Studying how gene regulation controls cell aging

The main repeating building block of chromatin is the nucleosome (Fig. 1) [6]. The structure of the canonical nucleosome core includes an octamer of proteins, wrapped by a ~ 147 bp segment or 1.7 turns of DNA [6]. The protein components of the nucleosome are histones, and the octamers are made up of two copies each of histones H2A, H2B, H3, and H4, also known as core histones [7]. The short and linear piece of DNA between two neighboring nucleosomes is called linker DNA and it is protected by the linker histone H1 [8]. Flexible regions extending from the surface and flanking both ends of the histone fold are termed “histone tails” [9]. Histone tails contribute to nucleosome stability and facilitate genomic transcription as they are able to bind transcription factors and enzymes [9]. Contrary to prior observations supporting the folding of a nucleosomal array into a 30 nm fiber, recent studies using super-resolution microscopy revealed that nucleosomes appear more flexible, grouped in discrete domains known as “nucleosome clutches” (Fig. 1) [10, 11]. Linker histones contribute to the stabilization of the nucleosome-nucleosome and nucleosome-DNA interactions [12].

Regulatory elements (promoters, enhancers, and silencers) comprise binding sites for transcription factors which may be positioned upstream or downstream of the genes they control [13]. Chromatin looping is widely acknowledged and explains the capacity of the above-mentioned elements to regulate transcription at a distance (Fig. 1) [14]. Within this context, the formation of enhanceosome loops, referring to specific DNA arrangements within a nucleoprotein complex comprised of multiple transcription factors, increases the level of gene expression [15]. Recent advancements in the field of chromosome biology indicate the existence of different types of chromatin loops depending on their structure and function. Chromatin looping is a key structural feature of chromatin organization and below are discussed the three major types of chromatin loops: mitotic loops, interphase loops and transcription loops [16].

As a result of loop extrusion, an additional key feature of eukaryotic genome folding is the organization into sub-megabase scale domains, also known as TADs (Fig. 1) [30, 31]. TADs are considered self-interacting genomic regions acting as functional units, with a low degree of interaction with regions outside of the domain [32]. TAD features appear to be highly conserved across mammalian species, with a key characteristic being the presence of the insulator-binding protein, CTCF, along with the SMC cohesin complex at most TAD boundaries [33, 34]. Within TADs, genes and their regulatory elements are often in close spatial proximity, ensuring efficient transcriptional regulation [2, 35]. In addition, specific active transcription marks, such as H3K4me3 and H3K36me3, are strongly enriched at TAD boundaries [33]. Of note, evidence indicates that TADs are enriched for housekeeping genes and transcription start sites (TSS), with depletion of insulatory TAD boundaries leading to ectopic gene expression in vitro and in vivo [32, 36]. Latest studies also demonstrate the importance of TADs, revealing that deletions of TAD boundaries disrupt normal genome function and cause abnormal gene expression, developmental defects, and tumorigenesis [31, 37, 38].

Advancements in techniques for studying chromatin organization have opened new horizons in our understanding of long-range interactions between TADs [2]. These interactions can occur between very distant areas on the linear genome, giving rise to nuclear compartments which separate active (euchromatin) and inactive (heterochromatin) chromatin (Fig. 1) [39]. In more depth, the human genome is partitioned into either “A” compartments, which are associated with open chromatin and subsequently actively transcribed genes and active histone marks, or “B” compartments, which in contrast, correlate with closed chromatin, inactive genes and repressive marks [40]. A and B nuclear compartments are further subdivided into six smaller contact domains known as subcompartments (two for the A compartment and four for the B compartment) which arise through the attraction and/or repulsion between individual TADs with analogous epigenetic marks (Fig. 1) [41, 42]. The “compartment switch” process in genomic regions can be influenced by various factors such as epigenetic state, gene expression and replication timing (RT) [43]. Although A/B compartments are less well understood, current knowledge of chromatin organization indicates that they are cell-type specific and less conserved compared to TADs [44].

