CpG island density and its correlations with genomic features in mammalian genomes

Background  

CpG islands, which are clusters of CpG dinucleotides in GC-rich regions, are considered gene markers and represent an important feature of mammalian genomes. Previous studies of CpG islands have largely been on specific loci or within one genome. To date, there seems to be no comparative analysis of CpG islands and their density at the DNA sequence level among mammalian genomes and of their correlations with other genome features.     

Gene-associated CpG islands in plants as revealed by analyses of genomic sequences

We screened plant genome sequences, primarily from rice and Arabidopsis thaliana, for CpG islands, and identified DNA segments rich in CpG dinucleotides within these sequences. These CpG-rich clusters appeared in the analysed sequences as discrete peaks and occurred at the frequencies of one per 4.7 kb in rice and one per 4.0 kb in A. thaliana. In rice and A. thaliana, most of the CpG-rich clusters were associated with genes, which suggests that these clusters are useful landmarks in genome sequences for identifying genes in plants with small genomes. In contrast, in plants with larger genomes, only a few of the clusters were associated with genes. These plant CpG-rich clusters satisfied the criteria used for identifying human CpG islands, which suggests that these CpG clusters may be regarded as plant CpG islands. The position of each island relative to the 5'-end of its associated gene varied considerably. Genes in the analysed sequences were grouped into five classes according to the position of the CpG islands within their associated genes. A large proportion of the genes belonged to one of two classes, in which a CpG island occurred near the 5'-end of the gene or covered the whole gene region. The position of a plant CpG island within its associated gene appeared to be related to the extent of tissue-specific expression of the gene; the CpG islands of most of the widely expressed rice genes occurred near the 5'-end of the genes.     

A species-generalized probabilistic model-based definition of CpG islands

The DNA of most vertebrates is depleted in CpG dinucleotides, the target for DNA methylation. The remaining CpGs tend to cluster in regions referred to as CpG islands (CGI). CGI have been useful as marking functionally relevant epigenetic loci for genome studies. For example, CGI are enriched in the promoters of vertebrate genes and thought to play an important role in regulation. Currently, CGI are defined algorithmically as an observed-to-expected ratio (O/E) of CpG greater than 0.6, G+C content greater than 0.5, and usually but not necessarily greater than a certain length. Here we find that the current definition leaves out important CpG clusters associated with epigenetic marks, relevant to development and disease, and does not apply at all to nonvertabrate genomes. We propose an alternative Hidden Markov model-based approach that solves these problems. We fit our model to genomes from 30 species, and the results support a new epigenomic view toward the development of DNA methylation in species diversity and evolution. The O/E of CpG in islands and nonislands segregated closely phylogenetically and showed substantial loss in both groups in animals of greater complexity, while maintaining a nearly constant difference in CpG O/E between islands and nonisland compartments. Lists of CGI for some species are available at http://www.rafalab.org.     

CpG islands: Algorithms and applications in methylation studies

Methylation occurs frequently at 5’-cytosine of the CpG dinucleotides in vertebrate genomes; however, this epigenetic feature is rarely observed in CpG islands (CGIs) or CpG clusters in the promoter regions of genes. Aberrant methylation of the promoter-associated CGIs might influence gene expression and cause carcinogenesis. Because of the functional importance, multiple algorithms have been available for identifying CGIs in a genome or a sequence. They can be categorized into the traditional algorithms (e.g., Gardiner-Garden and Frommer (1987), Takai and Jones (2002), and CpGPRoD (2002)) or statistical property based algorithms (CpGcluster (2006) and CG cluster (2007)). We reviewed the features of these algorithms and evaluated their performance on identifying functional CGIs using genome-wide methylation data. Moreover, identification of CGIs is an initial step in many recent studies for predicting methylation status as well as in the design of methylation detection platforms. We reviewed the benchmarks and features used in these studies.     

