SNP Formation Bias in the Murine Genome Provides Evidence for Parallel Evolution

Abstract

In this study, we show novel DNA motifs that promote single nucleotide polymorphism (SNP) formation and are conserved among exons, introns, and intergenic DNA from mice (Sanger Mouse Genomes Project), human genes (1000 Genomes), and tumor-specific somatic mutations (data from TCGA). We further characterize SNPs likely to be very recent in origin (i.e., formed in otherwise congenic mice) and show enrichment for both synonymous and parallel DNA variants occurring under circumstances not attributable to purifying selection. The findings provide insight regarding SNP contextual bias and eukaryotic codon usage as strategies that favor long-term exonic stability. The study also furnishes new information concerning rates of murine genomic evolution and features of DNA mutagenesis (at the time of SNP formation) that should be viewed as "adaptive."

Keywords: SNP formation bias; mutation; parallel evolution.

Fig. 1.—

Base frequency bias of murine A⇔G SNPs. (A) Base location frequency bias (Bias %) immediately surrounding (±4 bp) 20,603 exonic A⇔G homozygous SNPs, compiled from coding regions of murine chromosomes 1–8. Bias % was calculated as described in Materials and Methods. (B) Genomic base frequency bias (±4 bp) relative to 20,000 randomly chosen (not SNP-associated) adenine (A) and guanine (G) nucleotides within murine exons (chromosomes 1–4) studied in a fashion otherwise identical to Panel (A). (C) Base location frequency bias immediately surrounding (±4 bp) 50,244 A⇔G SNPs compiled from intronic regions of murine chromosomes 1–3. (D) Genomic base frequency bias relative to 50,000 randomly chosen “A” or “G” sites within murine introns. (E) Base location frequency bias relative to 67,663 A⇔G SNPs compiled from intergenic regions on murine chromosomes 1–3. (F) Genomic base frequency bias relative to 50,000 randomly chosen “A” or “G” sites from intergenic regions of murine chromosome 2. In all cases, standard deviation (as judged by bootstrap analysis) was very low (on the order of ∼0.1–0.3%).

Fig. 2.—

SNP-permissive and shielding quartets in human ORFs. (A) Most frequent SNP-permissive (yellow) and SNP-shielding (red) A⇔G quartet contexts within the cDNA of CFTR. (B) Same analysis for cDNA of dystrophin. A statistical (chi square 2 × 2 contingency table) analysis indicated overrepresentation of shielding dinucleotide quartets in both genes (P < 0.00001), and underrepresentation of quartets permissive for SNP formation (P < 0.00001).

Fig. 2.—

SNP-permissive and shielding quartets in human ORFs. (A) Most frequent SNP-permissive (yellow) and SNP-shielding (red) A⇔G quartet contexts within the cDNA of CFTR. (B) Same analysis for cDNA of dystrophin. A statistical (chi square 2 × 2 contingency table) analysis indicated overrepresentation of shielding dinucleotide quartets in both genes (P < 0.00001), and underrepresentation of quartets permissive for SNP formation (P < 0.00001).