- Open Access
Gradual transition from mosaic to global DNA methylation patterns during deuterostome evolution
© Okamura et al; licensee BioMed Central Ltd. 2010
Published: 15 October 2010
DNA methylation by the Dnmt family occurs in vertebrates and invertebrates, including ascidians, and is thought to play important roles in gene regulation and genome stability, especially in vertebrates. However, the global methylation patterns of vertebrates and invertebrates are distinctive. Whereas almost all CpG sites are methylated in vertebrates, with the exception of those in CpG islands, the ascidian genome contains approximately equal amounts of methylated and unmethylated regions. Curiously, methylation status can be reliably estimated from the local frequency of CpG dinucleotides in the ascidian genome. Methylated and unmethylated regions tend to have few and many CpG sites, respectively, consistent with our knowledge of the methylation status of CpG islands and other regions in mammals. However, DNA methylation patterns and levels in vertebrates and invertebrates have not been analyzed in the same way.
Using a new computational methodology based on the decomposition of the bimodal distributions of methylated and unmethylated regions, we estimated the extent of the global methylation patterns in a wide range of animals. We then examined the epigenetic changes in silico along the phylogenetic tree. We observed a gradual transition from fractional to global patterns of methylation in deuterostomes, rather than a clear demarcation between vertebrates and invertebrates. When we applied this methodology to six piscine genomes, some of which showed features similar to those of invertebrates.
The mammalian global DNA methylation pattern was probably not acquired at an early stage of vertebrate evolution, but gradually expanded from that of a more ancient organism.
In mammals, DNA methylation is achieved by the collaboration of several DNA methyltransferases; i.e., Dnmt1 and the Dnmt3 family . It plays important roles in general gene silencing, the repression of one of the two alleles of imprinted genes and X-linked genes in females, and the inactivation of various transposons. Whereas the Dnmt family is not found in some model organisms, such as the budding yeast, nematode, and fruit fly, in which DNA methylation is absent or below the limit of detection, it is well conserved among deuterostomes . However, functional analyses of the methylation have been limited to mammals; i.e., mouse and human. In contrast, the DNA methylation status has been investigated widely, from mammals to invertebrate deuterostomes, such as the sea urchin, revealing that large parts of these genomes are subject to stable methylation, even in invertebrates . Methylated and unmethylated DNA sequences coexist and pattern the genomes of these animals. Unmethylated stretches in vertebrates are mainly limited to CpG islands, which have been used as gene markers .
It is well documented that the global patterns and levels of DNA methylation are distinct between vertebrates and invertebrate deuterostomes . In the former group, almost every CpG site is methylated, with the exception of those in CpG islands. For instance, nearly 80% of CpG sites are methylated in the human genome . Because CpG sites are excessively concentrated in CpG islands, we may infer that almost all other parts are methylated regions. This is called global pattern of DNA methylation. In contrast, the genome of the sea squirt, an ascidian, for instance, contains roughly equal amounts of methylated and unmethylated regions . Relatively long genomic tracts of tens or hundreds of kilobases are hypermethylated, and other long tracts are hypomethylated. The alternation of these two distinctive types of tracts is called fractional or mosaic pattern of DNA methylation. DNA methylation analyses of the sea urchin, lancelet, ascidian, lamprey, and hagfish have suggested that the transition from the mosaic to the global methylation pattern occurred at an early stage of vertebrate evolution [5, 8].
Curiously, especially to computational biologists, methylated and unmethylated regions can be reliably predicted from the local frequency of CpG dinucleotides . Separated by relatively sharp boundaries, methylated and unmethylated regions tend to contain few and many CpG sites, respectively. This may be attributable to the inherent mutability of 5-methylcytosine  and seems to be consistent with our knowledge of the methylation status of CpG islands and other regions in the mammalian genomes .
