DNA Methylation in Eukaryotes: Regulation and Function
In this chapter we focus on the regulation and function of DNA methylation in mammals and especially in humans. We describe the main features of the enzymatic machinery generating 5-methylcytosine (5mC) that functions as an epigenetic mark in mammalian cells, and outline the active and passive mechanisms that can remove this reversible modification of DNA. We briefly introduce the characteristics of “maintenance” and “de novo” DNA-(cytosine-C5)-methyltransferases (DNMTs) and overview how their expression is regulated at the transcriptional, posttranscriptional, and posttranslational level. The interacting partners and chromatin marks involved in the targeting of DNMTs to the replication foci during S phase or to various chromatin domains during other phases of the cell cycle are also discussed. The enzymatic functions of DNMTs and their interactions with cellular macromolecules are involved in a series of cellular processes, some of them vital for mammals. Thus, DNA methylation has a role in the regulation of chromatin structure and promoter activity. It may silence the promoters of imprinted genes showing monoallelic expression as well as the promoters of transposons, and contributes to gene silencing on the inactive X chromosome, too. There are genome-wide demethylation and remethylation events during embryogenesis suggesting a regulatory role for DNA methylation in developmental processes, and both cytosine methylation and the active removal of 5mC from DNA is involved in the control of cell differentiation. DNA methylation plays a role in the preservation of genomic stability and gene body methylation affects the inclusion of certain exons into mature mRNA molecules by affecting – indirectly – the splicing of primary transcripts. Epigenetic regulatory mechanisms, including DNA methylation, are at the forefront of brain research these days. For this reason we outlined some of the most interesting results of this exciting new field in a separate subsection.
Modified bases occurring in bacterial and mammalian genomes
N6-methyladenine (6 mA)
N6-methyladenine (6 mA)
Oxidation products of 5mC
The expression of bacterial DNA MTases that modify the host cell genome is frequently accompanied by the expression of a sequence-specific endodeoxyribonuclease (restriction endonuclease) recognizing the same DNA sequence as the MTase. Such MTase/restriction endonuclease pairs may protect bacterial cells from invading foreign bacteriophage or plasmid DNA molecules that are either unmethylated or methylated at different recognition sites (Ishikawa et al. 2010). “Solitary” or “orphan” DNA methyltransferases that do not have a corresponding restriction endonuclease pair also modify bacterial genomes (reviewed by Casadesus 2016). In addition to the modified bases occurring in bacterial genomes, DNA genomes of bacteriophages may contain a series of additional modified purines and pyrimidines, possibly to avoid recognition and prevent degradation by host-encoded restriction endonucleases (Shabbir et al. 2016; Weigele and Raleigh 2016).
In contrast to bacteria, mammalian genomes typically do not contain detectable levels of the modified bases N4-methylcytosine and N6-methyladenine. N6-methyladenine was detected, however, in other organisms belonging to the domain Eukarya: it was enriched at active transcription start sites in the genome of Chlamydomonas reinhardtii, a flagellated protozoan, and it was also observed in the ciliate protozoa Oxytricha fallax, Paramecium aurelia, Stylonichia mytilius, and Tetrahymena pyriformis (Fu et al. 2015b; reviewed by Wion and Casadesus 2006). N6-methyladenine was also detected in the fungus Penicillium chrysogenum and in the genomes of the nematode Caenorhabditis elegans and the fruit fly Drosophila melanogaster (Rogers et al. 1986; Greer et al. 2015; Zhang et al. 2015a; reviewed by Heyn and Esteller 2015; Meyer and Jaffrey 2016). Moreover, in a recent study, Liu et al. detected N6-methyladenine, with the help of ultrahigh performance liquid chromatography coupled to triple quadrupole mass spectrometry (UHPLC-QQQ-MS/MS), during the early embryogenesis of vertebrate species, zebrafish (Danio rerio) and pig (Sus domesticus). Certain repetitive sequences of these vertebrate genomes were especially enriched, temporarily, in N6-methyladenine (Liu et al. 2016). Similar results were reported by Koziol et al. who also used UHPLC-MS/MS as well as DNA immunoprecipitation with N-6-methyl-deoxyadenosine specific antibodies followed by high throughput sequencing for the analysis of Xenopus laevis (African clawed frog) and Mus musculus (house mouse) DNA samples. Using dot blot, N-6-methyl-deoxyadenosine, the nucleoside corresponding to N6-methyladenine, was detected in the DNA of a human cell line (293T) as well (Koziol et al. 2016).
The enzymatic machinery involved in the deposition of N6-methyladenine marks on eukaryote genomes remains to be elucidated. It was observed, however, that during the embryogenesis of Drosophila melanogaster a DNA N6-methyladenine demethylase (DMAD) removed the methylation mark, especially from transposons, suggesting a role for DMAD in transposon suppression (Zhang et al. 2015a).
Notwithstanding the observations documenting the presence of N6-methyladenine in eukaryote genomes, the vast majority of data accumulated so far indicated that the most abundant modified base in mammalian cells is 5mC . Discovered as a component of mammalian DNA by Wyatt in 1951, 5mC occurs predominantly in the dinucleotide CpG (Wyatt 1951; reviewed by Li and Zhang 2014). Oxidation products of 5mC, including 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC), can also be detected in mammalian genomes (reviewed by Rasmussen and Helin 2016) (Table 1). In this chapter we wish to focus primarily on the regulation and function of CpG-methylation in mammals and especially in humans. DNA methylation in other eukaryotic taxa including protists, fungi, plants, invertebrates, and vertebrates other than mammals was covered in several recent papers and reviews (Wion and Casadesus 2006; Walsh et al. 2010; He et al. 2011; Ponts et al. 2013; Vu et al. 2013; Dabe et al. 2015; Jeon et al. 2015; Huang et al. 2016; Taskin et al. 2016; Vidalis et al. 2016; Zabet et al. 2017).
Prokaryotic DNA MTases modify, with a few exceptions, all of their target sequences, and it was suggested that local hypomethylation in bacterial genomes may be due to the competition of site-specific DNA-binding proteins with DNA MTases (Ardissone et al. 2016). In contrast, in large-genome eukaryotes, including mammals, the methylation pattern of the genome is typically discontinuous and changes during developmental processes, organogenesis, and cell differentiation (Kunnath and Locker 1982; Bird 1986; Frank et al. 1990; Frank et al. 1991; Deaton and Bird 2011; Yu et al. 2011; Hansen et al. 2014; Bestor et al. 2015; Keil and Vezina 2015; Farlik et al. 2016; Li et al. 2016; Zhou et al. 2017). In mammals, a typical feature of the genome is the presence of predominantly unmethylated, short interspersed sequences called CpG-islands that are regularly associated with promoters. CpG-islands have a high GC content and they are enriched in CpG-dinucleotides compared to the average genomic pattern (Deaton and Bird 2011; Jones 2012).
2 DNA Methylation in Eukaryotes: Basic Facts
DNA-(cytosine-C5)-methyltransferases and related proteins in human cells
De novo methylation
De novo methylation
Regulation of DNA methylation
Methylation of tRNAAsp
2.1 DNMT1: The Maintenance DNA-(Cytosine-C5)-Methyltransferase in Human Cells
Unlike the C-terminal domain that resembles bacterial methyltransferases, the larger N-terminal domain of DNMT1 is of unknown origin and it has no significant amino acid sequence homology with the corresponding bacterial enzymes (Bestor 2000). Comparison of the mouse Dnmt1 protein with the amino acid sequences of other eukaryotic methyltransferases revealed that the N-terminal domain itself is composed of two parts, A and B (Margot et al. 2000). Thus, the mouse Dnmt gene, and the ancestral methyltransferase gene of Metazoa, was possibly generated by two fusion events: joining of the sequences coding for part A and B of the N-terminus was followed by the fusion to a putative gene coding for the C-terminus.
The N-terminal domain has a dual role in the regulation of the C-terminal catalytic domain. On the one hand, it is indispensable for the activation of the enzyme; on the other hand its interaction with the DNA binding pocket of the catalytic domain blocks substrate binding (Syeda et al. 2011; reviewed by Qin et al. 2011; Jeltsch and Jurkowska 2016). Thus, the N-terminal sequence mediating the latter autoinhibitory function should be sterically rearranged – possibly with the help of UHRF1 that interacts with DNMT1 – for the activation of the enzyme (reviewed by Qin et al. 2011; Jeltsch and Jurkowska 2016).
In addition to its enzymatic function, DNMT1 affects a series of other cellular processes by the interactions of its N-terminal regulatory domain with numerous nuclear proteins (Qin et al. 2011). Changing a critical catalytic cysteine residue to serine in the C-terminal domain of Dnmt1 abolished, however, several important biological phenomena attributed to the enzyme, suggesting that the enzymatic activity is required for in vitro differentiation of embryonic stem cells, retrotransposon suppression, and proper localization of Dnmt1 (Damelin and Bestor 2007). The very same mutation resulted in developmental arrest and embryonal lethality in Dnmt1 ps/ps mice carrying the catalytic-defective allele (Takebayashi et al. 2007). The embryos died shortly after gastrulation, like the Dnmt1-null mutant (Dnmt c/c ) embryos and compound heterozygous (Dnmt1c/ps) embryos that expressed the mutant, inactive Dnmt1 protein only. In parallel, there was a significant decrease in the methyl-CpG content of repetitive DNA sequences of mutant embryos, compared to those of their wild-type counterparts (Takebayashi et al. 2007). These data suggested a vital role for the catalytic activity of the maintenance DNA-(cytosine-C5)-methyltransferase in mouse embryogenesis.
Major functions of the DNMT1 N-terminal domain
Transport of DNMT1 to the nucleus
Targeting of DNMT1 to the replication fork
Guided by PCNA and UHRF1
Formation of multiprotein repressor complexes
Interactions with MBPs, de novo DNMTs, PcG proteins, chromatin remodeling ATPases, tumor suppressors
In vitro, the purified human DNMT1 protein catalyzed the transfer of methyl groups not only to hemimethylated but also to unmethylated oligonucleotide duplexes, although it showed a strong, 7–21-fold preference for hemimethylated versus unmethylated substrates (Pradhan et al. 1999). It was also observed, that DNMT1 cooperated with the de novo DNA methyltransferases DNMT3A and DNMT3B (see Sect. 2.2) to ensure efficient methylation of unmethylated DNA duplexes, whereas DNMT3A and DNMT3B may contribute to maintenance methylation by acting on CpG-sites “missed” by DNMT1 (Fatemi et al. 2002; Liang et al. 2002; Jones and Liang 2009). Thus, DNMT1 may function both as maintenance as well as a “de novo” methyltransferase. Furthermore, DNMT1 was able to bind, in addition to CpG-dinucleotides, to non-CpG sequences as well, implicating that it may have a non-CpG methylase activity in vivo (Pradhan et al. 1999). Xu et al. speculated that induction of de novo methylation by DNMT1 by certain conditions may play a role in development or disease initiation (Xu et al. 2010).
2.2 “De Novo” Methyltransferases in Human Cells: DNMT3A and DNMT3B
DNMT3A and DNMT3B are expressed at high levels during embryogenesis and germ cell development, but they are typically downregulated in adult somatic cells (reviewed by Dan and Chen 2016). Unlike DNMT1, these enzymes do not discriminate between hemimethylated and unmethylated double stranded DNA substrates. They methylate, in addition to CpG-dinucleotides, non-CpG sites as well. Non-CpG-methylation was frequently observed in embryonic stem cells, and it was also detected in diverse human tissues and organ systems as well as in plants (Lister et al. 2009; Becker et al. 2011; Schultz et al. 2015). DNMT3A has two major isoforms with DNA-(cytosine-C5)-methyltransferase activity whereas there are both enzymatically active and inactive members among the more than 20 isoforms of DNMT3B (reviewed by Choi et al. 2011; Dan and Chen 2016).
DNMT3L (DNMT3-like), a member of the DNMT3 family, does not function as a DNA MTase but may act as a regulator of DNA methylation: two DNMT3L molecules form tetramers with two DNMT3A or two DNMT3B molecules, respectively. Interaction with DNMT3L within such a butterfly-like structure stimulates the DNA methylation activity of the de novo methyltransferases (Suetake et al. 2004). DNMT3A preferentially methylates CpG-dinucleotides located about one helical turn apart, and DNMT3A/DNMT3L complexes and their murine counterparts are involved in the establishment of methylation patterns during embryogenesis (reviewed by Xu et al. 2010; Chen and Chan 2014). The murine Dnmt3L has a dual function. By interacting with both Dnmt3a and the unmethylated lysine 4 of histone H3, it targets the methyltransferase enzyme to heterochromatic regions and – in parallel – it also activates the enzyme (Jia et al. 2007). In addition, the N-terminal regulatory domain of the murine Dnmt3a enzyme also binds to the histone 3 lysine 36 trimethylation mark (histone H3K36me3) and guides the enzyme to repressed nuclear regions (Dhayalan et al. 2010).
Transfection of human DNMT3 isoform constructs into HEK 293T cells revealed that DNMT3A1 preferentially targeted active genes marked with H3K4me3, whereas DNMT3B1 methylated inactive genes marked with H3K27me3 (Choi et al. 2011). In addition, the DNMT3B1, DNMT3B2, and DNMTΔ3B isoforms increased the methylation of LINE1 (Long Interspersed Nuclear Element 1) and Satellite-α (Sat-α) repeats, whereas the DNMT3A isoforms were ineffective or less effective (Choi et al. 2011). Even the DNMT3B isoforms which do not have catalytic activity may play a role in DNA methylation: Duymich et al. observed that such molecules recruited DNMT3A to gene bodies, and increased its activity in somatic cells (Duymich et al. 2016).
In murine somatic tissues, the transcriptional repressor E2F6 may recruit Dnmt3b to a set of developmental regulator genes such as Hoxa11 and Hoxa13 involved in the establishment of the thoracic and lumbar vertebral identity as well as germline-specific genes including Tex11 expressed only in male germ cells, the RNA helicase Ddx4, and others (Velasco et al. 2010). Thus, Dnmt3b-mediated methylation at E2F6-marked promoters may ensure coordinated transcriptional silencing of a gene battery (reviewed by Walton et al. 2014). Expression of Oct-4, a gene coding for octamer-binding transcription factor 4 that plays an important role in the renewal and pluripotency of embryonic stem cells (ES cells), is also silenced by Dnmt3b, in collaboration with Dnmt3a during ES cell differentiation: it was observed that de novo methylation initiated at two sites upstream from the murine Oct-4 promoter was later spread to the proximal region of the promoter (Athanasiadou et al. 2010). Lsh (lymphoid-specific helicase) , a member of the SNF2 family of chromatin-remodeling ATPases, aided spreading of CpG-methylation (Athanasiadou et al. 2010). In mouse cells, Lsh collaborated with de novo DNMTs in the methylation of repetitive elements as well (reviewed by Huang et al. 2004). These included satellite repeats located to pericentromeric regions as well as retroviral LTR elements. Based on experiments with Lsh−/− mouse embryo fibroblasts, Termanis et al. suggested that Lsh may interact with de novo DNMTs deposited to distinct chromatin domains and initiate local chromatin remodeling necessary for efficient de novo methylation (Termanis et al. 2016).
Zinc finger proteins containing a Krüppel-associated box (KRAB) domain may also attract de novo DNMTs to distinct DNA sequences. KRAB containing repressor proteins associate with KAP1 (KRAB-associated protein 1), a regulator of development and differentiation. Quenneville et al. observed that in embryonic stem cells the KRAB/KAP1 complex induced heterochromatin formation and de novo DNA methylation in the vicinity of KAP1 binding sites. In ES cells, CpG-methylation was spreading to the nearby CpG-islands, and the pattern of CpG-island methylation was maintained in differentiated hepatocytes (Quenneville et al. 2012).
