Forty primary pancreatic cancer (PC) tissues and 40 normal were collected at the First Affiliated Hospital of Wenzhou Medical University from 01/03/2017 to 03/08/2018. All specimens from resection surgery were frozen and stored in liquid nitrogen. This study was performed with the approval of the Ethic and Research Committees of the First Affiliated Hospital of Wenzhou Medical University and these studies was conducted in accordance with the Declaration of Helsinki Principles. Informed written consent was obtained from all subjects.

Human NK cells were isolated as described before using NK cell isolation kit (Miltenyi Biotech, Bergisch Gladbach, Germany) according to the manufacturer's instruction (19). The PBMC were collected according to the protocols approved by the First Affiliated Hospital of Wenzhou Medical University.

Human PC cells (Mia PaCa-2 and PANC-1) were purchased from the American Type Culture Collection (Rockville, USA). The cells were cultured in RPMI 1640 or DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 mg/mL streptomycin at 37°C under 5% CO₂ in a humidified chamber.

MiR-3607-3p mimics, inhibitor, and negative control were purchased from GenePharma (Shanghai, China). The transfection was performed using Lipofectamine 2000 (Invitrogen, Waltham, MA USA) following the manufacturer's instructions. The knockdown efficiency at the mRNA level was assessed using qRT-PCR assay.

PC cells were seeded in 96-well plates at a density of 5 × 10³ cells/well. At the indicated time points, viable cells were examined by cell counting kit-8 (CCK-8) according the manufacturer's instructions.

Cell migration and invasion ability were measured using transwell chambers (24-23ll insert, 8 μm, Millipore, Billerica, MA, USA). The chambers were coated with Matrigel Matrix (Sigma, St. Louis, MO). PC cells were seeded at a density of 5 × 10⁴ cells/well in upper chamber with 200 μL serum free medium. The lower chamber was filled with 600 μL of medium containing 10% FBS. After incubation for 48 h, the number of cells passing through the bottom membrane of the chamber was counted under a microscope.

PC cells transfected with NC-miRNA or miR-3607-3p were seeded in a 24-well plate at a density of 5 × 10⁴ cells/well and cultured for 24 h. Then, the cells were collected and seeded (1,000–1,500 cells/well) in a dish for 10 days. Surviving colonies were fixed, stained with 5% gentian violet (Sigma) and counted. The experiment was carried out in triplicate wells for three times.

Total protein was extracted from the PC cells using RIPA buffer (Invitrogen) according to the manufacturer's protocol. Equivalent amounts of proteins from each sample were subjected to SDS-PAGE electrophoresis and then transferred to a PVDF membrane (Millipore), blocked in 5% fat-free milk for 1 h at room temperature, incubated with specific primary antibodies overnight at 4°C with gentle shaking, and followed by detection with enhanced chemiluminescence system (Pierce, Waltham, MA). Primary antibodies were as follows: CD63, TSG101, IL-26, Actin (Abcam, Cambridge, MA).

Total RNA from PC tissues and cell lines was harvested using the TRIzol reagent (Invitrogen) following the manufacturer's instructions. RNAs were reverse-transcribed into cDNAs using the Prime Script RT reagent Kit (TaKaRa, Dalian, China) according to the manufacturer's protocol. MiR-3607-3p RT primer: GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACCATCA; U6 (internal control) RT primer: GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAAAATA TGGAA. SYBR Taq (TaKaRa) was used to measure the expression of miR-3607-3p and IL-26. The primers used were described below: miR-3607-3p Forward: ATGACTGTAAACGCTTTCTG, Reverse: GTGCAGGGTCCGAGGT;

U6 Forward: TGCGGGTGCTCGCTTCGGCAGC, Reverse: GTGCAGGGTCCGAGGT; IL-26, Forward: AAGCAACGATTCCAGAAGACC; Reverse: AAGTCCTCCACAAAGCGTATTTT; GAPDH, Forward: ACAACTTTGGTATCGTGGAAGG; Reverse: GCCATCACGCCACAGTTTC. Fold changes of miR-3607-3p and IL-26 were calculated by the equation 2^−ΔΔCt.

The dual luciferase reporter assay was performed using Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA). Briefly, luciferase reporter vector containing the wild-type (WT) and mutant (MT) miR-3607-3p binding site in the 3′-UTR of IL-26 were constructed transiently transfected into MIA PaCa-2 or PANC-1 cells, together with miR-3607-3p mimics or NC using Lipofectamine 2000. Primers used for cloning WT IL-26 3′UTR, Forward: CTACTCGAGACCAAAGCCAAGTACATT, Reverse: CTAGCGGCCGCGAAGGAAACCCAATTTA; Mutant IL-26 3′UTR, Forward: GCCAAGTACATTGATTCTCCCCT, Reverse: AGGGGAGAATCAATGTACTTGGC. Luciferase activity was measured 48 h later. The renilla luciferase activity were normalized to the firefly luciferase activity in the corresponding well.

