Mechanisms and functions of chromosome compartmentalization

Abstract

Active and inactive chromatin are spatially separated in the nucleus. In Hi-C data, this is reflected by the formation of compartments, whose interactions form a characteristic checkerboard pattern in chromatin interaction maps. Only recently have the mechanisms that drive this separation come into view. We discuss new insights into these mechanisms and possible functions in genome regulation. Compartmentalization can be understood as a microphase segregated block co-polymer. Microphase separation can be facilitated by chromatin factors that associate with compartment domains, and that can engage in liquid-liquid phase separation to form subnuclear bodies, as well as by acting as bridging factors between polymer sections. We then discuss how a spatially segregated state of the genome can contribute to gene regulation, and highlight experimental challenges for testing these structure function relationships.

Chromatin compartmentalization is an organizing principle of the nucleus

Within each eukaryotic nucleus, DNA is intricately folded. Chromosome organization is important both for resolving the physical problem of how to fit very long DNA molecules into a nucleus with a much smaller diameter, and to ensure that the DNA within the nucleus is correctly regulated [1–3]. There are two major types of chromatin which are spatially separated within the nucleus, euchromatin (see Glossary) and heterochromatin [1–3] (Figure 1). The different types of chromatin were first described by differential DNA staining in early microscopy experiments, and have since been extensively studied [2, 4]. Euchromatin is characterized by the presence of active genes, wider spacing between nucleosomes, higher accessibility, and histone marks such as H3K4me3, H3K27ac, H4K8ac, and H4K16ac, as well as histone variants H2A.Z and H3.3 [5, 6]. In contrast, heterochromatin is transcriptionally inactive, less accessible, and is decorated with H3K9me3 and H3K27me3 histone marks [5, 6]. Microscopy studies have also revealed that heterochromatin and euchromatin are spatially segregated, with heterochromatin mainly localized at the nuclear periphery and the region surrounding the nucleoli, while euchromatin is positioned in the interior of the nucleus [2] (Figure 1C).

Figure 1:

Active and inactive genomic regions form microphase separated compartments in the eukaryotic nucleus.

A. Schematic of a eukaryotic cell nucleus showing different chromosomes in different colors to illustrate chromosome territories.

B. Each chromosome is made up of both euchromatin or active (A, red) and heterochromatin or inactive (B, blue) chromatin which spatially cluster to form separate compartments within the nucleus.

C. Schematic of subnuclear localization of A and B compartments. B compartments (blue) are spatially localized to the nuclear periphery and surrounding the nucleoli, while A compartments (red) are more interior.

D. A and B compartments are visible in Hi-C matrices by a characteristic ‘checkerboard’ pattern (bottom left triangle), which, when subject to eigenvector decomposition, consist of alternating A and B compartments (top right triangle).

Genome-wide analysis of epigenetic marks using a wide range of methods including ChIP-Seq, CUT&RUN, ATAC-seq, DNAse-seq, and others have revealed that regions of euchromatin and heterochromatin alternate along the length of chromosomes [7–10]. Analysis of chromosome conformation using methods such as Hi-C, SPRITE, TSA-Seq, HiChIP, and GAM show that these domains are spatially segregated [11–16]. In a Hi-C pair-wise interaction heatmap, active and inactive regions of the genome exhibit a ‘plaid’ or ‘checkerboard’ pattern (Figure 1D), showing that not only do individual active and inactive regions remain spatially separated, but, in addition, each type of chromatin interacts with distal regions of the same type [11]. Eigenvector deconvolution analysis of Hi-C data can be used to identify regions that correspond to each type of chromatin, which are termed the ‘A’ (active) and ‘B’ (inactive) compartments, and which correspond to euchromatin and heterochromatin, respectively [11]. In addition, these nuclear compartments are correlated with replication timing, with A compartments replicating earlier than B compartments [17]. Further studies have shown that within A and B compartments, there are also sub-compartments corresponding to specific combinations of epigenetic marks [18].

