1.6 DNA methylation at intergenic regions and repetitive elements
Szulwach, K. E., Li, X., Li, Y., Song, C. X., Wu, H., Dai, Q., et al. (2011). 5-hmC-mediated epigenetic dynamics during postnatal neurodevelopment and aging. Nat. Neurosci. 14, 16071616. doi:The detection of 5hmC is technically more challenging than that of 5mC due to the low abundance of 5hmC, and standard bisulfite sequencing does not distinguish between 5mC and 5hmC because both are resistant to bisulfite treatment [77]. Hydroxymethylated DNA immunoprecipitation (hMeDIP), which is modified from MeDIP, characterizes the relative abundance of 5hmC at specific loci or throughout the entire mammalian genome. hMeDIP involves immunoselection and immunoprecipitation using anti-5hmC antibodies and subsequent analysis by qPCR, microarray hybridization or next-generation sequencing [78]. TET-assisted bisulfite sequencing (TAB-seq) has been used to generate genome-wide 5hmC profiles at a single-base resolution in human and mouse ESCs [81]. In TAB-seq, 5hmC is protected from TET protein-mediated oxidation by the addition of glucose to 5hmC using β-glucosyltransferase (β-GT) to generate β-glucosyl-5-hydroxymethylcytosine (g5hmC). 5mC is oxidized by the Tet1 enzyme to 5caC. 5caC and unmethylated C are susceptible to bisulfite conversion and thus are sequenced as T, whereas 5hmC is sequenced as C. TAB-seq measures 5hmC directly, and information regarding 5mC can be obtained using the same analysis pipeline as BS-seq.
The recent development of affordable technology for DNA methylome analysis made the investigation of human brain samples practicable as well (Figure 2). Large cohorts of human prefrontal cortex samples revealed the dynamic changes occurring during development and aging of the brain (Hernandez et al. , 2011; Numata et al. , 2012; Jaffe et al. , 2016). These studies indicate that, although methylation differences are occurring in different scales, either at individual CpGs, at differentially methylated regions (DMRs) or at even larger domains, most changes are established during development and childhood, while methylomes are less plastic later in life. These findings likely point to differences in cellular composition rather than developmental dynamics and thus demonstrate the predicaments when heterogeneous cell populations are examined. Analysis of more homogeneous cell populations allow deeper insights, e. g. , revealing how in the developing and adult frontal cortex 5mC patterns distinguish cell types (Lister et al. , 2013) or that methylated CpH sites are almost absent from (NeuN negative) non-neuronal cells (Lister et al. , 2013). Instead, CpH methylation is generated de novo during neuronal maturation both in mouse and human cells (Lister et al.
Another mechanism of transcription regulation by methylation of single sites involves methylation dependent binding proteins (MDBPs). MeCP1 described by Meehan et al. needs at least 7 methylated CpGs for efficient DNA binding and therefore is less important for genes without CpG islands in the promoters [35]. MeCP2 (MDBP-2) binds to a single methylated CpG and can inhibit the transcription of the gene [36]. In addition to this two proteins recently Hendrich and Bird described a family of MDBP which have high homology with MeCP2. All of them contain DNA binding domain as well as transcription inhibitory domain [37]. Although the functional significance of HOXD4 in atherosclerosis remains unclear, a previous study identified differential expression of this gene in the samples of human aorta with varying degrees of atherosclerosis [24]. On the other hand, the region studied for the DNA methylation encompasses four CpG-dinucleotides located within MIR10B gene sequence. MiR-10b is a microRNA associated with metastasis and/or invasiveness of various cancer types [25].
Recently, it was shown that HoxA5 is regulated by flow in a DNA methylation-dependent manner and that Hoxa5 regulates endothelial inflammation [30]. Induction of disturbed blood flow by partial carotid ligation surgery in a murine model resulted in hypermethylation within the promoter of HoxA5 and downregulation of gene expression in arterial endothelium [30]. Our study shows that HOXA5 is hypermethylated in great saphenous veins as compared with the internal mammary arteries that may contribute to aberrant remodeling of veins after coronary artery bypass grafting surgery. In 1948, Rollin Hotchkiss used paper chromatography to separate and quantify the components of DNA. To his surprise he detected not only the four nucleo-bases thymine, adenine, cytosine, and guanine, but also a minor constituent designated epicytosine [with] a migration rate somewhat greater than that of cytosine (Hotchkiss, 1948). As Hotchkiss had already suspected, epicytosine turned out to be a methylated form of cytosine.
