The extent of linkage disequilibrium caused by selection on G6PD in humans

Abstract

The gene coding for glucose-6-phosphate dehydrogenase (G6PD) is subject to positive selection by malaria in some human populations. The G6PD A- allele, which is common in sub-Saharan Africa, is associated with deficient enzyme activity and protection from severe malaria. To delimit the impact of selection on patterns of linkage disequilibrium (LD) and nucleotide diversity, we resequenced 5.1 kb at G6PD and approximately 2-3 kb at each of eight loci in a 2.5-Mb region roughly centered on G6PD in a diverse sub-Saharan African panel of 51 unrelated men (including 20 G6PD A-, 11 G6PD A+, and 20 G6PD B chromosomes). The signature of selection is evident in the absence of genetic variation at G6PD and at three neighboring loci within 0.9 Mb from G6PD among all individuals bearing G6PD A- alleles. A genomic region of approximately 1.6 Mb around G6PD was characterized by long-range LD associated with the A- alleles. These patterns of nucleotide variability and LD suggest that G6PD A- is younger than previous age estimates and has increased in frequency in sub-Saharan Africa due to strong selection (0.1 < s < 0.2). These results also show that selection can lead to nonrandom associations among SNPs over great physical and genetic distances, even in African populations.

Figure 1.

Ideogram of the human X chromosome and the genomic regions surveyed in this study. Approximate distances between each of the surveyed windows including G6PD are marked on the scale. Transcription orientations of the genic regions are marked with solid arrows. The exon/intron structure of G6PD is designated (displayed in inverted orientation relative to chromosomal orientation) along with the defining substitutions of the three allelic classes: A−, A+, and B (in box). Positions of amplification primers used to survey the 5.1-kb window of G6PD are marked with shaded arrows.

Figure 2.

Table of polymorphism for G6PD and surrounding loci. Fifty-one unrelated human males of sub-Saharan African descent were surveyed for nucleotide variability at G6PD and nine surrounding loci. Individual samples were selected on the basis of a priori allele type determination based on coding sites 202 and 376 of G6PD to define three allele classes: A−, A+, and B. Each segregating site (in columns) represents a biallelic marker (i.e., SNP or INDEL). Segregating sites are listed in numerical order. For exact alignment positions (in base pairs) of segregating sites and the identity of the polymorphic nucleotides (i.e., A, G, C, or T), sequences are available from GenBank or upon request from the author. At each segregating site one allelic state is marked with a blue box and the alternate allelic state is marked with a yellow box. Missing data for individual Ivc18 at locus F8 and for individual JR323 at locus G0.9MT are indicated with gray boxes. Boundaries between loci are marked by vertical white bars. S and N denote synonymous and nonsynonymous changes, respectively. G6PD coding site 202 (segregating site 90) is marked with an asterisk. Polymorphisms at sites 72, 73, 74, and 75 in G6PD are in the 3′-untranslated region of exon 13. All other polymorphisms are in introns or intergenic regions. INDELs are marked with an open triangle including the size of the INDEL in base pairs. At G1.5MC and BGN unsurveyed regions in the otherwise contiguous windows are marked by an arrow with numbers in box indicating the number of contiguous unsurveyed base pairs.

Figure 3.

Nucleotide variability for G6PD and nine surrounding loci for subset groups of the data set (G6PD A− alleles, G6PD A+ alleles, G6PD B alleles, and CRS): (a) nucleotide diversity (θπ); (b) haplotype diversity.

Figure 4.

Patterns of LD between segregating sites at G6PD and flanking loci. The table of polymorphism includes only segregating sites at which the less common allele (minor allele) is found in five or more individuals. At each segregating site, one allele is marked with a blue box and the alternate allele is marked with a yellow box. Missing data are marked with a gray box. Segregating sites are numbered and labeled according to Figure 2. Boundaries between loci surveyed are marked by vertical white bars. Below the table of polymorphism is a matrix of estimates of |D′| for all pairwise comparisons of sites. Values of |D′| are shown between 0.5 and 1.0 in accordance with the shading scale. Intergenic pairwise associations in significant LD (P < 0.05) by Fisher's exact test are marked in the matrix by circles.

Figure 5.

Results of the evolutionary analysis of the G6PD A− allele. Results were obtained by combining the importance sampling method of Slatkin (2001) for averaging over replicate sample paths with the method of Garner and Slatkin (2002) for computing the probability of a configuration of haplotype frequencies at two linked loci. A population frequency for G6PD A− of 0.1 and a constant population size of 10,000 individuals were assumed. The numbers of G6PD A− chromosomes with the two-locus haplotypes at L1CAM and G0.9MT were 14, 6, 0, and 0. For each point, 90,000 replicate sample paths were generated and 20 replicates of the Garner-Slatkin program were run for each sample path. The estimated recombination rates from G6PD were c = 0.008375 and 0.002775 M for L1CAM and G0.9MT, respectively. Other results shown were obtained by doubling and halving those values. (A) Log-likelihood of s, the hypothesized selective advantage of heterozygous carriers of the G6PD A− allele. Additive selection was assumed. (B) The posterior distribution of allele age (t₁) for two selection coefficients consistent with the observations. (C) The posterior distribution of allele age (t₁) for s = 0.2 for the three sets of recombination rates used.