Connecting multiway enhancer-promoter interactions to changes in gene expression in cancer

Abstract

Hi-C, and more recently Micro-C, experiments suggest that genome organization is altered as normal cells become cancerous, often resulting in aberrant gene expression. In prostate cancer, there is evidence that large topologically associating domains in normal cells split and are accompanied by shifts in the epigenetic marks from inactive to active states. However, it is unclear whether changes in genome organization alone can account for the gene expression increase in cancer cells. By combining polymer physics concepts and data-driven modeling, we first calculated an ensemble of three-dimensional (3D) chromatin structures using only the 2D contact map as input. The enhancer-promoter (E-P) distance distributions for the overexpressed Androgen Receptor (AR) and Forkhead Box Protein A1 (FOXA1) are broad, with mean values exceeding the threshold for direct E-P contact. Importantly, the average E-P distances decrease in the cancer cells, compared to normal cells in the AR locus, whereas in FOXA1 they are roughly constant or increase modestly. Similarly, the number of multiway contacts increases as cancer progresses across cancer cell lines in the AR locus. In contrast, the number of multiway contacts in FOXA1 is similar in normal and cancer cells. Because the 3D characteristics do not explain the enhanced gene expression in cancer cells, we developed Activity-by-Multiway-Contact (AMC) Model that integrates the multiway contacts with enhancer biochemical activity. The AMC model provides a plausible mechanism for the overexpression of both AR and FOXA1 in prostate cancer. Moreover, using Micro-C data for breast cancer, we show that the number of multiway enhancer-promoter contacts increases in four of five genes studied. When multiway contacts are combined with biochemical activity, changes in gene expression found in the experiment, positively correlate with the AMC score. The predictions of the AMC model not only account for overexpression of genes in prostate and breast cancer but also provide a basis for understanding gene expression variations in other genes as well.