Therefore activation of oncogenes was proposed as a possible role of decrease in DNA methylation in carcinogenesis. However no significant data was collected to test this hypothesis. Computational pipeline for genome-wide bisulfite sequencing data analysis. Reads from bisulfite sequencing are first aligned to the reference genome. The alignment data may be visualized in different tracks for comparison. After methylation calling, the bulk methylation level and genome-wide methylation level are calculated and plotted, and DMRs are determined. To perform an integrative analysis, DNA methylation data are coupled with gene expression, e. g. , differentially expressed genes (DEGs), to delineate the regulatory role of DNA methylation Previous studies of vascular tissue samples revealed alterations in methylation of ALOX15, ESR1, ESR2, MCT3 and TFPI2 genes in patients with atherosclerosis [59]. These genes were covered by Infinium Human Methylation27 BeadChip, but they were not significantly differentially methylated in our study.
Rather, a common theme is cell death, due to either the failure to fully differentiate and/or to repress repetitive elements (Ramesh et al. , 2016), quite similar to what had been described in the postnatal retina (Rhee et al. , 2012). Thus, the hypothesis that DNA methylation represses alternative fates has to be questioned, while the role in differentiation receives more support. Indeed, in the few studies of mouse mutants that examined the transcriptome, many aspects of immature cells, such as cell proliferation, fail to be repressed at later stages along with the failure to up-regulate genes involved in synaptic maturation. According to Wu et al. , Prc2 mediated mechanisms could be involved in these processes as they showed that Dnmt3a-mediated DNA methylation adjacent to H3K4me3 high promoters interferes with Prc2 binding and H3K27me3 and thereby mediates up-regulation of neuronal progenitor genes (Wu et al. , 2010). In addition or alternatively, Tet-mediated roles could be involved as described above from the Uhrf1 study (Ramesh et al. , 2016).
Rather DNA methylation appears generally required for repression of ERVs, even though with striking cell type specificity. A further general concept that emerged is its role in orchestrating cell differentiation, but within a given lineage (neurons, oligodendrocyte progenitors, Table 2). The involvement of splicing as effector of changes in DNA methylation is an exciting new angle to pursue with more precise epigenetic engineering tools. Distinguishing essential from specific, and causal from secondary marks will be essential for neuro-epigenetics. New approaches promise to answer long outstanding questions and will likely facilitate the discovery that DNA modifications might have new unexpected roles in the brain. Development continues to some extent also in the adult brain, both in adult neurogenesis but also in the ongoing synaptic plasticity that constantly re-forms new synaptic connections. DNA modifications have also accredited functional roles in these processes including information storage and providing (in adult NSC niches) new mature neurons (Ninkovic and Götz, 2013). In the late 60s, an open debate was started, how neurons would be able store memory information for life, while the stability of the molecular building blocks of these cells is many orders of magnitudes shorter.
Hereafter, we will discuss the evidence for the above criteria in establishing the previous model, namely a role of DNA-methylation in repressing alternative fates. Subsequently we will proceed to discuss experimental evidence testing this model. Data from pluripotent stem cell differentiation and mouse models in vivo (section Mouse Models) demonstrate that no fate switch to an alternative fate occurs even when most or all of methylation marks have gone (see section Mouse Models). Conversely, phenotypes appear late in brain development, often at postnatal stages, indicating rather that maturation processes are affected (Tables 1, 2). Figure 4. Suggested influences of DNA methylation on neurogenesis. (A) Temporal progression of DNA methylomes might influence the potential of neural stem and progenitor cells. (B) Cell specific methylomes, here 5mC for simplification, might be responsible for neural cell identities. They could not only influence lineage choices, but might also simultaneously block alternative fates. (C) Through controlling activity of transposon derived sequences, DNA methylation has been implicated in contributing to neuronal diversity. (D) Global alterations of DNA modifications often result in cell death during differentiation.
However, much remains to be understood about the repressive function of DNA-methylation in regard to differentiation and neuronal maturation. This is particularly evident from the poor correlation between changes in DNA-methylation and transcription. Further follow-up studies on the transcriptional changes that are crucial for the phenotypes aiming to correlate these to epigenetic mechanisms will hold the key to better mechanistic understanding of the mouse mutant phenotypes. In our study several vascular tissue samples were stained with hematoxylin and eosin or immunostained with antibodies against smooth muscle-specific α-actin and CD68. Smooth muscle cells were predominated in all analyzed vascular tissues. CAP, IMA and GSV showed a variable degree of infiltration with macrophages. Specimens taken from the CAP contained a large accumulation of macrophages to compare with IMA and GSV. For DNA methylation microarray profiling, we selected match-paired specimens (CAP, IMA, GSV) from six patients. For confirmation by pyrosequencing and replication, match-paired tissue samples from all 21 individuals were used.
