Aere perennius: how chromatin fidelity is maintained and lost in disease

Abstract

Multicellular organisms arise from a single genome template in the zygote, necessitating the cells of the developing embryo to up- and downregulate specific genes to establish and maintain their identity. This template is maintained, propagated, and interpreted as chromatin, a polymer of nucleic acids and associated structural and regulatory proteins. Recent genome-wide surveys documented a wealth of disease-associated mutations in chromatin factors, indicating their fundamental significance and potential for therapeutic targeting. However, chromatin factors exist in a complex balance, with a single deficiency often leading to pleiotropic downstream effects. Here, we review the mechanisms of chromatin regulation and partitioning, highlighting examples of how these processes are altered in human diseases. We argue that loss of chromatin fidelity, both locally at specific genes and regulatory elements, and globally at the megabase-scale, contributes to many pathological states and may thus represent an intriguing target for corrective interventions.

Graphical Abstract

Graphical Abstract.

Introduction

How the zygote gives rise to multiple cell types during development, while preserving largely identical genome across lineages has been studied since the dawn of modern developmental biology [1]. Both biochemistry and molecular biology rapidly converged on gene promoters as the defining regulatory features of the genome and initiated decades of discovery in their respective fields, including the identification of “master regulator” transcription factors (TFs), lineage-defining regulators sufficient to drive gene expression programs toward specific cell fates. From the discovery of MyoD by the Weintraub lab in the late 1980s as a key factor in myogenesis [2, 3], to the subsequent identification of the four factors sufficient for pluripotency reprograming by Takahashi and Yamanaka [4], TFs are rightly regarded as key drivers in cell fate decisions. A recent survey of the human genome suggests that ∼1,600 out of ∼23,000 genes encode TFs [5], allowing for nearly infinite combinations, limited only by the availability of specific TF binding sites on the DNA and the presence of the TFs themselves. However, even the groundbreaking Yamanaka experiment offers two interpretations: on the one hand, ectopic TFs dramatically altered cell fate in the experiment, thus fulfilling the sufficiency criteria. On the other hand, only an estimated 0.02% of fibroblasts expressing the four factors successfully converted into stem cells [4]. This together with later observations that additional factors alter cell fate plasticity [6, 7] suggests that the promoter-centric paradigm does not sufficiently account for all observations, and additional mechanisms ensuring the stability of transcriptional programming must exist.

The genome is propagated, maintained, and interpreted within the context of the chromatin fiber—a polymer complex whose main components are DNA and histone proteins. The ∼146 base pairs of DNA wrapped around the octamer of histone proteins, H3/H4 tetramer and two H2A/H2B dimers, form the nucleosome core particle, the basic repeating unit of chromatin [8–10]. Other factors, including linker histones (H1) and high mobility group (HMG) proteins, as well as non-coding RNAs, further contribute to chromatin fiber compaction [11–13]. Crucially, while the DNA sequence (with a few notable exceptions [14–16]) is largely invariant and forms the basis of Mendelian inheritance, dynamic modifications of the chromatin fiber, as well as the degree of compaction and three-dimensional topology within the nucleus, are both variable across cell lineages and instructive to the gene expression [17, 18]. This layer of non-sequence-dependent regulation in the cell is often referred to as epigenetic; however, the strict definition of the term implies transgenerational inheritance of acquired traits, which is the subject of some controversy [19], and we favor the use of “chromatin biology” over “epigenetics” to denote the broader field of study.

For the information to be encoded in the genome in a sequence-independent manner, three core conditions must be met. First, a robust mechanism for establishing such modifications must be present. These factors have been extensively studied since the transcriptional activator GCN5 was shown to be a histone lysine acetyltransferase [20], and dozens of post-translational modifications in all components of the chromatin fiber have been characterized in detail [21]. Enzymes implicated in establishment and removal of specific modifications are commonly referred to as “writers” and “erasers,” respectively. Importantly, these are frequently located downstream of the sequence-specific transcriptional activators and repressors, highlighting the close connection between sequence-dependent and -independent regulation [22, 23]. Second, these modifications must be interpreted by cellular machinery. Indeed, a vast array of “reader” domains has been identified since the discovery of p300/CBP-associated factor (PCAF) bromodomain interaction with acetyl-lysine on H4 N-terminal tail [24]. Combinations of multiple reader domains within a single protein or protein complex enable logical connections (“AND” and “NOT”) for specificity, local separation of function, and redundancy [25, 26]. Third, maintenance mechanisms are required to ensure that specific information is retained throughout development. These are the more enigmatic of the three mechanisms: while the DNA structure itself provides clues to its semiconservative replication, the maintenance of specific chromatin modifications in situ relies on many distinct mechanisms that vary depending on the cell type (e.g. mitotically active versus non-dividing cells), substrate (e.g. modifications of DNA or histone proteins), and chemical nature of modification. Thus, while the role of the DNA repair mechanism became clear soon after the Avery experiment [27] and remains the focus of intense investigation [28], the mechanisms ensuring fidelity of chromatin modifications are still less well understood.

Recent methodological advances open exciting opportunities to understand the chromatin maintenance mechanisms in development and disease. Here, we review select examples that contribute to the stability of transcriptional programing, focusing on the molecular basis of broad domain establishment and maintenance, and their relationship to the topological compartmentalization of the nucleus in three-dimensional (3D) space. We then discuss how these processes relate to human disease, particularly aging, neurodegenerative, and metabolic disorders.

Chromatin modifications are established and interpreted as ensembles

Genomic regions with specific activity and function, including architectural chromosomal elements and cis-regulatory modules, as well as transient events such as DNA replication and repair, are demarcated by distinct chromatin modifications. The incorporation of histone variants into the nucleosome core particle, DNA modifications, and post-translational histone modifications have been extensively studied in disease context, and we will refer the reader to many excellent reviews published in the recent years [29–31]. Here, we will only briefly describe recent studies of how distinct states of chromatin are established and maintained in the context of their nuclear surroundings, focusing mainly on developmental regulation by Polycomb group (PcG) complex.

A convenient classification of factors operating in chromatin into functionally distinct “readers” and “writers” of post-translational modifications is complicated by the close relationship between the two: writers frequently incorporate reader subunits and domains, and/or are sensitive to the modification state of the neighboring amino acid residues. A paradigmatic example of such a relationship is the PcG complex, which is best known for its developmentally regulated repressor function [32]. In humans, core Polycomb Repressor Complex 2 (PRC2) contains the critical histone H3 lysine 27 (H3 K27) methyltransferase “writer” subunit Enhancer of Zeste Homolog 1 or 2 (EZH1/2), two structurally critical factors: Suppressor of Zeste 12 homolog (SUZ12) and Retinoblastoma-Associated Protein 46/48 (RBAP46/48, also known as RBBP4/7), and an essential “reader” component Embryonic Ectoderm Development (EED) that recognizes methylated H3 K27. This arrangement enables a feed-forward mechanism in which H3 K27 methylation, established by PRC2, in turn recruits PRC2 for processive enzymatic activity. This is additionally supported by allosteric activation mechanisms, including via PRC2 subunits JARID2 and AEBP2 or PALI1 [33, 34] (Fig. 1A). Remarkable structures of PRC2 in complex with the chromatin substrate reveal how the EED “reader” anchors the core complex while EZH2 methyltransferase extends to the adjacent nucleosome [35]. Together with the single-molecule studies of Polycomb processivity [36], these results suggest that PRC2-dependent spreading of H3 K27me3 may be sensitive to the distance between neighboring nucleosomes and is thus regulated by pre-existing state of the chromatin substrate.

Figure 1.

