Quantitative Systems Pharmacology Modeling to Inform Preclinical-to-Clinical Translation of T-cell Engager Programs

T-cell engagers (TCEs) have recently surfaced as promising new cancer therapeutics. However, TCE- induced cytokine release syndrome (CRS) in patients presents a significant challenge in clinical TCE development. Furthermore, preclinical experiments used to advance candidates poorly mimic patient physiology and may not translate well to clinical success. To understand the most relevant factors in TCE preclinical-to-clinical translation, we developed a QSP model to probe how E:T ratio, TCE affinities, and receptor density control TCE potency and cytokine release. The model was trained on in vitro data for AMG 673, a CD33-targeting TCE for acute myeloid leukemia treatment, and the model includes meta- models for (1) TCE synapse formation; (2) Receptor trafficking; (3) T-cell activation, proliferation, and apoptosis; (4) drug-induced cancer cytotoxicity; and (5) Tumor necrosis factor α (TNFα) release. We then performed a sensitivity analysis to assess how E:T ratio and TCE arm affinities influence potency and cytokine release. The model achieved a reasonable 24.6% geometric mean percent error across observation data. We found that by increasing CD3 KD above 10 µM, lysis can remain high with low TNFα release across 4 orders of magnitude of TCE concentration. We also found that cytokine release is highly sensitive to E:T ratio and effector cell concentrations. This model is generalizable to assess affinity- potency/safety relationships of other TCEs, and model trends have been validated across different E:T ratios and TCE arm affinities for other TCE candidates.