Mechanistic drivers of chromatin organization into compartments

Abstract

The human genome is not just a simple string of DNA, it is a complex and dynamic entity intricately folded within the cell’s nucleus. This three-dimensional organization of chromatin, the combination of DNA and proteins in the nucleus, is crucial for many biological processes and has been prominently studied for its intricate relationship to gene expression. Indeed, the transcriptional machinery does not operate in isolation but interacts intimately with the folded chromatin structure. Techniques for chromatin conformation capture, including genome-wide sequencing approaches, have revealed key organizational features of chromatin, such as the formation of loops by CCCTC-binding factor (CTCF) and the division of loci into chromatin compartments. While much of the recent research and reviews have focused on CTCF loops, we discuss several new revelations that have emerged concerning chromatin compartments, with a particular focus on what is known about mechanistic drivers of compartmentalization. These insights challenge the traditional views of chromatin organization and reveal the complexity behind the formation and maintenance of chromatin compartments.

Introduction

The development of genome-wide chromatin conformation capture (3C) technologies, such as in situ Hi-C, Micro-C, and HiPore-C, has opened new horizons in our understanding of the intricate world of chromatin interactions [1–5]. These methods have unveiled a wealth of distinct organizational features beyond chromosome territories. Analysis of these contact maps reveals a distinctive plaid-like pattern, namely the general segregation, that is, compartmentalization, of loci in active (A) and inactive (B) states [6,7] (Figure 1a).

Figure 1.

Refining our understanding of compartments by increased resolution.Genome-wide contact maps display plaid-like patterns, representing the segregation of active and inactive chromatin into A and B compartments, respectively. (a) When the interaction signal is binned at 1 Mb, the intervals corresponding to compartments are annotated as multi-megabase features. (b) Recent RCMC and ultra-deeply sequenced Hi-C maps with signal to use 500-bp to 1-kb bins revealed that compartment intervals can exist that are only a couple kilobases in size. Promoters of genes (blue arrows) and active enhancers (green triangles) are overwhelmingly found in these small compartments, and the 5′- and 3′-ends of genes can lie in opposite compartment intervals.

Chromatin contact maps also exhibit conspicuously bright foci that signify robust long-range chromatin loops [4,8–11]. In mammals, these foci, or loops, can primarily be attributed to the concerted actions of CTCF and cohesin-mediated loop extrusion [11–15]. However, not all bright foci in Hi-C and Micro-C maps are due to CTCF looping, as similar foci have been attributed to polycomb in D. melanogaster [8–10], the dosage compensation complex in C. elegans [16–18], as well as chimeric proteins found in cancer cells [19,20].

Another prominent feature of Hi-C maps is ‘triangle’ contact patterns, indicating the interaction of proximal loci to form domains of enriched interactions along a chromosome [21]. These domains, often referred to as topologically associating domains (TADs), are frequently equated with CTCF loops, yet this interaction pattern can also be due to proximal interactions within a compartment interval, making compartment domains [11,22,23]. It is important to note that compartment domains and CTCF loop domains are not mutually exclusive, and compartment domains can even reside inside CTCF loop domains and vice versa [22]. Indeed, a recent study suggests that loop extrusion domains and compartment-like domains can co-exist at the single-molecule level [24].

While it is evident that aberrant genome organization can lead to cellular dysfunction and disease, the precise orchestration of 3D chromatin organization remains enigmatic [25,26]. Chromatin organization is often studied as a regulator of nuclear processes, particularly transcription. However, a growing body of compelling evidence suggests that there is a give-and-take relationship where nuclear processes themselves can impact 3D genome organization [27–30]. Nevertheless, there are several studies that investigate the impact of transcription on 3D genome organization and often have seemingly contradictory conclusions [30–34]. While some past studies reported the relationships between chromatin organization and either transcription or RNA polymerase, many others reported results that showed little-to-no impact after disruption of transcription. Interestingly, more recent studies used Micro-C after transcription inhibition or used auxin-mediated degradation of RNAPII or Mediator, and identified a strong impact, particularly on enhancer–promoter contacts [35–37]. When considering the field, it is imperative to keep in mind its infancy and that we are only on the cusp of understanding how genome-wide 2D contact maps relate to complex nuclear processes and structures. To highlight recent developments and explore current questions, we first explore the challenges imposed by resolution limitations and how they have potentially impacted our understanding of 3D genome organization and chromatin compartments. We then discuss recent work that challenges the previously held notions of genome organization and evaluate several potential drivers of genome compartmentalization.

