A guide to visualizing the spatial epigenome with super-resolution microscopy

Abstract

Genomic DNA in eukaryotic cells is tightly compacted with histone proteins into nucleosomes, which are further packaged into the higher-order chromatin structure. The physical structuring of chromatin is highly dynamic and regulated by a large number of epigenetic modifications in response to various environmental exposures, both in normal development and pathological processes such as aging and cancer. Higher-order chromatin structure has been indirectly inferred by conventional bulk biochemical assays on cell populations, which do not allow direct visualization of the spatial information of epigenomics (referred to as spatial epigenomics). With recent advances in super-resolution microscopy, the higher-order chromatin structure can now be visualized in vivo at an unprecedent resolution. This opens up new opportunities to study physical compaction of 3D chromatin structure in single cells, maintaining a well-preserved spatial context of tissue microenvironment. This review discusses the recent application of super-resolution fluorescence microscopy to investigate the higher-order chromatin structure of different epigenomic states. We also envision the synergistic integration of super-resolution microscopy and high-throughput genomic technologies for the analysis of spatial epigenomics to fully understand the genome function in normal biological processes and diseases.

Graphic Abstract

Higher-order chromatin structure plays an important role in regulating gene expression. Recent advances in super-resolution microscopy allow direct visualization of spatial epigenomics at an unprecedent resolution in single cells. This review highlights the recent progress, with a focus on the impact of epigenetic modifications on higher-order chromatin structure. We also envision the potential for synergistic integration of super-resolution microscopy and high-throughput genomic technologies for the analysis of spatial epigenomics.

Introduction

Epigenetics in a broader sense refers to additional changes beyond DNA sequence. Each mammalian cell contains about 6 billion base pairs of DNA, about 2 meters long if stretched out. They are wrapped around ~30 million nucleosomes, forming a massive macromolecular complex termed chromatin [1]. The nucleosomes, serving as the basic building block and the first level of chromatin compaction, are further packaged into the higher-order chromatin structure to fit into a small nucleus that has a diameter of just a few microns, a scale at one millionth the length of the stretched DNA. The highly packaged DNA is also compartmentalized in the nucleus, organized into open chromatin (euchromatin) that is transcriptionally active and condensed chromatin (heterochromatin) that is transcriptionally repressed.

Unlike DNA sequence, the higher-order chromatin structure is highly dynamic. It is believed that epigenetic processes are significantly involved in the regulation of gene expression through the 3D organization of the genome [2–4]. A large number of epigenetic modifications, such as covalent modifications of histones (post-translational modifications such as acetylation, methylation, phosphorylation and ubiquitylation) [4–7] and DNA (DNA methylation) [8], have been identified to regulate higher-order chromatin structure and influence the accessibility of genomic DNA to transcription machinery proteins important in many cellular processes, such as gene transcription, DNA replication, recombination, and repair [9, 10]. For example, the densely methylated DNA is generally found in heterochromatic region in the nucleus which is consistent with its known repressed transcription activity [11, 12]. Trimethylation on lysine 9 and lysine 27 on the histone H3 (H3K27me3 and H3K9me3) are enriched at the highly condensed heterochromatin associated with inactive transcription activity. While trimethylation of lysine 4 of histone H3 (H3K4me3) is highly enriched at open euchromatic regions associated with active transcription, and acetylated histone (e.g., histone H4-K16 acetylation and histone K9 acetylation) is associated with decondensed chromatin structures [7].

Our understanding of epigenetics is largely dependent on biochemical assays and powerful DNA sequencing technologies [13] such as chromatin immunoprecipitation (ChIP) based techniques [14] and bisulfite sequencing [15] to decipher the epigenetic “code” along the linear (1D) DNA sequence, but missing the important picture of the physical compaction of 3D higher-order chromatin structure. Other techniques such as micrococcal nuclease (MNase) digestion [16] and ATAC-seq [17] take advantage of enzymes that cleave the accessible DNA to derive the information about chromatin accessibility. Recent development in chromatin conformation capture (3C) based techniques (e.g., 3C, 4C, 5C, Hi-C) further advances our understanding of chromatin structure via a statistically derived picture of contact frequencies between genomic sites. These powerful high-throughput techniques provide genome-wide profiles marked by epigenetic modifications and chromatin structure, but these in-vitro analyses of fragmented DNA neither allow direct visualization of the physical compaction of 3D chromatin structure nor provide accurate spatial scale on the packaged 3D higher-order chromatin structure at the single cell level.