Distinct nuclear proteins binding to characteristic sequence motifs may block de novo methylation. In the female germline, a large number of CpG-islands acquired methylation during oogenesis, whereas a set of CpG-islands containing the sequence motif CGCGC, a binding site for the transcription factors E2f1 and E2f 2 involved in chromatin remodeling, resisted de novo methylation (Saadeh and Schulz 2014). Saadeh and Schulz suggested that binding of E2f1 and E2f 2 caused nucleosome depletion by recruiting a nucleosome remodeling complex; such a chromatin alteration could possibly hinder binding of Dnmt3a and Dnmt3b that associate with a subset of intact nucleosomes in mammalian nuclei (Jeong et al. 2009; Saadeh and Schulz 2014).
R-loops (R standing for RNA) formed at a subset of actively transcribed promoters located in CpG-islands may also hinder de novo methylation (Ginno et al. 2012; Ginno et al. 2013). R-loops are preferably formed during the formation of primary transcripts at genes characterized by “GC skew,” i.e., a significant strand asymmetry in the distribution of guanines and cytosines, immediately downstream from the transcription start site. The transcribed G-rich RNA forms a stable heteroduplex (RNA-DNA hybrid) with the C-rich template DNA strand, while forcing the nontemplate G-rich DNA strand into a largely single-stranded “looped” conformation. Ginno et al. detected R-loop formation at a significant number of human promoters and observed a reduced efficiency of DNMT3B1-mediated methylation upon R-loop formation (Ginno et al. 2012). They also described R-loop formation at the 3′-end of human genes, a phenomenon potentially involved in transcription termination (Ginno et al. 2013). In addition to DNMT3A and DNMT3B, the maintenance methylase DNMT1 may also perform de novo methylation (see Sect. 2.1). Fatemi et al. suggested that methylated or partially methylated DNA strands generated by DNMT3A may attract and activate DNMT1 to contribute to de novo methylation (Fatemi et al. 2002; Jeltsch 2006).
2.3 Active DNA Demethylation and Replication-Coupled Passive DNA Demethylation
Methylated DNA sequences in mammalian genomes may undergo local demethylation initiated by pioneer transcription factors that are capable to bind heterochromatic regions (Zaret et al. 2008; Serandour et al. 2011). In addition, general demethylation events affecting most of the 5mC -residues were also observed during mammalian development (reviewed by Clark 2015). These “global” DNA demethylation events occurring in preimplantation embryos and in primordial germ cells leave, however, distinct methylated loci unaffected: certain repetitive sequences as well as a series of imprinting control regions (ICRs , see 4.2), rare intragenic regions, and CpG-islands (see Sect. 4.1) are protected from demethylation (reviewed by Wu and Zhang 2014; Clark 2015; Li et al. 2016; Rasmussen and Helin 2016). Thus, in mouse embryos, the genome is transiently hypomethylated but not completely unmethylated, even in the inner cell mass of the preimplantation blastocyst where the lowest level of CpG-methylation was detected (Smith et al. 2012). Global hypomethylation was also observed in the human preimplantation embryo (Smith et al. 2014). Imprinting control regions and a set of CpG-island promoters were protected, however, from demethylation. A variable degree of hypomethylation was observed in species-specific repetitive elements (Smith et al. 2014).
It is also possible, however, that successive oxidation of 5mC is combined with passive demethylation, because oxidized cytosine bases (5hmC, 5fC, and 5caC) may interfere with the maintenance methylation machinery, resulting in the replication-dependent passive dilution of 5mC (Wu and Zhang 2014). Such a combined mechanism, i.e., “active modification (AM) followed by passive dilution” (AM-PD) may act during the erasure of paternal genome methylation in preimplantation embryos as well as in primordial germ cells (Wu and Zhang 2014). Alternatively, passive demethylation may also occur without the involvement of 5mC-oxidation, due to the inefficient recruitment of DNMT1 to the replication foci or due to the presence of DNMT1 inhibitors during successive rounds of DNA replication and cell division (reviewed by Smith and Meissner 2013b; Wu and Zhang 2014). In the zygote, a passive dilution mechanism that follows a slow kinetics may play a role in the 5mC depletion of the maternally derived genome (Hackett and Surani 2013). This phenomenon is possibly based on the exclusion of the oocyte-derived DNMT1o and UHRF1 from the DNA replication machinery (Seisenberger et al. 2013b; Wu and Zhang 2014).
3 Regulation of DNA Methylation in Eukaryotes
The methylation pattern of eukaryotic genomes depends on the expression levels of DNA-(cytosine-C5)-methyltransferase genes. In mammals, the transcriptional activity of these genes depends on the cell type, and the cytoplasmic level of the corresponding mRNAs is modulated by posttranscriptional regulatory mechanisms. After translation, targeting of DNA methyltransferase enzymes to various chromatin domains represents another level of DNA methylation control in mammalian cells. Targeting is affected by posttranslational modifications of DNMTs and by the interacting protein or RNA partners of these enzymes. In addition, certain activating or repressing chromatin marks recognized by DNA methyltransferases may also influence their localization to distinct chromatin domains.
3.1 Transcriptional Regulation of Human DNA-(Cytosine-C5)-Methyltransferase Genes
Expression of the human DNMT1 gene or its murine counterpart is modulated by the binding of nuclear proteins to the regulatory sequences of these genes. Sp1 and Sp3, members of the specificity protein (Sp) family of transcription factors, bound to a cis element in the 5′-flanking region of the mouse Dnmt1 promoter and stimulated its activity; Sp1 upregulated DNMT1 transcription in human cell lines as well (Kishikawa et al. 2002; Lin et al. 2010; reviewed by Lin and Wang 2014). DNMT1 transcription is also affected by BRCA1, a protein encoded by Breast cancer-associated gene 1. Shukla et al. observed that BRCA1, a pleiotropic regulator and breast tumor suppressor bound to the regulatory sequences of DNMT1 at a potential OCT1 (octamer binding protein 1) site, induced activating histone modifications in the region, and upregulated DNMT1 transcription (Shukla et al. 2010).
The activity of the human DNMT3A and DNMT3B genes coding for de novo methyltransferases is controlled by multiple promoters: there are three TATA-less promoters for DNMT3A and two TATA-less promoters for DNMT3B (Yanagisawa et al. 2002). Binding sites for the ubiquitously expressed Sp1 and Sp3 transcription factors were identified in the regulatory sequences of the DNMT3A third promoter and in the control regions of the first and second promoter of DNMT3B (Jinawath et al. 2005). Overexpression of Sp3 increased transcription of both DNMT3A and DNMT3B in a transformed human cell line, suggesting a role for Sp3 in the control of de novo MTase genes (Jinawath et al. 2005).
The regulatory region of the human DNMT1 promoter and its murine counterpart contain binding sites for E2F1, a member of the E2F family of transcription factors involved in control of the cell cycle (McCabe et al. 2005). Binding of E2F1 to these recognition sequences activated the transcription of the maintenance methyltransferase gene both in murine and human cell lines (McCabe et al. 2005). Because E2F1 is released from its complex formed with the tumor suppressor protein RB in the late G1 phase of the cell cycle, these data indicate that the gene for maintenance DNA MTase in human and murine cells is co-regulated with a series of other genes involved in in DNA synthesis and S-phase progression.
A binding site for E2F1 was located to the regulatory region of the DNMT3A gene as well (Tang et al. 2012). Complexing of the tumor suppressor protein RB with E2F1 bound to this site resulted, however, in silencing of the DNMT3A promoter (Tang et al. 2012). In contrast, binding of WT1 (Wilms’ tumor 1) , a developmental master regulator, to the upstream sequences of the first and second promoter of DNMT3A activated transcription (Szemes et al. 2013).
In addition to RB, the tumor suppressor protein p53 also decreased DNMT1 transcription, by binding to a region in exon 1 of the gene (Peterson et al. 2003; Lin et al. 2010). The opposite situation, repression of p53 transcription by Dnmt1, was observed, however, in the developing pancreas of mice; Dnmt1 bound to the p53 regulatory region in pancreatic progenitor cells (Georgia et al. 2013). Dnmt1-mediated suppression of the pro-apoptotic p53 gene contributed to progenitor cell survival during pancreatic organogenesis.
MDM2, a human protein homologous to the murine Mdm2 (murine double minute 2) E3 ubiquitin-protein ligase, ubiquitinated p53 and also promoted proteasomal degradation of RB (Moll and Petrenko 2003; Sdek et al. 2005). Accordingly, MDM2 relieved the RB-mediated suppression of the DNMT3A promoter in a human lung cancer cell line (Tang et al. 2012). MDM2 also induced the expression of DNMT3B gene by destabilizing the repressor protein FOXO3a bound to the regulatory region of the gene (Yang et al. 2014).
3.2 Expression of Human DNA-(Cytosine-C5)-Methyltransferase Genes: Posttranscriptional Regulation
Processing of the primary transcripts may result in splice variant mRNAs encoding DNA MTase isoforms, whereas microRNAs attaching to mRNA molecules may facilitate the breakdown of their mRNA targets. In addition, microRNAs binding to mRNAs encoding transcription factors involved in the upregulation of MTase gene expression may affect DNMT transcript and protein levels indirectly. HuR, an RNA binding regulatory protein, may also modulate stability and translation of DNMT mRNAs.
3.2.1 Alternative Splicing
In human and mouse cells either the oocyte specific or the somatic cell-specific isoform of the maintenance DNA methyltransferase was detected. Due to the incorporation of a somatic cell-specific exon 1 into the mRNA, a longer protein was expressed in somatic cells, whereas the utilization of a downstream initiation codon located to the oocyte-specific exon 1 generated an N-terminal truncated enzyme (Dnmt1o in mice and DNMT1o in humans, respectively) that was expressed in oocytes and eight-cell embryos (Ratnam et al. 2002; Petrussa et al. 2014; reviewed by Dan and Chen 2016). As described above (see Sect. 2.1), both of the major isoforms of the de novo methylase DNMT3A possess enzymatic activity, whereas there are enzymatically active as well as inactive members among the more than 20 isoforms of DNMT3B (Ostler et al. 2007; reviewed by Choi et al. 2011; Dan and Chen 2016). A study by Gordon et al. revealed that inactive DNMT3B isoforms bound to the catalytically active DNMT3A and DNMT3B isoforms (Gordon et al. 2013). Similarly to the enzymatically inactive DNMT3L that interacted with and stimulated the activity of de novo MTases, binding of the inactive DNMT3B3 isoform to DNMT3A2 resulted in a modest increase in the activity of the latter. DNMT3B3 counteracted, however, the stimulatory effect exerted on DNMT3A2 activity by purified DNMT3L (Gordon et al. 2013). Interaction of DNMT3B4, another inactive isoform, with either of the active DNMT3B2, DNMT3A1, or DNMT3A2 isoforms significantly decreased de novo DNA methylation (Gordon et al. 2013). These data suggest that dysregulated expression of inactive DNMT3B variants may induce pathologic alterations by the modulation of the methylome and transcriptome in various cell types.
MicroRNAs (miRNAs) are short, nontranslated, regulatory RNA molecules modulating the level of most messenger RNAs (mRNAs) and proteins in mammalian cells (reviewed by Lin and Gregory 2015). Their precursors are either transcribed from independent transcription units or generated from the introns of primary transcripts and undergo RNase processing in the nucleus and in the cytoplasm. One strand of a mature, double stranded miRNA molecule, associated with the Argonaute protein, interacts with a short complementary sequence of a target mRNA molecule resulting in mRNA degradation or translational repression (reviewed by Lin and Gregory 2015). The mRNAs coding for DNMTs or for proteins regulating the expression of methyltransferase genes are also targeted by miRNAs in various mammalian cell types including primordial germ cells and somatic cells (reviewed by Denis et al. 2011). The expression pattern of miRNAs and the proteins involved in their biogenesis may change in various diseases including malignant tumors (reviewed by Lin and Gregory 2015). The situation is complex, however, because a miRNA may have multiple mRNA targets, and the 3′-untranslated region (UTR) of a mRNA, where most of the miRNAs attach, can interact with several miRNAs.
3.2.3 Stabilization of mRNAs by the RNA Binding Protein HuR
HuR (human antigen R) , an RNA binding protein, is able to associate with a series of RNA molecules in the nucleus and in the cytoplasm (reviewed by Grammatikakis et al. 2017). HuR is a pleiotropic protein affecting important cellular events. One of its partners is DNMT3B mRNA: in a human carcinoma cell line, interaction with HuR stabilized and increased the expression of DNMT3B RNA (Lopez de Silanes et al. 2009; reviewed by Denis et al. 2011). López de Silanes et al. speculated that, in addition to HuR, other RNA binding proteins and miRNAs may also modulate, in concert with regulators of transcription and splicing, DNMT3B mRNA and protein levels in the cytoplasm (Lopez de Silanes et al. 2009).
3.2.4 Posttranslational Modifications (PTMs)
Posttranslational, reversible, covalent modifications, including acetylation, ubiquitination, phosphorylation, methylation, and sumoylation, are able to affect the stability and catalytic activity of DNMT1 (reviewed by Qin et al. 2011; Scott et al. 2014; Jeltsch and Jurkowska 2016). Although numerous PTMs were mapped on various parts of the enzyme, in most cases the functional significance of the modifications remains to be established. Acetylation of distinct lysine residues followed by UHRF1 -mediated ubiquitination marked the mouse Dnmt1 for degradation whereas deubiquitination and deacetylation increased the stability of the enzyme. The outcome of phosphorylation at various serine residues was variable: depending on the targeted residue, it affected the interaction between the regulatory and catalytic domain, altered DNA binding affinity, disrupted or altered the interactions with cellular partner proteins, or stabilized the enzyme. Monomethylation at lysine 142 (K142) in late S phase promoted proteasomal degradation of the human DNMT1 whereas a lysine-specific demethylase stabilized the murine Dnmt1 enzyme (reviewed by Qin et al. 2011; Scott et al. 2014; Jeltsch and Jurkowska 2016).
Phosphorylation of the murine Dnmt3a near to the PWWP domain (see Sect. 3.2.5) increased the targeting of the enzyme to heterochromatin but reduced its activity (Deplus et al. 2014a). In contrast, peptidylarginine deiminase 4 (PADI4) citrullinated a region containing five arginines in the N-terminal domain, stabilized the enzyme, and upregulated its activity (Deplus et al. 2014b).
3.2.5 Targeting of DNMTs
Although DNMT1 binds with high affinity to hemimethylated DNA duplexes in vitro, during the S phase of the cell cycle it is guided to such target molecules by interacting with PCNA and UHRF1 (see 2.1). In addition to its association with the DNA replication machinery at replication foci during S phase, DNMT1 is also loaded to the chromatin during the G2 and M phases of the cell cycle (Easwaran et al. 2004). In a culture of mouse myoblast cells, Schneider et al. observed by super-resolution 3D-structured illumination microscopy and fluorescence recovery after photobleaching (FRAP) methods that in early S phase a fraction of Dnmt1 molecules binds transiently, via the PCNA-binding domain of its N-terminal regulatory region, to PCNA rings immobilized at replication foci in euchromatic regions (Schneider et al. 2013). This interaction may facilitate complex formation with hemimethylated CpG-sites located nearby. In late S phase, however, a stronger interaction, mediated by TS (targeting sequence, also called RFTS, replication foci targeting sequence, located to the N-terminal part of Dnmt1), directed the enzyme to replication foci as well as to the pericentromeric heterochromatin (Schneider et al. 2013). Schneider et al. suggested that maintenance methylation in densely methylated regions may continue even during the G2 phase (Schneider et al. 2013). DNMT1 is expressed in nonproliferating, postmitotic (G1 or G0) cells as well. Using in situ hybridization, Dnmt1 mRNA was detected in mature neurons of adult mice, and Dnmt1 protein was present in the nuclear and cytoplasmic fraction of mouse brain and spinal cord (Goto et al. 1994; Chestnut et al. 2011). These data suggested that in neurons Dnmt1 may fulfill a unique biological function unrelated to maintenance DNA methylation (see Sect. 4.9).
