Genome folding dynamics during the M-to-G1-phase transition

Abstract

All measurable features of higher-order chromosomal architecture undergo drastic reorganization as cells enter and exit mitosis. During mitosis, gene transcription is temporarily halted, the nuclear envelope is dismantled, and chromosomes undergo condensation. At this time, chromatin compartments, topologically associating domains (TADs), and loops that connect enhancers with promoters as well as CTCF/cohesin loops are dissolved. Upon G1 entry, genome organization is rebuilt in the daughter nuclei to resemble that of the mother nucleus. We survey recent studies that traced these features in relation to gene expression during the mitosis-to-G1-phase transition at high temporal resolution. Dissection of fluctuating architectural features informed the hierarchical relationships of chromosomal organization, the mechanisms by which they are formed, and their mutual (in-) dependence. These studies highlight the importance of considering the cell cycle dynamics for studies of chromosomal organization.

Introduction

During mitosis, chromatin architecture undergoes dramatic reorganization [1,2]. In prophase, the nuclear envelope breaks down, leading to mixing of nucleoplasm and cytoplasm, and chromosomes begin to condense. Cohesin ring complexes that configure interphase structural loops and link sister chromatids in G2 phase are removed, followed by chromosome segregation [3,4]. During metaphase, chromosomes are arranged into consecutive randomly positioned loop arrays, facilitated by the condensin ring complexes I and II [1,2]. This mitotic folding state occurs uniformly across all chromosomes, across cell types and species, and is locus-in-dependent. At that time, the now-classical hallmarks of Hi-C chromosomal contact maps that are observable throughout interphase are almost entirely dissolved [1,2]. This includes A/B compartments, topologically associating domains (TADs), structural loops anchored by CTCF and cohesin, as well as loops between promoters and enhancers [1,2,5,6]. However, in some studies, traces of interphase-like architectural features can be observed in mitotic cells that are not accounted for by contamination with interphase cells [5]. This suggests that mitotic chromatin organization is not entirely incompatible with the maintenance of interphase chromatin structures. The massive reorganization in chromosome architecture during mitosis as well as its uniformity across tissues has called into question whether architectural features can contribute to epigenetic memory that maintains regulatory information throughout the cell cycle [7].

Mitosis is also accompanied by loss or strong reduction in chromatin binding of a large fraction of transcription factors and their co-regulators, RNA polymerases, as well as chromatin-modifying/remodeling complexes [8]. The architectural protein CTCF is variably evicted from mitotic chromatin depending on cell type, but is rapidly rebound in anaphase [5,9,10]. Rebinding to chromatin of nuclear factors may influence or be influenced by architectural reorganization in newborn nuclei. However, it is important to note that not all chromatin features are lost during mitosis, as exemplified by the persistence of most histone modifications [11–15,48], chromatin accessibility at promoters and enhancers [16,17], and a variety of DNA-binding transcription factors or cofactors [18,19].

It is thought that the maintenance of cell identity and cellular functions necessitates the faithful restoration in G1-phase of the mutually influential processes of gene transcription and genome folding patterns [20–23]. This review focuses on how higher-order chromatin architecture is reshaped during the mitosis-to-G1-phase transition, and the multilayered forces that drive this reconfiguration in relation to gene transcription.

Complex patterns of chromosomal reconfiguration after mitosis

Most perturbative experiments designed to study the role of architectural factors were carried out in asynchronously growing cells, which means that they largely considered the maintenance of the interphase chromatin state. In contrast, focusing on the mitosis-to-G1-phase transition enables inquiries into the establishment of architectural organization.

While global loss of compartments, TADs, and loops during prophase entry occurs within a few minutes [2], their full restoration in G1 phase extends over a time frame of hours, which facilitates highly time-resolved dissections of intermediate chromatin folding states [5,24–26].

Several recent reports that are based on timed chromo-some conformation capture-based methods described the dynamic restoration profiles of chromatin architecture after mitosis. However, when comparing time-scales between studies, variation in cell cycle length between cell lineages must be taken into account. For example, embryonic stem (ES) cells have a very short G1 phase (1–2 hours) [27], suggesting that rebuilding chromatin architecture may occur in an accelerated manner compared to cells with longer G1 phases. One of the first global chromosomal rearrangements after mitosis occurs during telophase and is characterized by a lack of condensin-driven loops. Notably, this stage precedes cohesin-dependent chromosomal restructuring [24]. The function, if any, of this short-lived more or less loop-depleted and poorly compartmentalized state is unknown, but might present a ‘clean slate’ for the cohesin machinery to act upon in the absence of potential interference by condensin loops. What follows during G1-phase entry is a complex emergence pattern of structural features with distinct reformation kinetics at different length scales, and with varying reliance on cohesin complexes, CTCF, and other nuclear factors. In this section, we summarize how chromatin compartments, TADs, and loops are re-established in newly formed nuclei.

