An overview of the analysis of DNA methylation in mammalian genomes

DNA methylation at position C5 of the pyrimidine ring of cytosine in mammalian genomes has received a great deal of research interest due to its importance in many biological phenomena. It is associated with events such as epigenetic gene silencing and the maintenance of genome integrity. Aberrant DNA methylation, particularly that of chromosomal regions called CpG islands, is an important step in carcinogenesis. In order to elucidate methylation profiling of complex genomes, various methods have been developed. Many of these methods are based on the differential reactivity of cytosine and 5-methylcytosine to various chemicals. The combined use of these chemical reactions and other preexisting methods has enabled the discrimination of cytosine and 5-methylcytosine in complex genomes. The use of proteins that preferentially bind to methylated DNA has also successfully been used to discriminate between methylated and unmethylated sites. The chemical and structural dissection of the in vivo processes of enzymatic methylation and the binding of methyl-CpG binding proteins provides evidence for the complex mechanisms that nature has acquired. In this review we summarize the methods available for the discrimination between cytosine and 5-methylcytosine in complex genomes.     

De novo DNA methylation at nonrandom founder sites 5' from an unmethylated minimal origin of DNA replication in latent Epstein-Barr virus genomes

Latent episomal genomes of Epstein-Barr virus, a human gammaherpesvirus, represent a suitable model system for studying replication and methylation of chromosomal DNA in mammals. We analyzed the methylation patterns of CpG dinucleotides in the latent origin of DNA replication of Epstein-Barr virus using automated fluorescent genomic sequencing of bisulfite-modified DNA samples. We observed that the minimal origin of DNA replication was unmethylated in 8 well-characterized human cell lines or clones carrying latent Epstein-Barr virus genomes as well as in a prototype virus producer marmoset cell line. This observation suggests that unmethylated DNA domains can function as initiation sites or zones of DNA replication in human cells. Furthermore, 5' from this unmethylated region we observed focal points of de novo DNA methylation in nonrandom positions in the majority of Burkitt's lymphoma cell lines and clones studied while the corresponding CpG dinucleotides in viral genomes carried by lymphoblastoid cell lines and marmoset cells were completely unmethylated. Clustering of highly methylated CpG dinucleotides suggests that de novo methylation of unmethylated double-stranded episomal viral genomes starts at discrete founder sites in vivo. This is the first comparative high-resolution methylation analysis of a latent viral origin of DNA replication in human cells.     

Studies on the biological role of DNA methylation: V. The pattern of E.coli DNA methylation.

The distribution of the methylatable sites GATC and CCATGG was studied by analyzing the molecular average size of restriction fragments of E. coli DNA. Both sites were found to be randomly distributed, reflecting a random pattern of methylation. The methylation pattern of specific sequences such as the origin of replication and rRNA genes has been studied in wild type E. coli and a methylation deficient (dam- dcm-) mutant. These sequences were found to be methylated in wild type cells and unmethylated in the mutant indicating that there is no effect of the state of methylation of these sequences on their expression. Analysis of the state of methylation of GATC sites in newly replicating DNA using the restriction enzyme Dpn I (cleaves only when both strands are methylated) revealed no detectable hemimethylated DNA suggesting that methylation occurs at the replication fork. Taking together the results presented here and previously published data (5), we arrive at the conclusion that the most likely function of E. coli DNA methylations is probably in preventing nuclease activity.     

Methylation of deoxycytidine incorporated by excision-repair synthesis of DNA

Methylation of deoxycytidine incorporated by DNA excision-repair was studied in human diploid fibroblasts following damage with ultraviolet radiation, N-methyl-N-nitrosourea, or N-acetoxy-2-acetylami-nofluorene. In confluent, nondividing cells, methylation in repair patches induced by all three agents is slow and incomplete. Whereas after DNA replication in logarithmic-phase cultures a steady state level of 3.4% 5-methylcytosine is reached in less than 2 hr after cells are labeled with 6-³H-deoxycytidine, following ultraviolet-stimulated repair synthesis in confluent cells it takes about 3 days to reach a level of ～2.0% 5-methylcytosine in the repair patch. In cells from cultures in logarithmic-phase growth, 5-methylcytosine formation in ultraviolet-induced repair patches occurs faster and to a greater extent, reaching a level of ～2.7% in 10–20 hr. Preexisting hypomethylated repair patches in confluent cells are methylated further when the cells are stimulated to divide; however, the repair patch may still not be fully methylated before cell division occurs. Thus DNA damage and repair may lead to heritable loss of methylation at some sites.     

