Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Inactive or moderately active human promoters are enriched for inter-individual epialleles

Inactive or moderately active human promoters are enriched for inter-individual epialleles Background: Inter-individual epigenetic variation, due to genetic, environmental or random influences, is observed in many eukaryotic species. In mammals, however, the molecular nature of epiallelic variation has been poorly defined, partly due to the restricted focus on DNA methylation. Here we report the first genome-scale investigation of mammalian epialleles that integrates genomic, methylomic, transcriptomic and histone state information. Results: First, in a small sample set, we demonstrate that non-genetically determined inter-individual differentially methylated regions (iiDMRs) can be temporally stable over at least 2 years. Then, we show that iiDMRs are associated with changes in chromatin state as measured by inter-individual differences in histone variant H2A.Z levels. However, the correlation of promoter iiDMRs with gene expression is negligible and not improved by integrating H2A.Z information. We find that most promoter epialleles, whether genetically or non-genetically determined, are associated with low levels of transcriptional activity, depleted for housekeeping genes, and either depleted for H3K4me3/enriched for H3K27me3 or lacking both these marks in human embryonic stem cells. The preferential enrichment of iiDMRs at regions of relative transcriptional inactivity validates in a larger independent cohort, and is reminiscent of observations previously made for promoters that undergo hypermethylation in various cancers, in vitro cell culture and ageing. Conclusions: Our work identifies potential key features of epiallelic variation in humans, including temporal stability of non-genetically determined epialleles, and concomitant perturbations of chromatin state. Furthermore, our work suggests a novel mechanistic link among inter-individual epialleles observed in the context of normal variation, cancer and ageing. Keywords: Epigenetics, DNA methylation, epialleles Background (2) exposure to a compromised in utero environment as Epialleles are genomic loci at which the epigenetic state has been shown in rodent and human studies [10-12]; (3) can stably vary among individuals in a given population or adult life-style associated factors such as smoking [13]. [1]. Although first described and still best understood in Despite these and other previous studies, the molecular plants [2-4], in recent years we have come to realise that nature of mammalian epialleles, in particular those epigenomic landscapes in mammals can also show con- induced by non-genetic factors, has remained controver- siderable inter-individual variation (reviewed in [1,5]). sial [14]. To a large extent, this is due to the DNA methy- lation-focus of previous investigations [5]. Incorporation Mammalian epialleles could arise through the action of cis-or trans-genetic influences [6,7], or have non-genetic of information about the chromatin state would refine origins as a result of: (1) potential stochastic events [8,9]; our understanding of the molecular nature and ultimately functionality of epialleles in the context of normal varia- tion or disease states. * Correspondence: [email protected]; [email protected] † Contributed equally Here we describe the first systematic interrogation of The Blizard Institute, Barts and The London School of Medicine and mammalian epialleles that integrates genomic, methylo- Dentistry, Queen Mary University of London, 4 Newark Street, London E1 mic, transcriptomic and chromatin state information. 2AT, UK Full list of author information is available at the end of the article © 2013 Gemma et al.; licensee 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. Gemma et al. Genome Biology 2013, 14:R43 Page 2 of 11 http://genomebiology.com/2013/14/5/R43 Using a combination of experimental and computational was not feasible to repeat all the functional genomic analyses we identify key features of epiallelic variation in assays on additional MZ pairs. humans, including demonstrating that even non-geneti- cally determined epialleles can be temporally stable, and Identification of temporally stable inter-individual DMRs that DNA methylation variability at epialleles is associated (iiDMRs) with concomitant perturbations in chromatin state. Most An initial low-level analysis confirmed anticipated geno- notably, we find that promoter-associated epiallelic varia- mic/epigenomic correlations. First, we confirmed known tion is predominantly associated with developmentally functional correlations between DNA methylation, H2A.Z important and/or tissue-restricted genes. A similar cate- levels, and gene expression: promoter DNA methylation gory of genes is preferentially hyper-methylated in various was negatively correlated with both H2A.Z and gene cancers, in vitro cellular transformation, and human expression levels, and promoter H2A.Z levels were strongly chronological ageing, potentially pointing to a novel positively correlated with gene expression (Figure 1A). mechanistic link between these processes and human Second, analysis of the SNP arrays did not reveal intra-MZ inter-individual epiallelic variation. pair SNP or copy number variation (approximately 640,000 to 660,000 SNPs observed between unrelated individuals, Results and 104 and 115 SNPs observed in 31/32 and 21/22 MZ Comprehensive genomic/epigenomic/transcriptomic pairs, respectively, and these are most likely false positives). profiling of monozygotic (MZ) twin pairs Finally, DNA methylation profiles were substantially more For the initial discovery phase of the study, we originally similar between co-twins compared with unrelated indivi- aimed to generate integrated genomic/epigenomic/tran- duals (Figure 1B). scriptomic profiles for CD14+ cells from three different We then called inter-individual DMRs (iiDMRs) healthy MZ twin pairs of European ancestry (MZ pair between all five possible pair-wise comparisons (two dif- 31/32: female, age at sampling 27 years; MZ pair 21/22: ferent MZ pairs and one ‘unpaired’ co-twin). Illumina450K female, age at sampling 27 years; MZ pair, male, 11/12, probe sequences were remapped using BLAT, and we only age at sampling 19 years Additional file 1, Table S2). We used probes that mapped exactly once. Furthermore, we focussed on CD14+ cells as they can be obtained to >90% ignored probes on chrX and Y, and also those that over- purity [17] and Additional file 1, Figure S1, and are less lapped known SNPs (based on information provided in likely to harbor post-differentiation, random epigenetic the Illumina 450K annotation file) to eliminate artefacts alterationsastheyhave a lifespan of onlyafew weeks. due to differential probe hybridization effects, resulting in Genetic profiles were obtained using the Illumina a final set of 369,908 different probes. Exclusion of these probes eliminates only cis-acting genetic variants present Omni2.5S array that interrogates approximately 2.5 mil- lion single nucleotide polymorphisms (SNPs) with a in the 50 bp sequence covered by the probe, and all other minor allele frequency of down to 1%. DNA methylation cis- and trans-genetic effects are retained. We considered was assayed by Illumina450K arrays that provide bisulfite only those iiDMRs with a directionally consistent ≥5% conversion-based, single base resolution methylation absolute methylation difference at ≥2 adjacent CpGs measurements at approximately 450,000 different cyto- within 500 bp of each other, as the biological relevance of sines associated with a range of genomic features such as very small methylation differences limited to single CpG promoters, enhancers, and CpG islands (CGIs) [15]. sites is currently unclear. The number of iiDMRs found in Gene expression was profiled using the standard Illumina each of the five different pairwise comparisons is shown in mRNA-seq protocol. For the analysis of chromatin state Table 1. Consistent with Kaminsky et al. [8], iiDMRs were we performed ChIP-seq on the histone variant H2A.Z, found across the genome, but enriched at non-promoter which is strongly associated with transcriptional activity non-exonic regions, and a greater number of iiDMRs were (but can also be found at transcriptionally silent promo- observed between unrelated individuals (Figure 2A). ters), and is thought to be environmentally responsive Furthermore, we generated triplicate Illumina 450K pro- [16]. We obtained 60-80 million mapped 36 bp paired files for CD14+ cells from two additional MZ twin pairs, end reads for the ChIP-seq and RNA-seq libraries. All different to those described above, and found that the genomic/epigenomic/transcriptomic profiles were gener- amount of biological variation seen between twins ated from the same single sampling of CD14+ cells for exceeded the technical variation (false positive iiDMR -7 each individual. Unfortunately, during the course of pro- calls) by a factor of approximately 5 (P <10 , permutation cessing the samples, the DNA sample that was to be used test) (Figure 2B). for subsequent DNA methylation analysis for individual The epigenomic landscape for any given individual is, ‘12’ was inadvertently lost. As these twins were recruited to an extent, constantly in flux, and many iiDMRs iden- because we had methylomic data from a previous time- tified from single time-point measurements probably do point to test for temporal stability, as discussed below, it not have long-term phenotypic consequences. Therefore Gemma et al. Genome Biology 2013, 14:R43 Page 3 of 11 http://genomebiology.com/2013/14/5/R43 -0.5 -1 -1.5 -2 -2.5 -3 -3.5 -4 0 10 20 30 40 50 60 70 80 90 100 Methylation% [Illumina] 0 10 20 30 40 50 60 70 80 90 100 Methylation% [Illumina] -0.2 -0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 -1.8 -2 -2.2 -2.4 0 20 40 60 80 100 H2A.z tag count 0.995 0.99 0.985 0.98 0.975 0.97 31/32 21/22 31/21 31/11 11/22 Figure 1 Epigenomic, genomic and transcriptomic profiling in CD14+ cells from MZ twins. (A) We analysed the following correlations at promoter regions: DNA methylation vs. gene expression, H2A.Z vs. gene expression and DNA methylation vs. H2A.Z. Gene expression levels are represented as log-transformed aligned average FPKM (fragments per kilobase of exon per million fragments mapped) from RNA-seq data. FPKM values were determined by TopHat and Cufflinks and assigned to TSSs. DNA methylation values were determined as b values using Illumina 450K array (see Materials and Methods) and are represented here as percentage of methylation bins. Probes on the Illumina 450K array were assigned to their closest TSS, discarding assignments >1 kb away. H2A.Z abundance is represented as normalised read counts from ChIP-seq -3 -3 data correlates positively with gene expression (Spearman’s r = 0.50, P <10 ), DNA methylation is anti-correlated with H2A.Z (r = -0.59, P <10 ). Promoters were defined as TSS ± 1,000 bp. (B) Correlation coefficients of inter-individual DNA methylation differences. 