Patterns of methylation heritability in a genome-wide analysis of four brain regions

Investigating the prediction of CpG methylation levels from SNP genotype data to help elucidate relationships between methylation, gene expression and complex traits

As popularised by PrediXcan (and related methods), transcriptome-wide association studies (TWAS), in which gene expression is imputed from single-nucleotide polymorphism (SNP) genotypes and tested for association with a phenotype, are a popular approach for investigating the role of gene expression in complex traits. Like gene expression, DNA methylation is an important biological process and, being under genetic regulation, may be imputable from SNP genotypes. Here, we investigate prediction of CpG methylation levels from SNP genotype data to help elucidate relationships between methylation, gene expression and complex traits. We start by examining how well CpG methylation can be predicted from SNP genotypes, comparing three penalised regression approaches and examining whether changing the window size improves prediction accuracy. Although methylation at most CpG sites cannot be accurately predicted from SNP genotypes, for a subset it can be predicted well. We next apply our methylation prediction models (trained using the optimal method and window size) to carry out a methylome-wide association study (MWAS) of primary biliary cholangitis. We intersect the regions identified via MWAS with those identified via TWAS, providing insight into the interplay between CpG methylation, gene expression and disease status. We conclude that MWAS has the potential to improve understanding of biological mechanisms in complex traits.

Elevation of DNA Methylation in the Promoter Regions of the Brain-Derived Neurotrophic Factor Gene is Associated with Heroin Addiction

To study the potential role of brain-derived neurotrophic factor (BDNF) methylation in heroin addiction, we first detected the methylation level of seven CpG islands that included 106 CpG sites in the promoter regions of BDNF from 120 people addicted to heroin and 113 controls. Methylation quantitative trait locus (mQTL) analysis was then employed to determine the association between the single-nucleotide polymorphism rs6265, a well-known locus shown to be correlated with heroin addiction, and the methylation levels of these CpG sites. Finally, we used the JASPAR database to predict whether transcription factors could bind to these CpG sites. We found that the methylation levels of CpG islands 6 and 7 and the methylation levels of BDNF_45 and BDNF_80 were significantly higher in the heroin addiction group than in the control group. We also found that rs6265 was an mQTL and was associated with the methylation level of BDNF_58. Using the JASPAR database, we found that ALX homeobox 3 (ALX3), achaete-scute family bHLH transcription factor 1 (ASCL1) and aryl hydrocarbon receptor nuclear translocator 2 (ARNT2) could bind to CpG island 6, and ALX3 could bind to CpG island 7. In summary, we showed that increased DNA methylation in the promoter regions of the BDNF gene was associated with heroin addiction in Han Chinese.

Genetic impacts on DNA methylation: research findings and future perspectives

Multiple recent studies highlight that genetic variants can have strong impacts on a significant proportion of the human DNA methylome. Methylation quantitative trait loci, or meQTLs, allow for the exploration of biological mechanisms that underlie complex human phenotypes, with potential insights for human disease onset and progression. In this review, we summarize recent milestones in characterizing the human genetic basis of DNA methylation variation over the last decade, including heritability findings and genome-wide identification of meQTLs. We also discuss challenges in this field and future areas of research geared to generate insights into molecular processes underlying human complex traits.

Role of DNA Methylation in Mediating Genetic Risk of Psychiatric Disorders

Psychiatric disorders are common, complex, and heritable conditions estimated to be the leading cause of disability worldwide. The last decade of research in genomics of psychiatry, performed by multinational, and multicenter collaborative efforts on hundreds of thousands of mental disorder cases and controls, provided invaluable insight into the genetic risk variants of these conditions. With increasing cohort sizes, more risk variants are predicted to be identified in the near future, but there appears to be a knowledge gap in understanding how these variants contribute to the pathophysiology of psychiatric disorders. Majority of the identified common risk single-nucleotide polymorphisms (SNPs) are non-coding but are enriched in regulatory regions of the genome. It is therefore of great interest to study the impact of identified psychiatric disorders' risk SNPs on DNA methylation, the best studied epigenetic modification, playing a pivotal role in the regulation of transcriptomic processes, brain development, and functioning. This work outlines the mechanisms through which risk SNPs can impact DNA methylation levels and provides a summary of current evidence on the role of DNA methylation in mediating the genetic risk of psychiatric disorders.

DNA methylation and cis-regulation of gene expression by prostate cancer risk SNPs

Genome-wide association studies have identified more than 100 SNPs that increase the risk of prostate cancer (PrCa). We identify and compare expression quantitative trait loci (eQTLs) and CpG methylation quantitative trait loci (meQTLs) among 147 established PrCa risk SNPs in primary prostate tumors (n = 355 from a Seattle-based study and n = 495 from The Cancer Genome Atlas, TCGA) and tumor-adjacent, histologically benign samples (n = 471 from a Mayo Clinic study). The role of DNA methylation in eQTL regulation of gene expression was investigated by data triangulation using several causal inference approaches, including a proposed adaptation of the Causal Inference Test (CIT) for causal direction. Comparing eQTLs between tumors and benign samples, we show that 98 of the 147 risk SNPs were identified as eQTLs in the tumor-adjacent benign samples, and almost all 34 eQTL identified in tumor sets were also eQTLs in the benign samples. Three lines of results support the causal role of DNA methylation. First, nearly 100 of the 147 risk SNPs were identified as meQTLs in one tumor set, and almost all eQTLs in tumors were meQTLs. Second, the loss of eQTLs in tumors relative to benign samples was associated with altered DNA methylation. Third, among risk SNPs identified as both eQTLs and meQTLs, mediation analyses suggest that over two-thirds have evidence of a causal role for DNA methylation, mostly mediating genetic influence on gene expression. In summary, we provide a comprehensive catalog of eQTLs, meQTLs and putative cancer genes for known PrCa risk SNPs. We observe that a substantial portion of germline eQTL regulatory mechanisms are maintained in the tumor development, despite somatic alterations in tumor genome. Finally, our mediation analyses illuminate the likely intermediary role of CpG methylation in eQTL regulation of gene expression.