Chromatin compartments

Dileep et al. used multiplexed 4C to interrogate the dynamics of chromatin recompartmentalization after mitosis in C127 cells [28] that had undergone nocodazole-induced prometaphase arrest, followed by mitotic shake-off and release. Chromatin compartmentalization emerged as early as 0.5 hours of G1-phase entry and intensified up to 4 hours, coincident with the establishment of early versus late DNA replication domains. Interactions between closely spaced compartmental domains resumed more rapidly than those separated by greater distances [28]. Similar findings were made in a single-cell Hi-C study on asynchronous murine ES cells, ruling out artifacts from nocodazole exposure [29]. More recent studies using high-resolution Hi-C in a variety of murine and human cell lines revealed a similar early emergence of compartments, followed by their progressive intensification and expansion [5,24,25]. The earliest stage of detectable chromatin recompartmentalization was ana/telophase in murine erythroblasts that were purified via fluorescence-activated cell sorting [5]. An independent study using locus-specific oligo paints in murine embryonic fibroblasts revealed that the repressive lamina-associated domains (LADs), which largely overlap with B-type compartments, aggregate as early as in anaphase [30]. Of note, in HeLa S3 cells, A-type (active) formed slower than B-type (repressive) compartments at smaller distances but faster at larger distances, suggesting the existence of distinct forces that drive chromatin segregation at different length scales [24]. In sum, compartment formation is one of the earliest events in newly formed nuclei and, in light of cohesin recruitment dynamics, is likely driven largely by cohesin-independent mechanisms, such as chromatin state (Figure 1a) (see also below).

Figure 1.

Illustration of chromatin structural reconfiguration after mitosis. (a) Schematic showing the gradual emergence, expansion, and intensification of the characteristic checkerboard pattern of A/B compartments in Hi-C contact maps. Note that compartments become visible already in ana/telophase. (b) Illustration depicting that TAD reformation follows a ‘bottom-up’ fashion. (c) E–P interactions can emerge more rapidly after mitosis than structural loops even when accounting for loop size. (d) Model showing transient E–P interactions in ana/telophase that are terminated by interfering structural loops. (e) Depletion of CTCF eliminates structural loop reformation, enabling persistence of transient E–P interactions. (f) Pre-establishment of gene domains before completion of the first round of PolII elongation along the gene body.

Hierarchical reorganization of topologically associating domains and structural loops during G1 phase entry

TADs are distinctive submegabase structures that may comprise nested sub-TADs. TADs and sub-TADs are frequently bordered by CTCF and cohesin, and are thought to result from a cohesin-driven, CTCF-arrested chromatid loop extrusion process. So-called ‘corner dots’ visible on Hi-C maps at many TADs indicate the presence of loops, typically flanked by CTCF/cohesin that are also referred to as structural loops [5]. These dots represent chromatin contacts that are longer-lived than those that can be seen throughout the TADs, the latter representing varying loop extrusion states and manifesting as flares or stripes on Hi-C contact maps. In murine erythroblasts and ES cells, TADs are rebuilt in a ‘bottom-up’ fashion such that small nested sub-TADs emerge in ana- or telophase and subsequently converge into TADs later in G1 (Figure 1b) [5,25]. TAD formation is accompanied by increased insulation at domain boundaries. TAD emergence slightly preceded compartmentalization in one study [24], while the inverse was true in another [5], likely due to technical or analytical reasons. Regardless, TADs and chromatin compartments reflect different organizational principles and can therefore be dynamically uncoupled.

A highly time-resolved characterization of structural loops, including assessment of CTCF and cohesin chromatin occupancy profiles, found that loop dynamics followed that of postmitotic accumulation of cohesin but not that of CTCF, which reoccupies chromatin very rapidly [5]. Moreover, large loops were slower to form than small ones. These findings support the model that CTCF is by itself insufficient to form a structural loop, and that its architectural role is mostly to stall cohesin-driven loop extrusion.

