Spatial transcriptomics identifies tumor-intrinsic and stromal programs associated with immune exclusion in NSCLC

Immune checkpoint blockade (ICB) has reshaped the treatment of advanced non-small cell lung cancer (NSCLC), yet durable benefit remains limited to a subset of patients. One major pattern of resistance is immune exclusion, in which CD8+ T cells accumulate in stromal regions but fail to penetrate tumor nests. Here, using discovery and non-overlapping internal validation cohorts of ICB-treated NSCLC patients, we show that spatial CD8+ T-cell exclusion is associated with lower objective response rates and shorter progression-free survival. To characterize the molecular programs linked to this phenotype, we integrated GeoMx spatial transcriptomics, bulk transcriptomic profiling, TCR-probe expression analysis, multiplex immunohistofluorescence, and single-cell spatial neighborhood analyses. Compared with infiltrated tumors, immune-excluded tumors showed reduced expression of genes encoding antigen-presentation machinery in the tumor compartment, including HLA class I genes and TAP1, with an altered relationship between IFN-γ-associated transcriptional signals and expression of antigen-presentation machinery. At the stromal level, excluded tumors were enriched for myofibroblastic cancer-associated fibroblast programs and peri-epithelial collagen-rich architecture, whereas infiltrated tumors displayed immune-supportive stromal signatures. CD8-enriched regions from excluded tumors showed reduced metabolic activation and lower TCR-probe expression diversity, without a marked increase in canonical exhaustion markers. These findings provide an integrated, compartment-resolved description of tumor-intrinsic and stromal features associated with immune exclusion in NSCLC and support further evaluation of spatial immune contexture as a complementary biomarker in ICB-treated patients.