Linking the epigenome to the genome: correlation of different features to DNA methylation of CpG islands

DNA methylation of CpG islands plays a crucial role in the regulation of gene expression. More than half of all human promoters contain CpG islands with a tissue-specific methylation pattern in differentiated cells. Still today, the whole process of how DNA methyltransferases determine which region should be methylated is not completely revealed. There are many hypotheses of which genomic features are correlated to the epigenome that have not yet been evaluated. Furthermore, many explorative approaches of measuring DNA methylation are limited to a subset of the genome and thus, cannot be employed, e.g., for genome-wide biomarker prediction methods. In this study, we evaluated the correlation of genetic, epigenetic and hypothesis-driven features to DNA methylation of CpG islands. To this end, various binary classifiers were trained and evaluated by cross-validation on a dataset comprising DNA methylation data for 190 CpG islands in HEPG2, HEK293, fibroblasts and leukocytes. We achieved an accuracy of up to 91% with an MCC of 0.8 using ten-fold cross-validation and ten repetitions. With these models, we extended the existing dataset to the whole genome and thus, predicted the methylation landscape for the given cell types. The method used for these predictions is also validated on another external whole-genome dataset. Our results reveal features correlated to DNA methylation and confirm or disprove various hypotheses of DNA methylation related features. This study confirms correlations between DNA methylation and histone modifications, DNA structure, DNA sequence, genomic attributes and CpG island properties. Furthermore, the method has been validated on a genome-wide dataset from the ENCODE consortium. The developed software, as well as the predicted datasets and a web-service to compare methylation states of CpG islands are available at http://www.cogsys.cs.uni-tuebingen.de/software/dna-methylation/.     

A sensitive new method for rapid detection of abnormal methylation patterns in global DNA and within CpG islands.

To assess alterations in DNA methylation density in both global DNA and within CpG islands, we have developed a simple method based on the use of methylation-sensitive restriction endonucleases that leave a 5' guanine overhang after DNA cleavage, with subsequent single nucleotide extension with radiolabeled [(3)H]dCTP. The methylation-sensitive restriction enzymes HpaII and AciI have relatively frequent recognition sequences at CpG sites that occur randomly throughout the genome. BssHII is a methylation sensitive enzyme that similarly leaves a guanine overhang, but the recognition sequence is nonrandom and occurs predominantly at unmethylated CpG sites within CpG islands. The selective use of these enzymes can be used to screen for alterations in genome-wide methylation and CpG island methylation status, respectively. The extent of [(3)H]dCTP incorporation opposite the exposed guanine after restriction enzyme treatment is directly proportional to the number of unmethylated (cleaved) CpG sites. The "cytosine-extension assay" has several advantages over existing methods because (a) radiolabel incorporation is independent of the integrity of the DNA, (b) methylation detection does not require PCR amplification or DNA methylase reactions, and (c) it is applicable to ng quantities of DNA. Using DNA extracted from normal human liver and from human hepatocellular carcinoma, the applicability of the assay is demonstrated by the detection of an increase in genome-wide hypomethylation and CpG island hypermethylation in the tumor DNA.     

DNA motifs associated with aberrant CpG island methylation

Epigenetic silencing involving the aberrant methylation of promoter region CpG islands is widely recognized as a tumor suppressor silencing mechanism in cancer. However, the molecular pathways underlying aberrant DNA methylation remain elusive. Recently we showed that, on a genome-wide level, CpG island loci differ in their intrinsic susceptibility to aberrant methylation and that this susceptibility can be predicted based on underlying sequence context. These data suggest that there are sequence/structural features that contribute to the protection from or susceptibility to aberrant methylation. Here we use motif elicitation coupled with classification techniques to identify DNA sequence motifs that selectively define methylation-prone or methylation-resistant CpG islands. Motifs common to 28 methylation-prone or 47 methylation-resistant CpG island-containing genomic fragments were determined using the MEME and MAST algorithms (). The five most discriminatory motifs derived from methylation-prone sequences were found to be associated with CpG islands in general and were nonrandomly distributed throughout the genome. In contrast, the eight most discriminatory motifs derived from the methylation-resistant CpG islands were randomly distributed throughout the genome. Interestingly, this latter group tended to associate with Alu and other repetitive sequences. Used together, the frequency of occurrence of these motifs successfully discriminated methylation-prone and methylation-resistant CpG island groups with an accuracy of 87% after 10-fold cross-validation. The motifs identified here are candidate methylation-targeting or methylation-protection DNA sequences.     

De novo methylation of CpG island sequences in human fibroblasts overexpressing DNA (cytosine-5-)-methyltransferase.

