ABSTRACT: DNA-level alterations (mutations) conferring selective cell growth advantages are well-established in cancer initiation and progression. Changes in cell states not due to alterations in DNA sequence (epimutations) are also proposed to contribute to cancer, but concrete examples involving specific genes are elusive. Endometrial cancer is a common, lethal malignancy of women, understood to arise through the stepwise selection of heritable changes that result first in the formation of a preinvasive precursor known as endometrioid intraepithelial neoplasia, followed by progression to cancer. PAX2 is required for the development of endometrium in embryos, and PAX2 endometrial expression persists throughout life. Per clinical studies, loss of PAX2 protein occurs in 80% of endometrial cancers. However, the molecular basis of PAX2 loss—and its functional significance—remain unknown. Here, we show that loss of PAX2 protein occurs through a specific transcriptional gene silencing event specifically targeting PAX2. This RNA-level silencing event occurs in early microscopic neoplastic clones preceding endometrioid intraepithelial neoplasia, linking PAX2 silencing with the initiation of most endometrial cancers. Integrated genomic, transcriptomic, epigenomic, 3D genomic, and machine learning approaches showed that loss of PAX2 expression is associated with replacement of open/active chromatin features (H3K27ac and H3K4me3) with inaccessible chromatin features (H3K27me3). The spread of the H3K27me3 signal resembles a pearl necklace, with its length modulated by cohesin loops, preventing transcriptional dysregulation of neighboring genes, thereby limiting PAX2 silencing as a focal epigenetic event. PAX2 broadly regulates enhancer activity and transcription in endometrial cancer cells including PGR, rationalizing its activity as an oncodevelopmental tumor suppressor. Functional studies, including of a novel mouse model and tumor-derived organoids based on endometrial-specific Pax2 inactivation, proved that Pax2 is a bona fide tumor suppressor in vivo that potently cooperates with Pten, the second most frequently inactivated tumor suppressor in endometrial cancer. PAX2 silencing was reversible in human cell lines, suggesting that PAX2 may be a druggable target. Taken together, our findings establish a new paradigm for cancer-driving epimutations and open new lines of investigation into the origins of endometrial cancer, with diverse implications for its diagnosis and treatment.