Detection resolution limits our understanding of chromatin organization

The technological advancements and explosion of new findings in genome organization have led to a new model of chromatin organization. The previously widely embraced ‘hierarchical model of chromatin organization’ delineated chromatin into distinct hierarchical levels and placed compartments, domains, and loops at different sized scales [38]. In this hierarchical framework, compartments represent the largest organizational units, such that multiple domains can reside within a single-compartment interval. These domains, often called TADs, were then composed of multiple loops, though the terms ‘TAD’ and ‘CTCF loop domain’ have recently been used interchangeably. This model garnered attention and support because it neatly categorizes features and extends nicely off previous models of DNA and chromatin packaging. However, recent work has challenged the applicability of the hierarchical model when examined at high resolution [22,33,37]. This reevaluation stems from the historical limitations imposed by low sequencing coverage, which has been a significant constraint for the resolution of contact maps [6]. For example, many Hi-C maps are analyzed using binned data from 40 kilobases (kb) to 1 megabase (Mb) for compartment detection (Figure 1a) [4,7,39]. Recent work indicates that these detection resolution limits have skewed our understanding of compartments [22,37].

Recently generated Region Capture Micro-C (RCMC) maps obtained exceptionally high coverage for specific loci, revealing the existence of microcompartments [37]. This feature corresponds to a plaid-like pattern of interactions between small genomic intervals, only a few kb in size. At the same time, a genome-wide Hi-C map with 33 billion usable contacts allowed annotation of compartments at 500 bp, which found that compartment intervals are sometimes as small as 2 kb [22]. These studies reveal that the segregation of genomes by compartmentalization happens at an exceptionally fine-scale (Figure 1b). Indeed, the 5′- and 3′-ends of individual genes can even localize to distinct compartments (Figure 1b) [22].

Compartment annotation often relies on the calculation of the eigenvector on the interaction pattern spanning the entire chromosome, and the low signal at long distances has made it difficult to annotate compartment intervals at high resolution [6]. For this reason, the analysis of compartments is often performed at a coarser resolution than the analysis of other features. For example, it is relatively common to annotate compartments using 100-kb to 1-Mb bins, while CTCF loops are annotated at < 25-kb resolution [2,4,6,11,29]. Advances in sequencing technology, 3C methodology, and bioinformatic tools now allow the analysis of compartments at resolutions that are similar to what is used for the analysis of CTCF loops [22,37,40]. Doing so revealed that compartment intervals can actually be smaller than CTCF loops [22,37]. Indeed, some CTCF loop domains contain small compartment intervals that segregate loci inside the loop and interact with loci outside the loop [22]. These discoveries directly challenge the previously popular hierarchical model and exemplify our rapidly evolving understanding of governing principles that contribute to 3D chromatin organization.

Drivers and modulators of chromatin compartments

Surprisingly, little is known about the mechanisms that govern the segregation of loci into chromatin compartments. Indeed, past work suggests a relationship to DNA sequence, nuclear structures, transcription, and epigenetic modifications.

Relationship to DNA sequence

DNA sequence plays a vital role for several aspects of chromatin, including the placement of DNA methylation at CpGs, nucleosome sequence preferences, and the recruitment of transcription factors to their recognition motifs [41,42]. Likewise, specific sequence contexts likely impact the features of 3D organization. Indeed, recent deep-learning models, such as DeepC, Akita, and Orca, use DNA sequence to obtain surprisingly high accuracy at predicting the specific features of 3D chromatin organization, such as CTCF loops, domains, or A- and B-chromatin compartments [43–45]. While these represent advanced machine learning approaches, the relationship between compartments and DNA sequence has long been seen by correlation with guanine-cytosine content [46]. Indeed, recent studies also suggest that a guanine-cytosine (GC) gradient might contribute to chromatin organization, such as the radial configuration of DNA with respect to the nucleus [47,48]. While useful to consider, particularly when using machine learning models to predict the impact of variants [49], the sequence of DNA alone cannot fully explain chromatin organization. Indeed, numerous studies have revealed that compartments are cell-type specific and are responsive to conditions.

