UC Berkeley

DNA is well protected and organized inside the living cell. The ‘architecture’ of the genetic material not only provides a “hard disk” for the storage of “gene files”, but also zips the meters-long DNA string into a micron size spatial volume that enables the positioning, exporting, and copying of the “files” at a high frequency and accuracy. Therefore, how DNA compacts itself and fits within a confined cellular volume and what mechanisms cells use to regulate the accessibility of the genetic material are fundamental questions that remain at the forefront of scientific research. As a high-resolution imaging tool, electron microscopy (EM) studies of the genetic material enabled the first glimpse of the ultrastructure of chromatin in mid-twentieth century (1970). Due to the limitation of sample preparation and 3D reconstruction techniques, these early studies were confined to a 2-dimensional (2D) characterization of chemically fixed or stained chromatin samples. In the last decades, along with the maturation of the cryo-EM single-particle averaging (SPA) reconstruction approach, it became possible to determine the high-resolution 3-dimensional (3D) structure of medium and large protein molecules in vitro. However, the flexible character and highly dynamic nature of nucleic acid polymers makes it difficult to take advantage of SPA to increase the signal-to-noise ratio of the resulting averaged image. As an alternative 3D structure determination method, cryo-electron tomography (cryo-ET) has been used to specifically solve the structure, at intermediate resolution, of biological samples that present large heterogeneity, from large-scale tissue structures, to subcellular organelles, to dynamic protein assemblies. Therefore, finding experimental conditions and optimizing cryo-ET 3D reconstruction workflows that improve the resolution and recognition of the target samples can pave the way to a better understanding of the detailed structure of genetic material organization.

For the eukaryotic system, the genetic material is organized into chromosomes, with each of them being composed of long, intertwined chromatin fibers that consists of repeated nucleoprotein units called nucleosomes (Fig. 1.1). This hierarchical, step-wise organization was assumed to be a key step in the regulation of gene expression in response to the stage of the cell cycle and environmental changes. The robust detection of the conformations of fibers is still a difficult task for structural biologists. Similarly, defining the transitions between the different packaging states has remained elusive to biophysicists. In this dissertation, I will first describe our efforts to push the boundaries of cryo-ET reconstruction. In order to achieve higher resolution in ET reconstructions, cryo-ET adopted a similar notion to the ensemble averaging as SPA, an approach called “sub-tomogram averaging”. Its workflow avoids the requirement of identifying millions of particles from 2D images by averaging fewer 3D particles from tomograms to obtain higher resolution. However, when investigating samples with large structural dynamics, the conformational heterogeneity still represents a challenge. Ideally, the reconstruction can be done just with the information collected from one particle, if images were clear enough. To explore the feasibility of this ultimate goal, and to better understand what its limiting factors are, I decided to use what I call “individual particle” cryo-ET reconstruction. In the process, I have performed a systematic characterization of a refined workflow that includes sample preparation, data collection, image processing, 3D reconstruction and refinement of mono-nucleosome samples with long flanking DNA arms as a test sample, in order to achieve an improved reconstruction. The reconstruction achieved via this “individual particle” cryo-ET technique indeed provides enough resolution to describe the DNA arm flexibility and 3D dynamics.

To address the question of chromatin fiber organization and dynamics, we adopt a “bottom up” strategy that begins by characterizing a simpler system from in vitro reconstituted nucleosome arrays under low salt condition to simulate the open or transcriptionally-active state of chromatin (euchromatin). Since conformational transitions in chromatin depend on how flexible the “joints” that connect the nucleosome units within the array are, mapping out the dynamics of those “joints” turns out to be critical for understanding the intrinsic structural regulatory mechanisms of chromatin fibers. As will be shown here, having established the “individual-particle” cryo-ET methodology, we were able to investigate the effect of different factors (including array length, ionic strength, or the involvement of protein condensation factors such as linker histone H1) on nucleosome array dynamics. By quantitatively analyzing the various array structures, we extracted useful statistics and identified key parameters that determine array conformation. Resampling of the statistics allowed us to describe and reconstruct a longer nucleosome array fiber structure in silico, which provides a possible explanation for some of the controversy surrounding the existence of various types of “30-nm” fiber structures in vivo.

Throughout the cell cycle, chromatin can undergo a transition from its open or active state into a highly condensed or transcriptionally silenced state (heterochromatin or mitotic chromatin). In recent years, evidence has been obtained that rather than being governed by a hierarchical packing process, the transition between the lightly packed fibers and the condensed chromatin is mediated via a reversible thermodynamic process that has been termed “liquid-liquid phase separation”. The interaction between the intrinsically disordered regions (IDR) of histones and DNA are thought to be mainly responsible for this transition. Though liquid-liquid phase separation has been described in numerous fluorescence microscopy studies of proteins possessing intrinsically disordered regions (IDRs), as well as RNAs, in vitro studies specifically addressing the liquid-liquid phase separation that has been proposed for chromatin are much fewer, and questions regarding the detailed physical mechanism and organization of the nucleosome units within “condensates” have remained unanswered. Accordingly, I set up to exploit the capabilities of cryo-ET for the study of the early stage formation of condensates from a homogenous nucleosomal phase. Our success in using high-quality 3D reconstruction to capture the condensates at various stages of their formation and with sufficient resolution to distinguish individual nucleosome features, has enabled us to analyze, for the first time, the nucleosome distribution, organization, and orientation within the condensates. The characterization of both the interior and the overall geometry of the condensates provided us with new insights into the mechanisms of how nascent condensates emerge from the solution. Overall, this advanced “individual-particle” methodology applied to samples at different times during the condensation process, allowed us to perform biophysical measurements to describe the chromatin dynamics during liquid-liquid phase separation. In a broader sense, our approach opens a number of future research directions to address, in a similar fashion, other heterogeneous biological systems.