Global analysis of exon creation versus loss and the role of alternative splicing in 17 vertebrate genomes

Abstract

Association of alternative splicing (AS) with accelerated rates of exon evolution in some organisms has recently aroused widespread interest in its role in evolution of eukaryotic gene structure. Previous studies were limited to analysis of exon creation or lost events in mouse and/or human only. Our multigenome approach provides a way for (1) distinguishing creation and loss events on the large scale; (2) uncovering details of the evolutionary mechanisms involved; (3) estimating the corresponding rates over a wide range of evolutionary times and organisms; and (4) assessing the impact of AS on those evolutionary rates. We use previously unpublished independent analyses of alternative splicing in five species (human, mouse, dog, cow, and zebrafish) from the ASAP database combined with genomewide multiple alignment of 17 genomes to analyze exon creation and loss of both constitutively and alternatively spliced exons in mammals, fish, and birds. Our analysis provides a comprehensive database of exon creation and loss events over 360 million years of vertebrate evolution, including tens of thousands of alternative and constitutive exons. We find that exon inclusion level is inversely related to the rate of exon creation. In addition, we provide a detailed in-depth analysis of mechanisms of exon creation and loss, which suggests that a large fraction of nonrepetitive created exons are results of ab initio creation from purely intronic sequences. Our data indicate an important role for alternative splicing in creation of new exons and provide a useful novel database resource for future genome evolution research.

FIGURE 1.

Outgroup method for distinguishing between creation and loss. (A) Exon is present in the source and outgroup organism, which implies that the ancestral state is also present. This means that the exon was lost from the target organism after the split with source. (B) Exon is not present in either the target or outgroup organism; therefore, it must have been created after the split of source and target.

FIGURE 2.

Typical alignments of exons to the flanked region of other organisms. In A, a skipped exon of the cdc37l gene provides an example of a well-conserved class of exons showing very good sequence conservation (mismatches underlined) with no insertions and the AG/GT splice signals (bold) are conserved, while in B, a skipped exon of the dnajc7 gene shows an alignment typical of the second class, with conservation similar to the surrounding intron including a large insertion (italic) and splice-sites mutated.

FIGURE 3.

Time course of exon creation and loss. The fraction of minor (light diamonds), medium (dark squares), major (light triangles), and constitutive (dark circles) created in cow (A), dog (B), human (C), and mouse (D) or lost (E–H, respectively) within the evolutionary distance. The amount of loss and creation in constitutive and major-form exons is the same, indicated by strong overlap of points on the graph. In the alternative set, the fraction of created exons is anticorrelated with the inclusion level. The amount of loss is similar for all times and inclusion categories. Similar trends are observed in data from different source organisms.

FIGURE 4.

Fraction of exons created and lost by inclusion category. The fraction of exons created (dark diamonds) or lost (light squares) after the split between (A) human–chimp and (B) mouse–rat in each of the four inclusion categories. The amount of loss is almost constant within time (5 my [A], 40 my [B]) and inclusion category. The amount of creation is anticorrelated with the inclusion level.

FIGURE 5.

Dataflow of the analyses. Analyses of two data sets are schematically depicted: (A) in-depth study, (B) 17-way exon conservation analysis. Roughly, the analysis consisted of the following steps. Source data collection included downloading and arranging data. In the preprocessing step, we identified the regions that should contain the exons in question by mapping the adjacent (flanking) exons onto the target genomes. Then we aligned (in A) introns using full dynamic programming and (in B) just the exons and splice sites using UCSC conservation track. Next, the alignments were scored using various metrics of conservation. The resulting conservation data sets make up the VEEDB database and were used for further outgroup analyses.