Epigenetics: DNA Methylation Analysis

• Epigenetics is the study of chemical modifications to the genome that can alter gene expression.
• It provides a potential link between environment and genome.
• It adds complexity in studying human biological variation.
• Epigenetic research has expanded rapidly over the past decade.
• The article provides an overview of current DNA methylation assay methods and addresses considerations for planning and conducting DNA methylation analysis.
• The topics covered include sample collection, lab analyses, statistical analyses, tissue specificity, measure stability, sample timing, batch effects, and data interpretation.

Introduction

• Conrad Waddington described epigenetics as the dynamic interaction of genetic variation and environmental exposures during development to produce a phenotype.
• Molecular biology defines epigenetics as mitotically or meiotically heritable changes in gene expression not caused by changes in DNA sequence.
• An individual's epigenome is dynamic, responding to environmental experience, while the genome is generally stable.
• Epigenetic patterns are sensitive to nutrition, toxicants, and psychosocial stress.
• Epigenetics is a mechanism linking environmental experiences with health outcomes and represents a bridge between nature and nurture.
• Epigenetic changes may facilitate developmental plasticity, adjusting offspring biology to environmental stimuli like malnutrition or psychosocial stress.
• Heritable epigenetic marks across generations suggest a mechanism for evolutionary change.
• Epigenetic modifications can provide a biological "memory" of experiences across generations, enabling organisms to adapt to rapid environmental changes.
• Epigenetics can serve as a short-term adaptation mechanism responding to selective pressures too short-lived for natural selection.
• Epigenetic modifications are markers of prior environmental exposures, like prenatal smoking, or ancestral toxin exposure, and can be useful when those exposures cannot otherwise be assessed.
• Environmental exposures correlate with epigenetic changes, which can allow for inference of prior exposure based on epigenetic mark patterns.
• This article focuses on providing an overview of approaches for epigenetic studies.
• Methylation is the most commonly researched type of epigenetic mechanism.
• Methylation is the attachment of a methyl group to a cytosine and guanine base pair (CpG site) or a histone tail; DNA methylation is more frequently researched due to its stability.
• The article focuses on DNA methylation assessment at CpG sites.
• The article covers: sample collection, laboratory analyses, and statistical analyses.

Methodology for Analyzing Methylation

• Genomes are consistent across cells, however, epigenetic marks vary dramatically across cell and tissue types.
• Epigenetic differences across cells are at least partly responsible for differentiating cells with the same genome into different tissues.
• The first step is determining which type(s) of tissue sample(s) to collect, with buccal, saliva, and blood being three common tissue samples for minimally invasive protocols.
• Figure 1 provides a summary of the pros and cons of each of these types of samples.

Choice of Appropriate Tissue

• Multiple methods for collecting buccal cells and saliva samples are available.
• Blood samples are typically collected with EDTA tubes (an anticoagulant).
• DNA can also be extracted from dried blood spots, opening new possibilities for population-level epigenetic studies of neonates.
• Placenta is of interest as it is easily accessed after birth and at the interface of maternal and fetal environments.
• Epithelial cells via tape-stripping and corneal cells via irrigation of the corneal surface may be of potential interest.
• Buccal cells collected using swabs are preferred due to cell type homogeneity; DNA methylation levels vary across cell types.
• Associations between DNA methylation and phenotype can be lost if levels are averaged across cells with varying methylation.
• Methylation patterns in buccal cells are similar to other tissues, such as liver, kidney, brain, and sperm, compared with blood cells.
• Researchers using buccal cells should swab cheeks long enough to get sufficient DNA without brushing too forcefully to prevent bleeding and contamination of the sample.
• Multiple swabs are recommended to account for relatively low yield with each individual swab.
• DNA collected in buccal swabs is stable for days at room temperature or months/years with the application of tablets or a buffer.
• DNA can be extracted from samples using standard protocols upon completing collection.

