Single-molecule m6A detection empowered by endogenous labeling unveils complexities across RNA isoforms.

作者：Wen-Bing Guo, Zhijun Ren, Xiang Huang, Jiayin Liu, Jingwen Shao, Xiaojun Ma, Chuanchuan Wei, Yixian Cun, Jia-Liang He, Jie Zhang, Ze-Hong Wu, Yang Guo, Zijun Zhang, Zheng-Ming Feng, Jiang-Bo He, Jin-Kai Wang · 发表于：Molecules and Cells · 年份：2025 · DOI：10.1016/j.molcel.2025.01.014 · 被引用次数：24 · 研究领域：Medicine

The landscape of N6-methyadenosine (m6A) on different RNA isoforms is still incompletely understood. Here, in HEK293T cells, we endogenously label the methylated m6A sites on single Oxford Nanopore Technology (ONT) direct RNA sequencing (DRS) reads by APOBEC1-YTH-induced C-to-U mutations 10-100 nt away, obtaining 1,020,237 5-mer single-read m6A signals. We then trained m6Aiso, a deep residual neural network model that accurately identifies and quantifies m6A at single-read resolution. Analyzing m6Aiso-determined m6A on single reads and isoforms uncovers distance-dependent linkages of m6A sites along single molecules. It also uncovers specific methylation of identical m6A sites on intron-retained isoforms, partly due to their differential distances to exon junctions and isoform-specific binding of TARBP2. Moreover, we find that transcription factor SMAD3 promotes m6A deposition on its transcribed RNA isoforms during epithelial-mesenchymal transition, resulting in isoform-specific regulation of m6A on isoforms with alternative promoters. Our study underscores the effectiveness of m6Aiso in elucidating the intricate dynamics and complexities of m6A across RNA isoforms.