Cell culture and treatment with cytokines
Mouse glucagonoma cell line αTC1-6 was obtained from the American Type Culture Collection (ATCC); it was cultured in complete Dulbecco’s modified Eagle’s medium (DMEM, Sigma-Aldrich®, Saint Louis, MO, USA), as described . Mouse insulinoma cell line βTC1 also was from ATCC; cells were grown in DMEM with 25 mM glucose (Sigma-Aldrich®), supplemented with 2 mM L-Glutamine, 15% heat inactivated (HI) horse serum, 2.5% HI fetal bovine serum (FBS), 1% penicillin/streptomycin, in 95% humidified air-5% CO2 at 37° C. Cells were passaged once a week after trypsinization and replaced with new medium twice weekly. Treatment with cytokines (recombinant murine IL-1β, specific activity 5 × 108 U/mg, Preprotech, London, UK, UE; recombinant murine IFN-γ, specific activity 1 × 107 U/mg, Preprotech; recombinant murine TNF-α, specific activity 1 × 107 U/mg, Preprotech) was as described . αTC1-6 cells (passages 20–40) were seeded the day before treatment in 60 mm dishes at a density of 4.5 × 105 cells.
Evaluation of apoptosis and necrosis
Percentage of apoptotic or necrotic cells was assessed through flow cytometry. Analysis was performed with a Beckman Coulter Epics XL-MCL flow cytometer (Beckman Coulter©, Inc., Hialeah, FL, USA). Cells were collected, washed with phosphate-buffered saline (PBS), and stained with Annexin V-FITC/propidium iodide (PI) (Sigma-Aldrich®) in Annexin-V binding buffer, as specified by the manufacturer. EXPO32 ADC Analysis™ software package (Beckman Coulter©, Inc.) was used for data analysis.
RNA extraction and HT quantitative RT-PCR
Total RNA was extracted with Trizol (Lifetechnologies™, Foster-City, CA, USA), according to manufacturer’s instructions. RNA quantification was performed with a Qubit® Fluorometer (Lifetechnologies™). RNA for HT miRNA expression profiling was reverse transcribed into cDNA of 519 mouse-specific and 68 rat-specific miRNAs through Megaplex™ RT Rodent Primer Pool sets, and preamplified through Megaplex™ PreAmp Rodent Primer Pool sets (Lifetechnologies™). Resulting cDNAs were loaded onto TaqMan Low Density miRNA Arrays (TLDA) cards, according to manufacturer’s instructions. TLDA cards were run on ABI 7900HT Real Time PCR system. RNA for Single TaqMan miRNA Assays was reverse transcribed into miRNA-specific cDNA through TaqMan® MicroRNA Reverse Transcription Kit (Lifetechnologies™) and amplified using TaqMan® Universal Master Mix (Lifetechnologies™), according to manufacturer’s instruction. RNA for analysis of miRNA targets was reverse transcribed into cDNA through High Capacity RNA-to-cDNA Kit (Lifetechnologies™) and amplified through Fast SYBR® Green Master Mix (Lifetechnologies™), according to manufacturer’s instruction. Primer sequences are available upon request.
Criteria for selecting downregulated or overexpressed miRNAs
Data quality and quantification were computed using Real-Time Statminer® software (http://www.integromics.com) (Integromics, Granada, Spain). Multiple reference genes were used to normalize data: Genorm (integrated into Real-Time Statminer®)  and DataAssist™ (Lifetechnologies™) softwares allowed to choose the best ones. Limma test (see below: Statistical analysis) was carried out by Real-Time Statminer® to assess statistically significant DE genes. DE miRNAs were ranked based on their p-values and adjusted p-values (Benjamini-Hochberg correction with False Discovery Rate, FDR, set at 5%). Relative quantities (RQ) of miRNAs between αTC1-6 and βTC1, at steady state as between treated αTC1-6 cells and matched untreated controls, were calculated according to 2-ΔΔCt method . Values are reported as average fold changes of three independent biological replicates; RQ values < 1 were converted to negative fold changes by the formula: -1/RQ. The data files for each array are publicly available at the Gene Expression Omnibus (GEO) database repository (http://www.ncbi.nlm.nih.gov/geo/) (GSE42970).
