Separating endogenous ancient DNA from modern day contamination in a Siberian Neandertal

Abstract

One of the main impediments for obtaining DNA sequences from ancient human skeletons is the presence of contaminating modern human DNA molecules in many fossil samples and laboratory reagents. However, DNA fragments isolated from ancient specimens show a characteristic DNA damage pattern caused by miscoding lesions that differs from present day DNA sequences. Here, we develop a framework for evaluating the likelihood of a sequence originating from a model with postmortem degradation-summarized in a postmortem degradation score-which allows the identification of DNA fragments that are unlikely to originate from present day sources. We apply this approach to a contaminated Neandertal specimen from Okladnikov Cave in Siberia to isolate its endogenous DNA from modern human contaminants and show that the reconstructed mitochondrial genome sequence is more closely related to the variation of Western Neandertals than what was discernible from previous analyses. Our method opens up the potential for genomic analysis of contaminated fossil material.

Keywords: human evolution; paleogenomics.

Fig. 1.

PMDS distribution in ancient and present day human sequences. (A) A present day human. (B) Hundred-year-old remains of an Australian individual. (C) Five thousand-year-old remains of a Scandinavian individual. (D) A Pleistocene Neandertal. Note that the x axes in C and D are truncated, and PMD scores reach ∼26.

Fig. 2.

Artificial mtDNA and autosomal contamination experiments. (A) Artificial contamination represented by sequences from a present day human (French) was added to a sequence dataset from a Croatian Neandertal (Vindija 33.16). Final per base pair contamination was estimated using diagnostic mtDNA positions that differentiate Neandertals from modern humans. Up to 90% per base pair contamination levels can be reduced to negligible levels using the PMDS approach. (B) Artificial contamination represented by sequences from a present day human (French) was added to a sequence dataset from Neolithic Scandinavian hunter-gatherers. SNPs were extracted with and without filtering for PMDS, and a principal component analysis together with European and Levantine populations was performed for the different levels of contamination. The PCs were Procrustes transformed for each separate analysis (8).

Fig. 3.

Reduction of artificial and genuine contamination using PMDSs. (A) Expected contamination fraction in the Vindija 33.16 Neandertal as a function of the PMDS threshold using sequence data from a present day French individual as the artificial contaminant. (B) Expected contamination fraction in the Ajv70 Scandinavian hunter-gatherer as a function of the PMDS threshold using sequence data from the 100-y-old remains of an Australian individual as the artificial contaminant. The different colored lines in A and B correspond to different contamination levels (before PMDS filtering, the initial contamination fraction is given at PMDS = −2). (C) Estimated modern human contamination in three genuinely contaminated Neandertal datasets from Feldhofer 2, Mezmaiskaya 2, and Okladnikov 2.

Fig. 4.

A mt genome sequence from a contaminated Neandertal sample from Okladnikov Cave, Siberia. (A) Coverage distribution and sequence features of the Okladnikov 2 mt genome. (B) Gene genealogy of 67 published complete hominid mitochondria and Okladnikov 2. Node support is only indicated for posterior probabilities ≥ 0.75. (C) Geographical sampling location of Neandertal and Denisovan individuals from which mtDNA genomes have been sequenced. Okladnikov Cave is marked by a red square. GC, Guanine-Cytosine content.