We used six malignant pleural mesothelioma (MPM) cell lines (ACC-MESO1, ACC-MESO4, MSTO-211H, L324, N407, and K921) and a normal pleural mesothelial cell line (MeT-5A). ACC-MESO1 and ACC-MESO4 were obtained from the Riken Cell Bank (Tsukuba, Japan). MSTO-211H was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). L324, N407, and K921 cells were generated in the laboratory and kindly donated by Dr Hidetaka Uramoto, Kanazawa Medical University, Ishikawa (16). MeT-5A cells were purchased from ATCC. Human prostate cancer PC3 cells were obtained as previously described (17). Immortalized human pulmonary fibroblast (IHPF) cells (T0490) were purchased from Applied Biological Materials Inc. (Richmond, BC, Canada). MPM, MeT-5A, PC3, and IHPF cells were maintained in RPMI-1640 medium, Medium 199, DMEM (Thermo Fisher Scientific, Inc.), and Prigrow III medium (Applied Biological Materials Inc.), respectively. All cells were cultured with GlutaMAX supplement (Thermo Fisher Scientific, Inc.), 10% heat-inactivated fetal bovine serum (FBS), and 1% (v/w) penicillin/streptomycin, and were maintained at 37°C in 5% CO₂.

CD63-green fluorescent protein (GFP) (CYTO122-PA-1), CD9-GFP (CYTO120-PA-1), and CD81-red fluorescent protein (RFP) (CYTO125-PA-1) expression plasmids were purchased from System Biosciences. The pNL1.1 vector containing nanoluciferase (Nluc) was purchased from Promega Corp. CD63-Nluc, CD9-Nluc, CD81-Nluc, and CD81-GFP expression plasmids were constructed with these plasmids. GFP fusion proteins were used for qualitative analysis of intracellular expression and intracellular uptake. Nluc fusion proteins were used for quantitative analysis of extracellular secretion and intracellular uptake. Double-stranded oligonucleotides against hsa-miR-4728-5p, hsa-miR-193a-5p, and hsa-miR-551b-5p were synthesized (Table SI) and ligated into multicloning sites of pSIH1-H1-GFP-T2A-Puro, which was originally constructed with pSIH1-H1-copGFP (SI501A-1, System Biosciences) and pCDH-EF1-MCS-BGH-PGK-GFP-T2A-Puro (CD550A-1, System Biosciences).

Lentivirus-containing expression cassettes of GFP/Nluc fusion genes and miRNAs were obtained using the System Biosciences lentivirus packaging system. Lentiviral transduction was performed according to the manufacturer's protocol. Briefly, after infecting the MeT-5A cells, cells expressing CD63-GFP/Nluc, CD9-GFP/Nluc, CD81-GFP/Nluc were selected using 10 µg/ml of puromycin. In the same way, MeT-4728-5p, MeT-193a-5p and MeT-551b-5p cells expressing hsa-miR-4728-5p, hsa-miR-193a-5p and hsa-miR-551b-5p, respectively, were obtained by infection and selection with MeT-5A cells. MeT-ctrl cells did not express miRNA. Fused GFPs or independently expressed GFP of the Nluc fusion protein and miRNAs were observed using an EVOS fluorescence microscope (Thermo Fisher Scientific, Inc.). After confirming that over 90% of the cells were expressing GFP using LUNA-FL™ (Logos Biosystems, Korea), they were used in the assay.

Syringe filters with sizes of 220 nm (SFPES013022N), 50 nm (SF16008), and 20 nm (2.CF7103.0001) were purchased from Membrane Solutions, TISCH Scientific, and ANPEL Laboratory Technologies, respectively. In our laboratory, FBS was passed through a 220-nm bottle-top vacuum filter (Thermo Fisher Scientific, Inc.) (FBSpas220) to maintain normal cells. To remove EVs from FBS, FBSpas220 was passed through 50-nm syringe filters (FBSpas50). When cells grown in 10-cm dishes containing maintenance medium were 80% confluent, they were collected and seeded into three 10-cm dishes containing FBSpas50. After 48 h, the culture medium was collected and passed sequentially through 220 nm and 50 nm syringe filters. EVs captured by the 50 nm syringe filter (EVcap50/220) were partially collected by reverse-flow with phosphate-buffered saline (PBS) (EVrev50/220). The EVrev50/220 and EVs that had passed through a 50 nm syringe filter (EVpas50) were analyzed by dynamic light scattering (DLS, Zetasizer, Malvern Panalytical, UK) and transmission electron microscopy (TEM, JEOL 1200EX; JEOL, Japan). EVcap50/220 was used directly for RNA extraction and protein extraction, as outlined in the procedures below.