Microscopy techniques, on the other hand, are ideally suited to visualize the physical compaction and chromatin structure at the single-cell level, but conventional light and electron microscopy techniques yield little information about higher-order chromatin structure, due to either limited resolution or insufficient molecular contrast to visualize spatial compaction of nucleosomes and DNA. Recent advances in super-resolution fluorescence microscopy and electron microscopy have dramatically enhanced our ability to visualize the higher-order chromatin structure in situ in live and fixed cells and tissue. An electron microscopy technique with enhanced contrast to genomic DNA (known as ChromEMT) allows direct visualization of chromatin ultrastructure at the highest spatial resolution [18]; a suite of super-resolution microscopy techniques [19–24] are now readily available to visualize the higher-order chromatin structure in vivo down to the resolution of ~20-30 nm, approaching the length scale of the packaged groups of nucleosomes [23]. The multi-color imaging capability of fluorescence microscopy allows us to visualize the packaged higher-order chromatin structure and their spatial relationship with histone modifications and other transcriptional machinery proteins [25, 26]. Further, the ability to image higher-order chromatin structure in live cells is especially powerful to study the dynamics of higher-order chromatin structure in various biological processes or as a response to environmental factors or drugs.

In this review, we discuss spatial epigenomics, or studies of spatial information of epigenomics such as genomic organization, higher-order chromatin structure and their spatial relationship with epigenetic modifications. Our discussion will focus on different technical approaches—with an emphasis on super-resolution fluorescence microscopy—to investigate the spatial epigenomics and gain new insights into biological processes (summarized in Figure 1). We will also provide our perspective on the synergistic integration between super-resolution microscopy and genomic technologies to understand spatial epigenomic landscape in normal development and diseases.

Figure 1.

Conventional epigenomic analysis. Schematic that illustrates conventional biochemical techniques for epigenomic analysis and microscopy techniques to visualize higher-order chromatin structure. ChromEMT images are modified from [18] with permission.

Epigenomic states and chromatin structure derived from biochemical assays

Biochemical assays generally assess the functional sate of chromatin marked by histone modifications (e.g., ChIP-seq), DNA accessibility (e.g., MNase-seq, ATAC-seq), DNA methylation (e.g., bisulfite sequencing) and chromatin conformation (e.g., 3C, 4C, 5C, Hi-C), based on analyzing a large amount of fragmented DNA. ChIP-based techniques are the most widely used techniques to assess how DNA-binding proteins (e.g., histone modifications and transcription factors) affect the functional state of the chromatin they mark [27]. Specific DNA sites that physically interact with histone proteins or other DNA-binding proteins can be isolated by chromatin immunoprecipitation. By mapping the enrichment of histone modifications (or other DNA-binding proteins such as CTCF) at a specific genome location, we can derive the functional state for the specific genomic sequence. Each histone mark is associated with a functional role in regulating gene expression. For example, H3K4me3 and H3K9Ac mark gene promoter regions, H3K4me1 and H3K27Ac mark transcriptional enhancers, H3K36me3 marks transcribed regions of the genome, H3K9me3 marks constitutive heterochromatin that is largely made of repetitive DNA sequences and H3K27me3 marks facultative heterochromatin that repress targeted genes. However, traditional ChIP-based assays depend on the availability of high-quality antibodies and often require a large amount of fragmented DNA via sonication or enzymatic digestion from pooled cell population (millions of cells) to generate “averaged” profiles insensitive to cellular heterogeneity. Ligation-free method (also known as DNA SMART method) was developed based on template-switching technology to accommodate ChIP-seq library preparation from small amount DNA [28] and this technique is now widely used for low-input and single-cell RNA-seq [29, 30]. Other ligation-free methods based on different schemas were also used to measure the spatial proximity between individual genomic loci, such as RNA-TRAP (RNA tagging and recovery of associated proteins) [31], Genome architecture mapping (GAM) [32], split-pool recognition of interactions by tag extension (SPRITE) [33], and ChIA-Drop (multiplex chromatin-interaction analysis via droplet-based and barcode-linked sequencing) [34]. The CRISPR/cas9 genome editing system was also applied to map locus-specific chromatin interactions [35]. Recently, low-input and single-cell ChIP-seq techniques are emerging to reveal epigenetic heterogeneity at the single cell resolution [36]. ChIP-based techniques map the spatial enrichment of specific DNA-binding proteins along the linear DNA sequence, or “one-dimensional” view of the epigenome, but do not reveal spatial compaction of chromatin or spatial proximity of the genomic DNA, limiting their ability to explain the complex control of gene expression programs.

