Exploratory spatial transcriptomic analysis of compartment-specific recurrence-associated gene expression in gastric cancer

Abstract

We hypothesized that the prognostic value of immune and stromal gene expression is spatially dependent—specifically, signals driving homeostasis in the superficial mucosa may differ functionally from those at the deep invasive front. We analyzed formalin-fixed paraffin-embedded specimens from patients who underwent curative-intent gastrectomy and profiled superficial tumor layer, deep tumor layer, and adjacent normal gastric mucosa tissues using the GeoMx Digital Spatial Profiler. Segmented PanCK+ (epithelial) and PanCK– (non-epithelial) compartments were assessed using an 84-gene immune pathways panel for spatial transcriptomics to unmask these compartment-specific drivers of recurrence. Seventeen patients met the inclusion criteria; seven experienced recurrences. Spatial univariate Cox analysis identified nominal compartment-specific markers associated with recurrence. In superficial layers, higher STAT2 expression in the PanCK– compartment showed a protective effect (HR0.27, 95%CI 0.09–0.86, nominal p = 0.026). In deep tumor regions, ICAM1 expression in the tumor core correlated with recurrence (HR1.64, 95%CI 1.03–2.59, nominal p = 0.036), whereas EPCAM in the PanCK– compartment indicated lower risk (HR0.05, 95%CI 0.00-0.74, nominal p = 0.028). Several immune-related genes within the normal-appearing mucosa, including VEGFA, CD47, BATF3, and AKT1 were associated with reduced recurrence. An exploratory five-gene LASSO Cox prognostic score showed favorable discrimination (apparent C-index 0.95; optimism-corrected C-index 0.85; integrated AUC 0.990 over 0–4 years). The score stratified patients into low- and high-risk with shorter time to recurrence in high-risk group (median ≈ 14 months; log-rank p < 0.001). This exploratory spatial transcriptomic study identified biologically coherent, compartment-specific gene expression associated with recurrence. These findings provide preliminary evidence and support further validation in future cohorts.