So far, however, DNA methylation patterns and methylation levels have not been analyzed in the same way in vertebrates and invertebrates. The prevalent view that the DNA methylation of the two groups differs seems to have hindered us from examining their common features, despite the use of orthologous enzymes in both systems. It is unlikely that the change in the methylation pattern happened suddenly at a specific evolutionary stage. An abrupt change in the pattern of methylation could be deleterious to organisms. If a gradual transition from the mosaic to the global methylation pattern occurred, traces of the transition might be apparent in extant organisms located evolutionarily near the transition zone; e.g., fishes and invertebrate chordates.
To test this possibility of a gradual transition, we developed a new computational methodology to estimate the distributions of methylated and unmethylated regions in various animal genomes. This methodology does not confidently predict the methylation status or level in a specific region, but predicts the proportion of methylated and unmethylated regions from local frequencies of CpG dinucleotides. It is well known that in genomic imprinting and X-inactivation, one member of a pair of alleles with identical sequences sometimes exhibits a different level of DNA methylation, and these variations in methylation levels are often dependent on the developmental stages, tissues, or cell types . The levels also vary among various cancer cells . Heavily methylated CpG islands  and long unmethylated tracts with low CpG contents  have also been documented. Recent research has suggested that specific sequence motifs are more important than CpG contents in the establishment of proper DNA methylation . These facts make us reluctant to predict methylation levels. However, because these observations are exceptional and infrequent compared with the general features observed in whole genomes, the proportion of DNA methylation may be stochastically estimated in each species. Experimental data for the human and ascidian genomes demonstrated the plausibility of this methodology. We applied the method to the analysis of deuterostomes, from the sea urchin through to the human to evaluate the extent of the global methylation pattern, and investigated the global changes in DNA methylation pattern in silico.
A similar analysis of CpG-score distributions has been performed for 3 invertebrate and 6 vertebrate genomes . Those authors showed markedly different distributions of CpG scores in the promoter and intronic sequences. Our results for whole genomes are congruent with their distributions in intronic sequences. It has been reported that invertebrate and vertebrate genomes show bimodal and unimodal distributions, respectively. Using the NOCOM software with an expectation maximization algorithm to fit the distribution data , Elango and Yi revealed that, in invertebrates, methylated and unmethylated regions show discrete normal distributions with low and high CpG scores, respectively. Although the vertebrate distributions were assumed to be unimodal, we noted that the bell shapes were not symmetrical. The right sides of the curves, which correspond to higher CpG scores, bulged slightly. This could be caused by the distribution of CpG scores in unmethylated regions. Indeed, CpG islands, which are considered unmethylated regions in general, cover only 0.69% and 0.39% of the human and mouse genomes, respectively (see Methods). When we applied the NOCOM software to each seemingly unimodal distribution to separate it compulsory into two distinct normal distributions, better fits were obtained (Figure 3).
Statistical data on the methylated and unmethylated regions in deuterostomes
Boundary (CpG score)
Ratio of lengths
Ratio of SDs
Three indices of fishes used to compare their global methylation patterns
Boundary (CpG score)
Ratio of lengths
Ratio of SDs
A bimodal distribution of CpG scores has been reported for mouse and human promoter sequences [11, 22, 23]. However, when we turn our attention to their distribution in the whole genome, the distribution of unmethylated regions, i.e., CpG islands, is negligible and is clearly covered by the distribution of the methylated regions (Figure 3). In the present study, we have demonstrated that the distributions of the methylated and unmethylated regions in whole genomes can be represented by the composition of the two discrete normal distributions. This is supported by experimental data (Figure 4) and overlaying the two distributions produces a better fit. It is not easy to predict DNA methylation levels from local genomic sequences only. Moreover, the status can differ depending on developmental stage, tissue, and cell type. Nevertheless, the implication of the present findings is that the overall patterns formed by sets of each methylation status can be stochastically estimated by fitting them to two normal distributions. This computational methodology allowed us to observe the transition to the global DNA methylation pattern during deuterostome evolution. Evolutionary changes in 5-methylcytocine levels have been investigated in detail and in a large number of species . However, what we have observed here are not average levels but the transition of patterns that are formed by methylated and unmethylated regions.