Mechanisms of genome partitioning by Polycomb Group complexes. (A) Core PRC2 complex consists of four subunits, EED and EZH1/2 representing “reader” and “writer” modules for H3 K27 methylation. Sequence context of H3 tail is shown. Subunits associated with PRC2.1 and PRC2.2 variants are listed. (B) Two major PRC1 complexes are characterized by distinct subunit composition: non-canonical PRC1 (ncPRC1, left) contains H2A K119 “reader” RYBP/YAF2, and canonical PRC1 (cPRC1, right) contains one of several Chromobox proteins (CBX2/4/6/7/8) associated with H3 K27me3. Both complexes include RING1 E3 ubiquityl ligase “writer” module, with ncPRC1 characterized by higher enzymatic activity, and distinct complement of additional factors, anchored by PCGF subunits, listed in corresponding boxes. Sequence context of H3 and H2A tails are shown. (C) Recruitment and spreading of PcG domains in the vertebrate genome: left, unmethylated CpG islands are recognized by ncPRC1.1 subunit KDM2B, leading to initial H2A K119 ubiquitylation (orange circles) recognized by PRC2.2 via JARID2 subunit; methylated H3 K27 is in turn recognized by cPRC1 for compaction and/or further H2A ubiquitylation, and processive spreading of H3 K27 methylation is established by coordinate EED “reader” and EZH2 “writer” activity of PRC2; right, unmethylated CpGs associate with PRC2.1 subunit MTF2; initial H3 K27 methylation in turn recruits cPRC1 via CBX subunits, and PRC2 recruitment is propagated via H2A K119 recognition via JARID2, or H3 K27me3 recognition via EED. (D) Positive feedback regulation of PcG is balanced by incorporation of negative regulator subunits (EZHIP, left), incorporation of histone modifications that prevent H3 K27 methylation in cis (green and magenta circles correspond to H3 K36 and H3 K4 methylations; other modifications, including H3 phosphorylations, may act in similar fashion [53], middle), and activity of DNA methyltransferases, recruited in part by H3 K36 di- and tri-methylation by NSD1/2 and SETD2 enzymes, respectively (right). Direct stimulation of PRC2 by H1 incorporation is thus reinforced by reducing NSD1/2 activity in chromatin [6, 52].

PRC1 is the second component of the PcG system and also possesses distinct “reader” and “writer” functions characterized by a number of self-reinforcing mechanisms. All PRC1 subcomplexes share the RING1 E3 ubiquityl ligase, which catalyzes H2A K119 monoubiquitylation. “Canonical” PRC1 contains one of five H3K27me3-binding Chromobox (CBX) 2/4/6/7/8 subunits, a self-assembling Polyhomeotic (PHC) 1/2/3 subunit, and PHC-binding PCGF subunit 2 or 4. “Non-canonical” PRC1, on the other hand, is characterized by the incorporation of the RYBP or YAF2 subunit, as well as PCGF 1/3/5 or 6, all of which stimulate H2A ubiquitylation by RING1. Additional PRC1 subcomplexes with distinct cell-specific accumulation profiles and developmental functions are known (reviewed in [37], Fig. 1B). Recent structural study of non-canonical PRC1 in complex with a dinucleosome substrate shows a mechanism remarkably similar to that of PRC2 processivity: while the paired EED “reader” and EZH2 “writer” mediate trans-nucleosomal engagement in PRC2, RYBP binds to H2A K119Ub, while RING1 deposits ubiquitylation on the adjacent nucleosome, thus similarly allowing for processive spread of Polycomb domain [38].

Several non-exclusive models of PcG recruitment in vertebrates have been proposed, employing both PRC1- and PRC2-dependent targeting mechanisms. PRC1-dependent targeting is initiated by the recruitment of variant PRC1 to unmethylated CpG islands, followed by demethylation of H3 K36 by KDM2B and RING1-dependent H2A K119 ubiquitylation. This modification is recognized by JARID2, a regulatory subunit of PRC2 that in turn stimulates its K27me3 activity (reviewed in [39]). PRC2-dependent targeting is established via MTF2, a subunit of Polycomb-like (PCL) family, which directly recruits PRC2 to primary target sites defined by high CpG density, low 5-methyl-cytosine (5-mC) content, and distinct DNA shape [40] (Fig. 1C). Methylation of histone H3 K27 is then recognized by canonical PRC1 complexes via Chromobox (CBX) “readers,” of which CBX2, 4, 6, and 8 are associated with high nucleosome compaction activity via a distinct Compaction and Phase-Separation (CaPS) domain, whereas CBX7 appears to completely lack CaPS [41]. Since expression of the CBX subunits varies depending on tissues and developmental stages, other mechanisms of Polycomb-dependent repression have been proposed, including PRC2-dependent compaction via EZH1-dependent dimerization [42]. Further, PRC1 and PRC2 have both overlapping and unique functions in developmental gene regulation: loss of PRC2 is dispensable for maintaining the pluripotency of embryonic stem cells in culture [43], while loss of PRC1 causes immediate loss of pluripotency due to derepression of the differentiation program [44]. Although both PRC1 and PRC2 appear together in the last eukaryotic common ancestor (LECA), the complexes evolved independently, suggesting that the convergence of regulatory function may be a secondary acquisition [45].

While several elegant positive feedback mechanisms explain PRC1/2 propagation, what negative regulatory mechanisms limit PcG activity and how the boundaries to PcG-demarcated domains are established, are critical questions that yet remain to be fully deciphered. The accumulation of PRC variant complexes with reduced enzymatic activity, exemplified in the extreme case by incorporation of the autoinhibitory PRC2 subunit EZHIP/CXorf67, provides a global, cell-type specific mechanism [46–48]. Locally, endogenous lysine 36 (K36) methylation on histone H3, another broadly distributed core histone modification, appears to play an important role (Fig. 1D). Somatic loss of H3K36me in rare tumors expressing dominant negative K36M “oncohistone” leads to expansion of PRC2-dependent K27me and loss of PRC1-mediated repression by “reader” dilution [49]. A similar mechanism drives a subset of head and neck squamous cell carcinomas, associated with either loss of H3 K36-specific dimethylase NSD1 or, rarely, H3 K36M mutations [50]. Conversely, expansion of H3 K36me2 by the MMSET (NSD2) methyltransferase alters H3 K27me landscape in multiple myeloma, restricting PRC2 activity to narrow regions that undergo stronger repression [51]. Aspects of this regulation are reflected in a class of lymphomas driven by missense mutations in H1 linker histones: global loss of H1 in germinal center B cells broadly increases chromatin accessibility, with concomitant expansion of H3 K36me2 and contraction of H3 K27me-demarcated domains. This, in turn, drives derepression of early stem cell programs in terminally differentiated B cells, with resulting transcriptional profiles remarkably similar to both multiple myeloma and hematopoietic stem cells [6]. While H1 incorporation stimulates PRC2 activity and counters NSD2 function in vitro [6, 52], the hierarchy of events in chromatin upon H1 loss in vivo remains unknown. In addition to K36 methylation, other H3 modifications regulate PRC2 activity, including serine 28 phosphorylation, representing a well-characterized methyl-phos “switch” [53], lysine 4 methylation, and tail arginine methylations [54]. Interestingly, cancer-associated missense mutations in H3 R2 and R26 inhibit PRC2 activity in cis [55], thus potentially destabilizing the transcriptional repression established by broad Polycomb-dependent domains during development.

DNA methylation plays both direct and indirect roles in Polycomb regulation, restricting the PcG seeding to unmethylated CpG islands and limiting spread to unmethylated regions through its connection to H3 K36 methylation (Fig. 1D). The latter mechanism is particularly intriguing, as vertebrate de novo DNA methyltransferases DNMT3A and DNMT3B are distinctly recruited to di- and tri-methylated H3 K36, respectively [56, 57]. Further, an alternative isoform of DNMT3A was recently reported to contain a ubiquitin-dependent recruitment region (UDR) within its N-terminus; while targeting of DNMT3A to H3 K36-methylated regions takes precedence under normal conditions, both global loss of K36me2 or disease-associated missense mutations re-target DNMT3A to PRC1-dependent H2A K119-ubiquitylated regions [58]. While the physiological significance of this mechanism in normal development is yet unclear, this cryptic UDR domain is a conserved feature of a wild type protein, and ability to target de novo DNA methyltransferase activity to Polycomb domains may provide a much-needed mechanism to interrupt the feed-forward regulation.

