RNA sequencing data integration reveals an miRNA interactome of osteoarthritis cartilage

Abstract

Objective: To uncover the microRNA (miRNA) interactome of the osteoarthritis (OA) pathophysiological process in the cartilage.

Methods: We performed RNA sequencing in 130 samples (n=35 and n=30 pairs for messenger RNA (mRNA) and miRNA, respectively) on macroscopically preserved and lesioned OA cartilage from the same patient and performed differential expression (DE) analysis of miRNA and mRNAs. To build an OA-specific miRNA interactome, a prioritisation scheme was applied based on inverse Pearson's correlations and inverse DE of miRNAs and mRNAs. Subsequently, these were filtered by those present in predicted (TargetScan/microT-CDS) and/or experimentally validated (miRTarBase/TarBase) public databases. Pathway enrichment analysis was applied to elucidate OA-related pathways likely mediated by miRNA regulatory mechanisms.

Results: We found 142 miRNAs and 2387 mRNAs to be differentially expressed between lesioned and preserved OA articular cartilage. After applying prioritisation towards likely miRNA-mRNA targets, a regulatory network of 62 miRNAs targeting 238 mRNAs was created. Subsequent pathway enrichment analysis of these mRNAs (or genes) elucidated that genes within the 'nervous system development' are likely mediated by miRNA regulatory mechanisms (familywise error=8.4×10^-5). Herein NTF3 encodes neurotrophin-3, which controls survival and differentiation of neurons and which is closely related to the nerve growth factor.

Conclusions: By an integrated approach of miRNA and mRNA sequencing data of OA cartilage, an OA miRNA interactome and related pathways were elucidated. Our functional data demonstrated interacting levels at which miRNA affects expression of genes in the cartilage and exemplified the complexity of functionally validating a network of genes that may be targeted by multiple miRNAs.

Keywords: data integration; mRNA; microRNAs; osteoarthritis; regulatory networks.

Figure 1

Data integration approach. Relative log normalised miRNA and mRNA expression matrices were concatenated. Next, differentially expressed miRNAs and genes were correlated and integrated according to the opposite direction of its FC. Further, miRNA–mRNA interactions from prediction tools and experimentally validated databases were integrated. Finally, target genes that followed these criteria were considered to build an OA miRNA–mRNA network. DE, differential expression; FC, fold change; miRNA, microRNA; mRNA, messenger RNA; OA, osteoarthritis.

Figure 2

Paired differential expression analyses between preserved and lesioned OA cartilage. (A) Volcano plot with the differentially expressed miRNAs. (B) Volcano plot with the differentially expressed genes. Blue circles represent downregulated miRNAs (A) or genes (B); circles are red when they are upregulated. Labelled are the top differentially expressed genes and miRNAs, as well as the known and novel discovered ones. FC, fold change; FDR, false discovery rate; miRNA, microRNA; OA, osteoarthritis.

Figure 3

OA miRNA–mRNA interactome. Network of differentially expressed miRNAs targeting differentially expressed genes between unaffected (preserved) and lesioned OA cartilage. Diamonds are miRNAs and circles genes; edges denote that an miRNA targets the connected gene. The size of the nodes is proportional to the number of edges (interactions). Node colour characterises the strength and the direction of the expression change between unaffected (preserved) and lesioned OA cartilage. Edge thickness corresponds to Pearson’s correlation between the miRNA and gene across all samples. miRNA, microRNA; mRNA, messenger RNA; OA, osteoarthritis.

Figure 4

Functional validation miRNA–mRNA interactions. (A) Expression of AMIGO1, SMAD3, GHR and DCAKD in primary chondrocytes transfected with miR-143–3 p mimic or antagomir compared with control as determined by RT-qPCR. (B) Expression of WNT9A, FZD1 and GDF6 in primary chondrocytes transfected with miR-329–3 p or miR-99a-3p mimic or antagomir compared with control as determined by RT-qPCR. Data shown are the average±SE of the mean for three independent donors (*p<0.05). miRNA, microRNA; mRNA, messenger RNA; RT-qPCR, reverse transcriptase-quantitative PCR.