Epigenetics: differential DNA methylation in mammalian somatic tissues

Authors


H. Nagase, Advanced Research Institute for the Sciences and Humanities, Nihon University, Nihon University Kaikan Daini Bekkan, 12-5, Goban-cho, Chiyoda-ku, Tokyo 102-8251, Japan
Fax: +81 3 3972 8337
Tel: +81 3 3972 8337
E-mail: nagase-hiroki@arish.nihon-u.ac.jp

Abstract

Epigenetics refers to heritable phenotypic alterations in the absence of DNA sequence changes, and DNA methylation is one of the extensively studied epigenetic alterations. DNA methylation is an evolutionally conserved mechanism to regulate gene expression in mammals. Because DNA methyation is preserved during DNA replication it can be inherited. Thus, DNA methylation could be a major mechanism by which to produce semi-stable changes in gene expression in somatic tissues. Although it remains controversial whether germ-line DNA methylation in mammalian genomes is stably heritable, frequent tissue-specific and disease-specific de novo methylation events are observed during somatic cell development/differentiation. In this minireview, we discuss the use of restriction landmark genomic scanning, together with in silico analysis, to identify differentially methylated regions in the mammalian genome. We then present a rough overview of quantitative DNA methylation patterns at 4600 NotI sites and more than 150 differentially methylated regions in several C57BL/6J mouse tissues. Comparative analysis between mice and humans suggests that some, but not all, tissue-specific differentially methylated regions are conserved. A deeper understanding of cell-type-specific differences in DNA methylation might lead to a better illustration of the mechanisms behind tissue-specific differentiation in mammals.

Abbreviations
DMR

differentially methylated region

RLGS

restriction landmark genomic scanning

T-DMR

tissue-specific differentially methylated region

Vi-RLGS

virtual-image restriction landmark genomic scanning

V-RLGS

virtual restriction landmark genomic scanning

Cytosine methylation of CpG dinucleotides is an important epigenetic modification that has profound roles in gene regulation, development and carcinogenesis [1,2]. Methylation of CpG clusters or CpG islands within gene promoters can silence gene expression [3,4]. Therefore, identifying changes in DNA methylation at CpG islands is expected to lead to a clearer understanding of the differentiation of normal tissues and the development of complex diseases including cancer [5]. The DNA methylation pattern is somatically heritable via the effect of the maintenance DNA methyltransferase, DNMT1 [6]. The error rate of maintaining DNA methylation is low (∼ 1% per division) at human CpG sites [7]. During embryonic development, both somatic and germ-cell DNA methylation patterns are erased and then re-established during cell differentiation. Once established, DNA methytlation patterns are thought to be stable. Although it has been reported that DNA methylation may play a role in the regulation of tissue-specific gene expression [8,9], differential DNA methylation patterns among adult tissues were not confirmed until recently.

The flowering plant Arabidopsis thaliana, with mutations in the cytosine–DNA-methyltransferase gene, MET1, which led to a global reduction in cytosine methylation, is viable, and the delay in its flowering time is observed only after several generations [10]. Embryos from DNA methyltransferase gene-deficient mice, which have reduced levels of cytosine methylation, develop until the stage of 8.5 days, when many tissues are already differentiated [11]. Furthermore, analysis of DNA methylation patterns in genes known to be expressed in a tissue-specific manner failed to confirm a major role for DNA methylation in differentiation [12,13]. Using microarray analysis, only five genes expressed in a tissue-specific manner showed a significant increase in expression level in an in vivo system lacking DNA methylation [14]. These data suggested that DNA methylation had no role in regulating gene expression during development.

Song et al. [15] reported 150 tissue-specific differentially methylated regions (T-DMRs) in the mouse genome using restriction landmark genomic scanning (RLGS) in conjunction with virtual-image restriction landmark genomic scanning (Vi-RLGS) and confirmed at least 14 T-DMRs by bisulfite sequencing. Some of the confirmed T-DMRs exhibited a tissue-specific expression pattern that is consistent with methylation status and may play a role in tissue differentiation [15]. Subsequent studies reported the existence of frequent tissue-specific methylation events in mice [16,17] and humans [9,18–20]. Thus, the extent of DNA methylation appears to change in a systematic way during mammalian development.