Hemimethylated CpG dinucleotides are re-methylated upon DNA replication by maintenance methyltransferase DNMT1 and its cofactor Ubiquitin-like PHD and RING finger domain-containing protein 1 (UHRF1), while de novo methyltransferases DNMT3A and 3B, together with the catalytically inactive DNMT3L or DNMT3B3 cofactors, as well as UHRF1 [59], establish new methylation patterns, and ensure fidelity of DNMT1-dependent remethylation [60]. Importantly, histone modifications represent a key role in targeting DNMTs to distinct genome compartments, thus ensuring the fidelity of methylation [60]. A remarkable recent structural study of DNMT3A2/3B3 heterotetramer with the dinucleosome substrate demonstrate both the targeting and allosteric mechanisms of the complex, where the enzymatic activity is regulated by simultaneous presence of H3 K36 methylation and absence of K4 methylation, and is restricted to the linker DNA outside the nucleosome core particle [61]. As DNMT3 complex prefers short internucleosomal linker DNA, these studies echo the observations that actively transcribed and H3 K36 methylated regions are characterized by shorter internucleosomal distances, while longer nucleosome repeat length in H3 K27-methylated chromatin allows for linker histone H1 incorporation and more efficient higher-order compaction [62]. Loss of H1 indeed reduced the nucleosome repeat length and increased the global H3 K36 dimethylation [6, 63]; whether DNMT3A activity was altered remains to be tested.

Fundamentally, the regulatory circuit between Polycomb complexes, DNA methyltransferases, and their cognate substrates functions as a neural network, with multiple nodes represented by specific modifications, edges corresponding to positive or negative regulation, and directionality imposed by biochemical constraints. Positive feedback regulation, prevalent in enzymes operating to establish broad chromatin domains, necessitates such constraint. In fact, in silico modeling of epigenetic modifications established by “reader-writer” complexes demonstrate that negative regulation—by limiting the enzyme concentration, or via negative regulation of its activity—is essential for the stable maintenance of heritable patterns [64]. Account of the negative regulators is thus essential to better understand how chromatin landscape is established and maintained.

Genome organization is maintained in three dimensions

Distinct chromatin modifications co-segregate within the three-dimensional nuclear space, mirroring the organization of genomic DNA. Both histone modifications established by “reader-writer” machinery, and non-histone proteins associated with regulatory elements partition into distinct compartments, mediated by a combination of localized enzymatic activity anchored by “reader” modules [25] and charge-based interactions between intrinsically disordered regions [65]. Importantly, these two mechanisms are not mutually exclusive, as exemplified by presence of both H3 K27 methyl-“reader” chromodomain and phase-separating CaPS domain in several CBX subunits of PRC1 complex [41], and observations that association of yeast Heterochromatin protein 1 (HP1) homolog Swi6 with H3 K9-methylated nucleosome alters the nucleosome core conformation to facilitate partitioning of HP1-bound chromatin fiber into phase-separated condensates [66]. These interactions are typically confined to megabase-scale topologically associated domains (TADs) anchored by architectural proteins, in mammals represented by the zinc finger insulator protein CTCF and Cohesin complex, implicated in chromatin loop extrusion in interphase chromosomes [67]. More fine-grained interactions between cis-regulatory elements and target promoters are established by additional factors, including Mediator complex [68].

With average enhancer–promoter distance in vertebrates estimated at 20–50 kb, and occasionally exceeding hundreds of kilobases [69], how the regulatory elements find their targets is a fundamental question. Remarkably, not only the arrays of regulatory elements are brought to target promoters, but co-transcribed genes themselves are coupled in three-dimensional space, assembling into “topological operons” [70]. Curiously, while acute depletion experiments demonstrated near-complete disappearance of TADs upon CTCF or Cohesin loss [71] only modest effects on enhancer–promoter communication and resulting gene expression were observed under steady state [71, 72]. These observations suggest that, once established, enhancer–promoter communication is maintained independent of TAD anchors. Further, multiply-rearranged “balancer” chromosomes widely used in fruit fly genetic studies demonstrate distinct TAD organization yet remarkably few inappropriate enhancer–promoter activations [73]. Together, these observations suggest that TAD organization may be redundant with other mechanisms of gene regulation and is not the sole driver of transcriptional fidelity.

Lysine methylations of the histone H3 N-terminal tail have been extensively implicated in 3D genome organization. While both H3 K9me and K27me were extensively documented at lamina-associated domains (LADs) [74, 75], recent studies integrating bisulfite conversion into HiC workflow to map protein–DNA interactions (LIMe-HiC) demonstrated antagonistic role for H3 K27me3 in LAD organization [76]. Loss of PRC2 resulted in bipartite reorganization of facultative heterochromatin: one subset of genes sustained repression in absence of PRC2 by moving into LADs and gaining H3 K9me3, while another subset retained at the nuclear interior was derepressed. These studies highlight multiple compensatory mechanisms of nuclear organization, with simultaneous loss of several factors required for fully penetrant phenotype. Additional nuclear landmarks were integrated into the functional analyses of genome organization using optical microscopy coupled with analyses of histone modifications and replication timing [77]. In these studies, speckle attachment regions were identified as additional zones correlating with genome activity, and dynamics of chromatin domains were determined by both histone modifications and nuclear lamina composition.

Chromatin conformation is linked to cell specialization, with closely related cell types demonstrating distinct three-dimensional organization. Extensive reorganization of chromatin topology upon activation of specific programs in pyramidal neurons in the hippocampus and dopaminergic neurons in ventral tegmental area preferentially impacted genes implicated in synaptic plasticity and addiction, respectively [78]. Intact function of Topoisomerase 1 (TOP1) is required for such reorganization, and long genes are preferentially sensitive to TOP1 inhibition [79]. Remarkably, many of such genes are implicated in neurodevelopmental disorders. Cell-specific chromatin (mis)organization may thus be both the underlying cause of disease and potential target of therapeutic interventions.

Chromatin is maintained across the cell cycle

Every nucleated human cell contains millions of nucleosomes, each presenting a barrier for all DNA-templated processes, and requiring a reassembly mechanism to ensure preservation of epigenetic information through cell divisions. Whereas semiconservative nature of DNA replication provides a robust template to establish methylation patterns at palindromic CpGs, the problem presented by nucleosome transfer is orders of magnitude more complex, both due to the vast number of histone variants and modifications, and a lack of sequence specificity encoded in the nucleosome structure itself. Many chaperones and histone modifications involved in local recycling of nucleosomes are documented, in particular within genetically tractable yeast models, from initial observations that labeled nucleosomes are largely retained at their loci during replication [80], to recent studies taking advantage of AI structure predictions to identify new chaperones and surfaces involved in this process [81–83].

Eukaryotic replication machinery, referred to as Replisome Progression Complex (RPC), contains molecular mechanisms for both DNA synthesis and chromatin assembly, including CMG helicase complex containing Cdc45, Minichromosome maintenance (Mcm)2–7 hexamer, and go ichi ni san (GINS) complex of Psf 1, 2, 3 and Sld5 subunits, as well as Pol α-primase coupled to CMG by Ctf4 trimer, checkpoint effector Mrc1, Tof1-Csm3 pause complex, H2A/H2B histone chaperone Facilitates Chromatin Transcription (FACT), and Topoisomerase I [84–86]. Single-stranded DNA is protected by Replication protein A (RPA) assemblies [87], and DNA Polymerases δ and ϵ are clamped to lagging and leading strands via Proliferating Cell Nuclear Antigen (PCNA) trimer [88], which directly interacts with DNMT1 for remethylation of nascent DNA strands [89, 90]. Anti-silencing factor 1 (Asf1) is associated with Chromatin Assembly Factor 1 (CAF-1) complex, and together with Rtt106 orchestrates reassembly of H3/H4 tetramer following replication [91, 92]. Another H3/H4 chaperone NPM1 is recruited to RPC via Mcm2, and recruits PRC2 complex to re-establish the integrity of H3 K27me-demarcated domains upon nucleosome reassembly [93] (Fig. 2). Importantly, even perfectly efficient redistribution of old histones between the leading and lagging strands would result in a 2-fold dilution of all pre-existing modifications apart from ones established on unincorporated histones; the cell-cycle specific regulation of maintenance “writers” is thus of significant interest.

