Detection of Cytosine Methylation in Ancient DNA from Five Native American Populations Using Bisulfite Sequencing

While cytosine methylation has been widely studied in extant populations, relatively few studies have analyzed methylation in ancient DNA. Most existing studies of epigenetic marks in ancient DNA have inferred patterns of methylation in highly degraded samples using post-mortem damage to cytosines as a proxy for cytosine methylation levels. However, this approach limits the inference of methylation compared with direct bisulfite sequencing, the current gold standard for analyzing cytosine methylation at single nucleotide resolution. In this study, we used direct bisulfite sequencing to assess cytosine methylation in ancient DNA from the skeletal remains of 30 Native Americans ranging in age from approximately 230 to 4500 years before present. Unmethylated cytosines were converted to uracils by treatment with sodium bisulfite, bisulfite products of a CpG-rich retrotransposon were pyrosequenced, and C-to-T ratios were quantified for a single CpG position. We found that cytosine methylation is readily recoverable from most samples, given adequate preservation of endogenous nuclear DNA. In addition, our results indicate that the precision of cytosine methylation estimates is inversely correlated with aDNA preservation, such that samples of low DNA concentration show higher variability in measures of percent methylation than samples of high DNA concentration. In particular, samples in this study with a DNA concentration above 0.015 ng/μL generated the most consistent measures of cytosine methylation. This study presents evidence of cytosine methylation in a large collection of ancient human remains, and indicates that it is possible to analyze epigenetic patterns in ancient populations using direct bisulfite sequencing approaches.     

Capillary electrophoretic analysis of methylation status in CpG-rich regions by single-base extension of primers modified with N6-methoxy-2,6-diaminopurine

Polymerase-mediated single-base extension (SBE) of primers using a fluorescently labeled 2′,3′-dideoxynucleotide triphosphate terminator was originally commercialized as SNaPshot for analysis of single-nucleotide polymorphisms by capillary electrophoresis (CE). Application of this general method to bisulfite-converted/PCR-amplified genomic DNA (gDNA) to analytically infer polymorphic methylation status (i.e., 5-methylcytosine [5mC] vs. cytosine [C]) in CpG-rich regions of gDNA has been noted previously by others to be limited by CE mobility-shifted peaks for SBE products derived from guanine (G)/adenine (A)-mixed-base primers used to hybridize to possible polymorphic sites (i.e., C vs. thymine [T] resulting from 5mC vs. C, respectively). This limitation is precluded in the current study by using novel SNaPshot primers modified with N⁶-methoxy-2,6-diaminopurine (K), which was originally described in 1991 by Brown and Lin as a unique adenine–guanine analog capable of participating in three H-bonds with C or T in DNA. Oligonucleotides modified by K as a bispecific complement for C/T are commercially available or can be readily synthesized, and they may have utility in other assay formats used to analyze CpG methylation status.     

Disclosing Bias in Bisulfite Assay: MethPrimers Underestimate High DNA Methylation

Discordant results obtained in bisulfite assays using MethPrimers (PCR primers designed using MethPrimer software or assuming that non-CpGs cytosines are non methylated) versus primers insensitive to cytosine methylation lead us to hypothesize a technical bias. We therefore used the two kinds of primers to study different experimental models and methylation statuses. We demonstrated that MethPrimers negatively select hypermethylated DNA sequences in the PCR step of the bisulfite assay, resulting in CpG methylation underestimation and non-CpG methylation masking, failing to evidence differential methylation statuses. We also describe the characteristics of “Methylation-Insensitive Primers” (MIPs), having degenerated bases (G/A) to cope with the uncertain C/U conversion. As CpG and non-CpG DNA methylation patterns are largely variable depending on the species, developmental stage, tissue and cell type, a variable extent of the bias is expected. The more the methylome is methylated, the greater is the extent of the bias, with a prevalent effect of non-CpG methylation. These findings suggest a revision of several DNA methylation patterns so far documented and also point out the necessity of applying unbiased analyses to the increasing number of epigenomic studies.     

