An m⁵C-associated NSUN5-YBX1-CENPF programme marks melanoma cell states and immune features

RNA 5-methylcytosine (m⁵C)-associated programmes may help characterise melanoma cell states and their immune context. We investigated an NSUN5-YBX1-CENPF (NYC) programme using patient-matched sequencing, experimental perturbation and multiscale transcriptomic analyses.

We analysed m⁵C-MeRIP-seq and matched Input RNA data from three melanoma-adjacent-normal patient pairs, using a paired edgeR interaction model for regional IP enrichment. Tissue and cell experiments included NSUN5 knockdown, CENPF-targeted MeRIP-qPCR and western blotting. Clinical analyses considered proliferation and sampling site. Trajectory and regulatory inference were restricted to malignant cells from GSE215120, with patient-level summaries. Treatment-associated changes were examined in 43 paired patients from GSE91061. Computational perturbation and transcriptomic signature reversal prioritised candidate regulators and compounds.

Two CENPF regions showed increased relative IP enrichment after Input adjustment (4.97- and 5.85-fold; BH q = 0.0457 and 0.0074). NSUN5 knockdown reduced CENPF m⁵C enrichment and protein abundance. NYC was associated with proliferation; its survival association did not persist after proliferation adjustment. Within malignant cells, NYC-High retained an association with an immune-resistance expression programme after G2M/E2F and depth adjustment (nine patients; BH q = 0.0156). Regulatory analyses prioritised MLX, whose simulated knockout and overexpression produced opposing reference-flow alignments (BH q = 0.0134 and 0.0201). MLX regulon activity was associated with glycolysis and oxidative phosphorylation after adjustment. In GSE91061, responders showed a greater treatment-associated NYC decrease than non-responders (mean change difference -0.209; permutation P = 0.0138). Drug screening generated candidates for experimental testing.

NYC provides an m⁵C-associated framework connecting patient-derived molecular observations to malignant-cell states and immune features. YBX1 remains a candidate reader, and regulatory, treatment-associated and drug-prioritisation findings do not establish a complete causal pathway or a clinically validated predictive biomarker.