31/32 and 21/22 plots correspond to MZ twin iiDMRs; 31/21, 31/11 and 11/22 correspond to iiDMRs found in unrelated individuals. Correlation (Spearman’s rho) Mean Expression (log FPKM) Mean Expression (log FPKM) Mean H2A.z tag-count 10 10 122/222 122/221 122/213 122/212 122/211 121/222 121/221 121/213 121/212 121/211 112/222 112/221 112/213 112/212 112/211 111/222 111/221 111/213 111/212 111/211 213/223 212/222 211/221 213/221 212/221 211/221 112/122 112/121 111/122 111/121 221/222 212/213 211/213 211/212 121/122 111/112 Gemma et al. Genome Biology 2013, 14:R43 Page 4 of 11 http://genomebiology.com/2013/14/5/R43 Table 1 Number of iiDMRs identified in the various obtaining some measure of iiDMR temporal stability is pairwise comparisons. critical, especially for non-genetically determined Comparison iiDMRs (n) iiDMRs. The CD14+ cells used in this study were obtained from individuals recruited in late 2010. We 31 vs. 32 5,077 previously sampled the same individuals in mid-2008 as 21 vs. 22 2,528 part of a separate study in which we profiled their CD14 31 vs. 21 9,512 + cells using the Illumina27K array [17]. This array con- 31 vs. 11 9,062 tains probes for 27,578 different CpG sites, largely pro- 11 vs. 22 10,526 moter-associated, and >90% of these CpGs are also Number of iiDMRs for each comparison was calculated as DNA methylation differences ≥5% at ≥2 adjacent CpGs within 500 bp of each other. represented on the Illumina450K array. For maximum Replicates Twins Unrelated 31/32 21/22 31/21 32/22 30 30 30 30 20 20 20 20 10 10 10 10 0 0 0 0 -10 -10 -10 -10 -20 -20 -20 -20 -30 -30 -30 -30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 Methylation% difference [current] Methylation% difference [current] Methylation% difference [current] Methylation% difference [current] Figure 2 Genomic characteristics and temporal stability of inter-individual DMRs (iiDMRs). (A) This plot shows the distribution of iiDMRs across different genomic features (promoters, regulatory regions (that is, Regulatory Features from Ensembl), exons, introns and all other regions (that is, other). iiDMRs were defined as differences >5% in DNA methylation between individuals present in two adjacent probes of the Illumina 450K array. The feint line shows the expected level if iiDMRs were found equally in all regions of the genome. (B) A comparison of technical vs. biological variation for iiDMRs. Shown is the number of iiDMR calls between Illumina 450K array for technical replicates, twin pairs and unrelated individuals. This analysis was done with CD14+ DNA from two MZ twin pairs in triplicate. The biological variation significantly exceeded the -7 technical variation (P <10 , permutation test). (C) Temporal stability of iiDMRs. The inter-individual DNA methylation differences found in this study were compared with the DNA methylation differences present in the same individuals 2 years prior. The previous DNA methylation differences were obtained from CD14+ DNA methylation profiles from the same individuals using Illumina 27K data as part of a separate study [17]. Shown are all current and historic inter-individual methylation differences from all the common probes present on Illumina 27K and Illumina 450K arrays. Methylation% difference [historic] Number of iiDMRs Methylation% difference [historic] Methylation% difference [historic] Methylation% difference [historic] Gemma et al. Genome Biology 2013, 14:R43 Page 5 of 11 http://genomebiology.com/2013/14/5/R43 power when comparing to the much less densely spaced >1 kb away from the promoter (Figure 3A). Therefore, our probes on the Illumina27K array, we considered all >5% data show that a significant number of iiDMRs, even those methylation differences, rather than just those in multi- that are non-genetically determined, represent genomic ple-probe iiDMRs. Comparison of the 2010 with 2008 sites that harbour perturbed chromatin states, not just methylomic profiles, for CpG sites common to both DNA methylation differences, and thus can be classified as platforms, revealed that a substantial proportion of the epialleles [19]. epigenetic variation seen in the 2010 samplings were also Interestingly, we observed only a very weak anti-corre- found in the 2008 samplings, that is, was temporally stable lation between promoter iiDMRs and RNA-seq gene expression levels (Figure 3B). This was true for both (Figure 2C). Although this is not surprising for compari- sons between unrelated individuals, since many of the intra- and inter-MZ pair iiDMRs, and not significantly methylation differences in such cases most likely represent improved even when considering just those iiDMRs that stable genetic effects, temporal stability in the intra-MZ were anti-correlated with H2A.Z, or iiDMRs with a pair comparisons is particularly noteworthy as it means large magnitude, or those found within CpG islands or that random or environmental events can induce perma- CpG shores (Figure 3B). Not surprisingly, a direct com- nent, or at least semi-permanent, epiallelic variation. parison between intra-/inter-MZ pair H2A.Z variation Furthermore, given that the lifespan of CD14+ cells is only and expression did not reveal any significant correla- a few weeks, compared with the approximately 2-year gap tions either (Additional file 1, Figure S4). between the first and second samplings, these epialleles The relationship between promoter DNA methylation most probably arise in blood progenitor cells. For the and gene expression is known to be complex, even in CpGs in common between the second (450K array) and contexts where large DNA methylation differences are first time-points (27K arrays), the proportion of methyla- generally observed, for example, genetically-encoded dif- tion differences found at the second time-point that were ferentiation programs or cancer [5]. In our case, we also present at the first time-point is shown in Additional found it surprising that even inclusion of H2A.Z informa- file 1, Figure S2. The number of sites where a difference in tion did not improve the strength of the correlations as, the same direction is seen at both sites significantly theoretically, integration of information from different exceeds what would be expected by chance in all cases components of the ‘epigenetic’ state of a region should -13 2 (P <3 ×10 , c test). This is the first genome-scale yieldmorerobustcorrelations.To furtherexplore the demonstration of temporally stable epiallelic variation in cause for these observations, we compared mean expres- mammals that cannot potentially be explained by stable sion levels of iiDMR promoter-associated genes with all genetic effects (for example, [18]). For further analyses we genes in our dataset. We found promoter iiDMR-asso- ciated genes to be expressed at significantly lower levels, did not restrict ourselves to CpG sites common to the Illu- mina450K and 27K arrays, otherwise we would have been relative to the other genes, in all intra- and inter-MZ pair left with too few DMRs for any meaningful analyses, but it comparisons (Figure 3C). In other words, it seems that is reasonable to assume that the same degree of temporal promoter-iiDMRs are associated with genes that are stability should be present across the entire set of iiDMRs lowly expressed or silent in CD14+ cells. found from the Illumina450K data obtained from the 2010 These observations were reminiscent of previous results samplings. and our own report on human ageing-associated DNA methylation dynamics [20,21]. In those studies, human iiDMRs at promoters anti-correlate with H2A.Z and promoters that become hypermethylated with chronologi- preferentially associate with lowly expressed cal age (aDMRs) were also associated with genes expressed or silent genes at relatively low levels in the analysed tissue. Notably, We next wanted to investigate whether iiDMRs are asso- these aDMRs were strongly enriched for promoters har- ciated with altered chromatin states or gene expression, bouring bivalent chromatin domains in embryonic stem focussing on promoter-associated iiDMRs as it is currently (ES) cells. Bivalent domains harbour both H3K4me3, gen- not straightforward to correlate the activity of gene distal erally considered an active mark, and H3K27me3, gener- regulatory elements with gene expression levels. Promoters ally considered an inactive mark [22,23]. Furthermore, were defined as a 1 kb window centred at the annotated bivalent domain promoters are associated with develop- transcriptional start site (TSS). We observed a statistically mentally important and tissue-restricted genes [22,23]. significant anti-correlation between DNA methylation and An analysis of previously published H3K4me3 and H2A.Z at both intra- and inter-MZ pair promoter iiDMRs H3K27me3 ChIP-seq profiles in human ES cells [24] (Figure 3A and Additional file 1: Figure S3). This was espe- revealed that promoter iiDMRs were only moderately enriched for bivalent chromatin domains (relative to gen- cially marked for iiDMRs associated with >5% absolute ome average, refer to Figure 3D). Surprisingly though, methylation differences. Significantly, the negative H2A.Z- strong statistically significant enrichment was observed for DNA methylation correlation was observed even at sites Gemma et al. Genome Biology 2013, 14:R43 Page 6 of 11 http://genomebiology.com/2013/14/5/R43 Near TSS >1kb from TSS All data H2A.z in expected direction CpG shores 0.08 0.08 0.1 0.1 0.15 0.1 0.06 0.06 A 0.05 0.05 0.05 0.04 0.04 0 0 0 0.02 0.02 -0.05 -0.05 -0.05 -0.1 0 0 -0.1 -0.1 -0.15 -0.02 -0.02 -0.2 -0.15 -0.15 -0.04 -0.04 -0.25 -0.06 -0.06 -0.2 -0.2 -0.3 31/32 21/22 31/21 31/11 11/22 31/32 21/22 31/21 31/11 11/22 31/32 21/22 31/21 31/11 11/21 31/32 21/22 31/21 31/11 11/21 31/32 21/22 31/21 31/11 11/21 31/32 21/22 31/32 21/22 1 1 DMRs DMRs 0.14 0.14 NonDMR NonDMR 0.9 0.9 E 0.12 0.12 0.8 0.8 0.1 0.1 0.08 0.08 0.7 0.7 0.06 0.06 0.6 0.6 0.04 0.04 0.5 0.5 0.02 0.02 0 0 0.4 0.4 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0.3 0.3 -4 -3 -2 -1 0 1 2 3 4 5 6 -4 -3 -2 -1 0 1 2 3 4 5 6 Log FPKM Log FPKM 10 10 31/21 31/11 31/21 31/11 1 1 0.14 0.14 DMRs DMRs NonDMR NonDMR 0.12 0.12 0.9 0.9 0.1 0.1 0.8 0.8 0.08 0.08 0.7 0.7 0.06 0.06 0.04 0.04 0.6 0.6 0.02 0.02 0.5 0.5 0 0 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0.4 0.4 0.3 0.3 -4 -3 -2 -1 0 1 2 3 4 5 6 -4 -3 -2 -1 0 1 2 3 4 5 6 Log FPKM Log FPKM 10 10 11/22 11/22 0.14 DMRs NonDMR 0.12 0.9 0.1 0.8 0.08 0.06 0.7 0.04 0.6 0.02 0.5 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0.4 0.3 -4 -3 -2 -1 0 1 2 3 4 5 6 Log FPKM Figure 3 Promoter iiDMRs anti-correlate with H2A.Z changes and are preferentially found at genes expressed at low levels. (A) Anti- correlation between inter-individual DNA methylation (for only >5% methylation differences) and H2A.Z differences in CD14+ cells. Data are shown as mean and 95% credible intervals. (B) Inter-individual promoter epiallelic differences are not correlated with gene expression differences. For each bin of inter-individual DNA methylation differences we calculated the ratio of log transformed RNA-seq reads (FPKM) in the first individual of each comparison respect to the other. Data are represented as 95% credible intervals on the mean. The left panel shows the correlations using DNA methylation data only, the middle panel uses those epialleles in which DNA methylation and H2AZ are anti-correlated, and the right panel looks at epiallelic variation at CpG shores (as defined in the Illumina450K array annotation file). (C) iiDMRs are predominantly found at genes expressed at low levels. Shown is the distribution RNA-seq reads (expressed as FPKM) for iiDMRs and non-iiDMRs. In all cases the curves corresponding to DMRs are shifted to the left compared to non-DMRs showing that DMRs occur preferentially at genes expressed at low levels. For all cases 31/32 and 21/22 plots correspond to MZ twin comparisons; 31/21, 31/11 and 11/22 correspond to unrelated individuals comparisons. (D) Promoter iiDMRs occur preferentially at regions depleted for H3K4me3 and enriched for H3K27me3 (K4lo/K27hi) or depleted for both these marks (K4lo/K27lo) in human embryonic stem (hES) cells. hES cell H3K4me3 and H3K27me3 ChIP-seq data are from [24]. The dashed line in each plot refers to the overall iiDMR fraction against the whole dataset. (E) Illumina450K probes were ranked by methylation-ageing correlation using data from [20], and the 500 most- age-associated probes were taken as aDMRs, while 500 randomly selected probes from the remainder of the dataset were taken as controls. For each set, we collected absolute methylation differences from each of the five possible pairings of individuals in this study, and plot 95% credible intervals on the mean. the high H3K27me3/low H3K4me3, or low H3K27me3/ common property between iiDMR and aDMR promoters low H3K4me3 states in ES cells. Both these chromatin seems to be a strong association with genes that are tis- states are also strongly associated with tissue-restricted sue-restricted, but are only moderately active, or inactive, genes and indeed we found iiDMR promoters were signifi- in the analysed tissue. cantly less likely to be associated with house keeping -5 genes (P <10 in all five possible pair-wise comparisons, Analysis of iiDMRs in an independent cohort Chi-squared test). Re-analysis of the 500 most age-corre- To independently validate the findings above, we re-ana- lated probes (that is, aDMRs) from [20] revealed that these lysed DNA methylomic data from our previously pub- probes display significantly more intra- and inter-MZ pair lished human ageing study, that is, the test set [20]. This variability than 500 randomly selected probes (P <0.01, larger and independent dataset consists of Illumina27K bootstrapped for all five comparisons, Figure 3E). So the methylome profiles ofwhole bloodobtainedfrom30 Mean log moderated ratio Cumulative frequency Cumulative frequency Cumulative frequency Mean log moderated ratio Cumulative frequency Cumulative frequency 10 Mean log moderated FPKM ratio Fraction of probes showing >5% difference Fraction of probes showing >5% difference Fraction of probes showing >5% difference K4+ K27+ K4+ K27+ K4+ K27+ K4+ K27- K4+ K27- K4+ K27- K4- K27+ K4- K27+ K4- K27+ K4- K27- K4- K27- K4- K27- Fraction of probes showing >5% difference Fraction of probes showing >5% difference Mean log moderated FPKM ratio K4+ K27+ K4+ K27+ K4+ K27- K4+ K27- K4- K27+ K4- K27+ K4- K27- K4- K27- Mean log moderated FPKM ratio 10 Gemma et al. Genome Biology 2013, 14:R43 Page 7 of 11 http://genomebiology.com/2013/14/5/R43 different healthy female MZ pairs of European ancestry, genetically determined epialleles can be temporally stable ranging in age from 25 to 79 years old [20]. Although the (at least over the course of 2 years). That is, a significant 450K array