Alternative mRNA fates identified in microRNA-associated transcriptome analysis

Abstract

Background: MicroRNA (miRNA) are small non-coding RNA molecules which function as nucleic acid-based specificity factors in the universal RNA binding complex known as the RNA induced silencing complex (RISC). In the canonical gene-silencing pathway, these activated RISC particles are associated with RNA decay and gene suppression, however, there is evidence to suggest that in some circumstances they may also stabilise their target RNA and even enhance translation. To further explore the role of miRNA in this context, we performed a genome-wide expression analysis to investigate the molecular consequences of bidirectional modulation of the disease-associated miRNAs miR-181b and miR-107 in multiple human cell lines.

Results: This data was subjected to pathways analysis and correlated against miRNA targets predicted through seed region homology. This revealed a large number of both conserved and non-conserved miRNA target genes, a selection of which were functionally validated through reporter gene assays. Contrary to expectation we also identified a significant proportion of predicted target genes with both conserved and non-conserved recognition elements that were positively correlated with the modulated miRNA. Finally, a large proportion of miR-181b associated genes devoid of the corresponding miRNA recognition element, were enriched with binding motifs for the E2F1 transcription factor, which is encoded by a miR-181b target gene.

Conclusions: These findings suggest that miRNA regulate target genes directly through interactions with both conserved and non-conserved target recognition elements, and can lead to both a decrease and increase in transcript abundance. They also multiply their influence through interaction with transcription factor genes exemplified by the observed miR-181b/E2F1 relationship.

Figure 1

KEGG pathways analysis of predicted miR-181b target genes. KEGG pathways analysis of predicted target genes for miR-181b revealed ten significantly enriched pathways: TGF-beta signalling (p=0.0023); prostate cancer (p=0.0037); neurodegenerative diseases (p=0.0061); melanogenesis (p=0.0062); long-term potentiation (p=0.0070); T-cell receptor signalling (p=0.0087); axon guidance (p=0.0106); MAPK signalling (p=0.0217); dorso-ventral axis formation (p=0.0236); and circadian rhythms (p=0.0410). MAPK: mitogen-activated protein kinase. Predicted target genes for miR-181b were generated from the miRGen database and submitted for pathways analysis to DAVID.

Figure 2

Biological processes affected by miR-181b over-expression in cell culture via miR-181b transfection. Panel A demonstrates the experimental design for the identification of genes subject to PTGS by increased miRNA concentrations. Canonical miRNA function results in a subsequent decrease in mRNA expression levels detected by whole-genome expression analysis using microarrays. These differentially expressed genes are subsequently utilised for DAVID pathways analysis and correlated against predicted miRNA targets. Panel B shows the increase in miR-181b expression levels in comparison to controls for HEK-293, HeLa and SH-SY5Y cell types. Panel C shows a clustered-by-gene heat map from whole genome expression microarray data from each cell model, with n=2 per condition. Panel D shows the significantly enriched KEGG pathways for each cell type in response to increased intracellular miR-181b levels. RI: receptor interaction; ECM: extracellular matrix; MAPK: mitogen-activated protein kinase.

Figure 3

Biological processes affected by inhibition of endogenous miR-181b in cell culture in response to anti-miR-181b transfection. Panel A illustrates the experimental design for the identification of genes subject to de-repression of PTGS by decreased endogenous miRNA concentrations. Genes elevated in response to a fall in miRNA were utilised for pathways analysis and correlated against predicted miRNA targets. Panel B shows the decrease in miR-181b expression levels in comparison to controls for HEK-293, HeLa and SH-SY5Y cell types. Panel C shows a clustered-by-gene heat map from whole genome expression microarray data from each cell model, with n=2 per condition. Panel D shows the significantly enriched KEGG pathways for each cell type in response to decreased intracellular miR-181b levels.

Figure 4

Analyses of bidirectionally modulated genes in multiple cell types. The intersection of bidirectionally-modulated genes identifies genes modulated by both increased miR-181b expression (miR treatment) and miR-181b inhibition (anti-miR-181b treatment) in each cell type. Genes modulated by either miR-181b over-expression or inhibition were considered for the union of modulated genes across multiple cell types. The subsequent KEGG pathways analyses on these genes of interest revealed significantly enriched pathways, as evident in the bottom half of this figure.

Figure 5

The performance of conserved and non-conserved target predictions across multiple biological datasets. Panel A illustrates the accuracy with which modulated genes were correctly predicted as either targets or non-targets by Targetscan. Error bars were calculated using the mean, N, and standard deviation across HEK-293, HeLa, and SH-SY5Y datasets. Panel B illustrates the false discovery rates associated with Targetscan’s prediction of genes modulated subsequent to altered miRNA expression. A false negative indicates a gene differentially expressed with miRNA modulation, but not a predicted miR-181b target; and a false positive indicates a predicted miR-181b target that is not differentially expressed with miRNA modulation. Error bars were calculated using the mean, N, and standard deviation across HEK-293, HeLa, and SH-SY5Y datasets. Panel C shows the conservation status of predicted target genes modulated in response to altered miR-181b expression. The values in this figure represent the average values across both miR-181b over-expression and inhibition in HEK-293, HeLa, and SH-SY5Y cell types. PCT: Probability of conserved targeting; the lower the probabilistic value, the poorer the conservation of the predicted binding site across multiple species.

Figure 6

Reporter gene analysis of miRNA recognition elements (MRE). Putative MREs from genes modulated by miRNA expression were cloned into the 3^′-UTR of the firefly luciferase gene in pMIR-REPORT. Responsiveness of the firefly luciferase reporter gene to increased miR-181b expression (miR-181b transfection) was analysed with respect to a pRL-TK renilla luciferase control. Data was normalised against mutant miR-181b miRNA control transfection. This data was obtained from n=4 experiments, each performed in triplicate, and analysed using a one-tailed T-test.

Figure 7

miRNA-mediated regulation of E2F1 3^′-UTR reporter gene expression. The sensitivity of the E2F1 3^′-UTR to intracellular 181b, miR-107, and miR-20a levels was determined by luciferase reporter gene expression in the presence of either synthetic miRNA or corresponding anti-miR inhibitor. In each case the response was normalised against the respective miRNA and anti-miR control oligos. This data was obtained from n=4 experiments, each performed in triplicate.

Figure 8

Comparison of canonical (left) and non-canonical (right) miRNA-mRNA relationship. Panel A, scheme. Panel B contains charts of accuracy and false discovery rates associated with Targetscan’s prediction of observed changes in mRNA expression. Panel C, pie charts illustrating the distribution of miR-181b and E2F1 target genes predicted using different algorithms and parameters in multiple cell types. Signal-to-noise ratio is shown to increase for both canonical and non-canonical function as stringency increases from genes modulated by either miR-181b over-expression or inhibition across at least two cell types; to genes modulated by either miR-181b over-expression or inhibition across all three cell types; to genes modulated by both miR-181b over-expression and inhibition across at least two cell types. Panel D contains charts of enriched KEGG pathways from genes modulated by either miR-181b over-expression or inhibition across at least two cell types. RI: receptor interaction. MAPK: mitogen-activated protein kinase.

Figure 9

Comparison of canonical (left) and non-canonical (right) miR-107 function. Figure legend as per Figure 8 except in respect to miR-107.