Abstract

Introduction

Staining patterns of diverse portions of the genome led Emil Heitz to coin the word “heterochromatin” to describe the appearance of dark stained domains of mitotic chromosomes, compared to unstained areas (euchromatin) [1]. Subsequently heterochromatin was observed to fall into two distinct categories: constitutive, compacted genomic areas formed in many cell types at centromeres and telomeres; and facultative, locus- or cell type-specific heterochromatin [1]. Constitutive heterochromatin typically marks repeat-rich sequences and prevents recombination of conserved genomic portions between chromosomes (Figure 1), while facultative heterochromatin is thought to silence expression of cell type-inappropriate protein-coding genes [2,3]. For comprehensive reports, see two recent reviews on heterochromatin structure and function [4,5]. This review will focus on a newly appreciated role of H3K9me3 heterochromatin involved in gene regulation.

Open in a separate window

Figure 1

H3K9me3 overview.

(A) Top-to-bottom: representation of H3K9me3 heterochromatin, its function, and the main HMTases responsible for H3K9me3 deposition.

Heterochromatin is also defined by the presence of distinct histone post-translational modifications (PTMs) [4,5], in particular, di- and tri-methylation of lysine 9 (H3K9me2/3) and tri-methylation of lysine 27 (H3K27me3) on histone H3. In many eukaryotes, H3K9me2 and H3K9me3 mark repeat-rich heterochromatin at telomeric and centromeric regions [6–8]. These modifications are established by specific SET domain-containing histone methyltransferases: G9a and GLP contributing to H3K9me1 and H3K9me2; and Eset/SETDB1, SUV39H1 and SUV39H2, catalyzing H3K9me3 [6,9–13]. The specific function of each enzyme has been inferred from in vitro studies, with potential redundancy and cooperation observed in vivo [14–16]. Suv39h1 and Suv39h2 are able to methylate H3K9me0, but prefer H3K9me1 to establish H3K9me3 [6,14]. Setdb1, instead, is able to mono-, di-, and tri-methylate H3K9me0, in vitro [12,14,17,18]. However in vivo, a triple knock-out of Suv39h1, Suv39h2 and Setdb1 led to a loss of H3K9me3 but not H3K9me2 [19]. G9a and GLP have been shown to mono- and di-methylate H3K9me0 in euchromatic regions [10,11,20]. H3K9me2/3 are recognized by heterochromatin protein 1 (HP1), which, through self-oligomerization and interaction with other repressive modifications, ensures compaction, spreading and inheritance of heterochromatin [21]. Recently, the contribution of HP1 to heterochromatin has been suggested to occur through the formation of phase-separated droplets (see “Different physical state of heterochromatin” below) [[22]–24]. The repressive platform set up by H3K9me2/3-histone methyltransferases and HP1 favors the establishment of DNA methylation and maintenance of low histone acetylation [25].

H3K27me3 can modulate cell type-specific repression, silencing lineage inappropriate genes [26,27]. Polycomb repressive complex 2 (PRC2) deposits methyl groups on lysine 27, which can block transcription initiation [28]. A major difference between H3K9me3- and H3K27me3-marked heterochromatin regards DNA accessibility: regions marked by H3K27me3 can remain accessible to binding by transcription factors and paused polymerase, while H3K9me3 domains are generally not [29,30]. In addition to H3K9me2/3 and H3K27me3, other heterochromatin marks include H4K20me3, H3K56me3 and H3K64me3 [31–34]. As will be described further below, recent studies show that H3K9me2/3 and H3K27me3 marks do not necessarily predict compacted chromatin [19,35,36].

Mechanism of H3K9me2/3 establishment

Heterochromatin can be established with non-coding RNA, components of the RNA interference machinery, and transcription factor-mediated recruitment of H3K9me3 HMTases [4,5,29,37,38]. Both RNAi-dependent and RNAi-independent mechanisms are utilized to assemble heterochromatin [3,39]. In the fission yeast S. pombe and plants, constitutive heterochromatin formation requires components of the RNA interference (RNAi) machinery and transcription of the locus targeted for silencing [40–43]. S. pombe facultative heterochromatin, instead, has been shown to be established by an RNA-dependent mechanism involving the action of the EMC (Erh1-Mmi1) [44] and by Taz1-dependent process [45]. H3K9me3 heterochromatin spreading and restriction have been studied in diverse eukaryotes [46–49]. Studies in fission yeast highlight how a proper balance between the reader-writer and eraser factors is important for H3K9 methylation maintenance and inheritance [50,51]. It has been proposed that repressive marks represent a true epigenetic mechanism, in opposition to transcription activating PTMs, which require the continuous presence of initiators to establish and maintain active states [52].