Direct detection of methylation in genomic DNA

The identification of methylated sites on bacterial genomic DNA would be a useful tool to study the major roles of DNA methylation in prokaryotes: distinction of self and nonself DNA, direction of post-replicative mismatch repair, control of DNA replication and cell cycle, and regulation of gene expression. Three types of methylated nucleobases are known: N⁶-methyladenine, 5-methylcytosine and N⁴-methylcytosine. The aim of this study was to develop a method to detect all three types of DNA methylation in complete genomic DNA. It was previously shown that N⁶-methyladenine and 5-methylcytosine in plasmid and viral DNA can be detected by intersequence trace comparison of methylated and unmethylated DNA. We extended this method to include N⁴-methylcytosine detection in both in vitro and in vivo methylated DNA. Furthermore, application of intersequence trace comparison was extended to bacterial genomic DNA. Finally, we present evidence that intrasequence comparison suffices to detect methylated sites in genomic DNA. In conclusion, we present a method to detect all three natural types of DNA methylation in bacterial genomic DNA. This provides the possibility to define the complete methylome of any prokaryote.     

Distribution of 5-methylcytosine in the DNA of somatic and germline cells from bovine tissues.

Genomic DNA of calf thymus contains 1.5 times as much 5-methylcytosine as similar sperm DNA, but the major EcoRI repeat fragment from satellite I of thymus contains ten times as much 5-methylcytosine as the corresponding fragment from sperm DNA. Restriction enzyme analyses of the total DNA and the satellite I fragment show that three HpaII sites in the fragment are completely unmethylated in sperm but fully methylated in thymus DNA. Under-methylation of many sites in the satellite DNAs can probably account for the lower level of methylation of sperm DNA rather than hemimethylation as previously suggested. These results are also discussed in relation to maintenance and de novo (initiation-type) methylases.     

DNA methylation. Inhibition of de novo and maintenance methylation in vitro by RNA and synthetic polynucleotides

A partially purified HeLa cell DNA methylase will methylate a totally unmethylated DNA (de novo methylation) at about 3-4% the rate it will methylate a hemimethylated DNA template (maintenance methylation). Our evidence suggests that many, if not most, dCpdG sequences in a natural or synthetic DNA can be methylated by the enzyme. There is a powerful inhibitor of DNA methylase activity in crude extracts which has been identified as RNA. The inhibition of DNA methylase by RNA may indicate that this enzyme is regulated in vivo by the presence of RNA at specific chromosomal sites. The pattern of binding of RNA to DNA in the nucleosome structure and the DNA replication complex may determine specific sites of DNA methylation. An even more potent inhibition of DNA methylase activity is observed with poly(G), but not poly(C), poly(A), or poly(U). The only other synthetic polynucleotides studied which inhibit DNA methylation as well as poly(G) are the homopolymers poly(dC).poly(dG) and poly (dA).poly(dT). These results point out the unique importance of the guanine residue itself in the binding of the DNA methylase to dCpdG, the site of cytosine methylation. The surprising inhibition of the methylation reaction by poly(dA).poly(dT), which is itself not methylated by the enzyme, suggests the possible involvement of adjacent A and T residues in influencing the choice of sites of methylation by the enzyme.     

Topographical distribution of 5-methylcytosine in animal and plant DNA.

The topographical distribution of 5-methylcytosine on animal and plant cell DNA has been examined with methyl-sensitive restriction enzymes and gel electrophoresis analysis. These DNAs digested with the enzyme HpaII have a partially bimodal size distribution, indicating the existence of clusters of methylated and unmethylated CCGG sites in the DNA. By analyzing the methylation state of all CG moieties in restricted DNA fractions, it was possible to show that these genomes are, in general, arranged as clusters of relatively highly methylated and undermethylated regions. Plant DNA also contains 5-methylcytosine in the prototype sequence C-X-G. Restriction of this DNA with EcoRII revealed that these methyl groups are also distributed in clusters, suggesting that this is a general phenomenon. The undermethylated areas may correspond to the active fraction of the genome.     

