Genome-wide analysis of enhancer RNA in gene regulation across 12 mouse tissues

Abstract

Enhancers play a crucial role in gene regulation but the participation of enhancer transcripts (i.e. enhancer RNA, eRNAs) in regulatory systems remains unclear. We provide a computational analysis on eRNAs using genome-wide data across 12 mouse tissues. The expression of genes targeted by transcribing enhancer is positively correlated with eRNA expression and significantly higher than expression of genes targeted by non-transcribing enhancers. This result implies eRNA transcription indicates a state of enhancer that further increases gene expression. This state of enhancer is tissue-specific, as the same enhancer differentially transcribes eRNAs across tissues. Therefore, the presence of eRNAs describes a tissue-specific state of enhancer that is generally associated with higher expressed target genes, surmising as to whether eRNAs have gene activation potential. We further found a large number of eRNAs contain regions in which sequences and secondary structures are similar to microRNAs. Interestingly, an increasing number of recent studies hypothesize that microRNAs may switch from their general repressive role to an activating role when targeting promoter sequences. Collectively, our results provide speculation that eRNAs may be associated with the selective activation of enhancer target genes.

Figure 1. Proportions of number of En_eRNA and En_no-eRNA.

Red denotes the percentage of enhancers transcribing eRNAs. Blue denotes those not transcribing eRNAs. The black line represents the 50% cutoff. The number of enhancers with and without eRNAs is listed to the left of the bar graph. The number of eRNAs is listed to the right of the bar graph.

Figure 2. Correlation of expression level of eRNAs with the enhancer’s target gene or flanking regions.

Each bar represents the Pearson correlation between the expression level of eRNA and target gene (red), eRNA and 0–1 k flanking region (dark green), eRNA and 1–2 k flanking region (green), or eRNA and 2–3 k flanking region (light green). The expression level is calculated by BPKM. Significance is denoted by ^*p-value < 0.001; ^***p-value < 10⁻⁵.

Figure 3. Expression level of target genes of En_eRNA and of En_no-eRNA.

Each box represents the distribution of the BPKM of the target genes of En_eRNA and of En_no-eRNA. Significance is determined by a one-sided Wilcoxon-rank sum test. ^*p-value < 0.001; ^***p-value < 10⁻⁵.

Figure 4. Expression level of the same target genes of the same positioned enhancers across tissues.

Z-scores of the BPKM were calculated for each tissue. Each dot represents the average expressions (z-scores) of the same gene targeted by the same positioned En_eRNA or En_no-eRNA (see Material and Methods). The red line is least rectangles regression. The grey line is a reference diagonal line with a slope of one. For presentation, the axes are limited to three times the standard deviation of the average z-scores.

Figure 5. eRNA matches to miRNAs and other RNAs.

Dots in each plot represent the significant test statistics of eRNAs matching to miRNAs (denoted by red circles) or other RNAs (denoted by blue triangles). Significances were estimated using a two-proportion t-test, which tests the eRNA matches to miRNA versus other RNAs against random intergenic control data. The trend lines are least rectangles regression for miRNA data (red) and other RNA data (blue). For presentation, all axes are limited to 0.35 and represent the ratio of hits to a specific RNA family to the total number of significant hits from Rfam.