Repertoires of the Nucleosome-Positioning Dinucleotides

It is generally accepted that the organization of eukaryotic DNA into chromatin is strongly governed by a code inherent in the genomic DNA sequence. This code, as well as other codes, is superposed on the triplets coding for amino acids. The history of the chromatin code started three decades ago with the discovery of the periodic appearance of certain dinucleotides, with AA/TT and RR/YY giving the strongest signals, all with a period of 10.4 bases. Every base-pair stack in the DNA duplex has specific deformation properties, thus favoring DNA bending in a specific direction. The appearance of the corresponding dinucleotide at the distance 10.4 xn bases will facilitate DNA bending in that direction, which corresponds to the minimum energy of DNA folding in the nucleosome. We have analyzed the periodic appearances of all 16 dinucleotides in the genomes of thirteen different eukaryotic organisms. Our data show that a large variety of dinucleotides (if not all) are, apparently, contributing to the nucleosome positioning code. The choice of the periodical dinucleotides differs considerably from one organism to another. Among other 10.4 base periodicities, a strong and very regular 10.4 base signal was observed for CG dinucleotides in the genome of the honey bee A. mellifera. Also, the dinucleotide CG appears as the only periodical component in the human genome. This observation seems especially relevant since CpG methylation is well known to modulate chromatin packing and regularity. Thus, the selection of the dinucleotides contributing to the chromatin code is species specific, and may differ from region to region, depending on the sequence context.     

Sequence context analysis of 8.2 million single nucleotide polymorphisms in the human genome

We analyzed n-mers (n=3-8) in the local environment of 8,249,446 human SNPs and compared their distribution with that in the genome reference sequences. The results revealed that the short sequences, which contained at least one CpG dinucleotide, occurred more frequently in the local SNP sequences than in the genome sequences. To exclude the hypermutability effect of the methylated CpG dinucleotides on the sequence context of SNPs, we examined the distribution patterns for each of the six categories of substitution. We observed the similar pattern (i.e., CpG-containing n-mers vs. non-CpG-containing n-mers) in SNP categories A/G, C/T and C/G but the opposite pattern in category A/T. We next identified 34,928 putative CpG islands in the human genome and located 133,591 SNPs within these islands. In the CpG islands, CpG SNPs were 3.92-fold less prevalent relative to the presence of CpG dinucleotides. Conversely, in the human genome, the frequency of CpG dinucleotides at the polymorphic sites was 6.09 times that in the genome reference sequences. These results support the previous views of mutational suppression at the CpG sites in the CpG islands and hypermutability of the methylated CpG dinucleotides that are prevalent in the non-CpG island sequences in the human genome. Our study represents a comprehensive investigation of the sequence context of SNPs in the human genome and in human CpG islands.     

Segmental distribution of genes harboring a CpG island-like region on rice chromosomes

In plant genomes, there exist discrete regions rich in CpG dinucleotides, namely CpG clusters. In rice, most of these CpG clusters are associated with genes. Rice genes are grouped into one of the five classes according to the position of an associated CpG cluster. Among them, class 1 genes, which harbor a CpG cluster at the 5′-terminus, share similarities with human genes having CpG islands. In the present study, by analyzing plant genome sequence data, primarily from rice, we investigated the chromosomal distribution of genes of each class, mainly class 1 genes. Class 1 genes were not uniformly distributed across the rice genome, but were clustered into discrete chromosomal segments. EST-based analysis of the distribution of expressed genes indicates that this segmental distribution of class 1 genes caused a preferential distribution of expressed genes within class 1 gene-rich segments. We then compared the methylation status of genes of each class to examine the possibility that differential DNA methylation, if any, is relevant to the observed differential expression level of genes inside and outside the class 1 segments. The difference in the methylation level between these genes was revealed to be fairly small, which does not support the above-mentioned possibility. Electronic supplementary material Supplementary material is available in the online version of this article at and is accessible for authorized users.     

The tendency to recreate ancestral CG dinucleotides in the human genome

Background  

The CG dinucleotides are known to be deficient in the human genome, due to a high mutation rate from 5-methylated CG to TG and its complementary pair CA. Meanwhile, many cellular functions rely on these CG dinucleotides, such as gene expression controlled by cytosine methylation status. Thus, CG dinucleotides that provide essential functional substrates should be retained in genomes. How these two conflicting processes regarding the fate of CG dinucleotides - i.e., high mutation rate destroying CG dinucleotides, vs. functional processes that require their preservation remains an unsolved question.     

Human nucleosomes: special role of CG dinucleotides and Alu-nucleosomes

Background

The periodical occurrence of dinucleotides with a period of 10.4 bases now is undeniably a hallmark of nucleosome positioning. Whereas many eukaryotic genomes contain visible and even strong signals for periodic distribution of dinucleotides, the human genome is rather featureless in this respect. The exact sequence features in the human genome that govern the nucleosome positioning remain largely unknown.