Thus, the first description of an epigenomic mark occurred only few years after DNA has been identified as the carrier of genetic information (Avery et al. , 1944) and years before its structure has been resolved (Watson and Crick, 1953). Coincidently to these biochemical insights, first conceptual ideas arose attempting, to explain, how a single set of genetic information could give rise to the pleiotropy of cellular phenotypes (Waddington, 1957). From these early days on, epigenomic marks and epigenetic phenotypes have been closely intertwined, which lead to great discoveries but also to misconceptions, such as the perception, these two terms, epigenetic (heritable traits that have their origin not in the DNA sequence) and epigenomic (reversible marks, modifications and features of DNA-implicated in epigenetic traits) would be equivalent. While much remains to be done, experimental tests propose already a revision of the concept that DNA methylation would repress alternative fates (Tables 1, 2).
These marks are then thought to be lost passively or removed by the thymine DNA glycosylase (TDG), a forerunner of the base excision repair (BER) (Yu et al. , 2012; Zhang et al. , 2012). Also other proteins and enzymes involved in DNA repair (e. g. , GADD45/AID/APOBEC) have frequently been implicated in active DNA de-methylation (Rai et al. , 2008; Bhutani et al. , 2010, 2011), although their contributions to global methylomic changes are still being discussed (Nabel et al. , 2012). Taken together, the above mentioned experimental tests on the role of DNA methylation in cerebral cortex development do not lend much support to the model that it serves to repress alternative fates (Table 2). Besides GFAP up-regulation (Kim et al. , 2016) there is not much evidence for aberrant glial fate instruction, including in genome-wide expression analysis (Hutnick et al. , 2009; Ramesh et al. , 2016), and no ectopic fate choices have been observed in any of the above mutants.
When comparing methylomes with weak differences, extending the testing scale from one C to a cluster of neighboring Cs can reduce the number of hypothesis tests to improve the statistical power [91] (e. g. , BiSeq takes spatial correlation into account in DMR prediction [115]). Weak DNA methylation differences can be better measured by estimating the standard deviation from biological replicates to obtain more robust P values [91]. Later loss of the expression, associated with hypermethylation of promoter CpG island was shown for retinoblastoma (Rb) gene in 10 of patients with sporadic form of retinoblastoma [62]. Several publications have documented de novo methylation of the CpG island for the cyclin dependent kinase inhibitor p16 in both cancer cell lines and primary tumours [63,64]. The aberrant hypermethylation correlated with the lack of p16 expression in these cells. Treatment of the cell lines with 5-aza-2'-deoxycytidine resulted in the demethylation of p16 promoter and reactivation of p16 expression [65]. Due to the much higher complexity of eukaryotic genome in comparison to prokaryotic one it is logical to presume some additional roles of methylated cytosine as a "fifth base". Indeed, there is number of experimental evidences for the involvement of cytosine methylation in the functional reorganisation of eukaryotic genome.
Genetically modified mouse models of all known writers of the DNA methylation machinery have been generated to functionally test the global relevance of this epigenomic modification. The full knockout for the de novo methyltransferase Dnmt3a for example appears overall normal at birth (Li et al. , 1992; Okano et al. , 1999), but mice die 4 weeks after birth due to multiple developmental defects (Okano et al. , 1999). It has been suggested that this is in part due to a disturbed neurogenesis in the SEZ of the forebrain and the hippocampal dentate gyrus, as NSCs loose DNA methylation on the gene bodies of neuronal genes and fail to activate those during differentiation (Wu et al. , 2010). While defects in adult neurogenesis are unlikely to cause death of the entire organism, these data did reveal a key role of DNA-methylation in NSC differentiation with a clear decrease in postnatal neurogenesis. The authors also suggest that this was due to an increase in gliogenesis and hence a fate switch, but this is less clear as postnatal and adult NSCs also express astroglial markers, such as GFAP and some level of S100b (Beckervordersandforth et al.
DNA methylation has been implied in regulation of gene transcription already in the late 60s (Harrisson, 1971; Scarano, 1971; Holliday and Pugh, 1975; Riggs, 1975) and often still is; although it has become clear that it likely plays a much less general role than believed originally. But why has DNA methylation become the one epigenomic mark most frequently connected to epigenetic gene silencing in the first place? There are plenty of answers to this question, which we are neither able to discuss fairly, nor to list comprehensively; we think however, that most of the concepts and experimental evidence gained during the decades can be grouped into four types, which we will address below. First, the biochemical features of DNA methylation, its life cycle and inheritance make it a prime candidate for a developmental epigenetic mark; second, global correlations between the presence of DNA methylation and the activity state of DNA in the nucleus do occur; third, DNA methylation is necessary for normal animal development and finally, on some individual model loci a functional effect of DNA methylation on restricting transcription is clearly evident.
