CELL INTELLIGENCE · 肿瘤细胞治疗研究

Glioblastoma (GBM) features profound immunosuppression and metabolic reprogramming that undermine immunotherapy.

We sought biomarkers that capture cytotoxic lymphocyte activity and nominate tractable targets. Using TCGA, we quantified NK/CD8+ T-cell infiltration (CIBERSORTx), identified differentially expressed genes and derived a three-gene prognostic index (CTLsTKPI: LDHA, TNIP1, PTPN2) via Cox and LASSO. Performance was assessed in REMBRANDT, CGGA325 and CGGA693, and CTLsTKPI was incorporated into a multivariable nomogram alongside clinical variables.

We examined associations with immune checkpoints, tumour mutational burden and stemness, and performed functional studies of LDHA in GBM cells, CAR-NK92 assays and an orthotopic mouse model.

We detected 3195 DEGs versus normal brain; 119 tracked with cytotoxic infiltration. CTLsTKPI stratified survival in TCGA (AUCs 0. 694/0. 656/0. 819 at 1/2/3 years) and validated across cohorts; the nomogram outperformed individual predictors (AUC 0. 78). High CTLsTKPI aligned with checkpoint pathway activation, Treg/macrophage enrichment and reduced activated NK/CD8 + T cells.

LDHA was the dominant adverse component, linked to glycolysis; its overexpression enhanced proliferation, migration and invasion, while lactate impaired CAR-NK cytotoxicity. In vivo, LDHA inhibition (GNE-140) synergised with PD-1 blockade to reduce tumour burden and prolong survival. CTLsTKPI provides a clinically actionable biomarker, and LDHA represents a rational therapeutic target.