Molecular evolution and the decline of purifying selection with age

Abstract

Life history theory predicts that the intensity of selection declines with age, and this trend should impact how genes expressed at different ages evolve. Here we find consistent relationships between a gene's age of expression and patterns of molecular evolution in two mammals (the human Homo sapiens and the mouse Mus musculus) and two insects (the malaria mosquito Anopheles gambiae and the fruit fly Drosophila melanogaster). When expressed later in life, genes fix nonsynonymous mutations more frequently, are more polymorphic for nonsynonymous mutations, and have shorter evolutionary lifespans, relative to those expressed early. The latter pattern is explained by a simple evolutionary model. Further, early-expressed genes tend to be enriched in similar gene ontology terms across species, while late-expressed genes show no such consistency. In humans, late-expressed genes are more likely to be linked to cancer and to segregate for dominant disease-causing mutations. Last, the effective strength of selection (N_e s) decreases and the fraction of beneficial mutations increases with a gene's age of expression. These results are consistent with the diminishing efficacy of purifying selection with age, as proposed by Medawar's classic hypothesis for the evolution of senescence, and provide links between life history theory and molecular evolution.

Fig. 1. Late genes fix nonsynonymous mutations at higher rates.

Genes expressed later in life fix nonsynonymous mutations at a significantly higher relative rates than those expressed early, as measured by the d_N/d_S ratio, in mosquitoes, flies, mice, and humans. Significance was determined by Spearman rank correlations, and p-values in bold are significant at p < 0.05 (two-sided tests, not corrected for multiple tests). The lines are least squares regressions, and the gray regions show the approximate 95% confidence intervals for the regressions. See also Supplementary Fig. 1 and Supplementary Table 1.

Fig. 2. Late genes are more polymorphic for nonsynonymous mutations.

Genes that are more highly expressed late in life are significantly more polymorphic for nonsynonymous mutations than are genes expressed early in all four species. Nonsynonymous polymorphism is measured here as the fraction of sites in a gene that segregate for nonsynonymous alleles (p_N) relative to the fraction segregating for synonymous alleles (p_S). Significance was determined by Spearman rank correlations, and p-values in bold are significant at p < 0.05 (two-sided tests, not corrected for multiple tests). The lines are least squares regressions, and the gray regions show the approximate 95% confidence intervals for the regressions. See also Supplementary Fig. 2 and Supplementary Table 1.

Fig. 3. Late genes are younger.

Genes expressed later in life originated significantly more recently than those expressed early in all four species. Significance was determined by Spearman rank correlations, and p-values in bold are significant at p < 0.05 (two-sided tests, not corrected for multiple tests). The lines are least squares regressions, and the gray regions show the approximate 95% confidence intervals for the regressions. See also Supplementary Fig. 3 and Supplementary Table 1.

Fig. 4. Late genes are more likely to be driver genes for adult cancers in humans.

The probability that a gene is associated with adult cancers is significantly correlated with its age of expression in humans. Significance was determined by a Spearman rank correlation (two-sided test). The line is the least squares regression, and the gray region shows the approximate 95% confidence interval for the regression. See also Supplementary Fig. 4.

Fig. 5. Selection on late genes is weaker.

The relative strength of selection is significantly weaker in late-expressed genes than early-expressed genes in three of the four species. The estimates of N_e s are based on a highly simplified model that assumes beneficial mutations have fitness effect s and deleterious mutations fitness –s. Significance was determined by Spearman rank correlations, and p-values in bold are significant at p < 0.05 (two-sided tests, not corrected for multiple tests). The lines are least squares regressions, and the gray regions show the approximate 95% confidence intervals for the regressions. See also Supplementary Fig. 5 and Supplementary Table 1.

Fig. 6. Late genes fix beneficial mutations more often.

The fraction of nonsynonymous mutations that are beneficial (α_m) is significantly greater in late-expressed than early-expressed genes in two of the four species. Significance was determined by Spearman rank correlations, and p-values in bold are significant at p < 0.05 (two-sided tests, not corrected for multiple tests). The lines are least squares regressions, and the gray regions show the approximate 95% confidence intervals for the regressions. See also Supplementary Fig. 6 and Supplementary Table 1.