Date Published: June 01, 2017
Publisher: International Union of Crystallography
Author(s): Marcus D. Wilson, Alessandro Costa.
This review article summarizes recent progress in our understanding of chromatin biology based on single-particle cryo-electron microscopy studies.
Each human cell contains over 2 m of DNA that is highly compacted by chromatin into the cell nucleus, which in turn measures only around 50 µm3. Structural biology approaches have started to reveal how DNA is compacted and modified in the cell. For example, early electron-microscopy (EM) work confirmed biochemical conclusions that the basic unit of chromatin is the nucleosome, which compacts DNA around a central discoid of tetrameric H3–H4 capped by H2A–H2B dimers on either face (Richmond et al., 1984 ▸; Klug et al., 1980 ▸). The nucleosome core particle (NCP) is roughly 10 nm in width, with 1.65 turns of DNA wrapping around the equator of the coin-shaped particle. The landmark publication of the 2.8 Å resolution structure of the nucleosome (Luger et al., 1997 ▸) revealed the key features of the histone fold and protein–DNA interactions. The near-atomic map was made possible in part by the use of entirely recombinant histones and strong-positioning DNA, reassembled to create a more homogenous population of NCPs than those isolated from cells. Each histone exhibits a characteristic three-helical dumbell shape, with largely unstructured N- and C-terminal tails (Fig. 1 ▸). The DNA contacts the octameric disc, with numerous basic residues that map onto the outer perimeter of the histone core and project into the DNA minor groove, engaging in non-sequence-specific interactions. The solvent-exposed upper and lower faces of the nucleosome form an undulating surface with distinct electrochemical features used for chromatin protein recognition. The histone tails are the major site of post-translational modification (Ng & Cheung, 2016 ▸; Ruthenburg et al., 2007 ▸); they have been described to be in multiple conformations and are likely to be highly flexible (Luger et al., 1997 ▸; Davey et al., 2002 ▸; Hansen et al., 2006 ▸) but can become ordered upon protein binding (Armache et al., 2011 ▸; Arita et al., 2012 ▸). The NCP provides a platform that facilitates the reading and copying of the bound DNA and helps to control the myriad of DNA-related processes in the cell. The relative scarcity of nucleosome structures represents a major obstacle in understanding how nucleosomes are modified, read, unwrapped, removed and deposited. X-ray crystallographic studies have revealed how different histone variants and DNA sequences affect the core NCP (reviewed in McGinty & Tan, 2015 ▸). However, despite recent progress, relatively few X-ray structures of NCP-bound complexes, epigenetically modified NCPs and higher order NCP arrays exist (McGinty & Tan, 2016 ▸).
Despite their relatively small size (∼200 kDa), nucleosomes are highly compact and provide relatively high contrast in cryo-EM owing to increased electron scattering from the wrapped nucleosomal DNA. Despite these advantages, identifying small NCP particles in vitreous ice can be challenging, and most cryo-EM NCP structures to date have been visualized with added mass, either from bulky post-transitional modifications (Wilson et al., 2016 ▸) or in complex with large protein assemblies (Maskell et al., 2015 ▸; Yamada et al., 2011 ▸; Xu et al., 2016 ▸). Chua and coworkers circumvented this problem by using a Volta phase plate, allowing increased contrast at low spatial frequencies and improved particle alignment (Chua et al., 2016 ▸). These authors were able to reconstruct a final cryo-EM map with a resolution of 3.9 Å, which agreed well with available crystallographic structures of NCPs (Fig. 1 ▸). By comparing the EM density with the higher resolution published crystal structures, subtle details of the histone tails can be resolved (Chua et al., 2016 ▸; Wilson et al., 2016 ▸; Fig. 1 ▸). Compared with the EM maps, the path and density for the histone H3 and H4 tails are better ordered in crystal structures. However, even from the lower resolution EM maps the N- and C-terminal tails of histone H2A can be observed, suggesting these may be well ordered in isolated particles in vitreous ice (Fig. 1 ▸).