While the presence of active and inactive nuclear compartments is now well established, the mechanisms that drive this organization and the functional relevance of spatial clustering of chromatin domains, e.g. for gene regulation, have been open questions in the field, and are currently the focus of intense study. In the past 2.5 years, the field has shifted dramatically due to publications showing that phase separation may be the mechanism that separates heterochromatin from euchromatin, and follow-up studies have shown that this mechanism may be more broadly applicable to other types of chromatin [19–21]. However, in the past year, new publications and a recent review on the topic argue that phase separation may not be the only explanation for clustering of genomic loci in vivo, and that other mechanisms may also be important [22, 23]. It is also clear that while compartments are correlated with gene regulation, the specific function of long-range compartment interactions is still not completely understood. This review will cover these recent advances on the topics of the mechanisms of chromosome compartmentalization, what is known about the function of long-range compartment clustering and how this may be tested in future studies, and the interplay between compartments and other levels of chromatin organization.

Compartments are proposed to be formed by microphase separation

The biophysical processes establishing and maintaining the long-range A to A and B to B interactions that define the active and inactive nuclear compartments are starting to be explored in molecular detail [4, 19, 20, 24–27]. Recent observations and analyses strongly suggest that one process contributing to compartmentalization is phase separation. The physical principles by which long polymers that are composed of alternating blocks of monomers of different type (such as A and Bcompartments along chromosomes) can phase separate are well known [28–30]. This type of polymer is termed a block copolymer, and can fold so that blocks of each type cluster together while displaying few interactions with blocks of the other type (Figure 2A). Importantly, such copolymers display microphase separation, and not macrophase separation. This means that the spatial clusters are relatively small and that the two types of blocks do not entirely separate in just two large domains. This is due to the covalent linkages between blocks of different types that prevent their complete macroscopic segregation. The ability to microphase separate is dependent on the length of the blocks (size of chromatin domains) and the strength of the preference of domains to interact with other domains of the same type [28–30].

Figure 2:

Mechanisms of compartmentalization.

A. The chromatin fiber is a block co-polymer consisting of alternating active (A, red) and inactive (B, blue) blocks (domains).

B. Each type of domain recruits separate multivalent binding factors, as depicted by fuzzy interactions.

C. Chromatin domains with similar marks and binding proteins localize together; this is proposed to be through a microphase separation mechanism. The binding factors may act as bridging factors mediating PPPS or participate in LLPS, depending on the context.

D. Euchromatic regions consist of the A1 (dark red) and A2 (lighter red) sub-compartments [18], have low chromatin density, and are bound by Pol II (red major sectors) and binding factors such as Bromodomain containing proteins (orange circles). A1 domains differ from A2 domains in that A1 domains are in close proximity to Nuclear Speckles (large red circle), while A2 is active chromatin more distant from Nuclear Speckles. Heterochromatin is split into the B1 (light blue), B2 (medium blue), and B3 (dark blue) sub-compartments, and has a higher DNA density than the A compartment. The B1 sub-compartment is bound by the Polycomb complex (light blue rounded triangle) while the B2 and B3 sub-compartments contain HP1 (dark blue bean shape). B2 is often near the nucleoli (orange semicircle), consisting of nucleolar associating domains (NADs), while B3 is frequently close to the nuclear lamina (green lines), and consists of lamina associating domains (LADs).

Microphase separation of DNA in the nucleus has been proposed to occur through two related mechanisms that differ in the properties of the factors that mediate interactions between chromatin fibers: polymer-polymer phase separation (PPPS) and liquid-liquid phase separation (LLPS) (reviewed in [21]). Briefly, PPPS occurs when bridging factors bind to a polymer, i.e. the chromatin fiber, such that each bridging factor crosslinks two polymer blocks in the same state, i.e. A to A, but does not necessarily interact with other bridging factors [21, 31]. In LLPS, multivalent interactions between factors such as proteins or nucleic acids form macro-molecular structures with liquid-like properties [21]. LLPS can occur in any part of the cell, and does not necessarily require the involvement of a polymer; however, in the context of chromosomes, the interactions can occur between the DNA polymer and binding factors, as well as between the binding factors themselves, mediating long-range genomic interactions [21]. A key distinction between these models is that in LLPS, the binding factors can form droplets even in the absence of the polymer, which is not the case for bridging factors involved in PPPS [21]. However, in the context of the chromatin polymer, it can be difficult to distinguish between these two mechanisms; some binding factors may act as either PPPS bridging factors or LLPS mediating factors depending on circumstances, such as local concentration or post-translational modifications of the binding factor. In addition, the chromatin polymer itself may contribute to long-range interactions due to different types of histone modifications that can self-associate [31, 32]. A third mechanism that could lead to microphase separation in the nucleus is liquid-solid phase separation (LSPS), where, in the context of a chromosome, the factors decorating one chromatin state would cause it to behave like a liquid, while those bound to the other would have the properties of a solid, thereby causing spatial separation [33, 34]. In vivo, it is likely that some combination of all three of these mechanisms is occurring, leading to the compartmentalized cell nucleus.

