Session Q06: Genome Organization and Subnuclear Phenomena I: From DNA-Protein Interactions to Higher-Order Structures

Light, CRISPR and DNA repair

We present multi-target CRISPR, an approach whereby a degenerate gRNA directs Cas9 to simultaneously target a large number of epigenetically diverse genomic locations. This multiplexing of CRISPR is enabled by targeting the highly abundant Alu repetitive elements. Combined with different next-generation

sequencing readouts, multi-target CRISPR offered new insights into Cas9 activity and the ensuing cellular response. First, by interrogating the same protospacer in different genomic contexts, multi-target CRISPR showed that Cas9 binding is more efficient at open chromatin regions, and Cas9 cleavage is enhanced near transcribed regions. We then combined multi-target CRISPR with ATAC-Seq to unveil an increase in chromatin accessibility around Cas9 DSBs, which spreads over a window of ~1 kb centered at the cut site.

We attribute this chromatin opening to nucleosomes being evicted from the damaged region. Notably, the extent of chromatin opening is independent of the chromatin context in which the lesions occurs, suggesting that nucleosome eviction is a robust part of the DNA damage response. By combining multi-target CRISPR with photo-activatable gRNAs, we determined the characteristic time-scales of DSB-induced chromatin relaxation and found that nucleosome removal occurs within ~30 min. Last, we combined multi-target CRISPR with a photo-deactivatable gRNA to measure chromatin dynamics after DSB repair completion. Once repair is completed, chromatin accessibility levels returned to their original pre-

cleavage state within 1 h, revealing a quick repositioning of nucleosomes around the repaired cut site. Based on these measurements, we propose a model in which chromatin undergoes fast and transient opening in response to Cas9 breaks to facilitate recruitment of repair factors. In conclusion, multi-target CRISPR enables high-throughput studies of Cas9 activity and DNA repair in cellulo, revealing new insights on the interplay between Cas9 activity and chromatin.   

Structural effects of H4K16 acylations characterized by mesoscale simulation

H4K16 acylations have been implicated in significant biological effects such as the formation of euchromatic nuclear sub-compartments and medical conditions such as Propionic Acidemia, but the molecular mechanisms underlying these effects are just beginning to be understood. Recent experimental analysis of key nucleosomal physics through single-molecule force spectrometry and structural probing has allowed the development and validation of fast mesoscale polymer models for chromatin structure which directly incorporate these acylations, providing a potential avenue for understanding their impact on local chromatin structure. In this work, we perform mesoscale simulations of 12-nucleosome chromatin fibers with three different H4K16 acylations, deriving key model parameters directly from experiments performed in concert, analyze the impact of these acylations on fiber structure, and compare our predictions to the available body of experimental data. Ultimately, we find that H4K16Ac and H4K16Bu significantly disrupt internucleosome interactions, resulting in significantly less compact fibers with dramatically increased accessible DNA surface area. Unexpectedly, we find that key properties of H4K16Pr change the local folding geometry of chromatin fibers, without causing significant decompaction. Our results provide unique insight into the range of possible structural impacts of H4K16 acylations, and may provide the scientific community with a starting point for understanding the novel biological effects of these post-translational modifications.   

A polymer model to infer 3D chromosome structures from Hi-C

We develop a polymer physics-based minimalist model, called heterogeneous loop model (HLM), and infer 3D organization of chromatin from Hi-C data. For a given Hi-C map, HLM is not only a computationally efficient and versatile modeling tool to model chromosome structures, but also offers analytical expressions for various experimental measurements. We demonstrate the utility of the HLM by discussing the multi-way chromatin contacts and co-segregation probability obtained from genome architecture mapping (GAM) as well as explicit 3D chromosome structure.   

