Evolving methodologies and concepts in 4D nucleome research

Abstract

The genome requires tight regulation in space and time to maintain viable cell functions. Advances in our understanding of the 3D genome show a complex hierarchical network of structures, involving compartments, membraneless bodies, topologically associating domains, LADs, protein- or RNA- mediated loops, enhancer - promoter contacts, and accessible chromatin regions, with chromatin state regulation through epigenetic and transcriptional mechanisms. Further technology developments are poised to increase genomic resolution, dissect single-cell behaviors, including in vivo dynamics of genome folding, and provide mechanistic perspectives that identify further 3D genome players by integrating multiomics information. We highlight recent key developments in 4D nucleome methodologies and give a perspective on their future directions.

Introduction

Technological advancements made in the last decade contributed to our understanding of the 4D genome organization. Among the most remarkable achievements are genome-wide mapping of open and closed chromatin and their preferred nuclear location, the discovery of topologically associating domains (TADs), and concepts of loop extrusion mechanisms [1—3]. The intricacies of promoter—enhancer contacts are increasingly appreciated, but their effects on subsequent gene regulation remain only partially understood.

Microscopy is a long-standing approach for studying genome organization from whole chromosomes to single genes combined with RNA or protein detection. Significant advances became possible with fluorescence microscopy first in fixed cells and lately in living cells [4,5]. Fluorescence in situ hybridization (FISH) typically allows simultaneous visualization of a small number of targeted genomic loci with single-cell resolution. FISH made possible the detection of genomic rearrangements, chromosome copy-number variations, and nuclear positioning of genomic regions in relation to nuclear landmarks. Limitations of FISH to the study of a few regions make it difficult to investigate higher order chromatin structures that cover extensive genomic regions. One major technological catalyst for FISH has been the development of Oligopaints, which are small synthesized oligonucleotides complementary to loci of interest [6]. Oligopaint oligos contain barcoded tails, on either side ofthe genomic target region, that can be hybridized sequentially by complementary fluorescently labeled oligos to collect the signal that paints whole genomic stretches. Oligopaints are highly programmable in design and are rapidly diversifying to detect chromatin topology at different scales and relate it to RNA detection [7].

Genomic methods to study chromatin contacts have rapidly expanded with the advent of next-generation sequencing (NGS). DNA adenine methyltransferase identification (DamID) was initially developed to identify DNA loci that interact with specific nuclear proteins and has since then become genome-wide with the inclusion of NGS [8,9]. Another set of technologies that developed with the NGS boom are the chromosome conformation capture (3C) methods [10]. These follow the initial 3C method principles of cross-linking, digesting, and ligating proximal DNA fragments and, additionally, integrate NGS to obtain sequence-based information. Among them, Hi-C is perhaps the most popular, gaining its reputation as a quick benchtop application to obtain genome-wide pair-wise chromatin contacts [11]. Hi-C gave a genome-wide understanding of euchromatin and heterochromatin compartmentalization, described TADs, and explained the dynamics of chromatin architecture throughout the cell cycle [12,13].

To overcome the intrinsic biases of the 3C methods, the innovation in 3D genomics is now expanding with the development of ligation-free techniques [14]. Further to confirming previous findings, ligation-free approaches have started to reveal new aspects of genome biology, such as formation of hubs of superenhancers around speckle domains and of complex multiway genomic interactions between active regions or the nucleolus [15—17].

The development of new technologies in 3D genome research prompts the field toward answering major questions (Figure 1). How does 3D genome topology relate to the mechanisms and specificity of gene expression? What are the determinants of the chromosome structure? Is the higher order chromatin structure cell-type specific and is it controlled during cellular programming? Which genes are refractory or controlled by 3D topology? How does genome organization change in single cells or in different alleles in response to changes in cell state? Can understanding genome topology help us learn about mechanisms of disease?

Figure 1. Challenges in 4D genome research.

