show Abstracthide AbstractThe three-dimensional chromatin architecture plays a critical role in the establishment of cell-type-specific gene regulatory networks in eukaryotic cells. How pluripotent stem cells (PSC) alter their chromatin architecture to direct cell fate specification remains to be elucidated. Here, using a human PSC cardiomyocyte differentiation model, we analyze the dynamic reorganization of chromatin structure and gene regulatory networks during key transitional stages of cardiomyocyte development. We show that many human PSC-specific topologically associating domains (TADs) are driven by active transcription of the primate-specific retroviruses HERVH. These HERVH are silenced at the earliest stages of differentiation , accompanied by loss of TAD borders and subsequent merging of cognate TADs during differentiation, which leads to repression of gene expression within these domains . In line with these findings, deletion of select HERVHs results in elimination of corresponding TAD boundaries in human PSCs. We further discovered developmental stage-specific chromatin loop interactions that predict target genes of cardiac-related trait/disease non-coding genetic variants. Overall, our results not only highlight a novel role for endogenous retroviruses in shaping species-specific PSC chromatin architecture during evolution but also provide a genomic blueprint for understanding the impact of non-coding variants in congenital and adult heart disease/traits. Overall design: hESCs are differentiated into cardiomyocyte. We used a transgenic H9 hESC line expressing a H2B-GFP fusion protein under control of the ventricular cardiomyocyte specific MYL2 promoter. We collected samples at six critical time points during differentiation: human embryonic stem cells (hESC) (Day 0), hESC mesodermal cells (Day 2), hESC-cardiac mesodermal cells (Day 5), hESC-cardiac progenitor cell (Day 7), hESC-primitive cardiomyocytes (Day 15) and hESC-ventricular cardiomyocytes (Day 80). We performed HiC, ATAC-seq, RNA-seq, and ChIP-seq for H3K27ac, H3K27me3, H3K4me1, H3K4me3, H3K9me3 and CTCF for every time point, each with two biological replicates. We also made CRISPR edited deletion at two HERV-H loci and performed RNA-seq and Hi-C, each with two replicates.