Methylation pattern of mouse mitochondrial DNA.

The pattern of methylation of mouse mitochondrial DNA (mtDNA) was studied using several techniques. By employing a sensitive analytical procedure it was possible to show that this DNA contains the modified base 5-methylcytosine (m5Cyt). This residue occurred exclusively at the dinucleotide sequence CpG at a frequency of 3 to 5%. The pattern of methylation was further investigated by determining the state of methylation of several MspI (HpaII) sites. Different sites were found to be methylated to a different extent, implying that methylation of mtDNA is nonrandom. Based on the known base composition and nucleotide sequence of mouse mtDNA, the dinucleotide sequence CpG was found to be underrepresented in this DNA. The features of mtDNA methylation (CpG methylation, partial methylation of specific sites and CpG underrepresentation) are also characteristic of vertebrate nuclear DNA. This resemblance may reflect functional relationship between the mitochondrial and nuclear genomes.     

Non-enzymatic DNA methylation by S-adenosylmethionine results in the formation of minor thymine residues and 5-methylcytosine from cytosine

It was found that nonenzymatic DNA methylation proceeds in water solution in the presence of S-adenosylmethionine (AdoMet). The main reaction products are thymine and 5-methylcytosine residues. It was shown that labelled thymine residues are formed also upon DNA incubation in the presence of [methyl-14C]methionine as well as [methyl-14C]cobalamine. Only cytosine reacts with AdoMet resulting in thymine production. AdoMet may be a potential mutagen that induces GC----AT transitions during DNA replication in the cell.     

Methylation of parental and progeny DNA strands in Physarum polycephalum

Although 5-methylcytosine comprises 4 to 8% of the cytosine residues in the major nuclear DNA of Physarum polycephalum (Evans &; Evans, 1970), only 1 % of the cytosine residues of progeny DNA become methylated during replication. Further methylation occurs during the same and subsequent mitotic cycles, so that 6 to 7 cycles after its synthesis, 5-methylcytosine comprises 5 to 7% of the DNA-cytosine residues of a single generation of DNA. The extent of methylation occurring during the S period has been measured by the determination of the specific activity of the precursor (S-adenosylmethionine) and the product (DNA-5-methylcytosine) and by comparison of the radioactivity in DNA-cytosine and DNA-5-methylcytosine after incorporation of [¹⁴C]deoxycytidine. Continuing methylation of parental DNA has been shown, by density shift experiments and by the conversion of prelabeled DNA-cytosine to DNA-5-methylcytosine. The DNA-5-methylcytosine once formed was found to be stable.     

Methylation-sensitive single-nucleotide primer extension (Ms-SNuPE) for quantitative measurement of DNA methylation

Methylation-sensitive single-nucleotide primer extension (Ms-SNuPE) is a technique that can be used for rapid quantitation of methylation at individual CpG sites. Treatment of genomic DNA with sodium bisulfite is used to convert unmethylated Cytosine to Uracil while leaving 5-methylcytosine unaltered. Strand-specific PCR is performed to generate a DNA template for quantitative methylation analysis using Ms-SNuPE. SNuPE is then performed with oligonucleotide(s) designed to hybridize immediately upstream of the CpG site(s) being interrogated. Reaction products are electrophoresed on polyacrylamide gels for visualization and quantitation by phosphorimage analysis. The Ms-SNuPE technique is similar to other quantitative assays that use bisulfite treatment of genomic DNA to discriminate unmethylated from methylated Cytosines (i.e., COBRA, pyrosequencing). Ms-SNuPE can be used for high-throughput methylation analysis and rapid quantitation of Cytosine methylation suitable for a wide range of biological investigations, such as checking aberrant methylation changes during tumorigenesis, monitoring methylation changes induced by DNA methylation inhibitors or for measuring hemimethylation. Approximately two to four CpG sites can be interrogated in up to 40 samples by Ms-SNuPE in less than 5 h, after PCR amplification of the desired target sequence and preparation of PCR amplicons.     