Characterization of DNA methylation-based markers for human body fluid identification in forensics: a critical review

Body fluid identification in crime scene investigations aids in reconstruction of crime scenes. Several studies have identified and reported differentially methylated sites (DMSs) and regions (DMRs) which differ between forensically relevant tissues (tDMRs) and body fluids. Diverse factors affect methylation patterns such as the environment, diets, lifestyle, disease, ethnicity, genetic variation, amongst others. Thus, it is important to analyse the stability of markers employed for forensic identification. Furthermore, even though epigenetic modifications are described as stable and heritable, epigenetic inheritance of potential markers for body fluid identification needs to be assessed in the long term. Here, we discuss the current status of reported DNA methylation-based markers and their verification studies. Such thorough investigation is crucial to develop a stable panel of DNA methylation-based markers for accurate body fluid identification.

Identification of 55,000 Replicated DNA Methylation QTL

DNA methylation plays an important role in the regulation of transcription. Genetic control of DNA methylation is a potential candidate for explaining the many identified SNP associations with disease that are not found in coding regions. We replicated 52,916 cis and 2,025 trans DNA methylation quantitative trait loci (mQTL) using methylation from whole blood measured on Illumina HumanMethylation450 arrays in the Brisbane Systems Genetics Study (n = 614 from 177 families) and the Lothian Birth Cohorts of 1921 and 1936 (combined n = 1366). The trans mQTL SNPs were found to be over-represented in 1 Mbp subtelomeric regions, and on chromosomes 16 and 19. There was a significant increase in trans mQTL DNA methylation sites in upstream and 5′ UTR regions. The genetic heritability of a number of complex traits and diseases was partitioned into components due to mQTL and the remainder of the genome. Significant enrichment was observed for height (p = 2.1 × 10⁻¹⁰), ulcerative colitis (p = 2 × 10⁻⁵), Crohn’s disease (p = 6 × 10⁻⁸) and coronary artery disease (p = 5.5 × 10⁻⁶) when compared to a random sample of SNPs with matched minor allele frequency, although this enrichment is explained by the genomic location of the mQTL SNPs.

Detecting heritable phenotypes without a model using fast permutation testing for heritability and set-tests

Testing for association between a set of genetic markers and a phenotype is a fundamental task in genetic studies. Standard approaches for heritability and set testing strongly rely on parametric models that make specific assumptions regarding phenotypic variability. Here, we show that resulting p-values may be inflated by up to 15 orders of magnitude, in a heritability study of methylation measurements, and in a heritability and expression quantitative trait loci analysis of gene expression profiles. We propose FEATHER, a method for fast permutation-based testing of marker sets and of heritability, which properly controls for false-positive results. FEATHER eliminated 47% of methylation sites found to be heritable by the parametric test, suggesting a substantial inflation of false-positive findings by alternative methods. Our approach can rapidly identify heritable phenotypes out of millions of phenotypes acquired via high-throughput technologies, does not suffer from model misspecification and is highly efficient.

Standard approaches for heritability and set testing in statistical genetics rely on parametric models that might not hold in reality and give inflated p-values. Here, the authors develop a fast method for permutation-based testing of marker sets and of heritability that does not suffer from model misspecification.

A Bayesian mixed modeling approach for estimating heritability

BACKGROUND

A Bayesian mixed model approach using integrated nested Laplace approximations (INLA) allows us to construct flexible models that can account for pedigree structure. Using these models, we estimate genome-wide patterns of DNA methylation heritability (h 2 ), which are currently not well understood, as well as h 2 of blood lipid measurements.

METHODS

We included individuals from the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study with Infinium 450 K cytosine-phosphate-guanine (CpG) methylation and blood lipid data pre- and posttreatment with fenofibrate in families with up to three-generation pedigrees. For genome-wide patterns, we constructed 1 model per CpG with methylation as the response variable, with a random effect to model kinship, and age and gender as fixed effects.

RESULTS

In total, 425,791 CpG sites pre-, but only 199,027 CpG sites posttreatment were found to have nonzero heritability. Across these CpG sites, the distributions of h 2 estimates are similar in pre- and posttreatment (pre: median = 0.31, interquartile range [IQR] = 0.16; post: median = 0.34, IQR = 0.20). Blood lipid h 2 estimates were similar pre- and posttreatment with overlapping 95% credibility intervals. Heritability was nonzero for treatment effect, that is, the difference between pre- and posttreatment blood lipids. Estimates for triglycerides h 2 are 0.48 (pre), 0.42 (post), and 0.21 (difference); likewise for high-density lipoprotein cholesterol h 2 the estimates are 0.61, 0.68, and 0.10.

CONCLUSIONS

We show that with INLA, a fully Bayesian approach to estimate DNA methylation h 2 is possible on a genome-wide scale. This provides uncertainty assessment of the estimates, and allows us to perform model selection via deviance information criterion (DIC) to identify CpGs with strong evidence for nonzero heritability.

Characterization of cross-tissue genetic-epigenetic effects and their patterns in schizophrenia

Background

One of the major challenges in current psychiatric epigenetic studies is the tissue specificity of epigenetic changes since access to brain samples is limited. Peripheral tissues have been studied as surrogates but the knowledge of cross-tissue genetic-epigenetic characteristics remains largely unknown. In this work, we conducted a comprehensive investigation of genetic influence on DNA methylation across brain and peripheral tissues with the aim to characterize cross-tissue genetic-epigenetic effects and their roles in the pathophysiology of psychiatric disorders.

Methods

Genome-wide methylation quantitative trait loci (meQTLs) from brain prefrontal cortex, whole blood, and saliva were identified separately and compared. Focusing on cis-acting effects, we tested the enrichment of cross-tissue meQTLs among cross-tissue expression QTLs and genetic risk loci of various diseases, including major psychiatric disorders. CpGs targeted by cross-tissue meQTLs were also tested for genomic distribution and functional enrichment as well as their contribution to methylation correlation across tissues. Finally, a consensus co-methylation network analysis on the cross-tissue meQTL targeted CpGs was performed on data of the three tissues collected from schizophrenia patients and controls.

