Detecting past and ongoing natural selection among ethnically Tibetan women at high altitude in Nepal

Abstract

Adaptive evolution in humans has rarely been characterized for its whole set of components, i.e. selective pressure, adaptive phenotype, beneficial alleles and realized fitness differential. We combined approaches for detecting polygenic adaptations and for mapping the genetic bases of physiological and fertility phenotypes in approximately 1000 indigenous ethnically Tibetan women from Nepal, adapted to high altitude. The results of genome-wide association analyses and tests for polygenic adaptations showed evidence of positive selection for alleles associated with more pregnancies and live births and evidence of negative selection for those associated with higher offspring mortality. Lower hemoglobin level did not show clear evidence for polygenic adaptation, despite its strong association with an EPAS1 haplotype carrying selective sweep signals.

Fig 1. A schematic summary of the genotype and phenotype data of ethnic Tibetans in this study.

(A) We array genotyped all individuals in several Illumina platforms and generated whole genome sequences for a representative subset without recent admixture. Then, all individuals went through genotype imputation using our high altitude sequence data (“high altitude panel”) and world-wide data (“1KG phase 3 panel”) as reference haplotype panels. (B) Three physiological phenotypes were directly measured in the field, and two additional ones (oxyHb and deoxyHb) were constructed from them. (C) Fertility phenotypes capture both fertility and viability selection components. For details, please see Materials & Methods.

Fig 2

Locuszoom plots of the genome-wide significant associations found in Tibetans: (A) oxyHb and rs372272284 in the EPAS1 gene, (B) the numbers of pregnancies or (C) live births and rs6711319 in the CCDC141 gene, (D) the number of stillbirths and rs1957819 near the PAPOLA gene, (E, F) the proportion of offsprings died before age 15 years among the born alive and rs9392394/rs1459385. (A-C, E) are tests with all samples and (D, F) are those with the continuously married subset.

Fig 3

Tests of polygenic adaptation of Hb-associated SNPs: (A-C) 35 SNPs from our Tibetan GWAS (p ≤ 10⁻⁴) after excluding the EPAS1 SNP rs372272284, and (D-F) 91 genome-wide significant SNPs from a large GWAS of mostly European cohorts. The mean frequency differences of trait-increasing alleles between Tibetans and CHB (A, D) and between Sherpa and CHB (B, E) were presented (solid red line) together with the empirical null distribution of 10,000 sets of matched random SNPs. (C, F) The genetic values of populations (filled dots) and of regions (horizontal lines) were plotted. The size of dots and the width of lines are proportional to the significance of the corresponding outlier test.

Fig 4

Signatures of polygenic adaptations in Tibetans shown for (A) deoxyHb, (B) the number of children surviving at 1 year, (C) the number of live births, (D) the number of children born alive but died before 15 years, (E) the number of miscarriages and (F) the number of twin births. The mean frequency difference of trait-increasing alleles was presented (solid red line) together with the empirical null distribution of 10,000 sets of matched random SNPs. (C-F) uses GWAS SNPs from the “CM” subset.