The 1000 mouse experiment: quantifying the variability of tumor response to CAR T cell therapy

Abstract

Abstract Chimeric Antigen Receptor (CAR) T cell therapy is a type of cancer immunotherapy designed to target specific antigens on the surface of tumor cells using engineered T cells. While effective, there is often variability in patient responses and even within an experiment given the same experimental conditions. Analyzing this variability may help identify key modalities to successful CAR T therapy. We present a platform to explore this variability by quantifying the luminescence signal from in vivo tumor growth over time in a mouse model of human CAR T response. We developed a Python toolkit (https://github.com/soorajachar/radianceQuantifier) to automate labeling, cropping, quantification, and plotting of bioluminescent tumor image data. Applying this toolkit to digitize dozens of prior in vivo experiments conducted over several years allowed us to generate a dataset consisting of over 1000 mice. We then performed statistical analyses and mathematically modeled the time dynamics of tumor responses to CAR T cells in this dataset. We identified three distinct phases of tumor response: initial tumor growth, tumor decay due to killing by CAR T cells, and tumor relapse. These phases were fit to differential equations to extract the growth and decay rates of the tumor for each mouse. Upon fitting our model to the experimental data, we found that the growth rate of relapsed tumors was slower than the initial tumor growth rate within individual mice, suggesting a selection of edited tumor cells post CAR T cell therapy. Future work will leverage our quantitative framework to explore the alterations between relapsed tumor cells and those that are eradicated by CAR T cells. Supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute (Bethesda, MD)