Recent chromosome compartment studies have focused mainly on LLPS as a mechanism of chromatin compartment formation, due to the discovery that some chromatin binding proteins can participate in LLPS or formation of condensates in vitro [19, 20, 26, 27]. This presents an attractive model for segregation of active and inactive chromatin within the nucleus, as self-association of these proteins and local concentration gradients may be able to drive separation of A and B compartments (Figure 2B–C). Nucleolar assembly is also proposed to occur through LLPS, and may represent a possible third compartment of highly repetitive DNA that forms a distinct nuclear body from the A and B compartments, but is difficult to study by high-throughput sequencing due to the repetitive nature of the rDNA sequence [35, 36].

LLPS occurs when both phases have liquid-like properties, such as high mobility of the individual molecules and can occur through low affinity, high valency interactions between disordered regions of proteins, such as a domain of a chromatin binding factor [19, 20, 26, 27]. When the local concentration of the protein with an intrinsically disordered region (IDR) is above a certain threshold, such as when bound to a region of chromatin with a particular epigenetic mark, this can lead to spontaneous aggregation of both the IDR and the binding partners [19, 20, 26, 27]. The resulting condensates can have different physical properties, depending on the components. It should be noted, however, that while there is increasing evidence for LLPS in chromosome compartmentalization, concerns have also been raised that other mechanisms may explain some of the observations attributed to LLPS, and that it will be important to establish robust tests for LLPS in vivo to rule out alternative mechanisms [22, 23]. For instance, it is possible that factors that by themselves can form liquid condensates, actually mediate chromatin compartmentalization through acting as bridging factors (see below).

Evidence for LLPS in the A and B compartments

LLPS in the B compartment is proposed to happen in two different ways. In one part of this model, LLPS on H3K9me3 marked chromatin is mediated through the heterochromatin protein HP1 (Figure 2D). The a/alpha isoform of HP1 has been shown to form condensates alone and with polynucleosomes in vitro, and in Drosophila melanogaster embryos in vivo [19, 20]. It has also been shown that binding of the Schizosaccharomyces pombe HP1 protein Swi6 deforms the nucleosome core octamer in vitro, resulting in increased solvent accessibility, but also increased nucleosome concentration within the condensates [37]. Phase separated compartments formed with a different HP1 isoform, HP1 beta, its binding partner SUV39H1, and H3K9me3 modified nucleosomes exclude active chromatin proteins in vitro [38]. In addition, ectopic targeting of HP1 alpha to the lac operator results in an increased density and condensation of the targeted chromatin [39]. However, while HP1 alpha is capable of forming liquid droplets, it is possible that in vivo HP1 alpha can also act directly as a bridging factor to establish and/or maintain heterochromatic microphases. Similarly, on H3K27me3 marked chromatin, LLPS is proposed to be mediated by the Polycomb complex, which can form droplets in vitro [40] (Figure 2D). Mutations in CBX2, a subunit of the Polycomb complex, can disrupt Polycomb puncta in cells [40–42]. Recently, short stretches of chromatin alone have also been shown to be able to form droplets via a nucleosome intrinsic mechanism using an in vitro system of polynucleosome arrays. It was shown that formation of these phase separated bodies in vitro required unacetylated histone tails [32]. In these experiments, droplet formation requires the linker histone H1 when linkers between nucleosomes are long [32].

