Integrative profiling of lactylation reveals prognostic biomarkers and an immunosuppressive niche in acute myeloid leukemia

Background: The overall survival rate of acute myeloid leukemia (AML) remains less than 30%. Metabolic reprogramming of leukemia cells, such as the Warburg effect, enables them to adapt to the microenvironment and thereby develop. Elucidating the landscape of lactate regulation in AML helps clarify the pathogenesis from the perspective of metabolic reprogramming and identify possibilities for optimizing current treatment modalities. Methods: RNA and single-cell sequencing data for AML were obtained from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases. Seurat, limma package algorithm and Weighted gene coexpression network analysis (WGCNA) were conducted to identify candidate lactylation-related genes (LRGs). Enrichment analyses and protein-to-protein interactions were used to clarify the functions. Univariate COX regression and machine learning algorithms (LASSO-logistic, SVM-RFE and Boruta) narrowed the range of LRGs.The DALEX package employed four machine learning models for validation. CIBERSORT analyzed the relationship between immune cell infiltration and key LRGs, while single-gene GSEA was utilized to evaluate the functions of LRGs. We evaluated the associations between hub LRGs and AML using a two-sample Mendelian randomization (MR) analysis. Molecular docking was used to screen for feasible drugs targeting the hub genes. Western blotting was performed to assess pan-lactylation levels in AML cell lines. qRT-PCR and immunohistochemistry were performed to detect GZMB/LSP1 expression in AML patients. Results: Seven hub LRGs were identified in the AML groups: LSP1, MPO, GZMB, SPINK2, HLA-DRB1, HLA-DRA and POU2F2, of which GZMB and LSP1 passed MR test. The seven hub genes were enriched in immune and inflammatory pathways. GLM ultimately emerged as the optimal model validated by GEO datasets. Compared with healthy controls, Kasumi-1 cells exhibited elevated lactylation levels, with exogenous lactate treatment further increasing lactylation levels, whereas sodium oxamate administration had the opposite effect. Exogenous lactate treatment significantly upregulated the mRNA expression of GZMB and LSP1. (-)-Gallocatechin gallate and indomethacin bound well to GZMB, while benzo(a)pyrene and benzo(e)pyrene had good binding potential with LSP1. Conclusions: We established lactylation as a critical regulator of AML, and GZMB and LSP1 were identified as lactylation-related clinical modeling indicators, which provides a foundation for choosing prognostic and therapeutic strategies for AML.