Defining cell-specific physical basis of gene expression and signatures of functional chromatin 3D interactions: integrated analysis of single-cell 3D chromatin models, RNA/ATAC-seq, and epigenetic data.

Abstract

Three-dimensional structures of chromatin play important roles in gene regulation. Hi-C measurement provided a wealth of information on chromatin interactions. Emerging evidence suggests functional chromatin interactions are related to gene expression. However, gene promoters and their regulatory elements can be separated by 103-106 bp apart. Therefore, examining 3D chromatin conformations and identifying functionally long-range interactions are important for deciphering the structure-function relationship of 3D genome. Our recent work of CHROMATIX deep sampling algorithms can identify non-random chromatin interactions (5-7% of total Hi-C). These can be used to generate large ensembles of 3D single-cell chromatin conformations, accurately reproducing both single-cell capture and imaging studies, as well as population Hi-C measurements. In this study, we examine predicted ensembles of single-cell 3D chromatin conformations of loci among leukemia K562 cells, lymphoblastoid GM12878 cells, and lung fibroblast IMR90 cells. We identified small subsets of non-random chromatin interactions from Hi-C data based on large random ensembles of 3D chromatin chains folded in confined nuclear volume. These non-random interactions were used to generate ensembles of 5.0×104 single-cell 3D conformations for individual loci. We further quantified the heterogeneity of 3D chromatin conformations by clustering, with each spatial cluster and its proportion defined. In addition, we carried out an integrated analysis of RNA-seq, ATAC-seq, and epigenetic data of these cell lines. Results show that cell-specific functional chromatin interactions provide 3D basis for differential gene expression. There are well-defined signatures of 3D subpopulations of chromatin conformations involving distinctive patterns of epigenetic marks, and these signatures are strongly associated with gene differential expression.