Population Genetics of Mutation Rate Evolution

Abstract

Mutation rates determine the pace of evolution, shape the burden of inherited disease, and ultimately constrain the complexity and adaptability of every species. They are not a fixed parameter of life, however, but a trait that evolves: mutation rates and mutational spectra vary by orders of magnitude across the tree of life, differ between continental human populations, and even differ measurably among families. Understanding what drives this variation requires connecting the biology of DNA damage and repair with the population-genetic forces that act on the genes responsible. This thesis approaches that problem at two complementary scales. Chapter 2 asks why long-lived vertebrates accumulate more germline mutations per generation than short-lived species, even though those same long-lived species have evolved unusually efficient somatic DNA repair. Decomposing the per-generation germline mutation rate into a prepuberty component mu_E and a per-year postpuberty rate arising in the spermatogonia and oocytes mu_O + mu_S, we show using parent-offspring trio data from eight mammalian species that mu_O + mu_S is significantly lower in long-generation-time species, while mu_E trends weakly upward. Long-lived species do invest in more accurate DNA repair, but only the postpuberty germ cell component reflects that investment; the per-generation rate climbs anyway because the longer reproductive lifespan more than compensates. This motivates a modified drift-barrier model in which selection against clocklike mutator alleles, whose excess mutation load scales with generation time, is strongest in long-lived species despite their small effective population sizes. Chapter 3 turns to within-species variation by asking whether weak-effect mutator alleles, the standing genetic variation in DNA repair that several lines of evidence suggest must exist in humans, can actually be detected in a present-day population. We develop an identity-by-descent (IBD) footprint method that counts mutations on the haplotypes co-inherited with rare candidate mutator variants, and apply it to whole-genome sequencing data from approximately 200,000 UK Biobank individuals across a panel of 8 DNA repair genes. The method resolves effect sizes as small as 1.5-fold, well below the floor of state-of-the-art trio scans. Despite this power, no weak-effect mutators are detected, even though standard drift-barrier theory predicts that mutators of this magnitude should have reached detectable frequencies given recent European demographic history. We argue that this absence is best explained by pleiotropic fitness costs of variants in DNA repair genes: the same machinery that suppresses germline mutations also suppresses somatic mutations that contribute to cancer and other late-onset disease, so the effective selection coefficient against a mutator allele in vivo is substantially larger than its direct germline mutation load alone would imply. Together, the two chapters point to a common conclusion. The germline and somatic compartments share most of their DNA repair machinery, and the selective forces shaping mutation rates in one compartment cannot be understood in isolation from the other. At the macroevolutionary scale, this shared machinery causes life history to reshape cross-species germline mutation rates; at the microevolutionary scale, it causes pleiotropy to dominate selection on segregating mutator variants.