Session Z06: Genome Organization and Subnuclear Phenomena III: Measuring and Manipulating Genome Organization

Nuclear mechanotransduction & stem cell fate regulation

Precise control of three-dimensional (3-D) nuclear organization is critical for regulation of gene expression and the establishment and maintenance of cell identities. Stem cells execute these identity transitions in dynamic environments where they constantly undergo state changes in response to forces inflicted by cell movements, cell divisions and tissue growth. Recent work from us and others implicate mechanical force, by activating biochemical signaling in the cytoplasm as well as triggering mechanotransduction through nuclear deformation, in remodeling nuclear architecture, chromatin state, and global gene expression patterns in both somatic and embryonic stem cells. Using quantitative imaging and sequencing approaches, we show that mechanical deformation of nuclei results in a stress response characterized by reduced peripheral H3K9me3-marked heterochromatin, attenuated global transcription and increased H3K27me3-mediated silencing of lineage commitment genes. To assess if these changes resulted from a tightly controlled, reversible and stereotypic process or rather represented a stochastic breakdown of homeostasis, we characterized deformation-triggered alterations in chromatin accessibility and transcriptional activity of human induced pluripotent stem cells (hiPS) on a single cell level in their pluripotent state and at exit from pluripotency. Analyses of these data and subsequent follow up experiments revealed a mechano-osmotic nuclear mechanotransduction pathway consisting of two parallel components: 1) an osmotic pathway regulated by nuclear volume loss resulting in global transcriptional repression and a subsequent H3K27me3-regulated mechanical memory that attenuated exit from pluripotency and 2) a nuclear membrane tension-triggered pathway that resulted in activation of mechanoadaptive gene expression driven by mechanosensitive transcription pathways. Collectively this work reveals a universal mechanism by which nuclear deformation triggers changes in chromatin architecture and gene expression, resulting in epigenetic memory affecting stem lineage progression.   

Epigenetic Reactions and Chromatin-lamina Interactions Synergistically Regulate the Morphology of Lamina-associated Domains

Lamina-associated domains (LADs) constitute a significant portion of transcriptionally silent chromatin in eucaryotic nuclei. The spatial organization of LADs exerts stimulating effects on gene repression and cell function. To elucidate the mechanisms underlying the morphological modulation of LADs, we propose a theoretical model incorporating chromatin-chromatin interactions, chromatin-lamina interactions, and epigenetic regulations. Our model predicts a phase-diagram of LAD thickness, which is determined by the synergistic contributions of both the strength of chromatin-lamina interactions and the level of histone methylation. The model is validated for in-vitro nuclei under the alternation of chemo-mechanical cues such as epigenetic suppression and substrate stiffening. Our model serves as a novel tool to use super-resolution images to quantitatively infer the histone methylation rate as well as the strength of chromatin-lamina interactions under different chemo-mechanical stimuli. Our findings suggest that histone deacetylase (HDAC) together with chromatin anchoring lamina-associated polypeptide 2β (LAP2β) modulates the LAD organization, serving as a potential mechanism driving the response to the microenvironmental stimuli. Our work has significant relevance in understanding the roles of microenvironmental factors in regulating chromatin morphology, providing insights into the cell response to vital biological processes, including development, cancer metastasis, and degenerative diseases.

    

Polymer model predicts history dependent epigenetic and lamin-associated domain sizes

The association of epigenetic markers with phase-separated genomic compartments is well known. Although correlated changes in 3D organization and epigenetics in response to stimuli have been observed, current physical models cannot explain this behavior. To tackle this problem, we develop a dynamic, reaction-based polymer model which takes input from ChIP-seq data to identify the initial active and repressed beads on a string to represent the chromatin. The interaction potentials between different epigenetic states are tuned against known spatial distance-genomic distance relationships and show established hierarchical structures in the obtained HiC contact map. The proposed polymer dynamics form finite-size domains and prevent spatial domain coarsening. We further show that the dynamics succeed in maintaining epigenetic domain stability and spatial distance relations. Chromatin compaction and expansion are seen in response to epigenetic spreading through methylation and acetylation reactions, respectively, which agrees with experimental STORM imaging observations. We also predict 3D chromatin reorganization in response to changing chromatin-lamin interactions. Lastly, we expose our system to changing stimuli and find that the genomic domains show memory-like features. The simplicity of the dynamics, along with use of biological inputs, is a significant step forward in understanding how chromatin dynamics affect both short-term and long-term cellular behavior.