Schematic representation of major features of the interphase genome organization that have been described so far. These include genomic features such as compartments, topologically associating domains (TADs), complex regulatory interactions, and how the genome is organized around nuclear bodies and the lamina. Further technological developments will potentially solve major challenges that remain the 4D genome (highlighted in pink boxes).

Here, we discuss the latest 4D nucleome technology developments that are providing significant insights into the functioning genome, with a main focus on genomic techniques. We also comment on future directions of current technologies to answer the next questions and to understand further the 4D genome.

Recent developments

Increasing genomic resolution

Multiple methods have reported to improve Hi-C resolution by increasing throughput, DNA recovery, and changing the restriction digestion step [18,19]. These improvements have facilitated research looking at close-range promoter—enhancer interactions and have pinpointed the regulatory regions responsible for a phenotypic interaction [20].

Micro-C, a derivative of Hi-C, has reported the highest resolution to date [21]. It was recently adapted to mammalian cells and can achieve 100 bps in resolution by using the MNase enzyme to fragment DNA down to the nucleosome level [22]. Micro-C provides orthogonal information on nucleosome positioning and can localize boundaries between contact domains precisely to nucleosome-depleted regions. Micro-C also detects a ‘two-start’ zigzag 30-nm chromatin fiber model, that contrasts with a lack of an ordered 30-nm fiber found with imaging techniques, such as cryoEMT, that preserve nuclear integrity and keep cells in isotonic conditions [23].

The push for single-cell Hi-C

Major efforts are ongoing to study cell variability of chromatin architecture. Single-cell Hi-C (scHi-C) shows that cis and trans contacts are highly variable between individual cells, and it also identifies cells at specific cell cycle stages [24,25]. A recent development in scHi-C, Dip-C, adopts an improved DNA recovery method termed ‘multiplex end-tagging amplification’ [26]. Dip-C analysis at 20-kb resolution on a cell line with SNP variability between alleles has enabled analysis of chromatin conformation of individual parental alleles and subsequently described significant heterogeneity at imprinted and regular autosomal regions. Another recent adaptation of scHi-C incorporates simultaneous methylated cytosine detection [27]. This method gives an extra layer of information on how cell- type—specific methylation profiles relate to fundamental genomic structures, finding that CTCF sites are increasingly hypomethylated as chromatin contact frequencies increase. Other promising developments rely on the use of computational approaches to remedy the sparsity of scHi-C data, a major current limitation [28].

Under the lens

An alternative platform that has the potential to offer a full depiction of chromatin organization in single cells is the use of Oligopaint technology in FISH. Oligopaints were recently used to reconstruct a 3D depiction of a 700-kb locus at sufficient depth to investigate TAD formation, insulation, and function in single cells [29]. A significant variability between cells was discovered at the level of TADs, and the TAD boundaries considered had a preference for CTCF sites, with TAD structures still present in CTCF-depleted cells. Although currently limited to investigating only a few regions simultaneously, Oligopaint techniques are rapidly becoming more affordable and have higher throughput. Recently, a scalable Oligopaint strategy was used to tile the whole genome of Caenorhabditis elegans down to 500-kb resolution [30].

The emergence of ligation-free genomic techniques

Despite rapid improvements in the 3C technologies, several intrinsic biases such as GC content, protein occupancy, and restriction site density remain a concern, although some are currently thought to be partially resolved via computational methods [14,31]. In addition, 3C technologies cannot report on other parameters of chromosome organization, such as the distance of genomic regions to the nuclear periphery or nuclear domains, or chromatin compaction, which are critical features of gene regulation.

A major limitation in 3C methods is the ligation step that selects for contacts that are within a limited 3D space and inevitably dilutes complex multiway long- range interactions [32]. The difference becomes more critical when considering that chromatin organizes itself around nuclear bodies with large dimensions, such as in the order of 0.5—2 pm for splicing speckles, which may be beyond the ligation distance [33]. This limitation may explain why proximity 3C methods miss interactions between chromosomes.