The Behaviour of 5-Hydroxymethylcytosine in Bisulfite Sequencing

Background

We recently showed that enzymes of the TET family convert 5-mC to 5-hydroxymethylcytosine (5-hmC) in DNA. 5-hmC is present at high levels in embryonic stem cells and Purkinje neurons. The methylation status of cytosines is typically assessed by reaction with sodium bisulfite followed by PCR amplification. Reaction with sodium bisulfite promotes cytosine deamination, whereas 5-methylcytosine (5-mC) reacts poorly with bisulfite and is resistant to deamination. Since 5-hmC reacts with bisulfite to yield cytosine 5-methylenesulfonate (CMS), we asked how DNA containing 5-hmC behaves in bisulfite sequencing.

Methodology/Principal Findings

We used synthetic oligonucleotides with different distributions of cytosine as templates for generation of DNAs containing C, 5-mC and 5-hmC. The resulting DNAs were subjected in parallel to bisulfite treatment, followed by exposure to conditions promoting cytosine deamination. The extent of conversion of 5-hmC to CMS was estimated to be 99.7%. Sequencing of PCR products showed that neither 5-mC nor 5-hmC undergo C-to-T transitions after bisulfite treatment, confirming that these two modified cytosine species are indistinguishable by the bisulfite technique. DNA in which CMS constituted a large fraction of all bases (28/201) was much less efficiently amplified than DNA in which those bases were 5-mC or uracil (the latter produced by cytosine deamination). Using a series of primer extension experiments, we traced the inefficient amplification of CMS-containing DNA to stalling of Taq polymerase at sites of CMS modification, especially when two CMS bases were either adjacent to one another or separated by 1–2 nucleotides.

Conclusions

We have confirmed that the widely used bisulfite sequencing technique does not distinguish between 5-mC and 5-hmC. Moreover, we show that CMS, the product of bisulfite conversion of 5-hmC, tends to stall DNA polymerases during PCR, suggesting that densely hydroxymethylated regions of DNA may be underrepresented in quantitative methylation analyses.     

MethylQuant: a sensitive method for quantifying methylation of specific cytosines within the genome

Here we present MethylQuant, a novel method that allows accurate quantification of the methylation level of a specific cytosine within a complex genome. This method relies on the well-established treatment of genomic DNA with sodium bisulfite, which converts cytosine into uracil without modifying 5-methyl cytosine. The region of interest is then PCR-amplified and quantification of the methylation status of a specific cytosine is performed by methylation-specific real-time PCR with SYBR Green I using one of the primers whose 3′ end discriminates between the methylation states of this cytosine. The presence of a locked nucleic acid at the 3′ end of the discriminative primer provides the specificity necessary for accurate and sensitive quantification, even when one of the methylation states is present at a level as low as 1% of the overall population. We demonstrate that accurate quantification of the methylation status of specific cytosines can be achieved in biological samples. The method is high-throughput, cost-effective, relatively simple and does not require any specific equipment other than a real-time PCR instrument.     