Transient transfection of αTC1-6 cells with mimics of miR-296-3p and miR-298-5p
For transfection, αTC1-6 cells were plated into 24-well plates at a density of 4×104 cells per well to obtain RNA, and into 100 mm dishes at a density of 1.15×106 cells to obtain proteins. Transfections were performed using siPORT™ NeoFX™ (Lifetechnologies™) with 30 nM mimics of miR-296-3p/miR-298-5p/scrambled sequence (Pre-miR™ miRNA Precursor Molecules—Negative Control #1, Lifetechnologies™). For each experiment, efficiency of transfection was measured through real-time PCR.
In silico identification of miRNA targets
Validated targets of DE miRNAs were retrieved from the literature and miRTarbase (release 2.5) . Predictions were performed through starBase (release 2.1) (http://starbase.sysu.edu.cn/). Among validated and predicted targets, only genes expressed in pancreatic cells (data from Beta Cell Gene Atlas, found at http://t1dbase.org/page/AtlasHome) and known to be functionally involved in cell survival or death were chosen for real-time PCR and western blot assays. Data on genomic position of genes encoding miRNAs and their assignment to specific families were from MiRBase (http://www.mirbase.org/).
Protein lysates and their quantification were obtained as previously described . 50 μg of total protein extract were loaded into 10% SDS polyacrylamide gel (Hoefer miniVE, GE Healthcare©, Amersham Place, Buckinghamshire, UK) and blotted to nitrocellulose membranes by iBlot Dry Blotting System (Lifetechnologies™). Membranes were probed with polyclonal antibodies to IGF1Rβ (Santa Cruz Biotechnology®, Inc., Dallas, TX, USA), p-IRS-1 (Santa Cruz Biotechnology®, Inc.), total-IRS-1 (EMD Millipore Corporation©, Billerica, MA, USA), p-p44/p42 MAPK (Cell Signaling Technology®, Inc., Danvers, MA, USA), total p44/p42 MAPK (Cell Signaling Technology®, Inc.), TNFα (Cell Signaling Technology®, Inc.), using β-actin (Sigma-Aldrich®) as loading control. Proteins were detected by using ECL™ Western Blotting Detection Reagents (GE Healthcare©). Densitometric analyses were performed by ImageJ software (http://rsbweb.nih.gov/ij/index.html)
Prediction of transcription factors regulating expression of miRNAs 296-3p and 298-5p
MatInspector from Genomatix (http://www.genomatix.de/) was used to identify Transcription Factors Binding Sites (TFBS) and their corresponding Transcription Factors (TFs) . By using Multi Experiment Matrix (MEM) (http://biit.cs.ut.ee/mem) (selected collection: Affymetrix GeneChip Mouse Genome 430 2.0 [Mouse_430_2] platform, used to analyze 1546 datasets), TFs prediction was interpolated with data on statistically significant expression correlation among TFs, which regulate DE miRNAs and their mRNA targets. Settings used in MEM are described in Additional file 14.
Identification of CpG islands upstream the genes encoding Nespas, miR 296-3p and miR 298-5p
CpG islands upstream genes encoding miR-296, miR-298, Nespas were identified through UCSC Genome Browser (http://genome.ucsc.edu/).
Biological networks comprising miRNAs, their predicted upstream regulators (TFs), their validated targets and first neighbours interactants, were generated by retrieving interactome data through MiMI Cytoscape plugin  and visualized by Cytoscape v. 2.8.1. Biological processes and pathways involving network nodes were analyzed through the tool DAVID (http://david.abcc.ncifcrf.gov/) and BiNGO Cytoscape plugin 
P-values were calculated by applying different methods: Limma test , associated with Benjamini-Hochberg correction for multiple comparison, was applied to identify DE miRNA genes between test and control samples in HT miRNA transcriptome analyses; Student’s t-test, associated with Bonferroni correction method, was used to statistically analyze data from single TaqMan assays and to assess significantly different apoptotic levels between αTC1-6 treated with cytokines and their matched untreated controls; Tukey HSD post-hoc one-way ANOVA test was used to evaluate significant differences in apoptosis among different transfection experimental conditions. For analysis of correlation between the expression of miR-296-3p and miR-298-5p in αTC1-6, Pearson correlation coefficient was calculated. Finally, χ
2-square test was used to establish if miR-296-3p and miR-298-5p have more common targets than expected by chance; Fisher’s exact test was applied to evaluate the enrichment in specific gene ontologies. All statistical tests and correction methods, used to calculate p-values, are described throughout the text and figure legends.