Cell lines used in the present study are listed in Table I. Cells were maintained in a humidified atmosphere (37°C, 5% CO₂) in RPMI-1640 medium (Sigma-Aldrich; Merck KGaA, Darmstadt, Germany) supplemented with 2 mM L-glutamine (Nissui Pharmaceutical, Co., Ltd., Tokyo, Japan), 100 U/ml penicillin-100 mg/ml streptomycin, and 10% heat-inactivated (FBS) (all from Thermo Fisher Scientific, Inc., Waltham, MA, USA).

Real-time cell analysis (RTCA) of in vitro cell migratory and invasive activities was performed using an xCELLigence RTCA DP instrument (Roche Diagnostics, Indianapolis, IN, USA) as previously described (15). Samples were analyzed in quadruplicate as technical replicates. Data analysis was performed using the RTCA software (version 1.2) supplied with the instrument.

Exosomes were isolated from the cell culture medium as previously described (15,16). Briefly, cells were cultured for 48 h at 37°C with 5% CO₂ in complete RPMI-1640 medium containing 10% FBS depleted of contaminating microvesicles by centrifugation at 100,000 × g for 18 h. Culture medium (CM) was collected and centrifuged at 800 × g for 5 min and at an additional 2,000 × g for 10 min to remove detached cells. The supernatant was then filtered through a 0.1-µm pore polyethersulfone membrane filter (Thermo Fisher Scientific, Inc.) to remove cell debris and large vesicles, then concentrated using a Centricon Plus-70 with a 100,000-MW cut-off membrane (EMD Millipore; Billerica, MA, USA). Concentrated CM was ultracentrifuged at 100,000 × g for 2 h at 4°C using a 70Ti rotor (Beckman Coulter, Inc., Brea, CA, USA). Resultant pellets were resuspended in 6 ml phosphate-buffered saline (PBS) and ultracentrifuged at 100,000 × g for 1 h at 4°C using a 100Ti rotor (Beckman Coulter, Inc.).

Exosomal proteome analysis was performed by LC-MS/MS (liquid chromatography-mass spectrometry) as previously described (15,17). Proteins (200 µg) from isolated exosomes were dissolved in lysis buffer containing 7.5 M urea and 2.5 M thiourea (both from Sigma-Aldrich; Merck KGaA), 12.5% glycerol (Chemical Industries, Osaka, Japan), 50 mM Tris, 2.5% n-octyl-β-d-glucoside, 6.25 mM Tris(2-carboxyethyl)phosphine hydrochloride, and 1.25 mM protease inhibitor (all from Sigma-Aldrich; Merck KGaA) before being rotated at 4°C for 60 min. After centrifugation at 14,000 × g for 60 min at 4°C, the supernatant was fractionated using the Agilent 1200 HPLC system (Agilent Technologies, Inc., Santa Clara, CA, USA) with an Intrada WP-RP column (0.46×25 cm, 3-µm particle size and 30-nm pore size; Imtakt, Kyoto, Japan). Collected fractions were digested with trypsin (Promega Corp., Madison, WI, USA) and analyzed by LC-MS/MS using a nanoflow LC-ESI linear ion trap-TOF NanoFrontier L mass spectrometer (Hitachi High-Technologies, Tokyo, Japan). Raw LC-ESI data were converted to peak list files using NanoFrontier L Data Processing software (Hitachi High-Technologies). The peak list files were used for protein identification with the MASCOT MS/MS ion search (http://www.matrixscience.com) and X! Tandem software (http://www.thegpm.org). Upon peptide sequence annotation, the UniProtKB/Swiss-Prot database (version 2016_10; Homo sapiens; http://www.uniprot.org/statistics/Swiss-Prot) was used with the following parameters: enzyme, trypsin or none (when used with the home-made dataset only); maximum number of missed cleavage, 1; peptide tolerance, 0.2 Da; MS/MS tolerance, 0.2 Da; variable modification, oxidation of methionine; and peptide charge, (1+, 2+ and 3+). All identified proteins with MASCOT threshold scores < 95% confidence level and peptide numbers <2 were then removed from the protein list using Scaffold software (http://www.proteomesoftware.com/products/scaffold/).

Exosomes and cells were lysed with 7.5 M urea-based lysis buffer as described above. Protein concentrations were determined by the Bradford assay (Bio-Rad Laboratories, Inc., Hercules, CA, USA). Proteins (5 or 10 µg) were subjected to 8% SDS-polyacrylamide gel electrophoresis (SDS-PAGE), and transferred onto an Immobilon-P polyvinylidene fluoride (PVDF) membrane (0.45-mm pore size; EMD Millipore). PVDF membranes were blocked for 1 h at room temperature in Tris-buffered saline (10 mM Tris-HCl, pH 7.5, 150 mM NaCl) containing 0.01% Tween-20 and 5% non-fat dried milk (Wako Pure Chemical Industries). Blocked membranes were then incubated overnight at 4°C with primary monoclonal antibodies (listed in Table II). Membranes were then incubated for 1 h at room temperature with anti-mouse IgG antibodies conjugated with horseradish peroxidase (Table II). Specific proteins were visualized using an ECL Plus western blotting detection system (GE Healthcare, Wauwatosa, WI, USA) and a Fujifilm Luminescent Image Analyzer LAS3000 (Fujifilm, Tokyo, Japan). The molecular weight of each protein was deduced using Precision Plus Protein All Blue Standards (Bio-Rad Laboratories, Inc.).