Assay Methodologies

• A wide selection of technologies for analysis of DNA methylation has been developed over the last decade.
• A brief overview of the most commonly used approaches along with key advantages and disadvantages of each method is presented.
• The next key decision after deciding what tissue(s) to analyze is to choose between a targeted locus-specific approach, an exploratory genome-wide approach, and a global methylation approach.
• Locus-specific approaches target a small genic or intergenic region.
• Genome-wide approaches analyze thousands to millions of individual CpG sites throughout the genome.
• Global methylation approaches quantify total methylated cytosines in a DNA sample without providing information on methylation levels at any individual locus, making them less informative than the other approaches.
• With its shortfalls relative to the locus-specific and genome-wide approaches, global methylation is not considered in the manuscript.
• Factors that should be considered in choosing between approaches include amount of DNA available, presence or lack of a priori hypotheses regarding biological pathways of interest, and cost.
• Advantages of the locus-specific approach include less input DNA, lower cost, and less concern over Type I error (false positives) as fewer loci are tested.
• Advantages of the locus-specific approach make it the best for investigating a specific biological pathway.
• The primary disadvantage of the locus-specific approach is that it is limited to previously studied gene regions.
• The genome-wide approach offers novel discovery of differentially methylated regions (DMRs) and genome-wide effects, without assuming that changes may only be seen at a specific locus.
• Challenges of the genome-wide approach include: larger volume of input DNA (≥500 ng), higher cost, large datasets, and complex statistical methods to account for multiple testing.
• Methylation of a single gene region can be assessed with genome-wide data.
• Test for genome-wide mean methylation levels for association with a phenotype with genome-wide data.
• It becomes more cost-effective to utilize a microarray if more than 30-50 different loci/individual are to be assayed, though existing microarrays may not always capture the exact CpG sites of interest.
• Using a genome-wide scan to identify loci of interest, which can then be verified using locus-specific methods is a promising approach.
• Verification entails replication of DNA methylation results using a quantitative locus-specific technique on the same samples.
• Results can be validated by replicating in samples from a different population.

Methyl-Dependent Treatments

• A methyl-dependent treatment is necessary for the majority of DNA methylation analyses to maintain DNA methylation marks prior to DNA amplification.
• Methyl transferases are not present during PCR so all methylation marks are erased during DNA amplification.
• Endonuclease digestion, affinity enrichment, and bisulfite conversion are the three primary methyl-dependent treatments.
• The most commonly used method, bisulfite conversion, will be the only method focused on for this article.
• Bisulfite conversion selectively deaminates all unmethylated cytosine nucleotides to uracils while leaving 5-methylcytosine unchanged.
• After the initial round of PCR of bisulfite converted DNA, all uracils become paired with adenines.
• After many rounds of PCR, adenines pair with thymines.
• The DNA is sequenced using a locus-specific approach or a genome-wide method, such as whole genome bisulfite sequencing, or genotyped on a microarray.
• Once sequenced, the presence of a thymine at a CpG site is interpreted as an unmethylated cytosine.
• Failed conversion of unmethylated cytosines can lead to false positive methylation detection.
• Inappropriate conversion of a methylated cytosine into thymine can lead to underestimates of methylated cytosines.
• Comparing conversion performance of commonly used bisulfite treatment kits found some variation among the performance of the kits, which can be improved by adjusting incubation times and temperature cycling.
• Duplicate bisulfite conversions, followed by independent PCRs of each DNA sample, are recommended to ensure replicability.
• After bisulfite treatment, a variety of approaches can be used to assay DNA methylation at individual loci or across whole epigenomes.

Locus-Specific Approach

• Researchers interested in pursuing a locus-specific approach should choose specific loci to assay as the next step after bisulfite conversion.
• Decisions on which loci to study can be based on published literature, and with the aid of epigenetic research browsers and databases.
• If no information is available, one could target functional regions of a gene, such as enhancers or regions within transcription factor binding sites.
• Regions of a few hundred BPs that are rich in CpG sites, and often overlap with the transcription start site of a gene is a region know as CpG islands that researches primarily targeted.
• Nonpromoter CpG islands, low CG-content promoter and enhancer regions, and intronic sequences are also receiving more attention.

