Chromatin dynamics and cellular energy homeostasis

Abstract

Metabolism is a complex consortium of chemical reactions, consuming and releasing energy. In addition, cellular respiration is a pivotal catabolic process of nutrients using oxygen to generate ATP. Dynamics of chromatin structure are fundamental to genomic events in all cellular processes. Therefore, chromatin consumes robust cellular energy formed in metabolism. To regulate cellular processes, chromatin functions and metabolism need to be tightly coordinated in response to environmental changes including nutrition availability. Indeed, molecular oxygen directly impacts DNA and chromatin modification and influences cell fate. Herein, we review fundamentals of chromatin dynamics, nucleosome modifications and roles of chromatin dynamics in spatiotemporal regulation of cellular energy homeostasis.

Graphical Abstract

Graphical Abstract.

Oxygen is an important molecule for eukaryotic life. In aerobic respiration, 90 to 95% of oxygen consumed in the body is used in mitochondria to generate cellular energy in the form of ATP (1,  2). ATP is in high demand for macromolecule biosynthesis including DNA and RNA synthesis (1,  3,  4). Oxygen has a high standard reduction potential; therefore, it readily oxidizes other substances. Overproduction of intermediate species of oxygen, reactive oxygen species (ROS), can be harmful for the structures of cells and tissues including lipids, membranes, proteins and DNA through disruption of redox events (1,  5,  6). At low or moderate concentrations, however, ROS turn out to be beneficial for obtaining immunity in individuals through their antimicrobial activity and as signalling molecules for oxidative stress responses (7,  8). Anaerobic respiration produces fewer ROS compared to aerobic respiration (9). Anaerobic species or tissues have alternative respiratory chain redox systems; however, they still produce ROS that damage proteins and DNA (9).

Metabolism is the sum of all chemical reactions in energy anabolic and catabolic processes and maintains energy homeostasis. Cellular respiration is a metabolic process generating ATP from glucose (10). Hence, metabolism influences cell fate by consuming molecular oxygen. Chromatin greatly consumes cellular energy by its dynamics including nucleosome modification. Therefore, coordination between chromatin functions and metabolism in response to environmental changes and cellular metabolic status is pivotal to establishing coherent energy homeostasis. Imbalanced energy distribution by disruption of this coordination leads to developmental malfunctions, disorders including neurodegenerative disease and tumourigenesis (3,  11–13).

Here, we review roles of chromatin dynamics in redox and energy homeostasis by recapturing features of (i) chromatin structure dynamics and nucleosome modification, (ii) fundamental functions of chromatin modifiers, (iii) the interplay between metabolite, molecular oxygen and chromatin dynamics, and (iv) chromatin dynamics in spatiotemporal events that result in transcription. We will also discuss metabolic changes impacting variations in DNA sequence, in which inappropriate changes may interrupt the process of essential chromatin dynamics by altering the nucleosome substrate itself.

Nucleosomes are key substrates in chromatin dynamics and cellular energy homeostasis

For proper cellular processes, chromatin regulators and cellular fitness need to be coordinated. Details of these complex mechanisms are built with chromatin as a substrate and contain many components, which regulate nucleosomes. Chromatin is composed of repeating units of nucleosomes. The nucleosome comprises a histone octamer consisting of two copies of each core histone H2A, H2B, H3 and H4, wrapped by 147 base pair (bp) of double-stranded DNA. In humans, there are eight variants of H2A, six variants of H3 and two testis-specific variants of H2B. Many histone variant genes lack introns, and their mRNAs are not polyadenylated as with canonical core histones (5,  14). These histone genes encode histone protein expressed during S-phase for packaging newly synthesized DNA into chromatin (replication-dependent histones) (15). By contrast, the genes of some other histone variants including CENPA and macro-H2A contain introns and their mRNAs are polyadenylated (15). Synthesis of these histones is replication independent (replication-independent histones) (5,  15). Preferences of histone chaperones for binding to distinct histones contribute to proper assembly or disassembly of nucleosomes (5). After DNA replication, ATP-dependent chromatin remodelling complexes regularly space nucleosomes as a part of the chromatin assembly process (16,  17). During aging, however, increased replication-independent nucleosome assembly leads to more ectopic and fuzzy nucleosome positioning (5).

