The use of external miRNA standards of a known copy number revealed that both methods possess a comparable estimation of the respective quantity of the three Arabidopsisspike-ins as the measurements of the spike-ins in non-enriched total RNA were within ± log(10) 0.5 miRNA copies of the expected values for both RT-qPCR technologies (Figure 1). As the Arabidopsis miRNAs were spiked into human total RNA and the serially diluted synthetic miRNAs used to generate the standard curves for interpolating miRNA copy number were in a yeast tRNA background, this would suggest that there was no significant impact of the different background RNA matrices for either assay. It would be interesting to investigate whether our results are consistent with other assays of a comparable design, for example whether Qiagen's miScript PCR System yields similar results to the Exiqon, since they both use a similar RT and qPCR priming strategy, or whether these findings are specific to the two assays we have assessed. This is something we will be addressing in future experiments. An important note is that we performed these experiments by replicating the whole reverse transcriptase and PCR process. The Life Technologies protocol advises a single reverse transcription to be performed, followed by replicate PCRs; however, this does not provide any estimation of the variation associated with the reverse transcription step (43). We therefore elected to replicate the reverse transcription step for both the Life Technologies and Exiqon assays.
Our experiments were designed to address the use of RT-qPCR assays that are appropriate for the analysis of a relatively small number of miRNAs in a given study. However, for studies which aim to measure global miRNA profiles, other technologies such as global RT-qPCR arrays can also be used. Interestingly, a cross-platform analysis has previously been performed between three global miRNA profiling technologies, including the Life Technologies’ Taqman Human MicroRNA Array and Exiqon's miRCURY Ready-to-use PCR, which are based on the same RT-qPCR technologies that we evaluate in this study (44). Despite this, the relative performance characteristics of the Exiqon and Life Technologies’ arrays were significantly different from those identified in our study. For example, the Exiqon array displayed a superior reproducibility and linearity compared with the Life Technologies’ array (44). Although the basic principles of the RT-qPCR assay and array platforms are similar, there are, however, many significant differences that could impact on the miRNA measurement. For example, the Life Technologies’ array uses a pre-amplification (pre-amp) step and involves the megaplexing of >300 miRNA-specific reverse transcription primers in each reaction. Indeed, it is possible that the pre-amp step in particular might cause an increase in the variability of measurement, as has been found with other platforms (45). Comparison of the findings of the Jensen et al. (44) study with ours would suggest that the differences between the experimental design of the array platforms and the RT-qPCR assays that use similar technologies are great enough to yield very different performance characteristics. This indicates that the measurement capabilities of RT-qPCR technologies should not be generalized to their use in global array platforms.
To evaluate the effects of short RNA enrichment on the miRNA measurement, we compared equivalent volumes of total and enriched RNA material as the enrichment protocol was performed by eluting in the same volume that was used as starting material (50 µL). Consequently the same effective volumes of enriched and total RNA samples were added to the RT-qPCR reactions. This method of comparison was chosen in preference to analyzing equal amounts of total and enriched RNA as this would not allow an accurate assessment of the miRNA yield after enrichment since the accuracies of total and short RNA quantification are not comparable. The enrichment reduced the copy number of the miRNAs (endogenous and Arabidopsis miRNA spike-ins) to approximately 25% of that present prior to enrichment on average (P < 0.0001; Figure 2, Supplementary Figure S1, Table 2; detailed model output for all miRNAs is shown in the Supplementary Material). This dramatic reduction in miRNA levels represents a significant finding and highlights that it is not advisable to rely on quantifications of RNA from pre-enriched samples as this would result in a significant underrepresentation of the miRNA amount. Importantly, it was also found that the enrichment procedure did not have an equal effect on all measured miRNAs given that the miRNA yield ranged between 12%–35% for different miRNAs, resulting in changes to relative miRNA levels and potential experimental bias. Therefore, this would suggest that miRNA levels cannot be accurately compared between total and short RNA enriched preparations using the miRVana miRNA isolation kit and this needs to be considered when using different enrichment methods. Indeed, there are several other methods for enriching short RNAs, including column-based technologies similar to the miRVana miRNA isolation kit but also techniques such as short RNA gel purification, which is often performed prior to NGS analysis (46). We focused on the miRVana miRNA isolation kit because it is one of the most commonly used, but it is important for future studies to investigate the potential bias that these other methods may also introduce and the impact of this bias on miRNA quantification. The majority of the human miRNAs we analyzed have recently been reported to be suitable endogenous controls for normalizing miRNA expression (22, 47), highlighting that these findings are relevant to both relative and absolute miRNA quantification strategies as the differential enrichment of housekeeping miRNAs compared with other target miRNAs in a population could have a significant effect on relative miRNA expression after normalization. For example, let-7a shows a 22-fold lower expression relative to miR-26b in the total RNA sample when measured with the Life Technologies assay, however, after short RNA enrichment, let-7a was 29-fold lower in expression than miR-26b (Table 2).