DNA methylation status may be influenced by environmental exposure [2,20,21]. In gastric mucosa, Helicobacter pylori infection potently induces the methylation of several CpG islands [21]. Young monozygotic twins are essentially indistinguishable in their epigenetic markings, whereas older monozygous twins exhibit remarkable differences in overall content and genomic distribution of 5-methylcytosine DNA and histone acetylation [22]. Thus, the previous hypothesis that DNA methylation patterns acquired during development in mammals were stable in adult somatic cells can be discarded in favor of the accumulated evidence of frequent appearances of differentially methylated genomic regions in various tissue environments.

RLGS method for the mammalian genome

RLGS is a method for the 2D display of end-labeled DNA restriction fragments [23]. This method involves digestion of high molecular mass genomic DNA with a ‘landmark’ enzyme. The landmark enzyme, such as the methylation-sensitive enzyme NotI or AscI, determines the sites of the genome that will be labeled by filling in enzyme half-sites with radioactive nucleotides. Because the NotI recognition site contains two CpGs, and > 90% of the NotI sites are thought to lie within CpG islands, RLGS (with NotI and similar restriction enzymes) displays the DNA methylation status of the CpG islands and associated regions [23]. For example, when comparing normal and cancer RLGS profiles, spot loss due to methylation occurs because the methylated NotI site is not cut by the enzyme and is therefore not labeled. By contrast, spot gain in cancer indicates ‘demethylation’ of a NotI site which was methylated in normal tissues. Using methylation-sensitive NotI as a landmark, ∼ 1500–2000 spots (end-labeled restriction fragments) can be resolved on a single gel. These methods have been used to identify imprinted sites, aberrant methylated sites in cancers and epigenetic remodeling of mammalian tissues [23–25].

The lack of a PCR step and hybridization in the RLGS procedure provides an important advantage over other methods in identifying aberrant DNA methylation. RLGS profiles are quantitative, and the sensitivity is such that methylation can be reliably detected when > 40% of the alleles are methylated. This level of sensitivity ensures that the major demethylation events present in the sample could be detected. By contrast, other approaches, such as bisulfite sequencing or chromatin/methyl-cytosine precipitation, would allow the detection of very rare methylation events (false positive) or the omission of a partial DNA methylation (false negative), due to the involvement of array hybridization, PCR amplification and affinity precipitation.

Computational approaches for RLGS (Vi-RLGS)

Despite its clear importance and successes, RLGS has some potential drawbacks, the most significant of which is the difficulty of cloning individual spots [26,27]. This is critically important because the sequence of the altered RLGS spot must be determined in order to identify the affected gene. Clearly, with the availability of the genome sequence for many organisms, it has become possible to use this information to identify specific restriction fragments within genomes and produce in silico size fractionations [28–30]. Several in silico analyses that have been made in this sense include the Virtual Genome Scan (http://dot.ped.med.umich.edu:2000/VGS/index.html) [30], in silico digests [28], Vi-RLGS [29] and virtual restriction landmark genomic scanning (Conime by R. Wenger; http://www.cse.ohio-state.edu/~wenger/research/conime/contact.html). These tools help to create a more complete map of genomic sites that are either methylated or demethylated in current human and mouse genomic DNA sequence data.

Application of Vi-RLGS to the C57BL/6J mouse genome

We applied the Vi-RLGS software directly to the mouse genome using a NotI–PstI–PvuII combination. Examinations of a sample field from C57BL/6J liver DNA identified ∼ 1460 unmethylated spots in real RLGS, compared with 2170 spots in the same field of the virtual pattern. The vast majority of the ‘extra spots’ in the virtual profile in the mouse are, in fact, derived from repetitive sequences, which would be expected to be methylated and absent in the real profile [15].

