Although a considerable proportion of acute myeloid leukemia (AML) patients achieve remission through chemotherapy, relapse remains a recurring and significant event leading to treatment failure. This study aims to investigate the immune landscape in AML and its potential implications for prognosis and chemo-/immune-therapy.
Integrated analyses based on multiple sequencing datasets of AML were performed. Various algorithms estimated immune infiltration in AML samples. A subgroup prediction model was developed, and comprehensive bioinformatics and machine learning algorithms were applied to compare immune-based subgroups in relation to clinical features, mutational landscapes, immune characterizations, drug sensitivities, and cellular hierarchies at the single-cell level.
Two immune-based AML subgroups, G1 and G2, were identified. G1 demonstrated higher immune infiltration, a more monocytic phenotype, increased proportions of monocytes/macrophages, and higher FLT3, DNMT3A, and NPM1 mutation frequencies. It was associated with a poorer prognosis, lower proportions of various immune cell types and a lower T cell infiltration score (TIS). AML T-cell-based immunotherapy target antigens, including CLEC12A, Folate receptor β, IL1RAP and TIM3, showed higher expression levels in G1, while CD117, CD244, CD96, WT and TERT exhibited higher expression levels in G2. G1 samples demonstrated higher sensitivity to elesclomol and panobinostat but increased resistance to venetoclax compared to G2 samples. Moreover, we observed a positive correlation between sample immune infiltration and sample resistance to elesclomol and panobinostat, whereas a negative correlation was found with venetoclax resistance.
Our study enriches the current AML risk stratification and provides guidance for precision medicine in AML.