Adaptation by Loss of Heterozygosity in Saccharomyces cerevisiae Clones Under Divergent Selection

Abstract

Loss of heterozygosity (LOH) is observed during vegetative growth and reproduction of diploid genotypes through mitotic crossovers, aneuploidy caused by nondisjunction, and gene conversion. We aimed to test the role that LOH plays during adaptation of two highly heterozygous Saccharomyces cerevisiae genotypes to multiple environments over a short time span in the laboratory. We hypothesized that adaptation would be observed through parallel LOH events across replicate populations. Using genome resequencing of 70 clones, we found that LOH was widespread with 5.2 LOH events per clone after ∼500 generations. The most common mode of LOH was gene conversion (51%) followed by crossing over consistent with either break-induced replication or double Holliday junction resolution. There was no evidence that LOH involved nondisjunction of whole chromosomes. We observed parallel LOH in both an environment-specific and environment-independent manner. LOH largely involved recombining existing variation between the parental genotypes, but also was observed after de novo, presumably beneficial, mutations occurred in the presence of canavanine, a toxic analog of arginine. One highly parallel LOH event involved the ENA salt efflux pump locus on chromosome IV, which showed repeated LOH to the allele from the European parent, an allele originally derived by introgression from S. paradoxus Using CRISPR-engineered LOH we showed that the fitness advantage provided by this single LOH event was 27%. Overall, we found extensive evidence that LOH could be adaptive and is likely to be a greater source of initial variation than de novo mutation for rapid evolution of diploid genotypes.

Keywords: Saccharomyces cerevisiae; experimental evolution; gene conversion; mitotic recombination; resequencing.

Figure 1

Overview of LOH across HHL evolved clones across the 16 S. cerevisiae chromosomes. Each sample represents an individual clone sequenced from an independent population with environment indicated and mapped to the S288c genome. Shown are genotypes at 10 kb resolution, plotting the majority genotype across SNPs. For the eight in vivo populations, only the first clone is shown. White regions represent unmapped, typically repetitive regions of the S288c genome, and are not shown to scale.

Figure 2

Overview of LOH across LHL evolved clones across the 16 S. cerevisiae chromosomes. Each sample represents an individual clone sequenced from an independent population with environment indicated and mapped to the S288c genome. Shown are genotypes at 10 kb resolution, plotting the majority genotype across SNPs. White regions represent unmapped, typically repetitive regions of the S288c genome, and are not shown to scale.

Figure 3

Number of LOH events (A) and total LOH base pairs (B) across environments for the two ancestral genotypes. Bars represent means across independent populations and error bars show SE. Wine and larval populations were not generated for LHL. Letters above bars indicate environments which are not statistically distinguishable (P > 0.05) by a Dunn test after adjusting P-values using the Benjamini–Hochberg method. No pairwise comparisons were significant for the number of events (A) following P-value adjustment. Tests were conducted using HHL and LHL genotypes combined.

Figure 4

Fitness gain for each environment shows variation across measures and background genotype. Values shown are relative to ancestral genotype (dashed line). (A) Fitness as measured using doubling time. (B) Fitness measured using efficiency of growth (EOG) (i.e., carrying capacity). (C) Fitness measured using relative competitive growth in the presence of a marked ancestral strain. Each dot represents an individual clone. Boxes indicate the 25–75% percentiles, bars indicate medians, and the whiskers extend to 1.5 × the interquartile range. Asterisks indicate environments/measures where there was a significant differences in fitness of evolved strains depending on genetic background (P < 0.01; Welch two-sample t-test). Wine strains were not tested by growth curve analysis.

Figure 5

Overview of heterozygosity demonstrates LOH occurred at CAN1 in most canavanine evolved lines. Shown is the region of chromosome V from the left telomere to 120 kb. Heterozygosity is shown as the dominant genotype frequency in windows of 500 bp. White gaps indicate regions that cannot be mapped uniquely, are not polymorphic between the two parental genotypes, or have mean coverage below the threshold.

Figure 6

Overview of heterozygosity demonstrates extensive LOH favoring the allele from the clinical parent at the ENA locus in most high-salt evolved lines. Shown is the region of chromosome IV from 435 to 590 kb. Heterozygosity is shown as the majority genotype frequency in windows of 500 bp. White gaps indicate regions that cannot be mapped uniquely, are not polymorphic between the two parental genotypes, or have mean coverage below the threshold.

Figure 7

Rapid appearance of LOH genotype homozygous for ENA clinical allele in high-salt environment is associated with fitness increase measured by growth rate. (A) Initial ancestral haploids [SSP24 (Chinese parent), SSP253 (clinical parent), SSP264 (oak parent)] and diploids [SSP272 (LHL ancestor), SSP309 (LHL ancestor)], and CRISPR-mediated LOH transformants [China-cr-HOMO (homozygous for SSP24 ENA allele), Clinical-cr-HOMO (homozygous for SSP253 ENA allele)] show a marked difference in growth rate in salt. Values shown are relative to SSP309. Error bars indicate ±1 SD across replicate cultures or independent transformants. (B) Frequency of the homozygous genotype at the ENA locus over time; 36–50 clones per time point were used. Error bars indicate confidence intervals based on binomial distribution. AN-1 is one population of HHL and BN-2 is a population of LHL. (C) Relative fitness of strains with (Homo-Clin) and without (Het) LOH ENA genotype at intermediate time points. Fitness is relative to the ancestor. Error bars indicate ±1 SD across clone means. AN-1 population time points at 5 and 25 days are shown, and BN-2 population is shown at time 10 days. (D) Relative fitness of individual clones derived from six independent 100-day populations grown in high salt are shown across three media types. Error bars indicate ±1 SD across clone means.

Figure 8

Approximately half of LOH is generated from gene conversion-type events. (A). Clones displaying LOH can be classified according to their patterns, following St Charles et al. (2012). (B). These patterns show gene conversion to be the predominant mode of LOH, and the pattern does not differ dependent on initial host genotype. Two strains, L193-R05C1 and L197-R07C1, were not included due to their unusual patterns of LOH suggestive of meiosis or abortive meiosis.