Topologically associating domains can arise from stochastic folding of heterogeneous fluidlike chromatin

SUMMARY Topologically associating domains (TADs) are critical for gene regulation. Current views attribute TAD formation to cohesin-mediated extrusion and ignore the role of physical properties of in vivo chromatin. Here, we demonstrate that the two universal properties: chromatin fluidlike behavior and heterogeneity in DNA-packing density along chromatin, can drive TAD formation. We use DNA-accessibility data to parameterize DNA-packing density along chromatin and simulate stochastic folding of the heterogeneous chromatin in nucleus to yield a conformation ensemble. Such an ensemble can be cross-validated by Hi-C and FISH data. Furthermore, the stochastic folding model allows de novo prediction of the establishment and disappearance of key TADs during early T cell differentiation. Together, our work demonstrates that the intrinsic stochastic folding of fluidlike chromatin leads to the prevalence of TAD-like domains in single cells and their cell-to-cell variation, while the heterogeneity in DNA-packing density along chromatin mediates the emergence of TADs at ensemble-averaged level. In brief A study based on polymer simulation reveals that the two universal physical properties of in vivo chromatin fiber: chromatin fluidlike behavior and heterogeneity in DNA-packing density along chromatin play a vital role in TAD formation. Highlights Intrinsic stochastic folding of fluidlike chromatin in nuclear space underlies the prevalence of TAD-like domains in single cells and their cell-to-cell variation Heterogeneity in DNA-packing density along chromatin causes the emergence of TADs at ensemble-averaged level The disappearance and establishment of key TADs during early T cell differentiation can occur through a stochastic folding process alone, without the need of any cohesin-mediated chromatin extrusion The stochastic folding model applies to diverse cell types and is thus able to de novo predict the dynamics of genome organization over time