Exosomal miRNA signatures of pancreatic lesions

A series of 155 plasma samples obtained from 58 PDAC, 42 PanNETs, 20 IPMN, 20 AVC (Additional file 1), and 15 CP patients were retrieved from the archives of the ARC-NET biobank of the Verona University Hospital. All experiments were performed in accordance with relevant guidelines and regulations. Clinicopathological features of PanNET, PDAC and AVC patients are reported in Table 1. Regarding IPMN samples, only 6 of 20 displayed associated invasive pancreatic ductal adenocarcinoma. All peripheral fasting blood samples were collected in 10-mL sodium heparin tubes (BD Vacutainer; Becton-Dickinson, Milan, Italy) and processed within 2 h by centrifugation at 2000 g at 4 °C for 10 min. The plasma obtained was then transferred to new tubes and centrifuged at 3000 g at 4 °C for 10 min to remove platelet contamination. Final plasma preparations were carefully collected from the upper portion of the supernatant and stored in aliquots at − 80 °C.

Plasma-derived exosomes were isolated as described in Casadei et al. [33] by using the ExoQuick system (System biosciences) according to the manufacturer’s protocol. The quality and size of particles were determined by Nanosight, qRT-PCR and western blot analyses (Additional file 2). The exosomal pellet was resuspended in 1 ml Trizol (Life Technologies), and total RNA was isolated with Norgen RNA clean-up and concentration kit (Norgen BioTek), following the provided instructions. Exosomal RNA was eluted in 20 μl of H₂O; the average RNA yield per group per 1 ml of plasma was: for the CP group ~ 4 ng/ml, for IPMN ~ 6 nm/ml, for PDAC ~ 6.3 ng/ml, for AVC ~ 20 ng/ml, for PanNET ~ 4.7 ng/ml. Nanodrop 1000 spectrophotometer (Thermo Scientific) was used to exclude within our samples the presence of 230 nm peaks, due to organics compounds or chaotropic salts, which could represent a potential issue for the ligase enzymatic step during NanoString profiling.

Concentration and particle size distribution in plasma-derived exosomal samples were determined by NTA using NanoSight NS300 system (Malvern Technologies, Malvern, UK). Nanosight was configured with a 532 nm green laser, a high sensitivity scientific CMOS camera, and samples were loaded by using a syringe. Samples were diluted in particle-free PBS (Sigma) to acceptable concentrations based off of the manufacturer’s recommendation. Sample data files were acquired under continuous flow (flow rate = 25–50) and at room temperature. Samples were recorded for 3 × 60 s successive videos and were captured with a camera level of 12. Data were analyzed using NTA 3.1.5 software with a detection threshold of 3–4.

A total of 155 samples were processed with Nanostring nCounter® Human v3 miRNA Expression Assay. NanoString analysis was performed as described in Drusco et al. [34]. Briefly, 3.5 μL of exosomal RNA were annealed with multiplexed DNA tags (miR-tag) and bridges target specifics. Mature miRNAs were then bond to specific miR-tags using a Ligase enzyme and all the tags in excess were removed by enzyme clean-up step. The tagged microRNAs product was diluted 1 to 5 and 5 μL were combined with 20 μL of Reported Probes in hybridization buffer and 5 μL of Capture probes. The overnight hybridization (16 to 20 h) at 65 °C allowed to complex probes sequence specific with targets. Probe excess was removed using two-step magnetic beads-based purification on an automated fluidic handling system (nCounter Prep Station) and target/probe complexes were immobilized on the cartridge for data collection. The nCounter Digital Analyzer collected the data by taking images of immobilized fluorescent reporters in the sample cartridge with a CCD camera through a microscope objective lens. For each cartridge, a high-density scan encompassing 325 fields of view was performed. Images were processed internally into a digital format (RCC files). NanoString raw data were analyzed with nSolver™, a software provided by NanoString Technologies. Negative controls were used to perform background subtraction. Positive controls were used to perform technical normalization to adjust any lane-by-lane variability due to differences in hybridization, purification or binding. After technical normalization, the data were biologically normalized by calculating the geometric mean of the top 100 miRNAs in all samples, as recommended by NanoString. p-values were calculated using the LIMMA package (Linear Models for Microarray Data) from the Bioconductor R project. The p-values were adjusted for multiple testing using the Benjamini and Hochberg method [35] to control the False Discovery Rate (FDR). For each pairwise analysis, we considered all those significant (p-value< 0.05) and expressed (≥20 counts in at least one condition, as 20 counts is approximately the average of the expression of the negative controls in the NanoString panel) deregulated miRNAs for the downstream analysis.

