EXPLORING CAR-T CELL IMMUNOTHERAPY IN PRECLINICAL AND CLINICAL STUDIES THROUGH MATHEMATICAL MODELS

INTRODUCTION AND OBJECTIVES: CAR-T cell immunotherapy consists of the genetic modification of T lymphocytes aiming to increase their ability to detect and kill tumor cells that express specific antigens. Currently, several CAR-T cell designs and combinations with other therapies have been studied in preclinical trials. The FDA-approved CAR-T therapies have shown expressive complete response rates, but many patients still suffer a relapse and/or become resistant to therapy within the first few months or years. The barriers that prevent the effective success of the therapy are not completely understood but are related to patient-specific and product heterogeneities, among other issues. The CAR-T cell dynamics is marked by a multiphasic behavior with great variability in duration and characteristics. There are still several challenges to be faced and in this context, the mathematical modeling may represent a useful tool to better understand the mechanisms and improve the effectiveness of the therapy. Thus, our objective is to develop mathematical models that help in the planning and in the analysis of preclinical and clinical experiments. Specifically, we aim at identifying key parameters associated with patient-specific responses to CAR-T immunotherapy. MATERIAL AND METHODS: We developed a mathematical model that describes CAR-T immunotherapy against hematological tumors in immunodeficient mice, considering populations of effector and memory CAR-T cells and tumor cells. An in silico platform, called CARTmath, was developed, allowing researchers from different areas to reproduce the obtained results and explore new tests with the model. The model was then extended to represent the dynamics in patients by including the multiphasic dynamics of CAR-T cells through phenotypic differentiation. The CAR-T cells were divided into functional (distributed and effector), memory, and exhausted phenotypes and we considered patient and infused product heterogeneities and antigen-dependent expansion. The model is tested against different hematological malignancies and therapy outcomes. RESULTS AND CONCLUSION: The model developed for preclinical scenarios was able to reproduce several cases reported in the literature, with different CAR receptors and tumor targets. In silico experiments yield insights on immune checkpoint inhibitors, dosing strategies, and uncertainties impacting treatment outcomes. In patients, a wide variety of dynamic behaviors were represented and potential markers of response to therapy were obtained. Specifically, the joint assessment of the area under the curve with the corresponding fraction of non-exhausted CAR-T cells was considered the most promising for outcome classification. The models developed have allowed the exploration of several questions regarding therapy, generating promising insights into underlying mechanisms. We hope that future developments can further contribute to the development of preclinical and clinical studies.