Dissecting Bcl-2 Sensitivity In Acute Myeloid Leukemia: A Quantitative Systems Pharmacology Model For Venetoclax-Azacitidine Therapy

Objectives: Combining Venetoclax (VEN, Bcl-2 inhibitor) and Azacitidine (AZA, hypomethylating agent)has recently become the standard of care for newly diagnosed acute myeloid leukemia(AML) in patients unfit for intensive chemotherapy. Nevertheless, challenges remain todescribe heterogeneity in response and development of relapse. This study aimed todevelop a QSP model to study the effect of relative levels of Bcl-2 family members onpatient response to VEN and VEN-AZA.Methods: We developed a QSP model of VEN pharmacodynamics (PD), including representations ofcellular uptake, binding to Bcl-2 and intrinsic apoptosis through mitochondrial outermembrane permeabilization. We combined it with a cell line model and aphenomenological model for AZA PD to obtain a model for in vitro viability assays.Including a mechanistic model of AML proliferation and a simple PK model for bothtreatments yielded a model for human AML treated with VEN-AZA. We derived sensitivityto relative levels of Bcl-2 family members from basic assumptions and informed variabilityin cell lines/population using recent clinical results on quantifying Bcl-2 sensitivity by MACscore [1]. We then integrated in vitro viability assay data for VEN [2] (ML-2 cells) and thecombination of VEN and AZA [3] (Oci-AML5 cells) to inform the corresponding model.Lastly, we used clinical trial data from VEN and AZA monotherapy trials [4,5] in order totranslate our in vitro model to an in vivo human model for AML treatment. Simulationswere run on the collaborative QSP simulation platform jinkō.Results: Our model of intrinsic apoptosis captures single drug and drug-drug interaction data forVEN and AZA in vitro viability assays. Clinical translation of the model allowed tosuccessfully match response variability in the context of VEN monotherapy (bone marrowblast count changes after four weeks) and long term behavior in AZA monotherapy (meanblast and white blood cell counts). In silico trials coupled with sensitivity analysis andstratification then allowed to study model predictions, showing qualitative agreement fordependence of response markers on MAC score with clinical observations.Conclusions: We present, to our knowledge, the first QSP model for VEN-AZA that reproduces in vitroviability and clinical trial data for VEN and AZA monotherapy while deriving variability inresponse from quantitative clinical data on Bcl-2 sensitivity. This paves the way towards anAML QSP platform including additional treatment combinations targeting intrinsicapoptosis.Citations: [1] Waclawiczek et al. (2023), Combinatorial BCL2 Family Expression in Acute MyeloidLeukemia Stem Cells Predicts Clinical Response to Azacitidine/Venetoclax[2] Seipel et al. (2022), Rationale for Combining the BCL2 Inhibitor Venetoclax with thePI3K Inhibitor Bimiralisib in the Treatment of IDH2- and FLT3-Mutated Acute MyeloidLeukemia[3] Cojocari et al. (2021), Pevonedistat and azacitidine upregulate NOXA (PMAIP1) toincrease sensitivity to venetoclax in preclinical models of acute myeloid leukemia[4] Konopleva et al. (2016), Efficacy and Biological Correlates of Response in a Phase IIStudy of Venetoclax Monotherapy in Patients with Acute Myelogenous Leukemia[5] PMDA of Japan, CTD Azacitidine 2.7.6