, 2007)] indicating that neuronal maturation or specific neuronal functions in particular neuronal plasticity might indeed be dependent on normal availability of the DNA modification machinery. Although atherosclerosis is caused by the interaction of multiple genetic and environmental factors, these explain only a portion of the total disease risk. Epigenetic mechanisms that underly this pathology have become a promising area of research [2, 3]. Compared to genetic factors, epigenetic variation is much more suitable to explain the progressive and age-related nature of atherosclerosis characterized by sex and tissue specificity. Aberrant epigenetic patterns can be acquired during developmental stages under environmental influence. Today we know that many more DNA modifications exist. Additionally to the mark usually meant by the phrase DNA methylation [the methylation of cytosine at position C5 (C5-methylcytosine, 5mC)], the same base can also occur methylated on other positions [e. g. , N3-methylcytosine (3mC)]. 3mC is, however, thought to represent rather a product of DNA damage than a bona fide information carrier (Sadakierska-Chudy et al. , 2015).
observed the abnormalities in chromosomal division during cell replication after decrease of overall methylation induced by 5-aza-2-deoxycytidine treatment [55]. This suggests that demethylation may influence the structural integrity of chromosomes leading to cell transformation [56]. However more studies are required to establish the consequence of DNA demethylation in neoplastic cells. For targeted bisulfite sequencing, the SeqCap Epi System from Roche enables the enrichment of a small fraction of the genome containing regions of interest after bisulfite conversion [68]. In addition, the SeqCap Epi CpGiant Enrichment Kit allows the interrogation of more than 5. 5 million CpGs in the human genome with a starting DNA input of 1µg. Roche also provides customization of probe pools according to the type of organism and regions of interest. The SureSelectXT Methyl-Seq Target Enrichment Kit from Agilent Technologies involves the hybridization and enrichment of sequencing libraries with oligonucleotide baits before bisulfite conversion [69]. This platform supports the enrichment of an 84-Mb target covering 3.
Developed by Pacific Biosystems, SMRT allows the direct detection of base modifications by monitoring the activity of DNA polymerase during the incorporation of different fluorescently labeled nucleotides into complementary DNA strands [85, 86]. The direct detection of various base modifications involves the measurement of the kinetics variation in the time between base incorporations. This technology has the following advantages over second-generation sequencing: (1) minimal chemical modification during library preparation; (2) the requirement for DNA amplification is eliminated; (3) reduced requirement for input DNA; (4) the ability to generate longer reads (average read length of 3000bp); and (5) the ability to detect different types of epigenetic modifications [86, 87]. SMRT has been used in the identification of 6mA in C. elegans, and the recently developed SMRT of chromatin immunoprecipitation enriched DNA (SMRT-ChIP) has resulted in the identification of 6mAand associated demethylase ALKBH1 in mouse ESCs [11, 88]. Some of the highly differentially methylated genes are also very interesting.
The identification of DMRs relies on both computational power for genome-wide screening and statistical testing. In Table2, we included tools for implementing statistical methods in DMR screening [110, 112115]. Generally, the DMR detection algorithm adopts a sliding window across the genome to survey candidate DMRs, and the most common approach is to perform Fishers exact test CpG-wise. To detect DMR, as the coverage of each sample may be different, only sites covered by all samples are comparable. To enable the comparison, the comparing statistics such as methylation difference, T-score from T test or P value is needed in the testing. In the BSmooth software, a beta-binomial is assumed to be the suitable model for replicated bisulfite sequencing data. The observation is assumed to be binomially distributed, whereas the methylated proportion at a particular site can vary across samples. The differences at an individual site could be small but may expand and persist across a region, which is a candidate DMR. Therefore, DMRs are determined with greater statistical power and are more informative.
First, in our study S100A10 gene was hypomethylated in right coronary artery in the area of advanced atherosclerotic plaques compared to atherosclerotic-resistant internal mammary arteries. S100A10 is highly expressed in endothelial cells, macrophages and foam cells of complicated carotid plaque segments [31]. This Ca2+-binding protein involved in the plasmin/plasminogen system regulating proteolytic activity and degradation of extracellular matrix, angiogenesis and macrophage invasion. Second, we observed changes in the methylation level of GLRX gene in both comparisons CAP vs. IMA and GSV vs. IMA. Expression of GLRX gene is enhanced in endothelial cells, smooth muscle cells and macrophages of human nonatherosclerotic and atherosclerotic coronary arteries [32]. Glutaredoxin-1 is a cytosolic enzyme that regulates diverse cellular functions. As speculated by Okuda et al. [32] glutaredoxin might be involved in the pathogenesis of atherosclerotic coronary heart disease via its antioxidant effect and/or its role as a signaling molecule. The TET family of dioxygenases catalyze the oxidation of 5mC to 5hmC. The detection of 5hmC gained much attention recently after this C modification was identified as an epigenetic mark in mammals (mouse brain and ESCs), and 5hmC has been reported to be an intermediate in DNA demethylation [75, 76].