Results

We found a significant overlap of cis meQTLs (45–73 %) and targeted CpG sites (31–68 %) among tissues. The majority of cross-tissue meQTLs showed consistent signs of cis-acting effects across tissues. They were significantly enriched in genetic risk loci of various diseases, especially schizophrenia, and also enriched in cross-tissue expression QTLs. Compared to CpG sites not targeted by any meQTLs, cross-tissue targeted CpGs were more distributed in CpG island shores and enhancer regions, and more likely had strong correlation with methylation levels across tissues. The targeted CpGs were also annotated to genes enriched in multiple psychiatric disorders and neurodevelopment-related pathways. Finally, we identified one co-methylation network shared between brain and blood showing significant schizophrenia association (p = 5.5 × 10⁻⁶).

Conclusions

Our results demonstrate prevalent cross-tissue meQTL effects and their contribution to the correlation of CpG methylation across tissues, while at the same time a large portion of meQTLs show tissue-specific characteristics, especially in brain. Significant enrichment of cross-tissue meQTLs in expression QTLs and genetic risk loci of schizophrenia suggests the potential of these cross-tissue meQTLs for studying the genetic effect on schizophrenia. The study provides compelling motivation for a well-designed experiment to further validate the use of surrogate tissues in the study of psychiatric disorders.

Electronic supplementary material

The online version of this article (10.1186/s13073-018-0519-4) contains supplementary material, which is available to authorized users.

Genome-wide mapping of genetic determinants influencing DNA methylation and gene expression in human hippocampus

Emerging evidence emphasizes the strong impact of regulatory genomic elements in neurodevelopmental processes and the complex pathways of brain disorders. The present genome-wide quantitative trait loci analyses explore the cis-regulatory effects of single-nucleotide polymorphisms (SNPs) on DNA methylation (meQTL) and gene expression (eQTL) in 110 human hippocampal biopsies. We identify cis-meQTLs at 14,118 CpG methylation sites and cis-eQTLs for 302 3'-mRNA transcripts of 288 genes. Hippocampal cis-meQTL-CpGs are enriched in flanking regions of active promoters, CpG island shores, binding sites of the transcription factor CTCF and brain eQTLs. Cis-acting SNPs of hippocampal meQTLs and eQTLs significantly overlap schizophrenia-associated SNPs. Correlations of CpG methylation and RNA expression are found for 34 genes. Our comprehensive maps of cis-acting hippocampal meQTLs and eQTLs provide a link between disease-associated SNPs and the regulatory genome that will improve the functional interpretation of non-coding genetic variants in the molecular genetic dissection of brain disorders.

RL-SKAT: An Exact and Efficient Score Test for Heritability and Set Tests

Testing for the existence of variance components in linear mixed models is a fundamental task in many applicative fields. In statistical genetics, the score test has recently become instrumental in the task of testing an association between a set of genetic markers and a phenotype. With few markers, this amounts to set-based variance component tests, which attempt to increase power in association studies by aggregating weak individual effects. When the entire genome is considered, it allows testing for the heritability of a phenotype, defined as the proportion of phenotypic variance explained by genetics. In the popular score-based Sequence Kernel Association Test (SKAT) method, the assumed distribution of the score test statistic is uncalibrated in small samples, with a correction being computationally expensive. This may cause severe inflation or deflation of P-values, even when the null hypothesis is true. Here, we characterize the conditions under which this discrepancy holds, and show it may occur also in large real datasets, such as a dataset from the Wellcome Trust Case Control Consortium 2 (n = 13,950) study, and, in particular, when the individuals in the sample are unrelated. In these cases, the SKAT approximation tends to be highly overconservative and therefore underpowered. To address this limitation, we suggest an efficient method to calculate exact P-values for the score test in the case of a single variance component and a continuous response vector, which can speed up the analysis by orders of magnitude. Our results enable fast and accurate application of the score test in heritability and in set-based association tests. Our method is available in http://github.com/cozygene/RL-SKAT.

Evidence of epigenetic admixture in the Colombian population

DNA methylation (DNAm) measured in lymphoblastoid cell lines has been repeatedly demonstrated to differ between various human populations. Due to the role that DNAm plays in controlling gene expression, these differences could significantly contribute to ethnic phenotypic differences. However, because previous studies have compared distinct ethnic groups where genetic and environmental context are confounded, their relative contribution to phenotypic differences between ethnicities remains unclear. Using DNAm assayed in whole blood and colorectal tissue of 132 admixed individuals from Colombia, we identified sites where differential DNAm levels were associated with the local ancestral genetic context. Our results are consistent with population specific DNAm being primarily driven by between population genetic differences in cis, with little environmental contribution, and with consistent effects across tissues. The findings offer new insights into a possible mechanism driving phenotypic differences among different ethnic groups, and could help explain ethnic differences in colorectal cancer incidence.

MeCP2 and CTCF: enhancing the cross-talk of silencers

This paper provides a brief introductory review of the most recent advances in our knowledge about the structural and functional aspects of two transcriptional regulators: MeCP2, a protein whose mutated forms are involved in Rett syndrome; and CTCF, a constitutive transcriptional insulator. This is followed by a description of the PTMs affecting these two proteins and an analysis of their known interacting partners. A special emphasis is placed on the recent studies connecting these two proteins, focusing on the still poorly understood potential structural and functional interactions between the two of them on the chromatin substrate. An overview is provided for some of the currently known genes that are dually regulated by these two proteins. Finally, a model is put forward to account for their possible involvement in their regulation of gene expression.