Figure 2.

Chromatin memory at the replication fork. Replication fork is shown progressing right to left, with lagging strand on the top and leading strand on the bottom. Parental histones are shown in light purple, with pre-existing post-translational modifications in dark purple. Nucleosomes are destabilized by FACT and parental H3/H4 tetramers are passed onto MCM2 and Claspin for symmetric recycling between the leading and lagging strands. Newly synthesized histones (white) are chaperoned by ASF1, NAP1 and CAF1 on both strands to maintain the nucleosome density. Histone modifications are reinstated by processive “reader”-“writer” activity of the cognate enzymes (light purple, only PRC2 is shown for clarity). DNA methylation is reestablished by maintenance methyltransferase DNMT1 via recognition of hemimethylated CpGs. Histone chaperones and DNMT1 operate on both strands but are omitted from lagging or leading strand for clarity.

How combinations of histone modifications regulate the engagement of replication machinery, and whether the decision to recycle paternal histones locally or replace the nucleosome with newly synthesized and thus minimally modified histones is driven by local chromatin landscape, are among the key questions in chromatin biology. Retention of both H3 K4 and K27 trimethylation post-replication was documented by a combination of chromatin immunoprecipitation (ChIP) with thymidine analog EdU pulse labeling to distinguish newly synthesized strands after replication (chromatin occupancy after replication, ChOR-Seq) [94]. In a complementary approach, local nucleosome labeling by biotin ligase fused to catalytically inactive Cas9 (dCas9) followed by extensive ChIP profiling documented local recycling of biotinylated histones at transcriptionally silent, but not actively transcribed regions [95]. Importantly, S-phase arrest blocked histone exchange at actively transcribed genes, indicating that DNA replication is the sole major contributor to nucleosome exchange, at least in the time span of a single cell cycle. Further, switching the gene activity rapidly altered the nucleosome retention profile, with induced genes losing parental nucleosomes within a single cell cycle—arguing for a local, and likely histone modification-dependent, mechanism of nucleosome eviction and reassembly [95]. A follow-up to these studies identified Nucleophosmin (NPM1) bridging the Mcm2 subunit of the replisome to PRC2, with acute NPM1 depletion resulting in loss of H3 K27me3, exacerbated by treatment with PRC2 inhibitors [93]. Mcm2 additionally interacts with Mrc1, which ensures symmetric distribution of parental H3/H4 tetramer between the leading and lagging strands [81, 82]. These components are remarkably conserved from yeast to vertebrates, and mutations in Mrc1 ortholog CLASPIN were recently linked to the defect in symmetric histone inheritance in mouse embryonic stem cells [83]. CryoEM structural studies of endogenous yeast replisomes isolated in early S phase demonstrated sequential disassembly of the histone octamer into H2A/H2B dimer, and hexamer containing H3/H4 tetramer core and residual dimer of H2A/H2B, which then gets chaperoned to the lagging strand by Mcm2 and Tof1, suggesting that at least some of the H2A/H2B modifications or variants may be more stably retained in situ than previously assumed [96]. Molecular dynamics simulations suggest that additional parameters of the leading and lagging strand, including DNA bending and amount of RPA, contribute to recycling of parental histones [97].

Transfer of parental H3 further extends to H3.3 variant, enabling variant-specific genome demarcation to persist through replication. This occurs independent of HIRA and is modestly reduced by DAXX deletion, thus indicating that H3.3 recycling is largely independent of variant-specific chaperones [98]. In contrast, inheritance of the more divergent centrosome-specific H3 variant CENP-A requires variant-specific chaperone HJURP at the replisome via association with Mcm2 subunit of the helicase [99], highlighting that distinct factors are recruited to the replisome to operate in specific genomic regions. Further, recycling of parental histones may be related to distinct replication timing between the transcriptionally active and repressed compartments [100, 101]. Temporal segregation of histone variants and modifications thus creates two pools of parental histones, further contributing to partitioning of distinct genome functions. As origin synchronization in vertebrates is established in early embryogenesis during zygotic genome activation (ZGA) [102], the first three cell divisions occurring prior to ZGA would lack this mechanism and thus may be uniquely suited to erase the nucleosome-dependent remnants of parental epigenetic memory, coincident with rapid erasure of DNA methylation post-fertilization [103]. Likewise, cellularization of a Drosophila embryo is achieved through fourteen synchronized cycles of mitotic divisions with S phase as short as 4 min [104]; as such, no distinction between “early” and “late” replicating chromatin yet exists, and histones are replaced near exclusively with maternally loaded copies carrying no position-specific information.

Asymmetric division with capacity for both self-renewal and differentiation is a fundamental property of tissue stem cell populations. Remarkably, mouse embryonic stem cells engineered to carry Mcm2 mutation asymmetrically redirecting parental histones to the leading strand (MCM-2A) demonstrate reduced differentiation competency, degradation of H3 K9me3-demarcated heterochromatin domains, and transcriptional signatures associated with two-cell-like state [105]. Yet while breaking the symmetry of leading and lagging strand distribution alters local retention of parental histones this mechanism does not account for asymmetric inheritance of the old or new histones globally, as equivalent amounts of DNA are synthesized from leading and lagging strands during each S phase.

Global asymmetry in histone inheritance was first documented in Drosophila male germline, where germline stem cells are anchored at the somatic niche and divide asymmetrically to produce one self-renewing stem cell and one differentiating gonialblast. Using genetically encoded labels, the authors documented preferential retention of parental (old) replication-dependent H3 in the stem cell pool, while differentiating gonialblast acquired newly synthesized histones [106]. Curiously, H3.3 variant, associated with active regulatory regions, was distributed symmetrically, which may be attributed to a distinct abundance of variant-specific modifications that contribute to this process [107]. Indeed, H3 threonine 3 (T3) phosphorylation by Haspin was identified in the same system as an asymmetrically distributed modification, with phosphorylated histones preferentially retained in the non-differentiating stem cell pool [108]. A similar phenomenon was recently reported in Drosophila intestinal stem cells [109] but not in cultured mouse embryonic stem cells [110] or Drosophila female germline [111], suggesting both organism- and tissue-specific mechanisms may be involved. Together, these studies demonstrate the emerging complexity of retention of epigenetic information during cell division, and diverse, organism- and cell-specific solutions to this problem.

While progressing through the S-phase offers a chance to reset minor deficiencies in chromatin fidelity, long-lived non-cycling cells face a different challenge, where replication-dependent histone synthesis is largely absent, and a distinct complement of histone isoforms and chaperones maintains the integrity of the chromatin fiber over the lifespan of the organism. As histones are among the slowest-exchanging proteins in both neurons and astrocytes [112], epigenetic information is continuously written, erased, and re-written again over the same stretch of chromatin palimpsest, both highlighting the need for efficient “erasers” in these cell types, and raising the question whether partial retention of previous states may eventually impair normal function. Remarkably, replication-independent H3.3 variant is accumulated in neurons and astrocytes, constituting from a third of the total H3 pool at birth to nearly 100% by 20 years of age in humans [113]. H3.3 pool is slowly exchanged during lifetime, and impairment of this process via depletion of H3.3 or loss of HIRA chaperone alters both transcriptional programs and functional output of mammalian brain, implicating histone turnover in fundamental neurologic functions. Further, these findings indicate that mature neurons and glia must have unique mechanisms of genome partitioning, as H3.1–H3.3 distinction is lost in these cells [113]. Whether mature neurons and glia employ additional factors to demarcate cis-regulatory elements typically delineated by higher H3.3 content or rely on stability of three-dimensional architecture established early in development, remains to be investigated.

