Deconvolution of DNA methylation data reveals cell type-specific epigenomic signatures in endometriosis and eutopic endometrium

Endometriosis (EM) is a debilitating disease involving the growth of endometrial glands and stroma outside the uterus. To further our understanding of epigenomic dysregulation in EM and search for disease biomarkers, we performed a comprehensive evaluation of DNA methylation in eutopic endometrium (EE) and EM lesions. We undertook deconvolution analysis of DNA methylation data previously generated from endometrial aspirate biopsies obtained from 637 EM cases and 347 controls using microarray-based DNA methylation analysis. We further analyzed DNA methylation in EM lesions and paired EE samples via solution phase hybridization and DNA sequencing. For analysis of microarray data, we used a reference-free approach (BCconf) to recover latent factors and found high correlation with EM status, suggesting EE cell proportions vary between EM cases and controls. Deconvolution revealed that epithelial cells (P-value = 9.1 × 10-3) and fibroblasts (P-value = 0.022) were reduced in EM cases. Deconvolution of the sequencing data identified differences between EM lesions and EE of cases and controls, including vascular endothelial cells (false discovery rate [FDR]-adjusted P-value = 6.1 × 10-3; more abundant in EM lesions), natural killer cells (FDR-adjusted P-value = 0.031; less abundant in EM lesions) and ovarian/endometrial epithelium (FDR-adjusted P-value = 3.2 × 10-3). We detected significant differences in cell type proportion between EE samples of cases and controls. Improved reference data for deconvolution to inform more intelligent region targeting approaches will provide further insight into the molecular phenotype of EM and may inform novel approaches for minimally invasive detection.