Tumor-Infiltrating Lymphoid Structures Predict Gastric Cancer Immunotherapy Response

Gastric cancer (GC) is the third leading cause of cancer death worldwide, and while immune checkpoint inhibitors have improved outcomes, roughly 25–40% of advanced patients fail first-line immunotherapy. Better biomarkers are urgently needed to identify who will respond. Tertiary lymphoid structures—organized immune aggregates resembling lymph nodes that form within and around tumors—have emerged as promising prognostic indicators, but their spatial location relative to the tumor appears to matter as much as their presence alone.

The researchers enrolled 110 GC patients for pathology and survival analysis, then profiled 25 specimens with single-cell RNA sequencing (88,100 high-quality cells) and 5 specimens with spatial transcriptomics. Tumors were stratified into three groups: intratumoral TLS (iTLS), peritumoral TLS (pTLS), and TLS-desert (dTLS). Kaplan–Meier analysis confirmed that iTLS-GC patients had significantly superior overall and progression-free survival compared with pTLS and dTLS groups, and a higher rate of immunotherapy response.

Single-cell and spatial analyses revealed that iTLS-rich tumors are selectively enriched for four immune cell populations: CXCL13+ T lymphocytes, CXCR5+ germinal center B cells, LAMP3+CD80+ activated dendritic cells, and SELP+ACKR1+ high endothelial venule (HEV) endothelial cells. Cell–cell interaction modeling and functional experiments delineated a mechanistic axis in which HEV cells expressing VCAM1 and ICAM1 recruit and activate CXCL13+ T cells via the CXCL13-ACKR1 pathway; these T cells then promote TLS maturation through CXCL13-CXCR5 crosstalk with B cells. Spatial transcriptomics confirmed this cellular architecture in situ, with TLS-associated gene modules (CCL19, CXCL13, CXCR4, MS4A1) concentrated in intratumoral spatial clusters largely absent from dTLS samples.

Building on these findings, the team constructed a single-cell/spatial TLS (ssTLS) signature capturing the iTLS cellular ecosystem. This multi-gene signature demonstrated predictive value for immunotherapy outcomes in independent GC cohorts, outperforming existing biomarkers such as PD-L1 CPS alone. The results suggest that spatially resolved TLS profiling, potentially operationalized through the ssTLS gene signature, could guide patient selection for immunotherapy and identify candidates for TLS-inducing combination strategies.

Caveats include the modest sample size for spatial transcriptomics (n=5), the retrospective nature of survival analyses, and the fact that all scRNA-seq samples were treatment-naive, which may not fully represent the dynamics of TLS under therapy. Prospective validation in larger, multi-ethnic cohorts receiving standardized immunotherapy regimens is needed before clinical implementation.