Human adaptation and population differentiation in the light of ancient genomes

Abstract

The influence of positive selection sweeps in human evolution is increasingly debated, although our ability to detect them is hampered by inherent uncertainties in the timing of past events. Ancient genomes provide snapshots of allele frequencies in the past and can help address this question. We combine modern and ancient genomic data in a simple statistic (DAnc) to time allele frequency changes, and investigate the role of drift and adaptation in population differentiation. Only 30% of the most strongly differentiated alleles between Africans and Eurasians changed in frequency during the colonization of Eurasia, but in Europe these alleles are enriched in genic and putatively functional alleles to an extent only compatible with local adaptation. Adaptive alleles--especially those associated with pigmentation--are mostly of hunter-gatherer origin, although lactose persistence arose in a haplotype present in farmers. These results provide evidence for a role of local adaptation in human population differentiation.

Figure 1. Analyses of population differentiation.

(a,b) Genic enrichment among alleles in different bins of population differentiation between (a) CHB or (b) GBR and YRI. The bootstrap 95% confidence interval is shown in yellow, and the level of significance of the bias in genic enrichment when comparing the two tails is shown on top (*P< 0.05, **P<0.01 and ***P<0.001, and NS non-significant). The number of genic sites in each tail is on top of the tails. (c,d) Genic enrichment in alleles in different bins of the DAnc statistic (YRI, P₂, Ust′-Ishim (UI)) using (c) CHB or (d) GBR as P₂. The 95% confidence interval, the significance of the bias, and the number of genic alleles are shown as above. (e,f) Genic enrichment in alleles in different bins of the DAnc statistic when considering each East Asian (e) or European (f) population as P₂, with significance of the bias between the tails indicated by asterisks (*<0.05, **<0.01 and ***<0.001). (g,h) Expectation of the genic enrichment across the DAnc distribution under non-adaptive forces, based on simulations with different strengths of background selection (measured by B scores) and with no background selection (B=1).

Figure 2. Tree of inferred relationships of present-day and ancient humans considered.

(a) Including the ancient Ust′-Ishim sample, and (b) including the ancient Europeans samples. Colours indicate the branches where alleles in the European (blue) and African (orange) tails of the DAnc distribution most likely changed in allele frequency. For demographic parameters used see Methods and Supplementary Fig. 7.

Figure 3. Conservation and putatively regulatory role of genic alleles in DAnc tails.

Conservation was measured with phastCons, with (a) phastCons >0.9 representing strongly conserved sites and (b) phastCons <0.1 non-conserved sites. Observed frequencies of genic alleles with different levels of phylogenetic conservation are shown for the DAnc(YRI, Europe, UI) tails and the DAnc(YRI, East Asia, UI) tails using a coloured dot (blue for European, green for East Asian, orange for African). On the basis of resampling from the entire distribution the mean expectation is shown as a black dot and the 95% confidence interval as black lines. Asterisks indicate significant differences between observations and expectation (*P<0.05, **P<0.01 and ***P<0.001). c and d show the same analysis for putative regulatory function according to regulomeDB, with (c) regulomeDB category 1 representing likely regulatory sites and (d) category 7 representing sites with no evidence of a regulatory role. AFR, African (YRI); EA, East Asia; EUR, Europe; UI, Ust′-Ishim.

Figure 4. Analysis of the presence of European DAnc tail alleles in ancient European genomes.

(a) Proportion of European alleles present in Loschbour (LB) or Stuttgart (STG) among those in the tails of the DAnc (YRI, GBR, Ust′-Ishim) distribution. Bootstrapping provided confidence intervals (95% confidence interval shown) and asterisks indicate significant differences between LB and STG (*P<0.05, ^**P<0.01 and ^***P<0.001). (b,c) Expectation of these proportions in the (b) European or (c) African tails, in the absence of adaptive evolution and with various intensities of background selection (measured by B scores). Genic alleles are simulated with B<1, and non-genic alleles with B=1. The 95% confidence interval shown is obtained via bootstrapping.

Figure 5. Genic alleles in the DAnc(YRI, Europe, UI) tail and overlapping genes.

Close-by alleles are combined in a region if they are <100-kb apart. The left side of the figure shows the chromosomal position (chr:start) and the length in bp of each region. The right side shows the genes that overlap with each region. The number of alleles in each genic DAnc tail (P₂: GBR, FIN, TSI, CEU) is shown with the heatmap. Columns are ordered by number of alleles starting left. Superscripts note regions with too many genes to fit the figure: ^A: ANKRD13D, SSH3, CLCF1, PPP1CA, TBC1D10C, CARNS1, KDM2A; ^B: CCDC142, PCGF1, DQX1, AUP1, DCTN1, C2orf81, RTKN, LBX2, TLX2, HTRA2, LOXL3, DOK1, SEMA4F, TTC31, M1AP, RP11-287D1.3, SLC4A5; ^C: ACAP3, PUSL1, GLTPD1, DVL1, MXRA8, CCNL2, AURKAIP1.

Figure 6. Haplotypes in ancient and present-day modern humans.

Haplotypes for (a) OCA2/HERC2, (b) SLC45A2 and (c) LCT/MCM6. SNPs were included if they show moderate to strong LD with the phenotypically relevant SNP (OCA2/HERC2: rs12913832, SLC45A2: rs16891982, LCT/MCM6: rs4988235), which is marked with an orange arrow. Black arrows indicate alleles that are present in at least one European DAnc tail (P₂: GBR, FIN, TSI or CEU). The heatmap shows the phased haplotype for the 1000 Genomes data set (red derived allele and blue ancestral allele) and the genotype for the ancient genomes (red homozygous derived, purple heterozygous and blue homozygous ancestral).