Cis-regulatory element contacts can be forged rapidly after mitosis

G1-phase entry offers the opportunity to examine the degree to which cis-regulatory element (CRE) contacts regulate gene transcription. In erythroid precursor cells, some contacts between CREs were detectable as early as ana/telophase before the complete clearance of condensin and before the reloading of cohesin onto chromatin, as determined by live-cell imaging (Figure 1c) [5,24]. Contacts between CREs heavily marked by active histone modifications tended to form faster. Similarly, in murine ES cells, chromatin contacts strongly marked by H3K27ac at both loop anchors tended to be built up faster than loops flanked by only CTCF and/or cohesin [25]. That CRE contacts emerge early in ana/telophase highlights the possibility that these interactions can be independent of cohesin-mediated loop extrusion. This was further confirmed via acute degradation of CTCF during the M-phase to G1-phase progression, which prevented formation of all loop domains but not most CRE contacts [31]. The underlying driving force of the rapid CRE contact formation remains to be explored. Potential mechanisms may include affinity aggregation between active chromatic regions or protein–protein interactions mediated by a compatible sets of rapidly recruited transcription factors.

Investigations into the function of CRE contacts were further helped by acutely degrading CTCF during the metaphase to G1 phase, as CTCF has been reported to both promote and inhibit select CRE contacts [32]. Degradation of CTCF during the prometaphase–G1-phase interval has limited impact on gene expression changes in erythroid cells and in ES cells [31,33]. Notably, downregulation of genes was not associated with a significant loss in looped contacts [31], suggesting that at these genes, CTCF might act as a conventional transcriptional activator. In contrast, a considerable fraction of genes activated upon CTCF loss in early G1 phase gained long-range enhancer contacts, in line with the idea that CTCF can protect against ‘illegitimate’ proximity with distal enhancers.

Intriguingly, highly time-resolved Hi-C experiments detected a small group of transient enhancer–promoter (E–P) loops that were formed early in ana/telophase but vanished again later in G1 phase (Figure 1d). Their disappearance was associated with the interposition by late-forming CTCF/cohesin-anchored structural loops [5]. Upon acute depletion of CTCF, a fraction of these transient loops persisted deep into G1 phase, further confirming the ability of CTCF to limit improper E–P pairing (Figure 1e). Notably, the ability of CTCF to exert this function was linked to its engagement in structural loops [31]. Whether the transient E–P loops have any biological function, such as controlling the kinetics of transcription reactivation after mitosis or during early gene ‘spiking’ remains an open question. Even though E–P contacts generally correlate with gene activity during G1 entry [25], attempts to correlate CRE contacts with gene activation early during exit from mitosis are complicated by the finding that as the genome is activated, it experiences a hyperactive state during which a large number of genes, including otherwise silent ones, transiently spike in activity [5,12,33,34]. Moreover, this transcriptional spiking may be promoter-driven and occur independently of distant enhancers [5]. In conclusion, CRE contacts can form without the assistance of cohesin-mediated loop extrusion, implicating as-yet-unknown means of chromatin contact formation. However, CTCF/cohesin loops can disrupt such contacts if loop anchors are interposed between the CREs, or promote them if flanking them [31].

Transcription reactivation in the context of nuclear architecture

Whether and how transcriptional activity modulates chromatin architecture and vice versa is under debate. Bringing cell cycle dynamics into the picture helps to shed light on this interplay.

Transcription is reactivated as early as ana/telophase as assessed by RNA PolII ChIP-seq experiments [5,12]. Similar observations have been made using independent techniques, including PRO-seq, RNA-seq, EU-labeling-seq, and RNA FISH in both somatic cells and pluripotent cells [12,25,33,34].

More generally, as cells progress into G1 phase, gene reactivation does not occur uniformly, and there has been substantial speculation surrounding the importance of classes of genes that are activated early. For example, it has been proposed that immediate reactivation of lineage-determining genes is required for the stable maintenance of cellular lineage identity through mitosis [19,23,25,35,36]. However, this idea, while attractive, remains challenging to prove.

As mentioned above, these studies revealed a widespread spiking of transcription after mitosis. Yet, such spiking of gene expression is not correlated with promoter–enhancer interactions. For example, the c-kit gene showed postmitotic transcriptional spiking, while the physical interaction between the c-kit promoter and its well-characterized distal enhancer did not, suggesting an additional layer of transcription regulation after mitosis [12]. More generally, genes associated with early- versus late-forming E–P loops displayed similar reactivation kinetics [5,25]. However, limitations of such analyses include the difficulty in designating enhancers to their target genes, and the limited sensitivity in detecting short-range E–P interactions in Hi-C experiments. Therefore, it remains possible that E–P interactions play a functional role during early transcription reactivation after mitosis.

