Multi-omics characterization of RNA modification enzymes identifies NAT10 as a functionally validated prognostic...

RNA modification enzymes (RMEs) are key post-transcriptional regulators that impact RNA stability, translation, and splicing. Dysregulation of RMEs is closely associated with tumor initiation and progression. However, their global regulatory patterns and clinical relevance across cancer types remain incompletely characterized.

We conducted an integrative multi-omics analysis of RME expression, copy number variation (CNV), and clinical outcomes across multiple cancers. Machine learning algorithms were employed to identify tumor-discriminating RME signatures. Single-cell RNA sequencing (scRNA-seq) characterized tumor microenvironmental heterogeneity. A LASSO-derived prognostic model was established and validated in independent cohorts. Drug sensitivity prediction and supportive functional assays (EdU assays, qRT-PCR, immunohistochemistry) were performed for representative RMEs.

RMEs were broadly upregulated across cancers and showed strong associations with CNV gains. Machine learning identified 12 RMEs that reliably discriminated tumor from normal tissues. Single-cell transcriptomic analysis showed that 10 of the 12 selected RMEs (DKC1, METTL1, NAT10, TRMT1, RPUSD1, PUS1, WDR4, TRMU, ADAT2, GTPBP3) exhibited higher expression in tumor-infiltrating cells compared with adjacent normal tissues. T-cell subpopulations displayed marked heterogeneity, with ADAT2 preferentially enriched in regulatory T cells. CellChat analysis revealed T cell subsets as key mediators of intercellular communication via multiple immune-related pathways. A 6-gene prognostic model exhibited independent prognostic power and was integrated into a well-calibrated nomogram. Drug-response prediction revealed that high-risk patients exhibited enhanced sensitivity to microtubule-targeting agents and kinase inhibitors, whereas low-risk patients showed preferential response to epigenetic modulators. Importantly, supportive functional assays showed that NAT10 knockdown, validated by qRT-PCR, was associated with reduced proliferative activity in HCC cells as evidenced by EdU assays, and IHC validation further corroborated its overexpression in clinical tumor specimens compared to adjacent normal tissues.

This study delineates a CNV-associated landscape of RME dysregulation across cancers and establishes a 12-RME diagnostic signature and a 6-gene prognostic model with robust predictive performance. Single-cell analyses reveal tumor- and cell-type-specific expression patterns of RMEs, while supportive functional data suggest a potential biological relevance of NAT10 in HCC. Collectively, these findings provide an association-based framework for understanding the potential roles of RNA modification programs in cancer progression and clinical stratification.