Approaches to Analysis of Locus-Specific Methylation

• Bisulfite pyrosequencing has become the gold standard technique for assay of DNA methylation at targeted gene regions, as it is quantitative and reproducible.
• The technique involves PCR amplification of bisulfite-treated DNA using one biotinylated primer.
• The single strand of biotinylated DNA can attach to a streptavidin bead and serve as a template to which nucleotide bases are systematically incorporated one at a time into a complementary strand.
• Upon incorporation of each nucleotide, pyrophosphate is released, and through a series of enzymatic reactions, converted into visible light that is visualized in a pyrogram.
• The emitted light is proportional to the amount of nucleotide incorporated into the sequence.
• The height of each peak in the pyrogram represents a quantitative measure of the proportion of each nucleotide.
• For CpG sites, this measure is a ratio of cytosine to thymine, from which can be inferred the proportion of methylated to unmethylated cytosines at each locus.
• Methylation level is estimated as an average proportion across cells and DNA strands in a sample, and thus it ranges from 0 to 1.
• Pyrosequencing can also detect allele-specific methylation by designing a sequencing primer that contains a single nucleotide polymorphism (SNP) in the 3' region.
• One limitation of pyrosequencing is that lengths of individual reads are relatively short (50-60 bases), though the new PyroMark Q24 Advanced system may be able to read up to 140 bases.
• Efficiency of bisulfite conversion can be easily assessed in bisulfite pyrosequencing by adding a cytosine in the dispensation order at a non-CpG site to detect the presence of unconverted cytosines.
• The most important step for successful pyrosequencing is to optimize primers that produce robust and specific amplification of the target region.
• Three primers are needed for each locus, including a forward, reverse, and sequencing primer.
• One of the forward or reverse primers must be biotinylated at the 5' end for attachment to the beads.
• Use of published primers is a good starting place, particularly as it allows for replication of results across studies.
• The Pyromark Assay Design Software 2.0 (Qiagen) is a useful tool for designing primers for pyrosequencing.
• It is an important consideration to avoid incorporating CpG sites or SNPs within the primer that can interfere with hybridization.
• If a CpG site in the primer cannot be avoided, the cytosine (or guanine if sequencing the reverse strand) should be replaced with a mismatched base unrelated to the methylation status.
• All assay designs should be tested with standards of known methylation levels to verify their quantitative performance prior to implementation.
• A methylation scale can be created via a dilution scale of unmethylated DNA and fully methylated DNA.
• Methylation-specific PCR (MSP) is another locus-specific approach the utilizes one set of primers that amplifies converted unmethylated DNA and one set that only amplifies converted methylated DNA.
• MSP is sensitive and cost-effective, but it can also be very labor intensive and does not provide a quantitative estimate of methylation levels, but rather a binary result of "methylated or not."
• Newer modifications to this technique, such as quantitative MSP and in situ MSP, have improved upon the original MSP design.

Additionally, Methods Utilizing Mass Spectrometry

• Mass spectrometry, such as matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry can be used to measure DNA methylation patterns of bisulfite-treated DNA.
• This method is advantageous because it directly and quantitatively measures the methylation as an increase in mass of the DNA, is nonradioactive.
• Can be automated for high-throughput samples (up to 6,000 samples/day) at a fraction of the cost of standard sequencing methods.
• Allows for relatively long individual read lengths (up to 600 bps on Sequenom MassArray).
• Such methods require more expensive and specialized equipment, and care must be taken in primer design to ensure that fragments are not of similar length (and therefore, mass) or they will not be distinguishable in the analysis.