Effect of trypsin on DNA methylation in isolated nuclei from developing sea urchin embryos

Nuclei isolated from the developing sea urchin embryo $\text{Paracentrotus lividus}$ and incubated in the presence of [³H-methyl] S-adenosylmethionine methylate their own DNA. Addition of small amounts of trypsin produces a 20-fold increase in DNA methylation. The time kinetics and the specificity of the trypsin activation of DNA methylation are described. The only products of the reaction are 5-methylcytosine and thymine. DNA adenine, guanine and cytosine are not labeled. The distribution of the counts between 5-methyl-cytosine and thymine is variable. While 5-methylcytosine originates by enzymatic methylation of DNA cytosines, the origin of the labeled thymine cannot be inferred with certainty.     

Thymine DNA glycosylase is essential for active DNA demethylation by linked deamination-base excision repair

DNA methylation is a major epigenetic mechanism for gene silencing. Whereas methyltransferases mediate cytosine methylation, it is less clear how unmethylated regions in mammalian genomes are protected from de novo methylation and whether an active demethylating activity is involved. Here, we show that either knockout or catalytic inactivation of the DNA repair enzyme thymine DNA glycosylase (TDG) leads to embryonic lethality in mice. TDG is necessary for recruiting p300 to retinoic acid (RA)-regulated promoters, protection of CpG islands from hypermethylation, and active demethylation of tissue-specific developmentally and hormonally regulated promoters and enhancers. TDG interacts with the deaminase AID and the damage response protein GADD45a. These findings highlight a dual role for TDG in promoting proper epigenetic states during development and suggest a two-step mechanism for DNA demethylation in mammals, whereby 5-methylcytosine and 5-hydroxymethylcytosine are first deaminated by AID to thymine and 5-hydroxymethyluracil, respectively, followed by TDG-mediated thymine and 5-hydroxymethyluracil excision repair.     

Arabidopsis DEMETER-LIKE proteins DML2 and DML3 are required for appropriate distribution of DNA methylation marks

Cytosine DNA methylation is a stable epigenetic mark for maintenance of gene silencing across cellular divisions, but it is a reversible modification. Genetic and biochemical studies have revealed that the Arabidopsis DNA glycosylase domain-containing proteins ROS1 (REPRESSOR OF SILENCING 1) and DME (DEMETER) initiate erasure of 5-methylcytosine through a base excision repair process. The Arabidopsis genome encodes two paralogs of ROS1 and DME, referred to as DEMETER-LIKE proteins DML2 and DML3. We have found that DML2 and DML3 are 5-methylcytosine DNA glycosylases that are expressed in a wide range of plant organs. We analyzed the distribution of methylation marks at two methylated loci in wild-type and dml mutant plants. Mutations in DML2 and/or DML3 lead to hypermethylation of cytosine residues that are unmethylated or weakly methylated in wild-type plants. In contrast, sites that are heavily methylated in wild-type plants are hypomethylated in mutants. These results suggest that DML2 and DML3 are required not only for removing DNA methylation marks from improperly-methylated cytosines, but also for maintenance of high methylation levels in properly targeted sites.     

Repair methylation of parental DNA in synchronized cultures of Novikoff hepatoma cells.

Parental and filial DNA strands were isolated from a Novikoff rat hepatoma cell line, synchronized by S-phase arrest with excess thymidine, that had completed up to one round of DNA replication in the presence of (14-C-methyl)methionine and (6-3-H) bromodeoxyuridine. Both strands were methylated, the proportion of total methyl label in parental DNA increasing slightly with time in S-phase. The studies were repeated with (14-C-methyl)methionine and (3-H)deoxycytidine to determine if parental methylation occurred on extant or repair-inserted cytosine residues. Both (14-C) and (3-H) were found in parental DNA. The (14-C)/(3-H) ration of parental DNA-5-methylcytosine was about twice that in filial DNA while the (3-H) data showed twice the concentration of 5-methylcytosine in parental compared to filial DNA. Thus parental methylation occurred on repair-inserted cytosine residues and resulted in overmethylation. That the DNA damage and repair was due to 5-phase arrest was shown by repeating the studies using a sequential mitotic-G1 arrest method. With this method little (14-C) or (3-H) was found in parental DNA. We conclude that S-phase arrest leads to DNA damage and repair with subsequent overmethylation of repair-inserted cytosines; that sequential mitotic-G1 arrest minimizes DNA damage; and, that the latter technique, suitable for synchronization of large quantities of cells, may prove useful in relatively artifact-free studies of eukaryotic DNA replication.     