On the other hand, a set of chromosome conformation capture (3C) based approaches have been used to investigate 3D chromatin organization by mapping the genome-wide pair-wise chromatin interactions [37]. The original 3C assay was devised to quantify the interactions between two genomic loci in the 3D space, for example, the promoter-enhancer interaction [38]. Over time, a few 3C-based techniques have been developed, including Chromosome conformation capture-on-chip (4C) [39], Chromosome conformation capture carbon copy (5C) [40], chromatin interaction analysis by paired-end tag sequencing (ChIA-PET) [41] and high-throughput sequencing (Hi-C) [42]. Compared to 3C assay that can only detect the interaction between a single pair of genomic loci (one against one), 4C captures interactions between one locus and all other genomic loci (one against all). While the 5C assay detects interactions between all restriction fragments within a defined region, which is usually < 1Mb (many against many). ChIA-PET analyzes the genome-wide long-range chromatin interactions bound by a specific protein. With high-throughput sequencing, Hi-C is a more powerful technique to study the genome-wide 3D chromatin organization and allows one to identify all interactions both in cis and in trans simultaneously (all against all). By deep sequencing, spatial proximity maps of the human genome quantified by contact frequencies between different genomic sites at a resolution of 1 megabase was constructed by Hi-C [42]. One significant feature of chromatin organization revealed by 3C based techniques is the widespread existence of topologically associated domains (TAD) [43–45], a self-interacting genomic region within which DNA sequences interact with each other more frequently. The mechanism of TAD formation and their function remain to be elucidated, but evidence has shown that TAD is related to many biological processes, such as chromatin organization and gene expression [46–51]. However, similar to ChIP-based techniques, the reconstructed maps of chromosome contact frequency reflect the population average of millions of cells, and the recent contradicting results from single-cell Hi-C lead to the debate that TADs might only reflect a statistical property of averaged cell population [52].

Visualizing spatial epigenomics with conventional fluorescence microscopy

Microscopy techniques naturally provide in-situ visualization of higher-order chromatin structure at the single cell level, while maintaining the spatial context. Using fluorescence in-situ hybridization (FISH) [53] or CRISPR based labeling [54, 55], the spatial arrangement of various genomic foci at the scale of more than 250 nm can be directly visualized with conventional fluorescence microscopy [56, 57]. The FISH labeling of a specific genomic locus has also been combined with Proximity Ligation Assays to study the epigenetic regulation on a specific gene [58] in the spatial context of tissue. Conventional FISH is limited by the small number of targets. Recent advances in sequential labeling and high-throughput imaging have enabled high-resolution and high-throughput imaging on a large number of cells [59] or hundreds of genomic targets in intact cells [60] that significantly improve our ability to investigate the spatial genome organization at the single-cell level. For example, a recent study using high-throughput imaging revealed significant cell-to-cell heterogeneity [61] obscured by the population average. This study takes advantage of the genome-wide nature of Hi-C dataset to identify several hundred pairs of genomic loci that then were imaged by high-throughput microscopy on a cell population. Another study used a multiplexed FISH method for sequential imaging of many genomic regions and identified distinct chromosome folding structures at the single-cell level that deviate from the fractal-globule model derived from the Hi-C dataset [60].

Super-resolution microscopy techniques

The higher-order chromatin structure from a few to tens of nanometers are below the diffraction-limited resolution (~200 nm) of conventional light microscopy. Conventional electron microscopy, despite its superior resolution, is not widely used to visualize higher-order chromatin structure or epigenomic states due to its limited molecular specificity. Recent advances of super-resolution fluorescence microscopy and chromatin electron microscopy overcome these limitations and these powerful tools are poised to study higher-order chromatin structure at unprecedented resolution and molecular details [26, 62–64].