Genome-Wide Approaches

• Epigenome-wide association studies (EWAS) have become an increasingly common approach for identification of novel sites of DNA methylation variation.
• EWAS methods examine the methylation state at tens of thousands of loci throughout the genome to determine if any of the loci are associated with a phenotype or disease.
• EWAS can be performed using microarrays or high throughput sequencing, following a methylation-dependent treatment of DNA.
• The most commonly used methods for assessment of genome-wide DNA methylation include the Illumina Infinium HumanMethylation (27 K or 450 K) microarrays, reduced-representation bisulfite sequencing (RRBS), and Methylation DNA immunoprecipitation, combined with sequencing or a microarray (meDIP-seq or meDIP-chip).
• None of these methods covers all of the CpG sites in the human genome, which can only be accomplished through WGBS.
• The method of WGBS assays over a billion reads per sample but is a cost-prohibitive technique for many studies with larger sample sizes.
• Illumina 450 K array has become a forerunner in recent years for generating genome-wide methylation data among larger population studies.
• Relatively low cost along with high throughput of hundreds of samples at once, high coverage of over 450,000 CpG sites, and very high quantitative accuracy are the benefits of Illumina 450 K array.
• The array covers many important features, including CpG islands, shores, shelves, promoters, gene bodies, and intergenic regions.
• However, the array only covers ~2% of the >28 million autosomal CpG sites in the human genome, many of which are not located in regions of known functional relevance.
• The Infinium assay detected only 20% of those detected by MeDIP or RRBS when comparing its ability to detect DMRs with other sequencing-based methods.
• Requires relatively large amounts of input DNA (~500 ng).
• RRBS is a promising newer method that provides greater coverage of CpG sites in the genome compared with the Infinium array, with lower requirements for input DNA.
• RRBS enriches CG-rich regions by isolating small restriction fragments generated by Mspl, a methylation-insensitive endonuclease.
• The DNA is then bisulfite converted, amplified, and sequenced.
• Fewer sequencing reads are required as the method enriches for the CpG regions of the genome.
• This method leaves out the less CpG dense regions of the genome, where DNA methylation may be more variable and, therefore, interesting to study.
• Can also be very expensive.
• Genome-wide method utilizing MeDIP is less commonly used today relative to the Illumina microarrays or RRBS.
• MEDIP is an immunoprecipitation method that uses an antibody against 5-methylcytosine to enrich for methylated DNA sequences.
• MeDIP has reduced in popularity, as the enrichment protocol can be susceptible to many sources of bias, including room temperature, humidity, and operator influences, and the results require complex bioinformatics methods for normalization and removal of batch effects.
• The most comprehensive method for analyzing methylation across the human genome is WGBS.
• This method assays DNA methylation across the entire methylome at single-base pair resolution via shotgun sequencing of bisulfite converted DNA.
• A sequencing depth of 30x coverage is generally recommended, which is very expensive and a relatively inefficient approach, as up to half the reads do not cover CpG sites.
• Also generally requires a large amount of input DNA, although newer techniques request only 50 ng of genomic DNA.
• While WGBS is the only available method to measure DNA methylation at every site in the genome, its performance can be biased by efficiency of amplification of methylated versus unmethylated DNA, and by incomplete bisulfite conversion.
• Complex bioinformatics techniques are also required to accurately align bisulfite-converted sequencing reads.

Challenges and Considerations

• There are many challenges to consider when analyzing DNA methylation, and some suggestions for addressing them in future analyses are below.

Tissue Specificity

• May be one of the greatest challenges in analyses of DNA methylation.
• Degree of DNA methylation at particular locus depends on the tissue under consideration as demonstrated by many studies.
• Tissues are differentially methylated because cellular differentiation is mediated and maintained, in part, through changes in methylation.
• Differences in methylation between tissues in a single individual can greatly exceed the amount of variation between individuals.
• The degree of correlation between tissues varies not only by tissue type but also by the placement and density of particular CpGs in the region of analysis.
• Tissue specificity must be accounted for when comparing across studies and inferring functional significance of methylation changes as methylation varies in complex ways across tissues.
• Most tissue samples are composed of heterogeneous cell types.
• Methylation readouts reflect an average across diverse cell types present within a particular tissue sample.
• Cell sorting to isolate individual cell types can be cost prohibitive, and requires fresh samples that are often unfeasible to obtain.
• Heterogeneity of cell types in a tissue can confound statistical analysis when there is a difference in cellular composition between cases and controls.
• There are a number of bioinformatic methods that can be used to control for variation across different cell types in blood samples, even in the absence of isolating cell types.
• The accuracy of these methods is yet to be fully validated in direct comparison with cell sorting, and it may not be appropriate to use these methods if the environmental exposures or health of the reference samples differ from those of the population under study.
• This point is of particular note for anthropologists who may be working with populations coming from ecological and social environments that vary substantially from Western samples upon which these methods are based.