7 million CpG sites with a DNA input as low as 1µg. The general workflow for the bioinformatics analysis of DNA methylation data includes data processing, the quantification of DNA methylation levels, general profiling, the identification of DMRs and the visualization of the methylome [91]. Array-based data, such as that from Illuminas HM450K, are fluorescence intensities that quantify the relative abundance of methylated and unmethylated loci. The data from other non-bisulfite-conversion methods, such as MRE-seq and MeDIP-seq, are usually analyzed by comparing the relative abundance of fragments. Bisulfite-converted data, such as those from WGBS and RRBS, involve methylation calling at individual Cs, and statistical testing is required to assess differential methylation. In this section, we focus on the bioinformatics analyses of bisulfite-converted data, in particular WGBS and RRBS (see Fig. 3 for a general bioinformatics pipeline).
But not only cytosine can be targeted by methylation, also adenine [N6-methyladenine, (6mA); (Wu et al. , 2016)]. On top, new DNA modifications on the position C5 have been discovered recently, which are generated by DNA demethylation pathways (Booth et al. , 2015, Figure 1). The first of these 5mC oxidation products to be reported was 5hmC (C5-hydroxymethylcytosine) (Kriaucionis and Heintz, 2009; Tahiliani et al. , 2009); 5fC (C5-formylcytosine), and 5caC (C5-carboxylcytosine) followed later. Although 5hmC has been described to occur in animal tissues (e. g. , mouse brains) already in the 70s (Penn et al. , 1972), its relevance was not recognized as it was widely interpreted as a product of DNA damage (Privat and Sowers, 1996). Today we know that 5hmC and 5caC are not necessarily transient marks occurring solely in a sequence of chemical reactions; instead they can appear quite stable at least under some circumstances (Bachman et al. , 2014, 2015). Schmidt et al.
The regions of the genome with a high number of methylated cytosine are usually transcriptionally inactive. The absence of DNA methylation is a prerequisite for transcriptionally active regions. Since DNA methylation is reversible and does not directly depend on the sequence context it was described as an epigenetic mechanism of gene regulation [8,9]. Schmitz et al. performed a large-scale WGBS analysis in which DMRs from many Arabidopsis methylomes were detected [116]. They used the R package methylPipe to scan the genome with 100-bp windows [114], and the methylation level of the sites within a window was compared across all samples using a KruskalWallis test. The P values were then adjusted for multiple testing using the BenjaminiHochburg method, and only DMRs with an adjusted P value less than were selected. In addition to the adjusted P value, a second criterion is used to ensure the differences, and the DMR has to exhibit an eightfold methylation difference between the two groups.
For example in humans this dinucleotide is present only 5 to 10 of its predicted frequency. In 70 to 80 these CpG dinucleotides are methylated. These methylated regions are typical of the bulk chromatin that constitutes most nontranscribed DNA (for review 26)[27]. These two types of DNA methylation patterns determined by either low content of CpG sequence or CpG islands represent two types of regulatory regions. Genes which contain CpG island in their promoter are usually "housekeeping" genes, which have a broad tissue pattern of expression. Many relatively tissue specific genes are also regulated by CpG island methylation [26]. It is important to note that nonmethylated CpG island within the promoter region is not always associated with actively transcribed gene. However the lack of methylation of the CpG island within the promoter region of the gene is required for transcription of the gene. This modulatory role of methylation is reflected by the fact that chemically induced demethylation of CpG islands associated with inactivated genes leads to their partial reactivation [29].
- Criteria for Publication
- Editorial and Peer Review Process
- Editor Center
- Resources for Editors
- Guidelines for Reviewers
- Accepted Manuscripts
- Comments
Although we mention several biological processes and all known DNA modifications in this review, we will focus on the role C5-methylcytosine plays in neurogenesis and neuronal maturation. This review provides an overview of the current techniques for the assessment of genome-wide DNA methylation and the identification of DMRs. The commonly used techniques are primarily based on restriction enzyme digestion, affinity enrichment and bisulfite treatment, coupled with either microarray or sequencing technologies. Because each technique has its own advantages and disadvantages, we summarize in Table1 a comprehensive evaluation of each technique. In Fig. 2, we provide an overview of these experimental pipelines and their required DNA input amounts. The selection of a technique strongly depends on the research questions, cost, amount of input DNA and the expected degree of methylation changes [118]. In Table1, readers can also learn from the biological examples in which the profiling techniques were used to determine the experiments that best fit their research topic. For example, for mammalian studies with large-scale samples, one should consider a targeted approach, such as MeDIP or RRBS rather than WGBS, which would allow multiple sample comparisons with limited cost and provide sufficient information from CpG-rich regions.
