Deciphering lactate/lactylation networks in AML: integrated scRNA-seq and transcriptomics reveal functions and prognostic model

Acute myeloid leukemia (AML) exhibits pronounced heterogeneity, necessitating deep molecular characterization for precision therapy. Lactate metabolism and histone lactylation, influencing tumor biology via epigenetic regulation and immune microenvironment remodeling, represent an emerging focus. This study combines single-cell RNA sequencing (scRNA-seq) and bulk RNA sequencing (bulk RNA-seq) data to investigate the prognostic value of lactate/lactylation-associated genes (LL-genes, defined as genes involved in lactate metabolism and histone lactylation regulation) in AML. Specifically, Seurat was utilized for scRNA-seq clustering with cell annotation/validation via the TISCH2 database. Gene Set Variation Analysis (GSVA) assessed lactate/lactylation pathway activity. In bulk RNA-seq, ConsensusClusterPlus enabled molecular subtyping, while ten machine learning algorithms constructed a prognostic model. scRNA-seq revealed specific LL-gene overexpression in malignant progenitors, concomitant with elevated lactate metabolism-lactylation activity (LML-CAS; Lactate Metabolism-Lactylation Modification Combined Activity Score), enhanced metabolic-inflammatory synergy, and immunosuppression (increased Tregs/M2 macrophages). Molecular subtyping identified two clusters (A/B) exhibiting divergent survival outcomes (Cluster A: poorer prognosis). An optimized 7-gene prognostic model demonstrated high accuracy, predicting reduced chemotherapy response among high-risk patients. Transcriptomic profiling indicated lactylation-associated immunosuppression (e.g., downregulated CXCL9/10-CXCR3 axis, enrichment of T cell exhaustion markers) and heightened in silico-predicted sensitivity to BCL-2/FGFR inhibitors (ABT-737/AZD4547) in high-risk patients. qRT-PCR confirmed RNA-level dysregulation of key LL-genes (IFI16, THOC2, HIST1H2BD, ARPP19), aligning with bioinformatic predictions. Western blot analysis further validated aberrant protein expression of IFI16 and THOC2 in AML specimens, reinforcing their dysregulation. Collectively, integrated analyses uncovered lactate/lactylation-associated heterogeneity in AML. Our machine learning-based prognostic model predicts survival, therapeutic response, and drug sensitivity, suggesting a potential strategy for precision therapeutics in AML.