Chromosome territories refer to the distinct regions occupied by chromosomes within the nucleus (Fig. 1) [45]. This major feature of nuclear architecture is conserved across evolution, appears cell-type specific and is characterized by the arrangement of chromosomes in a radial pattern within the nucleus [46]. The organization of chromosome territories is not random; instead, it is associated with both gene density and chromosome size [47]. The gene-rich chromosomes tend to be located in the interior of the nucleus, while chromosomes with a lower gene content are more commonly found at its periphery [48, 49]. In addition, chromosome territories are spheroid dynamic structures enabling gene relocation from the periphery towards the interior or the opposite [50, 51]. High-order spatial interactions between neighboring chromatin territories have been described using high-throughput techniques. These structural interactions affect gene expression and are crucial for cell fate determination [52].

Chromatin is not an inert structure. Instead, DNA can respond to external cues influencing various biological processes [53]. Histone modifications, along with DNA methylation and micro-RNAs (miRNAs) are the main means of the epi-information beyond the DNA sequence, collectively referred to as epigenetics, the study of molecular modifications on DNA that regulate gene expression, without altering the DNA sequence (Fig. 1) [3].

The N-terminal tails of the histones can undergo post-translational modifications, which are the core mechanism of epigenetic regulation [54]. Histone post-translational modifications (HPTMs) include histone methylation, histone acetylation and histone phosphorylation. The addition of methyl groups usually to lysine (K) residues on histones H3 and H4, catalyzed by histone methyltransferases (HMTs), leads to histone modifications such as H3K4, H3K9, H3K27, H3K36, H3K79 and H4K20 [54]. Among them, H3K4, H3K36 and H3K79 are usually considered active marks as they are enriched in coding regions. These marks, are commonly associated with the activation of transcription making chromatin more accessible to transcription factors [55, 56]. In contrast, H3K9, H3K27 and H4K20 are generally considered as repressive marks, as they are associated with gene repression, heterochromatin and the formation of broad domains at promoters of silenced genes [57–60]. However, the outcome of the modification depends on the methylation site and methylation level (number of added methyl groups). For instance, H4K20me exists in three distinct forms mono-, di-, and trimethylation. H4K20me1 facilitates increased chromatin accessibility promoting expression of housekeeping genes, while H4K20me3 drives the maintenance of DNA methylation, promoting gene silencing [61, 62]. Lysine residues of histones are also subjected to acetylation, catalyzed by histone acetyltransferases (HATs) [53]. The most studied modification is H3K27ac, which usually co-exists with H3K4me3 and is defined as an active enhancer mark [63]. More specifically, H3K27ac is correlated with transcriptionally active genes and is found at active promoters and enhancers or around TSS [64]. Beyond the fact that H3K27ac distinguishes active enhancers from poised ones, it also contributes to cell identity control [65]. Notably, biological processes such as cell fate determination and differentiation are regulated by epigenetic dependent events [66]. The regulatory regions of development-associated genes that are marked by both activating and repressive histone modifications (such as H3K4me3-activating or H3K27me3-repressing) are known as “bivalent” or “poised” promoters [67]. Poised chromatin is associated with pluripotency and was first identified in embryonic stem cells, where transition of a subset of poised genes to a fully active state (resolution) occurred in response to external signals [66]. Characteristically, poising takes place during gametogenesis, silencing genes that are later activated at the onset of embryogenesis, overall acting as an epigenetic control mechanism over somatic tissue patterning [66]. Lastly, the deposition of phosphate groups onto serine (S), threonine (T), or tyrosine (Y) residues on histones is known as histone phosphorylation. Although this type of modification is less well-understood, it plays a role in gene expression, cell cycle regulation and DNA damage response [68].

