基于微生物组的淋巴瘤 CAR-T 治疗反应建模：鸟枪法宏基因组测序的启示

The interplay between the commensal microbiota and the mammalian immune system may influence the outcomes of T cell-driven cancer immunotherapies.

However, clinical studies supporting microbiota-based interventions in chimeric antigen receptor T-cell (CAR-T) therapy remain scarce.

This study included 30 adult patients with B-cell lymphoma treated with axicabtagene ciloleucel (axi-cel) or 4-1BB investigational product. Shotgun metagenomics sequencing (SMS) of fecal samples, collected before lymphodepletion and 1 month post infusion, enabled species-level resolution.

We also trained 25 microbiome-based machine-learning (ML) models for response prediction. Neither prior "high-risk" antibiotics exposure nor alpha diversity influenced toxicity, response, or survival.

However, dysbiosis was observed between 11 healthy controls and patients, particularly in those treated with axi-cel. SMS identified species associated with clinical outcomes. Increased abundance of Alistipes senegalensis and Alistipes onderdonkii correlated with lower neurotoxicity and improved survival, respectively. Bifidobacterium longum was associated with reduced cytokine release syndrome, whereas Bifidobacterium adolescentis , Bifidobacterium bifidum , and Bifidobacterium breve correlated with poorer survival.

ML models demonstrated strong predictive performance, with some identifying non-responders using only six species selected by the Boruta method ( Bacteroides xylanisolvens , Bifidobacterium bifidum , Bifidobacterium breve , Eubacteriaceae bacterium Marseille-Q4139, Negativibacillus massiliensis, and Sellimonas intestinalis).

These findings deepen current knowledge and support prospective microbiota-based strategies in CAR-T therapy.