The diameter of EVs obtained by reverse flow was measured by DLS (Zetasizer) according to the manufacturer's instructions, and as previously described (18). Samples were prepared using an Exosome TEM-easy kit (101 Bio, Mountain View, CA, USA) with EVs obtained from reverse flow; observation was carried out on a TEM (JEOL 1200EX).

PC3 cells (1,000 cells) were seeded into 12-well plates. Cell number was measured the next day and was used as the baseline at 0 h. PC3 cells were washed twice with PBS, and were added to either medium without FBS, medium containing FBSpas220 (normal medium), or medium containing FBSpas50. Cells were collected every 24 h, and the number of cells was calculated by LUNA-FL™ (Logos Biosystems) according to manufacturer's instructions. Experiments were performed in triplicates.

When MeT-5A cells grown in 10-cm dishes containing maintenance medium were 50% confluent, they were washed twice with PBS, and added with FBSpas50-containing medium. After 48 h, cultured medium or fresh medium containing FBSpas50 (for the negative control) was collected and passed sequentially through 220 and 50 nm syringe filters. Proteins of EVcap50/220 were eluted with 2 ml CHAPS lysis buffer (Dojindo) containing 1 mM PMSF. Each eluted solution was concentrated using Amicon Ultra 10 kDa (Merck Millipore) until its volume was less than 30 µl. All samples were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and were transferred to polyvinylidene difluoride (PVDF) membranes. The membrane was immunoblotted first with anti-EpCAM antibody (EBA-1, dilution 1:1,000, Santa Cruz Biotechnology, Inc.), CD63 antibody (TS63, dilution 1:1,000, Thermo Fisher Scientific, Inc.), CD9 antibody (TS9, dilution 1:1,000, Thermo Fisher Scientific, Inc.), CD81 antibody (M38, dilution 1:1,000, Thermo Fisher Scientific, Inc.). Secondary antibodies conjugated with horseradish peroxidase were used for detection. Signal intensity was obtained using LAS 4000 Mini and Multi Gauge software version 3.0 (Fujifilm), as previously described (17).

MeT-5A cells expressing and secreting CD9-GFP/Nluc, CD63-GFP/Nluc, and CD81-GFP/Nluc were cultured in plates or dishes with medium containing FBSpas50. These cells are called donor cells because the fusion proteins secreted by these cells are taken up by other cells. After the indicated hours, cultured medium or cells were collected. Each medium was passed through a 220 nm or 50 nm syringe filter and was used for Nluc assays. Cells were washed twice with PBS and lysed with Reporter Lysis Buffer (Promega Corp.). The cell lysate was vortexed, centrifuged (21,000 × g for 2 min), and the supernatant was collected for Nluc assays. Nluc activity was measured using the Nano-Glo^® Luciferase Assay System (Promega Corp.) according to the manufacturer's instructions. In order to observe secreted EVs with the GFP fusion protein via an EVOS fluorescence microscope (Thermo Fisher Scientific, Inc.), EVs were concentrated using a Total Exosome Isolation kit (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions.

Culture medium of MeT-5A cells that expressed CD81-Nluc was passed through a 220-nm syringe filter (EVpas220). EVpas50/220 and EVpas20/50 were obtained by passing EVpas220 through a 50-nm syringe filter and by passing the EVpas50/220 through a 20-nm syringe filter, respectively. Nluc activities of EVpas220, EVpas50/220, and EVpas20/50 were measured as described above.

To obtain the GFP fusion protein, MeT-5A cells expressing CD9-GFP, CD63-GFP, and CD81-GFP (donor cells) were cultured with 10 ml Opti-MEM medium without FBS for 48 h. Each culture medium was passed through a 220-nm syringe filter and concentrated to 30 µl or less by Amicon Ultra-4 (Merck Millipore). They were added to the culture medium of recipient cells that were able to take up the GFP fusion protein. After 8 h, cells were washed twice with PBS, and were observed under an EVOS fluorescence microscope (Thermo Fisher Scientific, Inc.). For the Nluc fusion protein, MeT-5A cells expressing CD81-Nluc were cultured in medium containing FBSpas50. After 48 h, each culture medium was passed through 220 and 50-nm syringe filters, and was independently replaced with media of recipient cells. After the allocated hours, Nluc activities in cell lysates were measured as described above, and were standardized to cell protein concentration. Uptake of EVs with diameters between 50 and 220 nm was determined by subtraction.

