Evaluation of extraction kits and RT-qPCR systems adapted to high-throughput platform for circulating miRNAs

Abstract

MicroRNAs regulate gene expression at the post-transcriptional level. Differential expression of miRNAs can potentially be used as biomarkers for early diagnosis and prediction for outcomes. Failure in validation of miRNA profiles is often caused by variations in experimental parameters. In this study, the performance of five extraction kits and three RT-qPCR systems were evaluated using BioMark high-throughput platform and the effects of different experimental parameters on circulating miRNA levels were determined. Differences in the performance of extraction kits as well as varying accuracy, sensitivity and reproducibility in qPCR systems were observed. Normalisation of RT-qPCR data to spike-in controls can reduce extraction bias. However, the extent of correlation for different qPCR systems varies in different assays. At different time points, there was no significant fold change in eight of the plasma miRNAs that we evaluated. Higher level of miRNAs was detected in plasma as compared to serum of the same cohort. In summary, we demonstrated that high-throughput RT-qPCR with pre-amplification step had increased sensitivity and can be achieved with accuracy and high reproducibility through stringent experimental controls. The information provided here is useful for planning biomarker validation studies involving circulating miRNAs.

Figure 1. Overview of the experimental design for data depicted in Fig 2 to Fig 7.

Plasma samples from eight to ten healthy individuals were aliquoted into five portions for isolation using five different extraction kits. The expression levels of 14 endogeneous miRNAs and two spike-in controls in these samples were evaluated using 3 RT-qPCR systems which were adapted to the high-throughput platform. Q, miRNeasy Serum/Plasma kit; E, miRCURY™ RNA Isolation Kit - Biofluids; A, mirVana™ PARIS™ Kit; MN, NucleoSpin® miRNA Plasma; NB, Plasma/Serum Circulating RNA Purification Kit.

Figure 2. miRNA recovery of different extraction kits were evaluated by the C_q values for spike-in controls (cel-miR-39 and cel-miR-54) assayed in the (a) TaqMan and (b) miScript RT-qPCR system.

Box plots with whiskers show 5–95 percentile of the C_q values for spike-in controls and the corresponding CVs are shown. The extent of extraction bias across samples was reflected in the range and CV of C_q values for spike-in controls. Higher median C_q compared to other extraction kits indicate poorer performance in miRNA recovery. Q, miRNeasy Serum/Plasma kit; E, miRCURY™ RNA Isolation Kit - Biofluids; A, mirVana™ PARIS™ Kit; MN, NucleoSpin® miRNA Plasma; NB, Plasma/Serum Circulating RNA Purification Kit.

Figure 3. The effects of extraction kits on RT-qPCR results were evaluated using the TaqMan system.

Pairwise comparison of extraction kits is shown in scatter plots. Each data point represents the average C_q value obtained from duplicate or triplicate wells of qPCR. Our results indicated that the correlation of data from same set of samples extracted using different kits can be improved by normalisation to spike-in controls.

Figure 4

Examples of assays which showed linear amplification (miR-16 and miR-141) or non-linear amplification (miR-135a*) in serial dilution samples which were preamplified in the (a) TaqMan and (b) miScript systems. Significant correlation between the preamplified and non-preamplified data points was seen in (c) TaqMan (ICC = 0.888, p < 0.001) and (d) miScript (ICC = 0.862, p < 0.001).

Figure 5. Comparison of TaqMan and miScript RT-qPCR systems adapted to the high-throughput platform by performance parameters which include reproducibility, accuracy and sensitivity.

Transformation of data to z-score was carried out to facilitate direct comparison. A1, accuracy when measuring fold change of abundant miRNAs; A2, accuracy when measuring fold change of less abundant miRNAs, R1, reproducibility of RT replicates; R2, reproducibility of RT replicates after normalisation; R3, reproducibility of qPCR replicates; S1, sensitivity to detect the presence of plasma miRNAs and S2, sensitivity to detect the presence of miRNAs in titration points.

Figure 6. Evaluation of consistency between technical replicatesin RT-qPCR.

Consistencies among (a) RT replicates in TaqMan system, (b) RT replicates in miScript system, (c) qPCR replicates in TaqMan system and (d) qPCR replicates in miScript system were evaluated by measuring the levels of 11–16 miRNAs in a same set of RNA samples (n = 5–14). Each data point in scatter plot represents the average C_q value obtained from the duplicate or triplicate wells of qPCR. Overall, significant correlations (ICC = 0.712–0.996, p < 0.001) between replicates were achieved when data were normalised to spike-in controls.

Figure 7. Comparison of different RT-qPCR systems adapted to the BioMark high-throughput platform.

(a) Average C_q values for endogeneous plasma miRNAs were depicted in box plot with whiskers showing the 5–95 percentile. Irrespective of extraction kits being used, within the same set of samples, more undetermined C_q values were obtained in the miScript system. (b) Correlation of qPCR results from TaqMan and miScript systems may vary among different miRNA assays. Each data point represents the normalised C_q value transformed into z-score. Q, miRNeasy Serum/Plasma kit; E, miRCURY™ RNA Isolation Kit - Biofluids; A, mirVana™ PARIS™ Kit; MN, NucleoSpin® miRNA Plasma; NB, Plasma/Serum Circulating RNA Purification Kit.

Figure 8. Intra-individual variations of eight circulating miRNAs were measured in nine individuals.

(a) Significant correlation (ICC = 0.962, p < 0.001) and (b) no significant fold change (paired T test, p > 0.05) of circulating miRNAs was observed in plasma samples collected at different time points. In paired plasma and serum samples collected at the same time point, (c) significant correlation (ICC = 0.948, p < 0.001) and (d) significant fold change (paired T test, p < 0.05) of circulating miRNAs were observed.