The expression of an exosomal individual mature miRNA was assessed in triplicate by using the TaqMan Stem-loop miRNA assay, according to the manufacturer’s protocol (ThermoFisher). For retro-trancription, 0.5 ng of RNA were used per reaction in a final volume of 7.5 μl. For qRT-PCR, 1 μl of cDNA was used per reaction. For RNA normalization, ath-miR-159a, osa-miR-414, and cel-miR-248 synthetic oligos (Integrated DNA Technologies) were added to each sample right after the exosomal pellet was resuspended in Trizol. Normalization was performed with the 2^-Δct method. Results were analyzed by using a two-tailed Student t test.

For the western blot analysis, exosomal proteins were isolated from 500 μl of plasma (the yield of exosomal protein was 9 mg/sample). The exosomal pellet was resuspended in RIPA buffer (Cell Signaling technology), supplemented with phosphatase and protease inhibitors (Roche), and incubated in ice for 20′. Samples were then harvested at 14,000 x g for 10′, and the supernatant collected in a new eppendorf. Protein concentration was determined by using Bradford Assay (Bio-Rad), following the manufacturer’s instructions. 80 μg of exosomal lysate were then loaded on a Criterion Tris-HCl 4–20% pre-cast gel (Bio-Rad), transferred onto a nitrocellulose membrane (Bio-Rad) and probed with anti-Alix (1:1000), anti-TSG101 (1:1000), anti-Calnexin (Sigma) (1:2000), and anti-CD9 (Cell Signaling Technology) (1:1000) primary antibodies, followed by isotype matched, horseradish-peroxidase-conjugated secondary antibodies. Finally, the proteins of interest were detected through chemi-luminescence reaction.

Locked nucleic acid (LNA) probes with complementarity to miR-4454, miR-106a-5p, and miR-17-5p were labelled with 5′-biotin and synthesized using Exiqon (Vedbaek, Denmark). Tissue sections were digested with ISH protease 1 (Ventana Medical Systems, Milan, Italy) and ISH was performed as we previously described [36]. Positive (U6; Exiqon) and negative scrambled LNA probes (Exiqon) were used as controls. Only cytoplasmic miRNA staining was retained for scoring purposes.

Comprehensive miRNA profiling was performed in order to identify exosomal miRNAs differently expressed between pancreatic lesions (AVC, IPMN, PDAC and PanNET) and chronic pancreatitis (CP). Overall, we found 26, 23, 40 and 45 deregulated miRNAs between AVC vs CP, IPMN vs CP, PDAC vs CP and PanNET vs CP, respectively (Fig. 1 a-b-c-d). For each comparison, a linear fold change > 1.5 was used as threshold. Next, relevant miRNAs were filtered again considering only those with a number of counts greater than 20 (See Methods section, Table 2 and Additional file 3). In details, we found 5 deregulated miRNAs (3 upregulated and 2 downregulated miRNAs) between AVC and CP; 4 miRNAs between IPMN and CP (3 upregulated and 1 downregulated miRNAs); 9 miRNAs between PDAC and CP (3 upregulated and 6 downregulated miRNAs) and 11 miRNAs between PanNET and CP (6 upregulated and 5 downregulated miRNAs) (Table 2 and Additional file 3).

