Function relaxation followed by diversifying selection after whole-genome duplication in flowering plants

Abstract

Episodes of whole-genome duplication (WGD) followed by gene loss dominate the evolutionary history of flowering plants. Despite the importance of understanding gene evolution following WGD, little is known about the evolutionary dynamics of this process. In this study, we analyzed duplicated genes from three WGD events in the Arabidopsis (Arabidopsis thaliana) lineage using multiple data types. Most duplicated genes that have survived from the most recent WGD (α) are under purifying selection in modern Arabidopsis populations. Using the number of identified protein-protein interactions as a proxy for functional divergence, approximately 92.7% of α-duplicated genes were diverged in function from one another in modern Arabidopsis populations, indicating that their preservation is no longer explicable by dosage balance. Dosage-balanced retention declines with antiquity of duplication: 24.1% of α-duplicated gene pairs in Arabidopsis remain in dosage balance with interacting partners, versus 12.9% and 9.4% for the earlier β-duplication and γ-triplication. GO-slim (a cut-down version of gene ontologies) terms reinforce evidence from protein-protein interactions, showing that the putatively diverged gene pairs are adapted to different cellular components. We identified a group of α-duplicated genes that show higher than average single-nucleotide polymorphism density, indicating that a period of positive selection, potentially driving functional divergence, may have preceded the current phase of purifying selection. We propose three possible paths for the evolution of duplicated genes following WGD.

Figure 1.

Function divergence between duplicated Arabidopsis genes on the time scale measured by Ks. Blue dots represent duplicated gene pairs from the α-WGD event (most recent), green dots from the β-WGD event, and red dots from the γ-WGD event (most ancient). ppi, Protein-protein interaction.

Figure 2.

Categories of WGD-duplicated genes based on the divergence of protein interaction partners. Blue circles represent WGD-duplicated paralogous gene pairs. Orange squares are the original interacting partners shared by both copies. Red triangles are newly gained interacting partners indicating function divergence. Arrows suggest the process of function divergence of WGD-duplicated genes. The top and bottom paths refer to divergence by loss and gain of interaction partners, respectively.

Figure 3.

Function divergence indicated by GO-slim terms. Venn diagrams show the numbers of diverged gene pairs in three Gene Ontology (GO) function categories. The integer indicates the number of gene pairs that shows diverged GO annotations in the function category. The number in parentheses shows the percentage. The numbers outside all three circles represents the number (percentage) of nondiverged gene pairs in all three GO function categories. A, Conserved. B, Partially diverged. C, Fully diverged. D, Proportion of diverged gene pairs in each GO category.

Figure 4.

Comparison of Ka/Ks ratios among function-divergence groups. The mean Ka/Ks ratio in each group is shown with a black bar, with error bars indicating se. Two-sample independent t tests show that there is significant difference between partially and fully diverged groups (P = 0.0371). [See online article for color version of this figure.]

Figure 5.

Comparison of coding SNP density in modern Arabidopsis populations among different function-divergence groups. Black bars indicate means of coding SNP density, with error bars indicating se. Mann-Whitney U tests show that the SNP density of gene pairs in the partially diverged group is significantly lower than that of fully diverged and conserved groups (P = 2.619e-11 and P = 0.000349). There is no significant difference between the fully diverged and conserved groups. The red line indicates the genome-wide average of coding SNP density. [See online article for color version of this figure.]

Figure 6.

Comparison of the severity of nonsynonymous nucleotide polymorphism in modern Arabidopsis populations among function-divergence groups. Average BLOSUM80 score over all nonsynonymous SNPs in each duplicated gene is calculated as a measure of the severity of amino acid changes, with smaller BLOSUM80 scores indicating greater severity. The red lines indicate the genome-wide average of BLOSUM80 scores. A, Comparison of the mean BLOSUM80 scores among function-divergence groups, with error bars indicating se. Mann-Whitney U tests show that there are significantly larger numbers of severe nonsynonymous substitutions in the fully and partially diversified groups than in the conserved group (P = 0.00183 and P = 0.00233). B, Comparison of coding SNP severity and degrees of expression divergence for each duplicated gene pair. Linear regression of the mean BLOSUM80 score is shown as a broken line with the function of the regression line. [See online article for color version of this figure.]