A genome-wide survey of the prevalence and evolutionary forces acting on human nonsense SNPs

Abstract

Nonsense SNPs introduce premature termination codons into genes and can result in the absence of a gene product or in a truncated and potentially harmful protein, so they are often considered disadvantageous and are associated with disease susceptibility. As such, we might expect the disrupted allele to be rare and, in healthy people, observed only in a heterozygous state. However, some, like those in the CASP12 and ACTN3 genes, are known to be present at high frequencies and to occur often in a homozygous state and seem to have been advantageous in recent human evolution. To evaluate the selective forces acting on nonsense SNPs as a class, we have carried out a large-scale experimental survey of nonsense SNPs in the human genome by genotyping 805 of them (plus control synonymous SNPs) in 1,151 individuals from 56 worldwide populations. We identified 169 genes containing nonsense SNPs that were variable in our samples, of which 99 were found with both copies inactivated in at least one individual. We found that the sampled humans differ on average by 24 genes (out of about 20,000) because of these nonsense SNPs alone. As might be expected, nonsense SNPs as a class were found to be slightly disadvantageous over evolutionary timescales, but a few nevertheless showed signs of being possibly advantageous, as indicated by unusually high levels of population differentiation, long haplotypes, and/or high frequencies of derived alleles. This study underlines the extent of variation in gene content within humans and emphasizes the importance of understanding this type of variation.

Figure 1

Genome-wide Distribution of Nonsense SNPs on Chromosomes 1 to X in the Human Genome The nonsense SNPs that were variable in our samples are displayed in red, and all nonsense SNPs reported in the human genome (dbSNP127) are shown in blue.

Figure 2

Even Distribution of Truncations Truncations were calculated as the percentage of the ancestral ORF length. The 169 nonsense SNPs were sorted along the x axis according to the amount of peptide truncation, starting at 1 for the lowest truncation and ending at 169 for the highest truncation. The identifying number of the SNP displayed in the figure can be found in Table S4. Orange labels transcripts where NMD is predicted to be triggered with the complete loss of the gene product, whereas green refers to transcripts where NMD is evaded because the nonsense SNP is located either in the last exon or less than 50 nucleotides upstream of the last exon-exon boundary.

Figure 3

Frequencies of Stop Homozygotes, Normal Homozygotes, and Heterozygotes for Each Variable Nonsense SNP The genotype frequencies of normal homozygotes (green), heterozygotes (purple), and stop homozygotes (orange) were plotted on a logarithmic scale. The nonsense SNPs were sorted along the x axis according to the frequency of stop homozygotes. The identifying number can be found in Table S4.

Figure 4

MAGEE2 (A) Geographical distribution of stop (orange) and normal (blue) alleles in MAGEE2. HapMap populations are displayed separately because they do not all have precise geographic locations. Pie areas are proportional to sample sizes. (B) Median-joining network of inferred MAGEE2 haplotypes. Circle areas are proportional to the haplotype frequency and are color coded according to population: CEU in yellow, CHB in green, LWK in pink, and YRI in red. Lines represent mutational steps between them (one or two steps, according to length). An arrow shows the location of a nonsense SNP (rs1343879).