Recent studies showing a correlation between the levels of DNA (cytosine-5-)-methyltransferase (DNA MTase) enzyme activity and tumorigenicity have implicated this enzyme in the carcinogenic process. Moreover, hypermethylation of CpG island-containing promoters is associated with the inactivation of genes important to tumor initiation and progression. One proposed role for DNA MTase in tumorigenesis is therefore a direct role in the de novo methylation of these otherwise unmethylated CpG islands. In this study, we sought to determine whether increased levels of DNA MTase could directly affect CpG island methylation. A full-length cDNA for human DNA MTase driven by the cytomegalovirus promoter was constitutively expressed in human fibroblasts. Individual clones derived from cells transfected with DNA MTase (HMT) expressed 1- to 50-fold the level of DNA MTase protein and enzyme activity of the parental cell line or clones transfected with the control vector alone (Neo). To determine the effects of DNA MTase overexpression on CpG island methylation, we examined 12 endogenous CpG island loci in the HMT clones. HMT clones expressing > or = 9-fold the parental levels of DNA MTase activity were significantly hypermethylated relative to at least 11 Neo clones at five CpG island loci. In the HMT clones, methylation reached nearly 100% at susceptible CpG island loci with time in culture. In contrast, there was little change in the methylation status in the Neo clones over the same time frame. Taken together, the data indicate that overexpression of DNA MTase can drive the de novo methylation of susceptible CpG island loci, thus providing support for the idea that DNA MTase can contribute to tumor progression through CpG island methylation-mediated gene inactivation.     

Genome-wide profiling of DNA methylation reveals a class of normally methylated CpG island promoters

The role of CpG island methylation in normal development and cell differentiation is of keen interest, but remains poorly understood. We performed comprehensive DNA methylation profiling of promoter regions in normal peripheral blood by methylated CpG island amplification in combination with microarrays. This technique allowed us to simultaneously determine the methylation status of 6,177 genes, 92% of which include dense CpG islands. Among these 5,549 autosomal genes with dense CpG island promoters, we have identified 4.0% genes that are nearly completely methylated in normal blood, providing another exception to the general rule that CpG island methylation in normal tissue is limited to X inactivation and imprinted genes. We examined seven genes in detail, including ANKRD30A, FLJ40201, INSL6, SOHLH2, FTMT, C12orf12, and DPPA5. Dense promoter CpG island methylation and gene silencing were found in normal tissues studied except testis and sperm. In both tissues, bisulfite cloning and sequencing identified cells carrying unmethylated alleles. Interestingly, hypomethylation of several genes was associated with gene activation in cancer. Furthermore, reactivation of silenced genes could be induced after treatment with a DNA demethylating agent or in a cell line lacking DNMT1 and/or DNMT3b. Sequence analysis identified five motifs significantly enriched in this class of genes, suggesting that cis-regulatory elements may facilitate preferential methylation at these promoter CpG islands. We have identified a group of non-X-linked bona fide promoter CpG islands that are densely methylated in normal somatic tissues, escape methylation in germline cells, and for which DNA methylation is a primary mechanism of tissue-specific gene silencing.     

Methylation-mediated silencing of TMS1/ASC is accompanied by histone hypoacetylation and CpG island-localized changes in chromatin architecture.

Aberrant methylation of CpG-dense islands in the promoter regions of genes is an acquired epigenetic alteration associated with the silencing of tumor suppressor genes in human cancers. In a screen for endogenous targets of methylation-mediated gene silencing, we identified a novel CpG island-associated gene, TMS1, which is aberrantly methylated and silenced in response to the ectopic expression of DNA methyltransferase-1. TMS1 functions in the regulation of apoptosis and is frequently methylated and silenced in human breast cancers. In this study, we characterized the methylation pattern and chromatin architecture of the TMS1 locus in normal fibroblasts and determined the changes associated with its progressive methylation. In normal fibroblasts expressing TMS1, the CpG island is defined by an unmethylated domain that is separated from densely methylated flanking DNA by distinct 5' and 3' boundaries. Analysis of the nucleoprotein architecture of the locus in intact nuclei revealed three DNase I-hypersensitive sites that map within the CpG island. Strikingly, two of these sites coincided with the 5'- and 3'-methylation boundaries. Methylation of the TMS1 CpG island was accompanied by loss of hypersensitive site formation, hypoacetylation of histones H3 and H4, and gene silencing. This altered chromatin structure was confined to the CpG island and occurred without significant changes in methylation, histone acetylation, or hypersensitive site formation at a fourth DNase I-hypersensitive site 2 kb downstream of the TMS1 CpG island. The data indicate that there are sites of protein binding and/or structural transitions that define the boundaries of the unmethylated CpG island in normal cells and that aberrant methylation overcomes these boundaries to direct a local change in chromatin structure, resulting in gene silencing.     