Chromatin fidelity is lost in disease

The establishment and maintenance of chromatin modifications present a unique challenge, as many cell- and site-specific mechanisms must work in concert. While failsafe systems and layers of redundancy exist, these mechanisms are nonetheless sensitive to germline and somatic mutations, changes in metabolic state, and age-related degeneration. Emerging evidence implicates progressive disorganization of chromatin in human disease, and examples of chromatin alterations observed in specific disorders are discussed below.

Aging

Progressive alterations in chromatin composition are among the hallmarks of aging observed in diverse organisms from invertebrate models to humans [114]. Early studies noted that loss of histone chaperone ASF1 resulted in near four-fold decrease of replicative life span in Saccharomyces cerevisiae, accompanied by aberrant transcription in telomeric regions, loss of silencing, and increased sensitivity to hydroxyurea pulse-chase indicative of a chromatin assembly defect [115]. This phenotype was subsequently linked to reduced expression of histone genes in aging, with ectopic overexpression of just H3 and H4 core histones extending the median lifespan of wild-type yeast by 30% [116]. Increased histone accumulation did not confer significant resistance to hydroxyurea or peroxide treatments, suggesting that lifespan extension in this model is likely not due to additional protection from DNA damage and chromosomal instability. Of note, in Metazoa each histone isoform is encoded by several genes, from two for each of the core histones in yeast, to dozens of near-identical paralogs in vertebrates, regulated by dosage compensation mechanisms [63, 117]. As loss of dosage compensation fidelity was reported during aging in both invertebrate and vertebrate models [118, 119], it is conceivable that changes in histone gene expression may both indirectly result from, and directly contribute to, the age-related chromatin phenotypes (Fig. 3A).

Figure 3.

Chromatin alterations in aging. (A) Histone loss and stochastic increase in chromatin accessibility are frequent in aging. (B) Young and aged cells are characterized by distinct activity of “writer” and “eraser” enzymes, with balance shift from lysine methyltransferases (KMT) and deacetylases (HDAC) in young cells and organisms to lysine demethylases (KDM) and acetyltransferases (HAT) in aged subjects. (C) Aberrant DDR and endogenous retroviruses are derepressed in aging cells, with shedded retrovirus-like particles (RVLPs) and secreted pro-inflammatory cytokines contributing to overall tissue senescence.

Numerous alterations in chromatin modifications are associated with aging. Loss of H3 K9me-demarcated constitutive heterochromatin was documented in both progeroid Werner syndrome fibroblasts, and mesenchymal stem cells derived from healthy older subjects [120]. While constitutive loss of H3K9me “writer” enzymes severely impairs early development and viability in mouse model [121], recent studies employed conditional alleles to generate inducible knock-out of the three methyltransferases responsible for H3 K9me3 deposition [122]. Loss of H3 K9me “writer” enzymes in vivo induced behavioral and morphological features of premature aging, including degeneration of organs and tissues, emergence of senescence-associated transcriptional and DNA methylation signatures, and severely reduced lifespan. While the organism-wide deletion provided the first causative model of heterochromatin depletion in aging, future studies will refine the primary effectors of the phenotype using tissue- and cell-type-specific Cre drivers.

Age-associated changes in developmentally regulated gene expression were linked to aberrant H3 K27me accumulation in Drosophila intestinal stem cells [123]. Curiously, significant H3 K27me2 expansion was documented by mass spectrometry, with minimal effect on K27me3—whether this indicates defects in PRC2 targeting, or increased activity of K27-specific demethylases, is unclear. Transcriptional effects were alleviated by depletion of Pc, a fly homolog of PRC1-associated K27me3 “reader” Cbx, further implicating PcG dysfunction in progressive tissue dysplasia during aging. Studies in mouse model of ulcerative colitis documented loss of CBX3/HP1γ, an H3 K9me3 “reader” protein implicated in constitutive heterochromatin organization [124]. CBX3 loss triggered inflammatory phenotypes in the gut epithelia, and activated cryptic splicing sites at multiple transcripts, including progerin, an aberrant prelamin A variant accumulated in Hutchinson-Gilford progeria. Interestingly, progressive splicing defects are documented in several models of aging and directly contribute to loss of proteostasis [125, 126]. Alteration of the chromatin landscape thus affects cellular homeostasis through transcriptional programing, and by altering fidelity of co-transcriptional pre-mRNA processing.

Loss and redistribution of repressive H3 K9 and K27 methylations in aging cells are balanced by gain of modifications associated with active genome. Nearly 1/5th of the genome in senescent IMR90 fibroblasts was occupied by H3 K4me3, a mark canonically limited to active transcription start sites [127]. Effects of this expansion on genome regulation are unclear, as K4 methylation directly recruits multiple “reader” factors via Chromo, Tudor, or Plant Homeodomain (PHD) [128]. Expansion of H3 K4 methylation is predicted to reduce the spread of PRC2-dependent K27 methylation by competition for an allosteric site on EED subunit [129]. Further, the state of H3 K4 is “read” by ATRX-DNMT3-DNMT3L (ADD) domain, with unmethylated K4 critical to the local activation of de novo methyltransferases and recruitment of chromatin remodeler and histone chaperone ATRX [130–132]. As such, effects of ectopic H3 K4 methylation likely extend beyond promoter activation by K4me3 “reader” recruitment, further exacerbating the degradation of repressive chromatin.

Aberrant histone acetylation has been linked to aging phenotypes, with excess acetylation at multiple H3 and H4 residues generally associated with age. Loss of acetyltransferase activity has been correlated with reduced senescence and extended healthy life span [133], while increased levels or activity of “eraser” enzymes have extended lifespan in invertebrate and select vertebrate models [134], although the latter notion has been challenged [135, 136]. Determining the molecular cascades implicated in ectopic and aberrant acetylation is further complicated by generally promiscuous activity of related enzymes [137], redundancies of many paralogous “writers” and “erasers,” and concerns about the specificity of antibodies essential for genome-wide analyses of specific modifications [138]. It seems apparent though that age-related alterations in histone modifications are not driven by a single principal factor but rather represent a coregulated ensemble. This is supported by recent findings that age predictor models trained on one set of histone modifications retain predictive power for other histone modifications, e.g. a model trained on H3 K27 acetylation is capable of predicting the sample age based on H3 K9ac or K4me3 as well [139] (Fig. 3B). While histone modification-based predictors of biological age are limited by the quality of genome-wide data, chromatin maintenance in aging can be estimated by DNA methylation- [140] or accessibility-based “epigenetic clocks” [141], further demonstrating profound connections between the chromatin maintenance mechanisms.

Functional outcomes of these diverse alterations include concurrent aberrant transcriptional activation and repression; while transcriptional noise has long been postulated to increase with age, recent analyses show it may not be a universal hallmark of aging [142]. More consistent are reports of DNA damage, activation of aberrant secretory phenotypes and endogenous non-long terminal repeat LINE-1 retrotransposons and retroviruses (ERVs) [143–145]. Further, ERV particle release by senescent cells induces similar phenotypes in young mesenchymal progenitors via innate cyclic GMP-AMP synthase (cGAS)–stimulator of interferon genes (STING) pathway, suggesting cellular heterogeneity may dominantly exacerbate aging phenotypes [146] (Fig. 3C).