The nucleosomal structure of chromatin may limit the accessibility of DNA for DNMT1. In vitro model experiments with reconstituted nucleosomes revealed that Dnmt1 preferentially bound to and methylated linker DNA sequences at the entry and exit sites of the nucleosome (Schrader et al. 2015). In contrast, the nucleosomal core particle served as a substrate for Dnmt1 only in the presence of the chromatin remodeling ATPases Brg1 (Brahma-related gene 1) and ACF (ATP-dependent chromatin assembly factor) that regulate nucleosome movement and spacing (Schrader et al. 2015).
DNMT1 interacts with methyl-CpG binding proteins which bind to methylated CpG-sites with high affinity. These binding partners may target DNMT1 to highly methylated heterochromatic regions and may ensure promoter silencing in such regions by complexing with histone deacetylases (HDACs) (Qin et al. 2011). HDACs interact with DNMT1 as well, and by the deacetylation of lysine or arginine residues of histones, they facilitate the establishment of a more condensed chromatin structure. Thus, hypermethylated DNA sequences regularly associate with hypoacetylated histones in transcriptionally inactive chromatin domains (Qin et al. 2011). In addition, histone methyltransferases that deploy chromatin marks associated with transcriptional repression may also recruit DNMT1 to target genes (reviewed by Qin et al. 2011).
Unmethylated histone H3 lysine 4 (H3K4me0), a histone mark characteristic for heterochromatic regions, may affect target selection of the de novo DNA-(cytosine-C5)-methyltransferases DNMT3A and DNMT3B by binding to the ADD domain located to the N-terminal regulatory region of these enzymes (reviewed by Denis et al. 2011). The ADD (ATRX–DNMT3–DNMT3L) domain of ATRX, a chromatin remodeling protein mutated in the alpha-thalassemia and mental retardation X-linked syndrome (ATR-X), contains a H3K4me0-binding pocket, similarly to DNMT3 and DNMT3L (Dhayalan et al. 2011). Binding of the unmodified histone H3 tail to the DNMT3A ADD domain stimulates the activity of the enzyme by disrupting the interaction of the ADD domain with the catalytic domain (Guo et al. 2015c).
DNMT3L, the catalytically inactive regulatory factor of de novo DNMTs, also possesses an ADD domain in the N-terminal regulatory region. Introduction of a point mutation (D124A) into this domain resulted in the defect of spermatogenesis in mice homozygous for the mutant Dnmt3L gene (Vlachogiannis et al. 2015). The mutation disrupted histone H3 binding by DNMT3L and caused a reduction of both CpG- and non-CpG methylation in prospermatogonia. In these cells, hypomethylation of retrotransposons and endogenous retroviruses and alteration of gene expression pattern was also observed. In addition, defective spermatogenesis was characteristic for male mice (Vlachogiannis et al. 2015). Thus, DNMT3L, the adaptor protein of de novo DNMTs, plays an important role in the reprogramming of DNA methylation during spermiogenesis in mice (see Sect. 4.5).
Via its PWWP domain, characterized by a proline-tryptophan-tryptophan-proline motif, the N-terminal regulatory region of DNMT3A interacts with H3K36me3, another repressive histone mark. As outlined in Sect. 2.2, this interaction facilitates the association of DNMT3A with heterochromatin (Dhayalan et al. 2010). The PWWP domain of DNMT3B is also involved in directing the enzyme to repetitive sequences of pericentric heterochromatin (Chen et al. 2004).
DNMT3A and DNMT3B may directly interact with a series of sequence-specific transcription factors that target them to distinct sequences causing aberrant de novo DNA methylation during tumorigenesis (Hervouet et al. 2009; Blattler and Farnham 2013). Such interactions may also occur in normal cells. The exact mechanism for such targeted methylation events remains to be elucidated (Ficz 2015).
4 DNA Methylation in Mammals: Biological Functions
In bacterial cells, DNA methylation fulfills vital functions such as DNA repair, attachment of the bacterial chromosome to the cell membrane, regulation of DNA replication, and resistance to invading phage or plasmid DNA molecules (Kramer et al. 1984; Ogden et al. 1988; Wilson 1988; Waldminghaus et al. 2012; Makarova et al. 2013; Adhikari and Curtis 2016; Cohen et al. 2016). The latter function depends on the joint action of DNA MTase and restriction endonuclease pairs recognizing the same DNA sequence. To protect the bacterial genome from degradation by the corresponding endonucleases, prokaryotic methyltransferases modify essentially all of their recognition sequences. Unmethylated DNAs or DNAs methylated at other recognition sites are recognized as “foreign” by restriction endonucleases. This “protective” function of DNA methylation results in a monotonous, continuous, relatively stable modification of bacterial genomes. In contrast, in large-genome eukaryotes, including mammals, there are alternating methylated and unmethylated regions and certain sequences may gain or lose methyl groups, even within the same cell. Thus, mammalian and human genomes are characterized by discontinuous and changing patterns of DNA methylation (Bird 1986, 1987).
The genes coding for DNA-(cytosine-C5)-methyltransferase enzymes of eukaryotes and related genes encoding proteins with an altered enzymatic function or without enzymatic activity originated, most likely, from bacterial methyltransferases of restriction-modification systems (Rodriguez-Osorio et al. 2010; Iyer et al. 2011; Jurkowski and Jeltsch 2011). DNMT1 and the related eukaryotic enzymes had an independent origin from the family of enzymes where DNMT3A and DNMT3B belong. DNMT2, an RNA-(cytosine-C5)-methyltransferase with sequence homology to the DNA-(cytosine-C5)-methyltransferases apparently changed its substrate preference from DNA to RNA during evolution (Goll et al. 2006; Jurkowski and Jeltsch 2011). The catalytically inactive DNMT3L has sequence similarities to DNMT3A and DNMT3B but it is characterized by a shorter N-terminal and a truncated C-terminal domain. Comparison of DNMT1, DNMT3A, and DNMT3B with bacterial DNA-(cytosine-C5)-methyltransferases revealed that the eukaryotic enzymes acquired distinct N-terminal regulatory sequences that are absent from their bacterial counterparts. Based on the analysis of the mouse Dnmt1, Bestor proposed that acquisition of the N-terminal regulatory domain was associated with a novel function for DNA methylation: in large-genome eukaryotes cytosine methylation may play a role in the compartmentalization of the genome (Bestor 1990). This idea fits well to the presence of alternating domains of methylated and unmethylated DNA regions in mammalian genomes. Compartmentalization may facilitate the interaction of diffusible regulatory factors to the unmethylated domains, whereas the methylated regions would be inaccessible for such regulators (Bestor 1990).
Biological functions of DNA-(cytosine-C5)-methyltransferases in human and mouse cells
DNMT1, aided by DNMT3A and DNMT3B;
During preimplantation development: DNMT1s (somatic form);
8-cell stage: DNMT1o (oocyte-specific isoform)
De novo DNA methylation
DNMT3A and DNMT3B aided by DNMT1
De novo MTases; DNMT1o: maintenance of genomic imprinting in preimplantation embryos
DNMT1, DNMT3A, DNMT3B
X chromosome inactivation
De novo MTases are involved in stably silencing Xist on the active X chromosome (Xa);
Dnmt1 is involved in gene silencing on the inactive X chromosome (Xi)
Regulation of embryogenesis
DNMT1o, DNMT1s, DNMT3A, DNMT3B
Regulation of cell differentiation
DNMT3A, DNMT3B, DNMT1
Preservation of genome stability
DNMT1, DNMT3A, DNMT3B
Regulation of splicing
Methylation of weak exons by de novo DNMTs possibly excludes their transcripts from mRNAs
DNMT1, DNMT3A, DNMT3B
4.1 DNA Methylation as a Regulator of Chromatin Structure and Promoter Activity
In eukaryotic cells, epigenetic regulatory mechanisms ensure the heritable alterations of cellular states and the inheritance of gene expression patterns from cell generation to cell generation. The best studied epigenetic regulatory mechanisms include DNA methylation, histone modifications, and the Polycomb and Trithorax protein complexes that also modify histone molecules. Other epigenetic regulators that do not necessarily rely on reversible covalent modifications include variant histones, pioneer transcription factors , long noncoding RNAs, and nuclear proteins mediating long-distance chromatin interactions (reviewed by Minarovits et al. 2016). Furthermore, the localization of distinct promoters to open or closed chromatin domains (see below) within the nucleus also influences transcriptional activity, and a change in nuclear position may activate or silence promoters (reviewed by Gyory and Minarovits 2005). Thus, there is a complex, multilayered epigenetic regulatory system in mammalian cells.
DNA methylation, histone modifications, as well as Polycomb and Trithorax protein complexes modulate the structure of chromatin, i.e., the architecture of the DNA-protein complex characteristic for eukaryotes. Chromatin is built up from repeating units, called nucleosomes, that contain eight histone proteins (histone octamer) and a stretch of 147 bp long DNA wrapped around the histone octamer (Kornberg 1974). The nucleosomes are connected by linker DNA of variable length and stabilized by histone H1. Two molecules of each histone H2A, H2B, H3, and H4 form the octamer, which is organized as a heterotypic tetramer (H3–H4)2 with two associated dimers (H2A–H2B) (Kelley 1973; reviewed by Marino-Ramirez et al. 2005). Epigenetic modifications usually affect the long tails of histone H3 and H4 although modifications of H2A and H2B occur as well (reviewed by Minarovits et al. 2016).
In vertebrates, the typical epigenetic mark of DNA is cytosine methylation (see Sect. 2). In mammalian cells, methylated cytosines are frequently present in the control regions of inactive promoters and methylation of such sequences in reporter constructs suppresses transcription (reviewed by Robertson 2001). In contrast, a series of active promoters are typically located to unmethylated chromatin domains, called CpG-islands characterized by an elevated G + C content and an increased density of CpG-dinucleotides compared to the rest of the genome (Deaton and Bird 2011). DNA sequences within CpG-islands are kept methylation-free by the presence of histone H3 trimethylated at lysine 4 (H3K4me3), an activating chromatin mark associated with an “open” chromatin architecture. H3K4me3 may block de novo methylation by interfering with the association of the murine Dnmt3a to the histone H3 tail (Zhang et al. 2010). Active DNA demethylation is mediated by Tet1 in mouse cells (see Sect. 2.3). Tet1 may protect unmethylated CpG-rich sequences from stochastic, aberrant methylation by converting the newly methylated 5mC to its oxidation products that block Dnmt1 (Williams et al. 2011; Ji et al. 2014). The enrichment of histone H3K4me3 at CpG-islands is due to the preferential binding of CFP1 (CxxC finger protein 1), a component of the mammalian SETD1 (SET domain 1) complex involved in histone H3 K4 methylation, to unmethylated DNA sequences (Thomson et al. 2010). Tet2 and Tet3 also facilitate SET1 binding and transcriptional activation (Deplus et al. 2013; reviewed by Delatte et al. 2014).
Unmethylated CpG-islands and activating histone modifications mark euchromatic domains that are typically associated with active promoters. In contrast, highly methylated DNA sequences and repressive histone modifications are characteristic for heterochromatic regions where silent promoters are located. Promoter inactivation is frequently achieved by multiple epigenetic regulators that may form repressor complexes. Methyl-CpG binding domain proteins (MBDs) preferentially bind to DNA sequences containing one or more symmetrically methylated CpG-dinucleotides (reviewed by Bogdanovic and Veenstra 2009). MBDs interact with histone deacetylases, nucleosome remodeling complexes, histone methyltransferases, and Polycomb group complexes involved in heterochromatin formation and promoter silencing (Jones et al. 1998; Nan et al. 1998; Fujita et al. 2003; Sakamoto et al. 2007; Gopalakrishnan et al. 2008).
Recent studies established that the effect of DNA methylation on promoter activity depends on the density of CpG-dinucleotides in the promoter region. High CpG-density promoters are frequently located to unmethylated CpG-islands in various cell types and can be efficiently inactivated by DNA methylation (reviewed by Messerschmidt et al. 2014). We would like to note, however, that methylation-mediated silencing of CpG island-associated promoters is a frequent event during carcinogenesis. CpG-methylation can suppress the activity of intermediate CpG-density promoters as well. Such events may occur at the promoters of pluripotency genes or germ cell-specific genes in differentiated cells. CpG-methylation may not necessarily influence, however, the activity of low CpG-density promoters (reviewed by Messerschmidt et al. 2014). Jiang et al. argued that in vertebrate species high CpG-density promoters are typically associated with housekeeping genes that are active in a wide spectrum of tissues, whereas tissue-specific functions are characteristic for genes with low CpG-density promoters (Jiang et al. 2014).
In addition to the methylation pattern of promoters, CpG-methylation at enhancer sequences also affects gene transcription. Aran et al. observed that hypomethylated enhancers were associated with genes upregulated in cancer whereas enhancer hypermethylation correlated with the downregulation of gene activity (Aran et al. 2013). Aran et al. suggested that enhancers do not necessarily act in a cell-type-specific manner; instead, an unmethylated enhancer may regulate transcription levels in more than one cell type provided that the relevant promoter is unmethylated (the “dimmer switch” model for enhancer action) (Aran et al. 2013).
Although methylation of mammalian promoters is usually associated with gene silencing, there are examples for CpG methylation-mediated activation of gene expression, too. Bahar Halpern et al. observed that a core CpG-island (CpG2a) and a neighboring CpG-rich region located upstream of the FoxA2 promoter were highly methylated in the FoxA2 expressing human pancreatic islet cells and early human endoderm stage cells (CXCR4+ cells) derived from human embryonic stem cells (hES cells) (Bahar Halpern et al. 2014). In contrast, the very same sequences were hypomethylated in undifferentiated hES cells that did not express FoxA2. Two other CpG-islands located upstream or downstream from CpG2a were hypomethylated independently of FoxA2 promoter activity. Bahar Halpern et al. speculated that the paradoxically high level of CpG-methylation at CpG2a may prevent the binding of a repressor protein to FoxA2 regulatory sequences (Bahar Halpern et al. 2014). Thus, methylation-dependent gene activation – possibly mediated by DNMT3B – activated by an unknown signal – may switch on the expression of FoxA2, a master regulator of early endoderm formation.
In human osteoblast-like MG63 cells, CpG-methylation in the distal promoter region of the PDPN (podoplanin) gene also stimulated promoter activity (Hantusch et al. 2007). In breast carcinoma cell lines, high levels of IL-8 (interleukin 8) gene transcription correlated with methylation of two isolated CpG-sequences that were located outside CpG-islands, far upstream of the IL-8 promoter (De Larco et al. 2003). De Larco et al. suggested that methylation at these CpG-sites possibly prevented the binding of a repressor or affected chromatin remodeling (De Larco et al. 2003). Methylation of sequences located far downstream from the promoter may also upregulate transcription. Unoki et al. found that the methylated intron 1 of EGR2 (early growth response 2), a gene coding for a putative tumor suppressor protein, enhanced the activity of the EGR2 promoter (Unoki and Nakamura 2003).