overall has approximately 15X as many probes fraction of these epialleles are not just transient epige- as the 27K, the majority of these are outside of promoter netic perturbations with little prospect of influencing regions. Of the total annotated protein-coding genes in molecular function. Second, inter-individual DNA the human genome (21,665), 19,409 (89.6%) are associated methylation variants are associated with perturbations of with at least one promoter probe on the 450K array, chromatin state, a relationship observed for even small whereas 14,400 (66.5%) are associated with at least one differences, for example, down to approximately 5% promoter probe on the 27K array. We calculated root methylation difference, and therefore can be considered mean square (RMS) differences for each probe on the Illu- as bona fide epigenetic perturbations. Of course, future mina27K array across the 30 different MZ pairs for both studies using bigger sample numbers are needed to intra- and inter-MZ pair comparisons. The RMS deviation further explore our initial findings. is a measure of the inter-individual methylation variability, The most significant aspect of our study is the finding and is directly proportional to the level of variability that the correlation of iiDMRS with gene expression differ- observed in the cohort under study. RMS difference, as ences is very weak and that iiDMRs are preferentially opposed to mean difference, was used because the differ- found in regions of relative transcriptional inactivity. So ences will be in an arbitrary direction for each pair. It is what are the implications of this? First, it is possible that important to note that the RMS difference is not a mea- some promoter epialleles show inter-related DNA methy- sure of directional age-related changes. Analysis of the test lation and chromatin state perturbations, but may not set resulted in several key conclusions, applicable to both impact significantly on genome function, at least as mea- intra-MZ pair and inter-MZ pair comparisons, validating sured by steady state transcriptional activity. In the case of our findings from the discovery set. First, iiDMR-asso- non-genetically determined epialleles, maybe all promoters ciated promoters found in the test set were associated are potentially subject to epiallelic variation, but the more with genes expressed at significantly lower levels com- active ones are ‘cleared’ of aberrant epigenomic variants, pared with the non-iiDMR set of promoters in the CD14+ whereas the less active/silent promoters can accumulate RNA-seq data generated in our study, and in a previously epigenetic variation. But the enrichment of epialleles in published whole blood array-based expression dataset less active/silent promoters was also found in comparisons (Figure 4A and Additional file 1, Figure S5). Second, the between unrelated individuals. Although it is hard to say test-set iiDMRs were significantly depleted for housekeep- what proportion of epialleles between unrelated indivi- -5 ing genes (P <10 foreitherintra-MZ pairorinter-MZ duals are due to genetic as opposed to environmental dif- ferences from our data, the genetic influence on DNA pair comparisons, Chi-squared test). Third, iiDMRs identi- fied in the test set were significantly enriched at promoters methylation profiles is well documented [3,4,25]. Bell and that harbour either high levels of H3K27me3 and low colleagues measured genome-wide methylation in 77 Hap- levels of H3K4me3, or are devoid of both of these marks Map Yoruba individuals, for which gene expression and in ES cells (Figure 4B). The fact that our findings from genotype data were available, and found a strong genetic CD14+ cells in discovery set were validated by unsorted component to inter-individual variation in DNA methyla- whole blood cell data in the test set further supports the tion profiles [4]. Although they found a significant enrich- robustness of our key conclusions. We also determined ment of SNPs that affect both methylation and gene that for all five possible comparisons in the discovery data- expression, they also noted that the total number of genes set, there’s a strong correlation of both intra- and inter- showing such a signal is only a small proportion of the MZ pair variability between the discovery and test datasets total number of methylation variants they identified [4]. A (Figure 4C). Finally, we also asked if there is an overlap similar conclusion was reached by Myers and colleagues between intra- and inter-MZ pair iiDMRs. This was done who analysed genome-wide methylation in six members of using the test set since it has significantly more pairs. a three generation family and found that only 22% of We found that mostly the same probes show the highest genes harbouring genotype-dependent DNA methylation variability in both intra- and inter-MZ pair comparisons exhibited allele-specific gene expression (albeit more than (Figure 4D). Interestingly, inter-MZ pair comparisons show expected by chance) [25]. Therefore, in both cases the cor- bigger differences at the same sites relative to intra-pair relation between genetically determined DNA methylation comparisons but the great majority of these show greater and expression is at best modest, which would be consis- variability in the between-pair comparisons (Figure 4D). tent with our results regarding chromatin state. It is possible that epiallelic variation acts in a manner Discussion not evident from simple correlations with steady-state expression levels in a given tissue. First it is possible that Our data reveal several novel and important features of these correlations are tissue-restricted as has recently mammalian epialleles. First, we find that even non- Gemma et al. Genome Biology 2013, 14:R43 Page 8 of 11 http://genomebiology.com/2013/14/5/R43 Within pairs Between pairs 0.45 0.45 0.3 A BD 0.4 0.4 0.35 0.35 0.25 0.3 0.3 0.2 0.25 0.25 0.2 0.2 0.15 0.15 0.15 0.1 0.1 0.1 0.05 0.05 0.05 0 0 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0 0.05 0.1 0.15 0.2 0.25 0.3 Within-pair differences (RMS) Figure 4 Validation of iiDMRs in an independent cohort. (A) For this analysis we used 30 MZ twin pairs from [20] whose whole blood DNA methylation profiles were generated by Illumina 27K arrays. We defined low and high expression based on the RNA-seq data we generated in this study from CD14+ cells. High expression: >1 FPKM, Low expression: <1 FPKM. (B) Promoter iiDMRs in the test set are preferentially enriched at regions low in H3K4me3 and high in H3K27me3, and in regions that lack both of these marks in hES cells. The analysis was performed essentially as described in the legend for Figure 3D. (C) Intra- and inter-MZ pair iiDMRs are correlated in the discovery and validation cohorts. For the 30 MZ pairs from [20], we measured intra-pair methylation variability by taking the RMS (root-mean-square) methylation differences for each probe. We also calculated a similar inter-pair variability measure by permuting the pairs. For each of the five possible pairs in the current study, we bin probes by methylation difference, and plot the mean inter- and intra-pair validation methylation variabilities. (D) For the 30 MZ pairs from [20], we calculated intra-pair and inter-pair RMS methylation variability as above and show a scatter plot for all probes on the Illumina27K array (with exclusions as described in the Methods for the Illumina450K data). been shown for genetically determined tissue-restricted gradual hyper-methylation during in vitro cell culture is gene expression [26]. Alternatively, conclusions from two found at promoters associated with genes not expressed recent studies, although not focusing on DNA methyla- in that cell type [30]. Additionally, it has been found in a tion/chromatin state in mammals, hint at other potential variety of human cancers that bivalent chromatin mechanisms by which epialleles could act. Yvert and col- domains (associated with low transcriptional activity in leagues recently compared H3K14 acetylation profiles stem cells) are preferential targets of hyper-methylation between two strains of the yeast Saccharomyces cerevi- [31-33]. The common thread among these seemingly dis- siae, and found 5,442 sites that significantly differed in parate examples of inter-individual epigenetic variation is H3K14ac levels, which they called single nucleosome epi- promoters that are developmentally regulated and tissue- restricted, and are only moderately active, or inactive, in polymorphisms (SNEPs) [27]. However, higher acetyla- the analysed tissue. We propose that there could be a tion in one strain did not always mean higher expression of the relevant gene, for example, in one case the SNEP potentially important mechanistic link between normal/ was associated with the strength of gene activation upon stochastic epiallelic variation and the epigenetic perturba- stimulation by heat shock. Secondly, Lindgren and collea- tions observed in the context of cancer and ageing. gues recently assessed the effect of naturally occurring variation in miRNA expression levels on mRNA levels in Conclusions humans, but found little correlation [28]. The authors The existence of mammalian epialleles is not in doubt, concluded that their findings were more consistent with but the key challenge now is to characterise epialleles at the primary role of miRNAs being to buffer mRNA the molecular level. Our work reveals key and novel levels. A key conclusion therefore is that correlating properties of epiallelic variation in humans, and further epialleles with steady-state RNA dynamics, possibly the suggests important mechanistic links between normal most common analysis currently presented in papers on inter-individual epigenetic variation and epigenetic per- epiallelic investigations, may not be particularly fruitful. turbations observed in cancer and chronological ageing. Finally, and potentially most importantly, the broadly similar characteristics of iiDMRs and aDMRs (from our Materials and methods previous study [20] and [29]) may in fact be a general fea- Samples ture of mammalian epiallelic variation in a variety of con- Fresh venous blood was obtained from three pairs (six texts. Meissner and colleagues found that aberrant individuals) of healthy MZ twins (Additional file 1, Fraction of probes showing >5% RMS diff. Fraction of probes showing >5% RMS diff. Between-pair differences (RMS) Gemma et al. Genome Biology 2013, 14:R43 Page 9 of 11 http://genomebiology.com/2013/14/5/R43 Table S1). Blood was diluted 1:1 in RPMI media and theIllumina arrayand used that as aChIPscore for then peripheral blood mononuclear cells (PBMCs) were that probe. RNA-seq reads were mapped to the refer- separated by Ficoll-Hypaque gradient centrifugation. ence genome using Tophat 1.3.1, then expression levels CD14+ cells were isolated according to manufacturer’s (FPKM) were estimated for each Ensembl transcript instruction using magnetic bead-based positive selection using Cufflinks 1.0.3. For analyses comparing methyla- system (Miltenyi Biotech). The purity of the cells was tion data to expression, methylation array probes lying determined by FACS using CD14-FITC antibodies within 1 kb of an Ensembl TSS were assigned an (Additional file 1, Figure S1). All subjects gave informed ‘expression level’ equal to that of the transcript asso- consent and the study was approved by the Northern ciated with the nearest TSS. and Yorkshire Research Ethics committee (REC Refer- Statistical analyses ence Number: 06-MREO-3-22). Validation of iiDMRs Correlation between variables was performed using was done using whole blood Illumina27K data pre- Spearman’s rank test. Confidence intervals for all box/bar viously generated [20]. This cohort included 30 different plots are obtained by bootstrapping unless otherwise sta- healthy MZ female twin pairs recruited from within the ted. Confidence intervals for the hES cell H3K4me3/ UK as part of the TwinsUK registry. H3K27me3 bar charts are estimated from a binomial model. Probes associated with housekeeping genes were Illumina 450K array defined as in [20]. A total of 500 ng of DNA from CD14+ cells isolated using For the genomic location enrichment analyses, exon, QIAamp DNA Mini Kit was bisulfite converted using the intron and regulatory features were extracted from EZ DNA Methylation kit (Zymo Research). Arrays were Ensembl, and promoters were defined as regions within processed at the Barts and The London Genome Centre, 1 kb of the TSSs of an Ensembl gene. For each of these London, UK according to the manufacturer’s recommen- categories, we asked what fraction of the probes lying in dations. Methylation scores for each CpG site are called as the selected regions were called as iiDMRs, and plot ‘Beta’ values (using BeadStudio software from Illumina), 95% confidence intervals on this proportion, estimated that range from 0 (unmethylated (U)) to 1 (fully methy- using a binomial