microRNA network analysis identifies miR-29 cluster as key regulator of LAMA2 in ependymoma

Ependymomas (EPN) are enigmatic tumors which continue to present significant management challenges to clinicians as evidenced by the failure to cure up to 40% of cases [1]. Recent genomic and epigenomic studies have identified alterations in DNA copy number, gene expression [2,3], and methylation [4] and showed that EPN is a heterogeneous disease and consists of distinct molecular subtypes. However, the involvement of the microRNAs (miRNA) and their influence over mRNA translation into proteins, in EPN and their contribution to the complexity of the disease are still poorly understood. To identify miRNA - mRNA regulatory network, we systematically evaluated miRNA - mRNA associations using expression profiles of tumors from 64 EPN patients (mean age of 13.3 years, Additional file 1: Table S1 for more details). For each miRNA-mRNA pair, we measured the association between miRNA and mRNA using a Spearman rank correlation and filtered with sequence-based predicted miRNA-target interactions of miRanda and TargetScan databases Additional File 3: Materials and methods [5,6]. We selected miRNA-mRNA pairs with strong negative correlation (FDR 0.5). We used these thresholds to obtain a high-confidence list of candidate miRNAtarget interactions. The combination of the correlation and target prediction filters yielded 390 miRNA-mRNA pairs, significantly more than was expected by chance (P=5.95× 10 �08 , two-tailed binomial test, x= 390, n= 78,934, p= 3.71 × 10 �03 = 28,309 predicted targets/7,623,897 total pairs). These 390 putative target interactions from EPN consisted of 107 evolutionarily conserved miRNAs and 305 target mRNAs (Additional file 1: Table S2). Remarkably, these miRNAs are significantly enriched for oncomiRs/

Microarray Ependymomas (EPN) are enigmatic tumors which continue to present significant management challenges to clinicians as evidenced by the failure to cure up to 40% of cases [1]. Recent genomic and epigenomic studies have identified alterations in DNA copy number, gene expression [2,3], and methylation [4] and showed that EPN is a heterogeneous disease and consists of distinct molecular subtypes. However, the involvement of the microRNAs (miRNA) and their influence over mRNA translation into proteins, in EPN and their contribution to the complexity of the disease are still poorly understood.
Next, we asked whether the network represented by these putative target interactions captures the salient transcriptomic features of EPN subgroups, particularly from posterior fossa (PF). In line with earlier studies, a consensus clustering analysis of mRNA expression data among 64 EPN cases led to the identification of three transcriptional subtypes: largely supratentorial (ST, 37.5%), PF with spinal (SP) and ST (PF + SP + ST, 29.7%), and largely PF (PF, 32.8%) (Additional file 2: Figure S3). The group largely associated with PF (named as PFA) comprised predominantly of children with mean age of 3.5 years (range from 0.4 -9.2 years), whereas the PF clustered with SP and ST (named as PFB) consisted of adults with a mean age of 24.5 years (range from 8 -45 years) (Additional file 2: Figure S4). The differential expression analysis between the two PF subgroups detected a total of 46 mRNA targets as significant (FDR < 0.05), and 27 of these were also differentially expressed in an independent dataset (P = 6.20 × 10 −05 , Fisher's exact test) of 37 PF EPN from the previously published study (Additional file 1: Table S4). Remarkably, the direction of expression differences between two PF subgroups was same in both cohorts for all of the 27 overlapping differentially expressed genes (Additional file 2: Figure S4). Fifteen overlapping mRNA target genes showed high levels of expression in the PFA subgroup, of which four (ACSL6, ATP1B2, CAMK2B, and S1PR1) were regulated by miR-24, four (ADAMTS9, BLMH, LAMA2, and NUDT11) were regulated by the miR-29 family, three (MMD, PKIA, and STK39) were regulated by miR-27b, and PDGFRA was regulated by miR-34c and miR-218. Interestingly, LAMA2 has previously been identified as a candidate maker gene for PFA subgroup and associated with poor prognosis [3,10,11]. In addition to a strong miR29a:LAMA2 association (r = −0.62; FDR = 1.18 × 10 −05 ), we also observed an inverse correlation between LAMA2 and miR29c (r = −0.49; FDR = 1.61 × 10 −03 ) in EPN, but not in other cancer types ( Figure 2). The two miR-29 family miRNAs are encoded at two different genomic loci, yet they showed anti-correlation with LAMA2 in EPN, suggesting EPNspecific strong co-regulation of the miR-29 loci. Taken together, these results suggest that downregulation of Figure 1 Ependymoma network of miRNA-mRNA target interactions. Two highly connected sub-networks from the inferred network of ependymoma comprising 390 putative target interactions between 107 miRNAs and 305 target mRNAs. Edge width represents strength of Spearman rank correlation for a given miRNA-mRNA pair, red edge colour represents evidence of recurrence in multiple cancer types, and miRNAs from the same family are color coded while a single miRNA is shown in white.
miR-29 family expression is a potent mechanism by which LAMA2 expression is altered in PF ependymoma.
In summary, we identified miR-29a/c as novel regulators of LAMA2 in ependymoma based on miRNA-mRNA covariation and sequence-based target predictions.
The decreased expression of miR-29a/c and elevated LAMA2 expression are therefore defining features of PF ependymoma post-transcriptional regulation, indicating a key mechanism for molecular pathogenesis. Apart from changes in miRNA expression, other mechanisms (genetic or epigenetic) can contribute to LAMA2 expression in PF ependymoma, which now mandates further investigation.