In this study, 201 NSCLC patients were included. First, the recurrence-specific exosomal miRNAs of 6 NSCLC patients were profiled by miRNA array using a 3D-Gene Human miRNA Oligo chips ver.20 (Toray Industries, Inc., Kamakura, Japan). Next, the usefulness of selected miRNAs as biomarkers was clarified using another 195 NSCLC cases. This study was approved by the Institutional Review Board of the Ethics and Indications Committee of Teikyo University (Tokyo, Japan). Written informed consent was obtained from all patients. All patients underwent curative surgery, and the median follow-up period was 20 months (range, 0.5–37 months). NSCLC patients were diagnosed according to the current World Health Organization histological classification and were staged according to the TNM classification of the International Union Against Cancer (22,23). Patients were treated with a standard regimen (uracil/tegafur for TNM stage I patients and cisplatin plus vinorelbine for TNM stage II and III patients) in accordance with the Guidelines of the Japan Lung Cancer Society (24). Post-operative follow-up was performed as follows: Confirmation of recurrence in all patients was required to evaluate imaging or pathological diagnosis. Physical examination and chest X-ray were conducted every 3 months for 5 years, and computed tomography was repeated every 6 months for 5 years after surgery.

For recurrence-specific miRNA microarray profiling, 6 cases of NSCLC (TNM stage III) were selected from the records of the Department of Surgery, Teikyo University. Tumor recurrence was found in 3 cases of the 6 cases after surgery. Recurrence sites identified were the brain (2 cases) and the bones (1 case). As a normal control, blood samples from 3 healthy volunteers were examined. All blood samples were derived from patients who had not received any chemotherapy or radiotherapy prior to surgery. Plasma samples were separated from the blood, and exosomes were purified from the plasma samples. Exosomal miRNA expression profiles were examined using a 3D-Gene Human miRNA Oligo chips ver.20 (Toray Industries, Inc., Kamakura, Japan). Details of the measuring methods are described as follows.

For the validation analysis, 195 consecutive NSCLC patients and 30 healthy individuals treated between April 2012 and July 2015 at Teikyo University Hospital (Tokyo, Japan) were studied. These NSCLC patients included 129 cases with tumor-node metastases (TNM) stage I, 34 cases with stage II and 32 cases with stage III. All blood samples were derived from patients who had not received any chemotherapy or radiotherapy prior to surgery. Plasma samples were separated from the blood, and exosomes were purified from the plasma samples. Exosomal miRNA expression was analyzed using reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Details of the measuring method are described as follows.

A peripheral blood specimen (5 ml) was drawn from each patient prior to surgery. Ethylenediaminetetraacetic acid was used as an anticoagulant. Plasma and blood cells were separated by centrifugation at 1,200 × g for 10 min at 4°C. Plasma (1.0 ml) was used for microarray analysis, and 500 µl was used for RT-qPCR. The exosomes of the plasma were purified by the ultracentrifugation method, as described previously (21). In brief, the exosomes were separated by ultracentrifugation at 100,000 × g for 70 min at 4°C, and then the pellets were washed with phosphate-buffered saline (PBS) and stored at −80°C for microarray and RT-qPCR analysis.

The isolated exosomes were dissolved in PBS, and a drop of the suspension was placed on a carbon-coated copper grid for 10 sec. The grid was then removed and excess liquid was drained by filter paper. The grid was put in contact with a drop of 2% uranyl acetate and phosphotungstic acid for 5 sec, and excess liquid was then removed. The grid was allowed to dry for several min and was then examined using an electron microscope (HITACHI H-7600; Hitachi Ltd., Tokyo, Japan) at 100 kV.

RNA was isolated from exosomes using the miRNeasy serum/plasma kit (Qiagen, Hilden, Germany). Exosomes purified from specific volumes of plasma (1.0 ml for microarray analysis and 500 µl for RT-qPCR assay) were diluted with 1 ml Qiazol Lysis Reagent. Following incubation for 5 min, 3.5 µl of a spike-in control (cel-miR-39 mimic) was added to each sample. Subsequent extraction and cartridge work were performed according to the manufacturer's protocols. The RNA quality was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA).

Exosomal miRNA expression profiles were examined using 3D-Gene Human miRNA Oligo chips ver. 20 (Toray Industries, Inc. Tokyo, Japan), according to the manufacturer's protocol. Fluorescence signals were scanned and analyzed using the 3D-Gene Scanner (Toray Industries, Inc.). A total of 2578 genes were mounted on this chip. The raw data from each spot were normalized by subtraction of the background signal mean intensity, determined by the 95% confidence intervals of the signal intensities of all blank spots. Valid measurements were considered those in which the signal intensity of the duplicate spots was >2 standard deviations of the background signal intensity.

