-miR-199b-5p expression levels were compared with expression of exosomal RNA from T-MSCs and WJ-MSCs by RT-qPCR. Expression of miR-199a-3p, mir-214-3p, and miR-199-5p was higher in T-MSCs than in WJ-MSCs (Fig. 3A). The exosomes of T-MSCs had higher levels of hsa-miR-199a-3p than those of WJ-MSCs, as indicated by RT-qPCR and miRNA sequencing (Fig. 3B; Table I). To identify the effect of miR-199a-3p on HepG2 cell migration in vitro, transfected cells were treated with negative control or miR-199a-3p inhibitor. Treated cells were scratched and observed for 24 and 48 h following the addition of 10 µg/ml exosomes (Fig. 3C). The migrated distance in the control group (without exosomes) was greater and the wound area became smaller as time passed. By contrast, the wound area in the exosome-exposed group remained similar throughout the experiment. miR-199a-3p inhibitor reversed the effect of exosomes on migration (Fig. 3D). miR-199a-3p inhibitor reversed the effect of exosomes on migration after 48 h (Fig. 3E).

To investigate the molecular mechanism of how miR-199a-3p affects HepG2 cells, miR-199a-3p targets were predicted using the TargetScanHuman 7.2 database (targetscan.org/vert_72/). Hsa-miR-199a-3p was predicted to target Kelch-like family member 3, serpin family E member 2, pro-apoptotic WT1 regulator, vesicle-associated membrane protein 3, G protein subunit α12, BCAR3, CDK7, CD2-associated protein, CD151, FGF7, CXCL11, ITGB8, G3BP stress granule assembly factor 2, CDK17, mesenteric estrogen-dependent adipogenesis, collagen (COL) type IV α5 chain, RAB GTPase-activating protein 1, IL13RA1, ITGA3, SP1, TAO kinase 1, ITGA6, COL12A1 and histamine N-methyltransferase. The seed sequence of hsa-miR-199a-3p matched the sequence of the 3′ untranslated regions of CD151, ITGA3 and ITGA6, suggesting that CD151, ITGA3 and ITGA6 were potential targets of miR-199a-3p (Fig. 4A). To confirm the effect of miR-199a-3p on T-MSCs, HepG2 cells were transfected with miR-199a-3p inhibitor or negative control inhibitor. Transfection efficiency of miR-199a-3p inhibitor was confirmed by the 87% decrease of the miR-199a-3p expression (Fig. S6). Expression of CD151, ITGA3 and ITGA6 in HepG2 cells was downregulated by EV treatment (Fig. 4B).