A comparison of humans and baboons suggests germline mutation rates do not track cell divisions

Abstract

In humans, most germline mutations are inherited from the father. This observation has been widely interpreted as reflecting the replication errors that accrue during spermatogenesis. If so, the male bias in mutation should be substantially lower in a closely related species with similar rates of spermatogonial stem cell divisions but a shorter mean age of reproduction. To test this hypothesis, we resequenced two 3-4 generation nuclear families (totaling 29 individuals) of olive baboons (Papio anubis), who reproduce at approximately 10 years of age on average, and analyzed the data in parallel with three 3-generation human pedigrees (26 individuals). We estimated a mutation rate per generation in baboons of 0.57×10-8 per base pair, approximately half that of humans. Strikingly, however, the degree of male bias in germline mutations is approximately 4:1, similar to that of humans-indeed, a similar male bias is seen across mammals that reproduce months, years, or decades after birth. These results mirror the finding in humans that the male mutation bias is stable with parental ages and cast further doubt on the assumption that germline mutations track cell divisions. Our mutation rate estimates for baboons raise a further puzzle, suggesting a divergence time between apes and Old World monkeys of 65 million years, too old to be consistent with the fossil record; reconciling them now requires not only a slowdown of the mutation rate per generation in humans but also in baboons.

Fig 1. Pedigrees and mean depth of coverage of the sequenced individuals.

Relationships among individuals in 3 human families (top row, teal) and 2 extended baboon pedigrees (bottom row, orange) are shown. Squares and circles represent males and females, respectively, and are annotated for the mean depth of sequencing coverage of the individual. Filled shapes indicate the focal F₁ individuals in which mutations were called. Sample identifiers and birth dates for all individuals are specified in S2 Table and S3 Table.

Fig 2. Dependence of mutation count on sex and age.

Each human (in A) and baboon (in B) F₁ individual is represented by a blue circle and a red triangle, which indicate how many DNMs were estimated to have occurred on the paternal or maternal genomes, respectively. Maternal points for 2 baboons were omitted because of a lack of information about the age of the mothers. Solid blue and red lines indicate the best fit obtained from a Poisson regression for each sex, respectively. Blue- and red-shaded regions denote the 95% CIs associated with the regression coefficients. Gray dashed lines represent the sex-specific fits from Gao and colleagues [32], scaled to account for differences in genome size in [32] versus this study. In (A), the solid gray line denotes the exponential maternal age effect fit from Gao and colleagues. We did not find a significant difference between the parental slopes of the 2 species (LR test p-value = 0.19 in fathers, 0.37 in mothers). (C) Human and baboon mutation spectra, in our sequenced cohorts and previously published datasets. Spectra are shown for all transmitted DNMs in humans (teal); all transmitted baboon DNMs (orange); DNMs called in a large set of Icelandic 3-generation pedigrees [17] (blue); and low-frequency SNPs (doubleton and tripleton alleles) identified in a sample of distantly related baboons (red). Each point denotes the relative proportion of mutations that belong to 1 of 7 types of mutations, as indicated on the x-axis; reverse complement mutation types were collapsed together into a single type (e.g., the left-most column, C>A, includes G>T mutations). Transitions at cytosine sites were separated into those that occurred at a CpG (CpG>TpG) or a non-CpG (CpH>TpH) site. For all datasets, only DNMs and SNPs located within regions of the genome determined to be orthologous between humans and baboons were included. Vertical lines represent 95% CIs from bootstrap resampling of 50 cM blocks. CIs on the 2 reference datasets (blue and red) were constructed by bootstrap resampling of mutations but are small and hidden by the points. Underlying data for this figure can be found in S2 Data. cM, centimorgan; CpG, 5′-cytosine-phosphate-guanine-3′; DNM, de novo mutation; LR, likelihood ratio; SNP, single-nucleotide polymorphism.

Fig 3. Estimates of the sex bias in mutation across mammals.

