CAR-T with ICOS costimulation lacks therapeutic efficacy in patients due to strong trogocytosis and failure of expansion 2261203

Abstract Introduction CAR-T therapy targeting B-cell antigens is effective for relapsed hematologic malignancies, but antigen escape remains a challenge. We previously showed dual CD19/CD20-4-1BB CAR-T therapy was highly effective (Nat Med, 2020). We further developed trispecific CAR-T cells (CD19/CD20/CD22) incorporating ICOS signaling (Sci Transl Med, 2021). Despite promising preclinical activity, trispecific-ICOS CAR-T cells failed to expand and showed no efficacy in a Phase I trial (NCT05094206; ASH 2024). Here, we investigated mechanisms underlying this failed response. Methods Based on our IRB approved protocol and using patient-derived PBMCs and leftover samples from treated patients, we generated tested bispecific-4-1BB, bispecific-ICOS and trispecific-ICOS CAR-T cells. Results NSG mice receiving bispecific- or trispecific-ICOS CAR-T cells had worse survival than those with 4-1BB CAR-T cells. In vivo, ICOS CAR-T cells failed to expand and showed an exhaustion phenotype. When co-culture with Raji cells, ICOS CAR-T cells significantly enhanced surface-FMC63 loss compared to 4-1BB counterparts. Surface-FMC63 expression was mostly recovered at 40 hours on 4-1BB CAR-T cells, but not on ICOS CAR-T cells. Since trogocytosis can mediate CAR internalization and antigen loss on tumor, we examined CD19 transfer. ICOS CAR-T cells displayed higher and more sustained CD19 uptake and fratricide reflected by apoptosis than 4-1BB cells. Consistently, when co-culture with patient autologous B cells, ICOS CAR-T cells again showed greater CAR loss and apoptosis. Extended co-culture mimicking chronic antigen stimulation revealed expansion of 4-1BB but not ICOS CAR-T cells, the latter showing reduced proliferation (Ki67) and elevated exhaustion markers. Conclusion In conclusion, the ICOS domain promotes trogocytosis and fratricide, explaining the poor persistence and function of ICOS CAR-T cells observed in our trial. Thus, our findings raise significant caution for using CAR constructs that rely on ICOS for CAR activation. Funding Source n/a Topic Categories Tumor Immunology: Cellular Responses and Tumor Microevironment (TIME)