Hsuan-Yu Lin

From studies investigating the differences in evolutionary rates between genes, gene compactness and gene expression level have been identified as important determinants of gene-level protein evolutionary rate, as represented by nonsynonymous to synonymous substitution rate (d_N/d_S) ratio. However, the causes of exon-level variances in d_N/d_S are less understood. Here, we use principal component regression to examine to what extent 13 exon features explain the variance in d_N, d_S, and the d_N/d_S ratio of human–rhesus macaque or human–mouse orthologous exons. The exon features were grouped into six functional categories: expression features, mRNA splicing features, structural–functional features, compactness features, exon duplicability, and other features, including G + C content and exon length. Although expression features are important for determining d_N and d_N/d_S between exons of different genes, structural–functional features and splicing features explained more of the variance for exons of the same genes. Furthermore, we show that compactness features can explain only a relatively small percentage of variance in exon-level d_N or d_N/d_S in either between-gene or within-gene comparison. By contrast, d_S yielded inconsistent results in the human–mouse comparison and the human–rhesus macaque comparison. This inconsistency may suggest rapid evolutionary changes of the mutation landscape in mammals. Our results suggest that between-gene and within-gene variation in d_N/d_S (and d_N) are driven by different evolutionary forces and that the role of mRNA splicing in causing the variation in evolutionary rates of coding sequences may be underappreciated.

Alternative splicing (AS) is known to significantly affect exon-level protein evolutionary rates in mammals. Particularly, alternatively spliced exons (ASEs) have a higher nonsynonymous-to-synonymous substitution rate (dN/dS) ratio than constitutively spliced exons (CSEs), possibly because the former are required only occasionally for normal biological functions. Meanwhile, intrinsically disordered regions (IDRs), the protein regions lacking fixed 3D structures, are also reported to have an increased evolutionary rate due to lack of structural constraint. Interestingly, IDRs tend to be located in alternative protein regions. Yet which of these two factors is the major determinant of the increased dN/dS in mammalian ASEs remains unclear. By comparing human–macaque and human–mouse one-to-one orthologous genes, we demonstrate that AS and protein structural disorder have independent effects on mammalian exon evolution. We performed analyses of covariance to demonstrate that the slopes of the (dN/dS-percentage of IDR) regression lines differ significantly between CSEs and ASEs. In other words, the dN/dS ratios of both ASEs and CSEs increase with the proportion of IDR (PIDR), whereas ASEs have higher dN/dS ratios than CSEs when they have similar PIDRs. Since ASEs and IDRs may less frequently overlap with protein domains (which also affect dN/dS), we also examined the correlations between dN/dS ratio and exon type/PIDR by controlling for the density of protein domain. We found that the effects of exon type and PIDR on dN/dS are both independent of domain density. Our results imply that nature can select for different biological features with regard to ASEs and IDRs, even though the two biological features tend to be localized in the same protein regions.