While LLPS has been more extensively studied in the context of heterochromatin, it is important to recognize that the A compartment is not just a region that is excluded from the heterochromatic microphase. In fact, A compartments likely form their own euchromatic microphase separated bodies mediated by unstructured regions on certain transcriptional regulators such as BRD4, TAF15, and FUS, as well as RNAs and RNA binding proteins [27, 32, 43]. For example, in Amyotrophic Lateral Sclerosis (ALS), mutations in FUS can lead to aggregation, which may represent a LSPS transition [34]. On in vitro polynucleosome templates, acetylation disrupts the chromatin intrinsic LLPS, but microphase separation can be rescued on acetylated nucleosomes with the addition of multibromodomain containing proteins which can bind to the acetylated histone tails [32]. Interestingly, when added to the same reaction, both the unmodified and acetylated chromatin with bromodomain proteins formed droplets, but each droplet only contained one of the types of chromatin and did not mix with the other type [32]. RNA Polymerase II may also contribute to phase separation in active regions such as super-enhancers via the intrinsically disordered Pol II C-terminal domain (CTD) [44]. The phosphorylation state of the CTD determines the type of condensate, with hypo-phosphorylated Pol II CTD included in Mediator containing condensates that function in transcription initiation and hyper-phosphorylated Pol II CTD interacting instead with splicing factors in separate condensates [44, 45].

Nuclear speckles are regions of the nucleus characterized by the presence of splicing factors and other mRNA processing proteins, that tend to form a cluster of active genes [46]. Depletion of the nuclear speckle structural protein Srrm2 resulted in decreased A compartment strength and increased B compartment strength by Hi-C [47] (Figure 2D). Artificial LLPS droplets made using the CasDrop system with transcriptional regulators TAF15 and FUS exclude heterochromatin and HP1 alpha, and have lower histone density than surrounding chromatin regions [27]. These measurements of chromatin density at nuclear regions impacted by different types of LLPS are consistent with measurements of chromatin density and compaction within A and B compartments by super-resolution microscopy [48–50]. Overall, the trend is that B compartments or HP1 alpha LLPS droplets have increased chromatin density compared to A compartments or transcription factor mediated LLPS droplets [39, 48–50] (Figure 2D).

Sub-compartments suggest multiple types of microphase separation

Recent work using high-resolution Hi-C datasets, as well as liquid-chromatin Hi-C (LC-Hi-C), has shown that there are multiple sub-compartments within both active and inactive chromatin [18, 51]. The general principle appears to be that chromatin with similar types of histone modifications and chromatin binding proteins tends to self-interact, but that within heterochromatin or euchromatin, there are distinct types of each kind of chromatin that are distinguishable by the stability of the interactions and by long-range clustering [18, 51]. These sub-compartments are proposed to correlate with nuclear regions observed by microscopy such as nuclear speckles/transcription factories, lamin associating domains (LADs), or nucleolus associating domains (NADs) [51]. An argument can be made that the 6 sub-compartments identified by Hi-C or the 9–25 chromatin states identified by ChromHMM or similar tools may actually correspond to loci associating with distinct demixed liquid or solid phases within the nucleus [18, 33, 52–56] (Figure 2D). LLPS has been discussed above, but LSPS is also implicated in processes such as nucleolar stress response aggresome formation [33]. In addition, some types of sub-compartments may occur due to PPPS rather than LLPS if the binding factors do not form droplets on their own [31]. These phase separated bodies and the chromatin associated with them display different physical properties such as the dynamics of factor exchange and the stability of chromatin interactions. The biochemical composition of these phase separated bodies is now being explored, but which factors are key for driving the separation in each type of chromatin is in most cases not known [53, 54].

Computational models of chromosome folding

Computational polymer models have been useful for understanding the forces determining compartmentalization, and provide a theoretical framework to explore mechanisms and properties of spatial segregation of domains on long polymers such as chromosomes [11, 21, 57–64]. It is important to note that the models discussed here model physical microphase separation of a block copolymer regardless of the precise molecular mechanism, which in vivo could be occurring by multiple mechanisms, such as bridging factors and PPPS, condensate formation and LLPS or LSPS, or intrinsic attractive forces between specific types of chromatin due to specific histone marks or variants [32]. Modeling chromosomes as block copolymers that microphase separate through interactions between blocks of the same type recapitulates the characteristic checkerboard pattern of long-range interactions observed in Hi-C maps. [60, 65]. Further, by comparing coarse grained polymer models of microphase separation to experimental results, the relative strengths of the attractive forces between different compartments can be estimated.

