Fig. 1

T2C description, interaction mapping, and direct determination of the chromatin quasi-fibre and the aggregated loop/rosette 3D architecture of the human and mouse genomes: a Cell nuclei in a population of cells (transmission light and fluorescence microscopy, [89]) have an underlying chromatin architecture (simulated cell nucleus containing 1.2 million polymer segments; resolution 5.2 kbp, i.e. ~50 nucleosomes; Multi-Loop-Subcompartment (MLS) rosette model with 126 kbp loops and linkers; [5]). After crosslinking the DNA is restricted within the nucleus by a 1st restriction enzyme, before the crosslinked fragments are extracted and diluted such that intra-fragment re-ligation is favoured. After de-crosslinking, the re-ligated material is shortened by a 2nd restriction enzyme or sonication and purified by a capture array with oligos designed next to the 1st restriction enzyme, before paired-end-sequencing over the ligation position. After alignment to the reference genome, this results in interactions frequency matrices (b–d) and scaling curves (Fig. 2). b, c Interaction matrices (logarithmic and colour-coded scale; left and right) of the human IGF/H19 11p 15.5-15.4 region (b) in HB2, HEK293T TEV (intact cohesin) and HEK293T HRV (cleaved cohesin) as well as the mouse β-globin 7qE3-F1 region (c) for fetal brain (inactive β-globin) and liver cells (active β-globin) show the formation of subchromosomal domains separated by a linker (borders: pink lines, right; D1s, D1e: start and end of domains), which consist of loops (red lines; 8L: number of loops), representing due to the grid-like pattern loop aggregates/rosettes. A grid-like pattern is also visible in the interactions between the domains and corresponds to the interactions of loops and loop bases of interacting domains. Near the diagonal the aggregation into a chromatin quasi-fibre and loop internal structures are visible (zooming in and out the images can make this clearer). Between different cell types and functional states only some local differences are visible resulting in a consensus architecture and allowing simulation of the 3D architecture (middle; resolution <~1 kbp). Note that the simulation is driven by the dominant consensus architecture. d The interaction matrix of a 380 kbp subchromosomal domain in the mouse 12qF1–F2 region at high resolution clearly shows the regular rosette-like picture with a detailed structure of the loop base with in- and outgoing loop fibre stretches as seen in simulations (e, f). e Simulated Multi-Loop-Subcompartment (MLS) model with an averaged spatial distance map for exact spatial distances 〈R _S 〉 (left) and on the diagonal normalized interaction maps for interaction radii 〈d _i 〉 of 50 nm, 70 nm, and 150 nm (right), for an MLS model with 126 kbp loops and linkers [16 Mbp upper and 1.2 Mbp zoom-in (z) lower row), showing clearly the formation of domains connected by a linker, their interaction, and the underlying loop aggregates/rosette architecture, with (anti-)parallel fibre stretches at the loop base. The dependence on the interaction radii corresponding to different crosslink probabilities is also clearly visible. f Sketch of the different structures visible on different scales in the experimental and simulated interaction matrices (spatial distance matrix: left; simulated interaction matrix: upper) (from e): On the smallest scale, near the diagonal the compaction of nucleosomes into the quasi-fibre (yellow line) and the fibre regime (dark blue line) can be found. On the largest scale the domains are clearly bordered (pink lines) and connected by a linker. On medium scales the loop aggregate/rosette like structure is characterized by the loop bases (red circles: within domains, blue circles: between domains) as well as the loop interactions (green triangles). The fine structure of the loops representing the (anti-)parallel loop stretches at the base (red crosses) and within loops (green stretches near diagonal)