Transcriptomic development of a tumor-associated macrophage ferroptosis-glycolysis prognostic signature in diffuse large B-cell lymphoma
TL;DR
GCLC and SLC1A5 emerged as TAM-related candidate genes that warrant further mechanistic and prospective validation and computationally predicted drug sensitivity and associations between TAMFGS and immune-related transcriptomic estimates and computationally predicted drug sensitivity are identified.
Abstract
Diffuse large B-cell lymphoma (DLBCL) is biologically heterogeneous and is associated with variable clinical outcomes. We aimed to develop a tumor-associated macrophage (TAM)-related ferroptosis–glycolysis prognostic signature and to explore selected signature genes in preclinical models. This retrospective multi-cohort computational prognostic biomarker-development study integrated public single-cell and bulk transcriptomic datasets. GSE10846 was used for feature selection, model fitting, and parameter tuning, whereas GSE32918, GSE69051, and TCGA-DLBC were used as model-selection validation cohorts. Exploratory immune, genomic, and computational drug-sensitivity analyses were performed. Signature-gene expression was assessed in 27 archived DLBCL tissues, and GCLC and SLC1A5 were further examined in TAM-related preclinical models in vitro and in vivo. An 11-gene TAM-related ferroptosis–glycolysis signature (TAMFGS) was developed. The time-dependent AUCs (95% CIs) were 0.978 (0.966–0.990), 0.986 (0.976–0.995), and 0.982 (0.966–0.999) at 1, 3, and 5 years, respectively, in GSE10846; 0.622 (0.524–0.720), 0.608 (0.515–0.702), and 0.619 (0.524–0.715) in GSE32918; 0.516 (0.286–0.747), 0.741 (0.548–0.934), and 0.754 (0.516–0.993) at 1, 2, and 3 years, respectively, in GSE69051; and 0.929 (0.850–1.000), 0.674 (0.382–0.966), and 0.677 (0.402–0.952) at 1, 3, and 5 years, respectively, in TCGA-DLBC. In GSE10846, the continuous TAMFGS score remained associated with overall survival after adjustment for available covariates (HR, 1.094; 95% CI, 1.082–1.106; p < 0.001). Exploratory analyses identified associations between TAMFGS and immune-related transcriptomic estimates and computationally predicted drug sensitivity. GCLC or SLC1A5 knockdown was associated with ferroptosis-associated molecular changes and an M1-like inflammatory shift in the THP-1-derived TAM model and was associated with reduced DLBCL growth in preclinical models. In retrospective transcriptomic DLBCL cohorts, TAMFGS was associated with overall survival. GCLC and SLC1A5 emerged as TAM-related candidate genes that warrant further mechanistic and prospective validation.