Single-Cell Decoding of the Clinical Outcome-Associated Diversification and Dynamic Changes of CAR-T Cells in Patients with B-cell ALL

Cellular evolution and molecular programs of chimeric antigen receptor-engineered (CAR)-T cells post-infusion are pivotal for developing better treatment strategies. Here, we constructed a high-precision single-cell transcriptional dynamic landscape of 7,578 CAR-T cells from 26 B-ALL patients and evaluated their long-term therapeutic efficacy. We demonstrated that the cytotoxic profile rather than memory property was a favorable biomarker for long-term remission. At the single-cell resolution, we uncovered the vast heterogeneity of CAR-T cells in vivo and identified eight CAR-T cell subtypes. Remarkably, the dominance of cytotoxic subtypes was coincident with long-term remission, while the emergence of subtypes with B-cell transcriptional features could be detected early and predict relapse. Furthermore, we defined transcriptional hallmarks of distinct CAR-T cell populations and revealed molecular changes along computationally-inferred cellular evolution of CAR-T cells in vivo . Collectively, these results illuminated intrinsic properties for durable response and provided molecular cues for improving CAR-T immunotherapy. Herein, we comprehensively dissected the dynamical changes in the transcriptomic profile of CAR-T cells enriched from IPs and paired peripheral blood samples longitudinally, collected at serial time points post-infusion for up to 15 months, and explored their clinical implications in the long-term response or resistance in 26 B-ALL patients. In order to obtain highly reliable CAR-T cells, a two-step purification procedure was adopted to purify CAR-T cells from infusion products and peripheral blood samples. Firstly, immunophenotypic CAR-T cells (CD3+ CAR+) were isolated and individually sorted into 96-well plates by Moflo XDP (Beckman Coulter). Then, sorted CAR-T cells were further validated by the amplification of CAR sequence in single-cell polymerase chain reaction (scPCR) before library preparation.