DNA methylation is a heritable epigenetic mark in the mammalian genome, involving the addition of a methyl group to the C-5 position of the cytosine ring of DNA on genomic 5′-C-phosphate-G-3′ (CpG) dinucleotides, by a family of DNA methyltransferases (DNMTs) [69]. This direct chemical modification of the DNA in promoter regions is covalent and reversible and favors gene silencing by recruiting repressive proteins or inhibiting the binding of transcription factors to DNA [70]. Transcriptional regulation of the genome is ensured by the coordinated action of DNMTs, which often exhibit distinct methylation patterns [71]. DNA methylation regulates various cellular and organismal processes such as, chromatin-based transcriptional regulation, chromosome stability, X-chromosome inactivation and embryonic development. Nevertheless, disruptions in DNA methylation can result in the silencing of tumor suppressor genes or oncogene activation, thereby contributing to genomic instability in numerous human diseases, including cancer [72–75]. Moreover, there is a strong association between hypomethylation of retrotransposons (LINE-1 and Alu elements) with genomic instability in non-small cell lung cancer [76]. Hypomethylation seems to render tumors more susceptible to senescence signals. If so, cancer cells that are more sensitive than normal ones to chemotherapy could rely on senescence as a means to survive.

Discovered about 30 years ago, miRNAs are endogenous small non-coding RNAs that regulate genomic translation in higher eukaryotes [77]. In greater detail, miRNAs act as epigenetic modulators either by binding to a target mRNA, leading to its degradation or suppression, or by repressing transposons through heterochromatin formation [77]. Even though they do not cause changes in the DNA nucleotide sequence, the chromatin structure is impacted by their function, as miRNAs regulate histone modifier molecules including DNMTs (miR-29 family) and histone deacetylases (HDACs, e.g., miR-140) [78, 79]. miRNAs hold a fundamental role in biological processes such as cellular proliferation, differentiation, DNA repair and aging, and hence they can be used as validated diagnostic and prognostic biomarkers [80].

Discovered in the 1980s, FISH is a powerful technique allowing researchers to detect and map specific DNA sequences on chromosomes (Fig. 2) [86, 87]. The main purpose of this technique is to detect chromosomal abnormalities [88]. FISH employs specially designed fluorescent probes that hybridize to targeted chromosomal regions in fixed cells allowing for the results to be analyzed under the fluorescence microscope [89]. Genetic disorders caused by gene fusions, abnormal number of chromosomes or the loss of a chromosomal region can all be detected through FISH [90]. In addition, FISH has been essential for gene localization, significantly contributing to the success of the Human Genome Project and advancing our understanding of developmental biology [86]. Although FISH assists in the direct identification of the 3D positions of specific genomic loci, it is unable to systematically uncover the complex folding patterns of chromatin [87].

3 C is distinguished by its high-throughput ability to analyze the overall spatial organization of chromosomes (Fig. 2) [91]. In more detail, 3C-based methods utilize crosslinking and ligation to capture the proximity of two genomic loci in three-dimensional space [92]. Genome-wide 3 C (Hi-C), circularized 3 C (4-C) and carbon copy 3 C (5-C) are all 3 C variations that have greatly contributed to a better understanding of the structure of the genome [93–95]. Among them, Hi-C, a powerful and the most extensively studied derivative, combines 3 C and next-generation sequencing (NGS) and entails crosslinking of chromatin with formaldehyde, followed by restriction enzyme digestion, re-ligation and subsequent massively parallel deep sequencing [95]. Re-ligation occurs only for DNA fragments that are physically associated in 3D space, thereby providing insights into genome-wide chromatin interactions [95]. Notable discoveries based on Hi-C include chromatin loops, TADs and A/B compartments, thus advancing our understanding of chromatin’s hierarchical folding [1, 2, 96].

