Meta_B cells: a computationally identified candidate immunosuppressive driver of gastric cancer metastasis revealed by single-cell analysis and machine learning.

Background: Gastric cancer (GC) metastasis remains a major clinical challenge due to insufficient understanding of tumor microenvironment (TME) dynamics. While B cells are implicated in GC progression, their subset-specific roles in metastatic niches are poorly defined.

Methods: We analyzed gastric cancer (GC) single-cell RNA-seq data from the GEO database (GSE163558), complemented by bulk RNA-seq analysis of TCGA-STAD cohorts. Meta_B cells were identified through Seurat clustering and validated in colorectal cancer metastases (GSE166555). And we constructed a prognostic model via hdWGCNA and LASSO-Cox regression. Functional analyses included GSEA, pseudotime trajectory (Monocle2) and cell-cell communication (CellChat).

Results: We identified meta_B cells, a metastasis-enriched B cell subset, characterized by CLEC2B/YBX3 overexpression. Functional analyses suggested a potential immunosuppressive role associated with computational inference of BTLA-TNFRSF14 pathway activation, correlating with interactions with macrophages and other immune cells. A machine learning-derived 10-gene prognostic model effectively stratified high-risk patients with stromal-rich tumor microenvironments and predicted potential enhanced chemosensitivity to axitinib, dasatinib, olaparib, rapamycin, and ribociclib.

Conclusions: Meta_B cells may represent a novel B cell subset computationally associated with immunosuppression and GC metastasis potentially mediated by the BTLA axis. Our integrative transcriptomic framework provides hypothesis-generating insights into metastatic TME remodeling and a clinically actionable tool for prognostic prediction. Targeting meta_B cells can be explored as a strategy to potentially overcome immunotherapy resistance.