CELL INTELLIGENCE · 肿瘤细胞治疗研究

CD19-CAR-T cells are a major therapy for relapsed/refractory B-cell malignancies, yet ~50% of patients relapse after infusion. Clinical data suggest that tisa-cel treatment is associated with higher rates of CD19 mutation than axi-cel, though direct comparisons have been lacking. Although both products target the FMC63 epitope, they employ different structures of hinge/transmembrane/costimulatory domains (CD8 /CD8 /4-1BB in tisa-cel, CD28/CD28/CD28 in axi-cel). Here, we show that CAR structure is a critical determinant of CD19 loss in tumor B cells. Specifically, repeated exposure to CD19-4-1BB-based CAR-T cells, but not CD19-CD28-based counterparts, drives FMC63-epitope and total CD19 protein loss. Consistent with clinical observations, resistance to CD19-4-1BB-based treatment is associated with aberrant splicing, loss of heterozygosity, and frameshift/missense mutations in CD19 in exons encoding the FMC63 epitope. Mathematical modeling indicates that the failure of CD19-4-1BB-based CAR-T cells to eliminate CD19 low cells, unlike CD28-based CARs, promotes the expansion of resistant populations. These results suggest that optimizing recognition of CD19 low cells could prevent the emergence of resistant clones and reduce relapse rates. We also show that commonly used diagnostic anti-CD19 monoclonal antibodies detect epitopes distinct from FMC63 and may therefore give misleading CD19-positive results, underscoring the need for FMC63-specific assessment to guide diagnostic and therapeutic decisions.