Cells were cultured for 48 h, and total RNA was extracted using the miRNeasy Mini Kit (Qiagen, Hilden, Germany) as previously described (15). RNA samples were quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific, Inc.) and assessed using an Agilent 2100 Bioanalyzer and an RNA 6000 Nano Total RNA kit (both from Agilent Technologies, Inc.). Quantitative mRNA levels were determined using real-time RT-PCR using the Applied Biosystems 7900 HT Sequence Detection System, a TaqMan Gene Expression Assay for human EPS8 (assay ID Hs00610286_m1), and a Eukaryotic 18S rRNA Endogenous Control (Applied Biosystems; Thermo Fisher Scientific, Inc.). Only the probe sequence for EPS8 (TTGGATGAAAGCCAGAGCAGAGTGG) was provided by the manufacturer. The probes of EPS8 and 18S rRNA were labelled with FAM and VIC dyes, respectively. cDNA was generated using 100 ng of total RNA, and a High Capacity cDNA Reverse Transcription kit (Applied Biosystems; Thermo Fisher Scientific, Inc.). RT-PCR was carried out in a total volume of 20 µl containing 100 ng of cDNA, TaqMan Fast Advanced Master Mix (Applied Biosystems), and the respective TaqMan target gene reagents. The amplification conditions were 95°C for 20 sec followed by 40 cycles of 95°C for 1 sec and 60°C for 20 sec. Samples were analyzed in triplicate as technical replicates. The EPS8 mRNA levels were defined from the cycle threshold (Ct), using the comparative Ct method (18), and each sample was normalized by comparison to 18S rRNA levels. The fold change of EPS8 mRNA levels in each cell line was determined using SU.86.86 cell EPS8 mRNA levels as a reference.

Student's t-test for comparison of cell motility between SU.86.86 and MIA PaCa-2 cell lines, the Pearson correlation coefficient to compare two variables in five analyses (exosomal Eps8 protein, intracellular Eps8 protein, intracellular Eps8 mRNA, migration, and invasion), and Multiple t-tests with Bonferroni-correction for comparison of three different cell origins of metastasis, ascites and primary tumors were used. P-values <0.05 were considered to indicate a statistically significant difference.

SU.86.86 cells were derived from a liver metastasis of a pancreatic ductal carcinoma (19). MIA PaCa-2 cells were derived from a primary pancreatic adenocarcinoma (20). Before proteome analysis, we first performed in vitro cell migration and invasion assays to evaluate if the two cell lines exhibited differences in metastatic-potential. The impedance-based RTCA has shown a strong correlation with the conventional Boyden chamber Transwell endpoint assay (15,21). Using the RTCA assay system, SU.86.86 cells had 23-fold greater cell migratory activity than did MIA PaCa-2 cells (Fig. 1A). Additionally, using a Matrigel barrier, SU.86.86 cells were found to be 20-times more invasive than MIA PaCa-2 cells (Fig. 1B). Collectively, these results indicated that the in vitro cell migratory and invasive behaviors of SU.86.86 and MIA PaCa-2 cells were correlated with their metastatic and primary tumor cell origins, respectively.

The proteome profiles of SU.86.86 and MIA PaCa-2 cell-derived exosomes were analyzed using LC-MS/MS. After 48 h of cell growth, exosomes were isolated from culture media by a series of filtration and ultracentrifugation steps as previously described (15,16). Proteome data processing identified a total of 133 proteins from exosomes derived from both cell lines (Fig. 2A, Table III). Among them, 31 proteins were identified in the exosomes of both cell types. A total of 101 proteins were uniquely identified in SU.86.86 cell-derived exosomes, and a single unique protein, histone H2A type 2-B (H2A2B), was identified in MIA PaCa-2 cell-derived exosomes.

To identify the SU.86.86 cell-specific exosomal proteins, we compared the 101 SU.86.86 cell-specific proteins with those that had been previously identified in exosomes derived from human duodenal cancer cell lines AZ-521 and AZ-P7a (Fig. 2A) (15). This comparison identified 82 proteins that were unique to SU.86.86 cell-derived exosomes (Table IV). Of the 82 proteins unique to SU.86.86, Eps8 revealed relatively high MS/MS values, including the number of matched peptides, the rate of sequence coverage, and the total spectral count. Furthermore, the Eps8 expression was elevated in pancreatic cancer cells derived from ascites and metastasis (22). Therefore, we chose to validate the presence of Eps8 in exosomes by western blot analysis.