Cytosine-to-Uracil Deamination by SssI DNA Methyltransferase

The prokaryotic DNA(cytosine-5)methyltransferase M.SssI shares the specificity of eukaryotic DNA methyltransferases (CG) and is an important model and experimental tool in the study of eukaryotic DNA methylation. Previously, M.SssI was shown to be able to catalyze deamination of the target cytosine to uracil if the methyl donor S-adenosyl-methionine (SAM) was missing from the reaction. To test whether this side-activity of the enzyme can be used to distinguish between unmethylated and C5-methylated cytosines in CG dinucleotides, we re-investigated, using a sensitive genetic reversion assay, the cytosine deaminase activity of M.SssI. Confirming previous results we showed that M.SssI can deaminate cytosine to uracil in a slow reaction in the absence of SAM and that the rate of this reaction can be increased by the SAM analogue 5’-amino-5’-deoxyadenosine. We could not detect M.SssI-catalyzed deamination of C5-methylcytosine (^m5C). We found conditions where the rate of M.SssI mediated C-to-U deamination was at least 100-fold higher than the rate of ^m5C-to-T conversion. Although this difference in reactivities suggests that the enzyme could be used to identify C5-methylated cytosines in the epigenetically important CG dinucleotides, the rate of M.SssI mediated cytosine deamination is too low to become an enzymatic alternative to the bisulfite reaction. Amino acid replacements in the presumed SAM binding pocket of M.SssI (F17S and G19D) resulted in greatly reduced methyltransferase activity. The G19D variant showed cytosine deaminase activity in E. coli, at physiological SAM concentrations. Interestingly, the C-to-U deaminase activity was also detectable in an E. coli ung ⁺ host proficient in uracil excision repair.     

Evidence for non-CpG methylation in mammals

In mammals, the existence of cytosine methylation on non-CpG sequences is controversial. Here, we adapted a LuminoMetric-based Assay (LUMA) to determine global non-CpG methylation levels in rodent and human tissues. We observed that < 1% cytosines in non-CpG motifs were methylated in 3T3-L1 fibroblasts, whereas 7–13% cytosines in non-CpG motifs were methylated in mouse tissues or embryonic fibroblasts. Analysis of cytosine methylation in human, rat, and mouse tissues by bisulfite sequencing revealed non-CpG methylation levels up to 7.5% of all non-CpG cytosines. These levels dropped to 1.5% when a second round of PCR was performed prior to bisulfite sequencing, providing an explanation for the common underestimation of non-CpG methylation levels. Collectively, our results provide evidence that non-CpG methylation exists at substantial levels in mammals.     

Bisulfite genomic sequencing: systematic investigation of critical experimental parameters

Bisulfite genomic sequencing is the method of choice for the generation of methylation maps with single-base resolution. The method is based on the selective deamination of cytosine to uracil by treatment with bisulfite and the sequencing of subsequently generated PCR products. In contrast to cytosine, 5-methylcytosine does not react with bisulfite and can therefore be distinguished. In order to investigate the potential for optimization of the method and to determine the critical experimental parameters, we determined the influence of incubation time and incubation temperature on the deamination efficiency and measured the degree of DNA degradation during the bisulfite treatment. We found that maximum conversion rates of cytosine occurred at 55°C (4–18 h) and 95°C (1 h). Under these conditions at least 84–96% of the DNA is degraded. To study the impact of primer selection, homologous DNA templates were constructed possessing cytosine-containing and cytosine-free primer binding sites, respectively. The recognition rates for cytosine (≥97%) and 5-methylcytosine (≥94%) were found to be identical for both templates.     

Rapid quantitation of methylation differences at specific sites using methylation-sensitive single nucleotide primer extension (Ms-SNuPE).

We have developed a rapid quantitative method (Ms-SNuPE) for assessing methylation differences at specific CpG sites based on bisulfite treatment of DNA followed by single nucleotide primer extension. Genomic DNA was first reacted with sodium bisulfite to convert unmethylated cytosine to uracil while leaving 5-methylcytosine unchanged. Amplification of the desired target sequence was then performed using PCR primers specific for bisulfite-converted DNA and the resulting product isolated and used as a template for methylation analysis at the CpG site(s) of interest. This methylation-sensitive technique has several advantages over existing methods used for detection of methylation changes because small amounts of DNA can be analyzed including microdissected pathology sections and it avoids utilization of restriction enzymes for determining the methylation status at CpG sites.     

BSMAP: whole genome bisulfite sequence MAPping program

Background  

Bisulfite sequencing is a powerful technique to study DNA cytosine methylation. Bisulfite treatment followed by PCR amplification specifically converts unmethylated cytosines to thymine. Coupled with next generation sequencing technology, it is able to detect the methylation status of every cytosine in the genome. However, mapping high-throughput bisulfite reads to the reference genome remains a great challenge due to the increased searching space, reduced complexity of bisulfite sequence, asymmetric cytosine to thymine alignments, and multiple CpG heterogeneous methylation.     