This method has been applied to the mouse genome using six different tissues (testis, brain, colon, kidney, liver and muscle) [15]. The methylation status of ∼ 4600 genomic sites fell into one of the following three categories: constitutively methylated, constitutively demethylated and methylated in a tissue-specific manner. The frequency of T-DMRs is estimated to be ∼ 5% (836/15 500 CpG islands) in the mouse genome [15]. This estimate may be different because RLGS is strongly biased by the genomic location of the NotI sites. DNA methylation profiling of human chromosomes 6, 20 and 22 suggested that many T-DMRs are not in CpG islands [31], but recent global human genome searches have identified ∼ 700 T-DMRs for human promoter regions, many of which are included in CpG islands [9,19,20]. However, when the two recent human global analyses are compared, 283 gene promoters are identified as ‘testis-specific’ DMRs in one analysis [19] and 104 in another [20]. Among these gene promoters, only 18 were concurrently identified as ‘testis-specific’ DMRs by both studies [19,20]. Although potential sources of contradictions in these two publications may be the differences in type of tissue, DNA purification and methodology used, most of recent methodologies are not accurately quantitative. A reproducible method for quantitative DNA methylation detection is needed for the in vivo study, in which DNAs are often prepared from a mixed cell population.

The application of a quantitative whole-genome methylation analysis by RLGS to the mouse genome provides evidence for specific differences in the DNA methylation patterns during development, differentiation, aging and in diseases such as cancer. The Vi-RLGS application, together with a sophisticated image-matching and registration program and spot intensity analysis program may provide a new analytical tool to measure the global methylation patterns of cell population in specific tissue environments [32,33] (G. E. Bove and P. Rogersen, University at Buffalo, NY, USA; unpublished data). Analysis of deposited numbers of previously performed RLGS images may prove to be a treasure chest for understanding the methylation patterns of each tissue or disease state. Based on those efforts and a considerable number of RLGS experiments, a rough draft overview of quantitative DNA methylation patterns with a quantitative DNA measurement of the C57BL/6J mouse genome has been created by using the Vi-RLGS in silico analysis. Figure 1 shows a draft NotI methylation map of C57BL/6J strain based on two RLGS profiles of NotI–PstI–PvuII and NotI–PvuII–PstI enzyme combinations. Table 1 is a preliminary distribution pattern on each chromosome of T-DMRs, constitutively methylated and constitutively non- or partially methylated NotI sites using a virtual-image RLGS analysis (note that for most NotI sites the methylation pattern has yet to be confirmed by other methods). Interestingly, considering the gene-poor regions [34], a significantly high number of T-DMRs are located in gene-rich genomic region, while non-T-DMRs are located in both (Fig. 2). In addition, a relatively high percentage of NotI sites in T-DMRs are located in the non-promoter region (exons, introns and 3’ regions). This may suggest that T-DMRs are likely to modify gene expression through transcriptional regulation or may have other functions that are unrelated to transcription. However, the functional relationships between T-DMRs and regulation mechanisms involved in tissue differentiation are unknown. Intragenic DNA methylation is known to be capable of altering the chromatin structure and elongation efficiency in mammalian cells, depleting RNA polymerase II exclusively in the methylated region [35]. It has been suggested that the methylation of Alu elements could suppress transcription and contribute to differential gene expression [36–38]. A study of transgenic mice demonstrated that the epigenetic modification of transgenes under the control of the mouse mammary tumor virus LTR conferred a tissue-dependent influence on the transcription of the transgenes [39]. Recent evidence suggested that there are related regulation mechanisms between micro RNA and epigenetics [40]. Although it has been reported that 95% of mammalian genomes are transcribed and have some functional means [41], the localization bias of T-DMRs may suggest that DNA methylation is a critical tissue-specific regulation mechanism and that it modifies RNA transcription. The T-DMR located in gene-poor regions may also facilitate the identification of previously unidentified regulatory mechanisms.

Figure 1.

NotI methylation map of the C57BL/6J genome. The methylation status of NotI sites is plotted on 19 mouse autosomal chromosomes. Constitutively methylated, constitutively unmethylated or tissue-specific methylated sites are indicated by red bars (left-hand side of the chromosome), green bars (right-hand side of the chromosome) and blue bars (left-hand side of the chromosome), respectively. Vi-RLGS analysis was performed in conjunction with duplicate RLGS analyses of six tissues (liver, muscle, kidney, colon, testis and brain) using Mouse Aug. 2005 (mm7) assembly, and then the methylation pattern of each spot of RLGS autographs were analyzed.