Exosomal miRNA expression was assayed using RT-qPCR. cDNA was synthesized from total RNA using Taqman microRNA primers specific for hsa-miR-21 (assay ID 000397), has-miR-4257 (assay ID 244369) and has-miR-16a (assay ID 000391) (Thermo Fisher Scientific, Inc., Waltham, MA, USA) and a TaqMan Micro-RNA Reverse Transcription kit (Thermo Fisher Scientific, Inc.). miR-16a was used as an internal control, as it is reported to be a reliable endogenous control for miRNA analysis in RT-qPCR for humans (21). RT-qPCR was performed using Lightcycler-480 (Roche Applied Science, Basel, Switzerland) and Taqman Universal PCR Master Mix (Thermo Fisher Scientific, Inc.) following the manufacturer's protocols. Relative quantification of miRNA expression was calculated using the 2^−ΔΔCq method (25). ΔCq was calculated by subtracting the Cq values of miR-16 from the Cq values of miR-21 or miR-4257. ΔΔCq was then calculated by subtracting the ΔCq of the healthy control from the ΔCq of the sample. For healthy controls, plasma exosome samples of 30 healthy volunteers were used.

The data are expressed as the mean ± standard deviation. The correlations between microRNA expression and clinicopathological factors were analyzed using Student's t-test, the χ² test and an analysis of variance. Disease-free survival (DFS) curves were analyzed using the Kaplan-Meier survival curve method, and the differences were examined using log-rank tests. Cox proportional-hazards regression analysis was used to estimate univariate and multivariate hazard ratios for DFS. All P-values are two-sided, and P<0.05 was considered to indicate a statistically significant difference. Statistical analyses were performed using the JMP 9.0 software (SAS Institute, Inc., Tokyo, Japan).

To verify the isolation of the exosomes from the plasma, transmission electron microscopy results were examined of ultracentrifugation samples from the NSCLC patients. Images of round microvesicles with diameters of 50–100 nm were captured, as shown in a representative sample in Fig. 1.

To identify the recurrence-specific miRNA profiling in the plasma exosomes of the NSCLC patients, the miRNA microarray analyses of samples from 3 patients with recurrence after surgery, 3 patients without recurrence after surgery and 3 healthy volunteers were examined. The clinical pathological characteristics of the 6 NSCLC patients used for this analyses are shown in Table I. In these patients, the histological type and the TNM stages were the same in the recurrence and recurrence-free cases. In the 3 recurrence cases, the recurrent sites were the brain (2 cases) and the bones (1 case).

Table II shows the five most highly upregulated exosomal miRNAs in the patients with recurrence. In this analysis, miR-4257 (miRBase no. MIMAT0016878) showed the highest upregulation in the patients with recurrence compared with those without recurrence and the healthy controls. The mean fold-change of miR-4257 was 2.1 in comparison with that of the recurrence-free patients, and 2.9 in comparison with that of the healthy controls. miR-21 (miRBase no. MIMAT0000076) showed the highest level after miR-4257. The mean fold-change of miR-21 was 1.7 in comparison with that of recurrence-free patients, and 2.5 in comparison with that of the healthy controls. On the basis of these profiling results, miR-4257 and miR-21 were selected as the recurrence markers for NSCLC cases and were used in the present study.

The expression of miR-21 and miR-4257 was assessed by RT-qPCR in the plasma exosomal samples from 195 patients with NSCLC and 30 healthy volunteers. As shown in Fig. 2A and B, exosomal miR-21 and miR-4257 levels were significantly increased in NSCLC patients as compared with those of healthy controls (P<0.01).

This study includes 195 NSCLC patients (112 men and 83 women), with a mean age of 71 years (range, 32–88 years). These patients were drawn from a separate cohort to the miRNA microarray study. To evaluate the correlation between the exosomal miR-21 and miR-4257 expression levels and the clinicopathological characteristics, patients were divided into two groups (high and low). The cut-off level was determined as the average of the miR-21 and miR-4257 expression levels, as described previously (21). As shown in Table III, a statistically significant association was observed between the high miR-21 group and tumor size and TNM stage. A statistically significant association was observed between the high miR-4257 group and lymphatic invasion, histological type and TNM stage (Table IV).

The Kaplan-Meier DFS curves of the 195 NSCLC patients (TNM stage I, 129 cases; TNM stage II, 34 cases; TNM stage III, 32 cases) according to the status of miR-21 and miR-4257 levels were examined. Patients who had undergone curative surgery were included in this analysis.

The DFS of patients in the high miR-21 group showed a significantly worse survival rate than those in the low miR-21 group (log-rank, P=0.016; Fig. 3). The DFS analysis revealed that patients in the high miR-4257 group exhibited significantly worse survival rates than those in the low miR-4257 group (log-rank, P=0.005; Fig. 4). These results suggest that the high expression of miR-21and miR-4257 is associated with recurrence in NSCLC patients.

Table V shows the results of univariate and multivariate Cox proportional hazard regression analysis for DFS in NSCLC patients (n=195). Multivariate analysis was performed for all factors tested by univariate analysis. In the univariate analysis, tumor size, lymphatic invasion, TNM stage, and miR-21 and miR-4257 levels showed significance for DFS. In the multivariate analysis, TNM stage, and miR-21 and miR-4257 levels showed significance for DFS. These results suggest that miR-21 and miR-4257 levels in tumor tissues have an independent prognostic value for DFS in NSCLC patients.