Comparison of microarray and quantitative real-time PCR methods for measuring MicroRNA levels in MSC cultures

Abstract

The capacity for self-renewal and the multilineage potential of mesenchymal stromal cells (MSC) offer a therapeutic promise for regenerative medicine. MicroRNAs (miRNAs) are small noncoding RNAs that play a key regulatory role during differentiation both at the level of posttranslational modulation and epigenetic control. Studies on MSCs have just begun to identify miRNA profiles in MSC and differentiated MSC. While several methods are available for miRNA exploration, microarrays and quantitative real-time PCR (qPCR) are the most common. Since there are several microarray and qPCR platforms available for miRNA detection, it is valuable to explore how these methods compare. We used the NCode Multi-Species miRNA microarray (Invitrogen) and the TaqMan Human microRNA array (Applied Biosystems) to compare microRNA expression in undifferentiated MSCs and MSCs differentiated into early osteoblasts. We show that while there is a somewhat low correlation between these two methods, there is a subset of miRNA measurements that did correlate.

Figure 1. Correlation of NCode™ miRNA microarray and TaqMan^® miRNA array measurements for the 70 miRNAs detected by both assays is dependent on expression level

Out of 454 miRNA probes present on the NCode arrays, 93 miRNA measurements showed a positive mean intensity above background. Results are plotted as the log₂ fold change between MSC and nascent osteoblasts. Out of these 93 miRNAs, 70 were present in both the NCode and TaqMan arrays. (A) Comparison between the two miRNA platforms using all 70 probes indicated poor correlation between the two methods (Spearman’s rho=0.295). (B) However, using only the qPCR results filtered for a higher level of expression (a C_t or cycle threshold value of less than 30) produced increased correlation (Spearman’s rho=0.528). (C) In contrast, when qPCR results were filtered for low levels of expression (C_t value of 30 or more cycles), there was less of an association between the two methods (Spearman’s rho=0.068). Results indicate that fold-change measurements are somewhat similar between the two methods at higher levels of expression but that the methods disagree at lower expression levels. This is likely due to the greater sensitivity of qPCR.