model. For comparison, the feint line lated (M)) on a continuous scale, and are calculated from indicates the fraction of iiDMRs across the whole data- the intensity of the M and U alleles as the ratio of fluores- set, allowing enrichment or depletion to be assessed. cent signals. Accession codes Illumina Omni2.5S arrays All data are available on GEO [GSE46220]. The arrays were processed according to the manufac- turer’s instructions using 500 ng of DNA. Additional material ChIP- and RNA-seq The chromatin immunoprecipitation (ChIP) assay was Additional file 1: Additional Materials and methods, Tables S1 and S2, and Figures S1 to S5. performed on 5 × 10 CD14+ cells according to pre- viously published protocols with minor modifications [34]. Chromatin was sonicated to get fragments of 100 to 500 bp and immunoprecipitated with 10 uL of anti-H2A. List of abbreviations aDMR: ageing-associated differentially methylated region; CGI: CpG island; Z antibody (Active Motif, Cat no: 39113). ChIP-seq ChIP-seq: chromatin immunoprecipitation; ESC: embryonic stem cells; FPKM: libraries were prepared following the Illumina protocol fragments per kilobase of exon per million fragments mapped; iiDMR: inter- and ligated to standard PE adaptors and sequenced on an individual differentially methylated region; MZ: monozygotic; PBMC: peripheral blood mononuclear cells; RMS: root mean square; SNEP: single Illumina GAIIx instrument. For RNA-seq, 200 ng of total nucleosome epi-polymorphism; SNP: single nucleotide polymorphism; TSS: RNA was used to prepare RNA-seq libraries using the transcriptional start site. TruSeq RNA kit from Illumina following the instructions Competing interests provided in the supplier’s manual, and sequenced on an The authors declare that they have no competing interests. Illumina GAIIx instrument. Authors’ contributions CG performed the majority of the experimental work and helped to draft Sequence data processing the manuscript. TAD performed the bioinformatics analyses. HB and MIH ChIP-seq reads were mapped to the GRCh37 (hg19) collected and processed blood samples from the MZ twin pair participants. reference genome sequence using MAQ 0.6.6 and map- MLH assisted CG with various aspects of the experimental work. PJH assisted with the ChIP-seq. GG, RDL and GCE contributed reagents and materials. SVR pings with quality scores <10 were discarded. For and VKR conceived and designed the study, participated in its design, iiDMR-centric analyses, we counted the numbered of coordination and analysis, and helped to draft the manuscript. All authors paired end fragments overlapping each probe region on read and approved the final manuscript. Gemma et al. Genome Biology 2013, 14:R43 Page 10 of 11 http://genomebiology.com/2013/14/5/R43 14. Bird A: Putting the DNA back into DNA methylation. Nat Genet 2011, Acknowledgements 43:1050-1051. VKR, CG and RDL are supported by the BBSRC, UK (BB/H012494/1). VKR and 15. Sandoval J, Heyn H, Moran S, Serra-Musach J, Pujana MA, Bibikova M, RDL are also supported by the EU-FP7 ‘BLUEPRINT’ program (282510). RDL is Esteller M: Validation of a DNA methylation microarray for 450,000 CpG also supported by Juvenile Diabetes Research Foundation International sites in the human genome. Epigenetics 2011, 6:692-702. (JDRFI Award 5-2011-145). HB was supported by EFSD/Novo Nordisk 16. Conerly ML, Teves SS, Diolaiti D, Ulrich M, Eisenman RN, Henikoff S: Program Grant and Diabetes UK (10/0004107). SVR and GE are funded by Changes in H2A.Z occupancy and DNA methylation during B-cell the Multiple Sclerosis Society of the United Kingdom. SVR and GG are lymphomagenesis. Genome Res 2010, 20:1383-1390. funded by the Medical Research Council of the United Kingdom (G0801976). 17. Rakyan VK, Beyan H, Down TA, Hawa MI, Maslau S, Aden D, Daunay A, TAD is funded by the Wellcome Trust (083563). Busato F, Mein CA, Manfras B, Dias KR, Bell CG, Tost J, Boehm BO, Beck S, Leslie RD: Identification of type 1 diabetes-associated DNA methylation Authors’ details The Blizard Institute, Barts and The London School of Medicine and variable positions that precede disease diagnosis. PLoS Genet 2011, 7: Dentistry, Queen Mary University of London, 4 Newark Street, London E1 e1002300. 2AT, UK. Department of Physiology, Anatomy and Genetics and Medical 18. Feinberg AP, Irizarry RA, Fradin D, Aryee MJ, Murakami P, Aspelund T, Research Council Functional Genomics Unit, South Parks Road, Oxford, OX1 Eiriksdottir G, Harris TB, Launer L, Gudnason V, Fallin MD: Personalized 3PT, UK. The Gurdon Institute and Department of Genetics, University of epigenomic signatures that are stable over time and covary with body Cambridge, Tennis Court Road, Cambridge CB2 1QN, UK. School of mass index. Sci Transl Med 2010, 2:49ra67. Biological and Chemical Sciences, Queen Mary University of London, 19. Finer S, Holland ML, Nanty L, Rakyan VK: The hunt for the epiallele. Environ London, E1 4NS, UK. Wellcome Trust Centre for Human Genetics, University Mol Mutagen 2011, 52:1-11. of Oxford, Headington, Oxford OX3 7BN, UK. 20. Rakyan VK, Down TA, Maslau S, Andrew T, Yang TP, Beyan H, Whittaker P, McCann OT, Finer S, Valdes AM, Leslie RD, Deloukas P, Spector TD: Human Received: 9 July 2012 Revised: 24 April 2013 Accepted: 25 May 2013 aging-associated DNA hypermethylation occurs preferentially at bivalent Published: 25 May 2013 chromatin domains. Genome Res 2010, 20:434-439. 21. Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, Weisenberger DJ, Shen H, Campan M, Noushmehr H, Bell CG, Maxwell AP, Savage DA, References Mueller-Holzner E, Marth C, Kocjan G, Gayther SA, Jones A, Beck S, 1. Richards EJ: Population epigenetics. Curr Opin Genet Dev 2008, 18:221-226. Wagner W, Laird PW, Jacobs IJ, Widschwendter M: Age-dependent DNA 2. Miura K, Agetsuma M, Kitano H, Yoshimura A, Matsuoka M, Jacobsen SE, methylation of genes that are suppressed in stem cells is a hallmark of Ashikari M: A metastable DWARF1 epigenetic mutant affecting plant cancer. Genome Res 2010, 20:440-446. stature in rice. Proc Natl Acad Sci USA 2009, 106:11218-11223. 22. Bernstein BE, Mikkelsen TS, Xie X, Kamal M, Huebert DJ, Cuff J, Fry B, 3. Schmitz RJ, Schultz MD, Lewsey MG, O’Malley RC, Urich MA, Libiger O, Meissner A, Wernig M, Plath K, Jaenisch R, Wagschal A, Feil R, Schreiber SL, Schork NJ, Ecker JR: Transgenerational epigenetic instability is a source of Lander ES: A bivalent chromatin structure marks key developmental novel methylation variants. Science 2011, 334:369-373. genes in embryonic stem cells. Cell 2006, 125:315-326. 4. Becker C, Hagmann J, Müller J, Koenig D, Stegle O, Borgwardt K, Weigel D: 23. Azuara V, Perry P, Sauer S, Spivakov M, Jørgensen HF, John RM, Gouti M, Spontaneous epigenetic variation in the Arabidopsis thaliana Casanova M, Warnes G, Merkenschlager M, Fisher AG: Chromatin methylome. Nature 2011, 480:245-249. signatures of pluripotent cell lines. Nat Cell Biol 2006, 8:532-538. 5. Rakyan VK, Down TA, Balding DJ, Beck S: Epigenome-wide association 24. Zhao XD, Han X, Chew JL, Liu J, Chiu KP, Choo A, Orlov YL, Sung WK, studies for common human diseases. Nat Rev Genet 2011, 12:529-541. Shahab A, Kuznetsov VA, Bourque G, Oh S, Ruan Y, Ng HH, Wei CL: Whole- 6. Zhang D, Cheng L, Badner JA, Chen C, Chen Q, Luo W, Craig DW, genome mapping of histone H3 Lys4 and 27 trimethylations reveals Redman M, Gershon ES, Liu C: Genetic control of individual differences in distinct genomic compartments in human embryonic stem cells. Cell gene-specific methylation in human brain. Am J Hum Genet 2010, Stem Cell 2007, 1:286-298. 86:411-419. 25. Gertz J, Varley KE, Reddy TE, Bowling KM, Pauli F, Parker SL, Kucera KS, 7. Bell JT, Pai AA, Pickrell JK, Gaffney DJ, Pique-Regi R, Degner JF, Gilad Y, Willard HF, Myers RM: Analysis of DNA methylation in a three-generation Pritchard JK: DNA methylation patterns associate with genetic and gene family reveals widespread genetic influence on epigenetic regulation. expression variation in HapMap cell lines. Genome Biol 2011, 12:R10-R16. PLoS Genet 2011, 7:e1002228. 8. Kaminsky ZA, Tang T, Wang SC, Ptak C, Oh GH, Wong AH, Feldcamp LA, 26. Nica AC, Parts L, Glass D, Nisbet J, Barrett A, Sekowska M, Travers M, Virtanen C, Halfvarson J, Tysk C, McRae AF, Visscher PM, Montgomery GW, Potter S, Grundberg E, Small K, Hedman AK, Bataille V, Tzenova Bell J, Gottesman II, Martin NG, Petronis A: DNA methylation profiles in Surdulescu G, Dimas AS, Ingle C, Nestle FO, di Meglio P, Min JL, Wilk A, monozygotic and dizygotic twins. Nature Genet 2009, 41:240-245. Hammond CJ, Hassanali N, Yang TP, Montgomery SB, O’Rahilly S, 9. Fraga MF, Ballestar E, Paz MF, Ropero S, Setien F, Ballestar ML, Heine- Lindgren CM, Zondervan KT, Soranzo N, Barroso I, Durbin R, et al: The Suñer D, Cigudosa JC, Urioste M, Benitez J, Boix-Chornet M, Sanchez- architecture of gene regulatory variation across multiple human tissues: Aguilera A, Ling C, Carlsson E, Poulsen P, Vaag A, Stephan Z, Spector TD, the MuTHER study. PLoS Genet 2011, 7:e1002003. Wu YZ, Plass C, Esteller M: Epigenetic differences arise during the lifetime 27. Nagarajan M, Veyrieras JB, de Dieuleveult M, Bottin H, Fehrmann S, of monozygotic twins. Proc Natl Acad Sci USA 2005, 102:10604-10609. Abraham AL, Croze S, Steinmetz LM, Gidrol X, Yvert G: Natural single- 10. Sandovici I, Smith NH, Nitert MD, Ackers-Johnson M, Uribe-Lewis S, Ito Y, nucleosome epi-polymorphisms in yeast. PLoS Genet 2010, 6:e1000913. Jones RH, Marquez VE, Cairns W, Tadayyon M, O’Neill LP, Murrell A, Ling C, 28. Parts L, Hedman ÅK, Keildson S, Knights AJ, Abreu-Goodger C, van de Constância M, Ozanne SE: Maternal diet and aging alter the epigenetic Bunt M, Guerra-Assunção JA, Bartonicek N, van Dongen S, Mägi R, Nisbet J, control of a promoter-enhancer interaction at the Hnf4a gene in rat Barrett A, Rantalainen M, Nica AC, Quail MA, Small KS, Glass D, Enright AJ, pancreatic islets. Proc Natl Acad Sci USA 2011, 108:5449-5454. Winn J, MuTHER Consortium, Deloukas P, Dermitzakis ET, McCarthy MI, 11. Waterland RA, Kellermayer R, Laritsky E, Rayco-Solon P, Harris RA, Spector TD, Durbin R, Lindgren CM: Extent, causes, and consequences of Travisano M, Zhang W, Torskaya MS, Zhang J, Shen L, Manary MJ, small RNA expression variation in human adipose tissue. PLoS Genet Prentice AM: Season of conception in rural Gambia affects DNA 2012, 8:e1002704. methylation at putative human metastable epialleles. PLoS Genet 2011, 6: 29. Teschendorff AE, Jones A, Fiegl H, Sargent A, Zhuang JJ, Kitchener HC, e1001252. Widschwendter M: Epigenetic variability in cells of normal cytology is 12. Hoile SP, Lillycrop KA, Thomas NA, Hanson MA, Burdge GC: Dietary protein associated with the risk of future morphological transformation. Genome restriction during F0 pregnancy in rats induces transgenerational Med 2012, 4:24. changes in the hepatic transcriptome in female offspring. PLoS One 2011, 30. Meissner A, Mikkelsen TS, Gu H, Wernig M, Hanna J, Sivachenko A, Zhang X, 6:e21668. Bernstein BE, Nusbaum C, Jaffe DB, Gnirke A, Jaenisch R, Lander ES: 13. Breitling LP, Yang R, Korn B, Burwinkel B, Brenner H: Tobacco-smoking- Genome-scale DNA methylation maps of pluripotent and differentiated related differential DNA methylation: 27K discovery and replication. Am cells. Nature 2008, 454:766-770. J Hum Genet 2011, 88:450-457. Gemma et al. Genome Biology 2013, 14:R43 Page 11 of 11 http://genomebiology.com/2013/14/5/R43 31. Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L, Mohammad HP, Chen W, Daniel VC, Yu W, Berman DM, Jenuwein T, Pruitt K, Sharkis SJ, Watkins DN, Herman JG, Baylin SB: A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat Genet 2007, 39:237-242. 32. Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Zimmerman J, Eden E, Yakhini Z, Ben-Shushan E, Reubinoff BE, Bergman Y, Simon I, Cedar H: Polycomb-mediated methylation on Lys27 of histone H3 pre- marks genes for de novo methylation in cancer. Nat Genet 2007, 39:232-236. 33. Widschwendter M, Fiegl H, Egle D, Mueller-Holzner E, Spizzo G, Marth C, Weisenberger DJ, Campan M, Young J, Jacobs I, Laird PW: Epigenetic stem cell signature in cancer. Nat Genet 2007, 39:157-158. 34. Cuddapah S, Jothi R, Schones DE, Roh TY, Cui K, Zhao K: Global analysis of the insulator binding protein CTCF in chromatin barrier regions reveals demarcation of active and repressive domains. Genome Res 2009, 19:4-32. doi:10.1186/gb-2013-14-5-r43 Cite this article as: Gemma et al.: Inactive or moderately active human promoters are enriched for inter-individual epialleles. Genome Biology 2013 14:R43. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Genome Biology Springer Journals

Inactive or moderately active human promoters are enriched for inter-individual epialleles

Download PDF
11 pages

Loading next page...
 
/lp/springer-journals/inactive-or-moderately-active-human-promoters-are-enriched-for-inter-3VbkJJNXIS

References (68)

Publisher
Springer Journals
Copyright
Copyright © 2013 by Gemma et al.; licensee BioMed Central Ltd.