(A) Ratio of paternal-to-maternal mutations, α, among phased DNMs. Circles indicate the fraction of DNMs assigned to the paternal genome among all transmitted DNMs (i.e., DNMs observed in an F2 offspring); in our samples, the mean ages in humans were 34.3 years for males and 31.3 years for females and in baboons, 10.3 years and 9.2 years, respectively (so the ratio of male-to-female generation time was about 1.1 in both). Vertical lines indicate 95% CIs, which were determined by bootstrap resampling of 50 cM blocks. Squares denote the ratio of paternal-to-maternal mutations estimated for typical parental ages in the species, assuming ages 32.0 and 28.2 years for human males and females, respectively, and 10.7 and 10.2 years for baboon males and females. Vertical lines denote the 95% CIs of the age effect and intercept estimates, obtained assuming normally distributed errors. (B) The ratio of paternal-to-maternal DNMs estimated from pedigrees across 9 mammalian species as a function of the mean male generation times in the study (for references, see Table 1). The square and diamond points denote separate chimpanzee estimates from Besenbacher and colleagues and Tatsumoto and colleagues, respectively [22,23]. The black triangle denotes the estimate for DNMs phased by transmission in 225 3-generation human pedigrees from Jónsson and colleagues [17]. Human and baboon estimates reported in this study are colored teal and orange, respectively. Fitting a linear regression (dashed line) yields an estimated slope of only 2.8×10⁻³ (95% CI −9.2×10⁻⁴ to 6.5×10⁻³, p-value = 0.12). To avoid duplicates, values from one chimpanzee (Tatsumoto and colleagues [23]) and one human (Jónsson and colleagues [17]) study were not included in the regression. The shaded area denotes the 95% CI of the slope and intercept. Underlying data for this figure can be found in S2 Data. cM, centimorgan; DNM, de novo mutation; SNP, single-nucleotide polymorphism.

Fig 4. Implications of present-day human and baboon mutation rates for the evolution of the yearly mutation rate.

(A) Phylogenetic relationship between humans and baboons with a marmoset (New World monkey) outgroup. Branch lengths denote the autosomal substitution rate per bp since the OWM–marmoset split as measured using data from [29] for all mutation types at putatively neutral regions of the genome. The relative branch length difference between baboon and human lineages is indicated in purple. (B) Sex-averaged mutation rates per year. Mutation rates were based on fitted values for typical generation times (i.e., assuming 32.0 and 28.2 years in human males and females, respectively, and 10.7 and 10.2 years in baboon males and females) and turned into a sex-averaged per-year mutation rate following [42]. Vertical lines indicate the span covered by the 95% CIs of the intercept and slope of the age effect regressions. (C) The ratio of yearly mutation and substitution rates in baboon relative to human, as estimated for the different possible types involving combinations of strong (S: G/C) and weak (W: A/T) bp. Each point denotes a different type, and strong-to-weak (S>W) types were separated into those that occurred at a CpG or a non-CpG site. Points are colored according to whether GC-biased gene conversion is expected to favor (light red), disfavor (dark red), or have no effect (blue) on the mutation type. Horizontal lines denote 95% CIs on the mutation rate ratio computed by resampling of 50 cM blocks. The upper CI for CpG S>W extends out of frame to 2.8. Point estimates for the substitution rates in baboons and humans were taken from [29]. The identity line is drawn in gray for reference. (D) Predicted divergence times of humans and OWMs as a function of parental ages. Divergence times were predicted using mutation and autosomal substitution rates measured in humans (teal) and baboons (orange), across a span of plausible past generation times. Each point within the shaded areas represents the divergence time calculated at a particular paternal generation time (x-axis) and paternal-to-maternal generation time ratio (ranging from 0.8 to 1.2) as indicated in purple. The dashed gray line indicates a plausible upper bound for the split time inferred from the fossil record [49,50]. Underlying data for this figure can be found in S2 Data. BGC, biased gene conversion; bp, base pair; cM centimorgan; CpG, 5′-cytosine-phosphate-guanine-3′; OWM, Old World monkey.