The impact of genetic variation and cigarette smoke on DNA methylation in current and former smokers from the COPDGene study

DNA methylation can be affected by systemic exposures, such as cigarette smoking and genetic sequence variation; however, the relative impact of each on the epigenome is unknown. We aimed to assess if cigarette smoking and genetic variation are associated with overlapping or distinct sets of DNA methylation marks and pathways. We selected 85 Caucasian current and former smokers with genome-wide single nucleotide polymorphism (SNP) genotyping available from the COPDGene study. Genome-wide methylation was obtained on DNA from whole blood using the Illumina HumanMethylation27 platform. To determine the impact of local sequence variation on DNA methylation (mQTL), we examined the association between methylation and SNPs within 50 kb of each CpG site. To examine the impact of cigarette smoking on DNA methylation, we examined the differences in methylation by current cigarette smoking status. We detected 770 CpG sites annotated to 708 genes associated at an FDR < 0.05 in the cis-mQTL analysis and 1,287 CpG sites annotated to 1,242 genes, which were nominally associated in the smoking-CpG association analysis (P(unadjusted) < 0.05). Forty-three CpG sites annotated to 40 genes were associated with both SNP variation and current smoking; this overlap was not greater than that expected by chance. Our results suggest that cigarette smoking and genetic variants impact distinct sets of DNA methylation marks, the further elucidation of which may partially explain the variable susceptibility to the health effects of cigarette smoking. Ascertaining how genetic variation and systemic exposures differentially impact the human epigenome has relevance for both biomarker identification and therapeutic target development for smoking-related diseases.

Imputation of DNA Methylation Levels in the Brain Implicates a Risk Factor for Parkinson's Disease

Understanding how genetic variation affects intermediate phenotypes, like DNA methylation or gene expression, and how these in turn vary with complex human disease provides valuable insight into disease etiology. However, intermediate phenotypes are typically tissue and developmental stage specific, making relevant phenotypes difficult to assay. Assembling large case–control cohorts, necessary to achieve sufficient statistical power to assess associations between complex traits and relevant intermediate phenotypes, has therefore remained challenging. Imputation of such intermediate phenotypes represents a practical alternative in this context. We used a mixed linear model to impute DNA methylation (DNAm) levels of four brain tissues at up to 1826 methylome-wide sites in 6259 patients with Parkinson’s disease and 9452 controls from across five genome-wide association studies (GWAS). Six sites, in two regions, were found to associate with Parkinson’s disease for at least one tissue. While a majority of identified sites were within an established risk region for Parkinson’s disease, suggesting a role of DNAm in mediating previously observed genetic effects at this locus, we also identify an association with four CpG sites in chromosome 16p11.2. Direct measures of DNAm in the substantia nigra of 39 cases and 13 control samples were used to independently replicate these four associations. Only the association at cg10917602 replicated with a concordant direction of effect (P = 0.02). cg10917602 is 87 kb away from the closest reported GWAS hit. The employed imputation methodology implies that variation of DNAm levels at cg10917602 is predictive for Parkinson’s disease risk, suggesting a possible causal role for methylation at this locus. More generally this study demonstrates the feasibility of identifying predictive epigenetic markers of disease risk from readily available data sets.

Contribution of Heritability and Epigenetic Factors to Skeletal Muscle Mass Variation in United Kingdom Twins

CONTEXT

Skeletal muscle mass (SMM) is one of the major components of human body composition, with deviations from normal values often leading to sarcopenia.

OBJECTIVE

Our major aim was to conduct a genome-wide DNA methylation study in an attempt to identify potential genomic regions associated with SMM.

DESIGN

This was a mixed cross-sectional and longitudinal study.

SETTING

Community-based study.

PARTICIPANTS

A total of 1550 middle-aged United Kingdom twins (monozygotic [MZ] and dizygotic [DZ]), 297 of which were repeatedly measured participated in the study.

MAIN OUTCOME MEASURE

Appendicular lean mass assessed using dual-energy X-ray absorptiometry technology, and methylated DNA immunoprecipitation sequencing DNA methylation profiling genome-wide were obtained from each individual.

RESULTS

Heritability estimate of SMM, with simultaneous adjustment for covariates obtained using variance decomposition analysis, was h(2) = 0.809 ± 0.050. After quality control and analysis of longitudinal stability, the DNA methylation data comprised of 723 029 genomic sites, with positive correlations between repeated measurements (Rrepeated = 0.114-0.905). Correlations between MZ and DZ twins were 0.51 and 0.38 at a genome-wide average, respectively, and clearly increased with Rrepeated. Testing for DNA methylation association with SMM in 50 discordant MZ twins revealed 36 081 nominally significant results, of which the top-ranked 134 signals (P < .01 and Rrepeated > 0.40) were subjected to replication in the sample of 1196 individuals. Seven SMM methylation association signals replicated at a false discovery rate less than 0.1, and these were located in or near genes DNAH12, CAND1, CYP4F29P, and ZFP64, which have previously been highlighted in muscle-related studies. Adjusting for age, smoking, and blood cell heterogeneity did not alter significance of these associations.

CONCLUSION

This epigenome-wide study, testing longitudinally stable methylation sites, discovered and replicated a number of associations between DNA methylation at CpG loci and SMM. Four replicated signals were related to genes with potential muscle functions, suggesting that the methylome of whole blood may be informative of SMM variation.

The heritability and patterns of DNA methylation in normal human colorectum

DNA methylation (DNAm) has been linked to changes in chromatin structure, gene expression and disease. The DNAm level can be affected by genetic variation; although, how this differs by CpG dinucleotide density and genic location of the DNAm site is not well understood. Moreover, the effect of disease causing variants on the DNAm level in a tissue relevant to disease has yet to be fully elucidated. To this end, we investigated the phenotypic profiles, genetic effects and regional genomic heritability for 196080 DNAm sites in healthy colorectum tissue from 132 unrelated Colombian individuals. DNAm sites in regions of low-CpG density were more variable, on average more methylated and were more likely to be significantly heritable when compared with DNAm sites in regions of high-CpG density. DNAm sites located in intergenic regions had a higher mean DNAm level and were more likely to be heritable when compared with DNAm sites in the transcription start site (TSS) of a gene expressed in colon tissue. Within CpG-dense regions, the propensity of the DNAm level to be heritable was lower in the TSS of genes expressed in colon tissue than in the TSS of genes not expressed in colon tissue. In addition, regional genetic variation was associated with variation in local DNAm level no more frequently for DNAm sites within colorectal cancer risk regions than it was for DNAm sites outside such regions. Overall, DNAm sites located in different genomic contexts exhibited distinguishable profiles and may have a different biological function.