Subject
Life Sciences; Animal Genetics and Genomics; Human Genetics; Plant Genetics & Genomics; Microbial Genetics and Genomics; Bioinformatics; Evolutionary Biology
eISSN
1474-760X
DOI
10.1186/gb-2013-14-5-r43
pmid
23706135
Publisher site
See Article on Publisher Site

Abstract

Background: Inter-individual epigenetic variation, due to genetic, environmental or random influences, is observed in many eukaryotic species. In mammals, however, the molecular nature of epiallelic variation has been poorly defined, partly due to the restricted focus on DNA methylation. Here we report the first genome-scale investigation of mammalian epialleles that integrates genomic, methylomic, transcriptomic and histone state information. Results: First, in a small sample set, we demonstrate that non-genetically determined inter-individual differentially methylated regions (iiDMRs) can be temporally stable over at least 2 years. Then, we show that iiDMRs are associated with changes in chromatin state as measured by inter-individual differences in histone variant H2A.Z levels. However, the correlation of promoter iiDMRs with gene expression is negligible and not improved by integrating H2A.Z information. We find that most promoter epialleles, whether genetically or non-genetically determined, are associated with low levels of transcriptional activity, depleted for housekeeping genes, and either depleted for H3K4me3/enriched for H3K27me3 or lacking both these marks in human embryonic stem cells. The preferential enrichment of iiDMRs at regions of relative transcriptional inactivity validates in a larger independent cohort, and is reminiscent of observations previously made for promoters that undergo hypermethylation in various cancers, in vitro cell culture and ageing. Conclusions: Our work identifies potential key features of epiallelic variation in humans, including temporal stability of non-genetically determined epialleles, and concomitant perturbations of chromatin state. Furthermore, our work suggests a novel mechanistic link among inter-individual epialleles observed in the context of normal variation, cancer and ageing. Keywords: Epigenetics, DNA methylation, epialleles Background (2) exposure to a compromised in utero environment as Epialleles are genomic loci at which the epigenetic state has been shown in rodent and human studies [10-12]; (3) can stably vary among individuals in a given population or adult life-style associated factors such as smoking [13]. [1]. Although first described and still best understood in Despite these and other previous studies, the molecular plants [2-4], in recent years we have come to realise that nature of mammalian epialleles, in particular those epigenomic landscapes in mammals can also show con- induced by non-genetic factors, has remained controver- siderable inter-individual variation (reviewed in [1,5]). sial [14]. To a large extent, this is due to the DNA methy- lation-focus of previous investigations [5]. Incorporation Mammalian epialleles could arise through the action of cis-or trans-genetic influences [6,7], or have non-genetic of information about the chromatin state would refine origins as a result of: (1) potential stochastic events [8,9]; our understanding of the molecular nature and ultimately functionality of epialleles in the context of normal varia- tion or disease states. * Correspondence: [email protected]; [email protected] † Contributed equally Here we describe the first systematic interrogation of The Blizard Institute, Barts and The London School of Medicine and mammalian epialleles that integrates genomic, methylo- Dentistry, Queen Mary University of London, 4 Newark Street, London E1 mic, transcriptomic and chromatin state information. 2AT, UK Full list of author information is available at the end of the article © 2013 Gemma et al.; licensee 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. Gemma et al. Genome Biology 2013, 14:R43 Page 2 of 11 http://genomebiology.com/2013/14/5/R43 Using a combination of experimental and computational was not feasible to repeat all the functional genomic analyses we identify key features of epiallelic variation in assays on additional MZ pairs. humans, including demonstrating that even non-geneti- cally determined epialleles can be temporally stable, and Identification of temporally stable inter-individual DMRs that DNA methylation variability at epialleles is associated (iiDMRs) with concomitant perturbations in chromatin state. Most An initial low-level analysis confirmed anticipated geno- notably, we find that promoter-associated epiallelic varia- mic/epigenomic correlations. First, we confirmed known tion is predominantly associated with developmentally functional correlations between DNA methylation, H2A.Z important and/or tissue-restricted genes. A similar cate- levels, and gene expression: promoter DNA methylation gory of genes is preferentially hyper-methylated in various was negatively correlated with both H2A.Z and gene cancers, in vitro cellular transformation, and human expression levels, and promoter H2A.Z levels were strongly chronological ageing, potentially pointing to a novel positively correlated with gene expression (Figure 1A). mechanistic link between these processes and human Second, analysis of the SNP arrays did not reveal intra-MZ inter-individual epiallelic variation. pair SNP or copy number variation (approximately 640,000 to 660,000 SNPs observed between unrelated individuals, Results and 104 and 115 SNPs observed in 31/32 and 21/22 MZ Comprehensive genomic/epigenomic/transcriptomic pairs, respectively, and these are most likely false positives). profiling of monozygotic (MZ) twin pairs Finally, DNA methylation profiles were substantially more For the initial discovery phase of the study, we originally similar between co-twins compared with unrelated indivi- aimed to generate integrated genomic/epigenomic/tran- duals (Figure 1B). scriptomic profiles for CD14+ cells from three different We then called inter-individual DMRs (iiDMRs) healthy MZ twin pairs of European ancestry (MZ pair between all five possible pair-wise comparisons (two dif- 31/32: female, age at sampling 27 years; MZ pair 21/22: ferent MZ pairs and one ‘unpaired’ co-twin). Illumina450K female, age at sampling 27 years; MZ pair, male, 11/12, probe sequences were remapped using BLAT, and we only age at sampling 19 years Additional file 1, Table S2). We used probes that mapped exactly once. Furthermore, we focussed on CD14+ cells as they can be obtained to >90% ignored probes on chrX and Y, and also those that over- purity [17] and Additional file 1, Figure S1, and are less lapped known SNPs (based on information provided in likely to harbor post-differentiation, random epigenetic the Illumina 450K annotation file) to eliminate artefacts alterationsastheyhave a lifespan of onlyafew weeks. due to differential probe hybridization effects, resulting in Genetic profiles were obtained using the Illumina a final set of 369,908 different probes. Exclusion of these probes eliminates only cis-acting genetic variants present Omni2.5S array that interrogates approximately 2.5 mil- lion single nucleotide polymorphisms (SNPs) with a in the 50 bp sequence covered by the probe, and all other minor allele frequency of down to 1%. DNA methylation cis- and trans-genetic effects are retained. We considered was assayed by Illumina450K arrays that provide bisulfite only those iiDMRs with a directionally consistent ≥5% conversion-based, single base resolution methylation absolute methylation difference at ≥2 adjacent CpGs measurements at approximately 450,000 different cyto- within 500 bp of each other, as the biological relevance of sines associated with a range of genomic features such as very small methylation differences limited to single CpG promoters, enhancers, and CpG islands (CGIs) [15]. sites is currently unclear. The number of iiDMRs found in Gene expression was profiled using the standard Illumina each of the five different pairwise comparisons is shown in mRNA-seq protocol. For the analysis of chromatin state Table 1. Consistent with Kaminsky et al. [8], iiDMRs were we performed ChIP-seq on the histone variant H2A.Z, found across the genome, but enriched at non-promoter which is strongly associated with transcriptional activity non-exonic regions, and a greater number of iiDMRs were (but can also be found at transcriptionally silent promo- observed between unrelated individuals (Figure 2A). ters), and is thought to be environmentally responsive Furthermore, we generated triplicate Illumina 450K pro- [16]. We obtained 60-80 million mapped 36 bp paired files for CD14+ cells from two additional MZ twin pairs, end reads for the ChIP-seq and RNA-seq libraries. All different to those described above, and found that the genomic/epigenomic/transcriptomic profiles were gener- amount of biological variation seen between twins ated from the same single sampling of CD14+ cells for exceeded the technical variation (false positive iiDMR -7 each individual. Unfortunately, during the course of pro- calls) by a factor of approximately 5 (P <10 , permutation cessing the samples, the DNA sample that was to be used test) (Figure 2B). for subsequent DNA methylation analysis for individual The epigenomic landscape for any given individual is, ‘12’ was inadvertently lost. As these twins were recruited to an extent, constantly in flux, and many iiDMRs iden- because we had methylomic data from a previous time- tified from single time-point measurements probably do point to test for temporal stability, as discussed below, it not have long-term phenotypic consequences. Therefore Gemma et al. Genome Biology 2013, 14:R43 Page 3 of 11 http://genomebiology.com/2013/14/5/R43 -0.5 -1 -1.5 -2 -2.5 -3 -3.5 -4 0 10 20 30 40 50 60 70 80 90 100 Methylation% [Illumina] 0 10 20 30 40 50 60 70 80 90 100 Methylation% [Illumina] -0.2 -0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 -1.8 -2 -2.2 -2.4 0 20 40 60 80 100 H2A.z tag count 0.995 0.99 0.985 0.98 0.975 0.97 31/32 21/22 31/21 31/11 11/22 Figure 1 Epigenomic, genomic and transcriptomic profiling in CD14+ cells from MZ twins. (A) We analysed the following correlations at promoter regions: DNA methylation vs. gene expression, H2A.Z vs. gene expression and DNA methylation vs. H2A.Z. Gene expression levels are represented as log-transformed aligned average FPKM (fragments per kilobase of exon per million fragments mapped) from RNA-seq data. FPKM values were determined by TopHat and Cufflinks and assigned to TSSs. DNA methylation values were determined as b values using Illumina 450K array (see Materials and Methods) and are represented here as percentage of methylation bins. Probes on the Illumina 450K array were assigned to their closest TSS, discarding assignments >1 kb away. H2A.Z abundance is represented as normalised read counts from ChIP-seq -3 -3 data correlates positively with gene expression (Spearman’s r = 0.50, P <10 ), DNA methylation is anti-correlated with H2A.Z (r = -0.59, P <10 ). Promoters were defined as TSS ± 1,000 bp. (B) Correlation coefficients of inter-individual DNA methylation differences. 