The order and variability of bacterial chromosome organization, contained within the distribution of chromosome conformations, are unclear. Here, we develop a fully data-driven maximum entropy approach to extract single-cell 3D chromosome conformations from Hi–C experiments on the model organism Caulobacter crescentus. The predictive power of our model is validated by independent experiments. We find that on large genomic scales, organizational features are predominantly present along the long cell axis: chromosomal loci exhibit striking long-ranged two-point axial correlations, indicating emergent order. This organization is associated with large genomic clusters we term Super Domains (SuDs), whose existence we support with super-resolution microscopy. On smaller genomic scales, our model reveals chromosome extensions that correlate with transcriptional and loop extrusion activity. Finally, we quantify the information contained in chromosome organization that may guide cellular processes. Our approach can be extended to other species, providing a general strategy to resolve variability in single-cell chromosomal organization.

We investigate simple one-dimensional driven diffusive systems with open boundaries. We are interested in the average on-site residence time defined as the time a particle spends on a given site before moving on to the next site. Using mean-field theory, we obtain an analytical expression for the on-site residence times. By comparing the analytic predictions with numerics, we demonstrate that the mean-field significantly underestimates the residence time due to the neglect of time correlations in the local density of particles. The temporal correlations are particularly long-lived near the average shock position, where the density changes abruptly from low to high. By using Domain wall theory (DWT), we obtain highly accurate estimates of the residence time for different boundary conditions. We apply our analytical approach to residence times in a totally asymmetric exclusion process (TASEP), TASEP coupled to Langmuir kinetics (TASEP + LK), and TASEP coupled to mutually interactive LK (TASEP + MILK). The high accuracy of our predictions is verified by comparing these with detailed Monte Carlo simulations.

Whereas pairwise Hi-C methods have taught us much about chromosome organization, new multi-contact methods, such as single-cell Hi-C, hold promise for identifying higher-order loop structures. The presence of such high-order structure may be revealed by comparing multi-contact data with a theoretical prediction based on pairwise contact information. Here, we develop and compare three polymer-based prediction schemes for chromosomal three-point contact frequencies, based on a non-interacting polymer, a polymer with independent cross-linking, and a polymer with weak pairwise interactions between monomers. First, we test these predictions for two distinct simulation models of bacterial chromosome organization: a data-driven model inferred from a Hi-C map and bottom-up simulations of loop-extruding proteins. We find that the most predictive approximation is indicative of how contacts are primarily formed in a model. We then apply our prediction schemes to previously published super-resolution chromatin tracing data for human IMR90 cells. Strikingly, we find that the best prediction is given by the independent cross-linking approximation. This result is consistent with chromosomal contacts being dominantly caused by weakly interacting loop-extruders. Our work could have implications for developing models of chromosome organization from multi-contact data, and for better identifying higher-order loop structures.

No abstract