Several genome-wide methods that rely on NGS have been developed to overcome the need for ligation. These include genome architecture mapping (GAM) [15], split-pool recognition of interactions by tag extension (SPRITE) [16], and chromatin interaction analysis drop [17]. GAM relies on ultrathin cryo- sectioning of fixed single cells, followed by DNA sequencing of individual nuclear slices, and derives physical proximity between genomic loci from how frequently they cosegregate in the same nuclear slice. SPRITE and chromatin interaction analysis drop are based on barcoding fragments of fixed chromatin and calculating the proximity of DNA fragments by determining the frequency at which DNA is detected in the same chromatin—protein complex. These technologies capture not only known features of chromatin organization but also additional layers of information that are missed by the 3C methods. For example, GAM can report on radial positioning and the compaction of a locus, and SPRITE can retrieve simultaneously chromatin-bound RNA [15,16]. Interestingly, all these techniques report multiway chromatin contacts that span tens of megabases and suggest a potential functional role of these contacts, in line with previous microscopy observations.

Other ligation-free technologies have been reported to study specific aspects of chromatin organization. For example, tyramide signal amplification sequencing (TSA-Seq) can measure the relative cytological distance between the nuclear feature of interest and chromatin [34]. TSA-Seq uses horseradish peroxidase conjugated to an antibody targeting the nuclear feature of interest, which catalyzes biotin-conjugated tyramide-free radicals that bind DNA. All genomic DNA is first isolated, followed by a biotinylated pull down to avoid chromatin- related pull-down biases; the biotin-enriched DNA is then sequenced. Quantifying DNA positioning relative to nuclear features is vital for investigating the role of phase-separated bodies or/and lamina in controlling gene expression. Together with contact mapping methods, TSA-Seq will become even more essential when studying nondividing cell types, such as neutrophils and terminal neuronal cells, in which nuclear conformation and bodies take on more diverse and distinct patterns.

DamID technology is becoming increasingly versatile and an ideal tool for in vivo investigation of a feature of interest, with the caveat that it requires a Dam-fusion transgenic system [9]. Impressively, orthogonal DamID derivatives are being developed to integrate transcription and lamina/Polycomb targets at the single-cell level [35]. A recent adaptation to DamID technology is DamC, which provides in vivo 4C-like data by first inserting Tet operator sites at the genomic region of interest and expressing Dam-used with the reverse tetracycline receptor, which binds to the Tet operator sites and methylates genomic regions that come into spatial proximity. DamC has validated the presence of TADs and the occurrence of their boundaries at CTCF sites [36].

Retrieving orthogonal information

One of the main challenges in the field is to develop assays that can integrate orthogonal information to help understand the interplay between chromatin topology and regulation of gene expression. Optical reconstruction of chromatin architecture, a derivative of Oligopaint technology, can capture 3D topology coupled with RNA expression [7]. In the first study in the fly embryo, optical reconstruction of chromatin architecture was used to recreate 3D depictions of 100 to 700-kb regions at 10- to 20-kb resolution and coupled this with expression of thirty RNA species. This simultaneous detection allowed the investigation of cell-type—specific structures identified by cell-type—specific RNA species. Interestingly, at the time of active transcription, deter-mined by the presence of nascent RNA, promoters’ proximity to enhancers was present but weak, and surprisingly, many inactive promoters were in proximity to known enhancers. Separately imaging individually 700-kb regions or thirty RNA species is low throughput for modern standards; however, their simultaneous detection yields powerful insights, and further developments are likely to scale up the genomic lengths that can be covered in a single experiment.

Several genomic techniques have also integrated tran- scriptomic information. In the first study, SPRITE was used to detect chromatin-bound RNA and validate the location of A-compartments at splicing speckles, via the presence of Malatl, and of B-compartments at the nucleolus via rRNAs [16]. SPRITE is an excellent technique to study the preference of genomic loci for multiple nuclear landmarks owing to its orthogonal recovery of chromatin-bound RNA. Single-cell developments in SPRITE are also highly anticipated.

