Dynamic landscape and regulation of RNA editing in mammals

To construct a mammalian reference atlas of A-to-I editing, we first compiled a comprehensive list of editing sites in human and mouse (Supplementary Note 1) and then examined the RNA editome across tissues using 8,551 RNA-sequencing (RNA-seq) samples derived from 552 donors in the GTEx project (Supplementary Information 1). Notably, the editing profiles across different tissues were highly correlated (Fig. 1a) and the overall editing activities were also generally similar, except for skeletal muscle, in which editing was significantly lower than in other tissues (P < 2.2 × 10⁻¹⁶, Wilcoxon rank sum test; Fig. 1b). Nevertheless, principal component analysis (PCA) showed that the brain regions could still be resolved from non-brain tissues (Extended Data Fig. 1a). Within the brain, the cerebellum was clearly segregated from other brain parts (Extended Data Fig. 1b), possibly owing to higher expression of ADAR2 (also known as ADARB1) (Extended Data Fig. 1c). When we examined non-repetitive sites in coding regions only, the editing levels became more distinct among the various tissues (Fig. 1a). The different brain regions clustered together, as did heart and skeletal muscle. Unexpectedly, the artery was the most highly edited tissue type (Fig. 1c). The importance of RNA editing in vascular disease was demonstrated in a recent study¹⁸. We further validated the results obtained from the GTEx data by applying a targeted sequencing approach (microfluidics-based multiplex PCR and deep sequencing; mmPCR–seq)¹⁹ (Supplementary Note 2) to examine 1

-Abstract Truncated-