Impact of Early Environment on Children's Mental Health: Lessons From DNA Methylation Studies With Monozygotic Twins

Over the past decade, epigenetic analyses have made important contributions to our understanding of healthy development and a wide variety of adverse conditions such as cancer and psychopathology. There is increasing evidence that DNA methylation is a mechanism by which environmental factors influence gene transcription and, ultimately, phenotype. However, differentiating the effects of the environment from those of genetics on DNA methylation profiles remains a significant challenge. Monozygotic (MZ) twin study designs are unique in their ability to control for genetic differences because each pair of MZ twins shares essentially the same genetic sequence with the exception of a small number of de novo mutations and copy number variations. Thus, differences within twin pairs in gene expression and phenotype, including behavior, can be attributed in the majority of cases to environmental effects rather than genetic influence. In this article, we review the literature showing how MZ twin designs can be used to study basic epigenetic principles, contributing to understanding the role of early in utero and postnatal environmental factors on the development of psychopathology. We also highlight the importance of initiating longitudinal and experimental studies with MZ twins during pregnancy. This approach is especially important to identify: (1) critical time periods during which the early environment can impact brain and mental health development, and (2) the specific mechanisms through which early environmental effects may be mediated. These studies may inform the optimum timing and design for early preventive interventions aimed at reducing risk for psychopathology.

A Flexible, Efficient Binomial Mixed Model for Identifying Differential DNA Methylation in Bisulfite Sequencing Data

Identifying sources of variation in DNA methylation levels is important for understanding gene regulation. Recently, bisulfite sequencing has become a popular tool for investigating DNA methylation levels. However, modeling bisulfite sequencing data is complicated by dramatic variation in coverage across sites and individual samples, and because of the computational challenges of controlling for genetic covariance in count data. To address these challenges, we present a binomial mixed model and an efficient, sampling-based algorithm (MACAU: Mixed model association for count data via data augmentation) for approximate parameter estimation and p-value computation. This framework allows us to simultaneously account for both the over-dispersed, count-based nature of bisulfite sequencing data, as well as genetic relatedness among individuals. Using simulations and two real data sets (whole genome bisulfite sequencing (WGBS) data from Arabidopsis thaliana and reduced representation bisulfite sequencing (RRBS) data from baboons), we show that our method provides well-calibrated test statistics in the presence of population structure. Further, it improves power to detect differentially methylated sites: in the RRBS data set, MACAU detected 1.6-fold more age-associated CpG sites than a beta-binomial model (the next best approach). Changes in these sites are consistent with known age-related shifts in DNA methylation levels, and are enriched near genes that are differentially expressed with age in the same population. Taken together, our results indicate that MACAU is an efficient, effective tool for analyzing bisulfite sequencing data, with particular salience to analyses of structured populations. MACAU is freely available at www.xzlab.org/software.html.

Many obesity-associated SNPs strongly associate with DNA methylation changes at proximal promoters and enhancers

BACKGROUND

The mechanisms by which genetic variants, such as single nucleotide polymorphisms (SNPs), identified in genome-wide association studies act to influence body mass remain unknown for most of these SNPs, which continue to puzzle the scientific community. Recent evidence points to the epigenetic and chromatin states of the genome as having important roles.

METHODS

We genotyped 355 healthy young individuals for 52 known obesity-associated SNPs and obtained DNA methylation levels in their blood using the Illumina 450 K BeadChip. Associations between alleles and methylation at proximal cytosine residues were tested using a linear model adjusted for age, sex, weight category, and a proxy for blood cell type counts. For replication in other tissues, we used two open-access datasets (skin fibroblasts, n = 62; four brain regions, n = 121-133) and an additional dataset in subcutaneous and visceral fat (n = 149).

RESULTS

We found that alleles at 28 of these obesity-associated SNPs associate with methylation levels at 107 proximal CpG sites. Out of 107 CpG sites, 38 are located in gene promoters, including genes strongly implicated in obesity (MIR148A, BDNF, PTPMT1, NR1H3, MGAT1, SCGB3A1, HOXC12, PMAIP1, PSIP1, RPS10-NUDT3, RPS10, SKOR1, MAP2K5, SIX5, AGRN, IMMP1L, ELP4, ITIH4, SEMA3G, POMC, ADCY3, SSPN, LGR4, TUFM, MIR4721, SULT1A1, SULT1A2, APOBR, CLN3, SPNS1, SH2B1, ATXN2L, and IL27). Interestingly, the associated SNPs are in known eQTLs for some of these genes. We also found that the 107 CpGs are enriched in enhancers in peripheral blood mononuclear cells. Finally, our results indicate that some of these associations are not blood-specific as we successfully replicated four associations in skin fibroblasts.

CONCLUSIONS

Our results strongly suggest that many obesity-associated SNPs are associated with proximal gene regulation, which was reflected by association of obesity risk allele genotypes with differential DNA methylation. This study highlights the importance of DNA methylation and other chromatin marks as a way to understand the molecular basis of genetic variants associated with human diseases and traits.

HMPL: A Pipeline for Identifying Hemimethylation Patterns by Comparing Two Samples

DNA methylation (the addition of a methyl group to a cytosine) is an important epigenetic event in mammalian cells because it plays a key role in regulating gene expression. Most previous methylation studies assume that DNA methylation occurs on both positive and negative strands. However, a few studies have reported that in some genes, methylation occurs only on one strand (ie, hemimethylation) and has clustering patterns. These studies report that hemimethylation occurs on individual genes. It is unclear whether hemimethylation occurs genome-wide and whether there are hemimethylation differences between cancerous and noncancerous cells. To address these questions, we have developed the first-ever pipeline, named hemimethylation pipeline (HMPL), to identify hemimethylation patterns. Utilizing the available software and the newly developed Perl and R scripts, HMPL can identify hemimethylation patterns for a single sample and can also compare two different samples.

