A mechanistic model for dynamics between CAR T cells and target cells captures features that determine killing profiles

Abstract

Chimeric antigen receptor (CAR) T cells represent a potent, programmable therapeutic that repurposes T cell cytotoxicity toward target cell elimination. Direct killing of tumor cells has been demonstrated in several cancer contexts but deficits in destroying solid tumors have been attributed to molecular and mechanical complexity. Understanding and predicting the efficacy of CAR T cell therapy requires a rigorous framework to capture the mechanistic interactions between immune cells and tumor targets. Toward enumerating such features and to quantitatively describe tumor cell survival in response to CAR T cell visitation, we propose a parametric hazard rate model for the right-censored conditional lifetime of a tumor cell, having the form of a time-discounted integral of cell:cell engagement. The model, conditioned using data on the numbers, durations, and modes of contacts made during the stochastic encounters between the two cell types, extracts mechanistic details from the experiments. We suggest two types of parameters to encapsulate features of CAR T cell killing potency ({kappa}) and the resilience of the tumor ({gamma}), respectively. These parsimonious parameters, nevertheless, generate substantial insights. Firstly, phenotypic heterogeneity from interactions mediated by substrate-engaged CAR T cells dominate killing outcomes, while those initiated in suspension contribute minimally. This emphasizes that a subset of the CAR T cell population is dominating the population outcome. Secondly, the model can be used to predict the killing as a function of time upon perturbation of the system. In this case, biasing the arrival process, toward more engagement in the adherent mode, can tune the rate of killing. Collectively, insights from the model framework imply that CAR T cell accumulation and killing efficiency depend not on the sum of individual T cell interactions, but rather on the weighted, time-dependent sum of heterogeneous interactions that are environmentally modulated. The model facilitates predictive insights into how contact sequences modulate CAR T cell responses and may provide strategies to alter therapeutic regimens.