Table 1.   Preliminary distribution pattern of T-DMRs, on each chromosome showing constitutively methylated and constitutively non- or partially methylated NotI sites identified by virtual-image RLGS analysis system.
Chromosome12345678910111213141516171819Total
Constitutively unmethylated regions2243421842752711742522182282073161541711291431381741251243849
%78.088.481.191.186.980.286.085.886.083.889.086.581.479.185.681.784.182.287.385.9
Constitutively methylated regions58373517293228253136231535291825252214534
%20.29.615.45.69.314.79.69.811.714.66.58.416.717.810.814.812.114.59.911.8
T-DMRs5881012111311641694566854151
%1.72.13.53.33.85.14.44.32.31.64.55.11.93.13.63.63.93.32.83.3
Total2873872273023122172932542652473551782101631671692071521424534
%6.28.44.96.66.84.76.45.55.85.47.73.94.63.53.63.74.53.33.1 
Physical length Mb195181158154150150139127124130121115114118103979290602596
%7.57.06.15.95.85.85.44.94.85.04.74.44.44.54.03.73.53.52.3 
Figure 2.

 Genomic regions of unmethylated NotI sites and NotI sites showing tissue-specific differentially methylated regions (T-DMRs). A bar graph indicates the percent of genomic distribution of NotI sites separately analyzed between those located within constitutively unmethylated regions (white bars) and those in T-DMRs (gray bars) at each indicated genomic region. Intergenic region (Junk), intragenic region (promoter, exon, 3′ regions; non-Junk), CpG island and non-CpG island were evaluated by UCSC genome browser (mm8).

RLGS for cancer study

RLGS has been used to study the degree of hypomethylation as well as to identify the targets of hypermethylation in many different types of cancerous tissues [24,42]. These studies reveal that CpG island methylation in cancers is non-random and shows both inter- and intratumor heterogeneity. Furthermore, certain RLGS fragments were found to be methylated in many different cancers, whereas others were methylated exclusively in one. In the set of tumors described by Costello et al. [24], gliomas had 34 RLGS loci methylated in > 40% of the samples (n = 14), colon tumors had 23 (n = 8) and medulloblastomas had eight (n = 22). Even if the associated gene is not known or ever found to be associated with cancer, given that they are methylated at such high frequency, methylation of these loci could be used as biomarkers. Rarely known tumor-suppressor genes have been shown to be methylated at a frequency of > 40% of tumors, even using highly sensitive PCR-based techniques [43]. One remarkable exception to this is the GSTP1 gene, which was shown to be hypermethylated in 40/42 human prostate cancers [44] and shows promising signs of becoming an excellent biomarker for prostate cancer [45]. A quantitative analysis such as RLGS provides an opportunity to scan the genome for frequent targets of methylation in a mixed cell population such as tumor DNA. This may have even more potential than existing DNA methylation biomarkers. In addition, studying DNA methylation for loci whose primary route of inactivation is through CpG island methylation, rather than through genetic disruption, may be more likely to result in the identification of effective therapeutic targets.

Conclusion

The recent technical revolution in epigenetic detection is providing a clearer understanding of epigenetic marks in the genomes of every mammalian cell type, even though genome-wide quantitative DNA methylation analysis is yet to be completely established. RLGS, together with in silico analysis, is a useful technique for comparing genome-wide DNA methylation patterns between tissues or disease states. The RLGS analysis has provided evidence of frequent T-DMRs in mammalian genomes and epigenetic control of tissue-specific RNA transcription modification.

Acknowledgements

We thank C. J. Kemp and W. A. Held for his critical reading. This work is supported by the Nihon University Multidisciplinary Research Grant for 2006, the Academic Frontier Project for 2006 Project for Private Universities: matching fund subsidy from MEXT, National Cancer Institute Grant CA102423 and the National Cancer Institute Center Support Grant CA16056 (to Roswell Park Cancer Institute).

Ancillary

Advertisement