miRNA genes are transcribed by RNA polymerase II (Pol II) [4] in the nucleus. This process yields long primary transcripts of miRNA (pri-miRNAs). Nuclear RNase III Drosha cleaves pri-miRNA into a precursor of miRNA (pre-miRNA) [5]. Pre-miRNAs are then exported from the nucleus into the cytoplasm by exportin-5 (Exp5) [6]. After further processing by cytoplasmic RNase III Dicer [7,8], mature miRNA is integrated into the miRNA-containing RNA-induced silencing complex (miRISC) [9,10]. miRNA transcription and maturation is a precise controlled and collaborated process [11].

Diverse techniques are available for the purpose of miRNA expression profiling, e.g., Northern blotting [12], dot blotting [13], primer extension analysis [14], RT-qPCR [15] and next generation sequencing (NGS) [16]. RT-qPCR, microarray and NGS are the three most state-of-the-art methods applied in miRNA profiling [17]. RT-qPCR offers the most sensitive profiling signal with a reasonable cost. With a considerate design, RT-qPCR can also achieve high throughput miRNA profiling. The recent progress of NGS has led to rapid cost reduction, which lowered the impact of NGS application in miRNA profiling. Nonetheless, microarrays with immobilized DNA probes on solid substrate are still the choice of many laboratories with a moderate budget [18]. Hence, microarrays are widely applied also in miRNA profiling [19]. One of the challenges of miRNA expression profiling is to distinguish between mature miRNAs and their precursors and to detect the miRNA expression signal with high specificity. This problem can be solved with a precise and genuine probe design on microarray platforms.

Much more attention has been dedicated to mature miRNA expression profiling, although pri-miRNA and pre-miRNA are important intermediates during miRNA biogenesis [20]. Pre-miRNAs have also been acknowledged as useful disease biomarkers [21,22]. RT-qPCR has been, so far, the gold standard for pre-miRNA expression evaluation [23]. Recently, a global pre-miRNA landscape has also been successfully provided by applying a deep sequencing technique [24]. Since pre-miRNAs and mature miRNAs can both be detected in cytoplasm, we demonstrated in this paper the possibility of exploring the expression of mature and pre-miRNAs on a single microarray platform, simultaneously.

Pre-miRNAs showed lower expression in these data sets than their matured products. Boxplots of expression values of pre- and mature miRNAs in three human and three mouse studies, respectively, are shown (Figure 1). In all murine studies, median expressions of pre-miRNA are significantly lower than those of mature miRNA, wherein one of the studies exhibits a minor difference (GSE32352). For human samples, the differences are not so pronounced; however, they are also significant for two of the three studies analyzed (GSE42915, GSE33809). At the same time, the upper quartiles of mature miRNAs distribution are located more in a higher expression value region. It is possible that mature miRNA is more enriched by sample preparation. Besides, the high stability observed by mature miRNAs could also contribute to their higher expression signal [32]. Moreover, the rapid processing of the intermediate product pre-miRNA might have an impact on the fewer copy number of pre-miRNA compared to the stable accumulating signal of mature miRNA. The lower expression value of pre-miRNA compared to its corresponding mature miRNA was also reported in another study using the RT-qPCR method [33]. Independent to the variety of individual samples, three data sets that used the same RNA isolation kit (GSE34413, GSE36257 and GSE42915) showed a similar pattern in expression value distribution. This observation indicates that sample preparation, especially RNA isolation, is a crucial factor for successful miRNA profiling.

The general higher expression level of mature miRNA was also reflected in the detection call of present miRNAs. Present calls were made for pre-miRNA and mature miRNAs compared to background signals. Mature miRNAs have almost doubled present calls with respect to pre-miRNAs in all data sets (Figure 2).

Most of the pre-miRNAs and mature miRNAs share the same present/absent calls. That means, present pre-miRNAs have present mature counterparts, and absent pre-miRNAs have also absent mature corresponding miRNAs (Figure 3, red bars). However, there are also present pre-miRNAs, which have absent mature miRNA, and vice versa (Figure 3, blue and black bars). miRNAs that were detected as present in mature form, but as absent in pre-miRNA form, could support the hypothesis that mature miRNAs are more stable than their precursor counterparts. As demonstrated, there was a few number of miRNAs that were present in their premature form, but absent in their mature form. This phenomenon indicates a possible miRNA transcription regulation by degrading mature miRNA [34]. The reduction of specific mature miRNAs compared to their pre-miRNAs was also reported in human colorectal neoplasia by Northern analysis [35].

The expression of mature miRNA and pre-miRNA measured on the same arrays showed a positive interdependency (Figure 4). Notwithstanding different expression levels, a positive correlation did exist between expression of pre- and mature miRNAs on the same array. This is consistent with the results reported previously [36]. Correlation coefficients of pre- and mature miRNA on the same array in one data set were consistent, while the values clearly varied between different data sets. For example, GSE32352 demonstrated a much lower correlation coefficient between pre- and mature miRNA expression. GSE34413, GSE36257 and GSE42915 showed similar correlation coefficients. While in the studies, GSE34413, GSE36257 and GSE42915, the same RNA isolation kit (TRIzol from Invitrogen) was used; in the GSE32352 study, the mirVana RNA isolation kit (Ambion) was used. For this reason, we believe that the observed correlation coefficients were strongly affected by the choice of sample preparation. The influence of the kind of probe labeling can be excluded, because all probes were labeled with the same kit (FlashTag biotin HSR kit from Genisphere).

Correlation coefficients were also calculated for the expression profile of every mature miRNA to its corresponding pre-miRNA across the arrays in the data sets. These coefficients support the evidence of the dependency between pre-miRNA and mature miRNA expression regulation, which differs according to different tissue sources and/or treatments. Correlation coefficients for all miRNAs in each study have been calculated between the pre-miRNA expression pattern and the mature miRNA expression pattern (Figure 5). In all data sets analyzed in this paper, correlation coefficients were widely distributed between −1 and 1. In particular, not only positive coefficients were observed, but also extreme negative ones. That means that upregulation of mature miRNA can be observed in a study even when the corresponding pre-miRNA is downregulated. This result indicates that during maturation of pre-miRNA to mature miRNA, complex factors are involved in regulation, depending on the kind of treatment, tissue-specific development or different biological/pathological contexts.