Longitudinal metagenomic profiling of gut microbiome dynamics in B-cell lymphoma patients receiving CAR-T cell therapy

Abstract

Chimeric antigen receptor T-cell (CAR-T) therapy has improved outcomes in B-cell malignancies, but treatment response and immune-related toxicities remain heterogeneous. We characterized peri-infusion gut microbiome dynamics and examined their associations with subsequent response and toxicity. We performed shotgun metagenomic sequencing of 103 fecal samples from 21 patients with B-cell lymphoma collected on days −7, 0, 7, 14 and 28 relative to CAR-T infusion. We assessed pre-infusion stability, longitudinal microbiome remodeling, and associations with treatment response, cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). Longitudinal data were analyzed using mixed-effects models, and composition-aware sensitivity analyses incorporated centered log-ratio transformation and false-discovery-rate correction. Day −7 clinical and microbial features were also evaluated using exploratory nested cross-validation for response and CRS. The gut microbiome remained broadly stable at the community level before infusion but underwent early, patient-specific remodeling after CAR-T therapy, characterized by a transient decline in alpha diversity and increased within-patient compositional deviation from baseline. Against this dynamic background, pre-infusion microbial variation was associated with subsequent clinical outcomes. Responders differed from non-responders in baseline microbial diversity, community structure and selected Sutterella and Segatella taxa. Patients who developed CRS showed lower microbial diversity and distinct community composition, whereas no comparable association was detected for ICANS. Exploratory nested cross-validation yielded AUCs of 0.51, 0.55 and 0.72 for the clinical, microbiome and combined response models, respectively, and 0.44, 0.85 and 0.83 for the corresponding CRS models. CAR-T therapy was accompanied by early, patient-specific gut microbiome remodeling, while pre-infusion microbial variation was associated with subsequent treatment response and CRS occurrence. These hypothesis-generating findings require confirmation in larger, independent prospective cohorts.