Isolation and characterization of heterochromatin

While several assays have been established to detect and characterize open regions of the genome at a local level (e.g. FAIRE-seq, ATAC-seq, Sono-seq, [53–55]), few approaches directly map compacted heterochromatin [35,56]. A new method employed sucrose gradient sedimentation [56] to separate sonication-sensitive from sonication-resistant crosslinked chromatin (Figure 1A) [35]. The method, Gradient-seq, was coupled to genome sequencing and compared to gene expression and epigenetic marks. Gradient-seq characterized domains of sonication-resistant heterochromatin (srHC), largely overlapping with H3K9me3, in comparison to the euchromatic fraction of the genome (Figure 2A, ​,B)B) [35].

Open in a separate window

Figure 2

Isolation of srHC though Gradient-seq.

(A) Crosslinked, sonicated chromatin from BJ fibroblasts is separated in distinct fractions in a sucrose gradient to identify sonication-sensitive (euchromatin) and sonication-resistant (srHC) portions of the genome. Sixty-one (61) % of srHC domains are marked by H3K9me3, which is found also in euchromatin (3% of total H3K9me3). Thirty-two (32) and 29% of heterochromatic mark H3K27me3 is found in srHC and euchromatin, respectively. (B) Browser view of srHC and euchromatic fractions compared to H3K9me3 and H3K27me3 histone marks, and mRNA profiles. “DBRs” = differentially bound regions [30]; “srHC + H3K9me3” = H3K9me3 IP performed from srHC fraction. H3K27me3 data are from GSE16368. (C) left: euchromatic fraction and H3K9me3 IP from srHC fraction have been considered for proteomics analysis. Righ: Volcano plot showing 172 significantly identified H3K9me3 heterochromatin proteins (red).

SrHC strongly correlates with DNA methylation, Lamin B, DNase inaccessibility and late replication timing [35]. A combinatorial analysis with classical heterochromatin marks (i.e. H3K9me3 and H3K27me3) surprisingly revealed how a fraction of the repressive H3K9me3 (3% of the total H3K9me3) and H3K27me3 (~30% of the total H3K27me3) marks is clearly enriched on lowly transcribed genes at euchromatin sites [35]. Interestingly, a portion of srHC contains no post-translational modifications (PTMs), suggesting that other mechanisms of chromatin compaction might occur. Proteomic analysis of H3K9me3-marked srHC revealed 172 H3K9me3 heterochromatin proteins overlapping the srHC-enriched proteins (Figure 2C), among which the RNA Binding Motif Protein, X-linked (RBMX), which was shown to impede reprogramming of human fibroblasts into hepatocytes [35]. Interestingly, 6 of the 172 identified H3K9me3 heterochromatin proteins are linked to amyotrophic lateral sclerosis (ALS). Further genomic characterization of one of the ALS proteins, TDP-43, showed that germline mutation of the corresponding encoding gene, TARDBP, in ALS patient fibroblasts is associated with upregulation of genes normally buried in srHC domains [35]. The heterochromatic landscape emerging from these studies strongly indicates that biophysical features of chromatin represent a better predictor of heterochromatin than solely the histone PTMs.

Different physical state of heterochromatin

By focusing on H3K9me3 binding partner heterochromatin protein 1 (HP1), the Narlikar and Karpen laboratories published side-by-side studies inferring a novel role for HP1a protein in directing heterochromatin formation by liquid phase separation [22,23]. The presented observations suggest a model where heterochromatin exists in multiple physical states: a soluble one, representing the least repressed form and allowing DNA access; a liquid-droplet like state, regulating gene repression; and a gel-like state characterizing constitutive heterochromatin at centromeres and exerting structural functions during chromosome segregation [24]. While more studies are needed to elucidate the basis for phase-separated heterochromatin, distinct biophysical characteristics of heterochromatin impact the overall genome architecture.