Relationship to the nuclear lamina

It is well-known that distinct chromatin states correlate with distinct physical locations within the nucleus. Indeed, proximity to different nuclear features is likely one way of controlling gene expression [50]. Lamina-associated domains (LADs) are genomic intervals that are spatially positioned near the nuclear lamina, many near the periphery. LADs are often associated with heterochromatin, and, as such, correlate with chromatin in the B compartment. GP-seq, which measures the radial position of chromatin, confirmed this general relationship between nuclear position and chromatin compartments [48]. The recently developed method, Lamina-Inducible Methylation and Hi-C (LIMe-Hi-C), which simultaneously measures lamin association, DNA methylation, and chromatin conformation, also found an association between LADs and the B compartment [51]. Interestingly, this study found that not all B-compartment intervals associate strongly with the lamina. Polycomb-repressive chromatin, which they found in both A and B compartments, had weaker lamina association even when annotated as B [51]. Importantly, inhibition of EZH2, a subunit of polycomb, caused these B-compartment intervals to have a higher association with the nuclear lamina. This suggests a model in which polycomb-repressive chromatin is distinct from both LADs and interior euchromatin (Figure 2a).

Figure 2.

Potential drivers of compartments. (a) The nuclear lamina. One model posits that the nuclear lamina (green) and chromatin’s radial configuration could be a driver of compartmentalization. In this model, heterochromatin (blue) is found at the periphery, while euchromatin (orange) is more centrally located. Recent work indicates that Pc-repressive chromatin (purple) has no strong lamina association and may represent a distinct compartment. (b) Transcription. High-resolution mapping of chromatin contacts revealed that active enhancers and transcription start sites (TSSs) lie in the A compartment. Additionally, while the bodies of paused genes lie in the B compartment (left), the bodies of elongating genes lie in the A compartment (right). (c) Chromatin modifications. Chromatin marks, such as histone modifications, may help organize or maintain chromatin compartments. Instead of segregating chromatin into two states, compatibilities between chromatin status may make many distinct but overlapping flavors of compartment identities.

Despite these relationships, the question remains whether lamina association is a mechanistic driver of compartment organization. Considering the polymeric physical properties of chromatin, and given the recent revelations about compartment sizes [22,37], how can a microcompartment interval (e.g. a 2-kb region in the A compartment) be in a vastly distinct physical location from that of neighboring loci in the B compartment? Consider Hutchinson–Gilford progeria syndrome (HGPS), a genetic disorder commonly caused by variants in lamin A, and which displays an overall-disrupted nuclear structure [52]. Altered lamina in HGPS results in changes to chromatin accessibility, DNA methylation, and gene expression, especially at sites that were annotated as LADs [52]. Surprisingly, Hi-C experiments in the fibroblasts of HGPS patients revealed that compartmental organization was highly similar to that of their unaffected parents, albeit with increasing differences in later passages of the cells [53]. Interestingly, these late-passage changes reflected a difference in compartment strength rather than a complete switch in compartment identity [53].

Next, to shed light on this issue, let us consider experiments that examined compartments in cell types that display inverted nuclei [54]. In rod neurons and lamin-B receptor null thymocytes, euchromatin is located at the nuclear periphery with heterochromatin toward the center, which is the inverse of convention [54]. Despite this inverted localization, Hi-C in these cells identified the same compartment intervals as their non-rod and wild-type thymocytes, which have more conventional radial configuration [54]. Last, in further support against lamina as the mechanistic driver, a preprint studying the relative timing of LAD and compartment organization post mitosis suggests that B-compartment formation is an earlier event than localization to the lamina [55]. It is also important to note that, while here we focus on the nuclear lamina, there are several other nuclear features that correlate with compartments and may be important for forming or maintaining these interactions.

Relationship to transcription

Transcription may cause the relocation of genes to the A compartment (Figure 2b), and the interplay between transcription and chromatin organization has been subject to extensive investigation. Indeed, measurements of nascent transcription, such as those obtained from GRO-seq experiments, can semi-accurately predict compartment patterns, highlighting their close relationship [23]. Our recent work also revealed a correlation between paused genes and the segregation of 5′- and 3′-ends into A and B compartments, respectively (Figure 2b) [22]. Additionally, imaging of long genes by RNA and DNA FISH revealed that expressed genes exhibit a spatial organizational change termed ‘transcription loops’ [56]. However, an intriguing and complex dimension of this relationship revolves around the question of whether transcription can impact chromatin organization.

Compartments begin reemerging 30 min after nocodazole release (anaphase/telophase) and continue to progressively develop in early and mid-G1 (1 hour/2 hours post release) [57], which approximately matches the progression of transcription reactivation [58]. However, inhibiting transcription in M/G1-arrested cells did not prevent the development of compartments in G1 [57]. Likewise, contact maps after auxin-mediated depletion of RNAPII still have a pronounced compartment pattern [32,34]. Therefore, how compartments correlate so exceptionally with transcription elongation remains unclear.