Detection of differentially methylated loci and regions
Join for free and get personalized recommendations, updates and offers.
, 2014), its occurrence is highly linked to the neuronal expression of Dnmt3a, as knockdown of this de novo methyltransferase abolishes CpH methylation (but not CpG methylation, which is mainly dependent on Dnmt1) (Guo et al. , 2014). However, maybe the most surprising results stem from genetically modified, overexpression or knockdown mouse models of writer, reader, and eraser proteins of DNA modifications, resulting either in phenotypes affecting memory formation or consolidation [Tet1 (Kaas et al. , 2013; Rudenko et al. , 2013; Zhang et al. , 2013), Dnmt1, Dnmt3a, and Dnmt3a2 (Miller and Sweatt, 2007; Feng et al. , 2010; Oliveira et al. , 2012, 2016)], emotional or maternal behavior [Dnmt3a (LaPlant et al. , 2010), Mbd2 (Hendrich et al. , 2001)], LTP [Mbd1 (Zhao et al. , 2003)], or adult neurogenesis in the hippocampus [GADD45b (Ma et al. , 2009), Tet1 (Zhang et al. , 2013), Mbd1 (Zhao et al. , 2003), Mecp2 (Smrt et al.
Moreover, these are frequently overlapping with regulatory sequences of important cell fate factors (like Pax6) and are dependent on transcription factor activity in some tested cases, as DNA binding (of the neural repressor REST for example) is necessary and sufficient to evade high DNA methylation levels on its binding sites (Stadler et al. , 2011). In the current study we performed a comparative analysis of DNA methylation patterns in right coronary arteries in the area of advanced atherosclerotic plaques (CAP), internal mammary arteries (IMA), and great saphenous veins (GSV) derived from the same individuals with coronary artery atherosclerosis, using Illumina HumanMethylation27 BeadChip microarrays. We hypothesize that DNA methylation differences are the key to distinguish CAP from the IMA with respect to their liability to the development of atherosclerosis. A comparison of DNA methylation patterns between GSV and IMA can provide an explanation for the difference in graft patency after coronary artery bypass grafting surgery. Restriction enzyme-based methods take advantage of the differential digestion properties of isoschizomers and neoschizomers. A pair of isoschizomers recognizes the same sequence and has the same point of cleavage but exhibit different sensitivities to the DNA methylation state.
If the study aims to investigate the first methylome of an organism, then WGBS with deep sequencing would be a more suitable method to obtain detailed information in coding regions and intergenic regions. The input DNA amount should also be considered when rare cell types or tissues are studied. To reveal the methylation state of undifferentiated stem cells without heterogeneity, single-cell approaches would be the best choice. The sequencing depth is a key parameter in DMR discovery; the greater the depth, the more power to discover DMRs. However, for studies with a large sample size such as disease-centered research studies, the distribution of limited resources should be considered, e. g. , sequencing a few samples deeply or more samples less deeply. A balance may be reached by considering the profiling technique coupling with the data analysis that would provide precise and accurate DMR prediction with low coverage requirements. Hoxa2 and Hoxa5 have been cloned from an adult rat vascular smooth muscle cDNA library, although the functions of these proteins were not determined in this study [28]. Hoxa5 is normally expressed in quiescent vessels, blocked angiogenesis and increased vascular stability [29].
Additionally some other CpG-rich regions on 11p, which is known to contain multiple potential tumour suppresser genes [60], were simultaneously hypermethylated [59]. It was suggested that 11p region is a hot spot for CpG island methylation in neoplasia and that this DNA methylation change could be an important potential mechanism for inactivation of tumour suppresser genes [59]. First indications about the cell type specific distribution and dynamics of DNA methylation during neurogenesis (and its relation to other epigenomic marks and transcription factor binding) have been gained from differentiation of embryonic stem cells or neural progenitor cells (Meissner et al. , 2008; Stadler et al. , 2011). Profiling of pluripotent and neural stem cells revealed for example, that regions with low methylation show the most dynamic DNA methylation changes during development.