    

Chromosome Modeling on Downsampled Hi-C Enhances Compartmentalization Signal

The human genome is organized within a nucleus where chromosomes fold into an ensemble of different conformations. Chromosome conformation capture techniques such as Hi-C provide information about the genome architecture by creating a 2D heatmap. Initially, Hi-C maps experiments were performed in human interphase cell lines. Recently, efforts were expanded to several organisms, cell lines, tissues, and cell cycle phases where obtaining high-quality maps is challenging. Poor sampled Hi-C maps present highly sparse matrices where compartments located far from the main diagonal are difficult to observe. Aided by recently developed models for chromatin folding and structure, we develop a framework to enhance the compartments' information far from the diagonal observed in experimental sparse matrices. The simulations were performed using the Open-MiChroM platform aided by new trained parameters into the Minimal Chromatin Model (MiChroM) energy function. The simulations optimized on a downsampled experimental map (10% of the original data) allow the prediction of a similar contact frequency to the complete (100%) experimental Hi-C. The modeling results open a discussion on how simulations and modeling can increase the statistics and help fill in some Hi-C regions not captured by poor sampling experiments. Open-MiChroM simulations allow us to explore the 3D genome organization of different organisms, cell lines, and cell phases that often do not produce high-quality Hi-C maps.   

Contrasting the organization and dynamics in normal and cancer cell chromosomes

Chromatin is organized into multiple domains, such as phase-separated compartments and topologically associating domains (TADs). Although the role of TADs in gene regulation is still unclear, it has been shown that there are differences between normal and cancer cells. Hi-C experiments show that sizes and epigenetic states are different between normal and cancer cells. We used the knowledge-based potential in Chromosome Copolymer Model (CCM) to generate the structural ensemble of chromatin. We simulated the structures of chromosomes 12 and 15 in RWPE1 and C42B cell lines. We first demonstrate that the structures from simulations are in agreement with the structures from Hi-C (using HIPPS) at the resolution of TADs. In particular, for the TAD regions that differ in the Hi-C contact maps between normal and cancer cells, we found that the distributions of locus pair distance and radius of gyration obtained from the simulations show the differences qualitatively consistent with those expected from the Hi-C data. We highlight that these structural differences lead to the dynamical changes in the given TAD regions, which are possibly connected with the differences in gene expression levels between normal and cancer cells.   

Image-based Spatial Genomics in Health and Disease

The three-dimensional (3D) organization of the genome affects many genomic functions. Changes in genome architectures have been associated with normal development, aging, and a wide range of diseases. Despite its critical importance, understanding the spatial organization of genome in single cells, the variation of the organization in diverse cell types in complex mammalian tissue, and the regulatory mechanisms and functions of the organization in different biomedical processes remains a major challenge. To address these questions, we developed image-based 3D genomics and spatial multi-omics techniques termed chromatin tracing and MINA to map 3D genome and multi-ome in single cells and complex mammalian tissues. We are applying these developments to depict multiscale genomic architectures associated with gene expression and other genomic functions in different types of cells and tissues undergoing different biological processes in cancer, aging, and immunology. We are also systematically identifying novel regulators of 3D genome by developing high-content image-based CRISPR screens. These novel regulators may serve as a new class of therapeutic targets to halt or even reverse deleterious 3D genome changes in diseases. In sum, the talk will focus on our efforts on: 1) characterizing the 3D genome with new technology developments, 2) understanding 3D genomic functions in the native tissue context in health and disease, and 3) building the 3D genomic “regulatome” to manipulate the 3D genome.

    

A computational model for correcting intrinsic enzymatic biases in chromatin accessibility profiling data

Genome-wide profiling of chromatin accessibility using high-throughput sequencing-based techniques such as ATAC-seq (with Tn5 transposase) or DNase-seq (with DNase I) has been widely used for studying regulatory DNA elements and transcriptional regulation. Enzymatic DNA cleavage exhibits intrinsic sequence biases that confound data analysis. We developed a simplex encoding-based mathematical model for accurate estimation of such sequence biases. Here we present a computational method for correcting intrinsic biases from both bulk and single-cell sequencing data. We show that bias correction can improve transcription factor binding prediction from DNase footprints and can generate more accurate cell clustering from single-cell ATAC-seq data. This work demonstrates that innovative quantitative modeling can enhance conventional bioinformatics analysis and extract biologically meaningful information from complicated high-throughput data.   

Live-cell magnetic micromanipulation of a genomic locus reveals interphase chromatin mechanics

Our understanding of the physical principles organizing the genome in the nucleus is limited by the lack of tools to directly exert and measure forces on interphase chromosomes in vivo and probe their material nature. Here, we introduce an approach to actively manipulate a genomic locus using controlled magnetic forces inside the nucleus of a living human cell. We observed viscoelastic displacements over micrometers within minutes in response to near-piconewton forces, which are consistent with a Rouse polymer model. Our results highlight the fluidity of chromatin, with a moderate contribution of the surrounding material, revealing minor roles for cross-links and topological effects and challenging the view that interphase chromatin is a gel-like material. Our technology opens avenues for future research in areas from chromosome mechanics to genome functions.   

Bridging nano- and mesoscale nuclear organization with single-molecule microscopy

Chromatin in the cell nucleus is composed of nucleosomes wrapped with DNA, forming a “beads on a string” structure. At the microscale, chromatin self-organizes into dense domains with less-dense interchromatin space in between, which is posited to play a role in separating the DNA into “active” and “repressive” zones stochastically. However, evidence for the relation between the chromatin organization and its biological function is largely based on protein localization studies in fixed cells or extrapolated from population measurements using biochemical approaches. Here, we combine high-resolution lattice light sheet microscopy, super-resolution localization microscopy, and single particle tracking to investigate the relationship between chromatin structure, nucleosome dynamics, and biological functions. Interestingly, we observed that nucleosome motion and packing vary within different chromatin compartments, resemble a glassy dynamic material, and are consistent with a fractal globule model of chromatin.