It has been widely believed that cephalochordates, including lancelets, are more closely related to vertebrates than to urochordates, represented by ascidians . This view may have caused researchers to overlook the gradual transition in the DNA methylation pattern, even though they observed higher levels of methylation in the ascidian than in the lancelet . Recent studies of chordate genomes have changed the conventional consensus [26, 27]. Currently, both urochordates and vertebrates are thought to have evolved from a common ancestor of cephalochordate-like organisms. Consistent with this, our results indicate a firm direction in the changes in the methylation pattern during evolution, based on the assumption that lower organisms have relatively better retained their primitive features than higher ones.
To take a closer look at this transition, we focused on six fish species. The existence of CpG islands in fish genomes is unclear, but an analysis of evidence-based transcription start sites has clearly shown that, like mammals, there are a large number of promoter-associated CpG islands in the medaka genome . However, because sequence features, including the G+C and CpG contents, seem to vary considerably among fishes, the definition of fish CpG islands is still controversial . Adjustable criteria that do not rely on ad hoc thresholds may be required to define standardized CpG islands among highly divergent species. In the present study, CpG islands were not considered because we wanted to give organisms lacking CpG islands equal attention. Instead, the boundaries between methylated and unmethylated regions were defined by analyzing the CpG-score distributions, species by species. Although the indices obtained showed a wide range, they are consistent with the piscine phylogeny (Figure 2) , and are situated between those of invertebrates and tetrapods, supporting a gradual transition in the methylation pattern (Table 2). The two pufferfish genomes show exceptionally high ratios. This could be attributable to their excessively reduced genomes relative to those of other vertebrates . A similar tendency was also observed in platypus, which is a mammal that exhibits some characteristics of reptiles, probably because of the unique character of its genomic sequence . Exceptionally high ratios observed in cat could be ascribed to its low coverage of 1.9 fold (Figure 5) . Its finished sequences are eagerly awaited to know the bona fide causes. We intend to examine the causes of these discrepancies. The genomes of increasing numbers of organisms have been sequenced . The boundary and the indices used to estimate the extent of the global methylation pattern will be fundamental values in future comparative studies.
In mammals, DNA methylation is essential for normal development [12, 16, 35]. A wave of de novo methylation occurs globally around the time of implantation. Interestingly, we observed a transition to the global DNA methylation pattern from the invertebrates to mammals. These facts recall Haeckel's dictum "Ontogeny recapitulates phylogeny." From our computational analysis, it appears that the establishment of global DNA methylation during ontogeny recapitulates phylogeny. Although the methodology is in part supported by experimental data, our conclusion must be confirmed experimentally. The distributions of CpG scores, which are the fundamental data in the present study, might be the consequences of the spontaneous deamination of methylated CpG sites, rather than the cause of the established methylation levels. The methylation and unmethylation signals of various deuterostomes are yet to be identified in DNA methylation analyses. They are eagerly awaited. However, by using a new computational methodology, our study has shed light on the unexplored molecular evolution of epigenetics.
Genomic sequences of deuterostomes
We downloaded the genomic sequence data from the UCSC ftp site ftp://hgdownload.cse.ucsc.edu/. The names of the assemblies used in the analysis are strPur2 for Strongylocentrotus purpuratus, braFlo1 for Branchiostoma floridae, ci2 for Ciona intestinalis, petMar1 for Petromyzon marinus, tetNig1 for Tetraodon nigroviridis, fr2 for Takifugu rubripes, gasAcu1 for Gasterosteus aculeatus, oryLat2 for Oryzias latipes, danRer4 for Danio rerio, xenTro2 for Xenopus tropicalis, anoCar1 for Anolis carolinensis, taeGut1 for Taeniopygia guttata, galGal3 for Gallus gallus, ornAna1 for Ornithorhynchus anatinus, monDom4 for Monodelphis domestica, bosTau4 for Bos Taurus, equCab2 for Equus caballus, canFam2 for Canis familiaris, felCat3 for Felis catus, cavPor3 for Cavia porcellus, rn4 for Rattus norvegicus, mm8 for Mus musculus, calJac1 for Callithrix jacchus, rheMac2 for Macaca mulatta, ponAbe2 for Pongo pygmaeus, panTro2 for Pan troglodytes, and hg18 for Homo sapiens. The ratio of the observed number to the expected number of CpG, called the "CpG score" in this paper, was calculated using 2-kb fixed lengths, with no space between two adjacent windows . If a window contained more than 50 undetermined or ambiguous nucleotides, it was discarded. A large number of 2-kb sequences spanning gaps were also discarded under this criterion. Most of the analyses in this study were performed using Perl scripts, which are available upon request.
