深度学习模型与组学筛选揭示血管紧张素原为肝癌中 5-甲基胞嘧啶（m(5)C）调控的肿瘤微环境通讯介质

The m 5 C methylomes in wildtype and m 5 C-catalyzing enzyme NSUN2 -perturbed liver cancer cells were profiled via MeRIP-seq. GAT-MeRIP, a graph attention neural network-based algorithm, was developed to identify functional m 5 C-modified target genes from MeRIP-seq data. TME-related functional m 5 C targets were screened through cell-cell communication analysis of single-cell transcriptomic data. In vitro functional validation of the key target gene was performed via a combination of cell co-culture, qRT-PCR, MeRIP-qPCR, ELISA, and flow cytometry assays. Additionally, public liver cancer cohort data were used for clinical correlation and prognostic analysis.

Angiotensinogen (AGT) was identified as a key m 5 C-regulated secretory factor contributing to tumor-microenvironment communication in liver cancer. NSUN2 knock-down increased AGT's expression and enhanced cytotoxicity of co-cultured NK cells, which can be canceled by AGT-neutralizing antibody. Exogenous AGT treatment significantly enhanced NK cell cytotoxicity by upregulating IFN- , TNF- , and perforin, as well as the proportion of CD107a NK cells. Liver cancer patients with low NSUN2 and high AGT exhibited significantly improved overall survival rates and higher immune infiltration.

This study unveils novel regulatory function of m 5 C-modified AGT in modulating the liver TME that could be helpful for improving liver cancer prognosis and immunotherapy.