Structural basis for preservation of a subset of Topologically Associating Domains in Interphase Chromosomes upon cohesin depletion

Reviewer #1 (Public Review):

Summary:

In this paper, Jeong et al. investigate the prevalence and cause of TADs that are preserved in eukaryotic cells after cohesin depletion. The authors perform an extensive analysis of previously published Hi-C data, and find that roughly 15% of TADs are preserved in both mouse liver cells and in HCT-116 cells. They confirm previous findings that epigenetic mismatches across the boundaries of TADs can cause TAD preservation. However, the authors also find that not all preserved TADs can be explained this way. Jeong et al. provide an argument based on polymer simulations that "physical boundaries" in 3D structures provide an additional mechanism that can lead to TAD preservation. However, in its current form, we do not find the argumentation and evidence that leads to this claim to be fully compelling.

Strengths:

We appreciate the extensive statistical analysis performed by the authors on the extent to which TAD's are preserved; this does seem like a novel and valuable contribution to the field.

Weaknesses:

1. As the authors briefly note, the fact that compartmentalization due to epigenetic mismatches can cause TAD-like structures upon cohesin depletion has already been discussed in the literature; see for example Extended Data Figure 8 in (Schwarzer et al., 2017) or the simulation study (Nuebler et al., 2018). We are hence left with the impression that the novelty of this finding is somewhat overstated in this manuscript.
2. It is not quite clear what the authors conceptually mean by "physical boundaries" and how this could offer additional insight into preserved TADs. First, the authors use the CCM model to show that TAD boundaries correlate with peaks in the single cell boundary probability distribution of the model. This finding is consistent with previous reports that TAD-like structures are present in single cells, and that specific TAD boundaries only arise as a population average. The finding based on the CCM simulations hence seems to be that preserved TADs also arise as a population average in cohesin-depleted cells, but we do not follow what the term "physical boundaries" refers to in this context. The authors then use the Hi-C data to infer a maximum-entropy-based HIPPS model. They find that preserved TADs often have boundaries that correspond to peaks in the single cell boundary probabilities of the inferred model. The authors seem to imply that these peaks in the boundary probability correspond to "physical boundaries" that cause the preservation of TADs. This argument seems circular; the model is based on inferring interaction strengths between monomers, such that the model recreates the input Hi-C map. This means that the ensemble average of the model should have a TAD boundary where one is present in the input Hi-C data. A TAD boundary in the Hi-C data would then seem to imply a peak in the model's single-cell boundary probability. (The authors do display two examples where this is not the case in Fig.3h, but looking at these cases by eye, they do not seem to correspond to strong TAD boundaries.) "Physical boundaries" in the model are hence a consequence of the preserved TADs, rather than the other way around, as the authors seem to suggest. At the very least the boundary probability in the HIPPS model is not an independent statistic from the Hi-C map (on which their model is constrained), so we have concerns about using the physical boundaries idea to understand where some of the preserved TADs come from.

References:
Nuebler, J., Fudenberg, G., Imakaev, M., Abdennur, N., & Mirny, L. A. (2018). Chromatin organization by an interplay of loop extrusion and compartmental segregation. Proceedings of the National Academy of Sciences of the United States of America, 115(29), E6697-E6706. https://doi.org/10.1073/PNAS.1717730115/SUPPL_FILE/PNAS.1717730115.SAPP.PDF

Schwarzer, W., Abdennur, N., Goloborodko, A., Pekowska, A., Fudenberg, G., Loe-Mie, Y., Fonseca, N. A., Huber, W., Haering, C. H., Mirny, L., & Spitz, F. (2017). Two independent modes of chromatin organization revealed by cohesin removal. Nature 2017 551:7678, 551(7678), 51-56. https://doi.org/10.1038/nature24281

Reviewer #2 (Public Review):

Summary:

Here Jeong et al., use a combination of theoretical and experimental approaches to define molecular contexts that support specific chromatin conformations. They seek to define features that are associated with TADs that are retained after cohesin depletion (the authors refer to these TADs as P-TADs). They were motivated by differences between single cell data, which suggest that some TADs can be maintained in the absence of cohesin, whereas ensemble HiC data suggest complete loss of TADs. By reananalyzing a number of HiC datasets from different cell types, the authors observe that in ensemble methods, a significant subset of TADs are retained. They observe that P-TADs are associated with mismatches in epigenetic state across TAD boundaries. They further observe that "physical boundaries" are associated with P-TAD maintenance. Their structure/simulation based approach appears to be a powerful means to generate 3D structures from ensemble HiC data, and provide chromosome conformations that mimic the data from single-cell based experiments. Their results also challenge current dogma in the field about epigenetic state being more related to compartment formation rather than TAD boundaries. Their analysis is particularly important because limited amounts of imaging data are presently available for defining chromosome structure at the single-molecule level, however, vast amounts of HiC and ChIP-seq data are available. By using HiC data to generate high quality simulated structural data, they overcome this limitation. Overall, this manuscript is important for understanding chromosome organization, particularly for contacts that do not require cohesin for their maintenance, and for understanding how different levels of chromosome organization may be interconnected. I cannot comment on the validity of the provided simulation methods and hope that another reviewer is qualified to do this.

Reviewer #3 (Public Review):

This manuscript presents a comprehensive investigation into the mechanisms that explain the presence of TADs (P-TADs) in cells where cohesin has been removed. In particular, to study TADs in wildtype and cohesin depleted cells, the authors use a combination of polymer simulations to predict whole chromosome structures de novo and from Hi-C data. Interestingly, they find that those TADs that survive cohesin removal contain a switch in epigenetic marks (from compartment A to B or B to A) at the boundary. Additionally, they find that the P-TADs are needed to retain enhancer-promoter and promoter-promoter interactions.

Overall, the study is well-executed, and the evidence found provides interesting insights into genome folding and interpretations of conflicting results on the role of cohesin on TAD formation.

To strengthen their claims, the authors should compare their de-novo prediction approach to their data-driven predictions at the single cell level.