ATP-dependent chromatin remodelling complexes are classified into four families: imitation switch (ISWI), chromodomain helicase DNA-binding (CHD), switch/sucrose non-fermentable (SWI/SNF) and INO80 (18–23). The common mechanisms among ISWI, CHD and SWI/SNF are that they use ATP hydrolysis for executing DNA translocation from a fixed location, two DNA helical turns away from the nucleosome dyad (18). Nucleosome spacing is primarily determined by the ISWI and CHD subfamily remodellers (18,  24). ISWI binds linker DNA and then slides ~1 bp of a nucleosome per ATP hydrolysis (18,  25). Repeating the action draws a nucleosome to abut the adjacent nucleosome, establishing the next nucleosome at a fixed distance from the substrate nucleosome (18). There are two homologs of ISWI in yeast, ISW1 and ISW2, forming distinct complexes (26). Interestingly, ISW2 binding distorts the nucleosome before the ATPase motor moves on nucleosomal DNA (27). In contrast, the yeast SWI/SNF subfamily member, RSC complex, moves along one of the two strands of the DNA, moving in the 3′–5′ direction, leading to one helical rotation of DNA for every ∼10 bp that is translocated (28). RSC remodelling utilizes higher ATPase activity and coupling efficiency (translocated DNA base-pairs per ATP hydrolysis cycle), changing the nucleosome remodelling reactions from nucleosome sliding to histone ejection (29,  30). The common function of remodelling complexes is that the DNA propelling action of remodellers moves 1–2 bp of DNA per cycle of ATP binding–hydrolysis (18). Hence, ATP-dependent chromatin remodelling complexes can alter the chromatin structure to both transcriptional active and silent status by utilizing ATP.

ATP-dependent chromatin remodelling complexes play crucial roles in every level of chromatin organization. Even the organization of three-dimensional (3D) structures within large chromatin domains, topologically associated domains (TADs), is dependent on ATP-dependent chromatin remodellers (31–33) (Fig. 1A). TADs are characterized by a high frequency of intra-domain chromatin contact and play decisive roles in gene expression (34). A recent study of chromatin conformation capture (3C) mapping at base-pair resolution used a novel approach termed ‘Micro Capture-C ultra’ (MCCu). This study revealed that nucleosome-associated chromatin (densely condensed nucleosomes) forms nanoscale domains of interactions, within conventional TADs, which are interrupted by nucleosome-free regions (35) (Fig. 1A). The nucleosome-free regions extend above the chromatin surface, potentially forming phase-separated condensates that act as hubs for abundant transcription factors (TFs) and co-activators (35) (Fig. 1B). Using MCCu, it was found that the contacts between enhancers and promoters are facilitated by the nucleosome-free regions in each enhancer and promoter as well as by enhancer and promoter regulatory elements (35). The CCCTC-binding factor (CTCF) binding sites generally organize local fixed nucleosome positioning. Further, the CTCF binding sites specifically contact other CTCF-binding sites at the boundaries of the larger TADs rather than enhancers (35) (Fig. 1A). This is often observed at a CTCF-binding site that was closely positioned to another nucleosome-free region, which resulted in fixed positioning of multiple nucleosomes relative to the DNA sequence (35) (Fig. 1A). These observations suggest that CTCF restricts the transcriptionally active regions to minimize energy consumption. This is in addition to the original concept that CTCF functions as an insulator-binding protein, which prevents enhancer–promoter interaction, leading to suppression of gene expression (36–38). CTCF also organizes extruding DNA into loops via cohesin, which belongs to the structural maintenance of chromosomes (SMC) family of ATPase complexes, generating TADs (31–33). Notably, ATP is required for TAD organization. In yeast ISW1a, an ATP-dependent nucleosome remodelling complex, slides nucleosomes into spaced arrays (Fig. 2) and interacts with cohesin (39). ISW1a is also thought to promote translocation of cohesin from the centromeric sites of loading to neighbouring regions (39). Therefore, repositioning nucleosomes by ATP-dependent chromatin remodelling complexes/ATPases may generate the nucleosome-free regions in conjunction with CTCF (Fig. 2) (described below). This system may be energetically efficient as ATP consumption is linked to interactions with the DNA-binding sites of CTCF. However, the DNA binding activity of TFs (but not their amount), which can act as a boundary limiting nucleosome sliding activity to one direction, is influenced by variations in DNA sequences in the flanking regions of their binding sites (40,  41) (Fig. 2). This mechanism supports the concept of maintaining nucleosome-free regions by remodelling (sliding) complexes and explains the size differences of loops formed between enhancers and CTCFs (Figs 1A and 2).

