QSP model trained on preclinical data informs clinical dose selection for a T-cell engaging bispecific antibody for solid tumor indications

Objectives: T cell-dependent bispecific antibodies (TDBs) remain a promising modality for treating cancer by facilitating cell killing via immune synapse formation between T cells and target cells. However, understanding the quantitative relationship between serum pharmacokinetics (PK), CD8+ T cell expansion, and efficacy (tumor reduction) represents a critical question in the development of TDBs. Additionally, dose finding solely based on the clinical PK/PD biomarker relationship in circulation will likely present challenges. We built a quantitative systems pharmacology (QSP) model to capture the mechanism of action of a TDB targeting immune cells in murine solid tumors. We translated the model to inform clinical dose selection by estimating the dose level achieving potential maximum effect in patients. Methods: A minimal physiologically-based pharmacokinetic (mPBPK) model1 was developed to capture the TDB PK, CD8+ T cell expansion, and tumor efficacy in preclinical mouse models and to predict a clinically active dose. The model was calibrated to single-dose PK data in mice. In the model, the TDB distributes into the tumor, binds to its tumor target and CD8+ T cells to form immune synapses, thus driving expansion of CD8+ T cells and tumor efficacy. Clinical simulations, informed by allometric scaling of nonspecific clearance in cynomolgus monkeys, were performed to estimate the TDB dose achieving the maximum amount of immune synapses formed per CD8+ T cell in the tumor. Results: The mPBPK model was calibrated to the PK in mice. Nonlinear PK was captured by using both a nonspecific and specific clearance term. The model predicts the dynamics of the immune synapse formed in the tumor between the TDB, target cells, and CD8+ T cells based on reported in vitro binding affinities. The model is calibrated to mouse tumor CD8+ T cell data. The CD8+ T cell expansion drives tumor killing, which is captured by the model across three dose levels and a vehicle control group. The ability of the model to capture tumor inhibition data suggests that immune synapse formation may be a useful surrogate for tumor efficacy in clinical predictions for TDBs. Thus, we employed the model to identify the dose range that maximizes the amount of synapses per CD8+ T cell in the tumor. Sensitivity analyses of key model parameters demonstrated that while many parameters impact the amount of synapses per CD8+ T cell, the dose level driving the maximum amount of synapses was relatively insensitive to model parameters within reasonable ranges. Conclusion: Our results outline a modeling framework for capturing TDB PK, CD8+ T cell expansion, and efficacy in solid tumors. The extension of our model to guide clinical dose selection is paramount due to the assumed difficulty in ascertaining a peripheral PK/PD biomarker relationship in clinical data. This work highlights the impactfulness of translational QSP modeling to accelerate development of TDBs.Citations: [1] Cao Y et al. JPKPD, 2013.