31/32 and 21/22 plots correspond to MZ twin iiDMRs; 31/21, 31/11 and 11/22 correspond to iiDMRs found in unrelated individuals. Correlation (Spearman’s rho) Mean Expression (log FPKM) Mean Expression (log FPKM) Mean H2A.z tag-count 10 10 122/222 122/221 122/213 122/212 122/211 121/222 121/221 121/213 121/212 121/211 112/222 112/221 112/213 112/212 112/211 111/222 111/221 111/213 111/212 111/211 213/223 212/222 211/221 213/221 212/221 211/221 112/122 112/121 111/122 111/121 221/222 212/213 211/213 211/212 121/122 111/112 Gemma et al. Genome Biology 2013, 14:R43 Page 4 of 11 http://genomebiology.com/2013/14/5/R43 Table 1 Number of iiDMRs identified in the various obtaining some measure of iiDMR temporal stability is pairwise comparisons. critical, especially for non-genetically determined Comparison iiDMRs (n) iiDMRs. The CD14+ cells used in this study were obtained from individuals recruited in late 2010. We 31 vs. 32 5,077 previously sampled the same individuals in mid-2008 as 21 vs. 22 2,528 part of a separate study in which we profiled their CD14 31 vs. 21 9,512 + cells using the Illumina27K array [17]. This array con- 31 vs. 11 9,062 tains probes for 27,578 different CpG sites, largely pro- 11 vs. 22 10,526 moter-associated, and >90% of these CpGs are also Number of iiDMRs for each comparison was calculated as DNA methylation differences ≥5% at ≥2 adjacent CpGs within 500 bp of each other. represented on the Illumina450K array. For maximum Replicates Twins Unrelated 31/32 21/22 31/21 32/22 30 30 30 30 20 20 20 20 10 10 10 10 0 0 0 0 -10 -10 -10 -10 -20 -20 -20 -20 -30 -30 -30 -30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 Methylation% difference [current] Methylation% difference [current] Methylation% difference [current] Methylation% difference [current] Figure 2 Genomic characteristics and temporal stability of inter-individual DMRs (iiDMRs). (A) This plot shows the distribution of iiDMRs across different genomic features (promoters, regulatory regions (that is, Regulatory Features from Ensembl), exons, introns and all other regions (that is, other). iiDMRs were defined as differences >5% in DNA methylation between individuals present in two adjacent probes of the Illumina 450K array. The feint line shows the expected level if iiDMRs were found equally in all regions of the genome. (B) A comparison of technical vs. biological variation for iiDMRs. Shown is the number of iiDMR calls between Illumina 450K array for technical replicates, twin pairs and unrelated individuals. This analysis was done with CD14+ DNA from two MZ twin pairs in triplicate. The biological variation significantly exceeded the -7 technical variation (P <10 , permutation test). (C) Temporal stability of iiDMRs. The inter-individual DNA methylation differences found in this study were compared with the DNA methylation differences present in the same individuals 2 years prior. The previous DNA methylation differences were obtained from CD14+ DNA methylation profiles from the same individuals using Illumina 27K data as part of a separate study [17]. Shown are all current and historic inter-individual methylation differences from all the common probes present on Illumina 27K and Illumina 450K arrays. Methylation% difference [historic] Number of iiDMRs Methylation% difference [historic] Methylation% difference [historic] Methylation% difference [historic] Gemma et al. Genome Biology 2013, 14:R43 Page 5 of 11 http://genomebiology.com/2013/14/5/R43 power when comparing to the much less densely spaced >1 kb away from the promoter (Figure 3A). Therefore, our probes on the Illumina27K array, we considered all >5% data show that a significant number of iiDMRs, even those methylation differences, rather than just those in multi- that are non-genetically determined, represent genomic ple-probe iiDMRs. Comparison of the 2010 with 2008 sites that harbour perturbed chromatin states, not just methylomic profiles, for CpG sites common to both DNA methylation differences, and thus can be classified as platforms, revealed that a substantial proportion of the epialleles [19]. epigenetic variation seen in the 2010 samplings were also Interestingly, we observed only a very weak anti-corre- found in the 2008 samplings, that is, was temporally stable lation between promoter iiDMRs and RNA-seq gene expression levels (Figure 3B). This was true for both (Figure 2C). Although this is not surprising for compari- sons between unrelated individuals, since many of the intra- and inter-MZ pair iiDMRs, and not significantly methylation differences in such cases most likely represent improved even when considering just those iiDMRs that stable genetic effects, temporal stability in the intra-MZ were anti-correlated with H2A.Z, or iiDMRs with a pair comparisons is particularly noteworthy as it means large magnitude, or those found within CpG islands or that random or environmental events can induce perma- CpG shores (Figure 3B). Not surprisingly, a direct com- nent, or at least semi-permanent, epiallelic variation. parison between intra-/inter-MZ pair H2A.Z variation Furthermore, given that the lifespan of CD14+ cells is only and expression did not reveal any significant correla- a few weeks, compared with the approximately 2-year gap tions either (Additional file 1, Figure S4). between the first and second samplings, these epialleles The relationship between promoter DNA methylation most probably arise in blood progenitor cells. For the and gene expression is known to be complex, even in CpGs in common between the second (450K array) and contexts where large DNA methylation differences are first time-points (27K arrays), the proportion of methyla- generally observed, for example, genetically-encoded dif- tion differences found at the second time-point that were ferentiation programs or cancer [5]. In our case, we also present at the first time-point is shown in Additional found it surprising that even inclusion of H2A.Z informa- file 1, Figure S2. The number of sites where a difference in tion did not improve the strength of the correlations as, the same direction is seen at both sites significantly theoretically, integration of information from different exceeds what would be expected by chance in all cases components of the ‘epigenetic’ state of a region should -13 2 (P <3 ×10 , c test). This is the first genome-scale yieldmorerobustcorrelations.To furtherexplore the demonstration of temporally stable epiallelic variation in cause for these observations, we compared mean expres- mammals that cannot potentially be explained by stable sion levels of iiDMR promoter-associated genes with all genetic effects (for example, [18]). For further analyses we genes in our dataset. We found promoter iiDMR-asso- ciated genes to be expressed at significantly lower levels, did not restrict ourselves to CpG sites common to the Illu- mina450K and 27K arrays, otherwise we would have been relative to the other genes, in all intra- and inter-MZ pair left with too few DMRs for any meaningful analyses, but it comparisons (Figure 3C). In other words, it seems that is reasonable to assume that the same degree of temporal promoter-iiDMRs are associated with genes that are stability should be present across the entire set of iiDMRs lowly expressed or silent in CD14+ cells. found from the Illumina450K data obtained from the 2010 These observations were reminiscent of previous results samplings. and our own report on human ageing-associated DNA methylation dynamics [20,21]. In those studies, human iiDMRs at promoters anti-correlate with H2A.Z and promoters that become hypermethylated with chronologi- preferentially associate with lowly expressed cal age (aDMRs) were also associated with genes expressed or silent genes at relatively low levels in the analysed tissue. Notably, We next wanted to investigate whether iiDMRs are asso- these aDMRs were strongly enriched for promoters har- ciated with altered chromatin states or gene expression, bouring bivalent chromatin domains in embryonic stem focussing on promoter-associated iiDMRs as it is currently (ES) cells. Bivalent domains harbour both H3K4me3, gen- not straightforward to correlate the activity of gene distal erally considered an active mark, and H3K27me3, gener- regulatory elements with gene expression levels. Promoters ally considered an inactive mark [22,23]. Furthermore, were defined as a 1 kb window centred at the annotated bivalent domain promoters are associated with develop- transcriptional start site (TSS). We observed a statistically mentally important and tissue-restricted genes [22,23]. significant anti-correlation between DNA methylation and An analysis of previously published H3K4me3 and H2A.Z at both intra- and inter-MZ pair promoter iiDMRs H3K27me3 ChIP-seq profiles in human ES cells [24] (Figure 3A and Additional file 1: Figure S3). This was espe- revealed that promoter iiDMRs were only moderately enriched for bivalent chromatin domains (relative to gen- cially marked for iiDMRs associated with >5% absolute ome average, refer to Figure 3D). Surprisingly though, methylation differences. Significantly, the negative H2A.Z- strong statistically significant enrichment was observed for DNA methylation correlation was observed even at sites Gemma et al. Genome Biology 2013, 14:R43 Page 6 of 11 http://genomebiology.com/2013/14/5/R43 Near TSS >1kb from TSS All data H2A.z in expected direction CpG shores 0.08 0.08 0.1 0.1 0.15 0.1 0.06 0.06 A 0.05 0.05 0.05 0.04 0.04 0 0 0 0.02 0.02 -0.05 -0.05 -0.05 -0.1 0 0 -0.1 -0.1 -0.15 -0.02 -0.02 -0.2 -0.15 -0.15 -0.04 -0.04 -0.25 -0.06 -0.06 -0.2 -0.2 -0.3 31/32 21/22 31/21 31/11 11/22 31/32 21/22 31/21 31/11 11/22 31/32 21/22 31/21 31/11 11/21 31/32 21/22 31/21 31/11 11/21 31/32 21/22 31/21 31/11 11/21 31/32 21/22 31/32 21/22 1 1 DMRs DMRs 0.14 0.14 NonDMR NonDMR 0.9 0.9 E 0.12 0.12 0.8 0.8 0.1 0.1 0.08 0.08 0.7 0.7 0.06 0.06 0.6 0.6 0.04 0.04 0.5 0.5 0.02 0.02 0 0 0.4 0.4 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0.3 0.3 -4 -3 -2 -1 0 1 2 3 4 5 6 -4 -3 -2 -1 0 1 2 3 4 5 6 Log FPKM Log FPKM 10 10 31/21 31/11 31/21 31/11 1 1 0.14 0.14 DMRs DMRs NonDMR NonDMR 0.12 0.12 0.9 0.9 0.1 0.1 0.8 0.8 0.08 0.08 0.7 0.7 0.06 0.06 0.04 0.04 0.6 0.6 0.02 0.02 0.5 0.5 0 0 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0.4 0.4 0.3 0.3 -4 -3 -2 -1 0 1 2 3 4 5 6 -4 -3 -2 -1 0 1 2 3 4 5 6 Log FPKM Log FPKM 10 10 11/22 11/22 0.14 DMRs NonDMR 0.12 0.9 0.1 0.8 0.08 0.06 0.7 0.04 0.6 0.02 0.5 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0.4 0.3 -4 -3 -2 -1 0 1 2 3 4 5 6 Log FPKM Figure 3 Promoter iiDMRs anti-correlate with H2A.Z changes and are preferentially found at genes expressed at low levels. (A) Anti- correlation between inter-individual DNA methylation (for only >5% methylation differences) and H2A.Z differences in CD14+ cells. Data are shown as mean and 95% credible intervals. (B) Inter-individual promoter epiallelic differences are not correlated with gene expression differences. For each bin of inter-individual DNA methylation differences we calculated the ratio of log transformed RNA-seq reads (FPKM) in the first individual of each comparison respect to the other. Data are represented as 95% credible intervals on the mean. The left panel shows the correlations using DNA methylation data only, the middle panel uses those epialleles in which DNA methylation and H2AZ are anti-correlated, and the right panel looks at epiallelic variation at CpG shores (as defined in the Illumina450K array annotation file). (C) iiDMRs are predominantly found at genes expressed at low levels. Shown is the distribution RNA-seq reads (expressed as FPKM) for iiDMRs and non-iiDMRs. In all cases the curves corresponding to DMRs are shifted to the left compared to non-DMRs showing that DMRs occur preferentially at genes expressed at low levels. For all cases 31/32 and 21/22 plots correspond to MZ twin comparisons; 31/21, 31/11 and 11/22 correspond to unrelated individuals comparisons. (D) Promoter iiDMRs occur preferentially at regions depleted for H3K4me3 and enriched for H3K27me3 (K4lo/K27hi) or depleted for both these marks (K4lo/K27lo) in human embryonic stem (hES) cells. hES cell H3K4me3 and H3K27me3 ChIP-seq data are from [24]. The dashed line in each plot refers to the overall iiDMR fraction against the whole dataset. (E) Illumina450K probes were ranked by methylation-ageing correlation using data from [20], and the 500 most- age-associated probes were taken as aDMRs, while 500 randomly selected probes from the remainder of the dataset were taken as controls. For each set, we collected absolute methylation differences from each of the five possible pairings of individuals in this study, and plot 95% credible intervals on the mean. the high H3K27me3/low H3K4me3, or low H3K27me3/ common property between iiDMR and aDMR promoters low H3K4me3 states in ES cells. Both these chromatin seems to be a strong association with genes that are tis- states are also strongly associated with tissue-restricted sue-restricted, but are only moderately active, or inactive, genes and indeed we found iiDMR promoters were signifi- in the analysed tissue. cantly less likely to be associated with house keeping -5 genes (P <10 in all five possible pair-wise comparisons, Analysis of iiDMRs in an independent cohort Chi-squared test). Re-analysis of the 500 most age-corre- To independently validate the findings above, we re-ana- lated probes (that is, aDMRs) from [20] revealed that these lysed DNA methylomic data from our previously pub- probes display significantly more intra- and inter-MZ pair lished human ageing study, that is, the test set [20]. This variability than 500 randomly selected probes (P <0.01, larger and independent dataset consists of Illumina27K bootstrapped for all five comparisons, Figure 3E). So the methylome profiles ofwhole bloodobtainedfrom30 Mean log moderated ratio Cumulative frequency Cumulative frequency Cumulative frequency Mean log moderated ratio Cumulative frequency Cumulative frequency 10 Mean log moderated FPKM ratio Fraction of probes showing >5% difference Fraction of probes showing >5% difference Fraction of probes showing >5% difference K4+ K27+ K4+ K27+ K4+ K27+ K4+ K27- K4+ K27- K4+ K27- K4- K27+ K4- K27+ K4- K27+ K4- K27- K4- K27- K4- K27- Fraction of probes showing >5% difference Fraction of probes showing >5% difference Mean log moderated FPKM ratio K4+ K27+ K4+ K27+ K4+ K27- K4+ K27- K4- K27+ K4- K27+ K4- K27- K4- K27- Mean log moderated FPKM ratio 10 Gemma et al. Genome Biology 2013, 14:R43 Page 7 of 11 http://genomebiology.com/2013/14/5/R43 different healthy female MZ pairs of European ancestry, genetically determined epialleles can be temporally