Characterizing tumor-stroma interfaces in chemotherapy-treated ovarian cancer via spatial transcriptomics

The tumor-stroma interface (TSI) is a key microenvironmental region where malignant and non-malignant cells interact and influence therapeutic resistance and disease progression. Despite its importance, this area remains insufficiently characterized at the transcriptomic level, and the cellular features that connect its organization to treatment outcomes are not well defined. This gap limits the translational interpretation of spatial microenvironmental structure in the context of patient response. To address this need, we developed an integrative framework that identifies and characterizes the TSI using spatial transcriptomics. The approach combines spot-level tumor detection with a neighborhood analysis that delineates interface regions surrounding cancer spots. We applied this framework to a post-chemotherapy high-grade serous ovarian cancer dataset generated by Elena Denisenko et al. (n = 8; three good responders, two partial responders, and three poor responders) and defined tumor, TSI, and normal regions across all samples. Linear mixed-effects modeling identified significantly different proportions of EIF4A3+ cancer associated fibroblasts, endothelial cells, and myofibroblasts between outcome groups within the TSI. Among them, myofibroblasts were elevated specifically within the TSI of poor responders, but not in tumor or normal regions, indicating a localized rather than tissue-wide shift. Although T-cell proportions were not significantly different between outcome groups, bivariate Moran's I analysis showed strong spatial co-localization between T cells and tumor cells in good responders, whereas this spatial coupling was largely absent in poor responders. These findings suggest that both immune cell organization and abundance may contribute to more effective post-treatment activity in favorable outcomes. Pathway analysis further revealed distinct functional programs within the TSI. Good responders showed enrichment of immune activation and antigen-presentation pathways, whereas poor responders displayed increased extracellular matrix remodeling and stromal signaling. Regression analysis confirmed that myofibroblast proportion was tightly linked to these pro-resistance pathways and negatively correlated with humoral immune response activity, connecting stromal remodeling with reduced immune pathway engagement in poor responders. Collectively, these findings highlight the functional heterogeneity of the TSI and its potential role in modulating treatment response. By integrating tumor detection, interface mapping, and pathway-level analysis, our framework offers a scalable strategy to dissect interface biology and bridge spatial transcriptomics with treatment-related mechanisms. This framework may enable the discovery of spatially informed biomarkers for predicting therapeutic outcomes in ovarian cancer and beyond.