ChIP-seq is a highly powerful and valuable tool in epigenomic research [85, 97]. This technique enables genome-wide analysis of DNA-binding proteins, chemical modifications of histone proteins or nucleosome positioning, thereby providing profound insights into the topological organization of DNA-associated proteins and their influence on phenotypic outcomes (Fig. 2) [98]. ChIP-seq allows for the selective enrichment of DNA sequences bound by the target protein, followed by precipitation of the fragments by using specific antibodies (either for the target protein or the histone modification) and subsequent massive parallel high-throughput sequencing of the enriched fragments [99–102]. This indispensable method offers significantly improved data compared to ChIP followed by hybridization to microarray (ChIP-on-chip). Specifically, ChIP-seq offers important information about gene regulation and epigenetic mechanisms including post-translational modifications of chromatin and histone variants, while ChIP-on-chip is not efficient at discovering unknown protein binding sites [103, 104]. Despite its remarkable capabilities, ChIP-seq has certain limitations such as the need for a high amount of input material and the requirement for high-quality antibodies to ensure both specificity and accuracy [85]. In addition to ChIP-seq, similar methodologies have emerged to address the complex organization of chromatin. For instance, ChIP-loop, a combination of 3 C and ChIP, identifies the proteins involved in DNA loop organization [105]. Notably, chromatin immunoprecipitation with paired-end tag sequencing (ChIA-PET) reduces non-specific interaction noise, often observed in ChIP-seq, and provides insights into long-range chromatin interactions, thus transcription regulation [106]. Finally, chromatin immunoprecipitation-polymerase chain reaction (ChIP-PCR) offers the highest resolution and accuracy to characterize histone modifications within defined regions of the genome [107]. To gain a deeper insight into unique mechanisms occurring in certain subpopulations, two independent groups developed single-cell ChIP-seq (scChIP-seq) [108, 109]. scChIP-seq utilizes a microfluidics system combined with single-cell DNA barcoding technologies, acquiring single-cell chromatin data [108]. This method enables the identification of activating (H3K4me3) or repressive (H3K27me3) marks for transcription, among single cells, providing important information about cellular heterogeneity [109]. Such tools pave the way for the treatment of heterogeneous diseases including cancer at the patient level following precision medicine principles [109].

Physical access to DNA is defined as the extent to which chromatinized DNA is available for binding by DNA-binding factors [83]. This accessibility exists on a dynamic spectrum ranging from “closed chromatin” to “permissive chromatin”, the latter initiating chromatin remodeling to ultimately lead to the “open chromatin” state, where chromatin becomes entirely accessible [81]. Chromatin accessibility is measured by quantifying its sensitivity to either enzymatic cleavage or methylation by using endonucleases, which digest dsDNA, and ligation-mediated PCR [110].

DNase-seq, first described by Boyle et al., provides extremely high resolution for identifying open chromatin by combining traditional DNase I footprint with NGS (Fig. 2) [111]. The DNase enzyme cleaves the unprotected DNA sequences (euchromatin), while chromatin packed into nucleosomes (heterochromatin) remains intact [111]. The isolated fragments are then used as a template for library construction and subsequent sequencing via NGS [111]. In brief, DNase-seq effectively maps active gene regulatory elements, providing precise representation of their locations across the genome, including promoters and enhancers [112]. Similarly, formaldehyde-assisted isolation of regulatory elements (FAIRE-seq) provides insights into the regulatory activity of DNA regions in the genome [113]. In FAIRE-seq, DNA–protein complexes are crosslinked using formaldehyde [114]. Ultimately, sequencing identifies regions of low nucleosome density which are typically associated with genome regulatory areas, such as protein binding sites and promoters [113].

MNase-seq, first pioneered in the mid-2000s, utilizes micrococcal nuclease to cleave and eliminate the naked DNA, leaving chromatin wound into nucleosomes for NGS analysis (Fig. 2) [115, 116]. In fact, MNase-seq relies on the activity of the MNase enzyme which functions as an endo- and exo-nuclease and cleaves protein-unbound DNA [117]. Thus, DNA bound to histones or transcription factors remains intact [81]. Intact DNA is purified and NGS is performed to obtain the genomic sequences directly relevant to the epigenome [117, 118].