We would like to add that BZLF1, an immediate early protein of Epstein-Barr virus (EBV , a human gamma herpesvirus), preferentially associated to its methylated recognition sequences and activated Rp, one of its target promoters (Bhende et al. 2004). High affinity of BZLF1 to its methylated responsive elements permitted the induction of lytic, productive replication of highly methylated, latent EBV genomes. Similarly to BZLF1, RFX, a cellular regulatory protein, was also able to bind to and activate a methylated promoter – without inducing local demethylation – at the major histocompatibility complex (MHC) (Niesen et al. 2005).
4.2 The Role of DNA Methylation in Genomic Imprinting
Genomic imprinting is a complex and fast evolving area of research. Here we outline only some of the most essential points related to the topic and refer only to a selected number of papers and reviews covering the field. In mammals, a set of “imprinted” genes shows monoallelic expression, i.e., they are transcribed either from the maternal or the paternal allele in diploid cells. Imprinted genes play an important role in the control of embryogenesis and in the formation of the placenta that regulates nutrient transfer between mother and fetus. Various genes located to different chromosomes in mammals became independently imprinted during evolution (Edwards et al. 2007). DNA methylation plays an important role in the regulation of imprinted genes, as suggested by the data of Li et al., who observed activation or repression of various imprinted alleles in Dnmt1 deficient mice (Li et al. 1993). Imprinted genes frequently form gene clusters and their expression is controlled by regulatory sequences called imprinting control regions (ICRs) or germline differentially methylated regions (gDMRs). The latter name refers to the fact that the parental alleles are distinguished from each other by epigenetic marks deposited to the ICRs in gametes and indicates that DNA methylation is the key epigenetic mechanism involved in the establishment and maintenance of the imprinting signal (Tucker et al. 1996; Reik and Walter 1998). Thus, the methylation pattern of the ICR influences whether the maternal or paternal allele will be silenced in somatic cells (reviewed by Colot and Rossignol 1999; Bartolomei and Ferguson-Smith 2011; Renfree et al. 2013; Barlow and Bartolomei 2014). Maternally methylated gDMRs are typically located at CpG-rich promoters, whereas paternally marked gDMRs are situated in intergenic regions (Renfree et al. 2013).
A methylated gDMR may also hinder the expression of a regulatory RNA molecule involved in gene silencing: in the placenta the methylated gDMR directly blocked the activity of a promoter where the transcripts of the Airn lncRNA (long noncoding RNA) are initiated on the maternal chromosome (reviewed by Barlow and Bartolomei 2014). Switching off the Airn promoter permitted the expression of neighboring genes of the Igf 2r cluster on the maternal chromosome, whereas on the paternal chromosome the unmethylated Airn promoter was active and the Airn lncRNA “coated” and suppressed – in cis – the genes of the Igf 2r cluster (reviewed by Braidotti et al. 2004; Barlow and Bartolomei 2014). The Airn lncRNA may inactivate genes by helping to establish a repressive chromatin structure, similarly to the Xist lncRNA which mediates X-chromosome inactivation (Brockdorff and Turner 2015) (see Sect. 4.4).
Recent studies indicated that in the gametes the number of differentially methylated genes was much higher than the number of imprinted genes (reviewed by Kelsey and Feil 2013). These data suggested that DNA methylation at ICRs is perhaps not a specifically targeted process, but the DNA methylation marks deposited during gametogenesis are preserved after fertilization at the imprinted loci, even during a wave of genome-wide DNA demethylation occurring in the preimplantation embryo (reviewed by Kelsey and Feil 2013; Barlow and Bartolomei 2014) (see Sect. 4.5). DNMT1o, the maternal isoform of DNMT1, may play a role in the maintenance of genomic imprinting in preimplantation embryos (Howell et al. 2001).
4.3 Transposon Silencing by DNA Methylation
A predominant fraction of mammalian genomes consists of repetitive sequences including satellite repeats, tandem repeats, and mobile genetic elements. Mobile genetic elements or transposable elements are capable to change their genomic position that may cause genomic instability due to the insertions, deletions, and genomic rearrangements associated with transposition events. Transposable elements or transposons may relocate within the genome either via an RNA intermediate that is reverse transcribed to DNA, or with the help of transposase enzymes that cut out and reinsert them into a new target site. Retrotransposons or retrotransposable elements comprise around 50% of mammalian genomes (Munoz-Lopez and Garcia-Perez 2010; Zamudio and Bourc’his 2010). They are relocating via an RNA intermediate and include the class of LTR retrotransposons, such as the proviruses of endogenous retroviruses (ERVs) that are flanked with long terminal repeats (LTRs). The other class of retrotransposons called non-LTR retrotransposons lack LTRs and comprise long interspersed nuclear elements (LINEs) and short interspersed nuclear elements (SINEs). DNA transposons transpose via a nonreplicative “cut and paste” mechanism. They make up around 3% of the human genome (Zamudio and Bourc’his 2010). Because transposons impose a potential threat to genome integrity and most of the 5mC in mammalian genomes resides in transposons, Yoder et al. proposed that a major function of DNA methylation is the suppression of transposons which they regarded as “parasitic” sequence elements (Yoder et al. 1997). The idea of methylation-mediated transposon silencing was supported by experiments demonstrating the reactivation of retroelements in mice with various defects of the DNA methylation machinery. Inactivation of Dnmt1, Dnmt3A and Dnmt3B, as well as deficiency in Dnmt3L, the de novo MTase cofactor “reanimated,” i.e., increased the expression of LINE and IAP (intracysternal A particle, an endogenous retrovirus) retroelements (reviewed by Zamudio and Bourc’his 2010). Global hypomethylation in cancer cells may also induce aberrant expression of LINE-1 (Miousse and Koturbash 2015).
Apparently, there is more than one layer of protection against transposon reactivation in mammalian cells. Induction of DNA hypomethylation in embryonic stem cells activated transposon transcription but also resulted in the accumulation of the repressive histone mark H3K27me3 at various transposon families that became re-silenced by the Polycomb-mediated reconfiguration of the chromatin (Walter et al. 2016). In addition, RISC (RNA-induced silencing complex) and piRNA (piwi-interacting RNA) pathway proteins may also suppress transposon activity by the degradation of retroelement RNAs (reviewed by Goodier 2016). Although transposable elements may cause genetic instability, their expression was regularly detected in mammalian oocytes and early embryos (reviewed by Evsikov and Marin de Evsikova 2016). These observations suggested that transposons – regulated by the cellular epigenetic machinery – may have a physiological function in mammals, especially during oogenesis and early embryogenesis. In these developmental processes or in stem cells of adult organisms, transposons may serve as a tool for the control of co-expressed gene batteries (Peaston et al. 2004; Macfarlan et al. 2012; Gifford et al. 2013; Evsikov and Marin de Evsikova 2016).
4.4 X Chromosome Inactivation
In somatic cells of female mammals, random inactivation of one X chromosome ensures dosage compensation for the products of X-linked genes between males and females. The majority of genes carried by Xi, the inactive X chromosome, are silent due to the concerted action of several epigenetic regulatory mechanisms (Csankovszki et al. 2001; Chaligne and Heard 2014). X chromosome inactivation (XCI) results in heterochromatinization of Xi which is relocated to the nuclear periphery in the G1 phase of the cell cycle as a Barr body and replicates late during S phase (Chen et al. 2016a). During the initiation of XCI, transcription of XIST (X-inactive specific transcript) is switched on (Gilbert et al. 2000). XIST is active on Xi but silent on the active X chromosome, Xa. On Xa, de novo DNMTs are involved in stably silencing Xist (Kaneda et al. 2004). The XIST transcript, a long noncoding RNA (lncRNA) , spreads in cis with the aid of L1 repetitive elements, coats the entire X chromosome to be inactivated, and induces the deposition of heterochromatic marks on the future Xi (Engreitz et al. 2013; Bala Tannan et al. 2014; Cerase et al. 2015). Transcriptional silencing of numerous genes located on Xi is possibly due to the eviction of RNA polymerase II from the chromatin or blocking its access to the chromatin due to the collapse of interchromatin channels (Smeets et al. 2014; Cerase et al. 2015). Cerase et al. suggested that silencing of transcription may actually initiate removal of euchromatic histone marks, deposition of heterochromatic histone marks, and DNA methylation (Cerase et al. 2015). Xi is characterized by a unique three-dimensional structure which is smoother and more spherical than its active counterpart, Xa (Pandya-Jones and Plath 2016). Silenced promoters are associated with deacetylated histones and PRC2, the Polycomb repressor complex depositing a repressive chromatin mark by methylating histone H3 lysine 27 (Gilbert and Sharp 1999; Brockdorff 2013). Although the XIST transcript initiates facultative heterochromatinization and gene silencing during random XCI, maintenance of the silenced state is due to the concerted action of several regulatory mechanisms, including XIST lncRNA, histone hypoacetylation, and DNA methylation (Csankovszki et al. 2001). Methylation profiling of human Xi and Xa revealed an increased methylation of CpG-islands associated with most of the silenced genes on the Xi. A subset of CpG-islands were less methylated on the Xi. Accordingly, some of the associated genes escaped XCI and showed biallelic expression (Sharp et al. 2011).
4.5 DNA Methylation as a Regulatory Mechanism of Embryogenesis in Mammals
The catalytic activity of the maintenance DNA-(cytosine-C5)-methyltransferase plays a vital role in mouse embryogenesis (Takebayashi et al. 2007) (see Sect. 2.1). The activity of Dnmt3a and Dnmt3b genes coding for de novo DNA methyltransferases (see Sect. 2.2) was also detected in mouse embryos using an expression reporter cassette targeted to these genes: during the early phase of embryogenesis , Dnmt3a was expressed at a moderate level in embryonic ectoderm and at a low level in mesodermal cells, whereas there was a high level of Dnmt3b expression in the embryonic ectoderm, accompanied with a weak expression in mesodermal and endodermal cells (Okano et al. 1999). At a later stage there was a ubiquitous expression of Dnmt3a in the embryos and a predominant expression of Dnmt3b in the forebrain and in the eyes (Okano et al. 1999). Inactivation of both genes in Dnmt3a −/− , Dnmt3b−/− mice caused embryonic lethality due to an arrest of growth and morphogenesis shortly after gastrulation (Okano et al. 1999).
In parallel with genome-wide demethylation, the pluripotent genes as well as a set of “germline defence” genes implicated in the regulation of transposon activity are reactivated, and – in female mouse embryos – Xi, the inactive X chromosome is activated (Hackett et al. 2012; reviewed by Saitou and Yamaji 2012; Messerschmidt et al. 2014). After colonization of the genital ridges the PGCs continue to proliferate, until they enter meiotic prophase in females and mitotic arrest in males (Hackett and Surani 2013; Dean 2014).
In PGCs, 5mC-erasure extends to the genomic imprints. It was suggested that this may permit the reestablishment of methylation marks at ICRs during oogenesis and spermatogenesis according to the sex of the cell. Transposons are also demethylated in murine PGCs, with the exception of certain IAPs (Hackett and Surani 2013). The process of “epigenetic reprogramming” was defined as the erasure of DNA methylation followed by the establishment of a new methylome mediated by de novo DNMTs (Morgan et al. 2005). Morgan et al. suggested that during gametogenesis such a reprogramming may build up sex-specific epigenotypes in mature gametes (Morgan et al. 2005). Thus, after the erasure of the paternal imprints the newly established imprints would correspond to the sex of the individual.
A global demethylation was also observed in PGCs isolated from human embryos, although the relative timing of distinct demethylation events such as the erasure of imprints differed between humans and mice (Guo et al. 2015a; reviewed by von Meyenn and Reik 2015). In vitro studies of PGC-like cells also indicated species-specific differences in reprogramming (von Meyenn et al. 2016). Demethylation of PGC methylomes in mouse and human embryos did not necessarily correlate, however, with gene activation: loss of 5mC did not result in a significant increase in the transcription of transposable elements (reviewed by von Meyenn and Reik 2015).
After genome-wide demethylation in PGCs, de novo methylation starts at embryonic day E13.5 in prospermatogonia or gonocytes within the fetal gonads of male mice, and the buildup of the male methylome is completed before birth (Stewart et al. 2016). In the female germline of mice, de novo methylation starts somewhat later, after birth, and the oocyte methylome is established by day 21 of postnatal development (Stewart et al. 2016). The intricate details of the consecutive epigenetic events shaping the sperm and oocyte methylomes in mice and humans are discussed in the recent review by Stewart et al. 2016.
Although the pattern of DNA methylation is usually stable in adult somatic cells, there is a spectacular change in the 5mC-content of mammalian cells during development: the genome undergoes almost complete demethylation during early embryogenesis (Guibert et al. 2012; reviewed by Hackett and Surani 2013). Between fertilization and implantation, the lowest level of DNA methylation was detected in the inner cell mass of the blastocyst. It is important to note, however, that although the erasure of the methylation mark begins already in the zygote, distinct DNA sequences in the maternal genome, marked by H3K9me2 and by binding of the maternal factor STELLA, stay methylated even in preimplantation embryos (Nakamura et al. 2012; reviewed by Kang et al. 2013). Paternal imprinted regions bound by the maternal factor STELLA also keep their methylation marks, similarly to IAP retrotransposons and DNA sequences located to centromeric heterochromatin (reviewed by Seisenberger et al. 2013a; Messerschmidt et al. 2014). In contrast, 5mC in the bulk of the paternal genome is accessible for TET3-mediated oxidation and the products of this process are removed by passive replication-coupled dilution. Alternatively, the activation of the base excision repair (BER) pathway may contribute to the demethylation of the paternal genome, independently of TET3 activity (Amouroux et al. 2016). Although the maternal pronucleus is not subjected to active demethylation, it is undergoing passive demethylation during subsequent cell divisions (reviewed by Seisenberger et al. 2013a; Messerschmidt et al. 2014) (Fig. 5).
Erasure of DNA methylation during early embryogenesis may play a role in the acquisition of pluripotency by the cells of the inner cell mass and by embryonic stem cells. Demethylation of the genes encoding pluripotency transcription factors such as Nanog that is methylated in the sperm facilitated this process (reviewed by Feng et al. 2010). After implantation, de novo methylation reestablishes distinct methylomes in somatic cell precursors and in early primordial germ cells (PGCs) and the methylomes are further modified during lineage-specific differentiation (Ji et al. 2010; Hackett and Surani 2013).
In summary, embryogenesis in mammals is a complex process, viewed as alternating phases of cellular differentiation that usually restricts developmental potential, and reprogramming events that allow the reconfiguration of the epigenome in the zygote and in developing germ cells, permitting the reconstruction of pluripotent or totipotent states (Reik 2007; Hackett and Surani 2013). Hackett and Surani speculated that epigenetic mechanisms, including DNA methylation may prevent reversion to preceding cellular states during mammalian development and may contribute to the maintenance of cell identity (Hackett and Surani 2013). In addition, cellular methylomes regularly undergo dynamic changes that create specialized, hypomethylated epigenotypes to ensure the realization of complex developmental programs and the successful reproduction of mammalian species (Reik 2007; Dean 2014; Messerschmidt et al. 2014).
4.6 DNA Methylation as a Regulator of Cell Differentiation
Although inactivating the maintenance DNA-(cytosine-C5)-methyltransferase Dnmt1 or knocking out of both de novo methyltransferase genes, Dnmt3a and Dnmt3b, disturbed embryogenesis and caused embryonic lethality in mice, undifferentiated mouse embryonic stem cells (ES cells) lacking all three of these enzymes survived and proliferated in tissue culture (Li et al. 1992; Okano et al. 1999; Tsumura et al. 2006; Takebayashi et al. 2007). These in vitro growing “triple knockout” mammalian cells lacking CpG-methylation kept their stem cell properties, expressed typical markers of undifferentiated cells, and their growth characteristics were comparable to that of their wild-type counterparts in a competition assay. Upon induction of cell differentiation by embryoid body formation, however, the growth of the “triple knockout” cells was delayed compared to the wild-type embryonic stem cells (Tsumura et al. 2006). These data suggested that CpG-methylation is dispensable for the maintenance of stem cell properties and proliferative capacity of ES cells. The above mentioned in vivo and in vitro observations also indicated, however, that DNA methylation plays an important role in the establishment and maintenance of cell differentiation.
