The genome-wide rate and spectrum of spontaneous mutations differ between haploid and diploid yeast

Abstract

By altering the dynamics of DNA replication and repair, alternative ploidy states may experience different rates and types of new mutations, leading to divergent evolutionary outcomes. We report a direct comparison of the genome-wide spectrum of spontaneous mutations arising in haploids and diploids following a mutation-accumulation experiment in the budding yeast Saccharomyces cerevisiae Characterizing the number, types, locations, and effects of thousands of mutations revealed that haploids were more prone to single-nucleotide mutations (SNMs) and mitochondrial mutations, while larger structural changes were more common in diploids. Mutations were more likely to be detrimental in diploids, even after accounting for the large impact of structural changes, contrary to the prediction that mutations would have weaker effects, due to masking, in diploids. Haploidy is expected to reduce the opportunity for conservative DNA repair involving homologous chromosomes, increasing the insertion-deletion rate, but we found little support for this idea. Instead, haploids were more susceptible to SNMs in late-replicating genomic regions, resulting in a ploidy difference in the spectrum of substitutions. In diploids, we detect mutation rate variation among chromosomes in association with centromere location, a finding that is supported by published polymorphism data. Diploids are not simply doubled haploids; instead, our results predict that the spectrum of spontaneous mutations will substantially shape the dynamics of genome evolution in haploid and diploid populations.

Keywords: RDH54; aneuploidy; mitochondria; replication time.

Fig. 1.

Cell divisions during MA and growth rates following MA. Cell divisions per day (A, haploids; B, diploids) depended on the interaction of ploidy level, RDH54 status, and time (>3,200 colonies measured; linear model, ploidy × RDH54: P < 10⁻⁸; ploidy × time: P < 0.01). Solid (dashed) lines show linear regression for RDH54⁺ (rdh54Δ) treatments. (C) Violin plots of maximum growth rates after 100 transfers relative to ancestral controls, accounting for block effects. Points (error bars) show means (SEs). Steeper declines in cell divisions per day in diploids relative to haploids during MA (A and B) are consistent with relative growth rate estimates following MA (linear mixed model based on 4,000 growth curves: MA × ploidy × RDH54: P < 0.05). Diploid MA lines have significantly reduced growth rates relative to the ancestor (MA: P < 10⁻⁶, MA × RDH54: P < 0.05), unlike haploid lines (MA: P = 0.64). Growth-rate data for individual lines are provided in Dataset S2.

Fig. 2.

Mutation rates in each group of MA lines, with 95% CIs. Panels show SNMs, indels, MNMs, chromosome (Chrom.) gains and losses, mt (Mito.) SNMs, and Mito. indels. Rates represent events per base pair (bp) per generation (gen), except for whole-chromosome gains and losses (events per generation). For mt events, we consider all treatments effectively haploid. The absolute numbers of events observed (“Counts”) are given at the left of each panel; note that detection power differs among groups such that rates are not necessarily proportional to mutation counts. Statistically significant treatment effects are noted at the right of each panel (binomial tests accounting for detection power: *P < 0.05; **P < 0.01; ***P < 0.001). A numeric summary of rates and CIs is provided in Dataset S2.

Fig. 3.

Gains and losses of each chromosome during MA and the fitness consequences of aneuploidy. (A) There was no evidence that the distribution of aneuploidy events among chromosomes depended on RDH54 type (Fisher’s exact test: P = 0.56), but with few events in RDH54⁺ lines, we have low power to test this hypothesis. Pooling data between RDH54 types, gains were more common than losses (binomial test: P < 10⁻⁶) and rates vary among chromosomes (χ² = 42.1, simulated P < 0.001). Several cases of trisomy for chromosome 11 in RDH54⁺ diploids were determined to be ancestral and are not scored as gains; in one case, the extra copy was lost, restoring euploidy, and this is scored here as a loss. Aneuploid line identifications are provided in Dataset S2. (B) Genome size relative to controls was negatively correlated with maximum growth rate relative to controls in diploid MA lines (r = −0.74, df = 112, P < 10⁻¹⁵). Diploid RDH54⁺ controls had trisomy for chromosome 11, so MA lines with only this trisomy were scored as having zero genome size change and MA lines without this trisomy were scored as having a genome size reduction unless other aneuploidy was present. This correlation persists when only aneuploid lines are considered (r = −0.70, df = 78, P < 10⁻¹²) or when lines with no change are excluded (r = −0.84, df = 45, P < 10⁻¹²).

Fig. 4.

Spectrum of nuclear SNMs. (A) Fraction of SNMs of each type, including complementary changes. Ts, transition; Tv, transversion. RDH54 status did not significantly affect the SNM spectrum (G = 4.5, P = 0.48), but the SNM spectrum differed between ploidy levels (G = 19.8, P < 0.01), particularly among transversion mutations (G = 17.9, P < 0.001). By contrast, the spectrum of transition mutations and the overall transition-to-transversion ratio did not vary significantly by ploidy (G = 0.04, P = 0.84 and odds ratio = 0.88, P = 0.18, respectively). (B) Fraction of SNMs in each 3-bp context, including the complementary context, centered on the focal site, accounting for the frequency of each context in the genome. While the context of SNMs did not differ significantly between haploids and diploids when considering all contexts (G = 21.5, P = 0.90), the fraction of mutations at A/T versus C/G sites differed between ploidy levels (odds ratio = 1.22, P < 0.05; horizontal lines show means across A/T or C/G sites for each ploidy level). The overall rate of SNMs depended on the 3-bp context at G/C sites (G = 81.9, P < 10⁻¹⁰) but not at A/T sites (G = 20.9, P = 0.14). The mutation rate in XCG contexts (where X is any base) was elevated relative to the rate for C/G sites in other contexts (binomial test: P < 10⁻¹¹).

Fig. 5.

Genomic context of nuclear mutations. (A) DNA replication timing (data from ref. 28) of mutated sites differed significantly between haploids (red) and diploids (blue) (Wilcoxon test: P < 0.05). Haploid mutations arose in later-replicating positions relative to random callable sites (Wilcoxon test: P < 10⁻⁴), whereas diploid mutations did not differ significantly from the random expectation (Wilcoxon test: P = 0.09). (Inset) Mean replication timing for each group (also SI Appendix, Fig. S6). (B) Mutations versus distance from the centromere, as a fraction of maximum possible distance. All event types except indels occurred further from the centromere than random callable sites (Wilcoxon test, SNM and MNM: P < 0.05, Wilcoxon test, homozygous: P < 0.001), and homozygous diploid SNMs are further from the centromere than heterozygous SNMs (Wilcoxon test: P < 0.001). (Inset) Mean distance from the centromere for each event type. (C) GC content surrounding mutations relative to GC content surrounding equivalent random sites (A/T or C/G sites for substitutions, genic or nongenic sites for all mutation types). We consider only nongenic indels here because of possible selection against genic indels. MNM values represent the mean for the component sites. MNMs were associated with higher GC content within 50 bp (bootstrap P < 0.05), whereas nongenic indels were associated with lower GC content within 50 bp (bootstrap P < 0.01). (D) Variation in SNM rate among chromosomes (Left; diploids: G = 23.7, P = 0.07; haploids: G = 12.0, P = 0.68) was correlated with the relative distance between the centromere and the chromosome midpoint in diploids (Right), and this correlation differs significantly between ploidy levels (bootstrap P < 0.01). Centromere location was calculated as (L_q − L_p)/(L_q + L_p), where L_q is the length of the long arm and L_p is the length of the short arm. A summary of variants by chromosome is provided in Dataset S2.