Postdoctoral fellow National Cancer Institute, Center for Cancer Research, National Institutes of Health North Bethesda, Maryland, United States
Disclosure(s):
Andrew Sweatt, PhD: No financial relationships to disclose
Introduction/Rationale: CAR T cell immunotherapy has had clinical success for liquid tumors but has faced variable potency for solid tumors, encompassing complete tumor clearance to no response. CAR-T responses are known to depend on many biological and clinical factors, but we lack a mechanistic framework that moves these qualitative observations towards quantitative predictors of CAR-T efficacy. Our goal is to determine the mechanisms by which CAR T cells clear tumors, quantitatively model the predictive features, and develop clinical interventions to increase potency.
Methods: We developed an experimental model in neuroblastoma, a pediatric cancer of the neural crest, by screening for gene knockouts that confer resistance to CAR T cells. We co-cultured the resistant lines with CAR T cells and tracked tumor growth, prepared cells for flow cytometry, and collected supernatants for cytokine quantification. We integrated the multimodal data into a machine-learning model to determine predictors of tumor clearance.
Results: The model performed remarkably well, so we identified the features most important for model predictions. The top features included cytokines (TNF, IFNγ), death receptors (Fas), and tumor antigen levels. We asked how the neuroblastoma lines would respond to cytokines alone, and we found similar patterns of resistance as with the CAR T cell co-cultures, indicating resistance independent of ligand availability. We then explored the transcriptional differences between knockout and wildtype cells by RNA sequencing. Interestingly, the genetic knockouts that conferred CAR-T resistance caused neuroblastomas to adopt a distinct resistance phenotype with key ubiquitous pathways downstream.
Conclusion: The data indicate that neuroblastoma may escape CAR T cells by undergoing a phenotypic switch that confers resistance to cytokine death signals. We are excited by the clinical prospect of knowing how to supplement or block cytokines to increase CAR T efficacy by anticipating and countering the phenotypic switch.
