A functional overlap between actively transcribed genes and chromatin boundary elements

Abstract

<jats:title>Abstract</jats:title><jats:p>Mammalian genomes are subdivided into large (50-2000 kb) regions of chromatin referred to as Topologically Associating Domains (TADs or sub-TADs). Chromatin within an individual TAD contacts itself more frequently than with regions in surrounding TADs thereby directing enhancer-promoter interactions. In many cases, the borders of TADs are defined by convergently orientated boundary elements associated with CCCTC-binding factor (CTCF), which stabilises the cohesin complex on chromatin and prevents its translocation. This delimits chromatin loop extrusion which is thought to underlie the formation of TADs. However, not all CTCF-bound sites act as boundaries and, importantly, not all TADs are flanked by convergent CTCF sites. Here, we examined the CTCF binding sites within a ∼70 kb sub-TAD containing the duplicated mouse α-like globin genes and their five enhancers (5’-R1-R2-R3-Rm-R4-α1-α2-3’). The 5’ border of this sub-TAD is defined by a pair of CTCF sites. Surprisingly, we show that deletion of the CTCF binding sites within and downstream of the α-globin locus leaves the sub-TAD largely intact. The predominant 3’ border of the sub-TAD is defined by a steep reduction in contacts: this corresponds to the transcribed α2-globin gene rather than the CTCF sites at the 3’-end of the sub-TAD. Of interest, the almost identical α1- and α2-globin genes interact differently with the enhancers, resulting in preferential expression of the proximal α1-globin gene which behaves as a partial boundary between the enhancers and the distal α2-globin gene. Together, these observations provide direct evidence that actively transcribed genes can behave as boundary elements.</jats:p><jats:sec><jats:title>Significance Statement</jats:title><jats:p>Mammalian genomes are complex, organised 3D structures, partitioned into Topologically Associating Domains (TADs): chromatin regions that preferentially self-interact. These chromatin interactions are thought to be driven by a mechanism that continuously extrudes chromatin loops, forming structures delimited by chromatin boundary elements and reflecting the activity of enhancers and promoters. Boundary elements bind architectural proteins such as CCCTC-binding factor (CTCF). Previously, an overlap between the functional roles of enhancers and promoters has been shown. However, whether there is overlap between enhancers/promoters and boundary elements is not known. Here, we show that actively transcribed genes can also behave as boundary elements, similar to CTCF boundaries. In both cases, multi-protein complexes bound to these regions may stall the process of chromatin loop extrusion.</jats:p></jats:sec>

DynamicRunx1chromatin boundaries affect gene expression in hematopoietic development

“…The formation of sub-TADs/GADs within the gene-body of actively transcribed genes suggests that these structures, instead of being caused by a CTCF-dependent mechanism like loop extrusion, may instead be dependent on transcriptional processes (Zhang C. et al 2020). Indeed, our data indicates that, as reported for other promoters (Cho et al 2018, Harrold et al 2020, Schwessinger et al 2020), the Runx1 promoters may function as chromatin boundaries in a CTCF-independent manner, which would explain the residual sub-TADs observed in P1/P2-CTCF-KO cells. Together, our findings suggest that loop extrusion and transcription-related mechanisms may act in concert to produce dynamic chromatin structures during differentiation.…”

Cell-type specialization in the brain is encoded by specific long-range chromatin topologies

“…For example, we found extensive restructuring of topological domains, with 76% of TAD borders being different in at least one cell type. Unique borders in each cell type contained genes relevant for cell specialization, which may be functionally necessary for their interaction with cis-regulatory elements and transcriptional activation 57 .…”

preciseTAD: A transfer learning framework for 3D domain boundary prediction at base-pair resolution

“…The hierarchical nature of TAD boundaries [39,47,48] is not considered by preciseTAD due to the lack of gold standard of TAD hierarchy. preciseTAD also does not consider the directionality of CTCF binding [49] as it predicts individual boundaries in contrast to pairs of convergent CTCF motifs marking individual domains. Recent research distinguishes CTCF-associated boundaries, CTCF-negative YY1-enriched boundaries, CTCF-and YY1depleted promoter boundaries, and the fourth class of weak boundaries largely depleted of all three features [46].…”

“…Recent research distinguishes CTCF-associated boundaries, CTCF-negative YY1-enriched boundaries, CTCF-and YY1depleted promoter boundaries, and the fourth class of weak boundaries largely depleted of all three features [46]. Furthermore, actively transcribed regions can serve as TAD boundaries themselves, independently of CTCF binding [49]. This may lead to some TAD boundaries being undetected by preciseTAD despite being detected by domain callers.…”