Biotin-miRNA pulldown assay was carried out as described before (20); Briefly, PC cells were washed with cold PBS, harvested by a scraper and treated cell lysis buffer. Supernatant was collected and incubated with biotinylated miR-3607-3p or biotinylated control random RNA for 1 h at 30°C. After incubation, Steptavidin MagneSphere Paramagnetic beads (Promega) was added for 3 h at 4°C. The biotin-miRNA/mRNA complex was eluted and washed for mRNA detection.

Extracellular vesicle (EV) isolation and characterization were prepared as described before (21). Purified NK cells were cultured in a serum free medium for 24 h. Cell culture supernatant was centrifuged for 5 min at 500 g and 10 min at 1,500 g to eliminate the cells and debris. EVs were pelleted by ultracentrifuge step at 80, 000 g for 120 min and washed with PBS. EVs were evaluated and characterized by flow cytometry and Scanning Electron Microscopy (SEM). Briefly, isolated EVs were put on a copper grid coated with 0.1% Formvar in chloroform. The grids were stained with 1% (vol/vol) uranyl acetate in ddH2O, and thereafter the EVs were examined immediately by SEM.

Four to six weeks old Male BALB/c nude mice were purchased from SLAC (Shanghai, China). The animal care and experimental protocols were approved by the institutional guidelines of the First Affiliated Hospital of Wenzhou Medical University and by the Animal Care and Use Committee of The First Affiliated Hospital of Wenzhou Medical University. 1 × 10⁷ PC cells were resuspended in 200 μL PBS medium and were subcutaneously injected into the flank of each nude mouse. The tumors were measured weekly and the tumor volume was calculated following the formula length × width²/2. The mice were killed at 6 weeks after inoculation.

All results are shown as mean ± standard deviation (SD) and were analyzed using GraphPad Prism 5 (GraphPad Software, USA) from at least three independent experiments. The differences between groups were analyzed using Two-tail Student's t-test or ANOVA; Chi-squared test was used to analyze the frequency of lung metastases; Pearson correlation analysis was used to analyze the correlation between expressions of two genes; Kaplan–Meier's analysis was used for prognostic analyses. Data were considered statistically significant when p < 0.05.

To test the function of NK cells in the tumor microenvironment, we first cultured NK cells together with pancreatic cancer cell line Mia PaCa-2 and PANC-1. CCK-8 assay indicated that the proliferation of Mia PaCa-2 and PANC-1 cells co-cultured with NK cells was significantly suppressed compared with that of cells cultured alone (Figures 1A,B). Colony formation assay, as shown in Figure 1C, showed that cells co-cultured with NK cells had fewer colonies formed. In line with this, after co-culture with NK cells, Mia PaCa-2 and PANC-1 cells had prominent attenuated cell migration and invasive ability (Figures 1D,E). Furthermore, we employed in vivo tumor xenotransplantation mouse model. In situ tumor-bearing experiments showed that NK cell co-inoculation could inhibit the tumor growth ability of PANC-1 cells in vivo (Figures 1F,G). When intravenous transferred in vivo through tail vein, PANC-1 cells co-transferred with NK cells significantly inhibited lung metastasis, as demonstrated by in vivo luminescence imaging and H&E staining of the lung tissues (Figures 1H,I). In summary, NK cell can significantly inhibit the proliferation and metastasis of pancreatic cancer cells in vitro and in vivo.

To explore the underlying mechanisms by which NK cells inhibit malignant transformation of pancreatic cancer cells, we speculate that NK cells might act on pancreatic cancer cells through EVs. EVs were isolated and identified from the culture medium of NK cells (Figures 2A–C). In addition, the level of miR-3607-3p in NK cells and its EVs was significantly higher than that in Mia PaCa-2 and PANC-1 (Figure 2D). Furthermore, when co-cultured Mia PaCa-2 and PANC-1 with NK cells or NK cells EVs, we found that the levels of miR-3607-3p were significantly up-regulated in co-cultured Mia PaCa-2 and PANC-1 cells (Figures 2E,F). In summary, the results suggest that NK cells might transmit miR-3607-3p to pancreatic cancer cells via secreted EVs.