Comparison of promoter methylation patterns (“the promoter epigenome”) of pluripotent mouse ES cells and embryonic germ (EG) cells with the promoter epigenome of differentiated primary embryonic fibroblasts revealed that a set of pluripotency–related genes including Nanog, Lefty1, and Tdgf1 were hypomethylated in ES cells and EG cells, but hypermethylated in fibroblasts (Farthing et al. 2008). In ES cells, promoter hypomethylation correlated with an increased gene expression in most cases. In contrast, targeted de novo promoter methylation silenced Nanog, a gene actively transcribed in ES cells under normal circumstances. Analysis of the promoter methylome in trophoblast stem cells (TS cells) representing the extraembryonic lineage showed a distinct pattern that differed from that of the other cell types (Farthing et al. 2008). Fouse et al. also observed that the housekeeping and pluripotency-related genes were unmethylated in mouse ES cells whereas most of the differentiation associated genes were repressed and methylated (Fouse et al. 2008). Comparison of wild-type ES cells with “triple knockout “ES cells lacking CpG-methylation revealed that demethylation was able to upregulate the expression of certain X-linked genes and genes implicated in cell differentiation. A series of genes, however, were not upregulated in the absence of CpG-methylation, possibly due to the action of other epigenetic regulators (Fouse et al. 2008).
Analysis of CpG-island methylation showed that ES cells, embryoid bodies, and teratomas had a characteristic, complex methylation pattern and indicated that cellular differentiation was associated with both de novo methylation and demethylation processes (Kremenskoy et al. 2003). In line with this observation, Dawlaty et al. studied the importance of cytosine demethylation in cellular differentiation (Dawlaty et al. 2014). They observed that knocking out of all three genes, Tet1, Tet2, and Tet3 that code for dioxygenase enzymes converting 5mC to 5hmC resulted in the complete loss of 5hmC in mouse ES cells and impaired the capacity of the “triple knockout (TKO)” cells to differentiate. The impaired potential for embryoid body and teratoma formation correlated with the hypermethylation and dysregulation of promoters driving the expression of genes involved in developmental processes and cell differentiation (Dawlaty et al. 2014). Most of the analyzed genes that were silenced in TKO cells by hypermethylation were active in wild-type ES cells. Thus, Tet-mediated DNA demethylation is indispensable for the establishment of the proper epigenotype and gene expression pattern involved in ES cell differentiation.
Tet enzymes play a role in the maintenance of chromosomal stability as well by regulating sub-telomeric methylation levels (Yang et al. 2016b). In addition, Wiehle et al. observed that in mouse fibroblasts Tet dioxygenases prevented the methylation of distinct genomic domains called DNA methylation canyons or DNA methylation valleys (Wiehle et al. 2015). Because genes involved in cellular differentiation are frequently located to such genomic areas, Wiehle et al. suggested that Tet activity may ensure the undisturbed access of chromatin regulators to key topological domains controlling developmental processes (Wiehle et al. 2015).
4.7 The Role of DNA Methylation in the Preservation of Genome Stability
In addition to transposon silencing (see Sect. 4.3), DNA methylation may contribute to the maintenance of genomic integrity by the stabilization of microsatellite repeats and via the recruitment of the mismatch repair surveillance complex to hemimethylated DNA sequences at the replication fork (reviewed by Putiri and Robertson 2011; Wang et al. 2016). Furthermore, establishment of an appropriate methylation pattern of centromeric and pericentromeric DNA sequences decreases the probability of chromosomal segregation defects and chromosome rearrangements. Finally, DNA methylation-mediated transcriptional silencing may preserve genome stability by decreasing the frequency of mitotic recombination. We would like to mention, however, that local CpG-hypermethylation may promote genome instability by silencing the genes encoding DNA repair enzymes, and 5mC is a potential source of mutations in itself because of its propensity to spontaneous deamination that results in thymine (5mC to T transition) which also contributes to human inherited disease (Cooper et al. 2010; reviewed by Putiri and Robertson 2011; Meng et al. 2015).
4.7.1 Stabilization of Microsatellite Repeats by Cytosine Methylation
Microsatellites are genomic elements composed of short tandem repeats of simple DNA motifs that are one to nine nucleotides in length (Sawaya et al. 2013). Formation of single-stranded DNA during the processes of DNA replication, transcription, or repair facilitates the formation of various secondary DNA structures including hairpins, Z-DNA, H-DNA (a DNA triplex), and G-quadruplex (G4) (reviewed by Putiri and Robertson 2011; Sawaya et al. 2013; Sawaya et al. 2015). During transcription, distinct trinucleotide repeats may transiently form R-loops containing an 8 bp long RNA:DNA hybrid (Reddy et al. 2011). The presence of these secondary structures may facilitate genomic rearrangements due to faulty repair or to replication slippage by DNA polymerase. Changes in repeat number within protein coding sequences cause neurological diseases in humans, and repeat number variations in protein-noncoding regulatory regions and promoter elements were also implicated in pathogenesis (Pearson et al. 2005; Sawaya et al. 2013). DNA methylation may stabilize repeat length by transcriptional repression of genes carrying either CpG-containing or non-CpG-containing microsatellites (reviewed by Putiri and Robertson 2011). Association of 5mC-rich sequences with distinct methyl-CpG binding proteins and repair proteins may also contribute to microsatellite stabilization , although in malignant cells heterochromatic histone modifications and CpG-island hypermethylation correlated with a high level of microsatellite instability (reviewed by Putiri and Robertson 2011).
4.7.2 Recruitment of the Mismatch Repair Surveillance Complex to Hemimethylated DNA Sequences
Wang et al. observed that in murine cells the postreplicative DNA mismatch repair (MMR) machinery was recruited to a complex of Dnmt1 and Uhrf1, bound to hemimethylated CpG-sites in newly synthesized DNA. The authors argued that the collaboration of the maintenance MTase and the MutSα MMR complex probably preceded chromatin assembly on the newly synthesized DNA duplex, because nucleosome formation could potentially block mismatch recognition by the DNA mismatch repair system (Wang et al. 2016). Such a coordinated action may ensure the correct transmission of both epigenetic and genetic information from cell generation to cell generation in mammals.
4.7.3 Methylation Patterns of Centromeric and Pericentromeric DNA Sequences: Functional Consequences
Yamagata et al. observed that the centromeric and pericentromeric DNA was hypomethylated in male and female germline cells of mice but hypermethylated in adult tissues with the exception of testis and epididymal sperm. In contrast, the IAP1 and LINE1 endogenous retroviral repetitive elements were highly methylated in the testis and sperm, like in other adult tissues. The authors speculated that the germline-specific hypomethylation of satellite sequences from centric and pericentric regions was associated with the expression of germ cell-specific genes transcribed in the testis and ovary (Yamagata et al. 2007). As a matter of fact, hypomethylated “pockets” were identified within hypermethylated sequences of a functional centromere (Wong et al. 2006). The kinetochore protein CENP-B (centromere protein B) preferentially attached to nonmethylated recognition sequences located within extensively methylated regions (Tanaka et al. 2005).
CENP-C, another constitutive centromere protein, interacted with DNMT3B and facilitated methylation of centromeric and pericentromeric satellite sequences (Gopalakrishnan et al. 2009). Both CENP-B and CENP-C enhanced the formation of heterochromatin and interacted with the centromere-specific histone variant CENP-A and possibly with a network of centromere-associated proteins to maintain centromere integrity (Okada et al. 2007; Gopalakrishnan et al. 2009; Putiri and Robertson 2011; Giunta and Funabiki 2017). Local or general hypomethylation of satellite CpG-dinucleotides caused, however, either chromosome rearrangements or chromosomal segregation defects in a subset of patients with hepatocellular carcinoma or in patients with ICF syndrome (immunodeficiency, centromeric instability, facial anomalies syndrome) , respectively (reviewed by Putiri and Robertson 2011).
4.7.4 Inhibition of Homologous Recombination by DNA Methylation
Dnmt1 deficiency in mice enhanced the rate of mitotic recombination, activated the transcription and transposition of endogenous retroviral elements, and facilitated tumorigenesis (Eden et al. 2003; Gaudet et al. 2003; Howard et al. 2008). DNA methylation also suppressed meiotic recombination in several organisms (reviewed by Termolino et al. 2016). In contrast, Sigurdsson et al. observed, based on the analysis of data sets from various sources, that in human male germ cells there was an elevated level of DNA methylation at recombinational hot spots (Sigurdsson et al. 2009). Sigurdsson et al. speculated that methylated regions could be preferential sites of meiotic recombination or recombined sequences could perhaps be targeted by DNMTs (Sigurdsson et al. 2009).
4.8 Regulation of Alternative Splicing by Gene Body Methylation
In eukaryotes, most of the genes are composed of exons and introns and their primary transcripts are processed by spliceosomes that excise the protein-non coding introns and unite the protein-coding exons to form mRNA molecules. Spliceosomes are large ribonucleoprotein complexes assembled co-transcriptionally (Will and Luhrmann 2011). In higher eukaryotes, alternative splicing of the same primary transcript frequently generates mRNAs with different combinations of exons: the so-called constitutive exons are regularly included into the mRNA whereas – depending on the rate of transcription elongation – “alternative” or “weak” exons are either included or skipped. Pausing of RNA polymerase II favors inclusion of “alternative” exons into the mRNA (reviewed by Shukla and Oberdoerffer 2012). The choice of splice sites is also affected by the structure of the chromatin and by RNA binding proteins. Recently, DNA methylation emerged as an important regulator of splicing. Shukla et al. observed that binding of the zinc finger protein CTCF to the unmethylated exon 5 of the CD45 gene elicited RNA polymerase II pausing that promoted the inclusion of this “weak” exon into the mRNA (Shukla et al. 2011). CTCF does not bind DNA, however, when its recognition sequence is methylated (Hark et al. 2000). Accordingly, Shukla et al. found that CpG-methylation at the alternative exon 5 of CD45 resulted, during the processing of the primary transcript, in the exclusion of exon 5 from the mRNA (Shukla et al. 2011). A possible explanation for this phenomenon was that cytosine methylation blocked the binding of CTCF to its recognition sequence, and in the absence of a local barrier created for RNA polymerase II by CTCF, the rapid rate of elongation quickly resulted in the synthesis of competing “strong” downstream splice sites, before spliceosome assembly at the “weak” upstream splice site. Thus, spliceosome assembly occurred only at these downstream sites, skipping the weak upstream site (Shukla et al. 2011; Shukla and Oberdoerffer 2012). Using CTCF depleted cells and a combination of ChIP-seq and RNA-seq methods, Shukla et al. demonstrated that in addition to CD45, processing of a series of other cellular transcripts is also regulated by CTCF-mediated RNA polymerase II pausing (Shukla et al. 2011). In contrast, Maunakea et al. found that DNA methylation and binding of the methyl-CpG-binding protein MeCP2 to alternatively spliced exons (ASEs) facilitated the inclusion of ASEs into mRNAs during the processing of the primary transcripts (Maunakea et al. 2013). Further studies may illuminate how CpG-methylation in gene bodies, especially in exons, may affect the complex process of RNA splicing in trans.
In a recent work, Marina et al. described that 5hmC and 5caC – the TET generated oxidation products of 5mC – facilitated CTCF binding to the CD45 “model gene” in T lymphocytes and promoted the inclusion of the alternative exon 5 into mRNA. They also observed that in vitro activation of naïve CD4+ T cells induced exclusion of exon 5, in parallel with a decrease in 5hmC and increase in 5mC at exon 5. The authors argued that 5hmC and 5caC promoted CTCF association and exon inclusion whereas 5mC blocked CTCF binding and facilitated exon exclusion (Marina et al. 2016). Oxidation products of 5mC bind a set of distinct nuclear proteins (Spruijt et al. 2013). This observation suggests that they may act as signals in important biological processes including splicing (see Sect. 4.9.2).
4.9 The Role of DNA Methylation in Brain Function
Epigenetic regulatory mechanisms, among them DNA methylation, are at the forefront of brain research these days. For this reason, we wished to outline some of the most interesting results of this exciting new field in a separate subsection. Oxygen is required for demethylation processes, and thus it is believed that first the increase in atmospheric oxygen allowed the appearance of reversible methylation and therefore regulable enzymatic methylation systems for DNA and proteins. Regulated methylation in turn was required for the emergence of multicellular animals and for organ development. Therefore, the Cambrian radiation was probably dependent on the parallel increase in atmospheric oxygen (Jeltsch 2013). Regulable DNA methylation at CpG-dinucleotides plays also an important role in the delineation of the human phenotype in comparison with other primate lineages. There is an estimated number of ~1.19 million nonpolymorphic species-specific CpG-dinucleotides, termed CpG-beacons, part of them extremely densely clustered into 21 genomic locations which are functionally associated with CpG-island evolution, human traits and diseases. Beacon clusters have been predicted to colocalize with accessible chromatin (Bell et al. 2012). Comparative primate epigenomic data from prefrontal cortex neurons showed that beacon clusters are indeed enriched for human-specific H3K4me3 peaks. Both beacon clusters and permissive chromatin signatures were enriched at telomeric chromosomal regions supporting a predisposition for recombination. In summary, epigenomic analysis was able to pinpoint chromatin structures that contribute to the human-specific phenotype (Bell et al. 2014). Regulated methylation may be of even higher importance for highly adaptive organ systems, like the immune system and the brain, than for the function of other organs. Regulated methylation within the brain may, of all biological properties, even be the most discerning between species. Comparative methylome analysis of human and chimpanzee brain tissue uncovered 85 human-specific and 102 chimp-specific differentially methylated regions (DMRs), approximately half of them located in intergenic regions or gene bodies. DMRs were enriched in active chromatin loops suggesting a human-specific organization of higher-order chromatin structure (Mendizabal et al. 2016).
4.9.1 Dynamic Methylation Within the Brain During Development and Disease
Methylation-profiling of CpG-dinucleotides in fetal frontal cortices using 450 K methylation arrays showed that methylation was continuously increasing at 1767 positions and decreasing at 1149 positions with gestational age. These altogether 2916 developmentally regulated differentially methylated positions (dDMPs) were enriched in gene bodies but underrepresented in CpG-islands. Overall, during the first trimester the human fetal brain is globally hypomethylated with methylation increasing in the second and third trimester. Because genes associated with dDMPs are assumed to be important for brain function, it is remarkable that dDMPs were also enriched for regions that have been associated with schizophrenia and autism. The authors concluded that this gestational age-matching set of dDMPs may have been adopted for both brain evolution and ontogeny (Schneider et al. 2016).