    

Positional information transfer by sub-micron transcription factor clusters

Rapid and reproducible cell fate decisions require accurate interpretation of input signals. In the early fly embryo, cells gain information about their physical location by the reading of transcription factor (TF) molecules that are distributed in a long-range gradient spanning the major axis. Interpretation of information in the nucleus results from interaction of these diffuse transcription factor molecules with gene regulatory enhancer elements on the DNA. Recently clustering of the molecules inside the nucleus has been observed, but whether these clusters report directly on the extra-nuclear information is unclear. By imaging fluorescently-labelled TF molecules in live embryos, we show that clustering only occurs at target genes and this gives rise to the observed heterogeneity in nuclear TF distribution. The number of clusters detected per nucleus at various embryo positions is surprisingly reproducible across embryos. An exponential fit of the average intensity of the TF clusters shows that the cluster intensity decays twice as slowly as the nuclear intensity. However the sum of all cluster intensities within a nucleus, weighted by the cluster sizes has the same exponential decay constant as that of the nuclear intensity. Thus, we show that while clustering enables differential interpretation of the nuclear TF concentration at the gene loci, the clustered molecules cumulatively reflect the information in the protein concentration gradient and hence the positional identity of a cell.   

Single-molecule tracking reveals two low-mobility states for chromatin and transcriptional regulators within the nucleus

Transcription factors (TFs) scan the nucleus in search of their consensus binding motifs located within enhancers or promoter-proximal regions. The mechanism by which TFs navigate the complex nuclear environment to assemble the transcriptional machinery at specific genomic loci remains elusive. Using single-molecule tracking, coupled with machine learning, we examined the mobility of multiple transcription factors and coregulators. We show that chromatin (labeled by histone H2B), steroid hormone receptors, as well as other transcriptional coregulators, architectural proteins, and remodelers, all display two distinct low-mobility states. Our results indicate that both low-mobility states are intimately coupled with mobile chromatin. Ligand activation results in a dramatic increase in the proportion of steroid receptors in the lowest mobility state. Mutational analysis revealed that chromatin interactions in the lower mobility state require an intact DNA-binding domain as well as domains important for forming protein complexes with other binding partners. These domains are not necessary for engagement with the higher mobility fraction of chromatin. Importantly, these states are not spatially separated as previously believed but in fact, individual H2B and TF molecules can dynamically switch between them. Together, our results identify two unique and distinct low-mobility states of transcriptional regulators that appear to represent common pathways for transcription activation in mammalian cells.

    

Modeling pairing dynamics of homologous chromosomes during Prophase I in Saccharomyces cerivisae using polymer physics

The development of viable offspring during sexual reproduction requires properly formed egg and sperm cells—the products of meiosis. During meiosis, homologous chromosomes pair and exchange genetic material, and errors in this process can result in defective daughter cells associated with miscarriages and birth defects. Pairing occurs in Prophase I, during which linkages tether and align the homologs together. We explore the dynamics associated with this pairing process in Saccharomyces cerevisiae through a polymer physics model of two randomly linked flexible polymers. This model recapitulates the heterogeneous subdiffusive behavior of chromosomal loci observed experimentally, as well as the temporal behavioral shifts that signify the progressively increasing number of linkages throughout Prophase I. We develop a predictive model for these linkages by calculating the evolving probability distribution for their genomic positions. Furthermore, the effect of cohesin-induced looped chromosomal structures on linkage formation is implemented in our model to study the competition between the fundamental polymer physics of chromosomes and the various biological influences that dictate linkage formation.   

Interpretable and tractable probabilistic models for single-cell RNA sequencing

We outline a class of models of technical and biological variability for single-cell RNA sequencing (scRNA-seq). This technology can quantify the number of RNA molecules in millions of cells at a genome-wide scale. Due to the volume of the data, typical scRNA-seq analyses are ad hoc and descriptive, typically relying on normalization, background subtraction, and data filtering. We argue that the descriptive worldview is insufficient, as it oversimplifies the intrinsic variability in the transcriptional and experimental processes.

Our approach emphasizes a class of processes that are particularly suited to the analysis of scRNA-seq data. To facilitate biological discovery, these processes are interpretable: each parameter encodes the rate of a biophysical phenomenon. To enable use with large datasets, they are also tractable, translating the convolution of noise sources into the composition of generating functions. We present computationally facile solutions to generic biological systems, which couple stochastic transcriptional driving to downstream RNA processing. In addition, we outline probabilistic strategies for phenomena particular to scRNA-seq, such as non-ergodicity in developmental trajectories, background contamination, and variability in capturing individual molecules.