A suite of super-resolution fluorescence microscopy techniques has been developed to break the diffraction limit. The diffraction limit can generally be overcome in four ways: (i) those which spatially and/or temporally modulate the point spread function with patterned illumination such as structured illumination microscopy (SIM) [20], stimulated emission depletion (STED) microscopy [19] and the recently reported MINFLUX [65]; (ii) single molecule localization microscopy (SMLM) using photo-switching or other mechanisms to stochastically turn on a small fraction of the fluorophores in each imaging frame followed by precise localization of center positions of the emitters such as photoactivatable localization microscopy (PALM) [22] and stochastic optical reconstruction microscopy (STORM) [21]; (iii) temporal correlation of fluctuating fluorescent emitters such as super-resolution optical fluctuation imaging (SOFI) [66] and super-resolution radial fluctuations (SRRF) [67]; and (iv) physical expansion of the sample such as expansion microscopy (ExM) [68]. Each super-resolution microscopy technique has its advantages and limitations and the selection of a specific super-resolution fluorescent microscopy techniques often depends on the specific biological questions to be answered, the availability of fluorophores and the cost. SIM and SOFI have a limited spatial resolution of 100 – 120 nm, preventing the visualization of higher-order chromatin structure at the scale of tens of nanometers; but it is compatible with most fluorescent dyes or proteins in imaging fixed and live cells. STED can routinely achieve the resolution of 50-60 nm in both fixed and live cells, but its high laser power requires highly photostable dyes and may also induce damage in live cells. Further, as STED requires high-end lasers and mechanical scanning system, its high cost can be a limiting factor for many laboratories with limited resources. The SMLM-based super-resolution techniques use “switchable” fluorophores that are either photo-switchable or spatially switchable, followed by single molecule localization. The photoswitchable fluorescent proteins used in PALM or fluorescent dyes used in STORM are turned on and off by light; while spatially switchable dyes use the fluorogenic properties upon transient binding of fluorescent dyes to the imaging target to generate stochastic blinking events such as binding-activated localization microscopy (BALM) [69] or point accumulation for imaging in nanoscale topography (PAINT) [70]. The former case is often limited to the photoswitchable fluorescent proteins or dyes with excellent blinking properties (e.g., high photon number, multiple switching cycles) that only a small number of fluorescent dyes or proteins meet such requirement; while the latter case is applicable to a wide range of fluorophores, but often needs a much longer image sequence (e.g., tens to hundreds of thousands of imaging frames) to accumulate sufficient sampling points for image reconstruction. The biggest advantage of the SMLM-based super-resolution imaging techniques is the spatial resolution down to 10 nm. For example, STORM routinely achieves the resolution down to ~20-30 nm [26, 62, 71]; DNA-PAINT that uses transient binding of short fluorescently-labelled oligonucleotides has achieved sub-10 nm spatial resolution [72, 73]. Furthermore, the recent development of MINFLUX combines single molecule localization with patterned illumination to localize the position of molecules at local emission minima and has achieved 1 nm spatial resolution. Of note, the above-mentioned super-resolution microscopy techniques often require computational image reconstruction steps (e.g., deconvolution, Gaussian fitting), which significantly affect the quality of the super-resolution images.

Besides spatial resolution, many approaches have been developed to reduce the complexity and lower the cost. The techniques that use patterned illumination such as SIM and STED often require high-end lasers and/or mechanical scanning system at a high cost; while SMLM uses wide-field illumination and its super-resolved imaging capability is not sensitive to the slight degradation on the stability of the laser output and spectral linewidth, so industry-grade lasers and cameras can be used to construct a SMLM system with a fraction of the cost used in STED or SIM [74–76]. Other techniques such as SOFI/SRRF and ExM aim for achieving super-resolution imaging capability with simple conventional wide-field and confocal microscopy. For example, the X10 expansion microscopy has achieved ~25 nm resolution using conventional wide-field epifluorescence microscope [77]. Due to the limited space of this short review and its focus on the application of spatial epigenomics, we refer our readers to other reviews [25, 64, 78] for more details of super-resolution techniques.