Next, miRNA-3607-3p level was analyzed in different PC cell lines (AsPC-1, PANC-1, Capan-2, CFPAC-1, SW1990 and Mia PaCa-2) and a normal human pancreatic ductal cell control (hTERT-HPNE). As shown in Figure 3A, the level of miR-3607-3p was significantly lower in PC cell lines compared with that in control cell line. Consistently, the level of miR-3607-3p was significantly lower in PC tissues compared with that in normal tissues (Figure 3B). In addition, PC patients with lymph node metastasis (LNM⁺) further decreased the expression of miR-3607-3p compared to that in lymph node metastasis free (LNM⁻) patients (Figure 3C). Kaplan–Meier's analysis of the correlation between miR-3607-3p expression and the metastasis-free survival of PC patients indicated that low levels of miR-3607-3p were characterized by worse overall survival and metastasis-free survival rate (Figures 3D,E). We also revealed the similar expression pattern of miR-3607-3p in plasma exosomes in PC patients and plasma exosomal miR-3607-3p expression in PC patients with LNM⁺ was significantly lower than that in LNM⁻ PC patients (Figures 4A,B). Our results indicate that miR-3607-3p acts as a tumor suppressor in PC and is likely to be involved in tumor metastasis.

To investigate the function of miR-3607-3p in pancreatic cancer cells, miR-3607-3p mimics was transfected to overexpress miR-3607-3p in PC cell line Mia PaCa-2 and PANC-1. As shown in Figures 5A–D, CCK-8 assay and colony formation assays demonstrated that overexpression of miR-3607-3p inhibited the proliferation of Mia PaCa-2 and PANC-1 in vitro; Transwell assay results showed that overexpression of miR-3607-3p inhibited the migration and invasion of pancreatic cancer cells in vitro (Figures 5E, 4F). Interestingly, NK cells extracellular vesicles (NK EVs) treatment was found to be able to inhibit cell viability (Figures S1A,B), proliferation (Figure S1C), migration (Figures S1D–F) and inhibited IL-26 production (Figure S1G) in both Mia PaCa-2 and PANC-1 cells. In conclusion, miR-3607-3p inhibits the malignant transformation of pancreatic cancer cells.

To explores the mechanism how miR-3607-3p regulates malignant transformation of PC, we performed bioinformatics analysis and Targetscan (http://www.targetscan.org/vert_71/) was used to predict potential downstream targets of miR-3607-3p. IL-26 was predicted as a potential miR-3607-3p target binding to the IL-26 3′UTR (Figure 6A). Luciferase reporter vector containing IL-26 3' UTR (WT) or IL-26 mutated 3' UTR (MT) was cotransfected with miRNA control or miR-3607-3p mimics into MIA PaCa-2 and PANC-1 cells. The dual luciferase reporter assay demonstrated that miR-3607-3p bound to the WT IL-26 3′UTR to inhibit the luciferase activity, but not to the mutant IL-26 3′UTR (Figures 5C, 6B). Furthermore, using the biotin-labeled RNA pulldown technique, we found that miR-3607-3p directly interacted with IL-26 mRNA (Figure 5D). We confirmed that miR-3607-3p inhibited IL-26 expression at both mRNA and protein levels while miR-3607-3p inhibitor enhanced IL-26 mRNA and protein levels in MIA PaCa-2 and PANC-1 cells (Figures 6E–H). Finally, we examined the mRNA levels of IL-26 in 40 pancreatic cancer tissues and 40 normal tissues. IL-26 was highly expressed in pancreatic cancer tissues compared with that in normal tissue (Figure 6I). Consistently, IL-26 mRNA level was significantly higher in tissues from LNM⁺ PC patients that that in tissues from LNM⁻ PC patients (Figure 6J). Correlation analysis showed that there was a significant negative correlation between the expression levels of miR-3607-3p and IL-26 in pancreatic cancer tissues. In summary, IL-26 is a direct target of miR-3607-3p in pancreatic cancer cells.

To test whether the inhibition of malignant transformation of PC by NK cells depends on miR-3607-3p, we cultured Mia PaCa-2 or PANC-1 cells together with NK cells transfected NC-inhibitor or miR-3607-3p-inhibitor, respectively. The miR-3607-3p levels were confirmed by qPCR (Figure 7A). The results showed that Mia PaCa-2 and PANC-1 cells co-cultured with NK cells transfected with miR-3607-3p-inhibitor had similar levels of cell proliferation, colony formation and cell migration/invasion compared with that of cells co-cultured with only NC-inhibitor (Figures 7B–F). Inhibition of cell proliferation, migration and invasion of Mia PaCa-2 or PANC-1 cells co-cultured with NK cells transfected with NC-inhibitor was significantly restored in cells co-cultured with NK cells transfected with miR-3607-3p-inhibitor (Figures 7B–F). Thus, NK cells inhibited malignant transformation of pancreatic cancer cells at least partially depending on miR-3607-3p.

QH was employed by company Singlera Genomics Inc. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.