Profiling microRNA (miRNA) expression is of widespread interest given the critical role of miRNAs in many cellular functions. Profiling can be achieved via hybridization-based (microarrays), sequencing-based, or amplification-based (quantitative reverse transcription-PCR, qPCR) technologies. Among these, microarrays face the significant challenge of accurately distinguishing between mature and immature miRNA forms, and different vendors have developed different methods to meet this challenge. Here we measure differential miRNA expression using the Affymetrix, Agilent, and Illumina microarray platforms, as well as qPCR (Applied Biosystems) and ultra high–throughput sequencing (Illumina). We show that the differential expression measurements are more divergent when the three types of microarrays are compared than when the Agilent microarray, qPCR, and sequencing technology measurements are compared, which exhibit a good overall concordance.

Introduction

It has recently become standard practice to profile the expression levels of microRNAs (miRNAs). Researchers have many different technological options to comprehensively analyze miRNA expression, with each option having advantages and disadvantages. Digital gene expression (DGE) profiling—based on ultra high–throughput DNA sequencing—is increasingly popular since it allows for the discovery of new miRNAs along with quantitative expression analysis. In a recent communication, Linsen et al. showed that DGE profiling is strongly biased toward certain small RNAs, which makes DGE inappropriate for absolute quantification of miRNAs, but not for differential miRNA expression analysis (1). These biases were dependent on library preparation and were also observed in quantitative reverse transcription polymerase chain reaction (qPCR) amplifications. Therefore, when novel miRNA discovery is not a priority, alternative technologies to DGE can still be attractive.

Microarray-based techniques have the advantages of being relatively cost-effective, relatively quick from RNA labeling to data generation, and simple to use. Among the available commercial miRNA microarray platforms, the single-color array format is the most common. Agilent Technologies (Santa Clara, CA, USA) has developed a miRNA profiling platform that provides both sequence and size discrimination for mature miRNAs (2). This system generates results that are highly correlated with qPCR results and, therefore, is an excellent choice for miRNA profiling (3,4). Other major microarray manufacturers have produced single-color miRNA array platforms, including both Illumina and Affymetrix (5). Real-time quantitative PCR is another popular method for differential miRNA profiling. Compared with array platforms, it has superior sensitivity (6), and has recently been parallelized in an array-like format (Applied Biosystems, Foster City, CA, USA) allowing the profiling of 380 miRNAs in a single experiment (7). In this study, we compare miRNA expression in the same brain and heart RNA samples using three different array platforms, qPCR, and DGE.

Materials and methods

RNA samples

Human heart and brain total RNA were from Stratagene (MVP human normal adult tissue total RNA; La Jolla, CA, USA).

Microarray

To assay technical reproducibility, four technical replicates from brain and heart RNA were hybridized on microarrays.

Agilent. One hundred nanograms of total RNA from each sample were labeled and hybridized on human Agilent mi RNA v2 microarrays as described previously (3). Data were extracted and summarized using Agilent Feature Extraction Software. Then they were imported into GeneSpring GX10 software (Agilent Technologies), quantile-normalized and log₂-transformed.

Illumina. Five hundred nanograms of total RNA from each sample were labeled and hybridized on Human v2 MicroRNA Expression BeadChips (Cat. no. MI-102–1024; Illumina, San Diego, CA, USA), according to the manufacturers recommendations (Illumina MicroRNA Expression Profiling Assay Guide). BeadChips were scanned with the Illumina iScan Reader. Data were imported into GenomeStudio (Illumina), quantile-normalized and log₂-transformed in R (www.r-project.org).

Affymetrix. One microgram of total RNA from each sample was labeled with the FlashTag Biotin RNA Labeling Kit for Affymetrix GeneChip miRNA arrays (Genisphere, Hatfield, PA, USA). After labeling, the samples were hybridized on Affymetrix GeneChip miRNA arrays according to the manufacturer's recommendations (Affymetrix, Santa Clara, CA, USA). Hybridization, washing, and scanning of the slides were done according to Affymetrix's recommendations (Fluidics Protocol FS450_0003). Data were extracted from the images, quantile-normalized, summarized (median polish) and log₂-transformed with the miRNA QC tool software from Affymetrix.

Quantitative real-time PCR

Seven hundred nanograms of total RNA were reverse-transcribed with the megaplex RT primers human pool A (Applied Biosystems). This pool contains specific stem-loop primers for 377 human miRNAs, 3 small RNAs, and 1 negative control, and are all based on miRBase v. 10.1. The resulting cDNA was transfered to a TaqMan Human MicroRNA A Array v2.0 (Applied Biosystems) and qPCR was performed on an Applied Biosystems 7900HT Sequence Detection system. Cycling conditions were 50°C for 2 min, 94.5°C for 10 min, and 40 cycles of 97°C for 30 s and 59.7°C for 1 min. Two technical replicates were performed per sample. Quantification cycle (Cq; standard name for Ct or Cp value) values were recorded with SDS version 2.3 software. Cq values ≥36 were considered beyond the limit of detection (a Cq value of 35 represents a single molecule template detection). miRNAs for which both brain duplicate or heart duplicate Cq values were ≥36 were removed. Cq values were imported into qbase^PLUS version 1.3 software (Biogazelle, Ghent, Belgium), which is based on geNorm (8) and qBase (9). U6 and RNU48 were found to be the most stable reference genes and used to normalize the data. Mean RQs (relative quantities) were calculated for each tissue after removing the remaining undetected values.