To visualize molecular structure at the highest spatial resolution down to sub-nanometer scale, electron microscopy remains one of the best options. Recently chromatin electron microscopy tomography has used a fluorescent dye (DRAQ5) that stains DNA with an osmiophilic polymer to selectively enhance its contrast under electron microscopy [18]. As electron microscopy provides the best spatial resolution (0.2 nm) [79], the higher-order chromatin structure can be visualized down to sub-nanometer scale. This approach at its present form is mainly used to visualize the ultrastructure of genomic DNA. Multiplexed electron microscopy has also been achieved by different labels including immunogold labeling or lanthanides depositing [80–82]. Although the need for complex processing and significant expertise currently limits its use to a handful of research groups, modern electron microscopy has great potential for future studies of molecular-scale chromatin structure.

Super-resolution imaging of higher-order chromatin structure

Chromatin mainly consists of packaged nucleosomes, where each nucleosome is composed of ~146 base pairs of DNA wrapped around a histone octamer with two copies of core histones (H2A, H2B, H3, H4). The chemical modifications (e.g., acetylation, methylation) on histone tails regulate the compaction of nucleosomes into higher-order chromatin structure and reflect its functional state marked by each histone modification. Therefore, higher-order chromatin structure can be visualized via fluorescence microscopic imaging of three fluorescently labeled targets – DNA [83–86], core histone proteins [23, 87–89] and histone modifications [90–92] (summarized in Figure 2).

Figure 2.

Super-resolution imaging of genome-wide higher-order chromatin structure by labeling DNA, core histone, and histone modifications. A: Super-resolution images of DAPI-stained C2C12 cell nucleus obtained with structured illumination microscopy (SIM). B: DNA structure obtained from fluctuation-assisted BALM (fBALM) of HL-1 cell nucleus using YOYO-1. C: Super-resolution image of nucleus of Vero-B4 cell stained with Vybrants DyeCycle™ Violet. D: Super-resolution images of DNA incorporated with EdU and labeled by Alexa 647 obtained with stochastic optical reconstruction microscopy (STORM). E: STORM images of core histone H2B labeled by immunofluorescence staining with Alexa-647 in fixed human fibroblast cells. F: STORM images of core histone H2B labeled by TMP-ATTO655 in living HeLa cells. G: The super-resolution image of core histone H2B fused with PA-mCherry in live cells, obtained with photoactivated localization microscopy (PALM). H: STORM images of higher-order chromatin structure at different epigenomic states labeled by immunofluorescence staining with Alexa 647. Figures A-H are modified with permission from the following sources: [93, 69, 95, 90, 23, 88, 107, 90], respectively.

Super-resolution imaging of genomic DNA provides a global view of higher-order chromatin structure. A wide range of nucleic acid binding dyes (e.g., DAPI, YOYO, Hoechst) have been used in conventional fluorescence imaging of chromatin structure [83, 93] in fixed and live cells and most are also applicable for STED and/or SIM based super-resolution microscopy (Figure 2A). Due to the higher laser power used in super-resolution microscopy system, photostable fluorophores are often desirable [84, 94]. On the other hand, photoswitchable organic fluorescent dyes are required for SMLM under specific imaging conditions (e.g., higher excitation power density, presence of low oxygen, primary thiol environment). Many organic dyes have been reported to be suitable for SMLM, and many DNA binding dyes can be directly used for chromatin imaging (e.g., YoYo-1 [69], Vybrants DyeCycle™ Violet [95], Sytox Orange [96]), as illustrated in Fig. 2B–C. Alexa 647 (or structural analog Cy5) is among the best-performing photoswitchable organic fluorophores due to its high photon number, excellent blinking properties and high signal-to-background ratio that result in high-quality super-resolution images [97]. For super-resolution imaging of genome-wide chromatin structure, cellular DNA can also be labeled by DNA precursor analog (e.g., 5-ethynyl-2´-deoxyuridine (EdU)) and detected by azide-conjugated dye such as azide-Alexa 647 [90, 98] (Figure 2D). Such labeling approach also has the advantages of high labeling density and small label size (fluorescent dye of ~0.5-1 nm [99]) and can be easily used with other photo-switchable fluorophores (e.g., Cy3B) for multi-color super-resolution imaging. Super-resolution imaging of genomic DNA in mammalian cells revealed compartmentalized open and condensed DNA domains [90, 95], which are dynamically regulated as a response to physiochemical factors [92]. Another approach for super-resolution imaging of genomic DNA is based on DNA-binding based localization. As many DNA-binding dyes exhibit ~1000-fold enhancement in fluorescence signal upon binding to the double-stranded DNA, localization of transient binding events (or “on” state of the fluorescent emitters) can be used to generate super-resolution image, as used in BALM [69, 100, 101] and DNA-PAINT [72]. However, diffusing fluorophores often induce higher background, which can introduce significant image artifacts and degrade the resolution of super-resolution image [102]. Therefore, these binding-based localization approaches are often used with total internal reflection microscopy (TIRF) to naturally reject the background. Further, the recent development of label-free photon localization microscopy (PLM) uses the intrinsic contrast from native, unmodified DNA molecules [103], which offers great promise for label-free super-resolution imaging of intrinsic DNA structures.