Compulsory decomposition of the bimodal CpG-score distributions
The compulsory decomposition of the bimodal CpG-score distributions was performed with the NOCOM software . Our software, designed to calculate the extent of the global methylation pattern, also includes the software that was coded in FORTRAN 77. We modified its interface so that it can be called from our Perl scripts of both stand-alone and CGI versions.
Occupancy ratios of CpG islands in the whole genomes
We downloaded the CpG island annotation data from ftp://hgdownload.cse.ucsc.edu/apache/htdocs/goldenPath/hg18/database/cpgIslandExt.txt.gz and ftp://hgdownload.cse.ucsc.edu/apache/htdocs/goldenPath/mm8/database/cpgIslandExt.txt.gz and summed all the lengths. We then divided these by the whole lengths of hg18 (3,107,677,273 bp) and mm8 (2,664,455,088 bp), respectively.
DNA methylation data for the human and ascidian genomes
We used the DNA methylation data of Human Epigenome Project Release 26th June 2006 [19, 20]. We used sliding 2-kb windows without any overlap, to avoid including gapped regions. We considered windows with average DNA methylation levels 70%-100% and 0%-30% as hypermethylated and hypomethylated regions, respectively. For the ascidian genome, we used data obtained with methylation-sensitive PCR for two 1-Mb genomic regions . The genomic coordinates of the 2-kb windows and their methylation status are tabulated (Additional file 1: Table S1).
Estimation of the extent of the global DNA methylation pattern
We used the barycenter of the two decomposed normal distributions as the boundary of the methylated and unmethylated regions. In this way, every sliding window of 2 kb, moved in steps of 100 bp, was assigned to one of the two regions, producing binary data for each 100 bp covering the whole genomes. To smooth the binary data, a 2-kb sliding window analysis was performed again and the methylation status was decided based on whether methylation was dominant over unmethylation or vice versa. Consecutive statuses were clustered into a methylated or unmethylated region and the whole set of the lengths was used for the final statistical analysis to estimate the extent of global methylation pattern.
We thank Dr. Hideo Iwasaki (Waseda University), Dr. Kei Yura (Ochanomizu University), Dr. Riu Yamashita and other members of Dr. Nakai laboratory for helpful discussions. We are grateful to Dr. Jurg Ott (Beijing Institute of Genomics) for providing the source code for the NOCOM software. Computation time was provided by the super computer system at the Human Genome Centre, Institute of Medical Science, University of Tokyo. This work was supported by KAKENHI (20870008), the Global COE Program (Centre of Education and Research for Advanced Genome-Based Medicine), MEXT, and Institute for Bioinformatics Research and Development (BIRD), Japan Science and Technology Agency (JST).
This article has been published as part of BMC Bioinformatics Volume 11 Supplement 7, 2010: Ninth International Conference on Bioinformatics (InCoB2010): Bioinformatics. The full contents of the supplement are available online at http://www.biomedcentral.com/1471-2105/11?issue=S7.
