Abstract Acute myeloid leukemia (AML) is characterized by profound immune dysregulation, yet the mechanisms underlying impaired cytotoxicity remain unclear. By analyzing samples from 20 AML patients and 20 healthy donors, we integrated scRNA-seq (GSE223844), bulk transcriptomes (GSE37642, GSE71014), and TCGA-AML. Single-cell profiling, BayesPrism deconvolution, WGCNA, and multiple machine-learning algorithms were used to define immune alterations and construct a prognostic model, with peripheral blood validation through RT-qPCR and multicolor flow cytometry. We identified 11 immune and progenitor populations showing AML-specific shifts, including depletion of T/NK cells and expansion of stem-like compartments. Regulatory network and ligand–receptor analyses revealed broad immune suppression and disrupted cellular communication. Deconvolution showed extensive transcriptional remodeling of CD8+ T cells, and WGCNA identified a CD8+ T-cell gene module. Nine candidate genes were incorporated into 18 machine-learning models, with ridge regression generating a stable nine-gene prognostic signature. Enrichment analyses indicated activation of TGF-β, TNF receptor, and TCR pathways in high-risk patients. LILRB1 emerged as a central immunosuppressive hub in CD8+ T and NK cells, and experimental validation confirmed its elevated expression alongside impaired cytotoxicity in AML-derived lymphocytes. Overall, LILRB1 serves as a key immune checkpoint driving cytotoxic dysfunction, marking exhausted CD8+ T cells and CD16+ NK cells. The nine-gene signature links CD8+ T-cell impairment to poor prognosis, while NK-cell involvement positions LILRB1 as a promising therapeutic target for restoring anti-leukemic immunity.
Miao et al. (Fri,) studied this question.