Purpose: T-cell engager (TCE) molecules activate the immune system and direct it to kill tumor cells. The key mechanism of action of TCEs is to crosslink CD3 on T cells and a tumor associated antigen (TAA) on tumor cells. The formation of this trimolecular complex (i.e. trimer) mimics the immune synapse, leading to therapeutic-dependent T-cell activation and killing of tumor cells. Computational models supporting TCE development must predict trimer formation accurately.
Methods: Here, we present a next-generation two-step binding mathematical model for TCEs to describe trimer formation. Specifically, we propose to model the second binding step with trans-avidity and as a two-dimensional (2D) process where the reactants are modeled as the cell-surface density, and compared it with the 3D binding model where the reactants are described in terms of concentration.
Results: Same parameter values were applied to both 2D and 3D models whenever possible. The only difference is the description of the second binding step for trimer formation. Compared to the 3D binding model, the 2D model predicts less sensitivity of trimer formation to varying cell densities, which better matches changes in EC50 from in vitro cytotoxicity assay data with varying E:T ratios. In addition, when translating in vitro cytotoxicity data to predict in vivo active clinical dose for blinatumomab by predicting trimer formation in the bone marrow of leukemia patients. The choice of model leads to a notable difference in dose prediction. The 2D model projects that the active dose is 21–24 ug/day, close to the clinically approved dose of 28 ug/day. In contrast, the 3D model projects a dose of 0.006-0.0068 ug/day, more than three orders of magnitude below the clinical dose.
Conclusion: In conclusion, the 2D model with trans-avidity to describe trimer formation is an improved approach for TCEs and is likely to produce more accurate predictions to support TCE development.