Consistent patterns of rate asymmetry and gene loss indicate widespread neofunctionalization of yeast genes after whole-genome duplication

Abstract

We investigated patterns of rate asymmetry in sequence evolution among the gene pairs (ohnologs) formed by whole-genome duplication (WGD) in yeast species. By comparing three species (Saccharomyces cerevisiae, Candida glabrata, and S. castellii) that underwent WGD to a nonduplicated outgroup (Kluyveromyces lactis), and by using a synteny framework to establish orthology and paralogy relationships at each duplicated locus, we show that 56% of ohnolog pairs show significantly asymmetric protein sequence evolution. For ohnolog pairs that remain duplicated in two species there is a strong tendency for the faster-evolving copy in one species to be orthologous to the faster copy in the other species, which indicates that the evolutionary rate differences were established before speciation and hence soon after the WGD. We also present evidence that in cases where one ohnolog has been lost from the genome of a post-WGD species, the lost copy was likely to have been the faster-evolving member of the pair prior to its loss. These results suggest that a significant fraction of the retained ohnologs in yeast species underwent neofunctionalization soon after duplication.

Figure 1.—

Schematic relationships among genomes and locus types. (A) Phylogenetic relationship between the two copies of a WGD duplicate (ohnolog) in the three post-WGD genomes and their ortholog in the outgroup non-WGD genome. The paralogous copies are arbitrarily labeled copy A (red) and copy B (green). The WGD event is marked by a solid circle. (B) A locus included in the Scer–Scas pairwise comparison in the 2:2 category. AS and BS are shared evolutionary branches, and A1, A2, B1, and B2 are species-specific branches. (C) A locus included in the 2:1 category of the Scer–Cgla comparison.

Figure 2.—

The REG1/REG2 locus. The YGOB (Byrne and Wolfe 2005) screenshot in the background shows the syntenic context at the locus, identifying orthologs and paralogs. The topology of the maximum-likelihood tree, from the Scer–Scas comparison in the 2:2 category, agrees perfectly with the topology derived from synteny, so this locus passes our phylogenetic filter and is retained in set 2. Furthermore, the orthologs REG2 and Scas_718.54 are the faster-evolving ohnologs in each species, making this a locus with consistent asymmetric evolution. The “X” marks the loss of the “fast” copy of the gene from C. glabrata, when it is compared to either S. cerevisiae (Scer–Cgla, 2:1) or S. castellii (Cgla–Scas, 1:2). Branch lengths show the pronounced asymmetric protein sequence evolution on both the shared and the terminal branches for this locus.

Figure 3.—

The extent and significance of rate asymmetry. (A) Plot of the terminal branch asymmetry measure (R′) for the 166 S. cerevisiae ohnologs in the Scer–Scas data set (set 2). Loci are ordered by increasing R′ on the x-axis, with R′-values shown on the y-axis. Dashed lines show the number and percentage of loci with R′ > 1.1, 1.25, 1.5, and 2.0. (B) Correlation of R′-values for terminal branch asymmetry of S. cerevisiae ohnologs and the significance of asymmetry between those ohnologs (expressed as the negative logarithm, to base 10, of the P-value). All points to the right of the vertical dashed line are significantly asymmetric (P < 0.05) after correction for multiple testing. The correlation is highly significant for both Pearson's r and Spearman's s.

Figure 4.—

Consistent rate asymmetry across species. A plot of branch lengths (K_A units) for the 62 significantly asymmetric 2:2 loci from the Scer–Scas species comparison in set 3 (the same data set as in Table 1) is shown. Each circle shows the branch lengths for faster (x-axis) and slower (y-axis) gene copies in S. cerevisiae, corresponding to branches A1 and B1, respectively, in Figure 1B. Triangles connected by dashed lines to each circle show the corresponding data for S. castellii, but the value of branch A2 (i.e., the S. castellii branch orthologous to the faster copy in S. cerevisiae) is always plotted on the x-axis, and the value of branch B2 (i.e., the S. castellii branch orthologous to the slower copy in S. cerevisiae) is always plotted on the y-axis, regardless of which of A2 and B2 is actually the longer branch in S. castellii. Hence any triangle lying above the diagonal is a locus where a different ohnolog is faster evolving in each species. A small number of points with branch lengths >1 are not shown.

Figure 5.—

Association across species of fast-evolving and lost ohnologs. (A) Distributions of rates of amino acid divergence (K_A) on branches leading to orthologs of a lost duplicate (dark shading) are significantly greater than that on branches leading to the paralog of the lost duplicate/ortholog of the retained duplicate (light shading), in the three pairwise comparisons Scer–Cgla, Scer–Scas, and Cgla–Scas, over loci in 2:1 and 1:2 categories from set 3. P-values are from paired Wilcoxon's signed rank tests on the underlying data, with data binned for display only. We tested the hypothesis that the “lost duplicate's ortholog” K_A distribution is greater than the “lost duplicate's paralog” K_A distribution. (B) The 2:1 and 1:2 locus classes and the losses at them are illustrated under the distributions for each species comparison in A. The estimated K_A on branches with dark shading (leading to ortholog of loss) or light shading (leading to paralog of loss) becomes part of the corresponding darkly shaded or lightly shaded distribution. The numbers and percentages to the right of each box show the trend for the fast-evolving copy in the two-copy species to be lost from the single-copy species. “Fast ohnolog is lost” means the faster copy of the gene has been lost in the other species. “Slow ohnolog is lost” means the opposite. P-values are from Fisher's exact two-tail tests against neutral expectation.