- Bestor TH: The DNA methyltransferases of mammals. Hum Mol Genet 2000, 9: 2395–2402. 10.1093/hmg/9.16.2395View ArticlePubMedGoogle Scholar
- Ponger L, Li W: Evolutionary diversification of DNA methyltransferases in eukaryotic genomes. Mol Biol Evol 2005, 22: 1119–1128. 10.1093/molbev/msi098View ArticlePubMedGoogle Scholar
- Bird AP, Taggart MH, Smith BA: Methylated and unmethylated DNA compartments in the sea urchin genome. Cell 1979, 17: 889–901. 10.1016/0092-8674(79)90329-5View ArticlePubMedGoogle Scholar
- Bird AP: CpG island as gene markers in the vertebrate nucleus. Trends Genet 1987, 3: 342–347. 10.1016/0168-9525(87)90294-0View ArticleGoogle Scholar
- Tweedie S, Charlton J, Clark V, Bird A: Methylation of genomes and genes at the invertebrate-vertebrate boundary. Mol Cell Biol 1997, 17: 1469–1475.PubMed CentralPubMedGoogle Scholar
- Ehrlich M, et al.: Amount and distribution of 5-methylcytosine in human DNA from different types of tissues of cells. Nucleic Acids Res 1982, 10: 2709–2721. 10.1093/nar/10.8.2709PubMed CentralView ArticlePubMedGoogle Scholar
- Simmen MW, et al.: Nonmethylated transposable elements and methylated genes in a chordate genome. Science 1999, 283: 1164–1167. 10.1126/science.283.5405.1164View ArticlePubMedGoogle Scholar
- Hendrich B, Tweedie S: The methyl-CpG binding domain and the evolving role of DNA methylation in animals. Trends Genet 2003, 19:, 269–277. 10.1016/S0168-9525(03)00080-5View ArticlePubMedGoogle Scholar
- Suzuki MM, Kerr ARW, De Sousa D, Bird A: CpG methylation is targeted to transcription units in an invertebrate genome. Genome Res 2007, 17: 625–631. 10.1101/gr.6163007PubMed CentralView ArticlePubMedGoogle Scholar
- Barker D, Schafer M, White R: Restriction sites containing CpG show a higher frequency of polymorphism in human DNA. Cell 1984, 36: 131–138. 10.1016/0092-8674(84)90081-3View ArticlePubMedGoogle Scholar
- Weber M, et al.: Distribution, silencing potential and evolutionary impact of promoter DNA methylation in the human genome. Nat Genet 2007, 39: 457–466. 10.1038/ng1990View ArticlePubMedGoogle Scholar
- Reik W: Stability and flexibility of epigenetic gene regulation in mammalian development. Nature 2007, 447: 425–432. 10.1038/nature05918View ArticlePubMedGoogle Scholar
- Takeshima H, Yamashita S, Shimazu T, Niwa T, Ushijima T: The presence of RNA polymerase II, active or stalled, predicts epigenetic fate of promoter CpG islands. Genome Res 2009, 19: 1974–1982. 10.1101/gr.093310.109PubMed CentralView ArticlePubMedGoogle Scholar
- Strichman-Almashanu LZ, et al.: A genome-wide screen for normally methylated human CpG islands that can identify novel imprinted genes. Genome Res 2002, 12: 543–554.PubMed CentralView ArticlePubMedGoogle Scholar
- Meissner A, et al.: Genome-scale DNA methylation maps of pluripotent and differentiated cells. Nature 2008, 454: 766–770.PubMed CentralPubMedGoogle Scholar
- Straussman R, et al.: Developmental programming of CpG island methylation profiles in the human genome. Nat Struct Mol Biol 2009, 16: 564–571. 10.1038/nsmb.1594View ArticlePubMedGoogle Scholar
- Elango N, Yi SV: DNA methylation and structural and functional bimodality of vertebrate promoters. Mol Biol Evol 2008, 25: 1602–1608. 10.1093/molbev/msn110View ArticlePubMedGoogle Scholar
- Hasselblad V: Estimation of parameters for a mixture of normal distributions. Technometrics 1966, 8: 431–444. 10.2307/1266689View ArticleGoogle Scholar