Characteristics of enzymatic DNA methylation in cultured cells of human and hamster origin,and the effect of DNA replication inhibition

In mammalian cells, inhibitors of DNA replication have been shown to induce chromosomal aberrations, cell death and changes in gene control. Inhibition of DNA synthesis has been reported to induce hypermethylation of mammalian DNA (enzymatic postsynthetic formation of 5-methylcytosine). These 5-methylcytosines in mammalian DNA have variously been suggested to be important in gene control, DNA repair, and control of DNA replication. In establishing the normal characteristics of enzymatic DNA methylation, we have demonstrated that, in asynchronously growing cells of both human and hamster origin, some cytosine methylation is delayed for several hours after strand synthesis and that this delayed methylation is completed before the DNA strand acts as a template for DNA replication in the next S-phase. Further, in testing whether the deleterious effects on mammalian cells of DNA synthesis inhibitors might be mediated via changes in enzymatic DNA methylation, we have found, contrary to some previous findings, no evidence for any change in the level of DNA methylation in DNA strands synthesized during 6 h of treatment of cells of human origin with high concentrations of four different inhibitors of DNA replication or during the 4 h following the 6 h treatment. Almost totally blocking DNA replication had no effect on the small amount of delayed methylation of DNA strands not involved in semi-conservative replication during the time of the experiment. This lack of effect on DNA methylation was obtained when the labelling medium contained normal, undialysed serum. In contrast, if dialysed serum was used in the labelling medium in order to maximize l-[Me-³H]methionine utilization, highly variable, totally irreproducible patterns of apparent DNA hypermethylation were obtained.     

DNA methylation pattern in pig in vivo produced embryos

DNA methylation/demethylation pattern, determined by 5-methylcytosine (5-MeC) immunostaining, was evaluated in porcine “in vivo” produced embryos from zygote up to the blastocyst stage. In one-cell stage embryos, only the maternal pronucleus showed a positive labeling whilst the paternal pronucleus showed almost no labeling. The intensity of labeling is high until the late morula stage. Blastocysts containing less than 100 cells showed the same intensity of labeling in both the inner cell mass (ICM) nuclei and the trophectodermal (TE) cell nuclei. Interestingly, with further cell multiplication, cells of the ICM became more intensively labeled when compared to TE cells. This distinct methylation pattern is even more profound in blastocysts containing about 200–300 cells and is not caused by the difference in the cell volume of ICM and TE cells.An erratum to this article can be found at     

5-Methylcytosine content of rat hepatoma DNA substituted with bromodeoxyuridine.

The aim of these experiments was to test whether incorporation of bromodeoxyuridine into DNA affects DNA methylation. Rat hepatoma (HTC) cells in culture were labeled for two generations with [14C]bromodeoxyuridine and [3H]thymidine to yield DNA which was 2.1, 20.6, 52.6, and 95.0% bromodeoxyuridine-substituted in the newly made strands. The DNA then was fractionated into highly repetitive, moderately repetitive, and single copy sequences. As determined by a comparison of 14C and 3H counts per min, the percentage of substitution with bromodeoxyuridine was found to be the same in each repetition class. The 5-methylcytosine content of each fraction was determined using high pressure liquid chromatography. It was found that bromodeoxyuridine, even at a level of substitution into newly mad DNA of 95%, has no effect on the 5-methylcytosine content of DNA. At all levels of bromodeoxyuridine substitution, highly repetitive DNA has slightly more 5-methylcytosine (3.0% of total cytosine) than does single copy DNA or moderately repetitive DNA (2.3%). The 5-methylcytosine content of whole HTC DNA is the same as that of rat liver DNA (2.4%).