AsPC-1 and PANC-1 pancreatic cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA). The established cell lines were recently authenticated by a commercial service (Multiplexion GmbH, Heidelberg, Germany). The human primary pancreatic cancer cell line ASAN-PaCa, which has been described previously, was kindly provided by Dr N. Giese (22). To maintain the authenticity of the cell lines, we prepared frozen stocks from the initial stocks, and a new thawed stock was used every three months for experiments. Monthly testing ensured mycoplasma-negative cultures. Cells were cultured under standard conditions in DMEM (PAA Laboratories GmbH; GE Healthcare Life Sciences, Little Chalfont, UK) supplemented with 10% heat-inactivated fetal calf serum (FCS; Sigma-Aldrich; Merck KGaA, Darmstadt, Germany) and 25 mmol/l HEPES (PAA).

Tissue specimens were obtained from patients who had undergone surgery at the Department of General, Visceral and Transplant Surgery, University of Heidelberg, from January 2014 to December 2016. The Ethics Committee of the University of Heidelberg approved the study after receiving written informed consent from the patients. Clinical diagnoses were established by conventional clinical and histological criteria. Surgical resection was performed as indicated by the principles and practice of oncological therapy.

Stock solutions of DEX (25 mM, ≥98% pure), Sigma-Aldrich; Merck KGaA) were prepared in ethanol. A solution of 5′AZA-2′-deoxycytidine was freshly diluted with the cell culture medium to prepare a 10 µM stock solution (23).

Total RNA was extracted with the use of a Qiagen miRNeasy Mini Isolation Kit (Qiagen, Hilden, Germany). Agilent Human miRNA Microarray (Release 19.0) covering 2006 human microRNAs was used for microRNA profiling. The raw array data were analyzed by the Linear Regression Model for Microarrays (LIMMA) software version 3.24.15. The Benjamini and Hochberg (BH) algorithm was used to correct for multiple testing and false discovery rates (FDRs). Gene Ontology analysis was used to detect associations between the identified set of miRNA candidates and the associated specific biological processes and molecular functions. The microarray data were uploaded to Array Express (https://www.ebi.ac.uk/arrayexpress/help/accession_codes.html) under the accession number E-MTAB-4718.

Gene Ontology (www.geneontology.org) was used to analyze the miRNA array data for identification of a set of significantly and differentially regulated miRNAs with similar functions. The 100 most significantly deregulated miRNAs after 96 h of DEX treatment were uploaded, and the keywords EMT, Wnt and SMAD were entered. The commercially available Ingenuity Pathway Analysis (IPA) database (Qiagen) was used for identifying targets of the top differentially regulated miRNAs. In addition, the databases miRanda (http://www.microrna.org) (24), TargetScan (http://www.targetscan.org) (25), miRwalk (http://mirwalk.uni-hd.de/) (26) and PicTar (http://www.pictar.org) http://www.pictar.org (27) were used for miRNA target gene prediction. Potential candidates with mirSVR scores <−0.1 were considered. We then chose the most common predictions of the different databases; the results from miRwalk and IPA were preferred as their data were the most up-to-date. The binding site of miR-132 in the 3′UTR of TGF-β2 was identified with TargetScan.

cDNA was synthesized from 500 ng total RNA using TaqMan^® Reverse Transcription Reagents (Thermo Fisher Scientific, Inc.) according to the manufacturer’s instructions, and 1 µl cDNA was used for quantitative PCR. Quantitative PCR was performed using a TaqMan Assay with specific primers. RNU6 served as an internal control for hsa-miR-132-3p expression, and GAPDH was the internal control for vimentin and E-cadherin expression (Thermo Fisher Scientific, Inc.). The PCR was performed using a StepOne Real-Time PCR System (Thermo Fisher Scientific, Inc.), and changes in relative concentration were calculated with the second derivative maximum method 2^−ΔΔCq (28). The ΔCT value was calculated by subtracting the CT of the housekeeping gene from the CT of the gene of interest [ΔCT = Ct (gene of interest) - Ct (housekeeping gene)]. Fold change data were generated using the equation ΔΔCT = ΔCT (treated sample) - ΔCT (untreated sample).

An ELISA-based MethylFlash Methylated DNA 5-mC Quantification Kit (Colorimetric; Epigentek, distributed by BioCat, Heidelberg, Germany) was used for quantification of total 5-methylcytosine (5-mC). This assay detects the methylated fraction of DNA with antibodies followed by quantification through an ELISA-like reaction using the absorbance at 450 nm measured by a micro-plate spectrophotometer. After genomic DNA was prepared by the DNeasy^® blood and tissue kit (Qiagen), 100 ng DNA was bound to each well of the assay plates in biological and technical duplicates (n=4). After the plates were washed, the capture antibody was added; after another wash, the detection antibody and an enhancer solution were added. A color-developing solution was then added, and absorbance was measured at 450 nm. A standard curve was generated for absolute 5-mC quantification.