The extensive research on methylation was conducted on bacteria. In this lower forms, both adenine and cytosine can be methylated, and this modification is involved in DNA replication and arrangement. A series of DNA methyltransferases (DNA-MTases) which can catalyse cytosine methylation in different sequence context were identified [2]. The main function of DNA methylation in bacteria is to provide a mechanism, which protects the cell from the effect of foreign DNA introduction. Restriction endonucleases discriminate between endogenous and foreign DNA by its methylation pattern. Introduced DNA which is not protected by methylation is then eliminated by cleavage [2]. Baylin et al described that a CpG island in the promoter region of the calcitonin gene at chromosome 11p, which was unmethylated in all normal tissues tested, was densely methylated in human solid tumours [57], leukemias [58] and cells transformed with various viruses [58,59].
1.6 DNA methylation at intergenic regions and repetitive elements
It was shown that overexpression of miR-10b induces human microvascular endothelial cell migration and angiogenesis via down-regulation of homeobox D10 (HOXD10) [26]. This microRNA was also upregulated in advanced carotid plaques as compared with internal mammary artery [27]. Further studies will be required to elucidate the functional significance of methylation changes within MIR10B gene sequence in atherosclerosis. One of the most striking system used for the analysis of the role of DNA methylation in carcinogenesis was an APCMIN mouse model analysed by Laird et al. [71]. APCMIN mouse carries a germline mutation in APC gene and develop hundreds of intestinal polyps. The inhibition of methyltransferase with 5-aza-2'-deoxycytidine in these mouse resulted in the reduction of polyp number from 113 to only 2. Although the mechanism by which demethylation reduces the polyp formation remains unclear, these results provide an important evidence for the involvement of methylation in carcinogenesis [71].
, 2010), making it impossible to decide whether the increased cell population are NSCs or astrocytes. Conditional deletions of Dnmt3a in the developing nervous system (Nes1-Cre) have been reported to have a shortened lifespan as well, which has been attributed to postnatal motor neuron loss (Nguyen et al. , 2007). Mouse embryos lacking Dnmt3b exhibit multiple developmental abnormalities, including rostral neural tube defects, and are not delivered to term (Okano et al. , 1999). Thus, normal neural development is (at least partially) dependent on the presence of both de novo methyltransferases. Although full knockouts of Tet1 have been reported to be born overall normal (Dawlaty et al. , 2011), recently new mutant alleles have been generated that are lethal during embryogenesis when outbred, at least partially due to deformities in forebrain development associated with incomplete closure of the anterior neuropore (Khoueiry et al. , 2017).
, 2013; Guo et al. , 2014) and parallels synaptogenesis and neuronal diversity (Lister et al. , 2013; Mo et al. , 2015). Remarkably, studies also indicate that methylation marks occurring in regulatory regions are more indicative of transcriptional repression when falling on CpH rather than on CpG sites (Mo et al. , 2015). The first characterization of 5hmC dynamics was linked to the development of reliable methods mapping this mark epigenome-wide (Figure 2). Using hMeDIP for example has shown that in contrast to 5mC, the cellular amount of 5hmC is significantly increasing when neural stem and progenitor cells are differentiating to neurons (Hahn et al. , 2013). A similar developmental dynamic has also been detected during ex vivo analysis of mouse cortices and human brain samples (Szulwach et al. , 2011; Lister et al. , 2013; Wen et al. , 2014; Vogel Ciernia and LaSalle, 2016). Interestingly, newly acquired 5hmC often associates with regulatory elements of neuronal genes (Szulwach et al. , 2011; Wang et al.
, 2012) and are solely detectable at CpG sites (Lister et al. , 2013). Bisulfite sequencing of DNA derived from adult mouse dentate granule neurons before and after synchronous neuronal activation in vivo, revealed that some DNA methylation marks do not behave as stable as commonly expected and rather suggested that around 1 of analyzed 5mC sites fulfill the criteria of activity induced de-methylation (Guo et al. , 2011) with yet elusive function. Taken together profiling of DNA methylation in mammalian brain cells from both in vitro and ex vivo models indicate that diverse cell populations differ significantly in their methylome and that these changes can swiftly emerge at meaningful sites, indicating that they could contribute to shape cellular functions. In the following, we will give a short overview about the distribution of DNA modifications and discuss how they are established. We will then present the suggested roles for DNA modifications in gene expression control and review how those have been implicated into regulating lineage decisions during brain development. We finish with re-evaluating the scientific evidence for DNA methylation marks controlling neurogenesis and discuss recent technical advances to study their function at precise sites in the genome.