Fig. 1.

Summary model of TAD formation and chromatin dynamics in SE creation. (A) There are high-compacted nucleosomes and nucleosome-free regions within CTCF organized TADs. The nucleosome-free regions extend above the chromatin surface. Regions flanking the nucleosome-free regions are compartmentalized with compacted nucleosomes of which surfaces are enriched in H3K27-acetylated nucleosomes that form relaxed chromatin. To create nucleosome-free regions, which are critical for transcription activation, TFs bound to regions flanking nucleosomes may need to act as barriers to nucleosome sliding activity (Fig. 2). Sequence variations may lead to preferences of enhancers in enhancer-enhancer connections, which create loop formation (Fig. 2). (B) Our proposed model for creations of SE is illustrated. TFs’ low affinity regions do not contribute to maintaining nucleosome-free regions; however, these regions may contribute to eRNA transcription to function as promoter/enhancer-like regions in DNA loops that are partitioned between nanoscale nucleosome domains (A). Acetylated nucleosomes, acetylated TFs and eRNA are BRD4 bromodomain binding substrates. The high density of these macromolecules generates CTCF-mediated SE using phase separation. GTFs, general TFs.

Fig. 2.

Model illustrating the establishment of nucleosome-free regions based on DNA sequence variations in TFs- and CTCF-binding regions flanking regions of nucleosomes. ISW1 nucleosome sliding is important to maintain nucleosome-free regions (24). DNA binding of TFs to their binding regions are affected by variations in their binding sequences and compete with ISW1 sliding activity. The amount of TFs does not affect ISW1 activity in vitro (40), suggesting creation of nucleosome-free regions is strongly influenced by sequence of DNA motifs.

Regarding biological significance, CTCF suppresses the somatic gene expression program and promotes a part of pluripotency-associated gene expression in mice (36). Importantly, SMARCA5, a subunit of the mouse ISWI complex, interacts with CTCF (36,  42). Smarca5 knockdown leads to improper nucleosome positioning adjacent to CTCF-binding sites and reduces binding of CTCF and TFs at CTCF binding sites at the time when somatic cells are reprogrammed (36). Therefore, loss of SMARCA5 appears to reverse CTCF-insulator activity, maintaining somatic gene activation and impairing reprograming (36). The SWI/SNF ATP-dependent remodelling complex has been shown to bind to CTCF and localize on the loci overlapping with CTCF sites in human cells (43). Thus, SWI/SNF remodelling activity may influence CTCF activity and vice versa.

The reversible nature of chromatin structure commits signalling responses and epigenetic reprogramming. Nucleosome positioning becomes fuzzier in aged chromatin (44) as seen in lower nucleosome density in introns in late-passage compared to early-passage cultured human cells (45). It was shown that a low density of nucleosomes leads to increases in the speed of transcriptional elongation as measured by RNA polymerase II (Pol II) coverage during aging in five species including flies and humans. Moreover, reduction of elongation speeds by additional expression of nucleosomes extends lifespan in flies (45). Consistent with this, a reduction of histone H3Lys36 trimethylation (H3K36me3), a transcription elongation-linked histone modification, over open reading frames (ORFs) led to increased cryptic transcription and shorten replicative lifespan (RLS) in aged yeast cells and aged worms (46). H3K36me3 recruits the Rpd3S histone deacetylase complex and resets the nucleosomes with hypoacetylation and enriched H3K36me3 after Pol II passing in yeast (47,  48). This may permit multiple rounds of proper transcription under reduced acetyl-CoA levels due to a decline in metabolism during aging. However, the downside is that lower nucleosome density and increased nucleosome displacement (histone exchange) during aging may expose cryptic promoters inducing cryptic transcription, which is the unintended initiation of transcription from non-promoter regions within the gene body, rather than the normal start site (48–50).

