Synonymous SNPs Provide Evidence for Selective Constraint on Human Exonic Splicing Enhancers
The human SNP database was used to detect selection on 238 hexamers previously identified as exonic splicing enhancers (ESEs). We compared the distribution of the 238 putative ESEs in biallelic and triallelic SNPs within five different functional categories of the SNP database: synonymous, nonsynonymous, introns, UTRs, and nongenic SNPs. Since true ESEs do not function outside of exons, SNPs that disrupt ESE motifs were expected to be more common in nonexonic portions of the genome. Our results supported this expectation: ESEs were least prevalent within synonymous SNPs and most common in nongenic SNPs. There were ∼11% fewer ESEs within synonymous biallelic SNPs than expected under no selective constraint. We also compared the frequency of neutral SNPs, those where neither allele was an ESE, with deleterious SNPs, those where one or more alleles was an ESE, across the five different functional classes of SNPs. In comparison with the other functional classes of SNPs, synonymous SNPs contained an excess of neutral variants (+1.64% and +6.04% for biallelic and triallelic SNPs, respectively) and a dearth of deleterious variants (−13.11% and −52.39% for biallelic and triallelic SNPs, respectively). The observed patterns were consistent with purifying selection on the 238 hexamers to maintain their function as ESEs. However, in contrast to previous work, we did not find evidence for selection to maintain ESE function at nonsynonymous SNPs because selection at the protein level probably obscured any difference at the level of ESE function.
KeywordsExonic splicing enhancers Single nucleotide polymorphisms Synonymous substitutions
This research was supported in part by NSF Grant 0315468 awarded to D.B.C.