Metabolomic Profiling Outperforms Cytokines in Predicting CAR T-Cell Neurotoxicity
Metabolic pathway scores consistently distinguished patients who developed severe neurologic events after CAR T-cell therapy across six clinical trials, surpassing traditional inflammatory markers in predictive performance.
Metabolomic pathway-based scores outperformed traditional inflammatory protein markers in predicting severe neurologic events after chimeric antigen receptor (CAR) T-cell therapy, according to the largest metabolomic analysis to date in this setting, including B-cell acute lymphoblastic leukemia (ALL), diffuse large B-cell lymphoma (DLBCL), and other B-cell malignancies.
Across 6 clinical trials, metabolic signatures consistently distinguished patients who developed grade 3 or higher neurologic events from those who did not, whereas cytokines such as IL-6 and TNFα showed inferior predictive performance. These results were announced in a company press release.
Investigators analyzed more than 3800 longitudinal serum and plasma samples, along with cerebrospinal fluid (CSF) specimens collected during neurotoxic episodes, from patients treated with the anti-CD19 CAR T-cell therapies axicabtagene ciloleucel and brexucabtagene autoleucel. The multi-trial meta-cohort encompassed individuals with B-cell ALL, DLBCL, mantle cell lymphoma, and follicular lymphoma, reflecting the breadth of current CAR T-cell applications and suggesting potential relevance across additional indications.
High-grade neurologic events were strongly associated with increased tryptophan catabolism. Patients who developed severe toxicity exhibited elevated downstream metabolites in the kynurenine pathway, including quinolinate, both before and after infusion. These findings implicate heightened activation of the tryptophan-kynurenine axis and N-methyl-D-aspartate receptor-linked excitotoxic processes in the pathogenesis of CAR T-cell-associated neurotoxicity.
Alterations in arginine metabolism also characterized severe cases. Investigators observed increased urea cycle activity and accumulation of acetylated polyamines, including N1 and N12-diacetylspermine, consistent with amplified immune activation and cellular stress. These pathway-level perturbations were reproducible across studies, underscoring their robustness.
Analyses of CSF confirmed that metabolic disruptions extended to the central nervous system. During neurologic events, CSF samples demonstrated elevated glutamate and other stress-related metabolites paralleling those observed in peripheral blood, supporting a direct link between systemic metabolic reprogramming and cerebral toxicity.
Importantly, composite metabolite-derived pathway scores surpassed conventional inflammatory markers in identifying patients at risk for severe neurologic events, with statistically significant improvements (P <.05 across models). Several metabolites associated with neurotoxicity, including quinolinate and acetylated polyamines, also correlated with inferior disease outcomes.
Machine-learning models reinforced the central role of the tryptophan-kynurenine pathway. Together, these data suggest that integrating metabolomic profiling into CAR T-cell therapy, whether for DLBCL, ALL, or other future applications, may refine risk stratification and illuminate therapeutic targets to mitigate life-threatening neurotoxicity.