Organoids derived from pluripotent human cells recapitulated the very early three-dimensional organization of the fetal brain and were used to compare epigenomic and transcriptional programs and to establish suitability as an in vitro model for gene expression dynamics in the early to mid-term fetal human brain. Non-CpG methylation was enriched at super-enhancers in cerebral organoids and fetal brains, and non-CpG methylation in vitro was a predictor of impending super-enhancer repression in vivo. Overall, cerebral organoids recapitulated the large-scale epigenomic remodeling during the early fetal brain development. However, generally organoids tend to a higher methylation level at CpG-positions, and wide-spread pericentromeric demethylation was observed in the in vitro model but not in the fetal brain (Luo et al. 2016). The brain methylome undergoes a widespread reconfiguration from fetal to young adult age corresponding to synaptogenesis. Concomitantly, non-CpG methylation becomes the more prevalent form of methylation in neurons, but not glia. Interestingly, non-CpG methylation identified genes escaping X-chromosome inactivation (Lister et al. 2013). While above studies did not differentiate methylation states between gray and white matter, a first 450 K chip- and pyrosequencing-analysis of Brodmann area 9 which is located within the dorsolateral prefrontal cortex and is important for higher cognitive skills and affected in diverse neurological disorders demonstrated robust gray-white methylation differences. Gray matter corresponds predominantly to neuronal cells whereas white matter is mainly built up from glial cells. Especially cell type-specific markers were enriched among differentially methylated genes. A large number of the identified DMRs had previously been associated with degenerative neurological diseases, like Alzheimer’s (AD), Parkinson’s (PD), Huntington’s diseases (HD), amyotrophic lateral sclerosis (ALS), and multiple sclerosis (MS) (Sanchez-Mut et al. 2017).
It is commonplace that genetic diseases associated with aberrant methylation, e.g., fragile X syndrome, alpha-thalassemia X-linked mental retardation syndrome, Angelman syndrome, Beckwith-Wiedemann syndrome, Prader-Willi syndrome, ICF (immunodeficiency, centromere instability, facial abnormalities) syndrome, Rett syndrome, Rubinstein-Taybi syndrome, and others, are also associated with cognitive impairment (reviewed by Calfa et al. 2012; Eggermann 2012; Gatto et al. 2012). It is also to be expected that inherited dysregulation of methylation interferes with the differentiation of brain cells and therefore brain function. However, the role of Epigenetics in normal brain function and of Patho-Epigenetics of the previously functional diseased brain is only beginning to be uncovered (reviewed by Klausz et al. 2012; Zelena 2012).
4.9.2 Normal Brain Function and Memory
Evidence is mounting that epigenetic processes are involved in the everyday operation of the central nervous system, and that long-term memory formation and synaptic plasticity needs accompanying epigenetic modifications of DNA and histones. The role of epigenetic regulation in memory formation has recently been reviewed (Kim and Kaang 2017). Therefore, only a selection of illustrative examples with a focus on CpG-methylation is mentioned here. In order to understand the epigenetic changes associated with learning, genome-wide histone and methylation profiles from neuronal and nonneuronal cells of the CA1 field of the hippocampus and the anterior cingulate cortex of adult mice at three time points before and 1 h and 4 weeks after learning were established. This provided a physiological gene regulatory network of learning as a background for the analysis of neurological and psychiatric diseases and potential development of medical treatment (Centeno et al. 2016). In contrast to short-term memory, long-term memory formation requires de novo protein synthesis (Davis and Squire 1984; Igaz et al. 2002). Early seminal experiments on individual Aplysia californica (California sea hare, a mollusk) neurons demonstrated that facilitating signals through serotonin (5-HT, 5-hydroxytryptamin), which are needed for new synapse formation, led to acetylation at the promoter of the C/EBP transcription factor gene, while inhibitory input signals through FMRF amide caused deacetylation at the C/EBP promoter. If both signals were simultaneously applied, the inhibitory signal dominated and yielded histone deacetylation and long-term synaptic depression (Guan et al. 2002). Recent experimentation on Aplysia sensory neurons uncovered a mechanism for sequence-specific methylation in the serotonin-dependent synaptic facilitating signal pathway. The CREB2 protein is the major inhibitory constraint on memory formation in Aplysia. A Piwi/piRNA (piwi-interacting RNA) complex conveyed the local methylation of a conserved CpG-island at the CREB2 promoter (Rajasethupathy et al. 2012). However, it is still unclear whether piRNAs or other microRNAs are playing a general major role in mammalian sequence-specific methylation in the brain and interconnected memory formation (reviewed by Sweatt 2013).
After fear conditioning, i.e., learning, Dnmt gene expression was increased in the adult rat hippocampus leading to CpG-methylation at the memory suppressor gene PP1 (protein phosphatase 1) , while the promoter of the synaptic plasticity gene RELN (Reelin) was demethylated and expressed. These methylation changes were highly dynamic and back to normal within 24 hours. Accordingly, the gene for BDNF (brain-derived neurotrophic factor) responded with a quick change of exon-specific methylation, histone modifications and mRNA expression, which went back to normal within 24 hours. In addition, NMDA receptor blockade prevented these methylation changes and inhibited memory formation in rats (Lubin et al. 2008). Earlier yet, a hypomethylation at the Bdnf promoter region IV upon depolarization, the dissociation of the MeCP2-HDAC1-mSin3A repressive complex from the promoter, and a corresponding transcriptional increase of exon IV mRNA were observed in cortical mouse neurons cultured in vitro, 3 days after stimulation (Martinowich et al. 2003; reviewed by Grigorenko et al. 2016). Furthermore, Dnmt inhibition led to both a disturbed regulation of methylation and functional impairment of memory formation in the treated animals (Miller and Sweatt 2007). Transcriptional repression by methylation of another memory suppressor gene, PPP3CA (Calcineurin), in cortical neurons persisted up to 30 days after the learning experiments. Again, treatment with Dnmt inhibitors disrupted long-term fear memory (Miller et al. 2010).
Neurons are strongly expressing DNMT1, DNMT3A, and DNMT3B, and are highly enriched in 5hmC (reviewed by Delgado-Morales and Esteller 2017). The abundance of 5hmC in the brain suggested a functional role for 5hmC in the adult brain (Kriaucionis and Heintz 2009). It has been further suggested that 5hmC in the brain is not just an intermediate stage of the demethylation process operated by TET proteins, but also a key epigenetic mark for neurological disorders (Chen et al. 2014; Cheng et al. 2015). Whole genome analyses of 5hmC distribution with single-nucleotide resolution showed that 5hmC was even more enriched at constitutive exons of genes with synapse-related function, both in the human and the mouse brain (Khare et al. 2012). 5hmC in fetal brains was a marker for regulatory regions, which were poised for demethylation and activation in the adult brain. Demethylation was dependent on Tet2 (Lister et al. 2013). The regulation of methylation and demethylation relying on the interplay between DNMTs and TET-dioxygenases has been shown to play a major functional role in memory formation. Neuronal activity upon electro-convulsive treatment modified the neuronal DNA methylation landscape. About 1.4% of 219,991 CpG-sites of the adult mouse hippocampus showed a rapid active demethylation or de novo methylation. Regulated CpG-sites were significantly enriched in brain-specific genes related to neuronal plasticity (Guo et al. 2011; reviewed by Sweatt 2013; Kim-Ha and Kim 2016). Tet1 expression levels in the mouse hippocampus were decreased 3 to 4 h after neuronal activity, induced by either KCl depolarization, flurothyl-induced seizure, or fear conditioning. Following neuronal activation by flurothyl-induced seizure, 5mC and 5hmC levels underwent a decrease at 24 hours postseizure. Adeno-associated virus (AAV)-mediated overexpression of Tet1 in the dorsal hippocampus led to decreased 5mC and increased levels of 5hmC and of unmodified cytosines, 2 weeks after AAV injection, accompanied by an altered transcription of genes involved in memory formation and synaptic plasticity, and an impaired contextual fear memory formation at 24 hours after training (Kaas et al. 2013). Furthermore, Tet1-knockout mice suffer dysregulated hippocampal gene expression programs, impaired synaptic plasticity, and impaired memory extinction in learning experiments (Rudenko et al. 2013). Contrary to Tet1 expression, Tet3 expression was significantly increased shortly after neuronal stimulation and in learning experiments. Fear-extinction learning led to a Tet3-mediated accumulation of 5hmC, which was accompanied by changes in gene expression and rapid behavioral adaptation (Li et al. 2014). In mice, the cognitive decline associated with aging is accompanied by diminished expression of hippocampal Dnmt3a2 levels. After restoring Dnmt3a2 expression in old mice by injecting recombinant AAV carrying the Dnmt3a2 open reading frame, their cognitive ability was improved back to normal levels in fear conditioning experiments (Oliveira et al. 2012). Uhrf 2 is a bona-fide 5hmC reader protein. Uhrf 2-knockout mice were viable and exhibited no gross defects, but showed a reduced level of 5hmC and altered neuronal gene expression pattern and a partial impairment in spatial memory acquisition and retention. Therefore, Uhrf 2 plays a functional role in long-term spatial memory (Chen et al. 2017).
Experience-dependent epigenomic changes in the offspring remind us of Jean-Baptiste de Lamarck’s theory which is therefore reconsidered again. One aspect of Lamarck’s theory, frequently exemplified by the famous giraffe stretching its neck to reach leaves high up in a tree, postulates that exercise will modify and strengthen organs whose enhanced function, in this case a longer and stronger neck, will in turn be inherited by the offspring. Also Charles Darwin’s Pangenesis theory, developed to accommodate his predecessor Lamarck, and bashfully ignored for decades by evolutionary biologists, is being rescued from oblivion, due to the resemblance of Darwin’s “gemmules” with extracellular vesicles, exosomes, or small RNAs (reviewed by Liu and Li 2012; Chen et al. 2016c; Minarovits and Niller 2017). There are two types of transgenerationally transmitted epigenomic changes in the central nervous system, which are produced by experience, one of them culturally transmitted, but not biologically inherited, while the other one is apparently inherited via the germline (reviewed by Sweatt 2013; Bohacek and Mansuy 2015). Maternal nurturing behavior regarding newborn pups triggers durable methylation changes in the glucocorticoid receptor (GR) signaling pathway in the central nervous system which has profound and durable effects on stress response behavior into the adult life. “Stressed” mice transmit their disturbed stress response behavior to their offspring in a cultural way by lack of appropriate grooming behavior, thereby creating the next generation of stressed or depressed animals with the same methylation changes in the GR promoter within the hippocampus (Weaver et al. 2004; reviewed by Champagne and Curley 2009; Anacker et al. 2014). Interestingly however, learning effects can be passed on to the offspring of mice not only through early-life experience but also through durable epigenetic changes which are inherited via the germline (reviewed by Bohacek and Mansuy 2015). For example, in the case of parental traumatic olfactory exposure with acetophenone, the olfactory receptor gene Olfr151 became hypomethylated. Hypomethylation and enhanced neuroanatomical representation of the Olfr151 pathway persisted to the F2 generation. The transgenerational effects were inherited through the parental gametes (Dias and Ressler 2014). Early life traumatic experience led to an altered microRNA expression. The behavioral and epigenetic characteristics of the affected mice were transmitted through injection of sperm RNA into fertilized wild-type oocytes (Gapp et al. 2014).
An important mechanism of genetic neuronal plasticity interconnected with epigenetics is L1-retrotransposition which leads to genomic plasticity of high numbers of individual brain cells. This phenomenon resembles somewhat the somatic diversity of the adaptive immune system (reviewed by Sweatt 2013). Long interspersed nuclear elements-1 (LINE-1 or L1s) are abundant and comprise roughly 20% of the human genome. A massive somatic L1 insertional activity was observed in the normal adult human brain, but not other adult tissues, which leads to a physiological level of genetic mosaicism of individual neuronal genomes. Insertional activity is down-modulated by methyl-DNA binding protein MeCP2 which is mutated and therefore dysfunctional in patients with Rett syndrome who have a higher retrotranspositional activity in their brain cells (Muotri et al. 2010; reviewed by Erwin et al. 2014). Somatic L1-associated variants depending on either retrotransposition or deletion events occur in crucial neural genes and are hotspots for somatic copy number variation in the healthy human brain. An estimated 50% of all brain cells, including progenitor, glia, and neuronal cells, are affected by such somatic events (Erwin et al. 2016).
Besides and interconnected with CpG-methylation, a large number of posttranslational histone modifications and an abundance of specific microRNAs in the brain have been described in association with memory regulation (reviewed by Kim-Ha and Kim 2016; Kim and Kaang 2017). Furthermore, the regulation of the three-dimensional conformation and reconfiguration of neuronal chromatin in association with normal brain function and disease is just beginning to be explored and is probably most important (Juraeva et al. 2014; Flavahan et al. 2016; Mendizabal et al. 2016).
The rather new discipline of “neuroimaging epigenetics” tries to correlate epigenetic alterations of transcriptional promoters which are important for brain function, e.g., for neurotransmitter or brain receptor genes, with imaging techniques. The methylation status of promoters is correlated with visible changes in structural or functional magnetic resonance tomography (fMRT) or positron emission tomography (PET) images (reviewed by Nikolova and Hariri 2015). The exact mapping of the methylomes of different brain areas showed greater between-tissue variability than interindividual methylation variability. Nevertheless, some interindividual variability of brain methylation was reflected by the methylation status of peripheral DNA. Therefore, for practicability, neuroimaging epigenetics uses the peripheral methylation status, which is measured from blood or saliva as a proxy for brain methylation (Davies et al. 2012; Smith et al. 2015; reviewed by Nikolova and Hariri 2015).
4.9.3 Epigenetics of Psychiatric Disorders and Dementia
There is a broad overlap between epigenetic alterations associated with normal aging and dementia-associated epigenetic alterations. This fact is not surprising, because the most important single risk factor for dementia is aging (Heyn et al. 2012; Horvath et al. 2012). However, the distinction of the epigenomic dysregulation coming along with aging, mild cognitive impairment (MCI), or with Alzheimer dementia (AD) has been made. Paradoxically, in MCI patients, gene expression programs associated with synaptic plasticity and facilitation were connected with a lower score in mental performance testing (Berchtold et al. 2014). The overall methylation status seems to change bidirectionally with aging. While repetitive elements become hypomethylated in many tissues of mice and humans, many developmental genes are being hypermethylated with aging (Maegawa et al. 2010). Interestingly, the interindividual epigenomic profiles of the human brain become more similar in the later stages of life and also more similar between cerebral cortex and cerebellum. Furthermore, a loss of boundaries between transcriptional domains was observed, altogether resembling cell dedifferentiation (Oh et al. 2016; reviewed by Kim-Ha and Kim 2016).
Dementia is a disease trait common to various neurodegenerative disorders, like AD, Parkinson’s disease (PD), Lewy body dementia (LBD), and frontotemporal dementia (FTD , also named Pick’s disease). It has been postulated that the accumulation of epigenetic alterations and as a consequence alterations of gene expression during the lifetime might seed and sustain dementia-associated disorders which has been termed the “Latent Early-Life Associated Regulation” (LEARn) hypothesis. The LEARn model implies that incipient dementia may be diagnosed before the onset of symptoms and ideally, as a future perspective, be halted in time or even reverted (Lahiri et al. 2009; Maloney and Lahiri 2016). Interestingly, AD, PD, LBD, and also Down syndrome share a common set of genes which undergo aberrant methylation shifts (Sanchez-Mut et al. 2016). Future work on epigenomics of dementia must distinguish methylation patterns of different cell types and brain regions and also differentiate between 5mC, 5hmC, and all other forms of modified nucleotide bases. Because both 5mC and 5hmC protect cytosines from conversion to uracil, this cannot be accomplished through mere bisulfite sequencing. Interestingly, the epigenetic age of different brain regions differs. This applies strikingly to the cerebellum (Horvath et al. 2015). However, also the prefrontal cortex seems to age slower than other parts of the human brain and body (Klein and De Jager 2016). The current knowledge of the role of methylation and hydroxymethylation in the major dementia-related disorders AD , PD , LBD, and FTD has recently been comprehensively reviewed. Causality has still to be established for most of the alterations (Klein and De Jager 2016; Delgado-Morales and Esteller 2017). A genome-wide profiling for 5hmC in the AD brain identified 517 differentially hydroxymethylated regions (DhMRs) associated with neuritic plaques and 60 DhMRs associated with neurofibrillary tangles. Hydroxymethylated gene loci were significantly enriched for functioning in neurobiological processes (Zhao et al. 2017). Repeated gas-chromatographic measurements of overall 5mC and 5hmC levels from four different brain regions of different types and stages of dementia (AD, LBD, FTD) showed significant alterations especially in the preclinical stages of AD. This suggested that epigenetic alterations may play an early role in the progression of AD and other forms of dementia (Ellison et al. 2017). A transgenic mouse expressing the AD-associated protein p25 in dependence of doxycyclin regulation, exhibited memory deficits which were rescued by environmental enrichment. Memory restoration and synaptic function was accompanied by increased histone acetylation and hippocampal and cortical chromatin remodeling. Furthermore, HDAC inhibitor treatment of those mice induced neurite sprouting, increased synapse numbers and restored learning behavior and access to long-term memory as well (Fischer et al. 2007). In the inducible mouse model for AD, transcriptomic and epigenomic profiling of the hippocampus demonstrated a coordinated downregulation of synaptic plasticity and upregulation of immune response genes and regulatory regions over time. The situation was reflected in human brain tissue making the inducible AD mouse a useful model (Gjoneska et al. 2015).
