Perivascular immune organization and spatial architecture are associated with immunotherapy response in non-small cell lung cancer

Clinical responses to immune checkpoint inhibitors (ICIs) in non-small cell lung cancer (NSCLC) remain highly heterogeneous and PD-L1 expression alone has limited predictive performance. Emerging evidence suggests that therapeutic response depends not only on immune cell abundance but also on their spatial organization within the tumor microenvironment (TME). Here, we characterized the spatial architecture of the pre-treatment NSCLC TME and evaluated its association with response to ICIs in a real-world clinical cohort. Seventeen pre-treatment specimens (10 responders and 7 non-responders) were analyzed using 13-marker tissue-based cyclic immunofluorescence (t-CyCIF) panel with single-cell phenotyping. Cell density was assessed across all samples. Spatial analyses were performed on resection specimens with preserved architecture (4 responders and 3 non-responders). Spatial analyses included assessment of lymphoid organization, cellular neighborhoods, pairwise spatial relationships and higher-order tissue architecture. Global immune and tumor cell densities did not differ between groups. TLS-related metrics showed a trend toward higher values in responders, suggesting greater immune organization within the TME. Cellular neighborhood analysis further demonstrated enrichment of cytotoxic T cell- and vasculature-associated regions in responders, whereas non-responders displayed a predominantly tumor-centered spatial architecture. Responders also exhibited shorter vessel to cytotoxic T cell distances and greater perivascular enrichment of cytotoxic and PD-1⁺ cytotoxic T cells. Higher-order spatial analysis highlighted preferential organization of PD-1⁺ cytotoxic T cells within vascular-associated niches in responder tumors. These findings demonstrate that spatial organization of immune and vascular compartments provides information beyond cellular abundance alone and may serve as a complementary biomarker framework for predicting immunotherapy response in NSCLC.