CaTCHing the functional and structural properties of chromosome folding

Proper development requires that genes are expressed at the right time, in the right tissue, and at the right transcriptional level. In metazoans, this involves long-range cis-regulatory elements such as enhancers, which can be located up to hundreds of kilobases away from their target promoters. How enhancers find their target genes and avoid aberrant interactions with non-target genes is currently under intense investigations. The predominant model for enhancer function involves its direct physical looping between the enhancer and target promoter. The three-dimensional organization of chromatin, which accommodates promoter- enhancer interactions, therefore might play an important role in the specificity of these interactions. In the last decade, the development of a class of techniques called chromosome conformation capture (3C) and its derivatives have revolutionized the field of chromatin folding. In particular, the genome-wide version of 3C, Hi-C, revealed that mammalian chromosomes possess a rich hierarchy of folding layers, from multi-megabase compartments corresponding to mutually exclusive associations of active and inactive chromatin to topologically associating domains (TADs), which reflect regions with preferential internal interactions. Although the mechanisms that give rise to this hierarchy are still poorly understood, there is increasing evidence to suggest that TADs represent fundamental functional units for establishing the correct pattern of enhancer-promoter interactions. This is thought to occur through two complementary mechanisms: on the one hand, TADs are thought to increase the chances that regulatory elements meet each other by confining them within the same domain; on the other hand, by segregation of physical interactions across the boundary to avoid unwanted events to occur frequently.
It is however unclear whether the properties that have been attributed to TADs are specific to TADs, or rather common features among the whole hierarchy. To address this question, I have implemented an algorithm named Caller of Topological Chromosomal Hierarchies (CaTCH). CaTCH is able to detect nested hierarchies of domains, allowing a comprehensive analysis of structural and functional properties across the folding hierarchy. By applying CaTCH to published Hi-C data in mouse embryonic stem cells (ESCs) and neural progenitor cells (NPCs), I showed that TADs emerge as a functionally privileged scale. In particular, TADs appear to be the scale where accumulation of CTCF at domain boundaries and transcriptional co-regulation during differentiation is maximal. Moreover, TADs appear to be the folding scale where the partitioning of interactions within transcriptionally active domains (and notably between active enhancers and promoters) is optimized.
3C-based methods have enabled fundamental discoveries such as the existence of TADs and CTCF-mediated chromatin loops. 3C methods detect chromatin interactions as ligation products after crosslinking the DNA. Crosslinking and ligation have been often criticized as potential sources of experimental biases, raising the question of whether TADs and CTCF- mediated chromatin loops actually exist in living cells. To address this, in collaboration with Josef Redolfi, we developed a new method termed ‘DamC’ which combines DNA methylation with physical modeling to detect chromosomal interactions in living cells, at the molecular scale, without relying on crosslinking and ligation. By applying DamC to mouse ESCs, we provide the first in vivo and crosslinking- and ligation-free validation of chromosomal structures detected by 3C-methods, namely TADs and CTCF-mediated chromatin loops.
DamC, together with 3C-based methods, thus have shown that mammalian chromosomes possess a rich hierarchy of folding layers. An important challenge in the field is to understand the mechanisms that drive the establishment these folding layers. In this sense, polymer physics represent a powerful tool to gain mechanistic insights into the hierarchical folding of mammalian chromosomes. In polymer models, the scaling of contact probability, i.e. the contact probability as a function of genomic distance, has been often used to benchmark polymer simulations and test alternative models. However, the scaling of contact probability is only one of the many properties that characterize polymer models raising the question of whether it would be enough to discriminate alternative polymer models. To address this, I have built finite-size heteropolymer models characterized by random interactions. I showed that finite-size effects, together with the heterogeneity of the interactions, are sufficient to reproduce the observed range of scaling of contact probability. This suggests that one should be careful in discriminating polymer models of chromatin folding based solely on the scaling.
In conclusion, my findings have contributed to achieve a better understanding of chromatin folding, which is essential to really understand how enhancers act on promoters. The comprehensive analyses using CaTCH have provided conceptually new insights into how the architectural functionality of TADs may be established. My work on heteropolymer models has highlighted the fact that one should be careful in using solely scaling to discriminate physical models for chromatin folding. Finally, the ability to detect TADs and chromatin loops using DamC represents a fundamental result since it provides the first orthogonal in vivo validation of chromosomal structures that had essentially relied on a single technology.