A recent study used comparisons between a polymer model and in vivo experiments using both immunofluorescence and Hi-C analysis in mouse rod cells that display an inverted nucleus architecture [66]. In these cells, all heterochromatin is located at the center of the nucleus, unlike in conventional nuclei where heterochromatin lines the periphery and the nucleolus. In both inverted and conventional nuclei, Hi-C shows similar checkerboard patterns, indicating spatial separation of active and inactive chromatin. Modeling shows that an inverted nucleus organization can only be formed when interactions between centromeric chromatin are the strongest, while interactions between inactive loci along chromosome arms are somewhat weaker, and interactions between active chromatin are very weak. In order for a conventional nucleus to position heterochromatin at the periphery, relatively stable tethers between heterochromatin and the nuclear lamina had to be included in the model. These models start to explore quantitatively the relative forces mediating subnuclear compartment formation, which, in the future, can be compared to forces that can be exerted through the different microphase separation mechanisms [66]. In addition, the models are also useful for investigating chromatin positioning with respect to other nuclear structures such as the lamina [66].

What is the functional relevance of chromosome compartmentalization?

One important question that has been difficult to address is whether the spatial segregation of euchromatin and heterochromatin is important for nuclear function. To test whether compartmentalization per se contributes to silencing or activating genes, one would need to disrupt long-range clustering of loci while maintaining their histone modifications and local chromatin structure. And vice versa, such functional studies would require perturbation of local chromatin modifications while maintaining compartmentalized clustering to determine if gene activity is changed. Such experiments have proven to be very difficult given that local chromatin features appear to be directly involved in compartmentalization. Further, an additional reason that it has been difficult to test the importance of higher-order structure on nuclear function is that, in many cases, there are feedback mechanisms between the histone modifications and their binding proteins, so perturbing the binding protein can then affect the level of the histone modification itself. For example, in the case of HP1 alpha and H3K9me3, HP1 alpha binding recruits histone methyl transferases, leading to further H3K9 methylation [39]. However, this also points to the possible importance of the formation of segregated microphases and higher-order chromatin structure in spreading or maintaining chromatin marks as a positive feedback mechanism could then lead to enhanced repression or activation of the targeted regions if higher levels of the binding protein was recruited to a cluster of A or B regions within the nucleus.

Despite these intricate interrelationships between local chromatin state and compartmentalization, some recent perturbation studies have started to shed some light on the role of factors on chromosome compartmentalization. Focusing first on the heterochromatic B compartment, it has been shown that tethering of HP1a to ectopic sites in Drosophila using the LacI system leads to transcriptional silencing of most sites, except active promoters, and the establishment of physical connections between the ectopic HP1a site and other HP1a bound sites [67] (Figure 3A). In addition, HP1a RNA interference (RNAi) disrupts localization of HP4, a different heterochromatin protein [20]. However, it is not yet known whether large scale changes in the chromatin interactions that form A and B compartments will also occur in the context of HP1a tethering or depletion. In budding yeast, which lack H3K9 trimethylation and the HP1 protein, the Sir complex is the main heterochromatin silencing system. Using chromosome conformation capture on genetic mutants of the Sir complex, it was found that silencing of specific loci in this system can occur even without long-range clustering, although stability of the silent state is reduced in this genetic background [68].

Figure 3:

Experimental perturbations to compartmentalization to elucidate function

A. Tethering of HP1 (dark blue bean) to an A compartment (light red line) using LacO/LacI results in gene silencing and new interactions forming between the ectopic HP1 site and heterochromatic regions (dark blue lines), this forms an ectopic B2 or B3 sub-compartment [67].

B. Knockout of the Lamin B receptor in mouse thymocytes results in loss of heterochromatin tethering to the nuclear periphery; however, compartmentalization per se is preserved [66]. The B3 sub-compartment is disrupted in its nuclear localization, but not in its long-range interactions. By microscopy, the A compartment (red) is now observed to be closest to the nuclear periphery, while the B compartment (blue) is within the nuclear interior.