The medium that passed through the 220-nm syringe filter was passed through a 50-nm syringe filter to obtain EVcap50/220. Total RNA of EVcap50/220 was eluted with 1 ml ISOGEN (Nippongene). The eluted solution was mixed with 200 µl chloroform, and was centrifuged at 12,000 × g for 15 min. Total RNA was purified with 500 µl supernatant using a NucleoSpin^® miRNA Plasma Kit (Macherey-Nagel), which was used according to the manufacturer's instructions. Total cell RNA was eluted in 100 µl H₂O, and was precipitated with 10 µl of 3.0 M potassium acetate, 100 µl isopropanol, and 1 µl Pellet Paint. After washing and drying the pellet, RNA was dissolved in 2 µl or 10 µl H₂O for microarray or quantitative real-time polymerase chain reaction (qPCR), respectively. Verification of microRNA purification was performed with Eukaryote Total RNA Nano Series II using an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.). All RNAs, including the miRNAs from cells, were purified using the miRNeasy Mini Kit (Qiagen), according to the manufacturer's instructions.

Microarray analyses with the Human miRNA Oligo chip-4 plex (TORAY, Tokyo, Japan) containing 2,565 probe sets and the Human Oligo chip 25k (TORAY) containing 24,460 probe sets were carried out using a 3D-Gene array system, according to the manufacturer's protocol. For the miRNA array, 2 µl RNA solution obtained from 30 ml medium or 250 ng total cell RNA were hybridized with the Human miRNA Oligo chip. For the mRNA array, 1 µg total cell RNA was used to hybridize the Human Oligo Chip 25k.

For RNA obtained from EVcap50/220, the dry pellet was dissolved with 10 µl of H₂O, and 1 µl RNA was used for each probe. For RNA obtained from cell lysates, RNA was prepared according to the manufacturer's instructions. cDNA for the miRNA was synthesized using the TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific, Inc.) in a StepOne Plus real-time PCR system (Thermo Fisher Scientific, Inc.). qPCR was performed using a specifically designed TaqMan probe. Primer sequences used to analyze human miRNAs and gene expression via qPCR were provided by the following sources: miR-4728-5p, Thermo Fisher Scientific, Inc., 461811_mat; miR-193a-5p, Thermo Fisher Scientific, Inc., 002281; miR-551b-5p, Thermo Fisher Scientific, Inc., 002346; U6, Thermo Fisher Scientific, Inc., 001973. The cycling conditions were as follows: Denaturing, hold at 95°C for 20 sec, 40 cycles of amplification (denaturation at 95°C for 30 sec, annealing and extension at 60°C for 30 sec). U6 was used as an internal control. The Ct values for each miRNA were normalized to U6, and relative gene expression was calculated using the ∆∆Ct method. All samples were run in duplicates in each experiment.

Statistical analysis was performed using GraphPad Prism7 (GraphPad Software, Inc.). Results were compared using a two-tailed Student's t-test and one-way analysis of variance (ANOVA) followed by Turkey's test and Dunnett's test. Data are expressed as mean ± SD. P<0.05 was considered to indicate a statistically significant difference.