Establishing H3K9me3 domains during development

Upon fertilization, maternal and paternal pronuclei undergo dramatic rearrangements aimed to re-set oocyte and sperm epigenetic landscapes and acquire a similar configuration during zygotic genome activation [57–59]. Whereas the maternal chromosomes in the zygote retain H3K9me3 at pericentric regions, the paternal pericentromeres show a striking divergence, with the absence of H3K9me3, low levels of H3K9me2, and an enrichment for Polycomb-dependent H3K27me3 and H3K9me1 [60,61]. At this stage, HP1-beta is associated with H3K9me1 [60], and thus ensure genomic stability prior to H3K9me3 deposition and also pre-mark domains for further modifications [57,60,61]. Maternal egg cytoplasmatic components contribute to pericentromeric heterochromatin formation on the paternal pronucleus [60,62], assuring the required chromosome arrangement and separation during following nuclear divisions [60]. In summary, progressive acquisition of H3K9me3 landscapes in maternal and paternal chromosomes, around the time of zygote genome expression, occurs in early differentiation and lineage commitment [58,60,61,63].

Later in development, differences in the amount of heterochromatic regions across the genome emerge. Using electron spectroscopy imaging, Bazett-Jones and co-workers identified differences in chromatin compaction between embryonic epiblast (EPI) and the lineage-restricted primitive endoderm (PE) and trophectoderm (TE): EPI is characterized by highly dispersed, pluripotency-dependent chromatin, while PE and TE show domains of highly compacted fibers [64]. ES cells are characterized by elevated global transcriptional activity, consistent with their more decondensed chromatin and greater enrichment for active histone modifications [65]. Interestingly, ES cells express repetitive sequences and mobile elements, which are silenced in terminally differentiated cells. While H3K9me3 deposition at these elements might be required in post-implantation stages, pre-implantation in mouse (8-cell-to-blastocyst stage) is characterized by a progressive loss in repeat element expression, due to erasure of the activating mark H3K4me3, rather than acquisition of the repressive H3K9me3 [66].

By mapping the genome-wide distribution of H3K9me3 in mouse early embryos, Wang et al. provided the first description of H3K9me3-heterochromatin dynamics from pre-implantation to e8.5 embryos [67]. The data described substantial reprogramming of H3K9me3 in the parental genomes upon fertilization. Moreover, in pre-implantation embryos, the repressive modification is present at long terminal repeats (LTRs) and few protein-coding genes. Finally, the authors found lineage-inappropriate genes marked by H3K9me3 after embryonic implantation (e6.5 and e7.5; Figure 3A) [67].

Open in a separate window

Figure 3

H3K9me3 dynamics during development.

Main events characterizing dynamics of H3K9me3 during pre- and post-implantation, and after gastrulation. Images of zygote, 2-cell, morula and early blastocyst are from https://embryology.med.unsw.edu.au. Profiles of H3K9me3, DNA methylation and gene expression are from [67]. The panel describing dynamics of H3K9me3 and srHC at liver genes during hepatocytes differentiation from uncommitted definitive endodermal cells is form [19].

A new study of post-gastrulated embryos reveals dynamics of H3K9me3 at protein coding genes, upon morphogenesis. Using cell sorting, low number populations of cells (3*10⁴) were isolated from pooled e8.25 undifferentiated definitive endoderm and mesoderm, e9.5 hepatic and pancreatic progenitors, e18.5 immature beta cells, and 2 month-old mature hepatocytes and beta cells. Since Gradient-seq requires high cell number of cells (3*10⁷) and it could not be applied for low cell number samples, separation of deproteinized DNA sizes by magnetic beads defined srHC-seq [19,36], has been employed to discriminate between less and more sonication-resistant, crosslinked chromatin. Uncommitted endodermal and mesodermal cells from e8.25 embryos show the highest number of genes marked by both srHC and H3K9me3, while pre-gastrula stage cells [67] and e9.5 or older hepatic and pancreatic cells show a lower number of srHC and H3K9me3-marked genes (Figure 3B) [19]. Upon differentiation, cell function-specific genes, active in differentiated cells, and developmental-associated markers, expressed in uncommitted cells, lose and gain chromatin compaction, respectively. Coupling transcription profiles, chromatin compaction (srHC) and H3K9me3, genes in compacted chromatin (srHC) are more repressed when marked by H3K9me3 [19]. In accordance with dynamics in chromatin compaction, protein-coding genes expressed in adult hepatocytes and mature beta cells lose H3K9me3 from uncommitted endoderm to differentiated cells (Figure 3B). While the molecular mechanism by which H3K9me3-related histone methylstransferases are recruited to specific portions of the genome (and how srHC is established) to promote gene repression is not fully understood, overall these studies describe, for the first time, extensive dynamics of H3K9me3 in embryonic development.