Seven miRNAs were therefore selected for orthogonal validation by qRT-PCR on the same cohort of patients: miR-106-5p, miR-7975, miR-4454, miR-16-5p, miR-25-3p, miR-320e and miR-451a (Fig. 1 e-k and Table 2). In particular, since the NanoString assay is not able to discriminate between miR-106a-5p/miR-17-5p and miR-4454/miR-7975, we decided at first to perform a qRT-PCR analysis for these four miRNAs in order to understand their specific expression contribution within selected sample groups. At first, we found a significant upregulation of miR-106-5p in AVC and IPMN compared to CP (Fig. 1e) while miR-17-5p was not differentially deregulated between the same lesions (data not shown). The expression of both miRNAs was also investigated between PanNET and CP showing no different between groups (data not shown). Than we validated the downregulation of miR-7975 and miR-4454 in PDAC compared to CP while only miR-4454 was found downregulated in the comparison between PanNET and CP (Fig. 1 f-g). The upregulation of miR-16-5p was also validated by qRT-PCR in IPMN groups, while the upregulation of miR-25-3p and miR-451a and the downregulation of miR-320e were validated as deregulated in PanNET as compared to CP (Fig. 1 i-j-k).

Differential miRNA profiling was also performed in order to identify relevant deregulated miRNAs between different pancreatic lesions (AVC vs PDAC, IPMN vs PDAC and PanNET vs PDAC). As reported in Fig. 2, we were able to identify 63 deregulated miRNAs between AVC and PDAC, 54 miRNAs between IPMN and PDAC and 16 miRNAs between PanNET and PDAC (Fig. 2 a-b-c). A linear fold change> 1.5 was applied as threshold for each comparison. After data filtering, we found 22 upregulated miRNAs in AVC compared to PDAC, 21 upregulated miRNAs in IPMN compared to PDAC and 7 deregulated miRNAs between PanNET and PDAC (4 upregulated and 3 downregulated miRNAs in PanNETs compared to PDAC) (Table 2 and Additional file 4). Selected miRNAs were than validated by qRT-PCR between different pancreatic lesions as reported in Additional file 3. Consistent with NanoString data, we found miR-17-5p and miR-106-5p significantly upregulated in AVC and IPMN as compared to PDAC (Fig. 2 d-e and Table 2). The upregulation of miR-16-5p and miR-19-3p in IPMN compared to PDAC was also validated as reported in Fig. 2 (f-g) and Table 2.

We also focused on the most deregulated circulating miRNAs associated with different biological and clinical subtypes of the same histotype: metastatic PDAC Vs localized PDAC (Met. PDAC vs Loc. PDAC), pancreatobiliary AVC vs intestinal AVC (PB AVC vs INT AVC) and IPMN with associated invasive adenocarcinoma and IPMN without associated invasive adenocarcinoma (IPMN-C VS IPMN) (Fig. 3). Using a linear fold change > 1.5 as threshold, we identified overall 9 deregulated miRNAs between metastatic and localized PDAC, 11 miRNAs between PB AVC and INT AVC and 12 miRNAs between malignant IPMN and benign IPMN (Fig. 3 a-b-c). As for previous analysis, data were filtered again leading to the identification of 9 downregulated miRNAs in metastatic PDAC compared to the localized one, 11 deregulated miRNAs between pancreatobiliary and intestinal IPMN (6 upregulated and 5 downregulated miRNAs) and 12 miRNAs between malignant and benign IPMN (10 upregulated and 2 downregulated miRNAs) (Table 2 and Additional file 4).

To further support our findings, we performed in situ hybridization (ISH) assay on matched formalin-fixed paraffin-embedded (FFPE) tissue sections of CP, IPMN, AVC, PDAC patients and normal pancreas for miR-4454, miR-106a-5p and miR-17-5p (Fig. 4). Despite ISH analyses lack the required sensitivity to identify subtle changes in expression levels of miRNA, they largely reflected qRT-PCR results. MiR-4454, which was significantly upregulated in CP compared to PDAC through qRT-PCR, showed the same expression trend by ISH (Fig. 4). MiR-106-5p and miR-17-5p qRT-PCR data also well matched with ISH experiments. As reported in Fig. 4, miR-106-5p was more expressed in IPMN and AVC tissue section as compared to PDAC and CP, while the expression of miR-17-5p was more evident in AVC and IPMN compared to PDAC.