Abstract
Adoptive immunotherapies have transformed the treatment landscape for several hematologic malignancies, yet their clinical impact in Acute Myeloid Leukemia (AML) remains limited. Unlike lymphoid malignancies, AML is characterized by extensive clonal heterogeneity and lineage plasticity, which complicate durable immune control. A central obstacle is the scarcity of surface antigens that are both consistently expressed on leukemic blasts and absent from critical normal hematopoietic populations. Current targets such as CD33 and CD123 exhibit significant expression overlap with hematopoietic stem and progenitor cells (HSPCs), increasing the risk of on-target off-leukemia toxicity. These challenges highlight the need for systematic strategies to identify AML-restricted surface markers that could serve as safer and more selective immunotherapeutic targets.
To address this challenge, we deployed a multi-modal discovery pipeline integrating high-throughput single-cell transcriptomics, enhanced whole exome sequencing (eWES), and quantitative surface proteomics across a genetically diverse cohort of 28 primary AML patient samples and 3 healthy bone marrow controls. Unsupervised clustering and somatic mutation-based tumor assignment enabled hierarchical resolution of leukemic populations, spanning hematopoietic stem and progenitor cell (HSPC)-, GMP-, monocyte-, and dendritic cell-like subtypes.
To identify therapeutic vulnerabilities within this landscape, we systematically interrogated surface protein expression across malignant and normal compartments. A subset of membrane proteins emerged as differentially expressed, with limited or absent representation in healthy HSPCs. This included established and emerging targets such as CD123, CLEC12A, ADGRE2, and CD85d (ILT4/LILRB2). Notably, CD85d—a member of the leukocyte immunoglobulin-like receptor family traditionally associated with myeloid regulation—consistently emerged as a top candidate based on its prevalent expression across AML phenotypes and restricted presence in normal hematopoiesis. CD85d expression was validated through high-dimensional flow cytometry across primary AML samples and healthy donors. It was absent from immunophenotypically defined HSPC compartments, including long-term HSCs (LT-HSC), multipotent progenitors (MPP), common myeloid progenitors (CMP), granulocyte-monocyte progenitors (GMP), common lymphoid progenitors (CLP), and primitive HSC subsets. Additionally, CD85d remained undetectable across lymphoid tissues and showed negligible expression on peripheral myeloid subsets, including monocytes and dendritic cells. These findings were further corroborated by quantitative surface proteomics, reinforcing the antigen's restricted expression profile and selective enrichment within malignant compartments. To assess off-tumor expression, we screened representative non-hematopoietic tissues using publicly available transcriptomic datasets. CD85d expression remained confined to leukemic contexts, with no appreciable signal detected in vital organ systems, reinforcing its translational potential as a selectively targetable antigen. To explore translational potential, we engineered chimeric antigen receptor (CAR) T cells with specificity for CD85d-expressing populations. These CAR-T exhibited selective cytolytic activity toward CD85d⁺ leukemic cells while sparing CD85d⁻ counterparts, demonstrating antigen-restricted reactivity and supporting the feasibility of CD85d-directed cellular immunotherapy.
In summary, we identify CD85d as a promising immunotherapeutic target in AML, supported by multi-omic evidence of restricted expression in normal tissues and functional validation of selective killing. These findings provide a strong rationale for the continued development of CD85d-directed cellular therapies and highlight the value of integrative surfaceome profiling in the rational design of precision immunotherapies for myeloid malignancies.