Epigenetic mechanisms may be of particular importance for multifactorial diseases with low genetic penetrance (reviewed by Sweatt 2013). Epigenetic alterations are certainly associated with the development of psychosis, i.e., schizophrenia, bipolar disorder, and major depression which may also open up new avenues of future treatment (Oh et al. 2015; reviewed by Labrie et al. 2012). A recent genome-wide association study (GWAS) on schizophrenia (SCZ) described 108 distinct SCZ-associated loci, 83 of which had been newly detected (Schizophrenia Working Group of the Psychiatric Genomics Consortium 2014). Another genome-wide approach is the definition of methylation quantitative trait loci (meQTLs) by charting methylation levels at specific loci which are dependent on individual genetic differences, i.e., mostly SNPs, at other defined loci. Cis-acting and trans-acting meQTLs are distinguished and are enriched at regulatory sites. Several meQTL studies have been conducted on psychiatric diseases (reviewed by Hoffmann et al. 2016). A landmark study on the genomic landscapes of the brains of 335 nonaffected control individuals across all ages from the 14th gestational week fetal stage to 80 years of age, and of 191 patients with schizophrenia identified 2104 CpG-sites that differed between SCZ-patients and controls. SCZ-GWAS-risk loci were slightly enriched among those sites. The SCZ-related CpG-sites were strongly enriched for neurodevelopmental and differentiation genes. Furthermore, they strongly correlated with changes at the prenatal-to-postnatal transition, but not with later changes at the transition from adolescence to later adult life (Jaffe et al. 2016). Similarly, more than 16,000 meQTLs from 166 fetal brains were found to be fourfold enriched in SCZ-GWAS-risk loci (Hannon et al. 2016). While the effects of individual meQTLs are small, the location of meQTLs to genomic regions important for methylome reconfiguration during early brain development might hint at early vulnerable periods for the development of SCZ (reviewed by Hoffmann et al. 2016; Bale 2015). A study examining the epigenome of different postmortem brain regions of schizophrenic patients using the 450 K methylation array identified 139 differentially methylated CpG-sites and yielded a complex picture. Differences were located at known and novel candidate gene sequences, included gene bodies and CpG-islands, shores, and shelves, and were dependent on the brain area analyzed. Furthermore, the methylation states in brain tissue were largely not reflected by blood cell sampling underscoring the need to analyze the appropriate tissues for meaningful conclusions (Alelu-Paz et al. 2016). Valproic acid (VPA) has long been in use and still is for the treatment of epilepsy and bipolar disorder, depression and schizophrenia and more recently off-label for migraine-associated cluster-headaches. The mechanisms of action are assumed to be the blockade of neuronal ion channels, the increased synthesis of the inhibitory neurotransmitter GABA (gamma-aminobutyric acid), and HDAC inhibition. VPAs general long-term neuroprotective effect is ascribed to its epigenetic activity which prevents seizure-dependent aberrant neurogenesis in the adult hippocampus and the seizure-associated cognitive decline (Jessberger et al. 2007; reviewed by Monti et al. 2009). The implications of epigenetics in the workings of the central nervous system, especially of the above-mentioned “Lamarckian” experiments, of the wide-spread genetic mosaicism within the brain, and of the apparently very-early-in-life risk for schizophrenia are somewhat mind-boggling.
5 Research Needs
Research on epigenetics within many subdisciplines of biology has – figuratively speaking – taken off after the year 2000 (Sweatt 2013). Due to the development of powerful high-resolution and high-throughput methods, and due to the involvement of epigenetic regulation in all sorts of biological mechanisms (see Sect. 4), this great interest in epigenetics has not come by chance.
5.1 Philosophical Questions
From nineteenth-century philosophy, a notorious sentence by the then world-renowned physician and physiologist Jakob Moleschott was passed down to us: “Ohne Phosphor kein Gedanke” (no thinking without phosphorus). This sentence highlighting the philosophical direction of scientific materialism of the times was ridiculed by Arthur Schopenhauer as “Barbiergesellen-Philosophie” (barber-shop philosophy). Varying this phosphorus-sentence, today we might say “no thinking without methylation” or more precisely and less catchpenny-like “there may be no memory formation or thinking without the dynamic regulation of methylation states in the central nervous system.” Perhaps, Schopenhauer would not criticize us anymore for that latter phrase, as long as we would not imply any philosophical overstatement, because latter sentence represents a differentiated and experiment-based view on the functioning of the brain, as far as we know today. However, some of the fundamentals, especially the dichotomy between brain and mind, and the question of how both relate to each other, are still an unsolved riddle, just like 165 years ago. Beyond the fundamental mind-brain-problem, there is the not nearly as fundamental “nature versus nurture”-problem which has found the entrance to its solution (Sweatt 2009). Clearly, epigenetic regulation is the biological interface for the dynamic interplay between genes and environmental exposure or experience (Sweatt 2013).
Epigenetic mechanisms are at the center of cell differentiation and organ development, but – as we have seen above – are also involved in everyday function of the brain and, of course, also of other organs (Sweatt 2013). It will be interesting to continue to comprehensively map the epigenome of the developing brain and of the normally operating young and adult brain. This map should sort out the epigenetic marks and genes involved in cell fate determination and the epigenetic marks associated with organ function. The same should be tried for, let’s say, the liver and all other organs. An associated experimental question might be whether the epigenomes of specific single cells or structured cell groups in the liver are the same or very similar to the epigenomes of brain cells, neuronal or glial. To answer those questions, single cell methylomes of myriad cells would need to be established, and large amounts of bioinformatic power need to be tapped (Guo et al. 2013; Moroz and Kohn 2013; Smallwood et al. 2014; Farlik et al. 2015; Guo et al. 2015b; Angermueller et al. 2016; Gravina et al. 2016; Hou et al. 2016; Hu et al. 2016). After all, we might be in for a surprise about some large overlaps between the epigenetics of development and the epigenetics of normal function and, possibly, about some striking epigenetic similarities between different organs in their proper mode of functioning. Catchily speaking: why does the brain do the thinking and not the liver? Not a wee bit?
Furthermore, the act of thinking certainly relies in part on prior memory formation. The question then is whether quick thinking within seconds or decision-making within split-seconds may also rely on the (quick) dynamic regulation of CpG-methylation, of posttranslational modification of histones, or of three-dimensional chromatin structures in the brain. However, it may be hard to design experimental approaches to address such a question. For a further reading on neuroepigenetics and the associated major biological questions we refer to a most excellent perspective by J. David Sweatt (Sweatt 2013).
5.2 Sequence- or Locus-Specific Regulation of Methylation
A major question of epigenetic regulation is how CpG-methylation and other epigenetic marks may be targeted in a sequence-specific or locus-specific manner (see Sect. 3.2.5). While RNA-directed localization of epigenetic marks is an important mechanism in protozoa, plants, flies, and nematodes, mammalian cells in general seem to lack the respective biochemical machineries (Zhang and Zhu 2011; Yang et al. 2016a; reviewed by Joh et al. 2014). An important highly specific example of locus-specific methylation has been found by Rajasethupathy et al. who described a Piwi/piRNA (piwi-interacting RNA) complex which imparted the local methylation of a conserved CpG-island at the CREB2 promoter of Aplysia californica (Rajasethupathy et al. 2012). A mammalian example showing the requirement for components of the PIWI-interacting RNA (piRNA) pathway in local DNA methylation has been described for a differentially methylated region (DMR) of the paternally imprinted mouse Rasgrf1-locus (Watanabe et al. 2011). Nevertheless, it is still unclear whether piRNAs or other microRNAs are playing a general dominant role in mammalian sequence-specific methylation and what the exact mechanisms are (Stewart et al. 2016). However, since piRNAs have been described to play a role in reshaping the epigenome of cancer cells, and the epigenomes of mice were shaped by the injection of sperm RNAs or tRNA-derived small sperm RNAs (tsRNAs) into fertilized oozytes, a prominent role for small RNAs in the directed placement of epigenetic marks is to be expected in all likelihood (Gapp et al. 2014; Fu et al. 2015a; Chen et al. 2016b). Remarkably, the relative spatiotemporal distribution of small RNAs changes considerably during the development of primordial spermatogonia to mature spermatozoa (reviewed by Chen et al. 2016c). May the different amounts of piRNAs, miRNAs, and tsRNAs in different developmental stages of spermatogenesis reflect on a changing biochemical RNA-processing machinery available to the cell in the respective phase? Furthermore, the question may be asked whether sperm RNAs are under certain physiological circumstances able to influence the epigenomes of somatic cells of adult females.
5.3 Patho-Epigenetics of Disease and Perspectives for Medical Treatment
Not only has research addressing the basic mechanisms of methylation taken off in recent years, but also research on epigenetics in the context of diseases and their potential treatment. The description of epigenetic alterations and the accumulation of data on methylomes of diseases as diverse as cancer of many subtypes, autoimmune, infectious, genetic, imprinting, genetic, cardiovascular, psychiatric and neurologic disease, and as a subspecialty of malignant disease, the epigenetics of virus- or microbe-associated cancer has progressed quite far. This is exemplified by a large body of literature and a series of textbooks on the Patho-Epigenetics of Human Disease (Minarovits and Niller 2012; Tollefsbol 2012; Minarovits and Niller 2016). In some cases, especially in connection with infectious disease, e.g., HIV-infection, epigenetic alterations may imply the use of miRNAs and epigenetic drugs, like HDAC-inhibitors, as promising therapeutic approaches (Takacs et al. 2009; Ay et al. 2013; Szenthe et al. 2013; Minarovits and Niller 2017). A very interesting and promising gene-therapeutic approach to rewriting the epigenome may be the employment of CRISPR/Cas or other sequence-specific binding proteins to target localized methylation and demethylation enzymes at will (Xu et al. 2016; reviewed by Cano-Rodriguez and Rots 2016). Epigenetic approaches may be considered even for genetic syndromes, like trisomy 21 (Dekker et al. 2014).
This work was supported by the grant GINOP-2.3.2-15-2016-00011 to a consortium led by the University of Szeged, Szeged, Hungary (participants: the University of Debrecen, Debrecen, and the Biological Research Centre, Hungarian Academy of Sciences, Szeged, Hungary), project leader Janos Minarovits. The grant was funded by the European Regional Development Fund of the European Union and managed in the framework of Economic Development and Innovation Operational Programme by the Ministry of National Economy, National Research, Development and Innovation Office, Budapest, Hungary.
