Clinically relevant AML modeling with in vitro bone marrow niche and in vivo PDX approaches

Introduction Acute Myeloid Leukemia (AML) remains therapeutically challenging due to various genetic alterations, drug resistance, and relapse at the clinical stage. Preclinical evaluation is therefore critical to identify effective novel therapeutic strategies. Optimal ex vivo growth and in vivo engraftment of primary AML cells are challenging, and current preclinical screening of oncology drugs is often compromised by the lack of advanced models that mimic therapeutical response in the native human bone marrow niche (BMN). To address this challenge, we present the integration of our 3D BMN in vitro platform composed of hematopoietic cells cultured on the perivascular site to capture response and resistance patterns with in vivo AML Patient Derived Xenograft (PDX) models, giving translational preclinical insights for therapy development. Methods Three AML PDX models were systemically engrafted in immunodeficient mice and exposed to standard of care (SoC) drug panels containing FLT3 and IDH inhibitors, as well as classical chemotherapy. Endpoint readouts included survival, clinical signs, and peripheral leukemic burden by flow cytometry (blood, bone marrow, spleen). In parallel, human CD45+ cells from the same AML PDX models were integrated in 3D mesenchymal-endothelial networks (BMN platform) and exposed to SoC and targeted drug panels for 10 days. Quantitative assessment of the in vitro drug effects was performed using a proprietary automated high content imaging (HCI) analysis platform. Immunofluorescence (IF) staining, NSG, cytogenetics, and flow cytometry were used for in-depth characterization of AML PDX. Results Engraftment of AML PDX (AM9626, AM9627, and AM9628) cells into mice was validated upon evaluation of leukemia signs, immunophenotypic profiling, and histopathology. Tumor burden growth (% of hCD45+) assessment across AML PDX models harboring FLT3 and IDH1 mutations revealed sensitivity and resistance patterns to targeted inhibitors and SoC agents. HCI analysis showed decreased AML PDX cell counts at the highest doses of FLT3 and IDH1 inhibitors in comparison to the control in the BMN platform, identifying different drug sensitivities across the PDX models upon support and protection provided by human mesenchymal and endothelial cells. Conclusion Here, we present a combination of advanced, patient-relevant preclinical platforms that offer valuable information on how a drug performs under divergent conditions. The in vivo AML PDX system recapitulates disease progression and enables systemic pharmacokinetic and pharmacodynamic assessment. In parallel, the in vitro BMN platform provides a quantitative measurement of treatment response in leukemia cells attached to the perivascular site. These findings underscore the utility of integrating in vitro and in vivo preclinical models to support AML therapeutic development and prioritize compounds for clinical translation.