Results

When analyzing the human genome sequence with the positional autocorrelation method, we found that only the dinucleotide CG shows the 10.4 base periodicity, which is indicative of the presence of nucleosomes. There is a high occurrence of CG dinucleotides that are either 31 (10.4 × 3) or 62 (10.4 × 6) base pairs apart from one another - a sequence bias known to be characteristic of Alu-sequences. In a similar analysis with repetitive sequences removed, peaks of repeating CG motifs can be seen at positions 10, 21 and 31, the nearest integers of multiples of 10.4.

Conclusions

Although the CG dinucleotides are dominant, other elements of the standard nucleosome positioning pattern are present in the human genome as well. The positional autocorrelation analysis of the human genome demonstrates that the CG dinucleotide is, indeed, one visible element of the human nucleosome positioning pattern, which appears both in Alu sequences and in sequences without repeats. The dominant role that CG dinucleotides play in organizing human chromatin is to indicate the involvement of human nucleosomes in tuning the regulation of gene expression and chromatin structure, which is very likely due to cytosine-methylation/-demethylation in CG dinucleotides contained in the human nucleosomes. This is further confirmed by the positions of CG-periodical nucleosomes on Alu sequences. Alu repeats appear as monomers, dimers and trimers, harboring two to six nucleosomes in a run. Considering the exceptional role CG dinucleotides play in the nucleosome positioning, we hypothesize that Alu-nucleosomes, especially, those that form tightly positioned runs, could serve as "anchors" in organizing the chromatin in human cells.     

CpG islands or CpG clusters: how to identify functional GC-rich regions in a genome?

Background  

CpG islands (CGIs), clusters of CpG dinucleotides in GC-rich regions, are often located in the 5' end of genes and considered gene markers. Hackenberg et al. (2006) recently developed a new algorithm, CpGcluster, which uses a completely different mathematical approach from previous traditional algorithms. Their evaluation suggests that CpGcluster provides a much more efficient approach to detecting functional clusters or islands of CpGs.     

Sequence context at human single nucleotide polymorphisms: overrepresentation of CpG dinucleotide at polymorphic sites and suppression of variation in CpG islands

Human polymorphisms originate as mutations, and the influence of context on mutagenesis should be reflected in the distribution of sequences surrounding single nucleotide polymorphisms (SNPs). We have performed a computational survey of nearly two million human SNPs to determine if sequence-dependent hotspots for polymorphism exist in the human genome. Here we show that sequences containing CpG dinucleotides, which occur at low frequencies in the human genome, are 6.7-fold more abundant at polymorphic sites than expected. In contrast, polymorphisms in CpG sequences located within CpG islands, important regulatory regions that modulate gene expression, are 6.8-fold less prevalent than expected. The distribution of polymorphic alleles at CpGs in CpG islands is also significantly different from that in non-island regions. These data strongly support a role for 5-methylcytosine deamination in the generation of human variation, and suggest that variation at CpGs in islands is suppressed.     

DNA Sequence Explains Seemingly Disordered Methylation Levels in Partially Methylated Domains of Mammalian Genomes

For the most part metazoan genomes are highly methylated and harbor only small regions with low or absent methylation. In contrast, partially methylated domains (PMDs), recently discovered in a variety of cell lines and tissues, do not fit this paradigm as they show partial methylation for large portions (20%–40%) of the genome. While in PMDs methylation levels are reduced on average, we found that at single CpG resolution, they show extensive variability along the genome outside of CpG islands and DNase I hypersensitive sites (DHS). Methylation levels range from 0% to 100% in a roughly uniform fashion with only little similarity between neighboring CpGs. A comparison of various PMD-containing methylomes showed that these seemingly disordered states of methylation are strongly conserved across cell types for virtually every PMD. Comparative sequence analysis suggests that DNA sequence is a major determinant of these methylation states. This is further substantiated by a purely sequence based model which can predict 31% (R²) of the variation in methylation. The model revealed CpG density as the main driving feature promoting methylation, opposite to what has been shown for CpG islands, followed by various dinucleotides immediately flanking the CpG and a minor contribution from sequence preferences reflecting nucleosome positioning. Taken together we provide a reinterpretation for the nucleotide-specific methylation levels observed in PMDs, demonstrate their conservation across tissues and suggest that they are mainly determined by specific DNA sequence features.     