- Alelu-Paz R, Carmona FJ, Sanchez-Mut JV, Cariaga-Martinez A, Gonzalez-Corpas A, Ashour N, Orea MJ, Escanilla A, Monje A, Guerrero Marquez C, Saiz-Ruiz J, Esteller M, Ropero S (2016) Epigenetics in schizophrenia: a pilot study of global DNA methylation in different brain regions associated with higher cognitive functions. Front Psychol 7:1496PubMedPubMedCentralCrossRefGoogle Scholar
- Barlow DP, Bartolomei MS (2014) Genomic imprinting in mammals. Cold Spring Harb Perspect Biol 6:a018382Google Scholar
- Bartolomei MS, Ferguson-Smith AC (2011) Mammalian genomic imprinting. Cold Spring Harb Perspect Biol 3:a002592Google Scholar
- Berkyurek AC, Suetake I, Arita K, Takeshita K, Nakagawa A, Shirakawa M, Tajima S (2014) The DNA methyltransferase Dnmt1 directly interacts with the SET and RING finger-associated (SRA) domain of the multifunctional protein Uhrf1 to facilitate accession of the catalytic center to hemi-methylated DNA. J Biol Chem 289:379–386PubMedCrossRefGoogle Scholar
- Blow MJ, Clark TA, Daum CG, Deutschbauer AM, Fomenkov A, Fries R, Froula J, Kang DD, Malmstrom RR, Morgan RD, Posfai J, Singh K, Visel A, Wetmore K, Zhao Z, Rubin EM, Korlach J, Pennacchio LA, Roberts RJ (2016) The epigenomic landscape of prokaryotes. PLoS Genet 12:e1005854PubMedPubMedCentralCrossRefGoogle Scholar
- Chen, R., Zhang, Q., Duan, X., York, P., Chen, G.D., Yin, P., Zhu, H., Xu, M., Chen, P., Wu, Q., Li, D., Samarut, J., Xu, G., Zhang, P., Cao, X., Li, J., and Wong, J. (2017). The 5-hydroxymethylcytosine (5hmC) reader Uhrf2 is required for normal levels of 5hmC in mouse adult brain and spatial learning and memory. J Biol Chem.Google Scholar
- Cooper DN, Mort M, Stenson PD, Ball EV, Chuzhanova NA (2010) Methylation-mediated deamination of 5-methylcytosine appears to give rise to mutations causing human inherited disease in CpNpG trinucleotides, as well as in CpG dinucleotides. Hum Genomics 4:406–410PubMedPubMedCentralCrossRefGoogle Scholar
- Davies MN, Volta M, Pidsley R, Lunnon K, Dixit A, Lovestone S, Coarfa C, Harris RA, Milosavljevic A, Troakes C, Al-Sarraj S, Dobson R, Schalkwyk LC, Mill J (2012) Functional annotation of the human brain methylome identifies tissue-specific epigenetic variation across brain and blood. Genome Biol 13:R43PubMedPubMedCentralCrossRefGoogle Scholar
- Delgado-Morales R, Esteller M (2017) Opening up the DNA methylome of dementia. Mol Psychiatry 22:485–496. https://doi.org/10.1038/mp.2016.242
- Deplus R, Blanchon L, Rajavelu A, Boukaba A, Defrance M, Luciani J, Rothe F, Dedeurwaerder S, Denis H, Brinkman AB, Simmer F, Muller F, Bertin B, Berdasco M, Putmans P, Calonne E, Litchfield DW, De Launoit Y, Jurkowski TP, Stunnenberg HG, Bock C, Sotiriou C, Fraga MF, Esteller M, Jeltsch A, Fuks F (2014a) Regulation of DNA methylation patterns by CK2-mediated phosphorylation of Dnmt3a. Cell Rep 8:743–753PubMedCrossRefGoogle Scholar
- Deplus R, Delatte B, Schwinn MK, Defrance M, Mendez J, Murphy N, Dawson MA, Volkmar M, Putmans P, Calonne E, Shih AH, Levine RL, Bernard O, Mercher T, Solary E, Urh M, Daniels DL, Fuks F (2013) TET2 and TET3 regulate GlcNAcylation and H3K4 methylation through OGT and SET1/COMPASS. EMBO J 32:645–655PubMedPubMedCentralCrossRefGoogle Scholar
- Edwards CA, Rens W, Clarke O, Mungall AJ, Hore T, Graves JA, Dunham I, Ferguson-Smith AC, Ferguson-Smith MA (2007) The evolution of imprinting: chromosomal mapping of orthologues of mammalian imprinted domains in monotreme and marsupial mammals. BMC Evol Biol 7:157PubMedPubMedCentralCrossRefGoogle Scholar
- Erwin JA, Paquola AC, Singer T, Gallina I, Novotny M, Quayle C, Bedrosian TA, Alves FI, Butcher CR, Herdy JR, Sarkar A, Lasken RS, Muotri AR, Gage FH (2016) L1-associated genomic regions are deleted in somatic cells of the healthy human brain. Nat Neurosci 19:1583–1591PubMedPubMedCentralCrossRefGoogle Scholar
- Farlik M, Halbritter F, Muller F, Choudry FA, Ebert P, Klughammer J, Farrow S, Santoro A, Ciaurro V, Mathur A, Uppal R, Stunnenberg HG, Ouwehand WH, Laurenti E, Lengauer T, Frontini M, Bock C (2016) DNA methylation dynamics of human hematopoietic stem cell differentiation. Cell Stem Cell 19:808–822PubMedPubMedCentralCrossRefGoogle Scholar
- Grammatikakis I, Abdelmohsen K, Gorospe M (2017) Posttranslational control of HuR function. Wiley Interdiscip Rev RNA 8:e1372Google Scholar
- Guo F, Yan L, Guo H, Li L, Hu B, Zhao Y, Yong J, Hu Y, Wang X, Wei Y, Wang W, Li R, Yan J, Zhi X, Zhang Y, Jin H, Zhang W, Hou Y, Zhu P, Li J, Zhang L, Liu S, Ren Y, Zhu X, Wen L, Gao YQ, Tang F, Qiao J (2015b) The transcriptome and DNA methylome landscapes of human primordial germ cells. Cell 161:1437–1452PubMedCrossRefGoogle Scholar
- Heyn H, Li N, Ferreira HJ, Moran S, Pisano DG, Gomez A, Diez J, Sanchez-Mut JV, Setien F, Carmona FJ, Puca AA, Sayols S, Pujana MA, Serra-Musach J, Iglesias-Platas I, Formiga F, Fernandez AF, Fraga MF, Heath SC, Valencia A, Gut IG, Wang J, Esteller M (2012) Distinct DNA methylomes of newborns and centenarians. Proc Natl Acad Sci U S A 109:10522–10527PubMedPubMedCentralCrossRefGoogle Scholar
- Hoffmann A, Ziller M, Spengler D (2016) The future is the past: methylation QTLs in schizophrenia. Genes (Basel) 7:E104Google Scholar
- Jeong S, Liang G, Sharma S, Lin JC, Choi SH, Han H, Yoo CB, Egger G, Yang AS, Jones PA (2009) Selective anchoring of DNA methyltransferases 3A and 3B to nucleosomes containing methylated DNA. MolCell Biol 29:5366–5376Google Scholar
- Ji H, Ehrlich LI, Seita J, Murakami P, Doi A, Lindau P, Lee H, Aryee MJ, Irizarry RA, Kim K, Rossi DJ, Inlay MA, Serwold T, Karsunky H, Ho L, Daley GQ, Weissman IL, Feinberg AP (2010) Comprehensive methylome map of lineage commitment from haematopoietic progenitors. Nature 467:338–342PubMedPubMedCentralCrossRefGoogle Scholar
- Jones PA, Liang G (2009) Rethinking how DNA methylation patterns are maintained. NatRevGenet 10:805–811Google Scholar
- Juraeva D, Haenisch B, Zapatka M, Frank J, Investigators G, Group, P.-G.S.W, Witt SH, Muhleisen TW, Treutlein J, Strohmaier J, Meier S, Degenhardt F, Giegling I, Ripke S, Leber M, Lange C, Schulze TG, Mossner R, Nenadic I, Sauer H, Rujescu D, Maier W, Borglum A, Ophoff R, Cichon S, Nothen MM, Rietschel M, Mattheisen M, Brors B (2014) Integrated pathway-based approach identifies association between genomic regions at CTCF and CACNB2 and schizophrenia. PLoS Genet 10:e1004345PubMedPubMedCentralCrossRefGoogle Scholar
- Khare T, Pai S, Koncevicius K, Pal M, Kriukiene E, Liutkeviciute Z, Irimia M, Jia P, Ptak C, Xia M, Tice R, Tochigi M, Morera S, Nazarians A, Belsham D, Wong AH, Blencowe BJ, Wang SC, Kapranov P, Kustra R, Labrie V, Klimasauskas S, Petronis A (2012) 5-hmC in the brain is abundant in synaptic genes and shows differences at the exon-intron boundary. Nat Struct Mol Biol 19:1037–1043PubMedPubMedCentralCrossRefGoogle Scholar
- Kurukuti S, Tiwari VK, Tavoosidana G, Pugacheva E, Murrell A, Zhao Z, Lobanenkov V, Reik W, Ohlsson R (2006) CTCF binding at the H19 imprinting control region mediates maternally inherited higher-order chromatin conformation to restrict enhancer access to Igf2. Proc Natl Acad Sci U S A 103:10684–10689PubMedPubMedCentralCrossRefGoogle Scholar
- Lister R, Mukamel EA, Nery JR, Urich M, Puddifoot CA, Johnson ND, Lucero J, Huang Y, Dwork AJ, Schultz MD, Yu M, Tonti-Filippini J, Heyn H, Hu S, Wu JC, Rao A, Esteller M, He C, Haghighi FG, Sejnowski TJ, Behrens MM, Ecker JR (2013) Global epigenomic reconfiguration during mammalian brain development. Science 341:1237905PubMedPubMedCentralCrossRefGoogle Scholar
- Lister R, Pelizzola M, Dowen RH, Hawkins RD, Hon G, Tonti-Filippini J, Nery JR, Lee L, Ye Z, Ngo QM, Edsall L, Antosiewicz-Bourget J, Stewart R, Ruotti V, Millar AH, Thomson JA, Ren B, Ecker JR (2009) Human DNA methylomes at base resolution show widespread epigenomic differences. Nature 462:315–322PubMedPubMedCentralCrossRefGoogle Scholar
- Lopez De Silanes I, Gorospe M, Taniguchi H, Abdelmohsen K, Srikantan S, Alaminos M, Berdasco M, Urdinguio RG, Fraga MF, Jacinto FV, Esteller M (2009) The RNA-binding protein HuR regulates DNA methylation through stabilization of DNMT3b mRNA. Nucleic Acids Res 37:2658–2671PubMedPubMedCentralCrossRefGoogle Scholar
- Oh G, Wang SC, Pal M, Chen ZF, Khare T, Tochigi M, Ng C, Yang YA, Kwan A, Kaminsky ZA, Mill J, Gunasinghe C, Tackett JL, Gottesman I, Willemsen G, De Geus EJ, Vink JM, Slagboom PE, Wray NR, Heath AC, Montgomery GW, Turecki G, Martin NG, Boomsma DI, McGuffin P, Kustra R, Petronis A (2015) DNA modification study of major depressive disorder: beyond locus-by-locus comparisons. Biol Psychiatry 77:246–255PubMedCrossRefGoogle Scholar
- Pfaffeneder T, Spada F, Wagner M, Brandmayr C, Laube SK, Eisen D, Truss M, Steinbacher J, Hackner B, Kotljarova O, Schuermann D, Michalakis S, Kosmatchev O, Schiesser S, Steigenberger B, Raddaoui N, Kashiwazaki G, Muller U, Spruijt CG, Vermeulen M, Leonhardt H, Schar P, Muller M, Carell T (2014) Tet oxidizes thymine to 5-hydroxymethyluracil in mouse embryonic stem cell DNA. Nat Chem Biol 10:574–581PubMedCrossRefGoogle Scholar
- Ponts N, Fu L, Harris EY, Zhang J, Chung DW, Cervantes MC, Prudhomme J, Atanasova-Penichon V, Zehraoui E, Bunnik EM, Rodrigues EM, Lonardi S, Hicks GR, Wang Y, Le Roch KG (2013) Genome-wide mapping of DNA methylation in the human malaria parasite plasmodium falciparum. Cell Host Microbe 14:696–706PubMedPubMedCentralCrossRefGoogle Scholar
- Saitou M, Yamaji M (2012) Primordial germ cells in mice. Cold Spring Harb Perspect Biol 4:a008375Google Scholar
- Sakamoto Y, Watanabe S, Ichimura T, Kawasuji M, Koseki H, Baba H, Nakao M (2007) Overlapping roles of the methylated DNA-binding protein MBD1 and polycomb group proteins in transcriptional repression of HOXA genes and heterochromatin foci formation. J Biol Chem 282:16391–16400PubMedCrossRefGoogle Scholar
- Sanchez-Mut JV, Heyn H, Vidal E, Delgado-Morales R, Moran S, Sayols S, Sandoval J, Ferrer I, Esteller M, Graff J (2017) Whole genome grey and white matter DNA methylation profiles in dorsolateral prefrontal cortex. Synapse 71:e21959. https://doi.org/10.1002/syn.21959
- Sanchez-Mut JV, Heyn H, Vidal E, Moran S, Sayols S, Delgado-Morales R, Schultz MD, Ansoleaga B, Garcia-Esparcia P, Pons-Espinal M, De Lagran MM, Dopazo J, Rabano A, Avila J, Dierssen M, Lott I, Ferrer I, Ecker JR, Esteller M (2016) Human DNA methylomes of neurodegenerative diseases show common epigenomic patterns. Transl Psychiatry 6:e718PubMedPubMedCentralCrossRefGoogle Scholar
- Schneider K, Fuchs C, Dobay A, Rottach A, Qin W, Wolf P, Alvarez-Castro JM, Nalaskowski MM, Kremmer E, Schmid V, Leonhardt H, Schermelleh L (2013) Dissection of cell cycle-dependent dynamics of Dnmt1 by FRAP and diffusion-coupled modeling. Nucleic Acids Res 41:4860–4876PubMedPubMedCentralCrossRefGoogle Scholar
- Schultz MD, He Y, Whitaker JW, Hariharan M, Mukamel EA, Leung D, Rajagopal N, Nery JR, Urich MA, Chen H, Lin S, Lin Y, Jung I, Schmitt AD, Selvaraj S, Ren B, Sejnowski TJ, Wang W, Ecker JR (2015) Human body epigenome maps reveal noncanonical DNA methylation variation. Nature 523:212–216PubMedPubMedCentralCrossRefGoogle Scholar
- Serandour AA, Avner S, Percevault F, Demay F, Bizot M, Lucchetti-Miganeh C, Barloy-Hubler F, Brown M, Lupien M, Metivier R, Salbert G, Eeckhoute J (2011) Epigenetic switch involved in activation of pioneer factor FOXA1-dependent enhancers. Genome Res 21:555–565PubMedPubMedCentralCrossRefGoogle Scholar
- Smeets D, Markaki Y, Schmid VJ, Kraus F, Tattermusch A, Cerase A, Sterr M, Fiedler S, Demmerle J, Popken J, Leonhardt H, Brockdorff N, Cremer T, Schermelleh L, Cremer M (2014) Three-dimensional super-resolution microscopy of the inactive X chromosome territory reveals a collapse of its active nuclear compartment harboring distinct Xist RNA foci. Epigenetics Chromatin 7:8PubMedPubMedCentralCrossRefGoogle Scholar
- Spruijt CG, Gnerlich F, Smits AH, Pfaffeneder T, Jansen PW, Bauer C, Munzel M, Wagner M, Muller M, Khan F, Eberl HC, Mensinga A, Brinkman AB, Lephikov K, Muller U, Walter J, Boelens R, Van Ingen H, Leonhardt H, Carell T, Vermeulen M (2013) Dynamic readers for 5-(hydroxy)methylcytosine and its oxidized derivatives. Cell 152:1146–1159PubMedCrossRefGoogle Scholar
- Tsumura A, Hayakawa T, Kumaki Y, Takebayashi S, Sakaue M, Matsuoka C, Shimotohno K, Ishikawa F, Li E, Ueda HR, Nakayama J, Okano M (2006) Maintenance of self-renewal ability of mouse embryonic stem cells in the absence of DNA methyltransferases Dnmt1, Dnmt3a and Dnmt3b. Genes Cells 11:805–814PubMedCrossRefGoogle Scholar
- Velasco G, Hube F, Rollin J, Neuillet D, Philippe C, Bouzinba-Segard H, Galvani A, Viegas-Pequignot E, Francastel C (2010) Dnmt3b recruitment through E2F6 transcriptional repressor mediates germ-line gene silencing in murine somatic tissues. Proc Natl Acad Sci U S A 107:9281–9286PubMedPubMedCentralCrossRefGoogle Scholar
- Vlachogiannis G, Niederhuth CE, Tuna S, Stathopoulou A, Viiri K, De Rooij DG, Jenner RG, Schmitz RJ, Ooi SK (2015) The Dnmt3L ADD domain controls cytosine methylation establishment during spermatogenesis. Cell Rep 10:944–956. https://doi.org/10.1016/j.celrep.2015.01.021
- Walter M, Teissandier A, Perez-Palacios R, Bourc’his D (2016) An epigenetic switch ensures transposon repression upon dynamic loss of DNA methylation in embryonic stem cells. elife 5:e11418Google Scholar
- Walton EL, Francastel C, Velasco G (2014) Dnmt3b prefers germ line genes and centromeric regions: lessons from the ICF syndrome and cancer and implications for diseases. Biology (Basel) 3:578–605Google Scholar
- Watanabe T, Tomizawa S, Mitsuya K, Totoki Y, Yamamoto Y, Kuramochi-Miyagawa S, Iida N, Hoki Y, Murphy PJ, Toyoda A, Gotoh K, Hiura H, Arima T, Fujiyama A, Sado T, Shibata T, Nakano T, Lin H, Ichiyanagi K, Soloway PD, Sasaki H (2011) Role for piRNAs and noncoding RNA in de novo DNA methylation of the imprinted mouse Rasgrf1 locus. Science 332:848–852PubMedPubMedCentralCrossRefGoogle Scholar
- Will CL, Luhrmann R (2011) Spliceosome structure and function. Cold Spring Harb Perspect Biol 3:a003707Google Scholar
- Zabet NR, Catoni M, Prischi F, Paszkowski J (2017) Cytosine methylation at CpCpG sites triggers accumulation of non-CpG methylation in gene bodies. Nucleic Acids Res 45:3777–3784. https://doi.org/10.1093/nar/gkw1330
- Zhang Y, Jurkowska R, Soeroes S, Rajavelu A, Dhayalan A, Bock I, Rathert P, Brandt O, Reinhardt R, Fischle W, Jeltsch A (2010) Chromatin methylation activity of Dnmt3a and Dnmt3a/3L is guided by interaction of the ADD domain with the histone H3 tail. Nucleic Acids Res 38:4246–4253PubMedPubMedCentralCrossRefGoogle Scholar
- Zhao J, Zhu Y, Yang J, Li L, Wu H, De Jager PL, Jin P, Bennett DA (2017) A genome-wide profiling of brain DNA hydroxymethylation in Alzheimer’s disease. Alzheimers Dement 13:674–688. https://doi.org/10.1016/j.jalz.2016.10.004