The precision and accuracy of miRNA RT-qPCR quantification was evaluated by investigating variability of miRNA copy number measurements in human brain total RNA, which was spiked with the three Arabidopsis miRNAs across a range of physiologically relevant levels and subsequently short RNA enriched in three independent replicate experiments using the miRVana miRNA isolation kit (Table 2). The enrichment of short RNA (<200 nucleotides) molecules from total RNA preparations is commonly performed prior to miRNA expression analysis. However, to our knowledge the effects of this procedure on miRNA quantification by RT-qPCR have not been extensively investigated. As described in detail in the materials and methods section, the miRVana miRNA isolation kit can be used to isolate total RNA in which the short RNAs are retained and optionally enriched or, alternatively, it can be performed on total RNA material that has already been isolated, possibly by a different method. The latter procedure, which we perform in this study, involves the separation of short RNAs from larger (>200 nucleotides) RNA species in the sample. The short RNA molecules isolated as a result of this process should have a higher concentration relative to the remaining RNA present in the sample compared with the relative levels of short RNAs to total RNA in the non-enriched starting material. However, this process may also result in the loss of short RNAs compared to the amount present in the starting material, which is something we investigate in this study. RT-qPCR measurements were performed with the Life Technologies and Exiqon assays on the non-enriched total RNA and short RNA enriched material. Serial dilutions of synthetic miRNA molecules were run on each plate to estimate miRNA copy number values. Measurement accuracy was determined by comparing expected miRNA copy number to that measured for the Arabidopsis spike-ins (Figure 1).
Comparison of the two RT-qPCR assays revealed that the variability of measurement was higher for Exiqon compared with Life Technologies (Table 2), which was supported by statistical analysis (detailed in Supplementary Material). As these measurements were performed on RNA material for which the spiking of the Arabidopsis miRNAs and subsequent short RNA enrichments were replicated, the standard deviation values presented in Table 2 demonstrate the variability associated with the preparation of the human brain RNA material in addition to that associated with the RT-qPCR assays. However, given that the same template material was used for both RT-qPCR assays, the differences in measurement precision between them can be due only to the variability inherent in the performance of the assays themselves. In order to eliminate any additional operational variability, all experiments in this study were performed by a single analyst. One possible explanation for the difference in measurement variability between the two RT-qPCR technologies could be differences in the assay methodologies. The Exiqon assay requires the cDNA reaction product to be diluted 80-fold. Since the reverse transcription reaction volume is 20 µL, this does not allow the diluting water (1580 µL or 790 µL for 10 µL half volume reactions) to be added directly to the reverse transcription reaction because this would exceed the volume of most standard PCR reaction tubes. Therefore, the cDNA needs to be transferred to a second, larger vessel, thus introducing an additional pipetting step and increasing the likelihood of variability between repeats. By contrast, the Life Technologies assay protocol does not require any dilution of the cDNA (or a small amount of diluent can be added directly to the reverse transcription reaction tube if desired) and thus there is less manipulation. miRNA copy number estimates varied considerably between the two RT-qPCR assays, but there was no general trend. For example Let-7a showed a mean 1.73-fold lower measured copy number in the total RNA sample with the Exiqon compared to the Life Technologies assay (Table 2). In the majority of cases these differences were statistically significant (detailed model output shown in Supplementary Material). This would therefore suggest that the technical error (or measurement uncertainty) associated with RT-qPCR technology could have a significant impact on miRNA quantification. This therefore needs to be considered, particularly when comparing experiments where absolute quantification, such as in our study where miRNA copy number values have been obtained by interpolation from standard curves of serially diluted synthetic miRNAs, has been performed using different approaches, e.g., meta-analyses. In addition to the effects on absolute miRNA quantification, our findings also have implications for relative miRNA quantification strategies in which the relative expression of multiple miRNAs has been determined in comparison to each other or the same miRNAs in different experimental conditions, due to the significant uncertainty associated with the measurement of both experimental and ‘housekeeping’ miRNAs. Indeed, previous studies have reported significant differences in relative miRNA expression of as low as 1.5-fold (41, 42), but our findings suggest that such small differences in expression should be treated with caution.