Alu element mutation spectra: molecular clocks and the effect of DNA methylation

In primate genomes more than 40% of CpG islands are found within repetitive elements. With more than one million copies in the human genome, the Alu family of retrotransposons represents the most successful short interspersed element (SINE) in primates and CpG dinucleotides make up about 20% of Alu sequences. It is generally thought that CpG dinucleotides mutate approximately ten times faster than other dinucleotides due to cytosine methylation and the subsequent deamination and conversion of C-->T. However, the disparity of Alu subfamily age estimations based upon CpG or non-CpG substitution density indicates a more complex relationship between CpG and non-CpG substitutions within the Alu elements. Here we report an analysis of the mutation patterns for 5296 Alu elements comprising 20 subfamilies. Our results indicate a relatively constant CpG versus non-CpG substitution ratio of approximately 6 for the young (AluY) and intermediate (AluS) Alu subfamilies. However, a more complex non-linear relationship between CpG and non-CpG substitutions was observed when old (AluJ) subfamilies were included in the analysis. These patterns may be the result of the slowdown of the neutral mutation rate during primate evolution and/or an increase in the CpG mutation rate as the consequence of increased DNA methylation in response to a burst of retrotransposition activity approximately 35 million years ago.     

CpG islands, genes and isochores in the genomes of vertebrates

We have shown that human genes associated with CpG islands increase in number as they increase in % of guanine + cytosine (GC) levels, and that most genes associated with CpG islands are located in the GC-richest compartment of the human genome. This is an independent confirmation of the concentration gradient of CpG islands (detected as HpaII tiny fragments, or HTF) which was demonstrated in the genome of warm-blooded vertebrates [A?ssani and Bernardi, Gene 106 (1991) 173-183]. We then reassessed the location of CpG islands using the data currently available and confirmed that CpG islands are most frequently located in the 5'-flanking sequences of genes and that they overlap genes to variable extents. We have shown that such extents increase with the increasing GC levels of genes, the GC-richest genes being completely included in CpG islands. Under such circumstances, we have investigated the properties of the 'extragenic' CpG islands located in the 5'-flanking segments of homologous genes from both warm- and cold-blooded vertebrates. We have confirmed that, in cold-blooded vertebrates, CpG islands are often absent; when present, they have lower GC and CpG levels; the latter attain, however, statistically expected values. Finally, we have shown that CpG doublets increase with the increasing GC of exons, introns and intergenic sequences (including 'extragenic' CpG islands) in the genomes from both warm- and cold-blooded vertebrates. The correlations found are the same for both classes of vertebrates, and are similar for exons, introns and intergenic sequences (including 'extragenic' CpG islands). The findings just outlined indicate that the origin and evolution of CpG islands in the vertebrate genome are associated with compositional transitions (GC increases) in genes and isochores.     

CpG Dinucleotide Frequencies Reveal the Role of Host Methylation Capabilities in Parvovirus Evolution

Parvoviruses are rapidly evolving viruses that infect a wide range of hosts, including vertebrates and invertebrates. Extensive methylation of the parvovirus genome has been recently demonstrated. A global pattern of methylation of CpG dinucleotides is seen in vertebrate genomes, compared to “fractional” methylation patterns in invertebrate genomes. It remains unknown if the loss of CpG dinucleotides occurs in all viruses of a given DNA virus family that infect host species spanning across vertebrates and invertebrates. We investigated the link between the extent of CpG dinucleotide depletion among autonomous parvoviruses and the evolutionary lineage of the infected host. We demonstrate major differences in the relative abundance of CpG dinucleotides among autonomous parvoviruses which share similar genome organization and common ancestry, depending on the infected host species. Parvoviruses infecting vertebrate hosts had significantly lower relative abundance of CpG dinucleotides than parvoviruses infecting invertebrate hosts. The strong correlation of CpG dinucleotide depletion with the gain in TpG/CpA dinucleotides and the loss of TpA dinucleotides among parvoviruses suggests a major role for CpG methylation in the evolution of parvoviruses. Our data present evidence that links the relative abundance of CpG dinucleotides in parvoviruses to the methylation capabilities of the infected host. In sum, our findings support a novel perspective of host-driven evolution among autonomous parvoviruses.