In general, the cell culture medium was made by adding fetal bovine serum (FBS) to the basic culture medium. Since FBS contains EVs derived from bovine, it is necessary to remove them in order to analyze cultured cell-derived EVs. We usually use FBS passed through 220 nm (FBSpas220) for maintenance culture medium. In this experiment, to remove EVs that were under 50 nm in diameter, FBSpas220 were further passed through 50-nm filters (FBSpas50). The efficiency of EV removal was analyzed via dynamic light scattering (DLS) with EVs captured by 50-nm filters (EVcap50/220), as well as EVs that passed through those filters (EVpas50). As shown in Fig. 1A, the diameter of EVcap50/220 was approximately 170 nm; however, this peak disappeared as they flowed through the 50-nm filters (Fig. 1B). To evaluate the shape of EVcap50/220, the EVs were collected by backflow (EVrev50/220). As shown in Fig. 1C, EVrev50/220 was spherical when analyzed using transmission electron microscopy (TEM). These results suggest that EVs from 50 to 220 nm diameter can be captured with 50 nm filters by passing the liquid sequentially through 220 and 50 nm filters (Fig. S1). Next, we investigated the effects of using FBSpas50 medium on cell growth. For this analysis, we used hormone-insensitive PC3 cells (19) with markedly rapid cell growth. As shown in Fig. 1D, cell proliferation was increased with a culture medium containing FBSpas50 when compared to FBSpas220-containing medium. This result suggested that FBSpas50 can be used for analysis of cultured cells, and that bovine-derived EVs inhibited cell growth. In this study, when isolating cell-derived EVs, cells were cultured in medium containing FBSpas50, and EVs secreted from the cells were captured with filters of various diameters. By this method, FBS-derived EVs are not contaminated during analysis with cell culture medium when filters greater than 50 nm are used.

MeT-5A cells were cultured in medium with FBSpas50 for 48 h and EVs in the medium were captured by 50-nm filters. The diameter of EVcap50/220 was approximately 170 nm when measured by DLS (Fig. 1E). Proteins and RNAs of EVcap50/220 were analyzed. As shown in Fig. 1F, EVcap50/220 contained CD9, CD63, CD81, and EpCAM proteins, which are markers of enriched exosomes. EVcap50/220 also contained microRNAs (miRNAs) less than 30 bp in length (Fig. 1G). Particle diameter, shape, specific protein expression, and miRNA inclusion obtained from FBS and cell culture medium suggested that EVcap50/220 consisted of exosomes and microvesicles.

As EVs function as intercellular mediators, it is very important that EVs derived from cells (donor cells) are taken up by other cells (recipient cells). We generated MeT-5A cells that expressed CD9-green fluorescent protein (GFP), CD63-GFP, and CD81-GFP proteins, which are representative marker proteins of exosomes. Exosomes are membrane vesicles released by the fusion of organelles in the endocytic pathway, forming multivesicular bodies (20). Localization similar to multivesicular bodies was observed in cells expressing CD81-GFP (Fig. 2A, donor cells), CD9-GFP, and CD63-GFP (Figs. S2A and S3A). Next, we examined secretion of EVs containing CD81-GFP, CD9-GFP, and CD63-GFP. EVpas220 of culture medium was captured using the Total Exosome Isolation Kit as the EVpas220 was too small to be directly observed via a microscope. As shown in Fig. 2B, the GFP signal was observed in medium in which the cells expressed CD81-GFP, but not in the control medium. Results with CD9-GFP and CD63-GFP are shown in Figs. S2B and S3B. These results suggested that the EVs contained CD63-GFP, CD9-GFP, and CD81-GFP. Since we observed that EVs contained GFP fusion protein in the culture medium, we investigated whether MeT-5A cells could take up EVs in the medium. As shown in Fig. 2A (recipient cells) and Figs. S2A and S3A, multivesicular bodies were observed 8 h after addition of the medium. These results suggested that EVs in the medium could be taken up by other cells. To quantify extracellular secretion and endocytosis of EVs, nanoluciferase (Nluc) fusion proteins, CD63-Nluc, CD9-Nluc, and CD81-Nluc were expressed in cells and analyzed. The amount of extracellular secretion of each Nluc fusion protein was found to be increased in a time-dependent manner (Fig. 3A). Since the luciferase activity of CD81-Nluc was the highest among the three, the capture rate of the syringe filter was calculated using cells that expressed CD81-Nluc. As shown in Fig. 3B, 64% of EVpas220 were captured at 20 nm, of which 80% of the EVs were captured at 50 nm. Next, we investigated whether MeT-5A and IHPF cells were able to take up EVs with diameters between 50 and 220 nm. As shown in Fig. 3C, EV uptake increased in a time-dependent manner, but the amount varied between cells.

