Metabolon metabolomics identifies metabolic predictors of CAR T neurotoxicity

The metabolomics dataset revealed reproducible biochemical patterns that identify patients at high risk of severe neurologic events after CAR T therapy within a multi-trial cohort. Using an untargeted discovery workflow, investigators screened more than 3,800 longitudinal serum and plasma samples, plus a limited cerebrospinal fluid series, across a meta-cohort of 6 clinical studies to define these signatures.

Analyses highlighted accelerated degradation of tryptophan with accumulation of quinolinate-family metabolites, shifts in arginine-derived polyamine biochemistry including elevations in N1,N12-diacetylspermine, and blood-to-CSF concordance for excitatory compounds such as glutamate. These pathway perturbations aligned with episodes classified as high-grade neurologic toxicity (defined as grade ≥3), and they persisted before and after cellular therapy in several patients.

When translated into composite pathway scores, the metabolite readouts separated high-risk subjects more reliably than conventional circulating inflammatory readouts like IL-6 or TNFα in the same samples. Machine-learning models prioritized the tryptophan-kynurenine route as a dominant contributor to predictive performance and showed links between these metabolites and poorer clinical trajectories.

CSF measurements taken during neurotoxic episodes confirmed central nervous system involvement rather than peripheral-only signals, supporting a mechanistic role for excitotoxic and metabolic stress in toxicity. The study therefore provides both candidate biomarkers for earlier detection and biochemical targets for potential mitigation strategies in CAR T safety management.

• Samples analyzed: 3,800+
• Clinical studies pooled: 6
• Neurotoxicity threshold referenced: grade ≥3

The work directly connects Metabolon’s global discovery assays with Kite (a Gilead company) CAR T datasets that include FDA-approved products such as axi-cel and brexu-cel, demonstrating translational potential across approved regimens. For trialists and safety teams, these biochemical scores could refine monitoring windows, trigger targeted investigations, or inform prophylactic interventions before clinical deterioration.

Longer term, the findings suggest metabolomics can complement genomics and proteomics by providing a functional readout of host response, drug metabolism, and neuroimmune interaction during cell therapy. Validation in independent cohorts and prospective testing as a clinical-grade companion assay are needed to move from discovery to implementation.