(49g) Single-Cell Multi-Omics Landscape of CD19 CAR-T Cell Infusion Products Predicts Long-Term Responses in Pediatric B Cell Acute Lymphoblastic Leukemia (Invited Speaker)

Abstract: Chimeric antigen receptor (CAR)-engineered T cell therapy has demonstrated high response rate in the treatment of B cell acute lymphoblastic leukemia (B-ALL), but ~10% of patients still do not respond and more than 50% of patients relapse within a year. How to predict the clinical response prior to treatment remains a challenge for CAR-T cell therapy. We investigated single-cell transcriptome and surface epitome sequencing of CD19 CAR-T cell infusion products at the basal state, upon antigen-specific stimulation, upon CD3/CD28 TCR stimulation, and under a control stimulation with mesothelin antigen presenting cells. We sequenced 75 samples including replicates and total 101,326 single cell transcriptomes with 18 proteins from 12 pediatric ALL patients treated with CD19-BB-z CAR-T therapy at UPenn. Computational data analysis revealed a highly heterogenous activation landscape of infusion products and identified multiple molecular signatures to not only predict non-responders from responders but also long-term durable responders vs recurrent patients (relapse). These signatures are mainly associated distinct T cell subset cytokine profiles such as Th1, Th2, and inhibitory factors as well as T cell memory and metabolic activities. How to predict long term responses is a major challenge in the field of CAR-T therapy. Our study provides an informative platform to discover predictive biomarkers and identify the first cellular signature that can predict relapse in patients with B-ALL.