H3K9me3 is required for establishment and maintenance of cell identity and ensures lineage stability

In murine ES cells, Setdb1, a histone methyltransferase catalyzing H3K9me3, has been shown to bind and silence developmental regulatory genes [68] and to function as co-repressor of Oct3/4 to suppress trophoblast genes [69–71]. Similarly, Loh et al. provided evidence that in murine ES cells, Oct3/4 positively regulates the expression of two key demethylases, KDM3A and KDM4C, which remove the heterochromatin-associated marks H3K9me2 and H3K9me3, respectively, from Tcl1 and Nanog, ensuring the maintenance of ES cell renewal [72]. By contrast, upon ES cell differentiation in vitro and in post-implantation embryos, there is heterochromatinization of the pluripotency-associated genes Oct3/4, Nanog, Stella, and Rx-1 [25,73].

New genetic studies address the role of H3K9me-mediating histone methyltransferases and interacting partners in development [6,9–11,19,74–80]. Single KO murine strains for either Suv39h1 or Suv39h2 show no developmental defects, apparently due to their redundant functions. However, double null (dn) mice exhibit aneuploidy and show perinatal lethality associated with genome instability with an expansion of tumor cells in the lymph nodes [76]. Spermatogenesis was severely impaired in Suv39h1/h2 dn mice, due to delayed entry in meiotic prophase and increased apoptosis of spermatocytes [76]. The aberrant chromosomal segregation in Suv39h1/h2 dn mice could be a consequence of improper repression of repetitive elements, which have been shown to be dynamically regulated in pre-gastrulation embryogenesis [67].

Homozygous mutations of SETDB1 causes inner cell mass growth defects and embryonic lethality [77]. Similarly, the chaperone Chromatin assembly factor 1 subunit A (Chaf1a/Caf1A) is crucial for early embryonic development [67] by interacting with SETDB1 and silencing proviruses and endogenous retroviruses (ERVs) [81]. Tamoxifen-inducible SETDB1 depletion in murine stem cells and embryonic fibroblasts revealed proviral reactivation [78]. Germline ablation of SETDB1 led to a decreased levels of H3K9me3 (together with lower levels of H3K27me3 and aberrant DNA methylation) at marked ERVs and long interspersed elements (LINE1) [79], with a reduction in primordial germ cells (PGCs). Conditional ablation of SETDB1 in hematopoietic stem and progenitor cells (HSPCs) showed a reduction in bone marrow (BM) and hematopoietic cells [80] associated with an increased cell death and impaired cell proliferation. SETDB1 was also shown to restrict non-hematopoietic genes to maintain HSPCs [80]. Interestingly, reduction in H3K9me3 levels at ERVs in HSPCs were milder compared to the ones described in ESCs [77,78], suggesting that later developmental stages are less affected by repeat element activation than the earliest stages of development.

Mouse embryos that were genetically depleted for Suv39h1, Suv39h2 and SETDB1 (TKO) at the endoderm stage, lose the hepatic transcriptional profile, as evidenced by single cell RNA-seq data in isolated hepatoblasts; despite expressing Albumin, TKO hepatoblast transcriptional profile clusters separately compared to wt and Setdb1 single KO cells [19]. Adult TKO hepatocytes upregulate many non-liver genes [19]. Interestingly the profound genome instability scored in Suv39h1/h2 germline double null mice [76] was not detected in endoderm-driven TKO adult hepatocytes [19], again consistent with the notion that H3K9me3 suppression of repeat elements may be most important in very early embryogenesis. These results highlight the need to heterochromatinize many protein-coding genes to permit lineage fidelity, while allowing loss of heterochromatin at appropriate differentiation genes.

