Evidence for balancing selection from nucleotide sequence analyses of human G6PD

Abstract

Glucose-6-phosphate dehydrogenase (G6PD) mutations that result in reduced enzyme activity have been implicated in malarial resistance and constitute one of the best examples of selection in the human genome. In the present study, we characterize the nucleotide diversity across a 5.2-kb region of G6PD in a sample of 160 Africans and 56 non-Africans, to determine how selection has shaped patterns of DNA variation at this gene. Our global sample of enzymatically normal B alleles and A, A-, and Med alleles with reduced enzyme activities reveals many previously uncharacterized silent-site polymorphisms. In comparison with the absence of amino acid divergence between human and chimpanzee G6PD sequences, we find that the number of G6PD amino acid polymorphisms in human populations is significantly high. Unlike many other G6PD-activity alleles with reduced activity, we find that the age of the A variant, which is common in Africa, may not be consistent with the recent emergence of severe malaria and therefore may have originally had a historically different adaptive function. Overall, our observations strongly support previous genotype-phenotype association studies that proposed that balancing selection maintains G6PD deficiencies within human populations. The present study demonstrates that nucleotide sequence analyses can reveal signatures of both historical and recent selection in the genome and may elucidate the impact that infectious disease has had during human evolution.

Figure 1

Diagram of the G6PD locus, spanning ∼18 kb. Exons are shown as blackened boxes, and introns and noncoding regions are shown as unblackened boxes. Intron 2 is nearly 10 kb in length. The 5.2-kb region that was sequenced for this study is shown, as well as the location of the common A−, A, and Med replacement SNPs.

Figure 2

Summary data for 5.2-kb region of G6PD for 216 individuals. All polymorphisms are shown as derived changes as compared with the “chimp” sequence. Nucleotide positions start with the first base pair of exon 3 (the first SNP is in the second intron 135 bp before exon 3). Coding-region SNPs are labeled as “r” and “s,” for replacement and silent sites, respectively; deletions are labeled as “d1”–“d4”; and the row labeled as “tree” refers to the African SNPs that were used in the GENETREE analysis (fig. 5). Although haplotype H was originally typed as an A− allele, it does not have the site 177 SNP that is common to all other A− alleles (see the “Results” section).

Figure 3

Measurements and plots of LD and association. a, Relationship between the LD measure as R² and the distance (in bp) between 15 SNPs found at frequencies >3% in the African sample from 160 individuals. Blackened circles indicate the 25 associations, of the 105 possible, that are significant at the 0.01% level by a χ² test and Bonferroni correction. b, Plot of the association among the 15 African SNPs for 160 individuals. Blackened boxes indicate significant LD at the 0.01% level, and numbers denote the nucleotide positions of SNPs in figure 2. c, Plot of LD for the 112 African B alleles. Blackened boxes indicate significance at the 0.01% level for 9 of the possible 36 associations, for the nine SNPs found at frequencies >3% in the African sample.

Figure 4

Neighbor-joining tree, based on only silent substitutions among 216 individuals and rooted with chimpanzee G6PD sequence. Each haplotype is represented only once for each time it is found in a different population. Scale indicates one substitution. Non-African haplotypes are shown in boldface italics. The A and A− haplotypes differ by only a single replacement SNP; therefore, except for the single Tunisian A− allele that possesses a silent SNP, there is no distinct A− haplotype clade. CA = Cameroon; SL = Sierra Leone; NA = Nigeria; PY = Bakola Pygmies; HZ = Hadza; MS = Maasai; SW = Sandawe; BS = South African Bantu-speakers; TU = Tunisia; CY = Cyprus; LB = Lebanon; AM = South American Andean; PNG = Papua New Guinea.

Figure 5

GENETREE results for 158 African individuals (haplotype N is omitted; see the “Subjects and Methods” section). The number and letter codes for the SNPs and haplotypes, respectively, are found in figure 2. All replacement SNPs are shown in boxes. SNPs 2 and 7 designate the common A− and A variants, respectively. Although haplotype H was originally labeled as “an A− allele” on the basis of functional analyses, it does not have the common SNP 2 (see the “Results” section). The order of SNPs in time is arbitrary when they are found on the same branch (e.g., SNPs 1, 7, and 16), and ages for all variants are mean estimates.