Evidence that two main bottleneck events shaped modern human genetic diversity

Abstract

There is a strong consensus that modern humans originated in Africa and moved out to colonize the world approximately 50 000 years ago. During the process of expansion, variability was lost, creating a linear gradient of decreasing diversity with increasing distance from Africa. However, the exact way in which this loss occurred remains somewhat unclear: did it involve one, a few or a continuous series of population bottlenecks? We addressed this by analysing a large published dataset of 783 microsatellite loci genotyped in 53 worldwide populations, using the program 'Bottleneck'. Immediately following a sharp population decline, rare alleles are lost faster than heterozygosity, creating a transient excess of heterozygosity relative to allele number, a feature that is used by Bottleneck to infer historical events. We find evidence of two primary events, one 'out of Africa' and one placed around the Bering Strait, where an ancient land bridge allowed passage into the Americas. These findings agree well with the regions of the world where the largest founder events might have been expected, but contrast with the apparently smooth gradient of variability that is revealed when current heterozygosity is plotted against distance from Africa.

Figure 1.

Variation in strength of evidence of a bottleneck (fitted t-value) with distance from Africa and heterozygosity in Africa. (a) A three-dimensional plot of the fitted values from a general linear model (table 1). (b,c) Plots of how fitted t-values vary with distance from Africa (±s.e.m.) for (b) low (heterozygosity = 0.3) and (c) high (heterozygosity = 0.9) heterozygosity values. In all cases, the response variable is transformed t, and higher values indicate stronger evidence of a bottleneck.

Figure 2.

Variation in strength of evidence of a bottleneck (raw data) with distance from Africa and heterozygosity in Africa. This is the same plot as figure 1a except that fitted t-values have been replaced with the raw data and a surface fitted using local regression. Heterozygosity is taken as that in Africa, as estimated from the Biaka Pygmy sample.

Figure 3.

Variation in strength of evidence of a bottleneck (raw data) with distance from (a) <0.5 (n = 14), (b) 0.5–0.55 (n = 17), (c) 0.55–0.6 (n = 13), (d) 0.6–0.65 (n = 46), (e) 0.65–0.7 (n = 99), (f) 0.7–0.75 (n = 162), (g) 0.75–0.8 (n = 201) and (h) >0.8 (n = 231). Africa and heterozygosity in Africa. Each panel represents a slice through figure 2 for a different range of heterozygosity values: Bin boundaries were selected to lie 0.05 apart unless this embraced fewer than 10 observations, in which case they were enlarged. In each case local regression is used to fit a smoothed spline and the derive 95 per cent confidence intervals on that spline. The response variable is the transformed t-value, higher values indicating stronger evidence of a bottleneck.

Figure 4.

Variation in strength of evidence of a bottleneck with distance from Africa by individual population. Populations are coded: open circles, Africa; grey squares, Middle East; grey diamonds, Europe; black triangles, Central Southern Asia; open squares, East Asia; black circles, Oceania; grey triangles, America. The response variable is the transformed t-value, higher values indicating stronger evidence of a bottleneck. Error bars are ± 1 s.e.m.