H3K9me3 heterochromatin acts as an impediment to cell reprogramming

H3K9me3 domains have been shown to act as an impediment during the conversion of terminally differentiated cells into a different cell type [29,30,35,82]. Studying the binding patterns of the so-called “Yamanaka factors” (i.e. Oct4, Sox2, Klf4 and c-Myc) during the conversion of human fibroblasts into induced pluripotent stem cells [83,84], megabase-scale regions where the four factors could not initially bind the fibroblast genome, but could bind in pluripotent cells, were identified [30]. These domains, defined as “differentially bound regions” (DBRs), span important pluripotent genes and are enriched for H3K9me3. Depletion of the HMTases responsible for deposition of H3K9me3 increases Oct4 and Sox2 binding and the number of iPS colonies [30]. Reprogramming mediated by somatic cell nuclear transfer (SCNT, [85]) also is improved by suppressing H3K9me3 heterochromatin [82]. More recently, reduction of H3K9me3 via injection of human Kdm4d mRNA to the SCNT embryo, coupled to deacetylase inhibitor treatment with trichostatin A (TSA), have been shown to improve blastocyst development and the pregnancy rate of transplanted SCNT embryos in surrogate monkeys, leading to the first primate cloning [86]. Classifying genes according to their heterochromatic histone modification signature, Becker et al. found that hepatic genes which fail to be activated upon cell conversion are marked by H3K9me3 in fibroblasts, while H3K27me3-marked genes show a modest defect in activation [35]. Overall, these results indicate that H3K9me3 heterochromatin prevent cell conversion regardless of the methodology applied, and suggest that H3K9me3 helps maintain cellular identity.

The stability of lineage commitment and maintenance of cell identity have been also shown to be regulated by H3K9me3-mediate heterochromatin in terminally differentiated cells [87–89]. During immune CD4+ T cell differentiation, Suv39h1 is pivotal for ensuring repression of T helper 1 (Th1) genes in Th2 cells, by establishing high levels of H3K9me3 at the expense of H3K9ac, and recruiting HP1 [87]. Similarly, Liu et al. showed that silencing of pluripotency and neuronal-lineage genes is required during oligodendrocyte differentiation [88]. Inhibition of H3K9me3, but not the heterochromatic mark H3K27me3, impairs oligodendrocyte differentiation and functionality [88]. Repression of stem cell genes via Suv39h1-mediated H3K9me3 has been further confirmed in CD8+ T cells [89], where deposition of H3K9me3 is required to restrict potency and promote cell identity establishment [89]. In adult livers, conditional triple knock-out of Setdb1, Suv39h1 and Suv39h2 leads to derepression of non-hepatic genes and failure in activating mature hepatocyte markers, without affecting chromosome stability [19]. Overall, these studies highlight the repressive nature of H3K9me3 at protein-coding genes and describe a function for this modification in regulating cell fate determination and maintenance.

Concluding remarks

The precise mechanisms of H3K9me2/3 establishment, and specifically the role of RNA-dependent processes, need to be defined for the mammalian genome, presumably to be inspired by the advanced state of understanding of facultative heterochromatin dynamics in S. pombe [44,45]. Similarly, the molecular events responsible for targeting H3K9me3-related HMTases in comparison to H3K27me3-related enzymes, to specific genomic regions requires further investigation. In this regards, interaction of the different HMTases with specific DNA-biding proteins and ncRNAs [90], could confer specificity to the HMTases Following on the phase separation phenomenon in heterochromatin formation, it will be important to better understand what causes such aggregation and how the cell regulates the formation of HP1-mediated droplets [24]. Sonication resistance of crosslinked heterochromatin (srHC) has been shown to be a more reliable feature to identify compacted portion of the genome, compared to histone PTMs [19,35]. Yet various srHC domains are not enriched for H3K9me3 and H3K27me3 [19,35,36], so further studies are required to better understand how the domains become condensed and transcription is silenced. Proteomic analysis on H3K9me3-heterochromatin revealed pivotal candidates associated with human disease [35]. Targeting heterochromatin factors represents a novel and promising route in regenerative medicine, aiming at generating accurate and stable cell identity by both regulate proper gene repression and dynamics of repetitive elements in the genome. It is important to highlight that further approaches are also needed to elucidate and discriminate whether H3K9me3 dynamics represent the cause or the consequence of in vivo cell fate determination. The combinatorial use of the CRISPR-Cas9 technology, single cell approaches and sequencing technique provides an extremely valuable set of tools to precisely dissect biological questions. Furthermore a deeper understanding on how heterochromatin is initiated, established and maintained is needed to modulate heterochromatin factors and generate functional cell at will.

Acknowledgments

Footnotes

References