Chromatin accessibility and its epigenetic status can alternatively be assessed using NOMe-seq [119]. Specifically, NOMe-seq is based on the treatment of chromatin with a GpC methyltransferase (M.CviPI methyltransferase) to detect nucleosome positioning and the DNA methylation status (Fig. 2) [120]. Although DNA methylation naturally occurs at CpG promoter sites in the mammalian genome, GpC sites (methylated by M.CviPI methyltransferase during NOMe-seq) are also abundant throughout the genome [81, 121, 122]. By methylating only unprotected by nucleosomes or transcription factors DNA regions, NOMe-seq creates a nucleosome footprint and at the same time allows analysis of the endogenous methylation patterns [119]. Although NOMe-seq provides a quantitative view of chromatin accessibility, it requires a large number of sequencing reads to achieve sufficient depth across the genome [81].

Both DNase-seq and MNase-seq require large numbers of cells as input (in the order of millions), which limits their applicability [83]. In contrast, ATAC-seq can be carried out with significantly fewer cells (around 50,000) and does not require prior knowledge of epigenetic marks to determine chromatin accessibility across the genome [123]. The ATAC-seq methodology relies on the hyperactive transposase Tn5 to cleave DNA and ligate short oligonucleotides (adaptors) at the beginning or the end of the DNA fragments which are chromatin regions characterized by increased accessibility (Fig. 2). Genome-wide mapping is followed by PCR amplification, and ultimately, the resulting double-stranded fragments are sequenced via high-throughput sequencing [124]. Notably, data from ATAC-seq experiments can be classified into two main categories: shorter reads (< 147 bp) which indicate nucleosome-free regions representing accessible chromatin, and longer reads (~ 147 bp) which reflect nucleosome regions, blocking the activity of the transposase [125]. Thus, ATAC-seq allows for the determination of both nucleosome occupancy and positioning [126]. Since Buenrostro et al. introduced ATAC-seq in 2013, it has gained wide popularity due to its advantages, such as simplicity and faster execution compared to other chromatin accessibility methods [127]. The method was further advanced in 2015, to study the chromatin accessibility profile at the single-cell level [128]. scATAC-seq is a droplet-based method utilized in individually barcoded cells from a sample and is characterized by high sensitivity [128]. This novel and cost-effective technique surpasses previous limitations and reveals the unique features of each cell that contribute to tissue function, development and various diseases [129]. However, scATAC-seq data are highly noisy and sparse compared to single-cell RNA-seq (scRNA-seq), thus requiring new approaches and software tools for their analysis [130].

In recent years, two novel immunotethering methods using enzymatic fusion proteins have been introduced in the field of genomics and epigenomics (Fig. 2). Cleavage under targets and release using nuclease (CUT&RUN) and Cleavage under targets and tagmentation (CUT&Tag) address many limitations associated with traditional mapping methodologies such as ChIP-seq [131]. In more depth, CUT&RUN isolates DNA fragments associated with specific proteins and selectively cleaves antibody-bound chromatin [132, 133]. CUT&RUN has been used primarily to profile global transcription factor binding and histone modifications in mammalian cells [134, 135]. Similarly, CUT&Tag is another powerful assay for next-generation epigenomic profiling, overcoming the need for adapter ligation during library preparation, enabling single-cell and multiplexing applications [131].

In 1961, Hayflick and Moorhead discovered that human fibroblasts lose their ability to proliferate after a finite number of cell divisions, providing the first evidence of senescence, termed replicative senescence (RS) [140]. Subsequent investigations clarified that this phenomenon relied on telomere shortening and dysfunction (damage) activating the DNA damage response (DDR) pathway and eventually resulting in cell cycle arrest [141]. Nowadays, it is well established that many stress insults such as oncogenes, DNA damaging agents and drugs, oxidative stress and metabolic perturbations can trigger senescence independently from telomere erosion, known as stress-induced premature senescence (SIPS) [142]. Although SIPS is not characterized by telomere attrition, it commonly features DNA damage at telomeric regions that fuel a persistent DNA damage response, implying thus also the involvement of telomere-related processes [142]. Oncogene-induced senescence (OIS), the most widely studied subtype of SIPS, is a well-known antitumor barrier, inhibiting the propagation of incipient cancer cells at the early stages of cancer development [143–146]. Oncogenes that trigger OIS include RAS, BRAF, CDC6, AKT, CYCLIN E and E2F1 [147, 148]. In addition, loss of tumor suppressors such as PTEN also induces cellular senescence as a failsafe mechanism against cancer, through the activation of the Akt pathway [147].