C. Knockdown of Srrm2, a structural component of nuclear speckles (NS), results in loss of NS foci (see Figure 2D), and dispersion of the NS components (small dark red circles) throughout the nucleus. A compartment strength is decreased, while B compartment strength is increased [47]. This is specifically affecting the A1 subcompartment that is usually close to NS foci.

D. Ectopic induction of phase separated droplets using Intrinsically Disordered Region (IDR) containing proteins related to active transcription such as TAF15, FUS, or BRD4 using the CasDrop system results in droplets (fuzzy cluster of orange circles) that exclude heterochromatin, form in regions of the nucleus with low chromatin density, and can function to pull two distal active regions together by fusion of separate CasDrops, forming an ectopic A1 or A2 compartment [27].

Another feature of the B compartment is localization to the nuclear periphery by association with the nuclear lamina, which serve as nuclear envelope anchors for heterochromatin [69]. There is a large body of literature on the effects of perturbing interactions between the nuclear lamina and chromatin; however, recent results suggest that while nuclear lamins are important for proper location of heterochromatin by tethering it to the nuclear periphery, this peripheral localization is not actually required for formation of microphase separated heterochromatin [66, 70, 71]. Indeed, multiple studies have shown that while lamin depletions can lead to local changes in gene expression, these changes do not always correspond to large scale compartment changes. One of these studies tested the importance of Lamin A/C nuclear envelope anchors to B compartment strength using a model of cardiac laminopathy, where Lamin A/C was depleted in human induced pluripotent stem cells (hiPSCs) that were then differentiated into cardiomyocytes [70]. Some changes in compartmentalization were observed by Hi-C in the differentiated cells, however, the majority of gene expression changes observed in the Lamin A/C depleted cells were not correlated with changes in compartments by Hi-C [70]. In addition, Hi-C and microscopy experiments in combination with equilibrium modeling in conventional (B compartment more peripheral) and inverted (B compartment more central) nuclei, derived from cells with or without Lamin B receptor respectively showed that chromosomes can compartmentalize to similar extents as measured by Hi-C with or without nuclear envelope tethering of B compartments [66] (Figure 3B). This result supports the model that compartment formation is an intrinsic property of chromatin that does not rely on tethering to the nuclear envelope [66]. Tethering of B compartments to the Lamina leads to specific subnuclear localization of these domains, but is not required for their clustering per se.

A separate study which used live-cell microscopy and modeled chromatin interactions using previously published single-cell Hi-C data showed that as cells exit mitosis, B domains that will become LADs self-associate before nuclear lamina assembly, and then move as a group to the nuclear periphery [71, 72]. In contrast, in Hutchinson-Gilford progeria syndrome (HGPS) fibroblast cell lines, large changes in compartmentalization and H3K27me3 are observed in late passages, near premature cellular senescence in these cells [73]. Interestingly, changes in H3K27me3 levels and Lamin A/C association preceded changes in compartmentalization of the chromosomes in this experimental system [73]. Overall, these studies suggest that while tethering of heterochromatin to the nuclear periphery is required for setting up the spatial organization of a conventional nucleus, other factors are usually sufficient for microphase separation of active and inactive chromatin. However, there may be circumstances such as cell senescence which require Lamin association to maintain compartment identity, and in these cases loss of peripheral localization of heterochromatin can lead to large scale changes in compartmentalization.

Much less is known about the role of A compartment clustering. Nuclear bodies such as speckles appear to be related to compartmentalization, as deletion of Srrm2 in mouse hepatocytes, which disrupts nuclear speckles, reduces the strength of A-A clustering [47] (Figure 3C). In Drosophila, inhibition of transcription elongation has been shown to reduce A-A interactions, but seems to have little effect on shorter range interactions along the diagonal of a Hi-C map during development [74, 75]. However, it is difficult to determine if the changes in long-range interactions are important for function, as transcription is already inhibited in these experiments. An interesting study using the CasDrop system showed that transcriptional regulators that contain IDRs can form droplets in vivo, and that these droplets exclude HP1 and heterochromatin, and are surrounded by active chromatin [27] (Figure 3D). Additionally, this study showed that fusion of two droplets could bring distal active regions together, which may contribute to long-range clustering of similar compartments. However, it has not yet been shown whether this clustering causes changes in gene activity. Further studies have shown that acetylated nucleosomes can also form phase separated droplets in vitro in the presence of bromodomain containing proteins, and these droplets exclude unacetylated nucleosomes [32]. Future work building on these tools will determine if loss or ectopic localization of bromodomain containing proteins which can induce phase separation of acetylated histones also changes A compartment interactions, and how this affects transcriptional activity.