Nucleotide metabolism may be behind altered chromatin substrates

DNA is sensitive to redox homeostasis, and its changes alter chromatin modification. DNA alkylating agents add alkyl groups to oxygen or nitrogen atoms in the DNA bases including modifying Ο6-guanine (Ο6G) to Ο6-methylguanine (Ο6meG) (51). Mismatch repair proteins, MSH2 and MSH6, form the MutSα complex to bind O6meG:C (52). O6meG:C can lead to mismatching generating O6meG:T. Unphosphorylated MSH6 binds O6meG, leading to MutSα binding O6meG-T mismatches, inducing DNA damage signalling by recruiting MutLα (MLH1-PMS2) (52). Stabilization of MutSα in response to alkylation damage occurs through interactions with the molybdopterin synthase associating complex (MPTAC), which regulates sulphur metabolism, and the Ada2a-containing histone acetyltransferase complex (ATAC) (53,  54). These interactions have been lost in human lymphoblastoid cells from Fragile X-associated disorders (FXD) patients. FXD is caused by CGG repeat expansion in the 5′ untranslated region of the FMR1 (53–55). Remarkably, these cells have altered purine metabolism (56), indicating that redox signalling relies on sulphur metabolism and purine metabolism to stabilize MutSα. Expansion of nucleotide repeats, such as CAG/CGG/ATTCT repeat, is a part of the pathogenesis of neurodegenerative disorders including Huntington’s disease (HD) (57). Base editing with G•C > A•T and A•T > G•C within CAG and GAA repeats alleles in the central nervous system of HD model mice reduced repeat expansion (58).

A recent study revealed that alteration of the cellular environment changes the length of tandem DNA repeats in somatic cells in a tissue-specific manner (59). An analysis of CAG trinucleotide repeats in blood cells using whole-genome sequencing (WGS) of biobank cohorts revealed that most human genomes contain unstable repeat alleles at 29 loci carrying repeat elements with somatic expansion (59). These allele-specific repeats were increased with aging although the instability of the expansions is common in the germline (59). Further search for genetic modifiers of repeat expansion found seven genetic modifiers of TCF4 CAG repeat expansion including MSH3 and four genetic modifiers of non-CAG DNA repeats (more than three short tandem repeats), MSH2, MSH3, MLH3 and PMS2, which encode mismatch repair proteins (59). Thus, an alkylating environment can be a high-risk factor for trinucleotide repeat disorders.

There are multiple links between redox status and chromatin modification. Notably, the human MSH2-MSH6 heterodimer disturbs a nucleosome (reducing affinity between DNA and histone octamer), and it was facilitated by histone H3 acetylation in vivo (60). Hence, altered redox homeostasis may alter chromatin formation within long range DNA regions including cis-regulatory elements associated with nucleotide repeat expansions. Interestingly, some nucleotide metabolic enzymes directly modify nucleosomes (61–64). For instance, guanosine 5′-monophosphate synthase (GMPS) deubiquitinates histone H2Bub1 through association with USP7, and phosphoribosylaminoimidazolesuccinocarboxamide (SAICAR) induces H3T11 phosphorylation through association with pyruvate kinase M2 (PKM2) (61–64). In addition, as ATP is highly utilized in DNA synthesis, the nucleotide metabolic enzymes may mediate a functional link between energy homeostasis and chromatin modifications in response to alteration of redox status.

Metabolism Governs Chromatin Dynamics and Transcription for Redox Homeostasis

Chromatin modifiers establish chromatin dynamics by regulating DNA modifications and/or nucleosome modifications to achieve transcription, DNA replication, repair and recombination. Amino acids in the Ν-terminal tails of histones within nucleosomes are modified by attachment of functional groups of many types of acyls, methyl and phosphoryl moieties as well as ubiquitin, SUMO and others. This is termed post-translational modifications (PTMs) of histones (65–67). Most of these modifications are catalysed by chromatin modifiers, the majority of which form protein complexes that interact with DNA and/or nucleosomes and TFs and utilize cofactors in their enzymatic reactions (3,  11). Distinct protein domains of histone modifying complex subunit(s) themselves recognize histone PTMs. Thus, histone modifying complexes recognizing histone modifications can lead to crosstalk/connections among PTMs (65,  66,  68,  69). PTMs also influence how chromatin-modifying complexes, including ATP-dependent remodellers, bind to nucleosomes (3,  11,  30,  47,  68,  70–72). Importantly, most of the cofactors for PTMs are metabolites (3,  73). Therefore, the global levels of nucleosome PTMs may reflect metabolic status and energy availability. Moreover, ROS disrupt not only cofactor synthesis but also reactions of chromatin modifications in addition to damaging DNA (5,  12). Here, we illustrate the effects of cellular metabolism on nucleosome structure and enzymatic activities by focusing on central metabolites. We also discuss roles of chromatin modification in metabolism.