APOBEC3G cytosine deamination hotspots are defined by both sequence context and single-stranded DNA secondary structure

Apolipoprotein B mRNA-editing, enzyme-catalytic, polypeptide-like 3G (i.e., APOBEC3G or A3G) is an evolutionarily conserved cytosine deaminase that potently restricts human immunodeficiency virus type 1 (HIV-1), retrotransposons and other viruses. A3G has a nucleotide target site specificity for cytosine dinucleotides, though only certain cytosine dinucleotides are ‘hotspots’ for cytosine deamination, and others experience little or no editing by A3G. The factors that define these critical A3G hotspots are not fully understood. To investigate how A3G hotspots are defined, we used an in vitro fluorescence resonance energy transfer-based oligonucleotide assay to probe the site specificity of A3G. Our findings strongly suggest that the target single-stranded DNA (ssDNA) secondary structure as well as the bases directly 3′ and 5′ of the cytosine dinucleotide are critically important A3G recognition. For instance, A3G cannot readily deaminate a cytosine dinucleotide in ssDNA stem structures or in nucleotide base loops composed of three bases. Single-stranded nucleotide loops up to seven bases in length were poor targets for A3G activity unless cytosine residues flanked the cytosine dinucleotide. Furthermore, we observed that A3G favors adenines, cytosines and thymines flanking the cytosine dinucleotide target in unstructured regions of ssDNA. Low cytosine deaminase activity was detected when guanines flanked the cytosine dinucleotide. Taken together, our findings provide the first demonstration that A3G cytosine deamination hotspots are defined by both the sequence context of the cytosine dinucleotide target as well as the ssDNA secondary structure. This knowledge can be used to better trace the origins of mutations to A3G activity, and illuminate its impact on processes such as HIV-1 genetic variation.     

The CHH motif in sugar beet satellite DNA: a modulator for cytosine methylation

Methylation of DNA is important for the epigenetic silencing of repetitive DNA in plant genomes. Knowledge about the cytosine methylation status of satellite DNAs, a major class of repetitive DNA, is scarce. One reason for this is that arrays of tandemly arranged sequences are usually collapsed in next‐generation sequencing assemblies. We applied strategies to overcome this limitation and quantified the level of cytosine methylation and its pattern in three satellite families of sugar beet (Beta vulgaris) which differ in their abundance, chromosomal localization and monomer size. We visualized methylation levels along pachytene chromosomes with respect to small satellite loci at maximum resolution using chromosome‐wide fluorescent in situ hybridization complemented with immunostaining and super‐resolution microscopy. Only reduced methylation of many satellite arrays was obtained. To investigate methylation at the nucleotide level we performed bisulfite sequencing of 1569 satellite sequences. We found that the level of methylation of cytosine strongly depends on the sequence context: cytosines in the CHH motif show lower methylation (44–52%), while CG and CHG motifs are more strongly methylated. This affects the overall methylation of satellite sequences because CHH occurs frequently while CG and CHG are rare or even absent in the satellite arrays investigated. Evidently, CHH is the major target for modulation of the cytosine methylation level of adjacent monomers within individual arrays and contributes to their epigenetic function. This strongly indicates that asymmetric cytosine methylation plays a role in the epigenetic modification of satellite repeats in plant genomes.     

