Patterns of molecular evolution in Caenorhabditis preclude ancient origins of selfing

Abstract

The evolution of self-fertilization can mediate pronounced changes in genomes as a by-product of a drastic reduction in effective population size and the concomitant accumulation of slightly deleterious mutations by genetic drift. In the nematode genus Caenorhabditis, a highly selfing lifestyle has evolved twice independently, thus permitting an opportunity to test for the effects of mode of reproduction on patterns of molecular evolution on a genomic scale. Here we contrast rates of nucleotide substitution and codon usage bias among thousands of orthologous groups of genes in six species of Caenorhabditis, including the classic model organism Caenorhabditis elegans. Despite evidence that weak selection on synonymous codon usage is pervasive in the history of all species in this genus, we find little difference among species in the patterns of codon usage bias and in replacement-site substitution. Applying a model of relaxed selection on codon usage to the C. elegans and C. briggsae lineages suggests that self-fertilization is unlikely to have evolved more than approximately 4 million years ago, which is less than a quarter of the time since they shared a common ancestor with outcrossing species. We conclude that the profound changes in mating behavior, physiology, and developmental mechanisms that accompanied the transition from an obligately outcrossing to a primarily selfing mode of reproduction evolved in the not-too-distant past.

Figure 1.—

Phylogenetic topology of Caenorhabditis that is assumed in calculations of lineage-specific divergence from Kiontke et al. (2004), Kiontke and Sudhaus (2006), and Kiontke et al. (2007). Dashed lines for C. elegans and C. briggsae indicate that self-fertilization (shaded tips) evolved at some point along these branches. Branch lengths are not to scale.

Figure 2.—

Fraction of variance in the frequency of optimal codons (F_op) explained by base composition, gene length, gene expression, and their interactions. Codon usage bias is more extreme in highly expressed genes (F_op × log₁₀[EST hit count] all P < 0.0001) and in shorter genes (F_op × log₁₀[length] P < 0.0001 for C. elegans, C. brenneri, C. remanei, and C. briggsae; P > 0.05 in other species).

Figure 3.—

Difference in RSCU between highly and lowly expressed genes (ΔRSCU) and inferred optimal codon identity in Caenorhabditis. Codons are sorted by mean ΔRSCU across species. Optimal codons are indicated along the right, with adjacent asterisks indicating statistical significance for different species ordered as for ΔRSCU columns (dots indicate lack of significance for the respective species). Yellow corresponds to ΔRSCU = 0, with shading toward red indicating ΔRSCU > 0 (maximum 1.04) and shading toward blue indicating ΔRSCU < 0 (minimum −0.67).

Figure 4.—

Slopes from orthogonal regressions of the frequency of optimal codons (F_op) between pairs of species. Values below the diagonal correspond to species listed along the side as dependent variable; values above the diagonal correspond to species listed along the top as dependent variable. In all comparisons exhibiting a slope >1, the intercept is negative; all comparisons with slope <1 have a positive intercept. The combination of slopes <1 and intercepts >0 for C. elegans as independent variable (second column from left, white text) is indicative of reduced codon bias in this species. Asterisks indicate significant differences from a slope of 1 (*P = 0.017, **P < 0.01, ***P = 0.00023). Darker shades of gray represent slopes with greater deviation from 1.

Figure 5.—

The predicted time since the onset of relaxed selection on codon usage to yield 65.8% optimal codons, as observed for 63 C. elegans genes that are conserved across the genus, as a function of the ancestral level of codon bias. Provided that the evolution of selfing initiated the relaxed selection, this time is equivalent to the duration of selfing in C. elegans. Timescale in years assumes a 60-day generation time in nature; dashed line demarcates the value for ancestral codon bias assumed in point estimation of t.