Human Germline Mutation and the Erratic Evolutionary Clock

Abstract

Our understanding of the chronology of human evolution relies on the "molecular clock" provided by the steady accumulation of substitutions on an evolutionary lineage. Recent analyses of human pedigrees have called this understanding into question by revealing unexpectedly low germline mutation rates, which imply that substitutions accrue more slowly than previously believed. Translating mutation rates estimated from pedigrees into substitution rates is not as straightforward as it may seem, however. We dissect the steps involved, emphasizing that dating evolutionary events requires not "a mutation rate" but a precise characterization of how mutations accumulate in development in males and females-knowledge that remains elusive.

Fig 1. The many steps involved in the conversion of mutation rate estimates from pedigree studies into yearly substitution rates.

Fig 2. Schematic illustration of mutations occurring during embryonic development and gametogenesis.

For simplicity, we show only mutations that arose in the father and one offspring (child 1). Stars represent mutations that originate in different stages of embryogenesis and gametogenesis of the father and the offspring; solid stars are mutations that arise in the father, and hollow stars are those that occur in the offspring. Shown below each individual are the expected frequencies of the labeled mutations in his or her blood sample. Red, brown, and green stars are heritable and should be included in an estimate of germline mutation rates, whereas blue stars are somatic mutations present only in blood samples, which should be excluded. The detection of mutations that are mosaic in both soma and germline strongly suggests that, in the cell lineage tree of human development, soma and germline are not reciprocally monophyletic [46,64]. The standard pipelines require allelic balance in the child and no (or very low) read depths in the parents, leading to inclusion of some postzygotic mutations in the child and exclusion of a fraction of germline mutations in the parents. The two effects partially balance, so the overall mutation rate is unlikely to be greatly biased. However, there is a tendency to detect child-specific mutations and to miss ones shared among siblings. As a consequence, the mutation rates during early development are likely underestimated, with potentially important practical implications for predictions of recurrence risk of diseases caused by de novo mutations.

Fig 3. Variation in the estimated paternal age effect for autosomes.

We plot the de novo mutation rate as a function of the paternal age at conception of the child. The rate was obtained from the reported counts of de novo mutations divided by the fraction of the genome assayed in each study (shown in the title of each subplot, along with the mean sequence coverage per individual). The solid line denotes the fitted slope (i.e., the increase in the mutation rate for each additional year of father’s age). Following the approach of Rahbari et al. 2015 [46], for their study, we used the corrected counts of de novo mutations, which are extrapolated to a genome length of 3 Gb (thereby assuming the mutation rate in the inaccessible regions of the genome is the same as that in surveyed regions). The three colors used in this plot denote the three different families that were studied: blue, family 244; green, family 603; and red, family 569.