A methylation-specific PCR analysis of the miR-132 promoter CpG island located on chromosome 17, between base pairs 497 and 545 upstream of the transcriptional start site, was performed as described (29). Genomic DNA was extracted from PDA cells using a QIAamp DNA Mini Kit (Qiagen), and a 100-ng sample of DNA was modified with bisulfite using EZ DNA Methylation™ Kit following the manufacturer’s instructions (Zymo Research, Irvine, CA, USA). PCR was performed with 20 ng bisulfite-modified DNA using the HotStar Polymerase Kit (Qiagen). The PCR primers were as follows: methylated miR-132 forward, 5′-TTTTTTGGGATATTTTTGACGTTAC-3′, methylated miR-132 reverse, 5′-CCGACTAAAAACTCTACTACTCCG-3′, amplifying a 122-bp fragment; unmethylated miR-132 forward, 5′-TTTTTGGGATATTTTTGATGTTATG-3′, unmethylated miR-132 reverse, 5′-CCAACTAAAAACTCTACTACTCCAC-3′, amplifying a 121-bp fragment. PCR settings were 95°C for 5 min, followed by 40 cycles of 94°C for 30 sec, 62°C for 30 sec, and 72°C for 60 sec, with a final extension step at 72°C for 7 min.

A Renilla luciferase reporter construct expressing the wild-type (wt) 3′UTR TGF-β2 was purchased from BioCat. The complete putative 3′UTR binding region for miR-132 was exchanged using QuikChange Site-Directed Mutagenesis Kit to create a mutated (mt) site (Agilent Technologies, Waldbronn, Germany). The following primer sequences were created with the QuikChange Primer Design Program (Agilent Technologies) and ordered from Eurofins GATC Biotech GmbH (Konstanz, Germany): TGF-β-M1-3′UTR forward, 5′-GCC TAA GGA AGC TTC TTG TAA GGT CCA AAA ACT AAA ATC TGA CAT AAT AAA AGA AAA CTT TCA GTC AGA ATA AGT CTG TAA G-3′; TGF-β2-M1-3′UTR reverse, 5′-CTT ACA GAC TTA TTC TGA CTG AAA GTT TCT TTT ATT ATG TCA GAT TTT AGT TTT GGA CCT TAC AAG AAG CTT CCT TAG GC-3′. The sequences of the generated wt and mt miR-132 3′UTR plasmids were confirmed by sequencing (Eurofins GATC Biotech GmbH).

MirVana™ mimics [miR-132-3p and non-coding miRNA (miR-NC)] at 50 nM each were transfected into cells using Lipofectamine 2000 (both from Thermo Fisher Scientific, Inc.) using the reverse transfection method described in the manufacturer’s instructions.

Cells were seeded at a density of 1×10⁴ cells per well into a 96-well plate and co-transfected with 50 nM miR mimics, 25 ng/well Firefly luciferase plasmid and 50 ng/well Renilla luciferase reporter construct expressing TGF-β2 3′UTR (BioCat). The cells were lysed in Passive Lysis Buffer of the Dual-Luciferase^® Assay System (Promega Corporation, Mannheim, Germany) at 48 h post-transfection. Renilla and Firefly luciferase activities were measured using the FLUOstar OPTIMA instrument (BMG Labtech GmbH, Ortenberg, Germany).

Cells were harvested 24 h after transfection and at 12 h after DEX treatment, resuspended in DMEM supplemented with 10% FCS and plated at a density of 5×10⁵ cells/ml in ready-to-use culture-inserts in a µ-Dish (ibdiTreat, item #80241; ibidi GmbH, Martinsried, Germany). At 100% cell confluence, the inserts were removed, leaving a 500-µm gap in each dish; fresh medium was added. After 24 and 48 h, images were taken with a Nikon Eclipse TS 100-F inverted microscope equipped with a camera. Images (5 per treatment condition) were obtained, and analysis was performed with the WimScratch Quantitative Wound-Healing Image Analysis Software version 1 (item #30002; ibidi GmbH).