stable ranging in age from 25 to 79 years old [20]. Although the (at least over the course of 2 years). That is, a significant 450K array overall has approximately 15X as many probes fraction of these epialleles are not just transient epige- as the 27K, the majority of these are outside of promoter netic perturbations with little prospect of influencing regions. Of the total annotated protein-coding genes in molecular function. Second, inter-individual DNA the human genome (21,665), 19,409 (89.6%) are associated methylation variants are associated with perturbations of with at least one promoter probe on the 450K array, chromatin state, a relationship observed for even small whereas 14,400 (66.5%) are associated with at least one differences, for example, down to approximately 5% promoter probe on the 27K array. We calculated root methylation difference, and therefore can be considered mean square (RMS) differences for each probe on the Illu- as bona fide epigenetic perturbations. Of course, future mina27K array across the 30 different MZ pairs for both studies using bigger sample numbers are needed to intra- and inter-MZ pair comparisons. The RMS deviation further explore our initial findings. is a measure of the inter-individual methylation variability, The most significant aspect of our study is the finding and is directly proportional to the level of variability that the correlation of iiDMRS with gene expression differ- observed in the cohort under study. RMS difference, as ences is very weak and that iiDMRs are preferentially opposed to mean difference, was used because the differ- found in regions of relative transcriptional inactivity. So ences will be in an arbitrary direction for each pair. It is what are the implications of this? First, it is possible that important to note that the RMS difference is not a mea- some promoter epialleles show inter-related DNA methy- sure of directional age-related changes. Analysis of the test lation and chromatin state perturbations, but may not set resulted in several key conclusions, applicable to both impact significantly on genome function, at least as mea- intra-MZ pair and inter-MZ pair comparisons, validating sured by steady state transcriptional activity. In the case of our findings from the discovery set. First, iiDMR-asso- non-genetically determined epialleles, maybe all promoters ciated promoters found in the test set were associated are potentially subject to epiallelic variation, but the more with genes expressed at significantly lower levels com- active ones are ‘cleared’ of aberrant epigenomic variants, pared with the non-iiDMR set of promoters in the CD14+ whereas the less active/silent promoters can accumulate RNA-seq data generated in our study, and in a previously epigenetic variation. But the enrichment of epialleles in published whole blood array-based expression dataset less active/silent promoters was also found in comparisons (Figure 4A and Additional file 1, Figure S5). Second, the between unrelated individuals. Although it is hard to say test-set iiDMRs were significantly depleted for housekeep- what proportion of epialleles between unrelated indivi- -5 ing genes (P <10 foreitherintra-MZ pairorinter-MZ duals are due to genetic as opposed to environmental dif- ferences from our data, the genetic influence on DNA pair comparisons, Chi-squared test). Third, iiDMRs identi- fied in the test set were significantly enriched at promoters methylation profiles is well documented [3,4,25]. Bell and that harbour either high levels of H3K27me3 and low colleagues measured genome-wide methylation in 77 Hap- levels of H3K4me3, or are devoid of both of these marks Map Yoruba individuals, for which gene expression and in ES cells (Figure 4B). The fact that our findings from genotype data were available, and found a strong genetic CD14+ cells in discovery set were validated by unsorted component to inter-individual variation in DNA methyla- whole blood cell data in the test set further supports the tion profiles [4]. Although they found a significant enrich- robustness of our key conclusions. We also determined ment of SNPs that affect both methylation and gene that for all five possible comparisons in the discovery data- expression, they also noted that the total number of genes set, there’s a strong correlation of both intra- and inter- showing such a signal is only a small proportion of the MZ pair variability between the discovery and test datasets total number of methylation variants they identified [4]. A (Figure 4C). Finally, we also asked if there is an overlap similar conclusion was reached by Myers and colleagues between intra- and inter-MZ pair iiDMRs. This was done who analysed genome-wide methylation in six members of using the test set since it has significantly more pairs. a three generation family and found that only 22% of We found that mostly the same probes show the highest genes harbouring genotype-dependent DNA methylation variability in both intra- and inter-MZ pair comparisons exhibited allele-specific gene expression (albeit more than (Figure 4D). Interestingly, inter-MZ pair comparisons show expected by chance) [25]. Therefore, in both cases the cor- bigger differences at the same sites relative to intra-pair relation between genetically determined DNA methylation comparisons but the great majority of these show greater and expression is at best modest, which would be consis- variability in the between-pair comparisons (Figure 4D). tent with our results regarding chromatin state. It is possible that epiallelic variation acts in a manner Discussion not evident from simple correlations with steady-state expression levels in a given tissue. First it is possible that Our data reveal several novel and important features of these correlations are tissue-restricted as has recently mammalian epialleles. First, we find that even non- Gemma et al. Genome Biology 2013, 14:R43 Page 8 of 11 http://genomebiology.com/2013/14/5/R43 Within pairs Between pairs 0.45 0.45 0.3 A BD 0.4 0.4 0.35 0.35 0.25 0.3 0.3 0.2 0.25 0.25 0.2 0.2 0.15 0.15 0.15 0.1 0.1 0.1 0.05 0.05 0.05 0 0 K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo K4hiK27hi K4hiK27lo K4loK27hi K4loK27lo 0 0.05 0.1 0.15 0.2 0.25 0.3 Within-pair differences (RMS) Figure 4 Validation of iiDMRs in an independent cohort. (A) For this analysis we used 30 MZ twin pairs from [20] whose whole blood DNA methylation profiles were generated by Illumina 27K arrays. We defined low and high expression based on the RNA-seq data we generated in this study from CD14+ cells. High expression: >1 FPKM, Low expression: <1 FPKM. (B) Promoter iiDMRs in the test set are preferentially enriched at regions low in H3K4me3 and high in H3K27me3, and in regions that lack both of these marks in hES cells. The analysis was performed essentially as described in the legend for Figure 3D. (C) Intra- and inter-MZ pair iiDMRs are correlated in the discovery and validation cohorts. For the 30 MZ pairs from [20], we measured intra-pair methylation variability by taking the RMS (root-mean-square) methylation differences for each probe. We also calculated a similar inter-pair variability measure by permuting the pairs. For each of the five possible pairs in the current study, we bin probes by methylation difference, and plot the mean inter- and intra-pair validation methylation variabilities. (D) For the 30 MZ pairs from [20], we calculated intra-pair and inter-pair RMS methylation variability as above and show a scatter plot for all probes on the Illumina27K array (with exclusions as described in the Methods for the Illumina450K data). been shown for genetically determined tissue-restricted gradual hyper-methylation during in vitro cell culture is gene expression [26]. Alternatively, conclusions from two found at promoters associated with genes not expressed recent studies, although not focusing on DNA methyla- in that cell type [30]. Additionally, it has been found in a tion/chromatin state in mammals, hint at other potential variety of human cancers that bivalent chromatin mechanisms by which epialleles could act. Yvert and col- domains (associated with low transcriptional activity in leagues recently compared H3K14 acetylation profiles stem cells) are preferential targets of hyper-methylation between two strains of the yeast Saccharomyces cerevi- [31-33]. The common thread among these seemingly dis- siae, and found 5,442 sites that significantly differed in parate examples of inter-individual epigenetic variation is H3K14ac levels, which they called single nucleosome epi- promoters that are developmentally regulated and tissue- restricted, and are only moderately active, or inactive, in polymorphisms (SNEPs) [27]. However, higher acetyla- the analysed tissue. We propose that there could be a tion in one strain did not always mean higher expression of the relevant gene, for example, in one case the SNEP potentially important mechanistic link between normal/ was associated with the strength of gene activation upon stochastic epiallelic variation and the epigenetic perturba- stimulation by heat shock. Secondly, Lindgren and collea- tions observed in the context of cancer and ageing. gues recently assessed the effect of naturally occurring variation in miRNA expression levels on mRNA levels in Conclusions humans, but found little correlation [28]. The authors The existence of mammalian epialleles is not in doubt, concluded that their findings were more consistent with but the key challenge now is to characterise epialleles at the primary role of miRNAs being to buffer mRNA the molecular level. Our work reveals key and novel levels. A key conclusion therefore is that correlating properties of epiallelic variation in humans, and further epialleles with steady-state RNA dynamics, possibly the suggests important mechanistic links between normal most common analysis currently presented in papers on inter-individual epigenetic variation and epigenetic per- epiallelic investigations, may not be particularly fruitful. turbations observed in cancer and chronological ageing. Finally, and potentially most importantly, the broadly similar characteristics of iiDMRs and aDMRs (from our Materials and methods previous study [20] and [29]) may in fact be a general fea- Samples ture of mammalian epiallelic variation in a variety of con- Fresh venous blood was obtained from three pairs (six texts. Meissner and colleagues found that aberrant individuals) of healthy MZ twins (Additional file 1, Fraction of probes showing >5% RMS diff. Fraction of probes showing >5% RMS diff. Between-pair differences (RMS) Gemma et al. Genome Biology 2013, 14:R43 Page 9 of 11 http://genomebiology.com/2013/14/5/R43 Table S1). Blood was diluted 1:1 in RPMI media and theIllumina arrayand used that as aChIPscore for then peripheral blood mononuclear cells (PBMCs) were that probe. RNA-seq reads were mapped to the refer- separated by Ficoll-Hypaque gradient centrifugation. ence genome using Tophat 1.3.1, then expression levels CD14+ cells were isolated according to manufacturer’s (FPKM) were estimated for each Ensembl transcript instruction using magnetic bead-based positive selection using Cufflinks 1.0.3. For analyses comparing methyla- system (Miltenyi Biotech). The purity of the cells was tion data to expression, methylation array probes lying determined by FACS using CD14-FITC antibodies within 1 kb of an Ensembl TSS were assigned an (Additional file 1, Figure S1). All subjects gave informed ‘expression level’ equal to that of the transcript asso- consent and the study was approved by the Northern ciated with the nearest TSS. and Yorkshire Research Ethics committee (REC Refer- Statistical analyses ence Number: 06-MREO-3-22). Validation of iiDMRs Correlation between variables was performed using was done using whole blood Illumina27K data pre- Spearman’s rank test. Confidence intervals for all box/bar viously generated [20]. This cohort included 30 different plots are obtained by bootstrapping unless otherwise sta- healthy MZ female twin pairs recruited from within the ted. Confidence intervals for the hES cell H3K4me3/ UK as part of the TwinsUK registry. H3K27me3 bar charts are estimated from a binomial model. Probes associated with housekeeping genes were Illumina 450K array defined as in [20]. A total of 500 ng of DNA from CD14+ cells isolated using For the genomic location enrichment analyses, exon, QIAamp DNA Mini Kit was bisulfite converted using the intron and regulatory features were extracted from EZ DNA Methylation kit (Zymo Research). Arrays were Ensembl, and promoters were defined as regions within processed at the Barts and The London Genome Centre, 1 kb of the TSSs of an Ensembl gene. For each of these London, UK according to the manufacturer’s recommen- categories, we asked what fraction of the probes lying in dations. Methylation scores for each CpG site are called as the selected