Future developments

Advancing single-cell understanding through orthogonality

Currently, we are only able to sample a small fraction of the genomic architecture and the transcriptome using Oligopaint techniques. There is much to be understood about the direct relationship of higher order structures such as compartments, TAD insulation, and the action of distal enhancers with gene expression. Explaining these relationships is nontrivial for current techniques, without the direct retrieval of both RNA expression and genomic contacts in the same cell.

Few studies using in vivo systems highlight that there is still much to learn about cells in their native states [37]. Tissues are inherently complex and heterogeneous. Therefore, new in vivo techniques will require single-cell technologies and orthogonal dimensions to describe cell types and cell states, be it via epigenetic, protein, and/or RNA profiles. Further optimization of in vivo techniques will also strengthen our ability to study clinical patient samples and improve our understanding of 4D genomics in the context of disease. Therefore, genome-wide single-cell—based approaches that can recover both contact and transcriptomic information are primed to give significant progress, such as understanding the direct relationship between regulatory regions and gene expression and allowing for easier application of in vivo studies and patient samples (Figure 1).

Epigenetic and a protein-centric view

Exploring the interplay between chromatin-bound proteins and the 3D topology is a large area for technology to expand into. Recent developments with the flexible and efficient CUT&Tag-fused protein, Tn5-ProteinA, in combination with current assays may aid these efforts [38]. One attempt could be to merge ATAC-see [39] and CUT&Tag principles to incorporate fluorescently tagged DNA into genomic sites occupied by a protein or epigenetic mark and further combine this with Oligo- paint technology. This approach could determine the effect of different proteins, for example, CTCF or Polycomb complexes, on the state of chromatin architecture at the single-cell level.

A growing number of publications have focused on transcription factors (TFs) and their role in genome architecture (reviewed in the study by Kim and Shen- dure [40]). TFs can form hubs/condensates within the nucleus owing to their interacting low-complexity domains. The influence of TF concentration, binding site affinity, and their ability to open chromatin on chromosome architecture has been studied both experimentally [40,41] and by theoretical polymer modeling [42]. Recent studies also show that enhancer regions come together in hubs [15], in conjunction with TFs and RNA polymerase II [43]. In the future, techniques that can recover both genome-wide topology and TF binding/concentration within the nucleus will help clarify the role of TFs in genome organization (Figure 1).

Validating proposed mechanisms

Various mechanisms are currently proposed to enact enhancer—promoter function. A more complex picture is emerging beyond the canonical model in which gene expression is driven by specific enhancer—promoter proximity contacts [44,45]. New models of proximity between multiple enhancers and between different promoters are being investigated. The relationship between larger hierarchical genome structures and nuclear landmarks is also being mapped in greater detail, although their direct functional value with reference to the cell state and type remains poorly understood.

Major advancements are also being made to develop technologies that can investigate the function of chromatin organization. Current functional studies are mostly carried out using bulk data and focus on few interactions, but more systematic approaches are required to investigate the more complex picture of the genome topology that is currently emerging, such as multiway interaction hubs. CRISPR-based editing methods have been successfully used to modify multiple genomic regions at once and could be used to screen for functional and redundant candidate contacts [46]. Techniques to force DNA looping or nuclear repositioning are also being developed but remain to be expanded for larger screenings [47—49]. Integrating functional studies with single-cell orthogonal information, especially in complex heterogeneous tissues such as the heart or the brain, will give insights into genome regulation and could help our understanding of complex disease (Figure 1). One remaining challenge is to further develop the high-throughput potential of genome-wide genome architecture techniques to enable large functional screening.

Concluding remarks

There has been a boost in developing technologies to study the 4D genome. Multiomics techniques are continuously improved to give us a more complete and higher resolution picture of chromosome architecture, transcription, and protein interplay. Although in their incipient stages, single-cell technologies have already brought us invaluable information on the mechanisms of gene regulation. Developing single-cell chromatin architecture assays to include transcriptomic and proteo- mic information will be a strong approach for gaining scientific insights into how cells function. Currently proposed models of the 4D genome need to be validated for their mechanistic properties, and rapid developments in the genome engineering will aid these efforts.