Conserved expression patterns predict microRNA targets

Abstract

microRNAs (miRNAs) are major regulators of gene expression and thereby modulate many biological processes. Computational methods have been instrumental in understanding how miRNAs bind to mRNAs to induce their repression but have proven inaccurate. Here we describe a novel method that combines expression data from human and mouse to discover conserved patterns of expression between orthologous miRNAs and mRNA genes. This method allowed us to predict thousands of putative miRNA targets. Using the luciferase reporter assay, we confirmed 4 out of 6 of our predictions. In addition, this method predicted many miRNAs that act as expression enhancers. We show that many miRNA enhancer effects are mediated through the repression of negative transcriptional regulators and that this effect could be as common as the widely reported repression activity of miRNAs. Our findings suggest that the indirect enhancement of gene expression by miRNAs could be an important component of miRNA regulation that has been widely neglected to date.

Figure 1. miRNA/mRNA pairs with strong conserved negative correlation coefficients overlap with miRBase and TargetScan predictions.

The expression levels of 809,640 miRNA/mRNA pairs were compared across 35 human tissues and a correlation coefficient (r) for each pair was calculated. Pairs with a negative r (inverse correlation) were binned into 5 groups. The first bin contained pairs for which r was between 0 and −0.1 (poor negative correlation) and the last contained pairs for which r was below −0.4 (strong negative correlation). Because our analysis identified only one CNC pair with an r below −0.5, no bin was created for this category. For each bin, the relative overlap between these pairs and putative miRNA/target-mRNA pairs predicted by (A) TargetScan and (B) miRBase was calculated. The degree of overlap between pairs in each bin and pairs from the two databases are represented by vertical bars. The number of pairs in each bin and their percentage of the total number of pairs are shown above these bars. The threshold for significant overlaps (p-value<0.05) is represented by a horizontal dotted line. For example, pairs with negative correlation coefficients between −0.3 and −0.4 show a significant overlap with predictions from TargetScan but not with predictions from miRBase. A similar analysis was conducted in (C) and (D) except that conserved correlation scores between human and mouse were calculated. Conserved negative correlation scores provided a higher overlap with existing predictions from miRBase and TargetScan.

Figure 2. Experimental validation of predicted miRNA targets using a luciferase reporter assay.

Histograms showing the luciferase activity of reporter plasmids containing endogenous 3′UTR sequences of indicated genes (white) as a percentage of the activity of the corresponding mutated microRNA seed region (black). The endogenous sequence comprised ∼500 bp of the 3′UTRs inserted in the 3′UTR of the renilla luciferase gene. The mutant sequence was identical to the WT except that it had the predicted miRNA ‘seed’ binding region deleted or mutated. Each of the plasmids was co-transfected with the relevant pri-miRNA. A) Of the 5 putative targets discovered by considering high scoring CNC pairs tested, 3 showed significant repression. B) The putative target discovered by considering intermediate gene regulation (see text) showed significant repression. Asterisks indicate p-values<0.05, using the non-parametric Mann-Whitney U test.

Figure 3. Positive correlations between miRNA and mRNA can be explained by intermediates.

Solid lines represent published regulatory relationships, whereas dotted lines represent CPC pairs (+) and CNC pairs (−). A. hsa-miR-32 and hsa-miR-92 (Figure 2B) repress RFX1 via a 3′UTR sequence. RFX1 represses PCNA . This results in a positive correlation in expression between hsa-miR-32, hsa-miR-92 and PCNA. B. Systematic discovery of indirect regulation where a putative “intermediate” gene (“gene ?”) explains the positive correlation between the miR-X and gene-Y. For each pair, we search for a direct target of miR-X, which is also an inhibitor of gene-Y. This was achieved by identifying genes that were negatively correlated with both miR-X and gene-Y.