Senescence is an extremely dynamic and heterogeneous process [149, 150]. Early senescence pertains to a shift from a transient to a stable cell-cycle arrest where early senescent cells progress to ‘deep’ or ‘full’ senescence once SASP production is established [150]. Major differences have been identified between RS and SIPS at multiple levels such as mechanisms involved, gene regulation and time of occurrence during life [136]. For instance, one difference pertains to the differential accumulation of deregulated translation mechanisms such as ribosome stalling which appears absent in RS, in contrast to OIS [151]. Those differences may also be related to the epigenetic variability between senescence subtypes [152]. Despite the multi-faceted nature of the senescence phenotype, senescent cells exhibit four interconnected hallmarks: a) cell cycle withdrawal, b) altered metabolic function, c) macro-molecular damage, and d) a proinflammatory secretory property termed senescence associated secretory phenotype (SASP) [136, 153] (Fig. 3).

Senescence related cell cycle arrest is achieved through the sequential activation of the p53/p21^WAF1/Cip1 and the p16^INK4A/RB (retinoblastoma) axes [136]. p21^WAF1/Cip1 expression is transient and is considered essential for establishing the senescent phenotype, while its maintenance is sustained through the persistent expression of p16^INK4A [154]. SASP, describes a unique secretome of pro-inflammatory cytokines (e.g. IL-1α, IL-1β, IL-6 and IL-8), growth factors (e.g. EGF and TGFα), chemokines (e.g. CXCL, CCL) and other inflammatory players (e.g. IFN-γ), that are cell-type dependent [155] and can act both in an autocrine and paracrine manner [156]. Early SASP facilitates normal processes eventually recruiting immune cells for senescence elimination following tissue regeneration, wound healing and embryonic development, while when its presence is indefinitely retained (late SASP) it becomes strongly pro-inflammatory, promoting disease development [149, 150]. An additional prominent feature of senescent cells is accumulation of macromolecular damage, which is directly linked to metabolic rewiring [136]. Apart from telomere shortening and subsequent DNA lesion accumulation [141], some of the most common types of damage are associated with protein [157] and lipid oxidation [153]. In parallel, key metabolic pathways such as glycolysis, autophagy, mitochondrial, and lysosomal metabolism are altered resulting in a deregulated metabolic profile [158]. Macromolecular damage contributes to increased reactive oxygen species (ROS) production, which exacerbates protein misfolding and aggregation, subsequently leading to metabolic disorders related to cellular senescence [159].

The most constant feature of senescent cells that reflects macromolecular damage, altered metabolic traits and cell cycle arrest is lipofuscin accumulation (Fig. 3) [136, 160]. Lipofuscin is a heterogeneous fluorescent aggregate consisting of oxidized proteins, lipids and metals [136]. Its specific aggregation during senescence has been exploited for the development of novel reagents that have been adopted in guideline multi-marker detection approaches, surpassing existing challenges in the field and uncovering the impact of senescence in human diseases and aging [136, 161–166]. Along with lipofuscin detection that is a prerequisite initial step, p16^INK4A and p21^WAF1/Cip1 expression evaluation are subsequent steps of this algorithmic assessment [164, 165].

The 16 identified genes encode proteins that can be classified into the following categories: transcription factors/repressors (TP53, E2F1, SIN3B), histone modifiers (KAT7, KAT8, HDAC6, SIRT6, KDM5A, RBBP5), histone chaperones (HIRA), DNA methylation-associated proteins (DNMT3L), chromobox (CBX) family of proteins (CBX1, CBX2, CBX5) and chromatin remodelers (CHD3, SMARCB1). Among these genes, the expression levels of SIRT6, DNMT3L, HDAC6, CBX1 and CBX5 were found to be decreased while SIN3B, KAT7 and KDM5A were overexpressed in senescence. Regarding the expression status of TP53, E2F1, KAT8, RBBP5, CHD3, SMARCB1 and HIRA, it still remains elusive as they display context-dependent or dual roles in senescence (Fig. 4b).