- Rakyan VK, et al.: DNA methylation profiling of the human major histocompatibility complex: a pilot study for the human epigenome project. PLoS Biol 2004, 2: e405. 10.1371/journal.pbio.0020405PubMed CentralView ArticlePubMedGoogle Scholar
- Eckhardt F, et al.: DNA methylation profiling of human chromosomes 6, 20 and 22. Nat Genet 2006, 38: 1378–1385. 10.1038/ng1909PubMed CentralView ArticlePubMedGoogle Scholar
- Sodergren E, et al.: The genome of the sea urchin Strongylocentrotus purpuratus . Science 2006, 314: 941–952. 10.1126/science.1133609View ArticlePubMedGoogle Scholar
- Yamashita R, Suzuki Y, Sugano S, Nakai K: Genome-wide analysis reveals strong correlation between CpG islands with nearby transcription start sites of genes and their tissue specificity. Gene 2005, 350: 129–136. 10.1016/j.gene.2005.01.012View ArticlePubMedGoogle Scholar
- Saxonov S, Berg P, Brutlag DL: A genome-wide analysis of CpG dinucleotides in the human genome distinguishes two distinct classes of promoters. Proc Natl Acad Sci USA 2006, 103: 1412–1417. 10.1073/pnas.0510310103PubMed CentralView ArticlePubMedGoogle Scholar
- Jabbari K, Cacciò S, Païs de Barros JP, Desgrès J, Bernardi G: Evolutionary changes in CpG and methylation levels in the genome of vertebrates. Gene 1997, 205: 109–118. 10.1016/S0378-1119(97)00475-7View ArticlePubMedGoogle Scholar
- Satoh N: The ascidian tadpole larva: comparative molecular development and genomics. Nat Rev Genet 2003, 4: 285–295. 10.1038/nrg1042View ArticlePubMedGoogle Scholar
- Delsuc F, Brinkmann H, Philippe H: Tunicates and not cephalochordates are the closest living relatives of vertebrates. Nature 2006, 439: 965–968. 10.1038/nature04336View ArticlePubMedGoogle Scholar
- Putnam NH, et al.: The amphioxus genome and the evolution of the chordate karyotype. Nature 2008, 453: 1064–1071. 10.1038/nature06967View ArticlePubMedGoogle Scholar
- Kasahara M, et al.: The medaka draft genome and insights into vertebrate genome evolution. Nature 2007, 447: 714–719. 10.1038/nature05846View ArticlePubMedGoogle Scholar
- Han L, Zhao Z: Comparative analysis of CpG islands in four fish genomes. Comp Funct Genomics 2008, 565631.Google Scholar
- Venkatesh B, Gilligan P, Brenner S: Fugu: a compact vertebrate reference genome. FEBS Lett 2000, 476: 3–7. 10.1016/S0014-5793(00)01659-8View ArticlePubMedGoogle Scholar
- Hedges SB, Kumar S: Genomics. Vertebrate genomes compared. Science 2002, 297: 1283–1285. 10.1126/science.1076231View ArticlePubMedGoogle Scholar
- Warren WC, et al.: Genome analysis of the platypus reveals unique signatures of evolution. Nature 2008, 453: 175–183. 10.1038/nature06936PubMed CentralView ArticlePubMedGoogle Scholar
- Pontius JU, et al.: Initial sequence and comparative analysis of the cat genome. Genome Res 2007, 17: 1675–1689. 10.1101/gr.6380007PubMed CentralView ArticlePubMedGoogle Scholar
- Haussler , et al.: Genome 10K: a proposal to obtain whole-genome sequence for 10 000 vertebrate species. J Hered 2009, 100: 659–674. 10.1093/jhered/esp086View ArticleGoogle Scholar
- Cohen NM, et al.: DNA methylation programming and reprogramming in primate embryonic stem cells. Genome Res 2009, 2193–2201. 10.1101/gr.096685.109Google Scholar
- Gardiner-Garden M, Frommer M: CpG islands in vertebrate genomes. J Mol Biol 1987, 196: 261–282. 10.1016/0022-2836(87)90689-9View ArticlePubMedGoogle Scholar
- Ott J: Detection of rare major genes in lipid levels. Hum Genet 1979, 51: 79–91. 10.1007/BF00278296View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.