Selective enrichment of damaged DNA molecules for ancient genome sequencing

Contamination by present-day human and microbial DNA is one of the major hindrances for large-scale genomic studies using ancient biological material. We describe a new molecular method, U selection, which exploits one of the most distinctive features of ancient DNA—the presence of deoxyuracils—for selective enrichment of endogenous DNA against a complex background of contamination during DNA library preparation. By applying the method to Neanderthal DNA extracts that are heavily contaminated with present-day human DNA, we show that the fraction of useful sequence information increases ∼10-fold and that the resulting sequences are more efficiently depleted of human contamination than when using purely computational approaches. Furthermore, we show that U selection can lead to a four- to fivefold increase in the proportion of endogenous DNA sequences relative to those of microbial contaminants in some samples. U selection may thus help to lower the costs for ancient genome sequencing of nonhuman samples also.High-throughput DNA sequencing and the shift to library-based sample preparation techniques have greatly facilitated genetic research on human evolution in recent years. Increasing amounts of sequence data are becoming available not only from present-day humans but also from ancient human remains, helping to uncover the evolutionary histories of present-day human populations as well as their relationship to extinct archaic groups (Stoneking and Krause 2011). In ancient DNA studies, the most accessible target is mitochondrial (mt) DNA, which is present in several hundreds of copies per cell as opposed to the diploid-only nuclear genome. Consequently, complete mtDNA genomes have been sequenced from more than a dozen remains of ancient modern humans (Ermini et al. 2008; Gilbert et al. 2008; Green et al. 2010; Fu et al. 2013b; Raghavan et al. 2014) as well as representatives of two archaic hominin groups that went extinct during the Late Pleistocene, the Neanderthals (Green et al. 2008; Briggs et al. 2009; Prüfer et al. 2014), and Denisovans (Krause et al. 2010b; Reich et al. 2010). The recent recovery of mitochondrial sequences from an ∼400,000 yr-old hominin from Sima de los Huesos in Spain indicates that more comprehensive sequence data may soon become available even from Middle Pleistocene remains (Meyer et al. 2014). Retrieving nuclear sequences from ancient human material is generally more challenging, but such data have also been generated at various scales, ranging from a few megabases of sequences to full genome sequences determined with high accuracy (see Shapiro and Hofreiter 2014 for a recent summary).Despite these advances, sequencing ancient human DNA continues to be challenging for several reasons. First, only trace amounts of highly fragmented DNA are usually preserved in ancient bones and teeth, imposing limits on the number of sequences that can be recovered from ancient specimens. Second, DNA extracted from ancient material is in many cases dominated by microbial DNA, which often contributes to 99% or more of the sequences, making direct shotgun sequencing economically infeasible. This problem can be partially overcome by hybridization enrichment of hominin sequences, such as those of a small chromosome (Fu et al. 2013a) or, as recently proposed, of sequences from throughout the whole genome (Carpenter et al. 2013). Another approach is restriction digestion of GC-rich sequence motifs, which was performed to change the ratio between endogenous and microbial library molecules in the first study of the Neanderthal genome (Green et al. 2010). A third and particularly severe problem for working with ancient human samples is present-day human contamination, which is inevitably introduced during excavation and laboratory work. Fortunately, a solid framework for validating the authenticity of ancient human sequences can be established using the distinct pattern of substitutions caused by cytosine deamination in ancient DNA sequences (Hofreiter et al. 2001; Briggs et al. 2007). The deamination product of cytosine is uracil, which is read as thymine by most DNA polymerases. Resulting C to T substitutions (or G to A substitutions, depending on the orientation in which DNA strands are sequenced and the method used to prepare DNA libraries) are particularly frequent at the 5′ and 3′ ends of sequences due to the higher rate of cytosine deamination in single-stranded overhangs (Lindahl 1993). Importantly, the frequency of deamination-induced substitutions correlates with sample age (Sawyer et al. 2012) and is low in present-day human contamination (Krause et al. 2010a). These substitutions can thus be taken as evidence for the presence of authentic ancient human sequences. Deamination-induced substitutions have also been exploited for separating ancient sequences from present-day human contamination in silico (Skoglund et al. 2012, 2014a,b; Meyer et al. 2014; Raghavan et al. 2014). Although effective in principle, this approach is costly, because a large proportion of sequence data is excluded from downstream analysis. Furthermore, ancient DNA base damage is not determined directly and may occasionally be confounded with evolutionary sequence differences.Here we describe a novel laboratory technique, uracil selection (“U selection”), which enables physical separation of uracil-containing DNA strands from nondeaminated strands at the stage of DNA library preparation. Our method builds on a single-stranded library preparation method, which has been shown to be particularly efficient for retrieving sequences from highly degraded ancient DNA (Meyer et al. 2012; Gansauge and Meyer 2013). We apply U selection to several Neanderthal DNA extracts and show that it is a powerful tool for enriching Neanderthal DNA sequences against a background of present-day human contamination. We also report cases where U selection drastically increases the proportion of Neanderthal DNA relative to microbial DNA.