Stability of Epigenetic Marks and Timing of Sample Collection

• The stability of DNA methylation patterns over time is still unknown.
• DNA methylation marks are lost and re-established early in embryonic development, but it is not clear how stable they remain throughout the life course.
• DNA methylation marks are dynamic across time.
• Methylation was associated with age in 28% of CpG sites.
• The nature of this association may vary depending on the location of the CpG.
• Epigenetic studies must carefully consider when samples were collected in relation to the exposures and outcomes of interest.
• If samples are collected at the same time as phenotypes are assessed, it is inappropriate to infer causality of epigenetic changes, as the direction of the association is unknown.
• Prospective designs and experimental studies are preferred for establishing causal relationships.
• No published studies have demonstrated a diurnal rhythm in DNA methylation patterns as has been found with histone acetylation.
• Methylation may be more sensitive to acute exposures than previously considered, and that to the extent possible, such factors should be accounted for in study design and statistical analysis.

Batch Effects

• DNA methylation analyses are prone to batch effects.
• Confounding is a result of laboratory conditions, position on a plate, or experiment time.
• The best way to reduce batch effects in an EWAS is to distribute samples randomly across chips.
• Duplicates along with control samples should be included.
• Validation of findings is also recommended.
• Several statistical programs have been developed to aid in correcting for batch effects.

Statistical Considerations

• The continuous and finite nature of DNA methylation data results in statistical properties are different from genotype or gene expression data.
• DNA methylation data should be Beta distributed.
• DNA methylation marks are often non-normally distributed.
• Microarrays, many probes are nonspecific.
• It is important to note that power analyses for EWAS are more complex relative to those used for GWAS, given that there are so little data to draw upon regarding frequency of DNA methylation variants.
• Demographic and environmental factors could serve as potential confounders.

Analysis and Interpretation

• Studies will report a statistically significant difference of as little as a few percent change in methylation.
• The errors inherent in many of the measurement techniques should be cautioned for these small changes.
• It is difficult to find evidence regarding what a biologically significant difference in DNA methylation would be.
• It may be preferable to analyze regional changes in DNA methylation, rather than individual sites.
• Too narrow a focus on regional effects could miss detection of individual CpG sites that could be functionally relevant if located in key regulatory regions, such as a transcription factor binding site.
• Statistically significant associations should ideally be followed up with RNA or protein expression analyses.

Manage Expectations

• Genetics research was believed to be the key to solving the greatest mysteries in human variation and disease for many years.
• Epigenetics research is receiving a similar hype, and we must be careful to manage expectations.
• It is critical to have an understanding of the extent or importance of variation in DNA methylation within and among individuals or populations before we can appreciate the significance of changes in methylation associated with particular environmental exposures or phenotypes.

Additionally

• DNA methylation can occur outside of CpG contexts, and can occur as 5-hydroxymethylcytosine (5hmC or HmeC) as opposed to 5-methylcytosine (5meC).
• There are also many other types of epigenetic mechanisms, the stability and sensitivity of which remain poorly understood, therefore multi-level regulation is most likely to influence phenotype completely.
• As methods improve and costs are reduced for measurement of microRNAs and methylation or acetylation of histones, anthropologists and other human biologists should begin to consider these other mechanisms in their analyses.

A Study

• One study that demonstrates a typical workflow, from 450 k Illumina array, through pyrovalidation and candidate gene analyses, is a study of maternal depression/anxiety on genome-wide DNA methylation patterns in cord blood.
• The study demonstrates recent recommendations for genome-wide analyses as well as many of the challenges in interpreting epigenetic data in an observational study from a peripheral tissue.

Conclusion

• Epigenetic data have tremendous potential to transform the field of anthropology by adding a new layer of complexity to our understanding of human biological variation, and a new way to explore human adaptation to changing environments.
• This article has served as a primer to assist more investigators in incorporating epigenetic data into their research programs.