Identification of genomic features that uniquely impact estrogen receptor alpha binding and its effects on gene expression in endometrial cancer

Estrogen receptor alpha (ER) is an established oncogenic transcription factor in breast and endometrial cancer; however, more is known about the mechanisms controlling ER behavior in breast cancer, and therapies targeting ER have been much more successful in breast cancer. To address this disparity, we sought to identify genomic features that control ER in endometrial cancer and determine to what extent these factors differ from those in breast cancer. We focused on the locations of estrogen response elements (EREs), ERs preferred DNA binding motif, throughout the human genome. To identify factors that predict ER genomic binding and effects on target gene expression, we applied machine learning to genomic data for each ERE in Ishikawa cells (ER+ endometrial cancer) and T-47D cells (ER+ breast cancer). Many of these factors, such as chromatin accessibility and histone modifications, were predictive of ER activity in both cell lines. However, the transcription factors that predict ER activity were found to be cell type-specific, including FOXA1 and GATA3 in T-47D cells, and ETV4 and SOX17 in Ishikawa cells. In addition, the features that predict ER binding and effects on gene expression differed, with transcription at EREs in the absence of estrogen being predictive of regulatory activity. To verify our findings in the Ishikawa cells, we performed a CRISPR knockout screen, which confirmed the discovery that SOX17 controls ER activity in endometrial cancer cells. These results identify important genomic features of ER binding and regulatory activity and how these features differ between endometrial cancer and breast cancer cells.