Viral mimicry in AML is linked to immune activation but not BCL2 inhibitor sensitivity Free

Some cancers trigger a response that resembles a viral infection, called viral mimicry, which can stimulate the immune system. This process has been linked to better responses to certain drugs in solid tumors, but its role in Acute Myeloid Leukemia (AML) is not well understood. In this study, we explore whether viral mimicry in AML is associated with immune activation and its response to BCL2 inhibition

RNA-sequencing data from 707 AML samples from the BeatAML2 dataset was analyzed. Each patient was assigned a viral mimicry score based on the expression of interferon-stimulated genes. Immune infiltration was estimated using established gene signatures for T cells, monocytes, dendritic cells, and other immune populations. The expression of immune checkpoint genes such as PD-1, PD-L1, CTLA4, and TIM-3 were also measured. Drug sensitivity was assessed using ex-vivo area under the curve (AUC) values for azacytidine, venetoclax, and ABT-737. Spearman correlation (ρ) and FDR-adjusted p-values were used for statistical analysis.

Expression of 1,442 genes were significantly linked to viral mimicry scores (p<0.05, |ρ|>0.3). The strongest association was with IFI44L, an interferon-stimulated gene (ρ=0.91, p<0.005). Other top genes included OAS3, MX1, and RSAD2, all tied to interferon signaling. Viral mimicry was moderate to strongly correlated with immune cell markers, including monocytes (ρ=0.47), dendritic cells (ρ=0.37), cytotoxic lymphocytes (ρ=0.36), CD8⁺ T cells (ρ=0.31), and NK cells (ρ=0.29), all with p<0.005. These findings suggest that higher mimicry is associated with broader immune activation. Immune checkpoint gene expression increased alongside viral mimicry scores. The highest correlation was with PD-L1 (CD274) (ρ=0.46), followed by TIGIT, CTLA4, and HAVCR2, each with ρ>0.30 and p<0.005. Viral mimicry scores were not associated with ABT-737 response, and most immune checkpoints lacked predictive value. Modest correlations with azacytidine (ρ=-0.31) and venetoclax (ρ=-0.24) suggest partial enrichment for HMA and BCL2 sensitivity (both p<0.005), but not enough for clinical prediction.

This is the first study showing that viral mimicry in AML is linked to immune activation and higher expression of immune checkpoint genes, but it does not help predict response to BCL2 inhibitors. Although the clinical impact is not yet clear, these findings highlight a unique immune-related signature that could help guide future studies of immunotherapy or hypomethylating agents in AML. Further clinical validation is needed.