A High-Coverage Mesolithic Aurochs Genome and Effective Leveraging of Ancient Cattle Genomes Using Whole Genome Imputation

Ancient genomic analyses are often restricted to utilizing pseudohaploid data due to low genome coverage. Leveraging low-coverage data by imputation to calculate phased diploid genotypes that enables haplotype-based interrogation and single nucleotide polymorphism (SNP) calling at unsequenced positions is highly desirable. This has not been investigated for ancient cattle genomes despite these being compelling subjects for archeological, evolutionary, and economic reasons. Here, we test this approach by sequencing a Mesolithic European aurochs (18.49×; 9,852 to 9,376 calBCE) and an Early Medieval European cow (18.69×; 427 to 580 calCE) and combine these with published individuals: two ancient and three modern. We downsample these genomes (0.25×, 0.5×, 1.0×, and 2.0×) and impute diploid genotypes, utilizing a reference panel of 171 published modern cattle genomes that we curated for 21.7 million (Mn) phased SNPs. We recover high densities of correct calls with an accuracy of >99.1% at variant sites for the lowest downsample depth of 0.25×, increasing to >99.5% for 2.0× (transversions only, minor allele frequency [MAF] ≥ 2.5%). The recovery of SNPs correlates with coverage; on average, 58% of sites are recovered for 0.25× increasing to 87% for 2.0×, utilizing an average of 3.5 million (Mn) transversions (MAF ≥2.5%), even in the aurochs, despite the highest temporal distance from the modern reference panel. Our imputed genomes behave similarly to directly called data in allele frequency-based analyses, for example consistently identifying runs of homozygosity >2 Mb, including a long homozygous region in the Mesolithic European aurochs.