Identification of m5C RNA modification-related gene signature for predicting prognosis and immune microenvironment-related characteristics of heart failure.

BACKGROUND: Methylation of RNA is involved in many pathophysiological processes. The roles of N6-methyladenosine (m6A) and N7-methylguanosine (m7G) in heart failure (HF) have been established. However, the impact of 5-methylcytosine (m5C) on HF and its relationship with the immune microenvironment (IME) remains elusive. METHODS: GSE141910 (200 HF, 166 NFDs) was used as training cohort. Focusing on 9 identified m5C differently expressed genes (DEGs), random forests (RF), LASSO logistic regression, and SVM-RFE were employed to identify hub genes. ROC curves were plotted to confirm the predictive value in diagnostic model. ScRNA-seq revealed cell-type-specific m5C regulator expression patterns and HF IME. Hub genes were validated using HF rat models after myocardial infarction (MI) through quantitative reverse-transcription PCR (qRT-PCR) and western blot (WB). Consensus clustering algorithms identified two m5C-related HF subtypes. Single-sample gene-set enrichment analysis (ssGSEA) and CIBERSORT deconvolution algorithm analyzed the IME in HF. Finally, we employed WGCNA and PPI network to find m5C associated key genes and their clinical significance in HF subgroups. RESULTS: In HF samples, four m5C regulators (NSUN6, DNMT3A, DNMT3B and ALYREF) were greatly upregulated, while five (NOP2, NSUN3, NSUN7, DNMT1 and TRDMT1) were downregulated compared to NFDs in the training set. ALYREF positively correlated with activated NK cells and monocytes, whereas TRDMT1 and NSUN3 showed inverse correlations with these cells. Four hub genes were identified by machine-learning algorithms and all verified by validation model. Single-cell RNA-seq dataset GSE183852 examined the levels of 13 m5C regulators across 11 different cell types in HF. In vivo experiments including qRT-PCR and WB finally identified NSUN6 as the most remarkable regulator. The diagnostic model demonstrated excellent performance in distinguishing between HF and NFDs (AUC 0.869, 95%CI 0.832-0.906). The two m5C subtypes exhibited distinct modification patterns, immune cell infiltration, immune checkpoints, and HLA gene expression. Additionally, 138 differentially expressed genes were uncovered based on m5C subtypes, and GSEA revealed associations with key pathophysiological mechanisms of HF. By using WGCNA and PPI network, three m5C associated key genes (RPS21, RPL36 and RPS19) were identified significantly influencing cardiac function in clinical practice. CONCLUSION: HF diagnostic model is developed based on 4 robust m5C RNA modification biomarkers (DNMT3B, NOP2, NSUN6 and DNMT1). Two distinct m5C RNA modification patterns in HF are identified, illustrating different IME characteristics. Our findings underline the significance of m5C regulators in HF, offering new perspectives on HF mechanisms and potential diagnostic and therapeutic strategies.