Western blot analyses revealed that the Eps8 protein was abundant in SU.86.86 cell-derived exosomes, while no immunoreactive Eps8 signals were detected in MIA PaCa-2 cell-derived exosomes (Fig. 2B). Furthermore, intracellular Eps8 expression levels were much higher in SU.86.86 cells than in MIA PaCa-2 cells, indicating a positive correlation with tumor malignancy, as previously reported (22). Comparison of the house-keeping proteins in the exosomes of both cell types revealed variation in glyceraldehyde 3-phosphate dehydrogenase (GAPDH) levels and no immunoreactive signals for α-tubulin, making them unsuitable for normalizing exosomal protein levels previously described (15).

Eps8 was specifically detected in exosomes from metastatic-derived SU.86.86 cells. Therefore, we assessed exosomal Eps8 protein levels in other pancreatic cancer cell lines. Western blot analysis revealed positive immunoreactive Eps8 signals in exosomes from metastasis-derived pancreatic cancer cell lines, including CFPAC-1, KP-3, PK-45H, PK-8 and Capan-1 (Fig. 3A). Additionally, positive immunoreactive Eps8 signals were observed in exosomes from ascites-derived pancreatic cancer cell lines, including HuP-T3, HuP-T4 and AsPC-1. In contrast, Capan-2 was the only primary tumor cell line that exhibited Eps8 immunoreactivity. Densitometric analysis, using relative amounts of exosomal Eps8 protein, was used to quantify the Eps8 immunoreactivities observed (Fig. 4). The level of Eps8 immunoreactivity in the exosomes of different cell lines was assessed relative to that observed in SU.86.86 cell-derived exosomes, which was given a value of 1.0. The relative Eps8 immunoreactivity was 0.76, 0.15 and 0.17 in metastatic cell lines PK-45H, CFPAC-1 and KP-3, respectively. In ascites-derived cell lines HuP-T3, HuP-T4, and AsPC-1, and the Capan-2 primary tumor cell line, the relative Eps8 immunoreactivity was 0.49, 0.21, 0.15 and 0.55, respectively. Intracellular Eps8 levels varied among the cell lines, particularly those derived from metastasis (PK-1, PK-8, PK-59 and KLM-1) and primary tumors (MIA PaCa-2, BxPC-3, PANC-1 and Panc 10.05). Except for PK-8, there was either no Eps8 immunoreactivity, or less Eps8 immunoreactivity, relative to that observed intracellularly, observed in the exosomes of these cells. Also, distinct Eps8 immunoreactivity was observed in NUGC-4 and MKN45P stomach cancer cell lines and the LoVo colon cancer cell line (Fig. 3B). Furthermore, cells with relatively high intracellular Eps8 protein levels expressed greater amounts of EPS8 mRNA (Fig. 4). However, Eps8 protein and mRNA expression levels did not correlate with the amount of Eps8 in exosomes. Collectively, these results revealed that, particularly in pancreatic cancer cells derived from metastasis and ascites, Eps8 was secreted into the extracellular environment via exosomes.

It was revealed that Eps8 protein is present in the exosomes of several pancreatic cancer cell lines in addition to SU.86.86. Therefore, we evaluated the in vitro cell migratory and invasive activities in 12 pancreatic cancer cell lines. The highest levels of migratory and invasive activities were observed in SU.86.86 cells, and these were set at a value of 1 to allow for comparison (Fig. 5). The metastatic PK-45H cell line, with the relative exosomal Eps8 protein level of 0.76, revealed relatively high levels of cell migratory (0.64) and invasive activities (0.56). Additionally, in ascites-derived HuP-T4 cells, with the relative exosomal Eps8 protein level of 0.21, relatively high levels of cell migratory (0.36) and invasive activities (0.84) were observed. However, no cell motility was detected in stomach cancer-derived NUGC-4 cells and colon cancer-derived LoVo cells, which exhibited moderate levels of exosomal Eps8 immunoreactivities (Fig. 3).

Integrative comparison of the data revealed that similar to SU.86.86 cells, PK-45H cells consistently had the highest levels of in vitro cell migratory and invasive activities, exosomal and intracellular Eps8 protein, and EPS8 mRNA expression (Fig. 6A and B). Furthermore, using the Pearson correlation coefficient, we identified that exosomal Eps8 levels were significantly correlated with migratory cell levels (r=0.85, P=4.2×10⁻⁴) (Fig. 6C). Therefore, we proposed that exosomal Eps8 protein level is indicative of metastatic potential in human pancreatic cancer cells.