The Effect of an Extreme and Prolonged Population Bottleneck on Patterns of Deleterious Variation: Insights from the Greenlandic Inuit

Abstract

The genetic consequences of population bottlenecks on patterns of deleterious genetic variation in human populations are of tremendous interest. Based on exome sequencing of 18 Greenlandic Inuit we show that the Inuit have undergone a severe ∼20,000-year-long bottleneck. This has led to a markedly more extreme distribution of allele frequencies than seen for any other human population tested to date, making the Inuit the perfect population for investigating the effect of a bottleneck on patterns of deleterious variation. When comparing proxies for genetic load that assume an additive effect of deleterious alleles, the Inuit show, at most, a slight increase in load compared to European, East Asian, and African populations. Specifically, we observe <4% increase in the number of derived deleterious alleles in the Inuit. In contrast, proxies for genetic load under a recessive model suggest that the Inuit have a significantly higher load (20% increase or more) compared to other less bottlenecked human populations. Forward simulations under realistic models of demography support our empirical findings, showing up to a 6% increase in the genetic load for the Inuit population across all models of dominance. Further, the Inuit population carries fewer deleterious variants than other human populations, but those that are present tend to be at higher frequency than in other populations. Overall, our results show how recent demographic history has affected patterns of deleterious variants in human populations.

Keywords: disease mapping; founder population; genetic load; isolated human populations; neutral theory.

Figure 1

Inferred history of population size for the Greenlandic Inuit population. The dark pink line shows the estimated diploid population size changes in discrete increments for the last 50,000 years. The estimates were obtained with the method “Stairway plot,” which bases its estimates on the SFS. The estimates are based on an assumption of a mutation rate of 1.38 × 10⁻⁸ per site per generation and a generation time of 25 years. The light pink lines represent 95% CI based on bootstraps. This analysis was based on 41,222,102 sites.

Figure 2

Site frequency spectrum for six human populations. We used 11 randomly sampled individuals from each population to infer the site frequency spectrum and excluded fixed categories. Populations contain individuals with the following ancestry: Finns from Finland (FIN), Peruvians from Lima, Peru (PEL), Gujarati Indians from Houston, Texas (GIH), Utah Residents (CEPH) with Northern and Western Ancestry (CEU), Yoruba in Ibadan, Nigeria (YRI), and Han Chinese in Bejing, China (CHB). The GI population has fewer sites in the singleton category, but more in the remaining more “common” categories. Each population is followed by a π_var estimate per variable site. The idea for this figure was adopted from the SFS comparison of four Hapmap populations and GI in Moltke et al. (2015).

Figure 3

Comparison of the number of derived alleles and homozygous genotypes across populations for variants in different GERP score categories. For each of four GERP score categories two ratios calculated based on sites located in exons are shown: (A) the ratio of derived allele counts in GI vs. CHB (orange), CEU (red), and YRI (turquoise); and (B) the ratio of homozygous-derived genotype counts in GI vs. CHB (light green), CEU (blue), and YRI (dark green). The former ratio can be viewed as an approximation for the ratio of load between the populations under an additive model and latter ratio can be viewed as an approximation for the ratio of load between the populations under a recessive model. Standard error for each ratio is indicated by error bars. Additional information is available in Table 1, Table S1, and Table S2. The * indicates significance compared to the neutral GERP score category (GERP <2), for the CEU and YRI comparison this value is significant (P = 5.3 × 10⁻⁴ and P = 5.9 × 10⁻⁹, respectively).

Figure 4

GERP score load and stratification of deleterious alleles based on frequency. (A) GERP score load for GI and CEU calculated assuming a fully additive model (top two bars) and assuming a fully recessive model (bottom two bars). The GERP score load is the approximation to the genetic load based on annotated GERP scores converted to selection coefficients using the approach used in the review by Henn et al. (2015). We note that if the selection coefficients from Henn et al. (2016) are used instead, we see qualitatively similar results. Black error bars indicate 95% C.I. (B) The proportion of deleterious variants classified by GERP score (Moderate: 2 < GERP < 4, Large: 4 < GERP < 6, Extreme: 6 < GERP) that belong to different frequency-based categories in GI and CEU: low frequency SNPs vs. common SNPs. Low-frequency SNPs are here defined as singletons and doubletons (equivalent of a frequency of at most 1/18∼0.056), while common SNPs are defined as tripletons or more than that including fixed derived sites (equivalent of a frequency of 1/12 or above).

Figure 5

GI are predicted to have a higher genetic load than the CHB. The ratio of the true genetic load (black errors bars indicate SE) (see Materials and Methods) in the GI vs. CHB (blue bars) is >1, suggesting that the GI have a higher genetic load than CHB. However, the ratio of the number of derived alleles in the GI vs. CHB rarely shows any deviation from 1 (green bars), likely due to the inclusion of neutral sites (see text).

Figure 6

Probability of alleles being common among GI compared to among Finns. For alleles in different frequency categories in Europeans (x-axis) the points show how much more likely the alleles are to be common in GI than in Finns (y-axis). Here, common is defined in four different ways, each represented by a specific color point: e.g., the dark blue circles represent results for analyses made with common defined as >2 out of 36 alleles, which corresponds to a frequency of ∼0.056 (for the remaining definitions, see the figure legend).