While it is likely that not all changes listed above are directly causative to aging, substantial evidence points to the fundamental role of chromatin dysfunction in degeneration of nuclear function. In particular, Mendelian disorders recapitulate many, albeit not all, features of normal aging. Mesenchymal stem cells derived from patients with an adult-onset progeroid Werner syndrome (OMIM: 277700) are characterized by reduced replicative potential and premature senescence, accompanied by global disorganization of nuclear DNA density documented by confocal imaging analyses, and loss of constitutive and facultative heterochromatin in genic regions and near telomeres [120]. Likewise, progeroid Cockayne syndrome B (CSB, OMIM: 133540) resulting from ERCC6 protein implicated in transcription-coupled excision repair, is characterized by reduced H3 K9 methylation and decrease in cognate SUV39H1 and SETDB1 methyltransferase levels [147]. Curiously, reduced expression of histone H3 genes was also reported, mirroring the observation of reduced histone dosage in aging cells [116, 147]. On the contrary, progerias driven by nuclear lamina defects due to LMNA and BANF1 mutations show increased accumulation of H3 K9me and K27me, coupled with global loss of chromatin accessibility [148, 149]. These observations indicate that genomic mislocalization, but not the absolute abundance of aberrant histone modifications, is a likely driver of disease progression. Alternatively, these chromatin phenotypes may converge on the similar functional output by the “reader dilution” mechanism, wherein both loss and global gain of the modification results in similar functional outcomes [49, 51, 64].

Whether these alterations represent actionable therapeutic targets remains unclear, as multi-decade human trials are unlikely to proceed due to design complexity and cost, and studies in vertebrate models remain controversial [135, 150, 151]. Advances in functional genomics and organoid technology may alleviate some of the challenges presented by model organism research, and while we are cautiously optimistic about future advances in chromatin biology of aging, it is clear that no “silver bullet” will resolve every aspect of this complex process.

Neurodegeneration

Closely linked to pathological aging is neurodegeneration, the loss of neuronal function culminating in premature cell death and manifesting as progressive neurological impairment. This group of disorders with a complex and often unclear etiology and diverse pathogenesis can be triggered by primary autoimmune or cerebrovascular deficits or arise autonomously as a result of progressive dysfunction of lineage fidelity mechanisms in non-dividing brain cell types, including neurons, microglia and astrocytes.

Extreme longevity of neurons and astrocytes imposes distinct constraints on genome fidelity maintenance. Cytosine methylation, both in symmetrical CpG dinucleotides, and non-CpG (CpH) context, is remarkably high in adult brain, particularly in neurons [152]. Coincidentally both DNMT3A “writer” and Methyl-Cytosine binding Protein 2 (MeCP2) “reader” of CpH methylations emerged at the onset of vertebrate lineage, suggesting this may have contributed to anatomical innovations, in turn facilitating behavioral and cognitive complexity of the vertebrates [153]. Indeed, de novo DNA methylation is required for maturation of adult-born neurons, controlling dendrite branching and synaptogenesis formative to behavior and cognition [154]. Likewise, histone isoform and post-translational modification landscapes are distinct in the brain, with replication-independent histone variants progressively making up a greater proportion of total pool in aging neurons [113], and unique modifications acting in circuit-specific manner to facilitate distinct transcriptional programs [155, 156]. Neuronal function is thus particularly reliant on a unique complement of histone chaperones and “writer” enzymes [113, 157].

Single-cell analyses of transcriptional activity and chromatin accessibility landmarks altered in Alzheimer’s disease (AD) revealed erosion of epigenomic constraints and dysregulation of cell identity programs in multiple cell types in the prefrontal cortex [158]. These were accompanied by impaired DNA damage response (DDR) and altered compartmentalization of DDR foci, coupled with misexpression of cohesin subunits and chromatin chaperones and remodelers, including ATRX and subunits of the BAF complex [159]. While two core characteristics of AD are extracellular plaques of β-amyloid and neurofibrillary tangles of Tau protein in the neuron body, genome-wide analyses revealed a complex phenotype affecting multiple cell types and processes, including nuclear function [160]. Several studies suggest that aberrant DDR contributes to Tau neurotoxicity [161]. Interestingly, this effect was linked to the reentry of post-mitotic neurons into the cell cycle: reduction of ATM and Chk2 kinases and TF p53 exacerbated Tau pathology, whereas increased p53 accumulation had a surprising neuroprotective effect [162, 163]. The mechanism upstream of cell cycle reentry was determined in subsequent studies in Drosophila model, where global reduction of H3 K9me and the reader protein HP1 was induced by Tau overexpression [164]. Tau-induced toxicity was alleviated by restoration of repressive heterochromatin using genetic tools; remarkably, a similar rescue was achieved by manipulating the Ago3/PIWIL1 components of the piRNA biogenesis pathway, suggesting aberrant transposon reactivation in Tau pathology, observations subsequently validated in vertebrate models and patient samples [164, 165].

Oxidative stress

Mechanistic, etiological connection between neurodegeneration and advanced age has been postulated from the strong correlation between the two. Progressive accumulation of free radical-induced damage is thus likely to specifically disrupt the function of extremely long-lived proteins. Protein turnover rates are assessed by stable isotope labeling (SILAC)—a mass-spectrometry technique for quantification of distinct isotope incorporation into newly synthesized proteins. Experiments in cell culture and animal models, as well as remarkable studies of post-mortem human material matching carbon 13 (C¹³) levels in specific proteins to atmospheric levels that briefly rose during the widespread nuclear testing in the mid-twentieth century revealed remarkably slow exchange rate of histones and many histone-binding proteins, indicating these may be more susceptible to damage over long periods of time [112, 113, 166]. As chromatin in each human cell represents a polymer of an estimated 20 million nucleosome subunits with limited capacity for self-renewal in non-dividing cells, oxidative damage of even a small proportion of incorporated histones may lead to significant dysfunction over time. Further, exchange of replication-dependent histone H3.1/2 isoform for replication-independent H3.3 in postnatal neurons and glia may represent a curious vulnerability to free radical detoxification. Copper is a critical microelement for oxidative phosphorylation in mitochondria and reactive oxygen species (ROS) detoxification by superoxide dismutase 1 (SOD1) [167]. Mutations in copper homeostasis factors cause developmental disorders, including Menkes and Wilson’s diseases, with the central nervous system particularly susceptible to damage [168], and copper imbalance has been extensively studied in AD pathogenesis [169]. Remarkably, copper directly associates with the nucleosome, and several amino acids in the histone fold of H3, including C110, H113, and L126 are implicated in copper reductase activity of the H3/H4 tetramer, catalyzing the Cu²⁺ to Cu¹⁺ reaction [170–172]. H3.3, the predominant H3 variant in adult post-mitotic neurons and astrocytes, replaces two adjacent oxidizable amino acids, M90 and C96 of H3.1/2, with non-oxidizable G90 and S96 (Fig. 4). As G90 is critical for H3.3/H4 replacement by DAXX/ATRX chaperone [173], these observations raise the question whether post-mitotic cells have a reduced capacity for copper buffering and thus may be specifically vulnerable to damage by ROS. We predict that future studies will identify mechanisms of disease progression beyond the canonical protein aggregates targeted by current therapeutic interventions.

Figure 4.

Histone H3 variant accumulation may alter redox capacity of the nucleosome. Replication-dependent H3.1 and replication independent H3.3 are shown schematically, with α-helices of histone fold represented as thicker bars. Five amino acids unique to each isoform are indicated, and H3.1-specific M90 and C96 are highlighted in magenta. Replacement of H3.1 with H3.3 in postnatal brain may reduce the redox buffering capacity in the aging post-mitotic cells.

Metabolism

Catalytic activities of enzymes operating in chromatin are critically dependent on small molecule cofactors, linking metabolic pathways to genome regulation. While the evolutionary advantage of this connection is intuitive, allowing for the rewiring of global transcriptional and replicative states in response to nutrient abundance, pathological states characterized by metabolic dysfunction directly impact chromatin landscape as well.