Relationship to histone modifications

There is a clear correlation between histone modifications and compartments. Indeed, histone modification data are often used in machine learning models to predict chromatin organization [49,59]. However, whether histone modifications could be drivers of compartmentalization is less clear. In support of this model, recent work examined large swaths of loci that become hyperacetylated in cells that express a BRD4-NUT fusion protein [60]. These hyperacetylated regions formed strong interactions with each other in a compartment-like pattern. Importantly, these interactions were lost upon BRD4-NUT depletion. It is tempting to speculate a role for H3K27ac in driving compartmentalization, particularly considering that high-resolution Hi-C identified distal active enhancers (2 kb in size) able to interact in a compartmental fashion distinct from their local context [22]. H3K27ac can also act as a mitotic bookmark, making this an even more attractive model for rapidly re-establishing the A compartment [61]. However, recent work using p300-inhibition to deplete H3K27ac during mitosis found no changes to compartments after release into G1 [61].

H3K9me3 is found in the B compartment, and several studies have indicated that factors important for H3K9me3, such as HP1, can participate in phase separation proposed to explain chromatin compartments [62–64]. Indeed, Hi-C combined with H3K9me3 HiChIP revealed long-range compartment-like interactions that disappear during pancreatic differentiation, corresponding to a loss of H3K9me3 [65]. Additionally, another study found that targeting H3K9me3 to specific regions using dCas9-KRAB resulted in a switch from the A to B compartment, depending on the genomic context [66]. In contrast to the above results supporting a role for H3K9me3, another study depleted the H3K9 methyltransferase, Setdb1, and found that compartment organization was largely preserved [67]. Interestingly, depletion of HP1 during Drosophila development impacted compartment formation during zygotic genome activation but not when the depletion was done in somatic S2 cells, suggesting a role in the establishment but not maintenance of compartments [68]. This same study also expressed a histone mutant H3K9M, which cannot be methylated, and found that the compartment status was fairly preserved, which suggests that HP1 is more important than H3K9me3 for compartment establishment [68]. Despite these findings, it is important to acknowledge that, for many cell types, H3K9me3 loci comprise only a small subset of B-compartment intervals [69]. Additionally, although annotated as the B compartment, several studies suggest that they form a distinct interaction pattern compared with the rest of the B compartment [4,22,69].

H3K27me3, associated with polycomb-repressive chromatin, may also play a role in compartmentalization, and several groups have identified distinct interaction patterns correlated with H3K27me3. For example, late oocytes display polycomb-associated domains that interact in a compartment-like fashion, revealed by their plaid pattern in Hi-C maps [70]. Similarly, polycomb-associated chromatin can make unique interaction patterns from the rest of the B compartment [51,69–72]. Interestingly, EZH2 inhibitors led to decreased H3K72me3 but not a switch from B compartment to A [51]. Instead, these sites formed a stronger association with the rest of the B compartment [51]. Therefore, H3K27me3 may mark chromatin in a compartment distinct from A or B.

We used three frequently studied histone modifications to highlight the relationship to compartments: each may play a role in establishing or maintaining interactions. Their impacts on genome organization may be via the mark itself or via proteins that deposit/recognize these modifications. It is also important to note that there are many other marks, and that they could potentially contribute to genome organization (Figure 2c).

Conclusion

The work discussed above highlights recent advances that have changed our understanding of chromatin compartments but reveals the remaining lack of clarity. It is becoming increasingly clear that compartments are complex structural features that can segregate loci at exceptionally fine-scale [22,37]. We have also learned that compartments are able to rapidly reform after disruption, for example, after mitosis, and that they can be highly responsive to conditions [34,57,73]. However, we have not yet revealed a single concrete mechanistic driver of compartmentalization. We often discuss compartments from a two-state model, A or B, or as subcompartments that exist within those states [6]. However, compartment interactions are likely more complex than a simple two-state model can describe [51,69,72]. Given this complexity, it is likely that multiple factors cooperate to form long-range chromatin interaction preferences that comprise compartments. Indeed, recent work suggests that euchromatin, H3K9me3 heterochromatin, and H3K27me3 polycomb- repressed chromatin form distinct interaction patterns from each other [51,69,72]. However, even a three-state model likely cannot capture the complexity of the compartment interaction pattern, particularly given the large amount of DNA often labeled as ‘Quiescent’ because it does not contain these marks [74,75]. Instead of a specific number of states, chromatin compartments may come in many flavors based on shared compatibilities (Figure 2c).

Data Availability

No data were used for the research described in the article.