On the origin of the integrity of a subset of Topologically Associating Domains upon Cohesin loss in Interphase Chromosomes

Interphase chromosomes undergo phase separation between active and inactive loci to form the A and B compartments on a multi-megabase scale. Below the megabase scale, Topologically Associating Domains (TADs), which are created by the motor protein cohesin in conjunction with the architectural protein CTCF, appear as squares along the diagonal in the contact matrix. Hi-C experiments have revealed that the majority of the TADs at the ensemble level are abolished upon cohesin depeletion, while recent super-resolution imaging studies exhibit cohesin-independent TAD-like domain structures at the single cell level. However, a closer examination of the Hi-C data shows a non-negligible fraction of TADs is preserved (P-TADs) even after cohesin loss. In this study, using polymer simulations, we uncovered that TADs with epigenetic mismatches across their boundaries or with physical boundaries in their 3D structures survive upon loop deletion. Informed by the simulations, we analyzed the Hi-C maps (with and without cohesin) in mouse liver and HCT-116, which corroborated that epigenetic mismatch plays an important role for P-TADs. Interestingly, the single-cell structures, calculated only from wild/cohesin-depleted Hi-C contact maps using the HIPPS method, display various TAD-like domains as in imaging experiments. We found that the preferential positioning of some single-cell domain boundaries is not lost after cohesin deletion, leading to the preservation of the ensemble TADs.   

Investigating the epigenetic code through data-driven chromosome structure modeling.

The three-dimensional organization of chromatin plays a major role in pathways of gene regulation and disease. Increasing evidence indicates that features of the three-dimensional structure control the physical readout of genetic information and may ultimately establish programs of gene expression. Advances in experimental probes of chromatin organization (Hi-C) provide unprecedented insight into pairwise contact frequencies, but the three-dimensional structure is not furnished by these methods. Computational models have been advanced to predict chromatin structures consistent with these experiments, but have thus-far been limited to a resolution no finer than the resolution of state-of-the-art Hi-C experiments (~5 kbp). However, models of this type are unable to investigate the structure interior to individual genes which are crucial to regulatory pathways. We present a model for chromosome architecture at nucleosome resolution, using maximum entropy methods to recapitulate experimental Hi-C contact maps. We explore how models at this resolution can connect length scales from nucleosome interactions all the way to whole-chromosome organization. We expect that models of this resolution are necessary to investigate important features of epigenetic regulation that require information at multiple length scales.   

Interaction energy scales intrinsic to the structural ensemble of chromosomes

Knowledge-based or statistical potential has been successfully used to predict various aspects of folding of proteins and RNA. Here we extend the idea and calculate the statistical potential (SP) between various locus types in order to predict the organization of chromosomes. We estimated the SPs for the pair interactions between loci of binary types (namely, A-A, A-B, or B-B interactions) within a given interphase chromosome from the corresponding contact frequency data in Hi-C experiments. Using the inferred SP values for the energetic parameters in the Chromosome Copolymer Model (CCM), we simulated the structural ensembles for chromosomes 2 and 21 in IMR90 cell, which exhibit the feature of A/B compartments as a result of the phase separation between A and B loci. The interaction scales based on the SPs produce good agreement between the simulated structures and the super-resolution imaging results. At sufficiently high resolution, the SP-based CCM simulations, incorporated with CTCF loop anchors, reveal topologically associating domains (TADs). Each individual chromosome shows different SP values depending on its size and epigenetic sequence as well as the cell type. Our theoretical scheme, which is simple but general, can be applied to many different kinds of experimental data to address the effective chromosome interaction energy scales that govern the underlying structural ensemble.   

FRAP calibrated polymer model of chromatin with HP1a predicts heterochromatin domain sizes, 3D organization, and dynamics.