One of the top hits of our analysis for epigenetically modulated genes in cellular senescence was the tumor suppressor p53. p53 plays a pivotal role in safeguarding genomic stability by functioning as a transcription factor that activates or represses multiple target genes [170]. Upon its activation, in response to DNA damage, p53 displays a dual role by promoting or inhibiting cellular senescence, depending on the cellular context, as well as the intensity and the persistence of the DNA damage [171]. On one hand, p53 transcriptionally regulates p21^WAF1/Cip1 at the initial phases of cellular senescence [172]. On the other hand, p53 suppresses cytoplasmic chromatin fragments (CCF) that represent damaged nuclear chromatin marked by γ-H2A.X and are characteristic of senescent cells [173]. This contributes to restriction of the SASP phenotype via chromatin remodeling of SASP genetic loci [173, 174]. Ultimately, upon its activation, p53 rapidly induces 3D chromatin conformational changes, including changes in genome compartments, DNA loops and TADS [175]. Another transcription factor was E2F1 which is a critical modulator of cellular senescence as it regulates cell cycle progression [176]. During cellular senescence, the stable repression of E2F target genes is achieved through senescence-associated heterochromatin foci (SAHF) formation [177]. SAHFs are specialized domains of facultative heterochromatin, a type of heterochromatin capable of shifting from a condensed transcriptionally inactive to a loose active state depending on the cell’s needs [178]. SAHF formation coincides with the recruitment of heterochromatin proteins and of the retinoblastoma (Rb) tumor-suppressor to E2F-responsive promoters [177]. It was also shown that during cellular senescence, Rb recruits and cooperates with HDAC1 to repress E2F-regulated promoters of target genes involved in cell-cycle progression and mitosis [179]. SUV39H1, a histone methyltransferase, is also responsible for the repression of E2F activity in senescent cells [180]. In agreement with several studies, our analysis also highlights the role of the chromatin-associated SIN3 transcription regulator family member B (SIN3B) protein in both RS and OIS. Importantly, SIN3B overexpression induces senescence and promotes SAHF formation in Ras-induced senescence [181]. Specifically, upon oncogenic stress, SIN3B expression is found upregulated, leading to its recruitment to the promoters of E2F-target genes, which are consequently transcriptionally silenced [181].

KAT7 and KAT8 are key chromatin-modifying enzymes, both of which function as histone acetyltransferases. Interestingly, published evidence supports their capacity to modulate chromatin accessibility during cellular senescence. Genome-wide CRISPR-Cas9-based screens have identified KAT7 as a putative epigenetic driver of cellular senescence and its inactivation was linked to an extended lifespan in mice [182]. On the other hand, bulk RNA-seq data following the deletion of KAT8, an essential acetylating enzyme of histone H4K16 [183], uncovered the enrichment of pathways implicated in cellular senescence, such as p53 [183]. Additionally, the expression of another chromatin modifying enzyme, HDAC6, is significantly reduced in RS and its selective inhibition by N-acylhydrazone (NAH) induced senescence in carcinoma hepatocellular cells [184]. Sirtuin 6 (SIRT6), a nicotinamide adenine dinucleotide (NAD⁺)-dependent deacetylase was additionally identified in our bioinformatics analysis as an important epigenetic regulator of various cellular processes, including senescence [185]. Specifically, SIRT6 is capable of deacetylating specific sites of histones H3K9, H3K56 and H3K18, thereby inhibiting the activation of transcription factors implicated in senescence, ultimately preventing cells from entering senescence [185]. Consequently, senescent cells were found to exhibit decreased levels of SIRT6, leading to disruption of genomic integrity, telomere attrition and deregulated cellular homeostasis [185]. We further uncovered lysine demethylase 5 A (KDM5A), a transcriptional repressor responsible for H3K4 demethylation at tumor suppressor genes, which promotes the growth of multiple human cancer types [186]. By blocking the KDM5A-H3K4me3 interaction, p16^INK4A activated G1 phase cell cycle arrest and cellular senescence [186]. Our meta-analysis also identified RBBP5, an H3K4 methyltransferase that is a core component of mixed-lineage leukemia 1 (MLL1) complex. Upon incorporation into this complex, RBBP5 mediates p16^INK4A activation and the establishment of the senescent phenotype in fibroblasts [187].