Acetyl-coenzyme A

Acetyl-coenzyme A (Acetyl-CoA) is a key metabolite from carbohydrate, amino acid, and fatty acid catabolism and is the substrate for histone acetylation (74,  75). Histone acetyltransferases (HATs) use acetyl-CoA to transfer the acetyl group to the ε-amino group of lysine side chains of histones. Histone acetylation is thought to weaken the physical interactions between DNA and histones by neutralization of the positive charge on lysine, reducing the electrostatic interactions with the DNA backbone (66,  67). A recent study using Debye-length Hamiltonian replica exchange molecular dynamics (D-HREMD) simulations illustrated that histone H3K27 acetylation leads to relaxed nucleosome positioning (35). This supports conventional concepts that nucleosome acetylation promotes transcription by increasing accessibility of the transcription activation machinery to the DNA. Histone acetyl-lysine is recognized by specific protein domains including bromodomains, YEATS (Yaf9, ENL, AF9, Taf14 and Sas5) domains and double plant homeodomain (PHD) fingers, each of which has distinct structure.

The global levels of histone acetylation in the nucleus are influenced by acetyl-CoA availability in cells. Therefore, glucose and mitochondrial respiration play critical roles in transcription regulation. In yeast, acetyl-CoA is utilized by the Spt-Ada-Gcn5-Acetyltransferase (SAGA) complex for the transcription activation of growth-promoting genes in the presence of glucose; however, in the absence of glucose, acetyl-CoA utilization is switched to mitochondrial ATP synthesis and formation of ketone bodies as alternative energy sources. Indeed, acetyl-CoA can be recaptured by deacetylation of histones on growth-promoting genes by the Rpd3 histone deacetylase complex (HDAC) (50,  76,  77).

The NAD⁺-dependent Sirtuins (SIRT) family contains seven class III histone deacetylases (3). Notably, SIRTs utilize NAD⁺ by activating the nicotinamide-ribose bond for nucleophilic attack in their deacylation reaction (78). This is different from the function of NAD⁺ as a redox cofactor. Increased cellular levels of NAD⁺ promotes Sir2-dependent gene repression in yeast (79). On the other hand, Sir2 suppresses genes associated with NAD⁺ biosynthesis (80). Thus, NAD⁺ and Sir2 are connected with a feedback loop. Reversible reactions between histone acetylation and deacetylation contribute to redox homeostasis through transcription regulation and protein modification involved in redox signalling.

In addition to acetylation, eight kinds of short-chain lysine (Lys) acylations on histones have been found. These include Lys propionylation (Kpr), Lys butyrylation (Kbu), Lys 2-hydroxyisobutyrylation (Khib), Lys succinylation (Ksucc), Lys malonylation (Kma), Lys glutarylation (Kglu), Lys crotonylation (Kcr) and Lys β-hydroxybutyrylation (Kbhb) (81). Acyl-group-specific transferases have not been identified. However, several acetyltransferases can also catalyse these alternative acylations (81). At least three HAT families: the Gcn5-related N-acetyltransferases (GNAT) family, the MOZ-, Ybf2/Sas3-, Sas2- and Tip60-related HATs (MYST family), and the p300/CBP HATs have been shown to catalyse Kpr (67,  81). Histone deacylation is similar to histone deacetylation, some of the SIRTs and Zn2⁺-dependent HDACs have been found to catalyse removal of acyl groups with some specificities (81,  82).

