细胞外囊泡糖基化谱预测并影响靶向CLDN18.2的CAR-T细胞疗法对胃肠道癌的治疗结局

CLDN18.2-targeted CAR-T cell therapy has shown promising potential against gastrointestinal cancers (GIC). However, biomarkers for its patient selection and outcome prediction are urgently needed. This study aims to analyze plasma extracellular vesicle (EV) glycosylation profile and select key features to predict and monitor survival outcomes and therapeutic response of CLDN18.2-targeted CAR-T cell therapy and further investigate the underlying potential mechanisms. We depicted EV glycosylation profile via a 56-lectin array in 91 dynamic plasma samples from 53 GI cancer patients received CLDN18.2-CAR-T therapy (CT041-CG4006 trial). Based on the baseline EV glycosylation profile, we found that GalNAc/Gal related glycosylation on total EVs correlated with patients' survival outcomes. We selected the binding intensity of plasma total EVs to four key lectin features using machine learning methods which simultaneously predicted progression-free survival (PFS), overall survival (OS) and therapeutic response at baseline and all specifically recognized GalNAc/Gal. Besides the baseline analysis, dynamic changes in the binding intensity of key lectins were also related to treatment response and survival outcomes. Furthermore, high binding intensity of the key lectins to total EVs and high expression of genes mediating GalNAc/Gal glycosylation were associated with weaker anti-tumor immunity represented by fewer circulating CD8+CD28+ T cells, and correlated with poorer tumor progression indicated by elevated expression of abdominal tumor biomarkers. Mechanistically, decreasing the level of GalNAc on the surface of EVs and tumor cells could strengthen killing capability of CAR-T cells in vitro. Collectively, our work provided novel biomarkers for GIC patients undergoing CLDN18.2 CAR-T cell therapy in view of plasma EV glycosylation and revealed the potential mechanism by which the EV glycosylation affects the efficacy of CAR-T therapy.