Only a handful of chromatin-binding protein–NCP structures have been determined by X-ray crystallography (Barbera et al., 2006 ▸; Makde et al., 2010 ▸; Armache et al., 2011 ▸; McGinty et al., 2014 ▸; Girish et al., 2016 ▸; Morgan et al., 2016 ▸; Fang et al., 2016 ▸; Zhou et al., 2015 ▸). All structures to date utilize multiple elements of the surface of the NCP to garner nucleosomal specificity and affinity, engaging in a multivalent manner. Multivalency may be imparted via multiple contacts within the same domain. Examples of this type of interaction have been reported for the known crystal structures of Rcc1 and the Sir3 BAH1 domain (Makde et al., 2010 ▸; Arnaudo et al., 2013 ▸; Armache et al., 2011 ▸). Alternatively, cooperative binding to NCPs could be built up through the genetic linkage of different chromatin-binding domains into a single polypeptide or several reader domains within the same protein complex. Indeed, tandem-adjacent histone code-recognition modules have been found in multiple proteins (Ruthenburg et al., 2007 ▸; Ng & Cheung, 2016 ▸; Taverna et al., 2007 ▸). Intriguingly, many nucleosome binders utilize a region of high negative charge, termed the acidic patch, formed between residues in histones H2A and H2B (Fig. 1 ▸). A common arginine anchor motif has been described in all acidic patch interactors to date (McGinty & Tan, 2016 ▸; McGinty et al., 2014 ▸).
NCPs become decorated with a wide range of post-translational modifications, which directly control DNA accessibility and binding to specific interactors. In turn, these histone-binding factors can alter the structural properties of chromatin, helping to coordinate DNA-related processes in the cell. The available crystal structures focus on isolated domains bound to short stretches of modified peptide. Indeed, it is clear that many proteins exhibit a higher affinity for chromatin than would be expected from a simple binding event to a short linear primary sequence. Numerous studies have now shown the critical relevance of analysing modified chromatin interaction within the context of an NCP (Xu et al., 2016 ▸; Bartke et al., 2010 ▸; Nikolov et al., 2011 ▸; Ng & Cheung, 2016 ▸). This suggests that the common theme of multivalent binding of chromatin proteins to the nucleosome surface also extends to the recognition of post-translationally modified NCPs in the form of the ‘histone code’. The majority of post-translational modifications are found on the disordered histone tails, and little structural information is available on how covalent modifications affect an NCP.
In cryo-EM, the rapid freezing of proteins into vitreous ice hopes to recapitulate the status of proteins in solution. Indeed, a diverse set of conformational states of the same macromolecular assembly can be isolated from an EM data set in silico. The nominal reported resolution reflects a global estimate derived from the entire three-dimensional structure. Owing to the nature of single-particle averaging in electron microscopy, an EM structure can span a large resolution range, providing high-resolution information on a structured core as well as information on conformational variability at the particle periphery. As a result, in comparing EM and crystallographic structures it should be noted that the methods for estimating resolution are inherently different. The local resolution of EM maps can be calculated by ResMap (Kucukelbir et al., 2014 ▸), allowing direct quantitation of the fluctuations in local map resolution. This data can be displayed in the form of heat maps and allows comparison not only within a structure but also between structures of comparable resolution, often providing important mechanistic insights.
Early rotary shadowing EM studies of partially unfolded chromatin revealed a characteristic ‘beads-on-a-string pattern’ of regularly spaced NCP arrays connected by linker DNA (Thoma & Koller, 1977 ▸). How more than 2 m of DNA is further compacted in the nucleus of each human cell has been the subject of intense research efforts. Higher order chromatin is likely to be arranged in multiple mixed states (Kuznetsova & Sheval, 2016 ▸). Cryo-EM has helped to reveal how one model of chromatin compaction, the ‘30 nm fibre’, may occur. 30 nm-like structures can be formed using in vitro reconstituted nucleosome arrays incubated with linker histone (Song et al., 2014 ▸), similar to those observed in cells (Scheffer et al., 2011 ▸; Li et al., 2015 ▸; Finch & Klug, 1976 ▸).
We currently lack a molecular understanding of how most chromatin-binding proteins interact with nucleosomal DNA, making the study of chromatin superstructure an exciting emerging field. Cryo-EM is an important addition to the structural biologist’s toolkit and will enable us to visualize increasingly complex biological systems centred on chromatin. Indeed, as we have outlined, cryo-EM offers a unique tool to help to investigate previously intractable nucleosome-bound factors. Unlike other structural biology techniques, the visualization of macromolecules in cryo-EM is only limited by their biochemical formation, their stability and the ability to discern particle orientations in the micrographs. Nevertheless, optimizing grid freezing and imaging conditions in cryo-EM is still a laborious task that prevents high-throughput structure determination at present.