MicroRNA Expression Profiling by Bead Array -2

Materials and Methods

Cell Culture, Interferon Treatment, and RNA Precipitation

Melanoma cells (ME-15) were cultured in RPMI 1640 with L-Glutamine supplemented with non-essential amino acids and sodium pyruvate (1 mM) and hepatoma (HuH7) cells were cultured in DMEM + GlutaMAX. Both media contained 10% FBS. All cell culture reagents were purchased from Invitrogen (GIBCO®). Roferon (Interferon alpha2a, ROCHE) was diluted in fresh medium to a final concentration of 1,000 U/mL and control cultures were grown without cytokine. Cells were cultured at 37°C in a humidified atmosphere containing 5% CO₂. Total RNA preparation was carried out using TRIZOL (Invitrogen) total RNA extraction using 1/2 volume of 1-bromo-3-chloro-propane (molecular biology grade, SIGMA) as chloroform substitute. For efficient recovery of small RNAs, DNA LoBind tubes (Eppendorf) were used and all centrifugation steps were performed at maximum speed and 4°C in an Eppendorf 5417R centrifuge. Total RNA was precipitated with 2 vol of 2-propanol (Fluka) at −20°C for at least 16 h. The RNA pellet was washed with 75% ethanol (Merck), dried, and dissolved in DEPC-treated water (Ambion). The RNA was quantified with Quant-iT™ RiboGreen® RNA Assay (Invitrogen) as suggested by Illumina.

Illumina Bead Array MicroRNA Detection

Starting with 500 ng/sample of total RNA, mature microRNAs were amplified with the Illumina human v1 MicroRNA expression profiling kit containing primers for 743 human microRNAs. The resulting amplicons were hybridized to a 96 sample universal probe capture array and fluorescent signals were detected by confocal laser scanning. All steps were performed according to Illumina’s instructions manual.

Data Processing and Statistical Analysis

The data was processed with Beadstudio software (version 3.1.3, gene expression module 3.3.8) including the calculation of detection p values based on negative control bead signals. Log-transformation, loess normalization (13) and statistical analysis were performed with R (2.8.1) (14) using the package lumi (1.8.3) (15) and software contained therein, in particular limma (2.16.4) (16). Statistical models were chosen as follows: a linear model (limma t statistics) with two separate coefficients for HuH7 and ME-15 cells was used for the selection of differently expressed genes shown in Fig. 1 and Table 1. Statistics represented in the tables were calculated by testing the two indicated conditions as independent factors. In Table 1, p values were adjusted by the false discovery rate method (17). Treatment effects shown in Table 2b and Fig. 3a were modeled with two coefficients (cell line, treatment) for time point 4 h, p values arise from t statistics. Normalized relative fluorescence levels were calculated by 2^mean (of log2 transformed, loess normalized values). Change factors (CHF) were calculated as fold change on the linear scale minus 1 as previously described (2). Raw data, non-normalized, and normalized microRNA expression data have been submitted to the Gene Expression Omnibus with accession number GSE16421.

Fig. 1 Differential microRNA expression in human melanoma (ME-15) and hepatoma (HuH7) cells. microRNA expression levels were compared in two cell lines at two different time points and corrected for the treatment effect. The 50 most significant (p value below 10⁻¹²) microRNA expression values from untreated samples are shown in a heat diagram including hybridization controls as reference for technical variance. White indicates noise levels, yellow indicates the first quartile, orange the median, red the third quartile, and black maximum expression levels. The intensity data, significance values and the IFNα-dependent expression levels are summarized in Table 1.

Table 1 Cell line differences in microRNA expression

,>

364