Abstract PS2-10-18: Epithelial-mesenchymal transition (EMT) Signature Score of a Bulk Tumor May Not Reflect the EMT of Malignant Epithelial Cells

Abstract Background: Epithelial-mesenchymal transition (EMT) is a process in which polarized epithelial cells acquire mesenchymal features, contributing to tumor progression, metastasis, and treatment resistance. Our group has been showing the clinical relevance of pathway activity scores generated by Gene Set Variation Analysis (GSVA) of Hallmark gene sets in breast cancer. Here, we hypothesized that breast cancer with high EMT score is associated with aggressive features and poorer outcomes compared to EMT-low tumors. Material and Methods: We analyzed three independent large breast cancer cohorts: TCGA (n = 1,077), METABRIC (n = 1,904), and SCAN-B (n = 3,069). Survival outcomes (OS, DSS, DFS) were assessed. Tumor immune microenvironment (TIME) was analyzed using xCell and Thorsson et al.’s immune metrics. Chemosensitivity was examined using four independent datasets (GSE21094, GSE22358, GSE25066, GSE32646). Mutation burden and neoantigen load were also evaluated. To explore EMT signatures across cell types, single-sample GSEA was applied to single-cell RNA-seq data. Results: EMT score was generated by Gene Set Variation Analysis of Hallmark EMT gene set and each cohort was divided into high and low score groups by the median as the cutoff. Unexpectedly, high EMT tumors were associated with significantly lower Nottingham histological grade, and 4 cell proliferation-related Hallmark gene sets; E2F Targets, G2M Checkpoint, MYC Targets V1, and MYC Targets V2, were enriched to low EMT tumors consistently in all three cohorts, TCGA, METABRIC, and SCAN-B (all FDR 0.01). Notably, EMT-high tumors did not exhibit features of a tumor-permissive immune microenvironment, as assessed by xCell. Surprisingly, high EMT scores were not associated with lymph node (N) or distant (M) metastasis in any of the three cohorts. Indeed, high EMT tumors were not associated with any of the survival parameters in TCGA and METABRIC, nor SCAN-B. To investigate this unexpected lack of prognostic significance of EMT score, we assessed the relationship of EMT score and chemosensitivity in four independent cohorts but observed no significant associations. Similarly, no association was found between EMT score and mutation load nor neoantigen burden. xCell analysis revealed that EMT-high tumors exhibited significantly higher infiltration of stromal cells, including adipocytes, endothelial cells, pericytes, and fibroblasts (all p 0.001), in all three cohorts. With ssGSEA of single-cell RNA sequencing cohort, cancer-associated fibroblasts (CAFs) was found to express the highest EMT scores compared to the other cell populations (e.g., cancer cells, T cells, B cells, and myeloid cells). These findings suggest that EMT signatures derived from bulk tumor transcriptomes likely reflect the stromal compartment, particularly CAFs, rather than intrinsic properties of malignant epithelial cells. Conclusion: Our findings suggest that EMT signal from transcriptome of a bulk tumor largely represent stromal components, particularly CAFs, rather than intrinsic EMT activity in malignant cells. This highlights the need for analyses of single-cell resolution to accurately interpret EMT biology in the tumor microenvironment. Citation Format: Y. Tokumaru, R. Wu, M. Oshi, S. Ando, M. Takeuchi, T. Ishikawa, M. Nobuhisa, M. Futamura, K. Takabe. Epithelial-mesenchymal transition (EMT) Signature Score of a Bulk Tumor May Not Reflect the EMT of Malignant Epithelial Cells abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS2-10-18.