FadE: whole genome methylation analysis for multiple sequencing platforms

DNA methylation plays a central role in genomic regulation and disease. Sodium bisulfite treatment (SBT) causes unmethylated cytosines to be sequenced as thymine, which allows methylation levels to reflected in the number of ‘C’-‘C’ alignments covering reference cytosines. Di-base color reads produced by lifetech’s SOLiD sequencer provide unreliable results when translated to bases because single sequencing errors effect the downstream sequence. We describe FadE, an algorithm to accurately determine genome-wide methylation rates directly in color or nucleotide space. FadE uses SBT unmethylated and untreated data to determine background error rates and incorporate them into a model which uses Newton–Raphson optimization to estimate the methylation rate and provide a credible interval describing its distribution at every reference cytosine. We sequenced two slides of human fibroblast cell-line bisulfite-converted fragment library with the SOLiD sequencer to investigate genome-wide methylation levels. FadE reported widespread differences in methylation levels across CpG islands and a large number of differentially methylated regions adjacent to genes which compares favorably to the results of an investigation on the same cell-line using nucleotide-space reads at higher coverage levels, suggesting that FadE is an accurate method to estimate genome-wide methylation with color or nucleotide reads. http://code.google.com/p/fade/.     

Reduced representation bisulfite sequencing for comparative high-resolution DNA methylation analysis

We describe a large-scale random approach termed reduced representation bisulfite sequencing (RRBS) for analyzing and comparing genomic methylation patterns. BglII restriction fragments were size-selected to 500–600 bp, equipped with adapters, treated with bisulfite, PCR amplified, cloned and sequenced. We constructed RRBS libraries from murine ES cells and from ES cells lacking DNA methyltransferases Dnmt3a and 3b and with knocked-down (kd) levels of Dnmt1 (Dnmt[1^kd,3a^−/−,3b^−/−]). Sequencing of 960 RRBS clones from Dnmt[1^kd,3a^−/−,3b^−/−] cells generated 343 kb of non-redundant bisulfite sequence covering 66212 cytosines in the genome. All but 38 cytosines had been converted to uracil indicating a conversion rate of >99.9%. Of the remaining cytosines 35 were found in CpG and 3 in CpT dinucleotides. Non-CpG methylation was >250-fold reduced compared with wild-type ES cells, consistent with a role for Dnmt3a and/or Dnmt3b in CpA and CpT methylation. Closer inspection revealed neither a consensus sequence around the methylated sites nor evidence for clustering of residual methylation in the genome. Our findings indicate random loss rather than specific maintenance of methylation in Dnmt[1^kd,3a^−/−,3b^−/−] cells. Near-complete bisulfite conversion and largely unbiased representation of RRBS libraries suggest that random shotgun bisulfite sequencing can be scaled to a genome-wide approach.     

Characterization of the level,target sites and inheritance of cytosine methylation in tomato nuclear DNA

The tomato nuclear genome was determined to have a G+C content of 37% which is among the lowest reported for any plant species. Non-coding regions have a G+C content even lower (32% average) whereas coding regions are considerably richer in G+C (46%).5-methyl cytosine was the only modified base detected and on average 23% of the cytosine residues are methylated. Immature tissues and protoplasts have significantly lower levels of cytosine methylation (average 20%) than mature tissues (average 25%). Mature pollen has an intermediate level of methylation (22%). Seeds gave the highest value (27%), suggesting de novo methylation after pollination and during seed development.Based on isoschizomer studies we estimate 55% of the CpG target sites (detected by Msp I/Hpa II) and 85% of the CpNpG target sites (detected by Bst NI/Eco RI)are methylated. Unmethylated target sites (both CpG and CpNpG) are not randomly distributed throughout the genome, but frequently occur in clusters. These clusters resemble CpG islands recently reported in maize and tobacco.The low G+C content and high levels of cytosine methylation in tomato may be due to previous transitions of 5mCT. This is supported by the fact that G+C levels are lowest in non-coding portions of the genome in which selection is relaxed and thus transitions are more likely to be tolerated. This hypothesis is also supported by the general deficiency of methylation target sites in the tomato genome, especially in non-coding regions.Using methylation isoschizomers and RFLP analysis we have also determined that polymorphism between plants, for cytosine methylation at allelic sites, is common in tomato. Comparing DNA from two tomato species, 20% of the polymorphisms detected by Bst NI/Eco RII could be attributed to differential methylation at the CpNpG target sites. With Msp I/Hpa II, 50% of the polymorphisms were attributable to methylation (CpG and CpNpG sites). Moreover, these polymorphisms were demonstrated to be inherited in a mendelian fashion and to co-segregate with the methylation target site and thus do not represent variation for transacting factors that might be involved in methylation of DNA. The potential role of heritable methylation polymorphism in evolution of gene regulation and in RFLP studies is discussed.     