Western blot analysis was performed as previously described (30). Rabbit monoclonal antibodies against TGF-β2 (#710276; ABfinity™; Thermo Fisher Scientific, Inc., Waltham, MA USA), E-cadherin (#3195-S; Cell Signaling Technology, Inc., Danvers, MA, USA), and vimentin (#EPR3255; Abcam, Cambridge, UK) and a mouse monoclonal antibody against β-actin (#A1978; Sigma-Aldrich; Merck KGaA) were used. All primary antibodies were diluted 1:1,000 and secondary antibodies 1:5,000; the incubations were all performed at room temperature.

AsPC-1, PANC-1 and ASAN-PaCa cells were transfected with 50 nM miR-132 mimics or a negative miR control (NC). At 8 h later, the cells were treated with 1 µM DEX in the presence or absence of miR-132. After 48 h, the cells were resuspended in DMEM supplemented with 10% FCS and plated at a density of 400 (AsPC-1 and PANC-1) or 1,000 cells/well (ASAN-PaCa) in 6-well tissue culture plates. The cultures were maintained under standard culture conditions for 14 days, followed by an evaluation of those fixed and Coomassie-stained colonies consisting of ≥50 cells. The plating efficiency was calculated as a percentage: (number of colonies/number of seeded cells) ×100, as previously described (30).

Immunofluorescence staining was performed on 6-µm-thick frozen or paraffin-embedded tissue sections, as previously described (31). The following antibodies were used: Mouse monoclonal antibodies against TGF-β2 (#ab36495), vimentin (#ab8059) and 5-methylcytosine (#ab10805) (all from Abcam), and rabbit monoclonal antibodies against E-cadherin (#3195s; Cell Signaling Technology, Inc.) and Ki-67 (#ab92742; Abcam). Primary antibodies were diluted 1:50 and secondary antibodies 1:400; the incubations were all performed at room temperature. Signals were detected with a Leica DMRB fluorescence microscope (Leica Microsystems Ltd., Milton Keynes, UK). Images of representative fields were captured using a SPOTTM FLEX 15.2 64-Mp shifting pixel digital color camera and analyzed with SPOT Basic/Advanced 4.6 software (both from Diagnostic Instruments, Sterling Heights, Michigan, USA).

Detection of miR-132 expression in tissue sections was achieved using miRCURY LNA™ microRNA Detection Kit (Exiqon, Vedbaek, Denmark) as described (32). Staining was performed on formalin-fixed, paraffin-embedded patient tissues. Briefly, sections were dewaxed in Roti-Histol (Carl Roth GmbH & Co. KG, Karlsruhe, Germany), rehydrated in 2-propanol, treated with proteinase K (15 µg/ml) and air-dried. Hybridization was performed for 2 h at 58°C using a miR-132-specific, digoxigenin-labeled locked nucleic acid (LNA) detection probe and a scrambled miR as a negative control. After stringent washes, the bound LNA probes were detected with an alkaline phosphatase-coupled digoxigenin antibody (Sigma-Aldrich; Merck KGaA) and NBT/BCIP (Thermo Fisher Scientific, Inc.) as the substrate. The sections were mounted using Roti-Mount FluorCare (Carl Roth GmbH& Co. KG) containing 4′,6-diamidino-2-phenylin-dole (DAPI) as a counterstain.

ASAN-PaCa cells were transfected with 50 nM mirVana™ mimics of hsa-miR-132 or miR-NC using the Lipofectamine 2000^® reagent (both from Thermo Fisher Scientific, Inc.). At 8 h later, the cells were treated with DEX or left untreated. At 48 h later, the cells were transplanted to the chorioallantoic membrane (CAM) of fertilized chicken eggs on day 9 of embryonic development (n=25). On developmental day 14, 50 µl miR-Lipofectamine mixture containing 50 nM hsa-miR-132 mimics or saline as a control were injected into the CAM vessels supplying the tumor xenografts. Fertilized eggs from genetically identical hybrid Lohman Brown (LB) chickens were obtained from a local ecological hatchery (Geflügelzucht Hockenberger, Eppingen, Germany), and tumor xenotransplantation, treatment, and evaluations of tumor take, tumor growth and metastasis were performed as described recently (33).

Quantitative data are presented as the mean ± SD of experiments performed in triplicate. The data were analyzed using Student’s t-test for statistical significance. The data were evaluated by the Mann-Whitney test and by the Kruskal-Wallis test followed by Mann-Whitney and Bonferroni-Holm tests. The in vivo tumor growth data were evaluated by the Kruskal-Wallis test followed by the Mann-Whitney and Bonferroni-Holm tests. Statistical significance was considered to be indicated by values of P<0.05.