Acetylation of newly synthesized histones is established by type B acetyltransferases in the cytoplasm, followed by transfer into the nucleus where at least three distinct classes of type A acetyltransferases operate in context of chromatin [174]. Both enzyme types utilize acetyl-coenzyme A (ac-CoA) in the reaction, with ac-CoA availability directly related to cytoplasmic histone processing and specific lysine modifications in chromatin [175] (Fig. 5). Longer-chain acylations originate from short-chain fatty acids (SCFAs) incorporated into CoA by acyl-CoA synthetase ACSS2 [176]. While their abundance is orders of magnitude lower than that of acetyl-CoA in most contexts, microbiota-derived SCFAs in intestinal lumen provide a unique environment which shapes the chromatin landscape of intestinal epithelial cells [177]. Curiously, SCFAs additionally inhibit the expression of histone deacetylases (HDACs), thus indirectly increasing the levels of acetylated (and acylated) histones [178]. Acylation of lysine amino group causes two non-exclusive effects: first, it reversibly negates the positive charge of basic amino acid side chain thus altering electrostatic histone–DNA and histone–histone interactions, and, second, provides a platform for specific “reader” association [179]. While acyl-dependent chromatin decompaction is generally associated with active transcription, recent studies point to a more nuanced relationship, where specific acylations may recruit context-dependent effectors [179, 180].

Figure 5.

Metabolic pathways are essential for chromatin fidelity. Key metabolic pathways directly implicated in chromatin regulation are shown above, with enzymes and additional crosstalk omitted for clarity. Left, polyamines are implicated in nucleosome organization and histone tail accessibility; middle, methionine and coupled folate cycle are the source of S-adenosyl-methionine (SAM) donor for histone and DNA methylation; right, TCA cycle is the source of α-KG cofactor for histone and DNA demethylases, with the exception of FAD-dependent LSD1 histone demethylase. Reciprocal relationships between modifications are illustrated below. dcSAM, decarboxylated S-adenosyl-methionine; SAH, S-adenosyl-homocysteine; THF, tetrahydrofolate; B2, B6, and B12 are critical cofactors of the folate cycle, with B12 coupling methionine and THF biosynthesis; 5,10-CH₂-THF, 5,10-methylene-tetrahydrofolate; 5-CH₃-THF, 5-methyl-tetrahydrofolate; Ac-CoA, acetyl-Coenzyme A; KMTs, histone lysine methyltransferases; DNMTs, DNA methyltransferases; KDMs, histone lysine demethylases; TETs, ten–eleven translocation (DNA demethylases); HDACs, histone lysine deacetylases; NAD+, nicotinamide adenine dinucleotide (oxidized form); NIC, nicotinamide.

Ac-CoA is essential in carbon metabolism, including tricarboxylic acid (TCA) cycle and lipid biosynthesis, yet its distinct role in chromatin is highlighted by a nuclear pool of ac-CoA synthesis machinery, including pyruvate decarboxylase (PDC), ATP citrate lyase (ACLY), and ACSS2 [181–183]. In a recent study, non-uniform metabolic flux within the tissue was connected to distinct histone acetylation profiles of nuclei at distinct spatial locations. In a Drosophila wing disc model, enhanced availability of acetate in the layer at the tissue edge drives increased ac-CoA synthesis and histone acetylation [184]. Whether such positional signals are conserved in other developmental contexts or pathological conditions remains to be elucidated.

Erasure of acetylation and related acylations in histones is accomplished by several classes of deacetylase enzymes (HDACs), including class I, II, and IV “classical” HDACs and class III sirtuins, named after yeast homolog Silent mating-type information regulation 2 (Sir2). Sirtuins have been widely studied in context of lifespan and aging, and are colloquially referred to as “longevity proteins” [185]. Yet, the role of sirtuins in aging remains controversial, as overexpression of sirtuins in Caenorhabiditis elegans and Drosophila failed to increase the lifespan [186, 187], and effects of transgenic SIRT1 in mouse model were modest [188]. These findings have been explained by confounding effects of background mutations in laboratory animals, as well as the dependence of Sirtuins on NAD+ cofactor, such that overexpression alone may not be sufficient to substantially increase their enzymatic activity [189]. Increase in NAD+ levels via nicotinamide riboside supplementation indeed extended the replicative lifespan in the yeast model [190], although whether these effects are sirtuin-dependent remains a subject of debate [136].

Histone methylation represents another abundant modification with profound connections to metabolic pathways. While acylation is specific to lysines and negates the charge of the amino acid side chain, methyl groups can be installed on both lysines and arginines, and do not affect the charge. As such methylations exert their function primarily via recruitment of specific “reader” modules and are implicated in both activation and repression: histone H3 K9 and K27 methylations are foundational to formation of constitutive and facultative heterochromatin, respectively, while K4, K36, and K79 methylations are associated with active regulatory elements, highly interactive compartment A, and productive elongation [29]. Notably, relative quantification of methylation abundance by mass-spectrometry documented remarkable abundance of H3 K9, K27, and K36 methylations in cycling cells, reflecting broad distribution of these modifications in the genome. Further, as single lysine residue may carry up to three methyl groups, chromatin acts as a major rheostat in one-carbon metabolism pathways.

Both histone and DNA methylation reactions require S-adenosylmethionine (SAM), the universal methyl donor generated at the convergence of folate and methionine biosynthesis cycles (Fig. 5). Dietary or redox state-driven deficiency of vitamin B coenzymes, in particular B12 and B6, thus has profound effects on SAM availability and downstream activity of methyltransferase enzymes in chromatin [191], and vitamin B12 supplementation may have neuroprotective effects in models of traumatic brain injury and ischemic stroke [192]. Erasure of methylation is accomplished via oxidation of the methyl group, either by flavin adenine dinucleotide (FAD)-dependent KDM1 enzyme, or a family of α-ketoglutarate (α-KG) dependent Jumonji-domain enzymes, including KDM2-6 histone demethylases and Ten-eleven translocation (Tet) family DNA demethylases. Availability of α-KG, ratio of α-KG to succinate, and presence of coenzyme cofactors, including iron and vitamin C, are critical for histone and DNA demethylation, and act as major regulator of cell fate in development and disease. Efficient demethylation is essential for pluripotency maintenance, as α-KG supplementation improves stem cell self-renewal in culture [193]. Simultaneously, both vitamin C and α-KG supplementation antagonize tumor progression in leukemia and pancreatic carcinoma models via activation of Tet-dependent DNA demethylation [194, 195]. Recently, maternal iron deficiency was causally linked to loss of H3 K9me “eraser” KDM3A activity in developing gonads: depletion of Fe²⁺ ion in utero results in accumulation of H3 K9me2 upstream of the Sex-determining region Y (Sry) gene in male embryonic gonads, inducing male-to-female sex reversal by suppressing Sry expression [196]. Together, these data highlight complex and context-dependent effects of metabolic interventions.

SAM availability has further implications for the biosynthesis of polyamines, ubiquitously abundant polycationic metabolites (Fig. 5). Aminopropyl group from decarboxylated SAM is transferred onto ornithine and spermidine by spermidine synthase (SRM) or spermine synthase (SMS), respectively. Polyamine recycling in turn is initiated by spermidine/spermine N-acetyltransferase utilizing ac-CoA as a cofactor [197]. Polyamine imbalance due to SMS mutation results in mendelian Snyder Robinson syndrome (SRS, OMIM: 309583) characterized by neurological deficits, mesenchymal malformations, and premature aging. While the primary cause remains unclear, SMS deficiency alters mitochondrial function, impairing glucose metabolism, and drives broad transcriptional misregulation in mesenchymal progenitor cells, limiting their proliferative capacity [198]. On the contrary, several malignancies including pancreatic ductal adenocarcinoma, midline gliomas, and acute leukemias upregulate polyamine biosynthesis pathway to support tumor growth, and inhibitors of synthesis and reuptake have shown promise in early trials [199–201]. In addition to multiple cytoplasmic activities (reviewed in [202]), polyamines are abundant in the nucleus, their polycationic nature suggesting extensive electrostatic interactions with negatively charged backbone of nucleic acids. Indeed, in vitro experiments demonstrated extensive association of polyamines with DNA [203], and polyamine supplementation enhances homology-directed repair, stabilizing synaptic complex formation by the RAD51 and single stranded DNA filament [204]. Recent studies investigated the role of polyamines in regulation of chromatin structure and histone modification landscape, revealing global disorganization of chromatin condensation upon induced polyamine deficiency [205]. Locally, polyamine depletion collapses the histone tails onto the nucleosomal DNA, reducing their accessibility to “writer” and “eraser” enzymes; globally altered modification landscape affected differentiation and reprogramming efficiency in cell culture [206]. These observations echo findings in linker histone deficient cells, wherein chromatin decondensation altered substrate preference of several “writer” complexes [6, 207], and further expand the connections between metabolic states, chromatin organization, and lineage fidelity.

