(M-056) A Translational Quantitative Systems Pharmacology (QSP) Modeling Framework to Predict Interleukin-15 (IL-15) Cytokine Levels in Multiple Myeloma Patients after Anti-BCMA CAR-NK Cell Therapy

Masood Khaksar Toroghi, na – Associate Director, Takeda Development Center Americas Inc; Jia Li, na – Associate Director, Takeda Development Center Americas Inc; Yuri Kosinsky, na – na, Modeling and Simulation Decisions; Haiqing Wang, na – Associate Director, Takeda Development Center Americas Inc; Kathryn Fraser, na – Associate Director, Takeda Development Center Americas Inc; Michael Curley, na – Sr Dir Global Program Leader, Takeda Development Center Americas Inc; Ajeeta Dash, na – Senior Director, Takeda Development Center Americas Inc; Kirill Peskov, na – na, Modeling and Simulation Decisions; Girish Bende, na – Director, Takeda Development Center Americas Inc; Ganesh M. Mugundu, na – Head of Cellular Pharmacology and Modeling, Takeda Development Center Americas Inc

Background: Engineering chimeric antigen receptor natural killer cells (CAR-NK) with soluble interleukin-15 (IL-15) has been shown to enhance NK cell proliferation, proinflammatory cytokine secretion, and cytotoxic activity. While CAR-NK therapies have demonstrated safety in clinical trials, there is uncertainty in the levels of circulating IL-15 that may be secreted in patients due to engineered IL-15 on NK cells. The objective of this work was to develop a QSP modeling framework to predict the levels of circulating IL-15 and associated cellular kinetics (CK) of CAR NK cells in virtual multiple myeloma (MM) patients after administration of allogeneic anti-BCMA CAR-NK cells.

Methods: To develop the QSP model, first, short term cytotoxicity of untransduced NK and anti-BCMA CAR+NK cells were modeled to capture the impact of tumor target cell BCMA density on anti-BCMA CAR-NK mediated killing. Second, by integrating the cytotoxicity model, disease biology and drug mechanism of action (MOA), an in vivo drug-disease QSP model was constructed. The model was calibrated using in-house preclinical in vivo data (i.e., CK, IL-15 levels and bioluminescence intensity). The in vivo QSP model was further translated into a clinical setting based on available published data. The translational QSP framework included multiple biological components such as endogenous NK and T cell (i.e., CD4+ and CD8+) turnover, BCMA expression levels, the effect of preconditioning using lymphodepletion agents, human IL-15 turnover, cell therapy characteristics (e.g., CAR expression levels) and CAR-NK mechanism of action. Multiple simulations were performed to project blood IL-15 levels and the CK of CAR+ NK cell expansion in virtual MM patients in different dosing regimens.

Results: We developed a QSP modeling platform in a stepwise manner to establish an understanding of the impact of multiple factors, including BCMA density, drug MOA, IL-15 secretion, CAR expression, dose and disease biology on the proliferation and persistence of CAR+ NK cells and on circulating IL-15 levels in a clinical setting. The findings indicate a dose-dependent increase in CAR-NK cell expansion in virtual patients, similar to those observed in in-vivo studies. Also, the model prediction shows an increase in IL-15 levels as dose increases, which is consistent with the observed preclinical data. In clinical trial simulations, the model projection of peak IL-15 levels for various dosing scenarios such as 100 million, 300 million and 1 billion anti-BCMA CAR+ NK cells were approximately 254, 434 and 796 pg/mL, respectively, which were significantly lower than the reported peak levels (Cmax~5662 pg/mL) associated with toxicities in clinical studies evaluating recombinant human IL-15 in patients with metastatic malignancies (Conlon KC., et al, 2019).

Conclusions: This modeling framework has the potential to be used to support starting dose and dose escalation estimations for CAR-NK cell therapies.

Citations: Conlon KC., et al., Clin Cancer Res. 2019