Evolutionary Patterns and Processes: Lessons from Ancient DNA

Abstract

Ever since its emergence in 1984, the field of ancient DNA has struggled to overcome the challenges related to the decay of DNA molecules in the fossil record. With the recent development of high-throughput DNA sequencing technologies and molecular techniques tailored to ultra-damaged templates, it has now come of age, merging together approaches in phylogenomics, population genomics, epigenomics, and metagenomics. Leveraging on complete temporal sample series, ancient DNA provides direct access to the most important dimension in evolution—time, allowing a wealth of fundamental evolutionary processes to be addressed at unprecedented resolution. This review taps into the most recent findings in ancient DNA research to present analyses of ancient genomic and metagenomic data.

Keywords: Ancient DNA; metagenomics; palaeogenomics; population genomics; temporal sampling.

FIG. 1

Effect of ancient DNA degradation on phylogenetic reconstruction, when including modern and ancient sequences. The continuous lines show the hypothetic tree, the points mark the true chronological location of the ancient samples, while the dotted lines represent the branch length artificially increased by DNA degradation-related mutations.

FIG. 2

Skyline-plot methods. a) Overview on how skyline-plot methods are used to reconstruct the variation of effective population size through time. After the reconstruction of the gene genealogy (step 1), the effective population size is estimated over time, on the basis of the density of coalescent events present in each corresponding time interval (step 2). The thick solid line is the median estimate, and the coloured area shows the 95% bounds of the highest posterior density (HPD). b) Bayesian skyline plot reconstructed from mitochondrial genome data of 34 ancient (dotted line) and 105 modern (continuous line) horse samples. (Table S1 available as Supplementary material in Dryad as http://dx.doi.org/10.5061/dryad.tt78r), calibrated by adding the donkey (Orlando et al. 2013) as outgroup, considering a split time of 2.75 Myrs 650 kyr BP based on the most credible range estimate for the date when caballine (horse) and non-caballine (donkeys, asses, and zebras) equine lineages became genetically isolated (Jónsson et al. 2014). Bayesian phylogenetic inference was performed with BEAST 1.8.2 (Drummond et al. 2012), considering 5 site categories treated as unlinked partitions, following (Achilli et al. 2012). Divergence was estimated using a log-uncorrelated molecular clock model, including dating of ancient samples as calibration tips. The phylogenetic inference was run for 200 million generations, sampling the chain every 30,000, with 10% burn-in. Convergence was checked with Tracer v1.6.0 (Rambaut et al. 2014), also used for reconstructing the Bayesian skyline demographic profile, plotted with ggplot2 package (Wickham 2009), R (R et al. 2015).

FIG. 3

Likelihood Ratio Test for direct ancestry implemented in Rasmussen et al. (2014). The alternative and null model differ in a single parameter, , representing the genetic drift leading to the ancient lineage. The method fully models the 2D-SFS of the observed samples to infer the most likely parameter values, including . The null model, considering the ancient sample belonging to the population directly ancestral to the modern one, is a particular case of the alternative one, where is equal to 0.

FIG. 4

Global and local ancestry methods. a) -statistics correspond to a global ancestry method testing for shared drift. They assess if a test population (H descends from two parental populations (H and H. Here we represent the test carried out for the Neanderthal interbreeding with non-African populations, which estimated 1.5 to 2.1% Neanderthal contribution to the present non-Africans (Prüfer et al. 2014). b) Local ancestry methods infer the date of admixture based on the length distribution of introgressed tracts. Here we show the Neanderthal introgression into the modern non-African population. Three ancient AMH helped dating the admixture to 50–60 kyr ago (Seguin-Orlando et al. 2014; Fu et al. 2016): Ust’-Ishim, a -kyr-old AMH from Siberia (Fu et al. 2014); Kostenki K14, a -kyr-old individual from Russian Federation (Seguin-Orlando et al. 2014), and; Mal’ta 1, a -kyr-old individual from Siberia (Raghavan et al. 2014b). Similar estimates (40–65 kyrs ago) have been found with a different method based on recombination rates (Sankararaman et al. 2012; Moorjani et al. 2016) (see sections on Calibration and Divergence Estimates, and Local Ancestry and Admixture).

FIG. 5

-statistics. -statistics are built on loci showing patterns of incomplete lineage sorting (a). A -statistics value deviating from zero indicates an excess of ABBA or BABA events, pointing to a shared ancestry or admixture event with one of the populations tested (H and H. In case of admixture events, the -statistic and -ratio (b) allow quantification of the admixture proportions.