Abstract 3957: Surface protein heterogeneity of acute myeloid leukemia is associated with distinct CD8 T cell landscape alterations

Abstract Acute myeloid leukemia (AML) is a heterogenous, aggressive hematologic malignancy with a poor response to immunotherapy. We applied non-negative matrix factorization (NMF) to cytometry by time-of-flight (CyTOF) data to classify AML into subtypes based on surface protein expression. We hypothesized that the resulting protein-defined subtypes of AML correlate with alterations to the CD8 T cell landscape, providing insight into the ineffectiveness of immunotherapy in AML. Primary healthy and AML human bone marrows were stained with two CyTOF antibody panels to identify myeloid and T cell proteins. We performed NMF on the cell-protein counts matrix of the myeloid panel to determine latent protein factors present among AML samples, identifying five factors (NMF1 to NMF5) each enriched for proteins commonly associated with distinct phenotypic states. To compare to a supervised approach, the data were also clustered using the Leiden algorithm and annotated based on similarity to healthy phenotypes resulting in ten myeloblast clusters (HSC-like, monocyte-like, etc). This revealed an association among NMF factors and clusters identified through manual annotation. For example, NMF3 was defined by a primitive pattern whereas NMF5 had a pattern associated with mature AML cells. Thus, protein-defined NMF factors showed biological differences and correlated with myeloid differentiation states, allowing us to categorize AML protein heterogeneity into its predominant protein patterns. We then used the Leiden algorithm to cluster CD8 T cells. Differential abundance analysis showed an enrichment of effector memory T cells re-expressing CD45RA (TEMRAs) in AML along a gradient with only some samples being significantly different (sFDR < 0.1). Suspecting this range was related to AML heterogeneity, we correlated each T cell subset with the respective sample’s dominant AML NMF factors. We found several associations among NMF factors and enrichments of CD8 T cell subsets compared to healthy. NMF4, composed primarily of monocyte-like and dendritic-like cells, was significantly associated with an enrichment of CD57+ TEMRAs (p = 2.8e-05). These showed significantly increased expression of activation markers (CD95, p_adj = 5.53e-39; CD69, p_adj = 1.5e-73) and a significant reduction in GZMB (p_adj = 1.3e-62). We also observed significant marker differences among factors: CD57+ TEMRAs associated with NMF3 significantly upregulated activation markers (HLADR, p_adj = 1.7e-73; GZMB, p_adj = 9.3e-39) compared to those associated with NMF4. There was no significant difference in the abundance of naive T cells between healthy and promonocyte-like factor NMF1 (p = 0.81). These data suggest that AML phenotype may directly or indirectly influence T cell phenotypes and function, which could hold implications for immunotherapeutic use in AML. Citation Format: Justin A. Cartailler, Y. Emily Chu, Chad R. Potts, Brent Ferrell. Surface protein heterogeneity of acute myeloid leukemia is associated with distinct CD8 T cell landscape alterations [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3957.