Recent evidence that TADs and chromatin loops are dynamic structures

Abstract

Mammalian genomes are folded into spatial domains, which regulate gene expression by modulating enhancer-promoter contacts. Here, we review recent studies on the structure and function of Topologically Associating Domains (TADs) and chromatin loops. We discuss how loop extrusion models can explain TAD formation and evidence that TADs are formed by the ring-shaped protein complex, cohesin, and that TAD boundaries are established by the DNA-binding protein, CTCF. We discuss our recent genomic, biochemical and single-molecule imaging studies on CTCF and cohesin, which suggest that TADs and chromatin loops are dynamic structures. We highlight complementary polymer simulation studies and Hi-C studies employing acute depletion of CTCF and cohesin, which also support such a dynamic model. We discuss the limitations of each approach and conclude that in aggregate the available evidence argues against stable loops and supports a model where TADs are dynamic structures that continually form and break throughout the cell cycle.

Keywords: 3D genome; CTCF; FRAP; chromatin loops; cohesin; dynamics; loop extrusion; modeling; single-molecule imaging; topological domains.

Figure 1.

Chromosome structure and function is organized at multiple scales. At the smallest scale, DNA is folded into a double helix, which gets compacted into ∼11 nm nucleosomes, whereby 147 bp of DNA wrap around a histone octamer. Functionally, nucleosomes regulate access of DNA-binding proteins and serve as modules for epigenetic modifications, which regulate gene expression. At the intermediate scale of tens of kilobases to a few megabases, chromatin is organized into Topologically Associating Domains (TADs) with a median size of a few hundred kilobases. Functionally, TADs are characterized by preferential contact of loci within them, and critically control enhancer-promoter interactions, and relative insulation from adjacent TADs. At a similar scale of TADs, chromatin is also organized into epigenomic A/B “compartments”, whereby active chromatin (A) tends to contact with other segments of active chromatin and localize in proximity of certain nuclear bodies such as nuclear speckles, while inactive chromatin (B) tends to contact with inactive chromatin and to be associated with the nuclear lamina. At the largest scale, particular chromosomes tend to associate with other chromosomes and form stereotyped chromosome territories inside the cell nucleus. The image used to illustrate chromosome territories was generously provided by Stevens et al.

Figure 2.

TADs, chromatin loops and the role of CTCF and cohesin. (A) A simulated and simplified Hi-C map. The color scale corresponds to contact frequency (darker red, more frequent contacts). TADs appear as triangles, within which there are more frequent chromatin contacts and are often marked by “cornerpeaks”, suggesting that they are held together by a chromatin loop. Below, simulated CTCF and cohesin ChIP-Seq tracks illustrating that TAD and loop boundaries are almost always bound by CTCF and cohesin. At the bottom, DNA with the location and orientation of CTCF binding sites listed (red arrows denote CTCF binding site orientation). Note that Hi-C and ChIP-Seq data in this sketch is simulated and simplified. This sketch was inspired Fig 2a in Merkenschlager and Nora. (B) Sketch of CCCTC-binding factor, CTCF, an 11-zinc finger DNA binding protein with its consensus DNA binding sequence shown. (C) Sketch of the cohesin complex composed of the proteins Smc1, Smc3 and Rad21, which closes the ring, and the SA1/2 subunit which is involved in protein interactions. Cohesin topologically entraps chromatin within its lumen. Please note that whether cohesin functions as a single ring or a pair of rings remains a subject of debate. (D) The presence of CTCF and cohesin bound “corner peaks” in Hi-C maps are generally assumed to correspond to a chromatin loop held together by CTCF and cohesin as sketched. We refer to this protein complex holding together a loop as a Loop Maintenance Complex (LMC). (E) “Loop rosette” model, where TADs are held together by loops and TADs without cornerpeaks form passively from adjacent loop domains. This sketch was inspired by Fig 6F in Rao et al.. (F) TADs may emerge at the population level when averaged over many heterogeneous single-cell genome topologies. Cohesin is sketched as rings and CTCF as in (B). This picture assumes that TADs are formed by cohesin-mediated extrusion, which stops at occupied CTCF binding sites. This sketch was inspired by Fig 7A in Fudenberg et al. Panels (B-D) have been adapted and reproduced from Hansen et al.

Figure 3.

Dynamic LMC model. Cohesin functions to hold together two strands of chromatin and CTCF positions cohesin at its convergent cognate binding sites as previously proposed. Note that whether cohesin entraps DNA as a single ring or a pair of rings remains a subject of debate. In the dynamic LMC model, while cohesin holds together a chromatin loop, different CTCF molecules are frequently binding and unbinding, giving rise to a dynamic protein complex with a molecular stoichiometry that changes over time. Even though cohesin's average residence time (∼20–25 min) is much longer than CTCF's (∼1–4 min), cohesin does eventually dissociate. Since there is now nothing to hold together the chromatin loop, cohesin dissociation causes the loop to fall apart. Thus, we propose that TADs and the chromatin loops that hold them together are dynamic structures, and that CTCF and cohesin form a dynamic protein complex. This figure has been adapted and reproduced from Hansen et al.