Genome-wide higher-order chromatin structure can also be visualized by targeting the core histone proteins (H2A, H2B, H3 and H4). The core histone proteins can be fluorescently labeled via immunofluorescence staining, where the core histone is labeled by fluorophore-conjugated secondary antibody that binds to a primary antibody. Using this approach, Ricci et al. discovered that the nucleosomes are arranged into discrete heterogeneous groups of nanoclusters (or nanodomains) along the chromatin fiber in situ, and observed that the size and density of the nucleosome nanodomains are correlated to “open” or “closed” chromatin [23] (Figure 2E). Such approach is relatively easy and inexpensive to implement and can be widely used in fixed cells and tissue. But it largely relies on the quality of the antibody, which also limits the labeling density. Alternatively, the core histone can be labeled by fusing genetically-encoded protein tags or using self-labeling proteins, such as TMP tag [88] (Figure 2F), SNAP-tag [104] and Halo-tag [105]. Compared to immunofluorescence staining, self-labeling tags or self-labeling proteins do not depend on antibodies, and theoretically can label any known protein. Furthermore, genetically encoded proteins have smaller size (~ 3-5 nm) compared to antibodies, which is especially important in super-resolution microscopy. For immunofluorescence-based labeling, due to the large size of primary and secondary antibodies (150 kDa, ~10 nm in size), the fluorophore can be as far as 20 nm away from the target [106]. Such error may be negligible in most diffraction-limited imaging but can significantly reduce the resolving ability of super-resolution microscopy for resolution less than 100 nm. Another advantage of the genetically encoded proteins is their applicability for live-cell imaging, by infusing a photoconvertible fluorescent protein such as PA-mCherry [107] (Figure 2G), EosFP [22] or mEos2 [106]. Super-resolution imaging of chromatin structure labeled by core histones using self-labeling tags or proteins using STED, PALM or dSTORM has been reported by many groups to study dynamics of higher-order chromatin structure in live cells [87, 88, 107–111]. For example, Nozaki et al used PALM imaging and single-nucleosome tracking with photoactivatable (PA)-mCherry fused with H2B in live cells and discovered that chromatin domains move coherently, altered by various physicochemical factors and even maintained in mitotic chromosomes [107]. Interestingly, they found that the nanoscale chromatin domains even exist when chromosomes assume their most compact form during mitosis—acting as their building blocks—to maintain the genetic information throughout cell cycle.

Visualizing spatial epigenomics using super-resolution microscopy

Conventional epigenomic analysis only provides the enrichment of epigenetic modifications on the genomic location without spatial information; while direct imaging of genomic DNA or core histone provides the global view of genome-wide chromatin structure but lacks the detailed information of epigenetic modifications important for cellular functions and activities. We used the term spatial epigenomics here to describe the studies of genome-wide epigenetic modifications in their spatial context and higher-order chromatin structure. Super-resolution imaging together with quantitative image analysis is ideally suited for visualizing spatial epigenomics.

Super-resolution imaging based on SIM revealed the higher-order structure of inactive X chromosome in human and mouse somatic cells [112]. The super-resolution image of genomic DNA reveals the presence of chromatin domain clusters (CDCs), where H3K4me3 was found enriched at decondensed sites at the boundary of CDCs and interchromatin compartment (IC) in the Xist RNA decorated Barr body; while H3K27me3 was preferentially located at the more compact interior of CDCs. They also showed that Xist RNA spreading is followed by a loss of active H3K4me3 marks and enrichment of repressive H3K27me3 marks. STED has also been used to image histone epigenetic marks in stem cells to study chromatin dynamics and quantitative analysis of image features from various histone marks can be used to depict the emergent cell phenotypes from stem cell differentiation [113].

