Integrated analyses of zebrafish miRNA and mRNA expression profiles identify miR-29b and miR-223 as potential regulators of optic nerve regeneration

Abstract

Background: Unlike mammals, zebrafish have the ability to regenerate damaged parts of their central nervous system (CNS) and regain functionality of the affected area. A better understanding of the molecular mechanisms involved in zebrafish regeneration may therefore provide insight into how CNS repair might be induced in mammals. Although many studies have described differences in gene expression in zebrafish during CNS regeneration, the regulatory mechanisms underpinning the differential expression of these genes have not been examined.

Results: We used microarrays to analyse and integrate the mRNA and microRNA (miRNA) expression profiles of zebrafish retina after optic nerve crush to identify potential regulatory mechanisms that underpin central nerve regeneration. Bioinformatic analysis identified 3 miRNAs and 657 mRNAs that were differentially expressed after injury. We then combined inverse correlations between our miRNA expression and mRNA expression, and integrated these findings with target predictions from TargetScan Fish to identify putative miRNA-gene target pairs. We focused on two over-expressed miRNAs (miR-29b and miR-223), and functionally validated seven of their predicted gene targets using RT-qPCR and luciferase assays to confirm miRNA-mRNA binding. Gene ontology analysis placed the miRNA-regulated genes (eva1a, layna, nefmb, ina, si:ch211-51a6.2, smoc1, sb:cb252) in key biological processes that included cell survival/apoptosis, ECM-cytoskeleton signaling, and heparan sulfate proteoglycan binding,

Conclusion: Our results suggest a key role for miR-29b and miR-223 in zebrafish regeneration. The identification of miRNA regulation in a zebrafish injury model provides a framework for future studies in which to investigate not only the cellular processes required for CNS regeneration, but also how these mechanisms might be regulated to promote successful repair and return of function in the injured mammalian brain.

Fig. 1

Schematic of experimental outline. Schematic flow diagram of procedures used to identify inversely correlated putative target genes of the differentially expressed miRNAs

Fig. 2

Gene ontology of over-expressed genes. GO analysis of enriched (a) cellular components and (b) biological processes of the over-expressed gene set. Each part represents -log₂ of the p-value associated with the GO category from WebGestalt. WebGestalt results were filtered through GOTrim and presented only GO terms with ≥5 genes. Molecular function not shown as no terms passed this criteria.

Fig. 3

Gene ontology of under-expressed genes. GO analysis of enriched (a) cellular component, (b) biological processes, and (c) molecular functions of the under-expressed gene set. Each part represents -log₂ of the p-value associated with the GO category from WebGestalt. WebGestalt results were filtered through GOTrim and presented only GO terms with ≥5 genes.

Fig. 4

Enriched IPA canonical pathways of genes that were differentially expressed after injury. (a) Over-expressed and (b) under-expressed genes are shown

Fig. 5

Gene ontology of miR-29b predicted targets from Targetscan Fish. Representation of GO terms associated with (a) cellular component, (b) biological process, and (c) molecular function. Predicted gene targets contained at least one miRNA binding site with a context score of ≤ −0.30. Each part represents -log2 of the p-value of biological process and cellular component from the set of significant biological processes and cellular components. The p-values were retrieved from gene ontology analysis in WebGestalt. A list of genes in each GO category is in Additional file 6: Table S3

Fig. 6

Gene ontology of miR-223 predicted targets from Targetscan Fish. Representation of GO terms associated with (a) biological process, and (b) molecular function. Predicted gene targets contained at least one miRNA binding site with a context score of ≤ −0.30. Each part represents -log2 of the p-value of biological process and cellular component from the set of significant biological processes and cellular components. The p-values were retrieved from gene ontology analysis in WebGestalt. A list of genes in each GO category is in Additional file 7: Table S4

Fig. 7

Validation of miR-29b and miR-223 putative gene targets from integration analysis. Microarray and corresponding RT-qPCR expression of genes predicted to be targeted by miR-29b (a) or miR-223 (b). The latter is presented as expression fold change (2^-ΔΔCt) relative to PPIA (mean ± SD; n = 4 groups of pooled retinal RNA containing 4 animals)

Fig. 8

Predicted miRNA binding sites within 3’UTR of predicted target genes. Sequence of miR-29b (a) and miR-223 (b) binding sites within 3’UTR of predicted target genes (nt, nucleotide position in 3’ UTR). Seed region is bolded. Mutations predicted to disrupt miRNA-mRNA binding were made in the seed region, mutated nucleotides underlined

Fig. 9

Validation of miRNA binding to 3’ UTR of putative target genes. Luciferase reporter assay of HEK293 cells cotransfected with pmirGLO plasmid containing the WT or MT 3’UTR miRNA seed sequence from each gene, and either miR-29b, miR-223 or miR-NC (scrambled control). Samples were analysed 48 h after transfection and data normalised to the pmirGLO only transfection. Columns represent the luciferase activity of either WT or MT constructs with miR-29b (a) or miR-223 (b), relative to transfection with the same construct and miR-NC. Data represents the mean ± SEM, n = 3 independent experiments containing 4 replicates each. Student’s t test comparing WT or MT construct with miRNA to miR-NC indicated as *p < 0.05, **p < 0.01, ***p < 0.001. Student’s t test comparing WT with miRNA to corresponding MT construct indicated as #p < 0.05, ## p < 0.01, ###p < 0.001