Contrasting roles of PD-L1 expression in tumor and stroma: Biological and prognostic implications in esophageal squamous cell carcinoma.

Programmed death-ligand 1 (PD-L1) plays a central role in immune regulation in esophageal squamous cell carcinoma (ESCC) and has been widely used as a biomarker for immune checkpoint inhibitor therapy. However, the biological and clinical significance of PD-L1 expression remains controversial, partly due to its marked spatial heterogeneity and dynamic regulation within the tumor immune microenvironment (TIME). In particular, conventional assessment methods do not sufficiently distinguish between PD-L1 expression in tumor cells and that in the surrounding stromal compartment. Accumulating evidence indicates that tumoral and stromal PD-L1 expression reflect distinct biological processes and may have contrasting prognostic and therapeutic implications. Tumoral PD-L1 expression often represents adaptive immune resistance induced by cytotoxic T-cell-derived cytokines and may serve as a surrogate marker of an immune-active tumor. In contrast, stromal PD-L1 expression integrates signals from immune cells, cancer-associated fibroblasts, and other non-neoplastic components, frequently reflecting an immune-enriched microenvironment associated with restrained tumor progression and sustained chemotherapy efficacy. Recent advances in digital pathology and artificial intelligence-assisted image analysis have enabled quantitative, compartment-specific evaluation of PD-L1 expression across whole-slide images, providing new insights into its spatial distribution and clinical relevance. These approaches facilitate refined interpretation beyond binary scoring systems such as the combined positive score or tumor proportion score. In this review, we summarize current knowledge on the regulation and pathological assessment of PD-L1 in ESCC, with particular emphasis on its compartment-specific roles in tumor cells and the stroma. By integrating biological mechanisms, methodological advances, and clinical implications, we propose a conceptual framework for spatially resolved PD-L1 evaluation that may improve prognostic stratification and support more personalized treatment strategies in ESCC.