Pros and cons of methylation-based enrichment methods for ancient DNA

Abstract

The recent discovery that DNA methylation survives in fossil material provides an opportunity for novel molecular approaches in palaeogenomics. Here, we apply to ancient DNA extracts the probe-independent Methylated Binding Domains (MBD)-based enrichment method, which targets DNA molecules containing methylated CpGs. Using remains of a Palaeo-Eskimo Saqqaq individual, woolly mammoths, polar bears and two equine species, we confirm that DNA methylation survives in a variety of tissues, environmental contexts and over a large temporal range (4,000 to over 45,000 years before present). MBD enrichment, however, appears principally biased towards the recovery of CpG-rich and long DNA templates and is limited by the fast post-mortem cytosine deamination rates of methylated epialleles. This method, thus, appears only appropriate for the analysis of ancient methylomes from very well preserved samples, where both DNA fragmentation and deamination have been limited. This work represents an essential step toward the characterization of ancient methylation signatures, which will help understanding the role of epigenetic changes in past environmental and cultural transitions.

Figure 1. Endogenous content, clonality and number of non-unique reads of DNA libraries constructed from MBD+ and MBD− fractions.

(A) Clonality estimates, expressed as the percentage of hits that were PCR duplicates both for those not filtered for mapping quality (MQ0, darker shade) and those filtered for mapping quality (MQ25, lighter shade). (B) Endogenous DNA content, expressed as the natural logarithm of the fractions of unique hits over the total number of reads retained both for those not filtered for mapping quality (MQ0, lighter shade) and those filtered for mapping quality (MQ25, darker shade). (C) Fraction of nuclear DNA reads non-uniquely mapped to the reference genome calculated on the full-length reads (minimum 2,000 per sample) not filtered for mapping quality (MQ0). MBD+ = MBD enriched fraction. MBD− = uncaptured fraction.

Figure 2. Measure of regional methylation level based on the average CpG→TpG mismatch frequencies observed at read starts (Ms), as defined by Pedersen and colleagues.

Ms was calculated within 100 bp genomic windows and two minimal coverage thresholds were tested (50 and 60 per window, respectively).

Figure 3. Molecular characteristics of DNA library inserts and post-mortem DNA degradation.

(A) Percentage of CpGs in the nuclear DNA, normalized by read length. (B) Length of DNA library inserts. (C) Cytosine deamination probability in double-stranded DNA context (δ_d). (D) Cytosine deamination probability in single-stranded DNA contexts (δ_s). DNA degradation parameters were calculated using mapDamage 2. A minimum threshold of 2,000 reads was applied to all data shown (mapping quality filtered, MQ25) so as to retrieve unbiased distributions. MBD+ = MBD enriched fraction. MBD− = uncaptured fraction.

Figure 4. Relative amount of CpGs to TpGs transitions over all other kind of possible C->T mutations in CpN contexts.

A minimum threshold of 2,000 reads was applied.

Figure 5. Per-base coverage distributions.

The dominant class of uncovered bases is not shown and the distributions are restricted to bases covered at least once in the nuclear genome. A minimum threshold of 2,000 reads was applied.

Figure 6. Properties of MBD enrichment for ancient DNA.

The MBD+ fraction favors DNA templates with a longer size and a higher number of ^mCpGs. This represents a limited fraction of the molecules present in aDNA extracts. mtDNA is under-represented in the MBD+ fraction while higher coverage is observed for hyper-methylated regions of the nuclear genome, including TEs.