The recognition of Kpr, Kbu, Kcr and Ksucc in addition to Kac by bromodomains was comprehensively assessed by a crystal structure study using peptide binding (83). Interestingly, the preference of different acylations depends upon nutrition availability and determines which bromodomains can bind the acylation mark. A study using mice spermatogenic cells showed that both acetylation and butyrylation of histone H4K5 and H4K8 promote transcription; however, butyrylated H4K5 but not butyrylated H4K8 prevents binding of bromodomain 1 (BD1, essential binding domain) in the double bromodomain BET factor, Brdt (84). Since Brdt-BD1 directs the genome-wide replacement of histones (85), alternating acetylation vs butyrylation on H4K5 can change bromodomain-binding dynamics during spermatogenesis. Short-chain fatty acids have been suggested to be a source of acyl-CoA (86). Therefore, H4 acylation vs acetylation may alter upon a metabolic shift that changes the ratio of acyl-CoA/acetyl-CoA, or during a developmental event with an acyl-CoA requirement like during spermatogenesis (87).

One-carbon

One-carbon (1C) metabolism mediated by folate redistributes carbon groups from inputs (usually serine and glycine) in nutrients (88). Outputs of 1C metabolism, nucleotides, glutathione (GSH), S-adenosyl-L-methionine (SAMe) and other metabolites, maintain cellular redox status (88) (Fig. 3). SAMe donates methyl groups to methyltransferases including lysine methyltransferases (KMTs), protein arginine methyltransferases (PRMTs) and histidine Nτ-methyltransferases. Methyl groups are transferred to DNA, lysine (K)/arginine (R)/histidine (H) residues in histones, splicing factors and other proteins (54,  73,  89–93). Both histone lysine and arginine methylation are associated with either activation or suppression of transcription dependent on the positions of the amino acid in the histone N-terminal tail in higher eukaryotes (73,  94). Methylation of lysine residues occurs more frequently and is more abundant than methylation of arginine and histidine (89). Histone lysines can be mono-, di- or tri-methylated. Methyl-lysines are bound by distinct domains including chromodomain (CD), plant homeodomain (PHD) finger, Tudor, malignant brain tumour (MBT), proline-tryptophan-tryptophan-proline (PWWP), bromo adjacent homology (BAH), ankyrin repeats, WD40 repeats, ATRX-DNMT3A-DNMT3L (ADD) and zn-CW (95). There are three kinds of arginine methylation: monomethylarginine, asymmetric dimethylarginine and symmetric dimethylarginine (96). Tudor domains are the main readers of arginine methylation, and some of them bind both methyl-lysine and methyl-arginine (96).

Fig. 3.

The levels of methyl groups are regulated by coordination of nucleosomes, free histones and a phospholipid (PE). The methionine cycle and cysteine metabolism pathway are regulated by chromatin modification and transcription. In normal growth conditions, methyl groups from SAMe are utilized predominantly for phospholipid methylation and then nucleosome methylation in the nucleus. Upon amino acid depletion, methyl groups are still utilized by phospholipids and nucleosomes; however, further excess methyl groups are stored on free histones catalysed by histone MTs. Histone methylation preferences are ranked as H3K36me2/3 > H3K27me3 > H3K4me3. Thus, free histones function as a methyl sink to prevent alkylation damage. SAMe, S-adenosyl-L-methionine; SAH, S-adenosyl-L-homocysteine; HCY, homocysteine; MAT, S-adenosylmethionine synthase, also methionine adenosyltransferase; MT, methyltransferase; PC, phosphatidylcholine; PE, phosphatidylethanolamine; TAG, triacylglycerol.

Connections of 1C metabolism to the free radical scavenger system by free histone methylation and SWI/SNF are crucial for redox homeostasis. Methylation of a phospholipid, phosphatidylethanolamine (PE), is a major consumer of SAMe and promotes SAMe turnover for synthesis of cysteine and glutathione via transsulphuration in sulphur amino acid metabolism (Fig. 3). Sulphur metabolism functions as an oxygen-free radical scavenger (13) (Fig. 3). Interestingly, histone methylation, with a preference of H3K36 > H3K79 > H3K4, can also function as a methyl sink (e.g. consumes SAMe) to avoid detrimental damage from an excess of methyl groups upon loss of PE in yeast (97) (Fig. 3). SWI/SNF regulates transcription activation of sulphur metabolic genes by recruiting the sulphur-sensing TF Met4 at proper gene loci in yeast cultured in nutrient-rich media (98). Finally, as described above, destabilized MSH6 by dysfunction of sulphur metabolism has been suggested to underlie Fragile X disorders (13,  54) (described in the ‘Nucleotide Metabolism’ section).