Importantly, we highlight the possible role of additional genes like CBX1 (HP1β) and CBX2, in the epigenetic regulation of cellular senescence. Notably, senescent cells display lower levels of CBX1, while loss of CBX2, which is part of the polycomb repressive complex PRC1, leads to genomic instability as well as senescence-associated chromosomal rearrangements [188, 189]. CBX5 encodes the non-histone chromatin-associated heterochromatin protein 1α (HP1α), which serves a key role in maintaining heterochromatin higher-order organization [190]. Initially, HP1α acts as a multivalent architectural protein which ensures genome integrity by promoting chromatin remodeling, leading to the activation of the DDR pathway [191]. Interestingly, HP1α is also a well-established marker of SAHF [192]. Although SAHF formation is not universal (it is observed primarily in OIS), HP1α has been identified as crucial for maintaining chromatin structure during RS as well [190].

Among the identified genes, the histone chaperone cell cycle regulator (HIRA) is known to be required for histone and chromatin control in senescent cells [193]. Specifically, previous work has shown that HIRA co-localizes with heterochromatin protein 1 (HP1) proteins into promyelocytic leukemia (PML) nuclear bodies, prior to their incorporation into SAHF [194]. PML bodies participate in the sequestration, modification, and degradation of various proteins and their detailed description is provided in Section"Higher-order chromatin structures"[195].

DNMT3L, which interacts with DNMT3A and DNMT3B, and enhances de novo DNA methylation was identified by our analysis as a key regulator [196]. Interestingly, mouse embryonic fibroblasts (MEFs) derived from DNMT3L-knockout mice exerted premature senescence, with the latter being accompanied by reduced H3K9me3 and H3K27me3 levels, indicating chromatin alterations and derepression of senescence-associated genes [197]. Ultimately, SMARCB1, a tumor suppressor protein and subunit of the SWI/SNF chromatin modifying complex, may serve as a critical regulator of cellular senescence through chromatin remodeling, as its loss seems to promote senescence in cancer cells [198].

The nf-core/atacseq pipeline, a collection of standard industry tools, was utilized for each study with the default parameters, including the narrow_peaks flag, using as input the raw fastq files [340]. Downstream analysis of samples was performed in R programming language (version 4.4.2) using Seurat (version 5.2.1) and Signac (version 1.14.0) R packages [341–343]. Peak files from all 3 studies for each of the samples (52 samples) were combined into one list and processed using Seurat/Signac, removing zero-count peaks and calculating key metrics such as the strength of the nucleosome signal per cell (NucleosomeSignal function) as well as the TSS (TSSEnrichment function). Control samples were defined as samples with timepoints of 0 to 48 h from the OIS experiments as well as the htert-immortalized samples from the RS experiment, since they do not undergo replicative senescence. We estimated gene expression levels based on chromatin accessibility at promoter and gene body regions using GeneActivity function, while data were split based on the study of origin. Data were further processed by normalizing (NormalizeData function), identifying the top 2,000 Highly Variable Genes (FindVariableFeatures function), scaling (ScaleData function) and calculating Principal Component Analysis (PCA). Harmony integration strategy was used to integrate the data from the 3 different studies, while Uniform Manifold Approximation and Projection (UMAP) transformation was applied to the first 30 normalized Harmony PCA-derived components [344]. Samples, separated by study, were combined again using the JoinLayers function. Differentially expressed genes (DEG) were identified between control and senescent samples, as well as OIS versus RS senescent samples.