Novel epigenetic determinants of type 2 diabetes in Mexican-American families

Although DNA methylation is now recognized as an important mediator of complex diseases, the extent to which the genetic basis of such diseases is accounted for by DNA methylation is unknown. In the setting of large, extended families representing a minority, high-risk population of the USA, we aimed to characterize the role of epigenome-wide DNA methylation in type 2 diabetes (T2D). Using Illumina HumanMethylation450 BeadChip arrays, we tested for association of DNA methylation at 446 356 sites with age, sex and phenotypic traits related to T2D in 850 pedigreed Mexican-American individuals. Robust statistical analyses showed that (i) 15% of the methylome is significantly heritable, with a median heritability of 0.14; (ii) DNA methylation at 14% of CpG sites is associated with nearby sequence variants; (iii) 22% and 3% of the autosomal CpG sites are associated with age and sex, respectively; (iv) 53 CpG sites were significantly associated with liability to T2D, fasting blood glucose and insulin resistance; (v) DNA methylation levels at five CpG sites, mapping to three well-characterized genes (TXNIP, ABCG1 and SAMD12) independently explained 7.8% of the heritability of T2D (vi) methylation at these five sites was unlikely to be influenced by neighboring DNA sequence variation. Our study has identified novel epigenetic indicators of T2D risk in Mexican Americans who have increased risk for this disease. These results provide new insights into potential treatment targets of T2D.

Efficient set tests for the genetic analysis of correlated traits

Set tests are a powerful approach for genome-wide association testing between groups of genetic variants and quantitative traits. We describe mtSet (http://github.com/PMBio/limix), a mixed-model approach that enables joint analysis across multiple correlated traits while accounting for population structure and relatedness. mtSet effectively combines the benefits of set tests with multi-trait modeling and is computationally efficient, enabling genetic analysis of large cohorts (up to 500,000 individuals) and multiple traits.

Personalized medicine: from genotypes, molecular phenotypes and the quantified self, towards improved medicine

Advances in molecular profiling and sensor technologies are expanding the scope of personalized medicine beyond genotypes, providing new opportunities for developing richer and more dynamic multi-scale models of individual health. Recent studies demonstrate the value of scoring high-dimensional microbiome, immune, and metabolic traits from individuals to inform personalized medicine. Efforts to integrate multiple dimensions of clinical and molecular data towards predictive multi-scale models of individual health and wellness are already underway. Improved methods for mining and discovery of clinical phenotypes from electronic medical records and technological developments in wearable sensor technologies present new opportunities for mapping and exploring the critical yet poorly characterized "phenome" and "envirome" dimensions of personalized medicine. There are ambitious new projects underway to collect multi-scale molecular, sensor, clinical, behavioral, and environmental data streams from large population cohorts longitudinally to enable more comprehensive and dynamic models of individual biology and personalized health. Personalized medicine stands to benefit from inclusion of rich new sources and dimensions of data. However, realizing these improvements in care relies upon novel informatics methodologies, tools, and systems to make full use of these data to advance both the science and translational applications of personalized medicine.

Correlation between DNA methylation and gene expression in the brains of patients with bipolar disorder and schizophrenia

OBJECTIVES

Aberrant DNA methylation and gene expression have been reported in postmortem brain tissues of psychotic patients, but until now there has been no systematic evaluation of synergistic changes in methylation and expression on a genome-wide scale in brain tissue.

METHODS

In this study, genome-wide methylation and expression analyses were performed on cerebellum samples from 39 patients with schizophrenia, 36 patients with bipolar disorder, and 43 unaffected controls, to screen for a correlation between gene expression and CpG methylation.

RESULTS

Out of 71,753 CpG gene pairs (CGPs) tested across the genome, 204 were found to significantly correlate with gene expression after correction for multiple testing [p < 0.05, false discovery rate (FDR) q < 0.05]. The correlated CGPs were tested for disease-associated expression and methylation by comparing psychotic patients with bipolar disorder and schizophrenia to healthy controls. Four of the identified CGPs were found to significantly correlate with the differential expression and methylation of genes encoding phosphoinositide-3-kinase, regulatory subunit 1 (PIK3R1), butyrophilin, subfamily 3, member A3 (BTN3A3), nescient helix-loop-helix 1 (NHLH1), and solute carrier family 16, member 7 (SLC16A7) in psychotic patients (p < 0.05, FDR q < 0.2). Additional expression and methylation datasets were used to validate the relationship between DNA methylation, gene expression, and neuropsychiatric diseases.

CONCLUSIONS

These results suggest that the identified differentially expressed genes with an aberrant methylation pattern may represent novel candidate factors in the etiology and pathology of neuropsychiatric disorders.

Genome-wide associations between genetic and epigenetic variation influence mRNA expression and insulin secretion in human pancreatic islets

Genetic and epigenetic mechanisms may interact and together affect biological processes and disease development. However, most previous studies have investigated genetic and epigenetic mechanisms independently, and studies examining their interactions throughout the human genome are lacking. To identify genetic loci that interact with the epigenome, we performed the first genome-wide DNA methylation quantitative trait locus (mQTL) analysis in human pancreatic islets. We related 574,553 single nucleotide polymorphisms (SNPs) with genome-wide DNA methylation data of 468,787 CpG sites targeting 99% of RefSeq genes in islets from 89 donors. We identified 67,438 SNP-CpG pairs in cis, corresponding to 36,783 SNPs (6.4% of tested SNPs) and 11,735 CpG sites (2.5% of tested CpGs), and 2,562 significant SNP-CpG pairs in trans, corresponding to 1,465 SNPs (0.3% of tested SNPs) and 383 CpG sites (0.08% of tested CpGs), showing significant associations after correction for multiple testing. These include reported diabetes loci, e.g. ADCY5, KCNJ11, HLA-DQA1, INS, PDX1 and GRB10. CpGs of significant cis-mQTLs were overrepresented in the gene body and outside of CpG islands. Follow-up analyses further identified mQTLs associated with gene expression and insulin secretion in human islets. Causal inference test (CIT) identified SNP-CpG pairs where DNA methylation in human islets is the potential mediator of the genetic association with gene expression or insulin secretion. Functional analyses further demonstrated that identified candidate genes (GPX7, GSTT1 and SNX19) directly affect key biological processes such as proliferation and apoptosis in pancreatic β-cells. Finally, we found direct correlations between DNA methylation of 22,773 (4.9%) CpGs with mRNA expression of 4,876 genes, where 90% of the correlations were negative when CpGs were located in the region surrounding transcription start site. Our study demonstrates for the first time how genome-wide genetic and epigenetic variation interacts to influence gene expression, islet function and potential diabetes risk in humans.