The studies of spatial epigenomics using localization-based super-resolution microscopy also began to emerge in the past few years. Prakash et al. used SMLM to investigate the epigenomic landscape of meiotic chromosomes at the pachytene stage in mouse oocytes and identified three distinct spatial morphologies and distributions of chromatin domains marked by histone modifications in the context of packaged genomic DNA structure: tangential repressed chromatin domains marked by H3K27me3; radial active chromatin domains marked by H3K4me3 and centromeric polar chromatin domains marked by H3K9me3 (Figure 3A). Recently, Xu et al. used STORM to examine the genome-wide higher-order chromatin structures at different epigenomic states defined by a large number of histone modifications [90] in the interphase and mitotic nuclei of mammalian cells. They identified the key structural difference between histone acetylation and methylation marks, with three distinct structure types: segregated nanoclusters formed by acetylation, dispersed nanodomains formed by active methylation, and compact large aggregates formed by repressive methylation marks (Figure 2H). They also reported the active histone mark (e.g., H3K4me3) largely coincides with open DNA and repressive histone mark (e.g., H3K27me3) overlaps with highly condensed DNA regions (Figure 3B). Both studies demonstrated the distinct structural differences of histone acetylation and methylation in the context of packaged chromatin structure, but the functional link of these distinct structures remains to be elucidated. Although it is generally believed that decondensed chromatin is usually transcriptionally active and the condensed chromatin is transcriptionally silent, some evidence suggests a different picture in certain scenarios. For example, with 3D-SIM, Matsuda et al. observed that in fission yeast, silent chromatin actually has the least condensed chromatin (less condensed than euchromatin), while the most condensed regions were just next to the subtelomeric silent chromatin, and its condensation is regulated by H3K36me3, an epigenetic mark enriched at gene body [114].

Figure 3.

Visualizing spatial epigenomics using localization-based super-resolution microscopy. A: Two-color super-resolution images of DNA and histone marks via immunostained with anti-SYCP3 (Alexa 555) and anti-histone modifications (Alexa 488). B: Two-color STORM images showing the spatial relationship between DNA and different histone modifications. C: STORM images of differential DNA compaction of transcriptionally active (red), inactive (gray), or repressed (blue) epigenetic domains. Figures A-C are modified with permission from the following sources: [91, 90, 117], respectively.

The higher-order chromatin structure at specific genomic loci at the scale of tens of nanometers were also studied by super-resolution imaging combined with FISH-based labeling [115, 116]. Beliveau et al designed oligonucleotide-based Oligopaint probes with secondary oligonucleotides (oligos) to enhance fluorescent signals needed for high-quality STORM images. Boettiger et al. used 3D-STORM to image the long genomic regions (labelled with Oligopaint) enriched with different histone marks corresponding to active, inactive and repressed epigenetic states in Drosophila cells [117] (Figure 3C). By direct mapping of the 3D structure of specific genomic regions, the repressed chromatin showed the densest packing, and Polycomb-repressed domains showed a high degree of chromatin intermixing within the domain. This work supported the role of epigenetic modifications in shaping the 3D structure of the distinct chromatin domains with different biochemical properties. More recently, using a combination of Hi-C, 3D FISH labeling and super-resolution imaging, consistent TAD-based physical compartmentalization of chromatin was observed in Drosophila cell [118]. Multiplexed super-resolution FISH imaging was used for high-resolution chromatin tracing of multiple regions of a chromosome, which starts to reveal TADs-like structure at single cells [119]. The combination of chromatin conformation capture and super-resolution imaging is expected to be increasingly used to reveal the physical structures of chromatin domains in single cells.