A human CpG island randomly inserted into a plant genome is protected from methylation

In vertebrate genomes the dinucleotide CpG is heavily methylated, except in CpG islands, which are normally unmethylated. It is not clear why the CpG islands are such poor substrates for DNA methyltransferase. Plant genomes display methylation, but otherwise the genomes of plants and animals represent two very divergent evolutionary lines. To gain a further understanding of the resistance of CpG islands to methylation, we introduced a human CpG island from the proteasome-like subunit I gene into the genome of the plant Arabidopsis thaliana. Our results show that prevention of methylation is an intrinsic property of CpG islands, recognized even if a human CpG island is transferred to a plant genome. Two different parts of the human CpG island – the promoter region/ first exon and exon2–4 – both displayed resistance against methylation, but the promoter/ exon1 construct seemed to be most resistant. In contrast, certain sites in a plant CpG-rich region used as a control transgene were always methylated. The frequency of silencing of the adjacent nptII (Km^R) gene in the human CpG constructs was lower than observed for the plant CpG-rich region. These results have implications for understanding DNA methylation, and for construction of vectors that will reduce transgene silencing.     

Histone methylation marks play important roles in predicting the methylation status of CpG islands

The methylation status of CpG islands is highly correlated with gene expression. Current methods for computational prediction of DNA methylation only utilize DNA sequence features. In this study, besides 35 DNA sequence features, we added four histone methylation marks to predict the methylation status of CpG islands, and improved the accuracy to 89.94%. Also we applied our model to predict the methylation pattern of all the CpG islands in the human genome, and the results are consistent with the previous reports. Our results imply the important roles of histone methylation marks in affecting the methylation status of CpG islands. H3K4me enriched in the methylation-resistant CpG islands could disrupt the contacts between nucleosomes, unravel chromatin and make DNA sequences accessible. And the established open environment may be a prerequisite for or a consequence of the function implementation of zinc finger proteins that could protect CpG islands from DNA methylation.     

The hypervariable DXS255 locus contains a LINE-1 repetitive element with a CpG island that is extensively methylated only on the active X chromosome.

The DXS255 locus at Xp11.22 is highly polymorphic due to a 26-bp variable number of tandem repeats (VNTR) motif. In previous studies, one of the MspI sites flanking the VNTR manifested a correlation between methylation and X chromosome inactivation. Here we show, by DNA sequence analysis, that this MspI site is located within the CpG island at the 5' end of a LINE-1 element, which is 2.5 kb from the VNTR. The methylation status of the CpG island was assessed in Southern blotting experiments using the methylation-sensitive enzymes HpaII, HhaI, and BssHII. All these sites were completely methylated on active X chromosomes, consistent with previously reported findings of full methylation of LINE-1 elements throughout the genome. However, on inactive X chromosomes these sites were predominantly unmethylated, although patterns were found to be heterogeneous. The results suggest that LINE-1 elements on the inactive X chromosome are not suppressed by full methylation of their CpG islands. The differential methylation of the DXS255 CpG island provides the basis for a highly informative X inactivation analysis system.     

Alu repeated DNAs are differentially methylated in primate germ cells.

A significant fraction of Alu repeats in human sperm DNA, previously found to be unmethylated, is nearly completely methylated in DNA from many somatic tissues. A similar fraction of unmethylated Alus is observed here in sperm DNA from rhesus monkey. However, Alus are almost completely methylated at the restriction sites tested in monkey follicular oocyte DNA. The Alu methylation patterns in mature male and female monkey germ cells are consistent with Alu methylation in human germ cell tumors. Alu sequences are hypomethylated in seminoma DNAs and more methylated in a human ovarian dysgerminoma. These results contrast with methylation patterns reported for germ cell single-copy, CpG island, satellite, and L1 sequences. The function of Alu repeats is not known, but differential methylation of Alu repeats in the male and female germ lines suggests that they may serve as markers for genomic imprinting or in maintaining differences in male and female meiosis.     