Kismeth: Analyzer of plant methylation states through bisulfite sequencing

Background  

There is great interest in probing the temporal and spatial patterns of cytosine methylation states in genomes of a variety of organisms. It is hoped that this will shed light on the biological roles of DNA methylation in the epigenetic control of gene expression. Bisulfite sequencing refers to the treatment of isolated DNA with sodium bisulfite to convert unmethylated cytosine to uracil, with PCR converting the uracil to thymidine followed by sequencing of the resultant DNA to detect DNA methylation. For the study of DNA methylation, plants provide an excellent model system, since they can tolerate major changes in their DNA methylation patterns and have long been studied for the effects of DNA methylation on transposons and epimutations. However, in contrast to the situation in animals, there aren't many tools that analyze bisulfite data in plants, which can exhibit methylation of cytosines in a variety of sequence contexts (CG, CHG, and CHH).     

Errors in the bisulfite conversion of DNA: modulating inappropriate- and failed-conversion frequencies

Bisulfite treatment can be used to ascertain the methylation states of individual cytosines in DNA. Ideally, bisulfite treatment deaminates unmethylated cytosines to uracils, and leaves 5-methylcytosines unchanged. Two types of bisulfite-conversion error occur: inappropriate conversion of 5-methylcytosine to thymine, and failure to convert unmethylated cytosine to uracil. Conventional bisulfite treatment requires hours of exposure to low-molarity, low-temperature bisulfite (‘LowMT’) and, sometimes, thermal denaturation. An alternate, high-molarity, high-temperature (‘HighMT’) protocol has been reported to accelerate conversion and to reduce inappropriate conversion. We used molecular encoding to obtain validated, individual-molecule data on failed- and inappropriate-conversion frequencies for LowMT and HighMT treatments of both single-stranded and hairpin-linked oligonucleotides. After accounting for bisulfite-independent error, we found that: (i) inappropriate-conversion events accrue predominantly on molecules exposed to bisulfite after they have attained complete or near-complete conversion; (ii) the HighMT treatment is preferable because it yields greater homogeneity among sites and among molecules in conversion rates, and thus yields more reliable data; (iii) different durations of bisulfite treatment will yield data appropriate to address different experimental questions; and (iv) conversion errors can be used to assess the validity of methylation data collected without the benefit of molecular encoding.     

Methylation in eucaryotes influences the repair of G/T and A/C DNA basepair mismatches

Although methylation of DNA at some sites regulates gene expression, 5mC at many sites does not appear to have any effect. We present evidence that hemimethylation at many different sites can act as a discrimination signal in mismatch repair. Deamination of 5mC in a symmetrically methylated doublet CpG yields the mismatched base pair T/G in a hemi-methylated doublet pair. Because both bases in the mismatched pair are normal constituents of DNA, identifying the incorrect base is problematic. The only apparent distinction of the two is the methylation on the strand opposite the deamination event. Using available methylases we have produced hemi-methylated SV40 DNAs that are mismatched at a single T/G or A/C basepair in a sequence that mimics the lesion resulting from the deamination of a 5mCpG. Methylation at the adjacent cytosine results in the replacement of the T much more frequently than when no methylation is present in the heteroduplex. Cytosine methylation at sites farther removed from the mismatch is equally effective in replacing the incorrect T at the mismatch. Although methylation in vertebrates is almost exclusively on cytosine in the doublet CpG, methylation of cytosines in other doublets, as well as methylation of adenosine, also act as strand discrimination signals. Perhaps some of the excess methylation in vertebrate DNAs may serve to direct mismatch repair.