Evidence for widespread subfunctionalization of splice forms in vertebrate genomes

Abstract

Gene duplication and alternative splicing are important sources of proteomic diversity. Despite research indicating that gene duplication and alternative splicing are negatively correlated, the evolutionary relationship between the two remains unclear. One manner in which alternative splicing and gene duplication may be related is through the process of subfunctionalization, in which an alternatively spliced gene upon duplication divides distinct splice isoforms among the newly generated daughter genes, in this way reducing the number of alternatively spliced transcripts duplicate genes produce. Previously, it has been shown that splice form subfunctionalization will result in duplicate pairs with divergent exon structure when distinct isoforms become fixed in each paralog. However, the effects of exon structure divergence between paralogs have never before been studied on a genome-wide scale. Here, using genomic data from human, mouse, and zebrafish, we demonstrate that gene duplication followed by exon structure divergence between paralogs results in a significant reduction in levels of alternative splicing. In addition, by comparing the exon structure of zebrafish duplicates to the co-orthologous human gene, we have demonstrated that a considerable fraction of exon divergent duplicates maintain the structural signature of splice form subfunctionalization. Furthermore, we find that paralogs with divergent exon structure demonstrate reduced breadth of expression in a variety of tissues when compared to paralogs with identical exon structures and singletons. Taken together, our results are consistent with subfunctionalization partitioning alternatively spliced isoforms among duplicate genes and as such highlight the relationship between gene duplication and alternative splicing.

Figure 1.

Gene duplication and exon structure divergence. (A) Stacked bar chart indicating the proportions of exon divergent paralogs, nondivergent paralogs, and singletons in each species. A total of 14,671, 14,445, and 10,115 genes comprise the data sets for human, mouse, and zebrafish, respectively. (B) A model of exon structure divergence. An alternatively spliced gene prior to duplication codes for both a long and a short transcript. After duplication, the long transcript becomes fixed in one of the duplicates, while the short transcript becomes fixed in the other copy. In this way, each duplicate codes for only half of the ancestral alternative splicing repertoire. The dashed lines represent alternative splicing; solid lines, constitutive splicing events that have become fixed after duplication.

Figure 2.

The relationship between exon structure divergence among paralogs and levels of alternative splicing. The exon divergent paralogs demonstrate significantly lower levels of alternative splicing compared to nondivergent paralogs and singletons as measured by mean number of transcripts (A) and the proportion of genes that are alternatively spliced (AS) (B). Student's t-test was used to calculate the significance of the differences in mean number of transcripts. χ² was used to calculate the significance of the differences in proportion of genes that are alternatively spliced. The asterisks indicate the significance of the difference as compared to exon divergent paralogs: (**) P < 0.01. Error bars, SEM.

Figure 3.

Family size and duplication age do not explain the reduction in alternative splicing. (A) Histograms comparing gene family size between exon divergent and nondivergent paralogs. In each species the nondivergent paralogs have larger mean family sizes. Student's t-test was used to calculate the significance of the differences in mean number of paralogs. The asterisks indicate the significance of the difference as compared to exon divergent paralogs: (**) P < 0.01. Error bars, SEM. (B) Histograms showing the proportion of all exon divergent paralogs (black) created at each evolutionary epoch in each species. These values are compared against the proportion of nondivergent paralogs (white) produced at each epoch within each species. (Note, for each species the black bars sum to one, as do the white bars.) Duplicate age increases along the x-axis (not to scale). These indicate that the exon divergent paralogs are not consistently enriched in young duplicates and that time since duplication is not a confounding effect.

Figure 4.

The potential for a preduplication bias. Histograms comparing the levels of alternative splicing in single-copy orthologs that serve as a proxy for the ancestral state. (Left) Human orthologs of zebrafish exon divergent paralogs (black) display higher levels of alternative splicing compared to the human orthologs of zebrafish nondivergent paralogs (white) and zebrafish singletons (gray). (Middle) Human orthologs of mouse exon divergent paralogs (black), nondivergent paralogs (white), and singletons (gray). (Right) Mouse orthologs of human exon divergent paralogs (black), nondivergent paralogs (white), and singletons (gray). The asterisks indicate the significance of the difference as compared to the singleton orthologs of the exon divergent paralogs: (**) P < 0.01 (*) P < 0.05. Error bars, SEM.

Figure 5.

Exon divergent paralogs display restricted expression profiles. Exon divergent paralogs are expressed in fewer tissue types than nondivergent paralogs and singletons in each species. The P-values indicate the significance of the difference as compared to exon divergent paralogs. Student's t-test was used to calculate levels of significance. The asterisks indicate the significance of the difference as compared to exon divergent paralogs: (**) P < 0.01. Error bars, SEM.

Figure 6.

Structural evidence of splice form subfunctionalization. The zebrafish paralogs cnnm2a and cnnm2b correspond to distinct isoforms of the alternatively spliced orthologous human gene CNNM2. The dashed box highlights the alternatively spliced cassette exon from the human gene that has been subfunctionalized in the zebrafish paralogs. The dashed lines connect homologous exons. Exons lengths are indicated either below or above exons. Introns are not to scale.