Distribution of DNA methylation,CpGs, and CpG islands in human isochores

DNA methylation is a major epigenetic modification of the genome that affects basic biological functions, such as gene expression and cell development. We used the human genome sequences and the DNA methylation data that are available in order to establish a map of the levels of GC and methylation in isochores. We also looked for the correlations that hold between GC levels and the distribution of the (1) dinucleotide CpG, (2) ratio 5mC/CpG, and (3) CpG islands. Our results show that methylation levels, CpG frequencies, and the density of CpG islands are positively correlated with the GC level of isochores. In contrast, the correlation between the 5mC/CpG ratio and GC is a negative one because the increase in methylation lags behind that of CpG, to reach a plateau in the GC-richest, gene-richest isochore families H2 and H3. In conclusion, there are more CpG targets that remain unmethylated in the GC-richest, gene-richest isochores in comparison with the other isochores. This conclusion supports the idea that the widespread methylation under consideration here has a general inhibitory effect on gene expression.     

The Undermethylated State of a CpG Island Region in Igf2 Transgenes Is Dependent on the H19 Enhancers

CpG islands are GC-rich regions located in the promoter regions of housekeeping genes and many tissue-specific genes. While most CpG islands are normally unmethylated, island methylation can occur and is associated with silencing of the corresponding gene. Experiments with transgenic mice and DNA transfection in pluripotential embryonic cells have led to the conclusion that the information required for protecting the islands from methylation is contained within the CpG islands themselves and have identified Sp1 binding sites as an important element in establishing and/or maintaining the methylation-free state of CpG islands. To examine the generality of these observations, we analyzed the methylation of one of the mouse Igf2 CpG islands and its flanks in transgenic mice. We observed that the undermethylated state of this region is dependent on the presence of a separate cis-regulatory element, the H19 enhancers. These tissue-specific enhancers had a ubiquitous, non-tissue-specific effect on island region methylation. Structural alterations outside of the island and these enhancers also affected this region's methylation. These findings indicate that the methylation of some CpG island-containing regions is more sensitive than previously believed to the activity of distant cis-regulatory elements and to structural alterations in nonisland sequences in cis.     

Tissue specific DNA methylation of CpG islands in normal human adult somatic tissues distinguishes neural from non-neural tissues

Although most CpG islands are generally thought to remain unmethylated in all adult somatic tissues, recent genome-wide approaches have found that some CpG islands have distinct methylation patterns in various tissues, with most differences being seen between germ cells and somatic tissues. Few studies have addressed this among human somatic tissues and fewer still have studied the same sets of tissues from multiple individuals. In the current study, we used Restriction Landmark Genomic Scanning to study tissue specific methylation patterns in a set of 12 human tissues collected from multiple individuals. We identified 34 differentially methylated CpG islands among these tissues, many of which showed consistent patterns in multiple individuals. Of particular interest were striking differences in CpG island methylation, not only among brain regions, but also between white and grey matter of the same region. These findings were confirmed for selected loci by quantitative bisulfite sequencing. Cluster analysis of the RLGS data indicated that several tissues clustered together, but the strongest clustering was in brain. Tissues from different brain regions clustered together, and, as a group, brain tissues were distinct from either mesoderm or endoderm derived tissues which demonstrated limited clustering. These data demonstrate consistent tissue specific methylation for certain CpG islands, with clear differences between white and grey matter of the brain. Furthermore, there was an overall pattern of tissue specifically methylated CpG islands that distinguished neural tissues from non-neural.Key words: Tissue specific methylation, CpG island methylation, neural, brain tissue, grey matter, white matter     

Genome-wide analysis of DNA methylation status of CpG islands in embryoid bodies, teratomas, and fetuses

Differentiation of embryonic stem (ES) cells into embryoid bodies (EBs) provides an in vitro system for the study of early lineage determination during mammalian development. We have previously reported that there are 247 CpG islands that potentially have tissue-dependent and differentially methylated regions (T-DMRs). This provided evidence that the formation of DNA methylation patterns at CpG islands is a crucial epigenetic event underlying mammalian development. Here we present an analysis by the restriction landmark genomic scanning (RLGS) using NotI as a landmark enzyme of the genome-wide methylation status of CpG islands of ES cells and EBs and of teratomas produced from ES cells. These results are considered in relation to the methylation status of CpG islands of genomic DNA from normal fetus (10.5 dpc) and adult tissues. We have prepared a DNA methylation panel that consists of 259 T-DMRs and includes novel T-DMRs that are distinctly methylated or unmethylated in the teratomas. The DNA methylation pattern was complex and differed for the ES cells, EBs, and teratomas, providing evidence that differentiation of cells involves both de novo DNA methylation as well as demethylation. Comparison of the numbers of T-DMRs, that were differentially methylated or unmethylated among the cells and tissue types studied, revealed that the teratomas were the most epigenetically different from ES cells. Thus, analysis of the DNA methylation profiles prepared in this study provides new insights into the differentiation of ES cells and development of fetus, EB, teratoma, and somatic tissues.