To evaluate potential GC-regulated miRNAs, we treated the established PDA cell line AsPC-1 with DEX for 0.5, 24 or 96 h, or left the cells untreated as a control. Total RNA was isolated and examined by Agilent Human miRNA Microarray analysis (Release 19.0). Based on bioinformatics evaluation, 268 miRNAs were significantly (P<0.05) differentially regulated at 96 h, more than at any other time-point. Among these 268 miRNAs, 112 were upregulated and 85 downregulated. The 24 top deregulated miRNAs with the lowest standard deviations are shown as a heatmap in Fig. 1A. We performed Gene Ontology analysis to elucidate which pathways are regulated by the most significantly regulated miRNAs and used keywords ‘EMT’, ‘DEX’, ‘Wnt’ and ‘Smad’ to determine miRNAs associated with DEX-induced EMT. In total, 7 of the 100 deregulated miRNAs were found to share all of these pathways, as shown in a Venn diagram (Fig. 1B; Table I). Finally, TargetScan, miRanda, miRwalk and Ingenuity Pathway Analysis (IPA) tools were applied, resulting in the selection of miR-132-3p with its predicted target gene TGF-β2 as the most relevant miRNA candidate. Via RT-qPCR we confirmed that DEX induced differential regulation of miR-132-3p, the expression of which was upregulated at 0.5 h after DEX treatment but downregulated at 24 h and completely inhibited at 96 h (Fig. 1C).

We hypothesized that promoter hypermethylation may play a role in the observed DEX-induced downregulation of miR-132. Because cytosines in CpG dinucleotides are methylated to form 5-mC, we searched for the presence of CpG islands in the miR-132 promoter using the UCSC genome browser (https://genome.ucsc.edu) and found several upstream of the miR-132 transcription start site (data not shown). To evaluate the ability of DEX to induce DNA methylation, we quantified the amount of total 5-mC using a highly sensitive, colorimetric ELISA-based assay. PDA cells were treated or not with DEX, followed by isolation of genomic DNA after 96 h. After DNA capture and detection with specific antibodies, signal enhancement and color development, absorbance at 450 nm was measured using a microplate spectrophotometer. With the 5-mC level in the controls set to 1, DEX significantly enhanced 5-mC levels to ~2 in AsPC-1 and PANC-1 cells, and to 2.5 in ASAN-PaCa cells (Fig. 2A). The presence of 5-mC was also detected by IHC in the tissues of patients with PDA (Fig. 2B and C; Table II) who had taken GCs prior to surgery via inhalation (n=8) or oral (n=6) methods, and in the tissues of patients with PDA who had not taken GCs (n=18). A greater level of methylation was present in tissues from patients who were treated with GCs relative to in tissues from patients who did not receive GC therapy, with the highest level appearing with oral intake, confirming the in vitro results.

To examine whether DEX methylates the miR-132 promoter, we performed methylation-specific PCR using primers for the CpG-rich portion upstream of the miR-132 promoter located at 497-540 bp. Cells were treated with DEX or left untreated, followed by total DNA isolation 96 h later. The extracted DNA was treated with bisulfite to convert the cytokine residues to uracil, leaving 5-methylcytosine residues unaffected. For the PCR reaction, we used a specific methylation primer pair that selectively amplifies the CpG-rich region containing methylated cytosine residues. In parallel, an unmethylated primer pair was used to amplify this region only if the non-methylated cytokine residues were reverted to uracil residues by bisulfite. The PCR products were separated by agarose gel electrophoresis, and representative results are shown (Fig. 2D). Although untreated AsPC-1 and PANC-1 cells yielded a band when using the unmethylated primer pair, cells treated with DEX yielded a strong band when using the methylated primer pair, which was suggestive of DEX-induced hypermethylation of the miR-132 promoter. These data indicate that methylation of the miR-132 promoter plays a role in decreased miR-132 expression.