In the context of the mitosis-to-G1 transition, inhibition of transcription initiation through triptolide treatment had remarkably little effect on all levels of chromatin reorganization, including compartments, TADs, and loops [31]. Also, chromatin insulation at transcription start sites (TSS) was unperturbed upon triptolide treatment, suggesting that RNA polymerase II (PolII) does not contribute substantially to local insulation at TSS, and raising the question as to how potent PolII is as a loop extrusion barrier, or to what extent it contributes to cohesin loading [37–40]. Similar observations were made in ES cells depleted during mitosis of H3K27ac via chemical inhibition of p300/CBP, which impacted transcriptional reactivation in G1 phase but had minimal effects on 3D architecture [25].

In contrast, another study found that auxin-induced degradation of PolII during mitosis impacts many layers of chromatin organization, including compartment boundaries, TAD boundaries, and chromosome loops [41]. Some of these changes were attributed to diminished cohesin loading in the absence of PolII. In order to reconcile these findings, one needs to consider that in the latter study, cells were first synchronized in G2/M phase by exposure to a CDK1 inhibitor for 21 hours, followed by release into G1 phase with auxin treatment of up to 14 hours. This contrasts with the 3-hours chemical inhibition of PolII in the Zhang et al. study [31] that also did not use a CDK1 inhibitor but instead relied on nocodazole arrest (8 hours) release. None of the treatment regimens are immune to artifacts, but shorter term PolII inhibition may be less prone to confounding secondary effects. Degradation of RNA PolI and PolII in asynchronously growing cells also had little impact on large-scale chromatin architecture [42]. Moreover, in that study PolII degradation for up to 6 hours specifically during G1-phase entry did not impede the formation of compartments and TADs. Short-term Pol-II depletion also had little impact on cohesin recruitment. However, it remains possible that epigenetic changes imposed by PolII may persist beyond the time of degradation. Thus, we would not dismiss as secondary effects all long-term PolII inhibition-induced architectural changes.

Two more insights into the relationship between transcription and genome folding that are worth mentioning were afforded by examining cell cycle dynamics. The first is that small domains of interactions across the lengths of individual genes (‘gene domains’, represented as small triangles along the diagonal of Hi-C maps) that are marked by H3K36me3 can be re-established early in G1 phase before completion of the first round of transcription (Figure 1f) [31]. This led to the speculation that since H3K36me3 is stable through mitosis [11], these gene domains, perhaps disrupted in M phase via condensin extrusion, may constitute a form of epigenetic memory that facilitates faithful transcription elongation during the earliest round of transcription. The second is that when TADs and gene reactivation were stratified according to their kinetics, there was little correlation between them, suggesting the TAD organization has a limited influence on gene regulation [5]. But as is always the case with global studies, such lack of correlation does not necessarily exclude that some genes might still be influenced by their surrounding TADs.

Limitations of current studies and future perspectives

Discrepancies in the emergence kinetics of structural features between studies might be due to different cell types used. However, it seems unlikely that such fundamental and conserved mechanisms and their underlying physical principles should vary much between lineages. Other reasons are of technical nature and might relate to the way cells were purified or to what conditions they were exposed. Owing to the large input requirement for genomic studies (ChIP-seq, Pro-Seq, Hi-C, etc.), most experiments were performed on cell populations arrested by chemical treatment such as nocodazole and RO3306, which may introduce biases or artifacts. New methods that require smaller cellular input amounts may potentially allow studies on asynchronously growing cells. Also, cross-linking with formaldehyde can produce artifacts in terms of transcription factor occupancy [43,44] and perhaps Hi-C experiments as well. Although it is noteworthy that formaldehyde may mostly influence nonspecific chromatin-binding proteins and leave specific binding factors less perturbed [45]. However, in a carefully done Hi-C study, different formaldehyde concentrations produced similar results [1]. Another technical limitation is that loops less than ~50 kb in size are mostly undetectable in the Hi-C experiments discussed here, which affects the interpretation of any correlations between loops and gene activity. The development of higher-resolution technologies such as micro-capture-C [46] and region capture micro-C [47] will help address this limitation.

The studies considered here described a complex pattern of chromatin reconfiguration after mitosis. Most nuclear factors that govern this process and resculpt chromatin at different scales are still unknown. Assessing the role of any molecule specifically during postmitotic stages is difficult as it comprises a very short window of the cell cycle. Traditional knockout or knockdown strategies are not suitable because of the time it takes for proteins to be degraded. The use of the inducible protein degron techniques or small molecules that displace select proteins from chromatin will further such types of studies. The advent of evermore sophisticated imaging technologies that allow the detection of structures such as TADs and loops at the single-allele level will help overcome the limitations posed by population-based methods and shed light on their hierarchical organization during G1-phase entry.

Data Availability

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