Inter-individual variation in genetics and epigenetics affects biological processes and disease susceptibility. However, most studies have investigated genetic and epigenetic mechanisms independently and to uncover novel mechanisms affecting disease susceptibility there is a highlighted need to study interactions between these factors on a genome-wide scale. To identify novel loci affecting islet function and potentially diabetes, we performed the first genome-wide methylation quantitative trait locus (mQTL) analysis in human pancreatic islets including DNA methylation of 468,787 CpG sites located throughout the genome. Our results showed that DNA methylation of 11,735 CpGs in 4,504 unique genes is regulated by genetic factors located in cis (67,438 SNP-CpG pairs). Furthermore, significant mQTLs cover previously reported diabetes loci including KCNJ11, INS, HLA, PDX1 and GRB10. We also found mQTLs associated with gene expression and insulin secretion in human islets. By performing causality inference tests (CIT), we identified CpGs where DNA methylation potentially mediates the genetic impact on gene expression and insulin secretion. Our functional follow-up experiments further demonstrated that identified mQTLs/genes (GPX7, GSTT1 and SNX19) directly affect pancreatic β-cell function. Together, our study provides a detailed map of genome-wide associations between genetic and epigenetic variation, which affect gene expression and insulin secretion in human pancreatic islets.

Detecting epistasis in human complex traits

Genome-wide association studies (GWASs) have become the focus of the statistical analysis of complex traits in humans, successfully shedding light on several aspects of genetic architecture and biological aetiology. Single-nucleotide polymorphisms (SNPs) are usually modelled as having additive, cumulative and independent effects on the phenotype. Although evidently a useful approach, it is often argued that this is not a realistic biological model and that epistasis (that is, the statistical interaction between SNPs) should be included. The purpose of this Review is to summarize recent directions in methodology for detecting epistasis and to discuss evidence of the role of epistasis in human complex trait variation. We also discuss the relevance of epistasis in the context of GWASs and potential hazards in the interpretation of statistical interaction terms.

Racial/ethnic disparities in human DNA methylation

The racial/ethnic disparities in DNA methylation patterns indicate that molecular markers may play a role in determining the individual susceptibility to diseases in different ethnic groups. Racial disparities in DNA methylation patterns have been identified in prostate cancer, breast cancer and colorectal cancer and are related to racial differences in cancer prognosis and survival.

Identification of methylation quantitative trait loci (mQTLs) influencing promoter DNA methylation of alcohol dependence risk genes

Interaction of DNA methylation and sequence variants that are methylation quantitative trait loci (mQTLs) may influence susceptibility to diseases such as alcohol dependence (AD). We used genome-wide genotype data from 268 African Americans (AAs: 129 AD cases and 139 controls) and 143 European Americans (EAs: 129 AD cases and 14 controls) to identify mQTLs that were associated with promoter CpGs in 82 AD risk genes. 282 significant mQTL-CpG pairs (9.9 × 10(-100) ≤ P(nominal) ≤ 7.7 × 10(-8)) in AAs and 313 significant mQTL-CpG pairs (2.7 × 10(-53) ≤ P(nominal) ≤ 9.9 × 10(-8)) in EAs were identified [i.e., mQTL-CpG associations survived multiple-testing correction, q values (false discovery rate) ≤ 0.05]. The most significant mQTL was rs1800759, which was strongly associated with CpG cg12011299 in both AAs (P(nominal) = 9.9 × 10(-100); q = 6.7 × 10(-91)) and EAs (P(nominal) = 2.7 × 10(-53); q = 1.4 × 10(-44)). Rs1800759 (previously known to be associated to AD) and CpG cg12011299 (distance: 37 bp) are both located in alcohol dehydrogenase (ADH) 4 gene (ADH4) promoter region. In general, the strength of association between mQTLs and CpGs was inversely correlated with the distance between them. Association was also influenced by race and AD. Additionally, 48.3 % of the mQTLs identified in AAs and 65.6 % of the mQTLs identified in EAs were predicted to be expression QTLs. Three mQTLs (rs2173201, rs4147542, and rs4147541 in ADH1B-AHD1C gene cluster region) found in AAs were previously identified by our genome-wide association studies as being significantly associated with AD in AAs. Thus, DNA methylation, which can be influenced by sequence variants and is implicated in gene expression regulation, appears to at least partially underlie the association of genetic variation with AD.

Contribution of genetic variation to transgenerational inheritance of DNA methylation

Background

Despite the important role DNA methylation plays in transcriptional regulation, the transgenerational inheritance of DNA methylation is not well understood. The genetic heritability of DNA methylation has been estimated using twin pairs, although concern has been expressed whether the underlying assumption of equal common environmental effects are applicable due to intrauterine differences between monozygotic and dizygotic twins. We estimate the heritability of DNA methylation on peripheral blood leukocytes using Illumina HumanMethylation450 array using a family based sample of 614 people from 117 families, allowing comparison both within and across generations.

Results

The correlations from the various available relative pairs indicate that on average the similarity in DNA methylation between relatives is predominantly due to genetic effects with any common environmental or zygotic effects being limited. The average heritability of DNA methylation measured at probes with no known SNPs is estimated as 0.187. The ten most heritable methylation probes were investigated with a genome-wide association study, all showing highly statistically significant cis mQTLs. Further investigation of one of these cis mQTL, found in the MHC region of chromosome 6, showed the most significantly associated SNP was also associated with over 200 other DNA methylation probes in this region and the gene expression level of 9 genes.

Conclusions

The majority of transgenerational similarity in DNA methylation is attributable to genetic effects, and approximately 20% of individual differences in DNA methylation in the population are caused by DNA sequence variation that is not located within CpG sites.