To gain insight into the molecular mechanisms by which miR-132 regulates TGF-β2 expression, we examined whether miR-132 directly binds to the TGF-β2 3′UTR region (Fig. 3A), which contains a putative binding site for miR-132, as identified by TargetScan. Based on a wt pLightSwitch™- TGF-β2-3′UTR luciferase vector, we mutated the target site of the TGF-β2-3′UTR region to create the mt pLightSwitch™-TGF-β2-3′UTR luciferase plasmid. AsPC-1, PANC-1 and ASAN-PaCa cells were transfected with the wt and mt pLightSwitch- TGF-β2-3′UTR constructs in the presence of miR-132 mimics. At 48 h after transfection, a luciferase reaction was performed, and luciferase activity was quantified using a luminescence microplate reader. We found that in all cell lines, miR-132 markedly reduced the luciferase reporter activity of the wt- but not the mt-TGF-β2 3′UTR construct (Fig. 3B). Furthermore, miR-132 significantly inhibited expression of the TGF-β2 protein in PANC-1 cells, as shown by western blot analysis (Fig. 3C). To evaluate clinical significance, expression of TGF-β2 was examined in tissue samples from patients with PDA who had received inhaled (n=8) or oral (n=6) GCs prior to surgery, and in tissues from patients with PDA who had not received GCs (n=20). With the use of a semi-quantitative scoring system, we detected higher expression of TGF-β2 in the tissues of patients who had received GCs compared with tissues from those who had not. There was a slight though insignificant increase in expression in tissues from those who had received oral GCs compared with those who had received inhaled GCs (Fig. 3D; Table II). We also assessed the expression of miR-132 in patient tissues by in situ hybridization. Although expression of miR-132 was detectable in both groups (Fig. 3E), we were unable to perform a quantitative analysis. Due to a lack of suitable patient tissue, we could not evaluate miR-132 expression by RT-qPCR. Nevertheless, expression of miR-132 and TGF-β2 in patient tissue underlines the potential clinical significance of our in vitro data.

To examine the role of miR-132 and its target gene TGF-β2 in DEX-induced EMT and CSC signaling, we performed colony-forming assays, which allowed us to evaluate the influence of miR-132 on clonogenicity as a typical stem cell feature. Cells were transfected with miR-132 mimics or non-coding control mimics, and then either treated with DEX 8 h later or left untreated. At 48 h after DEX treatment, the cells were seeded at a clonogenic concentration, followed by evaluation of colony formation 14 days later. Untreated control cells and cells treated with control mimics formed colonies, and DEX enhanced the colony number. However, miR-132 mimics strongly inhibited both basal and DEX-induced colony formation (Fig. 4A). Similar results were obtained in wound-healing assays, wherein miR-132 completely inhibited basal and DEX-induced PANC-1 cell migration into the wounded region (Fig. 4B). Similarly, miR-132 attenuated basal and DEX-enhanced cell viability and cell number, as measured by a luminescence-based viability assay (Fig. 4C) and the quantification of cell numbers with a Coulter counter (Fig. 4D), respectively.

These results were also reflected by the expression patterns of EMT markers, as assessed by RT-qPCR. miR-132 mimics significantly inhibited DEX-induced expression of the mesenchymal marker vimentin and significantly activated basal and DEX-inhibited expression of the epithelial marker E-cadherin (Fig. 5A). By contrast, transfection with antisense mimics targeting miR-132 upregulated E-cadherin and downregulated vimentin expression, regardless of the presence of DEX. These results were confirmed by western blot analysis (Fig. 5B).

For in vivo evaluation, we used the CAM xenotransplantation model. Prior to xenotransplantation, ASAN-PaCa cells were transfected with miR-132 or negative control mimics, followed by the application of DEX 24 h later, which was added to both untreated and miR-132-transfected cells. At 48 h post-transfection, 5×10⁵ cells per egg were transplanted onto the CAM at day 8 of chick development (n=25 eggs per group). On day 14 of embryonic development, 50 nM miR-132 mimics, negative control mimics, DEX in saline or saline alone was injected into the CAM vessels supplying the tumor xenografts. On day 18, the chick embryos were humanely euthanized, followed by resection of the tumor xenografts and determination of their volume by calipers. Large tumor xenografts developed in the groups treated with the negative control mimics or DEX alone, whereas the tumor sizes in the miR-132 groups were significantly smaller (Fig. 6A). Notably, the mean weight of the chick embryos in the DEX-treated group was ~10 g and thus one-third lower than that of the control embryos (15 g) (Fig. 6B). This change in embryo weight suggests that DEX interferes with embryonic development, as already shown in humans (34). It may be assumed that the tumor volumes of the DEX-treated group would have been much higher when relative to a normal body weight. However, in vivo treatment with miR-132 normalized the body weight despite the presence of DEX. To highlight these results, we examined the xenograft sections by IHC. DEX treatment was associated with increased proliferation, as indicated by Ki-67 expression and enhanced TGF-β2 expression (Fig. 6C). Both the basal and DEX-induced expression patterns of Ki-67 and TGF-β2 were strongly suppressed in the presence of miR-132. Similarly, the low expression of E-cadherin was restored in the presence of miR-132, and the high expression of vimentin was inhibited. Nonetheless, the effect of DEX on the basal expression levels of E-cadherin and vimentin was difficult to quantify in vivo.