We performed Capture Hi-C experiments on mouse thymocyte populations and found that most TAD structures are robust to developmental change. However, some domains were remodelled and we showed using CRISPR-mediated gene induction a direct causal role of transcription in these remodelling events (Fig 1b; Chahar et al., in prep). We also interrogated the dynamics of chromatin loops with all gene promoters in the same thymocyte populations. In order to call interactions robustly, we developed a pipeline for analysing such Capture Hi-C datasets (Ben Zouari et al., 2019; Genome Biol). We identified thousands of both stable and highly dynamic promoter-centred interactions, many as expected carrying the hallmarks of enhancers, but many more with less obvious functional relevance, even though they are reproducible and highly cell type-specific. We also investigated chromatin topology at earlier thymocyte and haematopoietic stem cell stages, requiring us to modify the technology for use in limiting cell numbers. As proof of principle, we identified novel candidate enhancers specific to haematopoietic stem cells, which we validated with CRISPR deletions (Fig 1c; Karasu, Molitor et al., in prep).
To track chromatin topology in real-time in single cells, we labelled site-specific regions in mouse embryonic stem cells, in combination with tags for nascent transcription (Fig 1d). Surprisingly, we found that local chromatin dynamics are not uniform across the chromosome but are specific to genomic context (Oliveira et al., 2021; Nat Comm). We then labelled the promoter and distal enhancer of the pluripotency gene Sox2 in mouse embryonic stem cells to follow their spatial communication relative to transcriptional status. Although the two elements were frequently proximal, their exact distance did not correlate with transcription, and was unaltered on treatment with drugs inhibiting transcription. Instead, the local chromatin diffusion dynamics were more closely linked to transcription. Both promoters and enhancers are significantly more constrained than control sequences, perhaps by their confinement within nuclear foci dedicated to transcription, and this confinement was relaxed on perturbation of transcription. Moreover, diffusion of the Sox2 promoter is faster in experimental conditions permitting gene transcription (Platania et al., in prep).
We dissected the Sox2 enhancer with CRISPR-mediated allele-specific deletions and found a remarkable decoupling between transcription and chromatin topology. Whereas transcription was eliminated by deletion of a few transcription factor binding sites, these deletions had no effect on chromatin architecture. Instead, perturbation of the interaction with the gene, and of maintenance of the TAD border, required extensive deletions, suggesting that chromatin topology maintenance is modular and distributed over numerous genomic sites. As a complementary approach, we also perturbed promoter-enhancer interactions within the same locus by engineering new sequences of choice into an intervening region. By engineering CTCF-mediated chromatin loops, we were able to perturb the endogenous promoter-enhancer interaction, but these had no effect on Sox2 transcription (Fig 1e; Taylor, Sikorska, Shchuka et al., 2021; submitted to Genes Dev).