The benefits of selecting phenotype-specific variants for applications of mixed models in genomics

Applications of linear mixed models (LMMs) to problems in genomics include phenotype prediction, correction for confounding in genome-wide association studies, estimation of narrow sense heritability, and testing sets of variants (e.g., rare variants) for association. In each of these applications, the LMM uses a genetic similarity matrix, which encodes the pairwise similarity between every two individuals in a cohort. Although ideally these similarities would be estimated using strictly variants relevant to the given phenotype, the identity of such variants is typically unknown. Consequently, relevant variants are excluded and irrelevant variants are included, both having deleterious effects. For each application of the LMM, we review known effects and describe new effects showing how variable selection can be used to mitigate them.

Characterization of the DNA methylome and its interindividual variation in human peripheral blood monocytes

AIM

Peripheral blood monocytes (PBMs) play multiple and critical roles in the immune response, and abnormalities in PBMs have been linked to a variety of human disorders. However, the DNA methylation landscape in PBMs is largely unknown. In this study, we characterized epigenome-wide DNA methylation profiles in purified PBMs.

MATERIALS & METHODS

PBMs were isolated from freshly collected peripheral blood from 18 unrelated healthy postmenopausal Caucasian females. Epigenome-wide DNA methylation profiles (the methylome) were characterized by using methylated DNA immunoprecipitation combined with high-throughput sequencing.

RESULTS

Distinct patterns were revealed at different genomic features. For instance, promoters were commonly (∼58%) found to be unmethylated; whereas protein coding regions were largely (∼84%) methylated. Although CpG-rich and -poor promoters showed distinct methylation patterns, interestingly, a negative correlation between promoter methylation levels and gene transcription levels was consistently observed across promoters with high to low CpG densities. Importantly, we observed substantial interindividual variation in DNA methylation across the individual PBM methylomes and the pattern of this interindividual variation varied between different genomic features, with highly variable regions enriched for repetitive DNA elements. Furthermore, we observed a modest but significant excess (p < 2.2 × 10(-16)) of genes showing a negative correlation between interindividual promoter methylation and transcription levels. These significant genes were enriched in biological processes that are closely related to PBM functions, suggesting that alteration in DNA methylation is likely to be an important mechanism contributing to the interindividual variation in PBM function, and PBM-related phenotypic and disease-susceptibility variation in humans.

CONCLUSION

This study represents a comprehensive analysis of the human PBM methylome and its interindividual variation. Our data provide a valuable resource for future epigenomic and multiomic studies, exploring biological and disease-related regulatory mechanisms in PBMs.

Influence of DNMT genotype on global and site specific DNA methylation patterns in neonates and pregnant women

This study examines the relationship between common genetic variation within DNA methyltransferase genes and inter-individual variation in DNA methylation. Eleven polymorphisms spanning DNMT1 and DNMT3B were genotyped. Global and gene specific (IGF2, IGFBP3, ZNT5) DNA methylation was quantified by LUMA and bisulfite Pyrosequencing assays, respectively, in neonatal cord blood and in maternal peripheral blood. Associations between maternal genotype and maternal methylation (n (≈) 333), neonatal genotype and neonatal methylation (n (≈) 454), and maternal genotype and neonatal methylation (n (≈) 137) were assessed. The findings of this study provide some support to the hypothesis that genetic variation in DNA methylating enzymes influence DNA methylation at global and gene-specific levels; however observations were not robust to correction for multiple testing. More comprehensive analysis of the influence of genetic variation on global and site specific DNA methylation is warranted.

Genetic effects on DNA methylation and its potential relevance for obesity in Mexican Americans

Several studies have identified effects of genetic variation on DNA methylation patterns and associated heritability, with research primarily focused on Caucasian individuals. In this paper, we examine the evidence for genetic effects on DNA methylation in a Mexican American cohort, a population burdened by a high prevalence of obesity. Using an Illumina-based platform and following stringent quality control procedures, we assessed a total of 395 CpG sites in peripheral blood samples obtained from 183 Mexican American individuals for evidence of heritability, proximal genetic regulation and association with age, sex and obesity measures (i.e. waist circumference and body mass index). We identified 16 CpG sites (∼4%) that were significantly heritable after Bonferroni correction for multiple testing and 27 CpG sites (∼6.9%) that showed evidence of genetic effects. Six CpG sites (∼2%) were associated with age, primarily exhibiting positive relationships, including CpG sites in two genes that have been implicated in previous genome-wide methylation studies of age (FZD9 and MYOD1). In addition, we identified significant associations between three CpG sites (∼1%) and sex, including DNA methylation in CASP6, a gene that may respond to estradiol treatment, and in HSD17B12, which encodes a sex steroid hormone. Although we did not identify any significant associations between DNA methylation and the obesity measures, several nominally significant results were observed in genes related to adipogenesis, obesity, energy homeostasis and glucose homeostasis (ARHGAP9, CDKN2A, FRZB, HOXA5, JAK3, MEST, NPY, PEG3 and SMARCB1). In conclusion, we were able to replicate several findings from previous studies in our Mexican American cohort, supporting an important role for genetic effects on DNA methylation. In addition, we found a significant influence of age and sex on DNA methylation, and report on trend-level, novel associations between DNA methylation and measures of obesity.

A powerful and efficient set test for genetic markers that handles confounders

MOTIVATION

Approaches for testing sets of variants, such as a set of rare or common variants within a gene or pathway, for association with complex traits are important. In particular, set tests allow for aggregation of weak signal within a set, can capture interplay among variants and reduce the burden of multiple hypothesis testing. Until now, these approaches did not address confounding by family relatedness and population structure, a problem that is becoming more important as larger datasets are used to increase power.

RESULTS

We introduce a new approach for set tests that handles confounders. Our model is based on the linear mixed model and uses two random effects-one to capture the set association signal and one to capture confounders. We also introduce a computational speedup for two random-effects models that makes this approach feasible even for extremely large cohorts. Using this model with both the likelihood ratio test and score test, we find that the former yields more power while controlling type I error. Application of our approach to richly structured Genetic Analysis Workshop 14 data demonstrates that our method successfully corrects for population structure and family relatedness, whereas application of our method to a 15 000 individual Crohn's disease case-control cohort demonstrates that it additionally recovers genes not recoverable by univariate analysis.

AVAILABILITY

A Python-based library implementing our approach is available at http://mscompbio.codeplex.com.