In mice, a methionine adenosyltransferase (MAT) isozyme, MATIIα, promotes H3K9me3 at the COX-2 (Cyclooxygenase-2) loci by its interaction with the SETDB1 H3K9 methyltransferase (99) (Fig. 3). MATII synthesizes SAMe from methionine and ATP (100). Importantly, MafK is a transcription corepressor, which requires the catalytic activity of MATII for the transcription repression of genes including HO1 (heme oxygenase-1). These target genes are oxidative stress-inducible genes (101). Thus, MATIIα suppresses oxidative stress responsive genes when nutrition is sufficient so that sulphur metabolism is not urgently required. It is not yet known whether MATII changes the methyltransferase partner(s) or changes target genes in conjunction with MafK in the absence of glucose.

Notably, cysteine(s) in nucleosomal histone H3 has been shown to be glutathionylated. Glutathionylation of H3 opens nucleosome structure (102). GSH is a major sulphur-containing antioxidant (3,  97,  103). It remains unclear whether antioxidant activity of glutathionylation on cysteine(s) affects modification of nearby lysines (i.e. methylation or acetylation).

Oxygen

Oxygen availability directly modifies the catalytic activity of JmjC lysine demethylases (KDMs), which remove methyl groups from lysines within histone tails (104,  105). JmjC are 2-oxoglutarate (2OG)/α-ketoglutarate-dependent oxygenases (2OGXs) and their reaction requires molecular oxygen and iron (II) (104–106). The oxygen dependency of H3K9me3 demethylation by KDM4E/JMJD2E was initially found in vitro (107). Subsequent studies revealed that demethylation of H3K9me2/3 and H3K36me2/3 by KDM4A/JMJD2A (108) observed under normoxia (21% O₂) is suppressed upon hypoxia (1.0% O₂) (105). KDM5A/JARID1 activity on H3K4me3/H3K36me3 and KDM6A/UTX activity on H3K27me3 follow the same patterns (109,  110). 2OGX activity is dysregulated in cancer due to increased levels of metabolites from the tricarboxylic acid (TCA) cycle and presence of other metabolites, such as (R)-2-hydroxyglutarate (2HG) (111,  112). Hypoxic environments at the inside of solid tumours may reduce 2OGX activity. Interestingly, hydroxylation of hypoxia inducible factor 1 subunit alpha (HIF1A) by prolyl hydroxylase domain enzymes destabilizes HIF1A and requires 2OG (113). 2-oxo acids have been suggested to selectively inhibit 2OG oxygenase activity (112). In addition to over expression of KDMs in cancer, such as KDM3B/JMJD1B, there are additional mechanisms by which KDMs contribute to oncogenesis (114). Thus, compounds, which are 2-oxo acid derivatives, may be applicable as enzyme inhibitors for treatment of cancers that have aberrant JmjC demethylase activities (115).

Spatiotemporal Regulation of Chromatin Events Enables Cells to Minimize Cellular Energy Consumption

The involvement of liquid–liquid phase separation (LLPS) in transcription regulation and its possible effects on genome remodelling have been proposed (5,  116,  117). Classical phase separation of a homogeneous mixture is a process where the non-miscible two components separate by either nucleation and growth or spinodal decomposition until they form two different phases (118). There is viscoelastic phase separation that creates the dynamic asymmetry between two phases to form a network-like pattern of the slower phase composed with high-molecular-weight unstructured polymers (e.g. messenger RNA) (119,  120).