Another approach to study the mechanism and function of nuclear compartments is to analyze this phenomenon across different cellular stages, tissue types, or organisms. One example of how this can be useful is considering the case of mitotic versus interphase cells, as mitotic cells have been shown by proteomics to retain many epigenetic marks though long-range compartment interactions are not present [76, 77]. In this case, many of the protein-protein interactions required for phase separation are likely disrupted by transient cell-cycle regulated phosphorylation and other post-translational modifications that occur during mitosis, such as H3S10 phosphorylation [76]. H3S10 phosphorylation disrupts HP1 binding to H3K9me3 marked histones [78]. By identifying events that regulate compartmentalization in different cell cycle states, we may be able to use these processes to perturb compartments in other contexts to gain greater understanding of the function of long-range interactions. In addition, cells derived from different tissues have been shown to have very different compartment strengths, providing a natural set of conditions to analyze the quantitative effects of compartment formation on gene activity [46, 66].

The antagonistic relationship between compartments and TADs

Within each nucleus, compartments exist in the same space and can overlap on the same polymers with topologically associating domains (TADs), but appear to be formed by different mechanisms. The balance between compartment and TAD formation may be important for correct regulation of chromosome function [61, 79, 80]. TADs are thought to be formed by cohesin mediated loop extrusion, are highly dynamic, and while they can form nested TAD structures, they do not generally have long range interactions beyond 1 Mb [61, 79, 80]. In contrast, compartments are thought to be formed by phase separation, and it is unknown if they are dynamic or stable structures. Compartment domains can associate over very large genomic distances (100s of Mb), and also with domains located on other chromosomes [11].

Cohesin loading and unloading proteins can be manipulated to modulate the level of loop extrusion experimentally. In experiments with decreased loop length, such as cohesin subunit depletions (Rad21-AID; Nipbl KO, Scc4 KO), compartment strength is increased, and small compartments become apparent within what is a TAD in the WT background, suggesting that loop extrusion usually prevents these small compartments from being formed [81–85]. In contrast, in cell lines lacking the cohesin unloader, Wapl, loop size increases, and the strength of compartmentalization is reduced [81, 86]. One hypothesis is that loop extrusion can pull sections of chromatin into a compartment that is inconsistent with their chromatin state. In effect, when loop extrusion is sufficiently strong, each TAD becomes a unit of phase separation. When loop extrusion is then inactivated, the intrinsic compartmentalization driven only by local chromatin state emerges, consistent with formation of new small compartment domains in the Nipbl mutant [85]. Modulating compartment strength separately from changing loop extrusion will be an important next line of experimentation, to determine if compartmentalization affects TAD formation as well.

This interplay between TADs and compartments has also been modeled, and these modeling experiments suggest that the balance between the strengths of these two processes is important to allow both to occur simultaneously on the DNA polymer [61]. Examples of cell types or organisms with extremely strong compartments or TADs would be interesting to study in this context to determine the effects of imbalance in the relationship between TADs and compartments on genomic function. For instance, studies in Drosophila suggest that in this species, the genome may be organized only by compartments, and not by the loop extrusion mechanism that is thought to contribute to TAD structure in mammalian nuclei. Therefore, Drosophila will be an interesting model system for further study of compartment function [74, 75].

