An immune exhaustion signature predicts prognosis and identifies patients with diffuse large B-cell lymphoma (DLBCL) who derive preferential benefit from chimeric antigen receptor (CAR)-T cell therapy.

Background: The tumor microenvironment (TME) is a key determinant of prognosis in diffuse large B-cell lymphoma (DLBCL). While T-cell exhaustion is implicated in therapeutic failure, its precise molecular hallmarks and utility for predicting response to modern immunotherapies, such as chimeric antigen receptor (CAR)-T cell therapy, remain unclear. Methods: We performed an integrative analysis of transcriptomic and clinical data from multiple DLBCL cohorts (The Cancer Genome Atlas [TCGA], GSE181063, GSE10846, GSE248835, GSE182434). We used unsupervised clustering, exploratory analysis of single-cell RNA sequencing data, and the least absolute shrinkage and selection operator for variable selection (LASSO-Cox) regression to characterize the exhausted TME, construct a prognostic model, and evaluate its predictive value for CAR-T cell therapy. The model's dynamic behavior was assessed in a proof-of-concept longitudinal cohort of patients treated with the T-cell-engaging bispecific antibody glofitamab. Results: We identified a "high-exhaustion" subtype associated with significantly poorer overall survival (OS; log-rank Conclusions: A state of active T-cell exhaustion and a suppressive TME drive the adverse immune phenotype in DLBCL. Our IERPS model captures this dysfunctional state, acting as a powerful prognostic tool and, more importantly, as a potential predictive biomarker to identify high-risk patients who appear to overcome their inherently poor prognosis through CAR-T cell therapy.