Some cis-regulatory elements, enhancers, may be located far away from transcription start sites (TSSs). There is a class of enhancers called super enhancers (SEs), which are clustered with a high density of transcription and chromatin regulators. These can include RNA Pol II, general TFs, the bromodomain containing transcription coactivator p300 histone acetyltransferase, bromodomain containing 4 (BRD4), cohesin, the mediator complex, bidirectional transcripts (enhancer RNAs (eRNAs), a class of non-coding RNAs), acetylated histone H3K27 and multiple TFs (117,  121–124) (Fig. 1B). These phenomena have been described as a phase separation model, and the effects of the phase separation on transcription have been studied. An increased concentration of transcriptional activators at SEs correlates with sigmoidal increases in the transcripts levels of SEs-corresponding genes (117,  121). In addition, expression of eRNAs transcribed from enhancer regions (123) tightly correlates with transcription activation of corresponding active SEs (117). For instance, rapid and drastic de novo clustering of chromatin activators including accumulation of nuclear factor-kappa beta (NF-κΒ) (a TF) at enhancers (pre-existing enhancer motifs) creates inflammatory SEs upon tumour necrosis factor alpha (TNFα) stimulation and dominantly promotes inflammatory gene transcription. Meanwhile H3K27 acetylation is also increased at basal-endothelial-SEs in human endothelial cells (125). This might explain how asymmetric partitioning of different high-molecular-weight polymers can regulate the transcription levels.

Regarding the creation of phase separation, hyper H3K27ac at SEs suggests nucleosomes at SEs are relaxed. This is contrary to the notion that compaction of chromatin might execute better contributions to increases in the density of macromolecules at SEs. However, it may be explained by a model of enhancer-promoter contacts from 3C experiments suggesting that active nucleosome-free regions are extended above the chromatin surface (35) (described above) (Fig. 1B). Considering the 3C experiments, the total amount of nucleosome substrates for p300 must depend on the total number of de novo nucleosomes located on the chromatin surface at cis-regulatory elements. However, increased BRD4, whose bromodomains bind acetyl histone H3 and H4 in nucleosomes promiscuously (126–128), suggests that BRD4 mediates acetylation of nucleosomes over a broad region beyond those at cis-regulatory elements in SEs (Fig. 1B). A study of inflammatory SEs in human endothelial cells revealed that BRD4-bromodomains are crucial for the enhancer remodelling in the transition to inflammation (125), suggesting that nucleosome acetylation contributes to enhancer remodelling. Notably, the binding of BRD4 to p300-acetylated NF-κΒ promotes the transcription of NF-κΒ target genes in human cells (129). Another example of acetylation of TFs observed in SEs is MYC. MYC interacts with SEs through long-range interaction with eRNAs (130,  131) and is acetylated by p300 and GCN5 in rat cells (132). Furthermore, eRNA directly interacts with the BRD4 bromodomain (133). Thus, acetylated non-histone molecules including TFs and eRNA may facilitate BRD4 recruitment, resulting in increases in the concentration of macromolecules (Fig. 1B). Importantly, many biomolecules participating in phase separation including p300 and NF-κΒ contain intrinsically disordered regions (IDRs). The multivalency of IDRs, which is the simultaneous binding of multiple ligands on one biological or chemical entity to multiple binding sites, often weak or dynamic, on another, can provide a platform for phase separation (134–136). In terms of energy homeostasis, phase separation may be an efficient method to achieve crucial events with minimum energy consumption within a minimum time period.

The contacts between distinct enhancers over long distances are established through nucleosome-free regions (35) (described above) (Fig. 1A). In vitro studies have found that BRD4 binding to acetylated lysine of human mononucleosomes requires flanking DNA (128). As BRD4 prefers binding nucleosomes with flanking DNA, it is possible that de novo creation of SEs may be initiated by BRD4 nucleosome binding and p300 nucleosome acetylation at the border of nucleosome-free regions before spreading acetylation. It may also require ATP-dependent chromatin remodelling to create nucleosome-free regions (Fig. 2).

Future Perspective

Maintaining nucleosome-free regions, for instance blocking of nucleosome sliding via binding of TFs (Fig. 2), can be a key mechanism of chromatin dynamics for gene regulation. It also suggests a novel role of TFs in transcription regulation. Systematic analysis combining data on the binding affinity (dissociation constant) of TFs with natural sequences of TFs binding regions within chromatin and 3C analysis may predict the probability of enhancer–enhancer interactions and the probability of SEs. In vitro nucleosome remodelling assay can prove these sites and factors as barriers. There are also critical factors that change nucleosome structure, such as RNA Pol II (137,  138) and natural G-rich sequences (139). Understanding the features of the interplay between chromatin dynamics and chromatin modifiers throughout the transcription activation process will reveal further mechanisms. Additionally, the energy requirements for a single transcription cycle may be measured. This will help quantify the relationships between chromatin dynamics, transcription and cellular metabolism.