Phase-separated heterochromatin domains are enriched for H3K9me2/3 with an abundance of heterochromatin protein 1 (HP1a). In vitro, HP1a exhibits liquid-liquid phase separation (LLPS), however, in vivo, its effects on dynamics and structure of heterochromatin domains remain unclear. We develop a simple model containing spherical particles to mimic HP1a protein and a single-chain polymer representing the chromatin. We account for HP1a self-interaction and its attraction to methylation markers and analyze the thermodynamic behavior of the system. We show that the heterochromatin domains are stabilized by the interplay of HP1a self-interactions and HP1-H3k9me2/3 interactions. We reveal that HP1a concentrations inside heterochromatin domains and nucleoplasm depend on the total concentration of HP1a in the nucleus which is consistent with the thermodynamics of multicomponent LLPS. Unlike other polymer models which mostly are tuned on structural experimental data (e.g. Hi-C contact map), we calibrate our model parameters to reproduce the dynamics of HP1a consistent with the in-vivo fluorescence recovery after photobleaching experiment (FRAP) of HP1a. The tuned model predicts the 3D genome organization and heterochromatin domain size verifiable through Hi-C and super-resolution imaging experimental data. Furthermore, our model shows that the transition from compact heterochromatin domains to coil-state euchromatin domains is accompanied by hysteresis, suggesting metastability of both phases. Therefore, our model predicts that substantial perturbation of heterochromatin domains by obliteration of methylation marks (as during replication) does not result in the system leaving that epigenetic state; in other words, cells have a memory of epigenetic states during reorganization.

    

Extended and dynamic linker histone-DNA interactions control chromatosome compaction

Chromatosomes play a fundamental role in chromatin regulation, but a detailed understanding of their structure is lacking, partially due to their complex dynamics. Using single-molecule optical tweezers to address the thermodynamics and kinetics of protein-DNA interactions, we reveal that linker histone interactions with DNA are remarkably extended, with the C-terminal domain binding both DNA linkers as far as ~ ±140 bp from the dyad. In addition to a symmetrical compaction of the nucleosome core governed by globular domain contacts at the dyad, the C-terminal domain compacts the nucleosome's entry and exit. These interactions are dynamic, exhibiting rapid binding and dissociation, sensitive to phosphorylation of a specific residue, and crucial to determining the symmetry of the chromatosome's core. Extensive unzipping of the linker DNA, which mimics its invasion by motor proteins, shifts H1 into an asymmetric, off-dyad configuration and triggers nucleosome decompaction, highlighting the plasticity of the chromatosome structure and its potential regulatory role.   

Quantitative Link Between Chromatin Compaction and Epigenetic Memory at the Single-cell Level

Histone modifications are often correlated to epigenetic memory: permanent changes in gene expression inherited over generations to pass information to descendants or maintain cellular identity. Repressive histone modifications such as H3K9me3 are believed to compact chromatin so that it becomes inaccessible to transcriptional machinery. However, it is not clear if compaction is needed for gene silencing or epigenetic memory. Moreover, quantitative analysis of the direct relationship between chromatin compaction and epigenetic memory is still lacking. Here, we measured 3D chromatin structure changes using multiplexed DNA FISH upon targeted epigenetic perturbations by direct recruitment of chromatin regulators to a reporter gene. We found that KRAB recruitment, known to cause H3K9me3 modifications and epigenetic memory, leads to chromatin compaction around 20kb, and that compaction is retained in permanently silenced cells even after releasing KRAB from the genome. However, recruitment of histone deacetylase HDAC4 can silence the reporter gene without resulting in chromatin compaction. More generally, by testing perturbations that cause different fractions of cells in the population to maintain epigenetic memory, we found that chromatin compaction is correlated with the amount of epigenetic memory rather than gene silencing. Finally, we found that this quantitative connection also holds in a natural context associated with irreversible transitions: epigenetic silencing and chromatin compaction at the Nanog locus in mouse ES cells during differentiation and fate commitment.   

