e17104 Background: Gynecologic carcinosarcomas (GCS) are rare tumors with poor prognosis. Reasons include a high percentage of advanced stage at diagnosis and a low response to conventional treatments. GCS constitutes a model for research in both tumoral heterogeneity and the epithelial-mesenchymal transition (EMT) process. Our aim is to define molecular expression heterogeneity in GCS distinct morphologic components. Methods: A retrospective, single institution, IRB approved study of 13 patients diagnosed with GCS was undertaken. Total RNA was extracted from representative FFPE tissue blocks of both the epithelial and mesenchymal components. The expression profile for each component (n = 26) was determined using the GeneChip WT Pico Reagent Kit and the Clariom D Array (Affymetrix Inc., Santa Clara, CA, USA). Robust multi-array normalization (RMA) and t-statistics was used for detecting differentially expressed genes between the studied conditions. Genes with a p-value < 0.05 and with an absolute fold change (FC) value > 1.5 were selected as significant. Results: A total of 13 cases representing 26 distinct samples, 9 uterine (UCS) and 4 ovarian carcinosarcoma with a median age of 68 (range: 45-81), 38% presented FIGO IIIC-IV stage at diagnosis. Among UCS, 5 women had a previous personal history of breast cancer. A total of 101 genes appeared as differentially expressed between epithelial and mesenchymal components, highlighting 5 of them: HMGA2 (FC = 2.15, p = 0.04) and ERBB4 (FC = 2.14, p = 0.005) overexpressed in epithelial component and ANX2 (FC = 1.95, p = 0.0006), SPP1 (FC = 2.15, p = 0,005) and ERRFI1 (FC = 1.95, p = 0.001) overexpressed in mesenchymal component. Conclusions: This is the first expression profiling in GCS that helps identify candidate genes that show a distinct expression in mesenchymal and epithelial components that could have a potential prognostic and predictive role.