regions were called as iiDMRs, and plot ‘Beta’ values (using BeadStudio software from Illumina), 95% confidence intervals on this proportion, estimated that range from 0 (unmethylated (U)) to 1 (fully methy- using a binomial model. For comparison, the feint line lated (M)) on a continuous scale, and are calculated from indicates the fraction of iiDMRs across the whole data- the intensity of the M and U alleles as the ratio of fluores- set, allowing enrichment or depletion to be assessed. cent signals. Accession codes Illumina Omni2.5S arrays All data are available on GEO [GSE46220]. The arrays were processed according to the manufac- turer’s instructions using 500 ng of DNA. Additional material ChIP- and RNA-seq The chromatin immunoprecipitation (ChIP) assay was Additional file 1: Additional Materials and methods, Tables S1 and S2, and Figures S1 to S5. performed on 5 × 10 CD14+ cells according to pre- viously published protocols with minor modifications [34]. Chromatin was sonicated to get fragments of 100 to 500 bp and immunoprecipitated with 10 uL of anti-H2A. List of abbreviations aDMR: ageing-associated differentially methylated region; CGI: CpG island; Z antibody (Active Motif, Cat no: 39113). ChIP-seq ChIP-seq: chromatin immunoprecipitation; ESC: embryonic stem cells; FPKM: libraries were prepared following the Illumina protocol fragments per kilobase of exon per million fragments mapped; iiDMR: inter- and ligated to standard PE adaptors and sequenced on an individual differentially methylated region; MZ: monozygotic; PBMC: peripheral blood mononuclear cells; RMS: root mean square; SNEP: single Illumina GAIIx instrument. For RNA-seq, 200 ng of total nucleosome epi-polymorphism; SNP: single nucleotide polymorphism; TSS: RNA was used to prepare RNA-seq libraries using the transcriptional start site. TruSeq RNA kit from Illumina following the instructions Competing interests provided in the supplier’s manual, and sequenced on an The authors declare that they have no competing interests. Illumina GAIIx instrument. Authors’ contributions CG performed the majority of the experimental work and helped to draft Sequence data processing the manuscript. TAD performed the bioinformatics analyses. HB and MIH ChIP-seq reads were mapped to the GRCh37 (hg19) collected and processed blood samples from the MZ twin pair participants. reference genome sequence using MAQ 0.6.6 and map- MLH assisted CG with various aspects of the experimental work. PJH assisted with the ChIP-seq. GG, RDL and GCE contributed reagents and materials. SVR pings with quality scores <10 were discarded. For and VKR conceived and designed the study, participated in its design, iiDMR-centric analyses, we counted the numbered of coordination and analysis, and helped to draft the manuscript. All authors paired end fragments overlapping each probe region on read and approved the final manuscript. Gemma et al. Genome Biology 2013, 14:R43 Page 10 of 11 http://genomebiology.com/2013/14/5/R43 14. Bird A: Putting the DNA back into DNA methylation. Nat Genet 2011, Acknowledgements 43:1050-1051. VKR, CG and RDL are supported by the BBSRC, UK (BB/H012494/1). VKR and 15. Sandoval J, Heyn H, Moran S, Serra-Musach J, Pujana MA, Bibikova M, RDL are also supported by the EU-FP7 ‘BLUEPRINT’ program (282510). RDL is Esteller M: Validation of a DNA methylation microarray for 450,000 CpG also supported by Juvenile Diabetes Research Foundation International sites in the human genome. Epigenetics 2011, 6:692-702. (JDRFI Award 5-2011-145). HB was supported by EFSD/Novo Nordisk 16. Conerly ML, Teves SS, Diolaiti D, Ulrich M, Eisenman RN, Henikoff S: Program Grant and Diabetes UK (10/0004107). SVR and GE are funded by Changes in H2A.Z occupancy and DNA methylation during B-cell the Multiple Sclerosis Society of the United Kingdom. SVR and GG are lymphomagenesis. Genome Res 2010, 20:1383-1390. funded by the Medical Research Council of the United Kingdom (G0801976). 17. Rakyan VK, Beyan H, Down TA, Hawa MI, Maslau S, Aden D, Daunay A, TAD is funded by the Wellcome Trust (083563). Busato F, Mein CA, Manfras B, Dias KR, Bell CG, Tost J, Boehm BO, Beck S, Leslie RD: Identification of type 1 diabetes-associated DNA methylation Authors’ details The Blizard Institute, Barts and The London School of Medicine and variable positions that precede disease diagnosis. PLoS Genet 2011, 7: Dentistry, Queen Mary University of London, 4 Newark Street, London E1 e1002300. 2AT, UK. Department of Physiology, Anatomy and Genetics and Medical 18. Feinberg AP, Irizarry RA, Fradin D, Aryee MJ, Murakami P, Aspelund T, Research Council Functional Genomics Unit, South Parks Road, Oxford, OX1 Eiriksdottir G, Harris TB, Launer L, Gudnason V, Fallin MD: Personalized 3PT, UK. The Gurdon Institute and Department of Genetics, University of epigenomic signatures that are stable over time and covary with body Cambridge, Tennis Court Road, Cambridge CB2 1QN, UK. School of mass index. Sci Transl Med 2010, 2:49ra67. Biological and Chemical Sciences, Queen Mary University of London, 19. Finer S, Holland ML, Nanty L, Rakyan VK: The hunt for the epiallele. Environ London, E1 4NS, UK. Wellcome Trust Centre for Human Genetics, University Mol Mutagen 2011, 52:1-11. of Oxford, Headington, Oxford OX3 7BN, UK. 20. Rakyan VK, Down TA, Maslau S, Andrew T, Yang TP, Beyan H, Whittaker P, McCann OT, Finer S, Valdes AM, Leslie RD, Deloukas P, Spector TD: Human Received: 9 July 2012 Revised: 24 April 2013 Accepted: 25 May 2013 aging-associated DNA hypermethylation occurs preferentially at bivalent Published: 25 May 2013 chromatin domains. Genome Res 2010, 20:434-439. 21. Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, Weisenberger DJ, Shen H, Campan M, Noushmehr H, Bell CG, Maxwell AP, Savage DA, References Mueller-Holzner E, Marth C, Kocjan G, Gayther SA, Jones A, Beck S, 1. Richards EJ: Population epigenetics. Curr Opin Genet Dev 2008, 18:221-226. Wagner W, Laird PW, Jacobs IJ, Widschwendter M: Age-dependent DNA 2. Miura K, Agetsuma M, Kitano H, Yoshimura A, Matsuoka M, Jacobsen SE, methylation of genes that are suppressed in stem cells is a hallmark of Ashikari M: A metastable DWARF1 epigenetic mutant affecting plant cancer. Genome Res 2010, 20:440-446. stature in rice. Proc Natl Acad Sci USA 2009, 106:11218-11223. 22. Bernstein BE, Mikkelsen TS, Xie X, Kamal M, Huebert DJ, Cuff J, Fry B, 3. Schmitz RJ, Schultz MD, Lewsey MG, O’Malley RC, Urich MA, Libiger O, Meissner A, Wernig M, Plath K, Jaenisch R, Wagschal A, Feil R, Schreiber SL, Schork NJ, Ecker JR: Transgenerational epigenetic instability is a source of Lander ES: A bivalent chromatin structure marks key developmental novel methylation variants. Science 2011, 334:369-373. genes in embryonic stem cells. Cell 2006, 125:315-326. 4. Becker C, Hagmann J, Müller J, Koenig D, Stegle O, Borgwardt K, Weigel D: 23. Azuara V, Perry P, Sauer S, Spivakov M, Jørgensen HF, John RM, Gouti M, Spontaneous epigenetic variation in the Arabidopsis thaliana Casanova M, Warnes G, Merkenschlager M, Fisher AG: Chromatin methylome. Nature 2011, 480:245-249. signatures of pluripotent cell lines. Nat Cell Biol 2006, 8:532-538. 5. Rakyan VK, Down TA, Balding DJ, Beck S: Epigenome-wide association 24. Zhao XD, Han X, Chew JL, Liu J, Chiu KP, Choo A, Orlov YL, Sung WK, studies for common human diseases. Nat Rev Genet 2011, 12:529-541. Shahab A, Kuznetsov VA, Bourque G, Oh S, Ruan Y, Ng HH, Wei CL: Whole- 6. Zhang D, Cheng L, Badner JA, Chen C, Chen Q, Luo W, Craig DW, genome mapping of histone H3 Lys4 and 27 trimethylations reveals Redman M, Gershon ES, Liu C: Genetic control of individual differences in distinct genomic compartments in human embryonic stem cells. Cell gene-specific methylation in human brain. Am J Hum Genet 2010, Stem Cell 2007, 1:286-298. 86:411-419. 25. Gertz J, Varley KE, Reddy TE, Bowling KM, Pauli F, Parker SL, Kucera KS, 7. Bell JT, Pai AA, Pickrell JK, Gaffney DJ, Pique-Regi R, Degner JF, Gilad Y, Willard HF, Myers RM: Analysis of DNA methylation in a three-generation Pritchard JK: DNA methylation patterns associate with genetic and gene family reveals widespread genetic influence on epigenetic regulation. expression variation in HapMap cell lines. Genome Biol 2011, 12:R10-R16. PLoS Genet 2011, 7:e1002228. 8. Kaminsky ZA, Tang T, Wang SC, Ptak C, Oh GH, Wong AH, Feldcamp LA, 26. Nica AC, Parts L, Glass D, Nisbet J, Barrett A, Sekowska M, Travers M, Virtanen C, Halfvarson J, Tysk C, McRae AF, Visscher PM, Montgomery GW, Potter S, Grundberg E, Small K, Hedman AK, Bataille V, Tzenova Bell J, Gottesman II, Martin NG, Petronis A: DNA methylation profiles in Surdulescu G, Dimas AS, Ingle C, Nestle FO, di Meglio P, Min JL, Wilk A, monozygotic and dizygotic twins. Nature Genet 2009, 41:240-245. Hammond CJ, Hassanali N, Yang TP, Montgomery SB, O’Rahilly S, 9. Fraga MF, Ballestar E, Paz MF, Ropero S, Setien F, Ballestar ML, Heine- Lindgren CM, Zondervan KT, Soranzo N, Barroso I, Durbin R, et al: The Suñer D, Cigudosa JC, Urioste M, Benitez J, Boix-Chornet M, Sanchez- architecture of gene regulatory variation across multiple human tissues: Aguilera A, Ling C, Carlsson E, Poulsen P, Vaag A, Stephan Z, Spector TD, the MuTHER study. PLoS Genet 2011, 7:e1002003. Wu YZ, Plass C, Esteller M: Epigenetic differences arise during the lifetime 27. Nagarajan M, Veyrieras JB, de Dieuleveult M, Bottin H, Fehrmann S, of monozygotic twins. Proc Natl Acad Sci USA 2005, 102:10604-10609. Abraham AL, Croze S, Steinmetz LM, Gidrol X, Yvert G: Natural single- 10. Sandovici I, Smith NH, Nitert MD, Ackers-Johnson M, Uribe-Lewis S, Ito Y, nucleosome epi-polymorphisms in yeast. PLoS Genet 2010, 6:e1000913. Jones RH, Marquez VE, Cairns W, Tadayyon M, O’Neill LP, Murrell A, Ling C, 28. Parts L, Hedman ÅK, Keildson S, Knights AJ, Abreu-Goodger C, van de Constância M, Ozanne SE: Maternal diet and aging alter the epigenetic Bunt M, Guerra-Assunção JA, Bartonicek N, van Dongen S, Mägi R, Nisbet J, control of a promoter-enhancer interaction at the Hnf4a gene in rat Barrett A, Rantalainen M, Nica AC, Quail MA, Small KS, Glass D, Enright AJ, pancreatic islets. Proc Natl Acad Sci USA 2011, 108:5449-5454. Winn J, MuTHER Consortium, Deloukas P, Dermitzakis ET, McCarthy MI, 11. Waterland RA, Kellermayer R, Laritsky E, Rayco-Solon P, Harris RA, Spector TD, Durbin R, Lindgren CM: Extent, causes, and consequences of Travisano M, Zhang W, Torskaya MS, Zhang J, Shen L, Manary MJ, small RNA expression variation in human adipose tissue. PLoS Genet Prentice AM: Season of conception in rural Gambia affects DNA 2012, 8:e1002704. methylation at putative human metastable epialleles. PLoS Genet 2011, 6: 29. Teschendorff AE, Jones A, Fiegl H, Sargent A, Zhuang JJ, Kitchener HC, e1001252. Widschwendter M: Epigenetic variability in cells of normal cytology is 12. Hoile SP, Lillycrop KA, Thomas NA, Hanson MA, Burdge GC: Dietary protein associated with the risk of future morphological transformation. Genome restriction during F0 pregnancy in rats induces transgenerational Med 2012, 4:24. changes in the hepatic transcriptome in female offspring. PLoS One 2011, 30. Meissner A, Mikkelsen TS, Gu H, Wernig M, Hanna J, Sivachenko A, Zhang X, 6:e21668. Bernstein BE, Nusbaum C, Jaffe DB, Gnirke A, Jaenisch R, Lander ES: 13. Breitling LP, Yang R, Korn B, Burwinkel B, Brenner H: Tobacco-smoking- Genome-scale DNA methylation maps of pluripotent and differentiated related differential DNA methylation: 27K discovery and replication. Am cells. Nature 2008, 454:766-770. J Hum Genet 2011, 88:450-457. Gemma et al. Genome Biology 2013, 14:R43 Page 11 of 11 http://genomebiology.com/2013/14/5/R43 31. Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L, Mohammad HP, Chen W, Daniel VC, Yu W, Berman DM, Jenuwein T, Pruitt K, Sharkis SJ, Watkins DN, Herman JG, Baylin SB: A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat Genet 2007, 39:237-242. 32. Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Zimmerman J, Eden E, Yakhini Z, Ben-Shushan E, Reubinoff BE, Bergman Y, Simon I, Cedar H: Polycomb-mediated methylation on Lys27 of histone H3 pre- marks genes for de novo methylation in cancer. Nat Genet 2007, 39:232-236. 33. Widschwendter M, Fiegl H, Egle D, Mueller-Holzner E, Spizzo G, Marth C, Weisenberger DJ, Campan M, Young J, Jacobs I, Laird PW: Epigenetic stem cell signature in cancer. Nat Genet 2007, 39:157-158. 34. Cuddapah S, Jothi R, Schones DE, Roh TY, Cui K, Zhao K: Global analysis of the insulator binding protein CTCF in chromatin barrier regions reveals demarcation of active and repressive domains. Genome Res 2009, 19:4-32. doi:10.1186/gb-2013-14-5-r43 Cite this article as: Gemma et al.: Inactive or moderately active human promoters are enriched for inter-individual epialleles. Genome Biology 2013 14:R43. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit

Journal

Genome Biology – Springer Journals

Published: May 25, 2013

There are no references for this article.