Concluding remarks

Chromosome compartmentalization separates active and inactive regions of the genome, and is proposed to occur via a microphase separation mechanism. Both the A and B compartments contain putative LLPS factors, and polymer models of microphase separation recapitulate Hi-C compartment data. However, it is important to note that it is not yet known whether all of the protein-protein interactions mediating compartment interactions are actually engaged in LLPS or PPPS in vivo, or if some other mechanism is occurring such as the formation of hubs of specific activity based on concentration of protein binding sites. As has been recently discussed, such hubs do not always have to form by the process of LLPS [22, 23]. This is still a topic of active debate in the field. Another interesting avenue of future work is to look beyond eukaryotic genome organization. Although compartments have generally been studied in eukaryotes, very recent work suggests that compartmentalization occurs in some Archaea as well, but whether the mechanism is similar or different from that in eukaryotes is as yet unknown [87].

Future research should also focus on the relationship between long-range compartment clustering and chromatin function (see Outstanding Questions). These are difficult questions to answer, as many of the perturbations that would be predicted to globally affect compartmentalization will also have an effect on gene expression directly, so it will be useful to identify new ways of inducing or disrupting phase separation or long-range chromatin interactions independent of gene expression or replication timing. The optical tools developed by the Brangwynne lab could provide new ways to address this question [27]. Finally, little is known about the kinetics and dynamics of establishment and maintenance of chromosome compartmentalization, e.g. during the cell cycle or development, and how compartments vary from cell to cell, and we expect these topics to be the focus of intense studies in the coming years.

Outstanding Questions.

Is clustering of distal A or B compartments important for chromatin function?

Is gene expression or DNA replication more efficiently regulated when regions of similar activity are clustered in space?

Is clustering of different types of chromatin a consequence or a cause of chromatin regulation?

What are the phase separation factors for each of the sub-compartments?

Does any mixing occur between different sub-compartments within either heterochromatin or euchromatin?

What are the dynamics of establishment of compartmentalization, e.g. during the cell cycle and differentiation?

How much cell to cell variation is there to compartmentalization, and how does this effect chromosome activity at the single cell level?

What are the mechanistic interplays between compartmentalization and other chromosome folding processes such as loop extrusion?

Highlights.

• Recent studies have shown that chromosome and nuclear compartmentalization may be driven by a process of phase separation

• Subnuclear positioning of compartments can be achieved by tethering to the lamina, the nucleoli, and other subnuclear bodies

• Compartmentalization can be counteracted by loop extrusion

• Phase separated compartments may function in transcriptional control by in- and ex-cluding specific co-factors required for gene regulation

• Recently developed tools will allow researchers to interrogate the function of long-range compartment interactions and phase separated regions of the nucleus

Glossary

Block copolymer

polymer that is composed of alternating blocks of monomers of different types

Bridging Factor

a factor that binds to two regions of the same type in a block copolymer to form a crosslink, resulting in polymer-polymer phase separation (PPPS). In the case of chromatin, a bridging factor would be a protein or protein complex and/or RNA that binds to two genomic regions of the same compartment, but that do not necessarily bind to other bridging factors to form multivalent interactions

CasDrop System

CRISPR-Cas9 based optogenetic technology developed by the Brangwynne lab that can induce localized condensation of liquid droplets at specific genomic loci

Chromosome Territory

region of the nucleus that is occupied by chromatin from just one chromosome. By Hi-C, chromosomes tend to have more intra-chromosomal interactions than intra-chromosomal interactions even for loci separated by very large genomic distances, which reflects this organization that was first observed by microscopy

Euchromatin

type of chromatin characterized by the presence of active genes, low chromatin density, and active histone marks and variants. Corresponds to the A compartment

Heterochromatin

type of chromatin characterized by being transcriptionally silent, with a higher chromatin density, and inactive histone marks. Corresponds to the B compartment

Intrinsically Disordered Region

region of a protein that does not have any defined secondary or tertiary structure due to the amino-acid sequence

Macrophase

type of phase separation where the phases become completely separated into one large region per phase. Usually applied to large scale ‘macroscopic’ processes, but can also apply to microscopic processes where the phase separation leads to complete segregation of the materials

Microphase

type of phase separation where the spatial clusters are relatively small and the two types of material do not entirely separate in just two large domains, but into many small domains for each type

Phase Separation

separation of two substances in a mixture based on their physical properties. Eg. Oil and water

Polymer

large macromolecule that is made up of a string of smaller repeating subunits called monomers

Super-Enhancer

cluster of enhancers with very high levels of transcription activating factors, often regulate genes important for cell identity