As shown in Fig. 3B, EVcap50/220 captured 80% of EVs with diameters between 20 and 220 nm, which included exosomes and microvesicles, but not apoptotic bodies. Considering the collection efficiency and filter handling (pressure power, flow volume), miRNAs in EVs were purified from EV50/220 (EV50-miRNAs). Expression profiles of the EV50-miRNAs were obtained by a 3D-Gene microarray system, and were compared between 6 types of malignant pleural mesothelioma (MPM) and MeT-5A cells. This chip contained a set of 2,565 probes derived from human cells, and signals detected for each miRNA were normalized by a global normalization method that adjusted the median of the detected signal intensity to 25. From the results of the microarray, 390 and 502 EV50-miRNAs were detected in MeT-5A and all six MPM cell lines, respectively. As shown in Fig. 4A, 11 EV50-miRNAs were detected exclusively in MeT-5A cells, but not in MPM cells. On the other hand, 123 EV50-miRNAs were detected in MPM cells but not in MeT-5A cells. These EV50-miRNAs may be markers of MPM. Among the 123 EV50-miRNAs, the 20 most common variants are shown in Table I in descending order of expression. The ratio was calculated using the mean values of MPM cells. Expression of these EV50-miRNAs are also shown on a heat map (Fig. 4B). Since the expression value of hsa-miR-4728-5p was the highest among all EV50-miRNAs, we focused on this miRNA and searched for its target genes.

First, we confirmed the expression of hsa-miR-4728-5p in EVcap50/220 of MPM and MeT-5A cells by qPCR. As shown in Fig. 5A, expression levels of hsa-miR-4728-5p in MPM cells were 5–17 times higher as compared with those of MeT-5A cells. To investigate the target genes of hsa-miR-4728-5p in normal cells, we established MeT-5A cells that overexpressed this miRNA (MeT-4728 cells) as well as control miRNA-free cells (MeT-ctrl). Since it was confirmed that hsa-miR-4728-5p was overexpressed in the MeT-4728-5p cells (Fig. 5B), microarray analysis was performed using mRNA obtained from MeT-4728-5p and MeT-ctrl cells. Based on the result, only two mRNAs, which came from EIF4H and TMEM87A, satisfied the following two conditions: i) Expression level of genes after global normalization was higher than 100 in MeT-ctrl cells; ii) expression levels of genes in MeT-4728 cells decreased to less than 0.5 when compared to MeT-ctrl cells. Since there are few mRNAs whose expression is suppressed by hsa-miR-4728-5p, the intracellular expression of hsa-miR-4728-5p in each cell was analyzed by qPCR. Contrary to our expectations, its expression was found to be highest in MeT-5A cells (Fig. 5C). These data suggest that MeT-5A cells express hsa-miR-4728-5p, but do not secrete it extracellularly, and function only in the cell.

As described above, hsa-miR-4728-5p, which showed the highest expression in EV50-miRNAs of MPM cells, was highly expressed in MeT-5A cells. Therefore, we selected EV50-miRNAs that were not found inside MeT-5A cells from 123 EV50-miRNAs via microarray analysis. As a result, 6 miRNAs were listed as candidates (Table II). Based on these results, we selected hsa-miR-193a-5p and hsa-miR-551b-5p, which exhibited the highest expression. It was confirmed that these miRNAs were contained in EV50-miRNAs of many MPM cells, and were not often present in the MeT-5A cells (Fig. 6A). We also examined the expression of hsa-miR-193a-5p and hsa-miR-551b-5p in cells, but found no correlation in miRNA expression between MPM and MeT-5A cells (Fig. 6A and B). Next, we established MeT-5A cells that overexpressed hsa-miR-193a-5p (MeT-193a cells) or hsa-miR-551b-5p (MeT-551b cells) to investigate their influence on gene expression (Fig. 7A). There were 288 mRNAs in the MeT-193a-5p cells and 146 mRNAs in the MeT-551b-5p cells that satisfied the following two conditions: i) Expression levels of genes after global normalization were higher than 100 in MeT-ctrl cells; ii) expression levels of genes in MeT-193a-5p cells or MeT-551b-5p cells decreased to less than 0.2 when compared to MeT-ctrl cells. Interestingly, the expression of 122 mRNAs was found to be suppressed (Fig. 7B). The function of these genes was analyzed with the GeneCoDis3 software (http://genecodis.cnb.csic.es/), and it was found that these genes are involved in several pathways (Table III). Among these functions, we focused on genes involved in the binding between cells as well as between the cell and the extracellular matrix, as MPM cells may reduce the adhesion ability of normal mesothelial cells. Invading the space between these cells is important for malignant transformation. A summary of each gene is shown in Table IV.