miRNAs, acting as negative regulators of gene expression, regulate various biological processes via inhibiting the expression of their target genes. Increasing evidence shows that some miRNAs play critical roles in tumorigenesis of CRC and have a potential clinical value in diagnosis, treatment and prognosis evaluation for CRC [4–9]. It is of great importance to identify the targets of miRNAs and elucidate their complex regulatory networks.
In this study, a multi-step approach to identify the mRNA targets of dysregulated miRNAs in CRC was adopted. Firstly, in the same cohort of samples containing CRC tissues and their corresponding adjacent tissues, the expression profiles of miRNAs and mRNAs were screened. Secondly, all possible pairs of miRNA-mRNA with an anti-correlationship were identified by Pearson’s correlation analysis, implying that these pairs of miRNA-mRNA may contain both the direct target-relationship and indirect anti-correlationship pairs. Thirdly, the reliability of the identified target-relationship pairs of miRNAs-mRNAs was assessed and confirmed by using both TargetScan and miRanda databases. Finally, the differential expression and their anti-correlationship of 6 miRNAs and 4 mRNAs were validated by qRT-PCR and the regulatory effect of mir-29a on the expression of target of KLF4 mRNA was validated by in-vitro analysis. Through the multi-step approach, 72 miRNA-mRNA pairs containing 22 dysregulated miRNAs and their 58 target mRNAs in CRC were identified, and the Wnt signaling pathway was the only affected pathway based on pathway analysis for the miRNA targets.
Compared with previous studies on the miRNA expression profile of CRC, including miRNA microarray [12–15], miRNA cloning and SAGE , most dysregualted miRNAs detected in this study were consistent with at least one of the previous studies. However, some discrepancy and inconsistency of the miRNA expression profile between studies were found. For instance, the elevated expression of mir-29b in CRC tissues in this study and the recent report  contradicted findings in Cummins’ report  and the elevated expression of mir-424 in CRC tissues in this study was in consistent with the other two studies based on Chinese patients’ samples[11, 40]. However, no other studies reported the altered expression level of mir-424 in CRC. The discrepancy of miRNA expression profile between studies might be attributed to the sample size, genetic background, clinical characteristics or different technique platforms. Strikingly, it was found that mir-7, an up-regulated miRNA with high fold -change in this study (fold-change = 3.60) was not reported as a dysregulated miRNA in CRC in previous studies. In studies with other tumors, an elevated expression level of mir-7 was an indicator of poor disease-free survival in non-small cell lung carcinoma , and correlated with a larger size of tumor and poor tumor grade in breast cancer tissues  while some reports stated that mir-7 inhibits EGFR and the AKT pathway, and the expression level of mir-7 was down-regulated expression in gilobastoma [43, 44]. It seems that the role of mir-7 differs in various types of cancer, suggesting that the expression and function of miRNAs is tissue cell specific. The expression level of mir-7 and its potential targets in CRC are worthwhile for further investigation
In this study, an important note should be added that a very strict criterion was adopted to identify the miRNA targets. Briefly, the target mRNA should be differentially expressed, significantly anti-correlated with its miRNA regulator, and predicted by two databases, TargetScan and miRanda. Based on this strict criterion,our final identified miRNA-mRNA pairs have a high probability of involvement in CRC tumorigenesis. The following analysis will support this high probability. On one hand, comparing the mRNA targets (n = 4338) of the dysregualted miRNAs (n = 32) predicted just by TargetScan, only a small portion of the target genes (244/4338, 5.6 %) were significantly differentially expressed in CRC, and were selected for further analysis, while a great number of mRNAs (4094/4338, 94.4 %) were not differentially expressed in CRC tissue, and were not selected for further analysis. The fact that these 4094 genes were predicted as miRNAs targets but not significantly differentially expressed, suggested that the expression of these 4094 mRNAs might not be regulated by the 32 dysregulated miRNAs, but by other mechanisms such as regulation by other miRNA, methylation, cis-regulatory elements, etc. which may act and counteract the regulation of miRNA. On the other hand, Pearson’s correlation analysis selected 17360 miRNA-mRNA pairs whose expression levels were significantly inversely correlated, however only 72 target relationships of miRNA were finally identified. The existence of the rest of 17288 miRNA-mRNA pairs may be caused by the indirect regulation.
Another interesting finding was that the expression values of mir-29a showed a narrower range of change than those of its predicted target, KLF4. To rationalize this finding, it is hypothesized that the KLF4 mRNA expression was regulated by many other factors including other miRNAs. Besides mir-29a, KLF4 was also identified as a target gene of mir-7, an up-regulated miRNA with high fold-change. The unparalleled change of miRNA and its target mRNA expression levels could be caused by the two miRNA (mir-29a and mir-7) targeting one gene (KLF4). Inspired by this finding, the relationship between fold-change of mRNAs expression and the number of corresponding miRNA regulators were further analyzed. In the identified 72 miRNA-mRNA pairs, three groups can be classified: 1) 2 mRNAs were regulated by 3 miRNAs, 2)10 mRNAs were regulated by 2 miRNAs, and 3) the rest (60 mRNAs) were regulated by 1 miRNA. However, the fold-changes of the gene mRNA expression between these three groups do not reach the statistical significance, indicating the gene expression regulation is very complex and dynamic with impact from multi-factors.
miRNAs regulate diverse biological processes via regulating the expression of target genes. In this study, only the Wnt signaling pathway, which had been proved to be aberrantly activated in CRC [45, 46], was found to be affected through pathway analysis of the miRNA targets. Our study revealed that 4 down-regulated mRNA and an up-regulated miRNA regulated by 5 dysregulated miRNA in CRC respectively were involved in the Wnt signaling pathway, and shed light on the miRNA regulation of Wnt signaling pathway.
Nevertheless, the orchestration of these regulatory pairs of miRNA-mRNA differed in different diseases, presenting the disease-specific expression profiles and composing a complex regulatory network. Although many methods had been used to identify the regulatory pairs of miRNA-mRNA in diseases such as microarray and software-prediction, no method possessed the capacity to determine all miRNA-mRNA regulatory pairs. Moreover, since the anti-correlationship of the miRNAs-mRNA expression levels was used as the screening criteria, only those target genes regulated through mRNA degradation could be identified here, target genes regulated by miRNAs through translation inhibition would be lost.
In summary, these new findings of pairs of miRNA-mRNA provide some hints for the mechanisms of CRC tumorigenesis and were worthwhile for further functional analysis. The more complex regulatory networks call for more research in the future.