The hierarchical folding dynamics of topologically associating domains are closely related to transcriptional abnormalities in cancers

Abstract

Graphical abstract

TADfit is a multivariate linear regression model for profiling hierarchical chromatin domains on replicate Hi-C data

“…Herein the level of a TAD boundary is determined using the terminology given by An et al 25 , that is, a TAD boundary belonging to a single TAD is regarded as a first-level boundary, and the second-level and third-level boundaries correspond to the ones that are shared by two and three hierarchical TADs, respectively. It is shown that the number of ChIP-seq peaks at the second-level boundaries is significantly greater than that at the first-level boundaries (right-tailed paired t -test, p -value <5.04e−4), based on an assumption that the enrichment differences between two different levels are normally distributed irrespective of the type of ChIP bindings and cell lines, and the same trend can also be seen at the TAD boundaries between the higher levels and second-level (right-tailed paired t -test, p -value <6.22e−5), thus, the enrichment of these ChIP bindings at TAD boundaries is significantly enhanced as the level of boundaries increases, which is in line with the observations of other studies 21 , 36 , 41 where the TAD boundaries with higher levels are believed to be more biologically meaningful. These results tell that the regression coefficients and hierarchical level of TADs called by TADfit present a reasonable biological relevance to some extent.…”

“…Generally, an enrichment difference can be seen between the two types of regions by examining all the results throughout different cell lines (GM12878 and K562), that is, the average signal of active histone marks (H3K36me3, H3K4me3, and H3K27ac) and the density of active genes in partially overlapping regions trend to be higher than that of the other regions, while the repressive histone mark (H3K27me3) does the opposite. That can be explained, since active epigenetic states and highly expressed genes are reported to be more enriched in inner TADs than in outer TADs 21 , 36 , 41 , 46 , and the partially overlapping regions are usually inner parts in a hierarchy of TADs. Furthermore, although the mechanisms underlying hierarchical TADs, especially partially overlapping ones, in gene expression regulation remain unclear 47 , 48 , the promoters and enhancers, especially active promoters and strong enhancers, are found to be favored within partially overlapping regions (Supplementary Fig.…”

MIR retrotransposons link the epigenome and the transcriptome of coding genes in acute myeloid leukemia

“…For this purpose, we performed an unbiased correlation analysis between the methylation status of individual CpGs (mCpGs) and expression levels of genes within the same TAD. Given that the major biological variation in TADs and TAD boundaries at the mega-base scale stems from differences in cell type 16 , 17 and that although malignant transformation is accompanied by extensive intra-TAD rearrangements, TAD boundaries at the mega-base scale are largely constant 18 , we performed HiC analysis of primary human CD34 + stem and progenitor cells to first define TAD boundaries in these cells (average TAD size = 1.4 Mb; Supplementary Data 1 ), followed by a correlation analysis of the mCpG status and gene expression levels within these boundaries. Two patient cohorts were included, our previously published cohort 5 (herein the Glass et al cohort) and the TCGA 1 .…”

Multilevel view on chromatin architecture alterations in cancer

“…In the altered regions, there are more TADs with smaller sizes (Taberlay et al, 2016;Nagai et al, 2019) (Figure 1C). The fact that the total majority of the boundaries presented in the normal cells are retained after malignancy (Taberlay et al, 2016;Du et al, 2021) may indicate that there is a partitioning of the existing TADs, rather than the arising of new ones. A clear confirmation of this statement is the isolation of 520 large TADs in normal prostate cells, which corresponds to 850 smaller TADs in cancer cells (Rhie et al, 2019).…”