Quantitative analysis of higher-order chromatin structure from super-resolution images

Besides direct visualization of spatial epigenomics, quantitative image analysis allows us to extract physical parameters of higher-order chromatin structures, such as categorization of different chromatin compaction levels and size and distribution of chromatin nanodomains. Many image processing and statistical methods have been applied in analyzing super-resolution images of higher-order chromatin structure [23, 69, 90, 107, 120]. The commonly used approaches can be generalized into three categories – quantification of the size and density of clustered chromatin nanodomains based on segmentation or clustering analysis such as K-means clustering [23], Gaussian mixed model (GMM) based clustering [121], Density-Based Spatial Clustering of Applications with Noise (DBSCAN) [122], Bayesian model based approach [123], and Voronoi tessellation [124, 125]; a global overview for heterogeneous chromatin nanodomains with multiple length scales based on radial distribution function (a.k.a. pair correlation function) or Ripley’s K function [107, 126]; and the analysis of spatial proximity between DNA and epigenetic modifications or other functional proteins based on co-localization analysis [127]. The advantages and limitations of different clustering techniques have been reviewed in Ref [128], which can be used as general guidance to select the proper approach of quantitative analysis based on the image features in different applications.

Perspectives

Epigenetic regulation plays an important role in normal development and pathological processes such as aging and cancer. Although higher-order chromatin structure is an important form of epigenetic information, conventional epigenomic biochemical assays limit direct visualization of such structure at the molecular scale in single cells with a well-preserved spatial context. With the advances of super-resolution microscopy, it is now possible to visualize and quantify the higher-order chromatin structure at the scale of supra-nucleosomes. Spatial epigenomics integrates the structural information of chromatin and genomic organization in its spatial context. There is a growing interest to integrate higher-order chromatin structure with conventional multiscale analysis of the epigenome to improve our understanding of functional implication of epigenome in normal development and various disease processes. As epigenomic function can be dynamically regulated by environmental stimuli [129, 130], such as injury, diet and metabolism, we speculate that super-resolution imaging will further reveal distinct changes in higher-order chromatin structure that may provide novel insights into the epigenetic regulation or serve as an indicator of environmental exposure.

Although super-resolution microscopy reveals the previously unobserved higher-order chromatin structure, the functional implication of the distinct structures and the crosstalk between epigenetic modifications and other functional proteins at the nucleosome level remain to be elucidated. Most super-resolution imaging studies of higher-order chromatin structure focus on basic understanding of biological processes, but few studies have directly interrogated the functional role of higher-order chromatin structure in various diseases such as cancer. Super-resolution imaging of dysregulated higher-order chromatin structure in human diseases may improve our understanding of how disease driver genes alter local and genome-wide chromatin structure and how disrupted chromatin structure in turn contribute to the genome function. Another area of great interest is the impact of epidrugs that target either the epigenome or an enzyme with epigenetic activities [131] on higher-order chromatin structure. Although it is well accepted that many epidrugs (e.g., histone deacetylase inhibitors (HDACi)) “open up” higher-order chromatin structure to activate gene expression, their impact on the local and global higher-order chromatin structure is not routinely studied, nor the functional implication of altered higher-order chromatin structure in vivo, due to the limited information provided by conventional microscopy. Now with super-resolution microscopy, the alteration of higher-order chromatin structure as a response to epidrugs may provide new perspectives to understand their therapeutic effect or serve as new markers for drug screening. Finally, as epigenetic markers have been used as valuable diagnostic and prognostic tools in cancer and microscopic nuclear architecture has long been used for clinical diagnosis, whether higher-order chromatin structure can eventually be used to address the limitation of current clinical diagnosis and prognosis such as distinguishing aggressive from indolent pre-cancerous lesions or predicting disease progression risk remains to be revealed.

Abbreviation:

ChIP

chromatin immunoprecipitation

3C

chromosome conformation capture

ATAC-seq

Assay for Transposase-Accessible Chromatin using sequencing

ChIA-PET

chromatin interaction analysis by paired-end tag sequencing

STORM

stochastic optical reconstruction microscopy

PALM

photoactivatable localization microscopy

SMLM

single-molecule localization microscopy

SIM

structured illumination microscopy

STED

stimulated emission depletion microscopy

SOFI

super-resolution optical fluctuation imaging

SRRF

super-resolution radial fluctuation

BALM

binding-activated localization microscopy

(PAINT)

point accumulation for imaging in nanoscale topography

RNA-TRAP

RNA tagging and recovery of associated proteins

GAM

Genome architecture mapping

SPRITE

split-pool recognition of interactions by tag extension

ChIA-Drop

multiplex chromatin-interaction analysis via droplet-based and barcode-linked sequencing