Genome-scale sequencing and analysis of human, wolf, and bison DNA from 25,000-year-old sediment

Abstract

Cave sediments have been shown to preserve ancient DNA but so far have not yielded the genome-scale information of skeletal remains. We retrieved and analyzed human and mammalian nuclear and mitochondrial environmental "shotgun" genomes from a single 25,000-year-old Upper Paleolithic sediment sample from Satsurblia cave, western Georgia:first, a human environmental genome with substantial basal Eurasian ancestry, which was an ancestral component of the majority of post-Ice Age people in the Near East, North Africa, and parts of Europe; second, a wolf environmental genome that is basal to extant Eurasian wolves and dogs and represents a previously unknown, likely extinct, Caucasian lineage; and third, a European bison environmental genome that is basal to present-day populations, suggesting that population structure has been substantially reshaped since the Last Glacial Maximum. Our results provide new insights into the Late Pleistocene genetic histories of these three species and demonstrate that direct shotgun sequencing of sediment DNA, without target enrichment methods, can yield genome-wide data informative of ancestry and phylogenetic relationships.

Keywords: Canis; Caucasus; Upper Paleolithic; bison; enviromental DNA; human; shotgun; soil sequencing.

Figure 1

Site description: an overview of the location of Satsurblia cave with contextual information (A) Map of the Caucasus region showing relevant sites that have yielded ancient DNA from humans (blue dots), animals (red dot), or both (purple dot). Only sites with remains from the Mesolithic or older are shown. (B) Layer B of Satsurblia cave with the SAT29 sampling location shown. (C) Two microphotographs (A and B) from block sample SAT 15 14. The microphotographs were taken adjacent to SAT 16 LS29, in cross- (XPL) and plane-polarized light (PPL), showing dominant components and processes. Typical anthropogenic components are bone (B), burnt bone (BB), charcoal (Ch), and flint (Fl). Note also the occurrence of rounded soil aggregates (yellow circles) that were transported into the cave from soils forming outside and that exhibit cross-striations of the clay component resulting from repeated wetting and drying cycles during their formation. Similarly, clay coatings in voids (blue arrow) result from water percolating through the sediment. See also Data S1.

Figure 2

Sequencing data properties (A) Metagenomic prediction with centrifuge. (B) Deamination patterns for the filtered reads mapped against the four reference genomes: Ovis aries, Homo sapiens, Bos taurus, and Canis lupus. (C) Fragment length distributions for the four aligned mammalian species. (D) X chromosome read proportions. For each of the three species, the fraction of nuclear reads mapping to the X chromosome is displayed for the SAT29 sample as well as a number of previously published genomes for comparison. For wolf and bison, the comparative genomes are the modern and ancient genomes used for the ancestry analyses, while for human they are 101 ancient genomes from a study of the Eurasian steppe. Bars denote 95% binomial confidence intervals. See also Figures S1 and S6 and Data S1.

Figure 3

SAT29 human genomics (A) A principal component analysis built with 2,335 modern Eurasian samples into which 82 ancient samples were projected. SAT29 appears closest to Dzudzuana2. (B) In ADMIXTURE clustering with 82 ancient genomes, SAT29 displays a profile similar to that of Dzudzuana2. (C) Bayesian tree built with 69 ancient genomes and 167 modern mitochondrial genomes. SAT29 falls close to the Dzudzuana2 genome and within the N haplogroup diversity. The values in the tree nodes represent the posterior probabilities. See also Figures S3 and S4 and Tables S2 and S3.

Figure 4

SAT29 wolf genomics (A) In ADMIXTURE clustering with wolves, dogs, and other canid species, the SAT29 sample clusters with Eurasian wolves. (B) Population history model relating the SAT29 sample to modern wolves, dogs, and Pleistocene Siberian wolves. Inferred admixture proportions from the dog reference genome (REF) to account for reference bias, are shown in blue. The trifurcation point indicates that it cannot confidently be determined on which side of this point the SAT29 sample falls. (C) Bayesian tree of 68 modern and 39 ancient wolf mitochondrial genomes. See also Figures S3 and S5 and Table S4.

Figure 5

SAT29 bison genomics (A) In ADMIXTURE clustering with cattle, aurochs, bison, gayals, and bantengs, the SAT29 sample clusters with bison. (B) A population history model relating the SAT29 sample to present-day American and historical (early 20^th century) European bison from Poland and the Caucasus. (C) Bayesian tree built with SAT29 and other 70 bovid mitochondrial genomes. See also Figure S3 and Table S5.