Metabolic disorders thus have profound connections to chromatin-dependent regulation. While causal connections between chromatin alterations and metabolic disease are largely limited to Mendelian disorders, “metabolic imprinting” is increasingly seen as a potential cause of metabolic disorders in adults and seniors [208]. Interestingly, recent studies identified two populations of pancreatic endocrine β-cells characterized by starkly distinct H3 K27me3 levels and capacity for insulin secretion, suggesting epigenome modulation may have a role in metabolic syndrome [209]. Chromatin dysfunction as a consequence of metabolic abnormalities appears a more common occurrence, whereby imbalance of metabolites produced by intracellular pathways and microbiota alters activity of many “writer” and “eraser” enzymes, in turn exacerbating the pathologic state [210].

Malignancy

Somatic, i.e. tumor-specific, mutations in “reader” and “writer” factors are among the most prevalent oncogenic drivers [211, 212] and represent a promising class of therapeutic targets [213]. More recently, a number of mutations in core histone genes have been identified, initially in a few rare tumors [49, 214], and later broadly across multiple diverse malignancies [215–217]. These mutants, aptly termed “oncohistones,” appear to broadly fall into three classes, based on the mechanism of action and prevalence in specific disorders (Fig. 6). First, specific missense mutations affecting key residues in the N-terminus of H3 and either dominantly inhibiting cognate SET domain methyltransferases in trans (H3 K27M/I and H3 K36M) [49, 214] or inhibit H3 K36 methylation in cis resulting in altered DNMT3A recruitment (H3 G34 mutations) [218]. These mutations are found in specific malignancies, including pediatric gliomas (DIPG) and sarcomas, and drive unique transcriptional and cellular phenotypes associated with significant, albeit not complete, loss of specific “writer” activity. The latter observation prompted a hypothesis that residual activity of PRC2 represents a specific vulnerability in H3 K27M mutant gliomas [219]. Second, mutations in specific residues within the histone fold domain alter structural stability of the nucleosome core particle, enhance nucleosome sliding, and impair stable cell differentiation in vitro [215–217, 220]. These mutations show limited tumor specificity, reminiscent of BAF remodeler complex mutations prevalent across many malignancies [211]. Similarly, several cancer-associated missense variants in H3 N-terminal tail appear to act in cis, impairing local PRC2 activity and lineage fidelity [55]. Third, broadly distributed missense mutations in linker histone H1 appear to largely represent loss-of-function and are uniquely frequent in peripheral B-cell malignancies [6, 52, 207]. Chromatin decompaction upon H1 loss results in ectopic reactivation of stem-cell-like transcriptional programs associated with early hematopoiesis, constituting a distinct class of lymphomas. Remarkably, H1 heterogeneity was reported in solid tumors, and likewise correlates inversely with accumulation of cancer stem cell markers [221]. Together, these studies exemplify the role of chromatin in enforcing lineage fidelity during cell division and differentiation.

Figure 6.

Chromatin partitioning is altered in malignancy. (A) balanced activity of “writer” enzymes (A and B, left) establishes distinct chromatin domains (A and B, right). (B) inhibition of “writer” B by type 1 “oncohistones” in trans leads to expansion of domain A and may dilute the density of modifications established by A or associated “readers.” (C) reduced local activity of “writer” B due to effects of type 2 “oncohistones” in cis leads to stochastic disruption of domain B, with spurious accumulation of ectopic modifications. (D) altered compaction of chromatin fiber in H1 mutant malignancies drives redistribution of A and B-dependent modifications.

Aberrant differentiation phenotypes induced by histone mutations in malignancy are echoed by observations of altered 3D chromatin topology driving tumorigenesis. Hypermethylation of CTCF binding sites in CpG island methylator phenotype gliomas carrying Isocitrate Dehydrogenase (IDH) “oncometabolite” mutations alters insulation between TADs and promotes aberrant long-distance interaction between ectopic enhancer and PDFRA oncogene [222]. Likewise, reactivation of transposable elements due to failure of chromatin-dependent repression enables oncogene expression via cryptic regulatory elements in many malignancies [223], a process mimicked by hemizygous loss of cohesin cofactor NIPBL and thus driven by broad dysregulation of genome topology [224]. It stands to reason that such defects similarly contribute to other progressive disorders beyond malignancy.

Concluding remarks

The exponential adoption of genomic technologies, population-wide descriptive studies, and high-throughput functional analyses, together with emerging in vivo tools for chemical biology and single-molecule techniques, have allowed chromatin biology to look far beyond the textbook “on/off switch” of transcriptional regulation. While interdisciplinary collaborations enabling truly “greater than the sum of parts” insights are to be celebrated, the new challenge lies in making highly complex results accessible to a broader audience while retaining sufficient information about controls, data transformations, and biologically relevant background. Likewise, the growing number of studies coupled with the increasing technical complexity of experimental and analytical approaches inevitably results in occasionally contradictory observations. The role of sirtuins in aging and disease [134, 136, 186], the contribution of three-dimensional chromatin organization to development and cancer [67, 72, 73], the connections between metabolism and histone modifications [225, 226], and the very nature of transcriptional regulation by chromatin [227] are debated. We believe these conversations represent a critical part of healthy scientific process and are a positive sign that the field is not dominated by a single dogma. As new technologies permeate chromatin biology, we expect that new findings will continue to refute, improve, and extend existing foundations.

Although chromatin dysfunction had been extensively documented in virtually early stage of disease state—from developmental disorders to malignancy to acute viral infections [228]—translating these mechanistic insights into the clinic has been challenging. The introduction of genotype-guided approaches, for example, through the design of basket studies, and subsampling of patient groups with distinct molecular vulnerabilities have proven promising in specific cases [229, 230]. More comprehensive interventions, including restriction of polyamine availability by simultaneously restricting synthesis and uptake, as well as targeting histone acetylation by HDAC inhibition, have shown positive results in combination therapies for various solid tumors and lymphoid malignancies [231, 232]. The discovery of novel regulatory pathways may further expand the repertoire of available interventions. For example, understanding the mechanisms of dosage compensation by histone genes, where the loss of a gene drives compensatory upregulation of its paralogs, could enable targeted upregulation of histone accumulation in aging cells [233]. Likewise, chromatin dysfunctions in malignancy and neurodevelopmental disorders might prove amenable to manipulation of compaction states [207, 234] or specific enzymes associated with distinct vulnerabilities [49, 219]. However, with the exception of rare disorders driven by a singular deficit, we do not view chromatin targeting in therapy as a panacea, but rather as one facet of combinatorial interventions where a multipronged approach would offer synergistic benefits. Understanding the subtle mechanisms of chromatin fidelity will undoubtedly help address this challenge.

Notes

Present address: The David Rockefeller Graduate Program in Bioscience, The Rockefeller University, 1230 York Ave, New York City, NY 10065, United States

Conflict of interest

None declared.

Funding

The Soshnev laboratory is supported by the National Institutes of Health [NCI R01CA234561], American Cancer Society [IRG-21-147-01-IRG], Cancer Prevention and Research Institute of Texas [IIRA RP240068, HIHR RP240446], UT Rising STARs, UTSA NDRB seed funding, and UTSA startup. C.M.C. is supported by the David Rockefeller Graduate Program at The Rockefeller University and a Rockefeller University Women & Science Graduate Fellowship. Mohamadzadeh laboratory is supported by NIH R01 DK109560C, NIH R01 AI154630-01, and VA BX006310-01A.

Data availability

No new data were generated or analysed in support of this research.