Microarray-Based miRNA Profiling
Microarrays are an established high-throughput method that can facilitate the assessment of multiple known miRNAs in a cost-effective and timely manner. However, due to the short length of miRNAs, optimal probe design for miRNA microarrays is an important consideration, and critical modifications have to be applied to ensure specificity and accuracy of the signal. The capture probes, synthetic oligonucleotides or cDNA fragments that display high specificity and affinity for individual transcripts, have to be adapted to facilitate binding to miRNAs. Due to the short length of miRNAs, probes’ melting temperatures (Tm) may vary between 45 and 74°C. At a medium hybridization temperature, capture probes with lower Tm values will yield lower signals, while capture probes with higher Tm values will display impaired nucleotide discrimination and lower specificity (Wang et al. 2007). Thus, a single hybridization temperature is suboptimal for most miRNA targets unless the probe length is adjusted accordingly, so that high specificity of detection is obtained for closely related mature miRNAs. The enzymatic labelling had little bias as it includes attachment of a single fluorophore-labelled nucleotide to the 3' end of each miRNA with high yield and minimal sample manipulation. Hybridization to the microarray is carried out under conditions that result in near-equilibrium binding and high hybridization yields for most miRNAs (Wang et al. 2007). This platform was shown to be accurate, although some variability can be observed at low concentrations (Ach et al. 2008; Sah et al. 2010). No fractionation or amplification is required, and 100 ng of total RNA input has been successfully used to assess the miRNA profile in plasma samples (Wang et al. 2009).
The inclusion of locked nucleic acids (LNA) in the capture probes can further increase the sensitivity and specificity of detection. LNAs are nucleotide analogues that are constrained in the ideal conformation for Watson-Crick binding and enable more rapid and stable pairing with the complementary nucleotide. This approach resulted in a significantly higher accuracy. By modifying the LNA content and length of probes, their Tm can be adjusted to facilitate stringent hybridization conditions while maintaining equal affinity. Hence, miRNA profiling is possible with as little as 30 ng of total RNA (Castoldi et al. 2008, 2006). LNA miRNA arrays have been applied to screen plasma samples from hypertensive patients (Li et al. 2011).
A totally different concept was applied in bead-based chips, where a single miRNA-specific oligo (MSO) was used to assess each miRNA on the panel. RNA samples were polyadenylated, reverse transcribed to cDNA; MSOs were hybridised to the sample, and a solid phase primer extension step was included to increase the sensitivity. Following amplification of the extended products and fluorescence labelling, these unique MSO sequences were utilised to identify the specific miRNA content in the sample (Chen et al. 2008). The address sequence from each MSO was used to hybridise specific miRNA products to specific locations on the BeadArray substrate for readout. The BeadArray Reader measured the signal intensity at each address location corresponding to the quantity of the respective miRNA in the original samples. A total of 100-200 ng of total RNA from each sample was sufficient for this quantification. This method improved accuracy and can be modified to include additional miRNA capture beads to the mixture, allowing detection of newly discovered miRNAs (Jay et al. 2007; Lu et al. 2005). However, it required enrichment of small RNAs by fractionation, a step that may introduce bias (Liu et al. 2008). It may also be prone to false-positive results as indicated by a study that utilised this platform for screening differentially expressed miRNAs in a cohort of heart failure patients (Tijsen et al. 2010). In the following extensive validation, the authors reported several discrepancies between the levels of miRNA expression as measured by microarrays and qPCR.
High sensitivity in the screening has also been obtained using a combination of conventional hybridization and an elongation step. The isolated RNA was labelled using the ‘klenow’ fragment of DNA polymerase I that is added in the channels of a biochip for specific elongation and labelling of hybridised miRNAs (Vorwerk et al. 2008). This eliminates the need for a labelling step prior to the hybridization and reduces the RNA requirements to as little as 20 ng of total RNA. This method was successfully applied for screening of circulating miRNAs in patients with coronary artery disease (Fichtlscherer et al. 2010).