Higher-order genetic interactions in sequence-function relationships

Pairwise interaction models, such as the Potts model, have been extensively applied to study sequence-function relationships. However, modern high-throughput phenotyping assays have shown that genetic interactions among three or more loci also frequently occur. In this talk, I will present ongoing work on modeling these higher-order interactions based on either high-throughput experimental assays or the analysis of collections of functional sequences. We will first present a semi-parametric generalization of the Potts model for estimating probability distributions over sequence space. This generalization is based on a family of linear operators constructed using spectral transformations of the graph Laplacian for the Hamming graph. Such linear operators can then be used to define a prior over the type of higher-order associations that are assumed to be absent in typical maximum entropy models such as the Potts and independent sites models. Next, we will show how higher-order interactions influence the distance correlation structure of local epistatic coefficients, and use these observations to build Gaussian process priors for reconstructing full fitness landscapes from noisy and incomplete experimental data. Finally, we will present new results showing how the above approaches can be generalized to accommodate diploid genotypes as well as a method for building priors over sequence-function relationships with variable allelic weights in order to incorporate a greater degree of anisotropy.   

Multi-scale simulations to assess chromatin organization along the nuclear periphery

Chromatin's complex and heterogeneous structure along the nuclear periphery plays a crucial role in regulating gene expression and function. As peripheral chromatin aggregates in lamin-associated domains (LADs), diverse interactions with lamina proteins, interior chromatin, and other LADs make modeling difficult. Here, we develop a multi-scale model leveraging the natural organization of chromosomes into territories to navigate chromatin's glassy energetic landscape. This work provides accurate chromatin maps showing different types of LADs, highlighting chromosome conformation and positioning as important factors in LAD strength.   

Author not Attending

Both theoreticians and experimentalists have been dedicating significant effort to studying chromosome organization and its relation to gene expression and regulation. Many experimental procedures were designed, such as Hi-C or ChIA-PET, to capture structural information of chromosomes. Several theoretical models were proposed to understand the principles underlying chromosome compartmentalization. MiChroM is a model that describes the compartmentalization behavior of interphase chromosomes. Compared to experimental data, the agreement of the in silico contact maps suggested that phase separation of A/B compartments is a key feature determining the structure of individual chromosomes. In this work, we use the Open-MiChroM software to simulate the 46 chromosomes of a GM12878 human cell at 50 kb resolution, seeking to understand whether the nuclear organization of the whole genome would rely on the same principles observed for the organization of a single chromosome. The ensemble of structures generated by our simulations captures chromosome organization within territories and phase separation of chromatin compartments. Inside a chromosome territory, active chromatin localizes preferentially in the periphery, while inactive chromatin lies on the inside. The intra-chromosomal contact maps and the shape metrics of each territory are consistent with experimental data. The resulting genome architecture resembles the inverted nucleus configuration, with active chromatin towards the periphery. In the interface between chromosomes, we also observe phase separation of compartments. However, the inter-chromosomal maps present higher contact probabilities for heterochromatin loci when compared to the experimental data. These results suggest that the interactions of chromatin with other nuclear features like the lamina greatly influence inter-chromosomal interactions while also affecting the overall positioning of chromosomes in the nucleus.   

Statistical physics models of transcriptional state

The state of a cell can be defined in part by the expression levels of all its genes. New experiments combine super-resolution microscopy with combinatorial probes to count each RNA molecule transcribed from each of many genes, giving a snapshot of the cell state in its high dimensional space. We use ideas from statistical physics to describe the distribution of ~100 genes from ~80,000 human cells. If we treat each gene as on or off (expressed above or below its mean level), we can build maximum entropy models for the resulting binary variables, matching their means and pairwise correlations. These are Ising models, and provide surprisingly good predictions of higher-order correlations in the network. However, if we try to match the mean and pairwise correlations of the full range of molecular counts, the resulting models fail completely, since the observed (co)variances fall outside the bounds of what these models can achieve. This leads to the problem of building maximum entropy models that match higher moments, or even full marginal distributions, along with pairwise correlations. In a different direction, we ask how the distribution of expression levels evolves as we coarse grain our description of the system, removing modes that make the smallest contribution to the total variance. Distributions appear to approach a fixed non-Gaussian form with fewer modes remaining, pointing toward simpler but still high dimensional dynamics.