Methylation-sensitive restriction enzymes (MREs), such as BstUI, HpaII, NotI and SmaI, cleave only their unmethylated target sequences (see [31] for lists of MREs) and leave the methylated DNA intact. MRE digestion has been coupled with sequencing technologies to predict genome-wide DNA methylation levels [32]. In the workflow of MRE digestion followed by sequencing (MRE-seq), the MRE cleaves the unmethylated CpG sites of genomic DNA, and the resulting DNA fragments are size-selected and sequenced. The sequencing results reveal the locations of the unmethylated CpG sites within the recognition sites of the enzyme utilized [33]. MRE-seq allows the estimation of relative DNA methylation levels but has relatively low coverage of the genome because the CpG-containing recognition sites are limited. The distribution of CpG sites in the genome is as important as the role of DNMT1 activity. During the evolution, the sequence CpG has been progressively eliminated from the genome due to deamination of methylcytosines to thymines.
Highly active TET proteins are required for the efficient conversion of 5mC to 5caC (more than 96), or else the incomplete conversion of 5mC might lead to false identification as 5hmC sites [81]. Both oxidative bisulfite conversion and TET-assisted bisulfite conversion are compatible with microarray and sequencing platforms to generate the 5hmC methylation profile for a whole genome or targeted regions [82]. The relatively low levels of 5hmC and the subtraction step demand an increase in the sequencing coverage and the number of replicates. A study of human PGC epigenome used TAB-seq to reveal the demethylation during epigenetic reprogramming between 57 and 113 days, and the heterogeneity of 5hmc in both individual loci and at individual cells has been identified [83]. Despite the clear association of DNA hypomethylation with both spontaneous and experimentally derived tumours, the exact role of this change is poorly understood. In 1983 Feinberg and Vogelstein reported a decrease of methylation in the promoter regions of c-Ha-ras and c-Ki-ras in lung and colon carcinomas [54].
That Dnmt1 constantly antagonizes passive DNA demethylation is widely accepted. Whether there are any active processes selectively removing DNA methylation marks from certain epigenomic locations has been a controversial issue for a long time. Over the last decades there have been a series of reported findings of DNA demethylases (wittily summarized by Ooi and Bestor, 2008). In contrast to those, recent candidates have been received more favorably (Wu and Zhang, 2010). Today it is widely accepted, that a number of enzymes contribute on the de-methylation of 5mC. First of all the members of the ten-eleven translocation family of enzymes (Tet1, Tet2 and Tet3) oxidize 5mC to 5hmC. But Tet activity does not necessarily stop at this point, as these enzymes can further oxidize 5hmC to 5fC and subsequently to 5caC (Figure 1) (He et al. , 2011; Ito et al. , 2011).
Interestingly, DNA modifications, due to their mode of inheritance, have been frequently suggested as prime candidates for memory storage (Griffith and Mahler, 1969; Crick, 1984). Already in 1969 J. S. Griffith suggested that the physical basis of memory could lie in the enzymatic modification of the DNA of nerve cells. It might be worth looking to see if there are unusual bases specific to nerve cell DNA, but in the absence of evidence to that effect, a plausible suggestion would be that the modification consists of methylation (or demethylation) (Griffith and Mahler, 1969). During the last decades this concept has been regularly revived (Meagher, 2014). Indeed we know now, that the brain is, compared to other organs, especially active in remodeling DNA methylation patterns and a prominent source of scarce DNA modifications. For example, non-CpG methylation is common in neurons in contrast to other differentiated cell types (Guo et al.
The comprehensive high-throughput arrays for relative methylation (CHARM) method first uses McrBC, an enzyme that digests methylated DNA, to fractionate DNA and subsequently utilizes array hybridization [34]. McrBC recognizes RmC(N)55103RmC and cleaves half of the methylated DNA and all the methylated CGIs [35], and thus, relatively unmethylated DNA will be size-selected and hybridized to the array. Using CHARM, Irizarry et al. discovered that most DNA methylation differences between colon cancer and adjacent normal tissues occurred in sequences up to 2kb away from CGIs, termed CpG island shores (CGI shores) [36]. Unexpectedly, differentially methylated regions (DMRs) in CGI shores have a strong inverse relationship with differential gene expression. CHARM, as a restriction enzyme-based method, is able to detect DMRs at CGI shores, which are otherwise not detectable with CpG-directed enrichment methods such as methylated DNA immunoprecipitation (MeDIP).
Bachman, M., Uribe-Lewis, S., Yang, X., Williams, M., Murrell, A., and Balasubramanian, S. (2014). 5-Hydroxymethylcytosine is a predominantly stable DNA modification. Nat. Chem. 6, 10491055. doi:Source: https://www.shutterstock.com