The genetic prehistory of southern Africa

Abstract

Southern and eastern African populations that speak non-Bantu languages with click consonants are known to harbour some of the most ancient genetic lineages in humans, but their relationships are poorly understood. Here, we report data from 23 populations analysed at over half a million single-nucleotide polymorphisms, using a genome-wide array designed for studying human history. The southern African Khoisan fall into two genetic groups, loosely corresponding to the northwestern and southeastern Kalahari, which we show separated within the last 30,000 years. We find that all individuals derive at least a few percent of their genomes from admixture with non-Khoisan populations that began ∼1,200 years ago. In addition, the East African Hadza and Sandawe derive a fraction of their ancestry from admixture with a population related to the Khoisan, supporting the hypothesis of an ancient link between southern and eastern Africa.

Figure 1. Population structure in southern Africa.

(a) Approximate locations of sampled populations. Populations are coloured according to linguistic affiliation, as indicated in the legend and Supplementary Fig. S1. The speckled region is the Kalahari semi-desert. (b) PCA on SNPs ascertained in a Ju|'hoan individual. Shown are the positions of each individual along the first and second axes of genetic variation, with symbols denoting the individual's population and linguistic affiliation using the same colour coding as in panel a.

Figure 2. All Khoisan populations are admixed.

(a) Admixture linkage disequilibrium (LD) in the Ju|'hoan_North. For each pair of SNPs in the Ju|'hoan_North (black) or the Yoruba (grey), we estimate the LD and the product of the differences in allele frequency between the Ju|'hoan_North and the Yoruba. (We use the Yoruba as a proxy for the non-Khoisan, presumably Bantu-speaking ancestral population because there has been very little change in allele frequencies between Niger-Congo-speaking groups.) We then binned pairs of SNPs by the genetic distance between them. For each bin, we plot the regression coefficient (over SNP pairs in the bin) from regressing the level of LD on the product of the allele frequency differences. The rate at which this curve decays is informative about the date of admixture, whereas the amplitude of the curve is informative about the proportion of admixture (Supplementary Methods). In black is the curve if we assume the Ju|'hoan_North are admixed; in grey is the curve if we assume the Yoruba are admixed (which serves as a negative control). The red line is the exponential curve fitted to the black points. (b) Estimates of mixture proportions. We used the modified f₄ ratio (Supplementary Methods) to estimate the fraction of non-Khoisan ancestry in each southern African population. (c) Estimates of mixture dates. We used the rate at which admixture LD decays to estimate dates of admixture for all southern African populations (Supplementary Methods). We plot the means, with ranges representing one standard error. Not shown are the Wambo, who have no detectable curve and hence may be unadmixed. The estimates of the mixture proportions and dates are also presented in Supplementary Table S4.

Figure 3. Relationships among Khoisan and eastern Africans after removing non-Khoisan admixture.

We extended TreeMix to build trees after subtracting out the effect of known admixture (Supplementary Methods) and then applied it to the Khoisan (excluding the Damara, who are genetically close to non-Khoisan). Populations are coloured according to their linguistic affiliation (Khoisan) or geographic location (dark grey=non-Khoisan African, light grey=Eurasian), and the chimpanzee was used as an out-group. The bar chart next to each population shows the estimated ancestry proportions for each population: blue is the proportion of Khoisan ancestry, and red is the proportion of non-Khoisan ancestry. Note that the actual source of these two ancestries may vary among populations. The proportions are not identical to those presented in Fig. 2b because of small differences in how they are estimated. The black dots show splits supported by more than 95% of bootstrap replicates, and the grey dots those supported by more than 80% of bootstrap replicates.