Long-acting GM-CSF-based drug conjugate effectively targets leukemia stem cells in Acute Myeloid Leukemia with limited toxicity

Abstract Acute myeloid leukemia (AML) is a heterogeneous hematopoietic malignancy with poor prognosis and frequent relapse, largely driven by therapy-resistant subpopulations such as slow-cycling leukemia stem cells (LSCs). These quiescent cells evade conventional chemotherapy and contribute to disease persistence. GM-CSF receptors (GM-CSFR) are broadly expressed on AML cells, including LSC-enriched populations, making them attractive targets for ligand-drug conjugates. However, prior GM-CSF-based fusion toxins showed limited efficacy and considerable toxicity due to suboptimal payload potency and drug design. To address these challenges, we developed GM-DM1, a next-generation conjugate that couples the potent microtubule inhibitor DM1 to a long-acting recombinant GM-CSF fusion protein. GM-CSF drives receptor-mediated uptake and promotes cell-cycle re-entry in quiescent LSCs, sensitizing them to DM1-induced mitotic disruption. Specifically, GM-DM1 consists of GM-CSF fused to a human serum albumin–binding nanobody (NbHSA), enhancing pharmacokinetics and stability. The fusion protein (rhGM-CSF-NbHSA) was efficiently expressed in E. coli, enabling scalable, cost-effective production. After purification, DM1 was conjugated to the fusion protein via a Sulfo-SMCC linker to generate the final GM-DM1 conjugate. GM-DM1 retained high-affinity receptor binding and exhibited an approximately 50-fold increase in serum half-life compared to native GM-CSF. It exhibited potent and selective cytotoxicity against GM-CSFR-high AML cell lines (e.g., THP-1, MV4-11, MOLM13) and primary patient samples ex vivo, inducing G2/M arrest, apoptosis, and microtubule disruption, while sparing GM-CSFR-low or GM-CSFR-negative leukemia cells such as K-562 and MOLT-4. Specificity was further validated in engineered HCD-57 cells, where GM-DM1 selectively inhibited human GM-CSFR-expressing cells but not parental GM-CSFR-negative cells, confirming the dependency of GM-DM1 activity on GM-CSF receptor–mediated uptake. In vivo, GM-DM1 significantly reduced leukemic burden and prolonged survival in both AML cell line xenografts and patient-derived xenograft (PDX) models. Histology and flow cytometry confirmed reduced AML infiltration and human CD45⁺ cells in peripheral blood, bone marrow, spleen, and liver. Limiting dilution assays in PDX models demonstrated reduced functional LSC frequency, confirming stem cell-targeting capacity. Importantly, GM-DM1 demonstrated minimal off-target toxicity across multiple models. In colony-forming assays, GM-DM1 selectively suppressed AML-derived CD34⁺ progenitor cells while sparing healthy donor hematopoietic progenitors. In a humanized mouse model reconstituted with human CD34⁺ hematopoietic stem cells, GM-DM1 caused only moderate myeloid suppression, with preservation of bone marrow architecture, lymphoid populations, and peripheral blood counts—supporting a favorable safety profile. To determine whether GM-DM1 selectively targets proliferating leukemic cells without affecting terminally differentiated, non-dividing immune cells that also express GM-CSF receptors, we compared its cytotoxicity in proliferating THP-1 cells versus PMA-differentiated, macrophage-like THP-1 cells. GM-DM1 induced marked apoptosis, chromatin condensation, and viability loss in proliferating THP-1 cells, but had minimal effect on differentiated, growth-arrested macrophage-like cells. These findings confirm that GM-DM1 preferentially targets actively cycling AML cells while sparing mature, quiescent myeloid populations, even in the presence of GM-CSF receptor expression. In summary, GM-DM1 is a receptor-targeted, long-acting, and potent drug conjugate that targets AML blasts and LSCs while sparing normal hematopoietic cells. This strategy, integrating ligand-directed uptake with mitotic disruption, represents a promising therapeutic approach to prevent relapse and improve outcomes in AML.