Graph-Based Pan-Genome Reveals the Pattern of Deleterious Mutations during the Domestication of Saccharomyces cerevisiae

Abstract

The "cost of domestication" hypothesis suggests that the domestication of wild species increases the number, frequency, and/or proportion of deleterious genetic variants, potentially reducing their fitness in the wild. While extensively studied in domesticated species, this phenomenon remains understudied in fungi. Here, we used Saccharomyces cerevisiae, the world's oldest domesticated fungus, as a model to investigate the genomic characteristics of deleterious variants arising from fungal domestication. Employing a graph-based pan-genome approach, we identified 1,297,761 single nucleotide polymorphisms (SNPs), 278,147 insertion/deletion events (indels; <30 bp), and 19,967 non-redundant structural variants (SVs; ≥30 bp) across 687 S. cerevisiae isolates. Comparing these variants with synonymous SNPs (sSNPs) as neutral controls, we found that the majority of the derived nonsynonymous SNPs (nSNPs), indels, and SVs were deleterious. Heterozygosity was positively correlated with the impact of deleterious SNPs, suggesting a role of genetic diversity in mitigating their effects. The domesticated isolates exhibited a higher additive burden of deleterious SNPs (dSNPs) than the wild isolates, but a lower burden of indels and SVs. Moreover, the domesticated S. cerevisiae showed reduced rates of adaptive evolution relative to the wild S. cerevisiae. In summary, deleterious variants tend to be heterozygous, which may mitigate their harmful effects, but they also constrain breeding potential. Addressing deleterious alleles and minimizing the genetic load are crucial considerations for future S. cerevisiae breeding efforts.

Keywords: Saccharomyces cerevisiae; deleterious mutations; domestication; graph-based pan-genome; heterozygosity.

Figure 1

Pan-genomic landscape based on 41 third-generation genomes and the reference genome, S288C. (A) Network diagram of inter-chromosomal recombination relationships in 41 S. cerevisiae strains. (B) Pan-genome variation landscapes.

Figure 2

Genomic landscape of S. cerevisiae variation. (A) Distribution of structural variations across chromosomes in each strain. (B) NJ tree of 687 isolates based on the presence or absence of all the variants. (C) Ratio of heterozygous variations in the wild, SSF, and LSF groups. The middle bars represent the median, while the bottom and top of each box represent the 25th and 75th percentiles, respectively, and the whiskers extend to 1.5 times the interquartile range. Dots are outliers. *** p = 0.001 (t-test).

Figure 3

SFS spectra, fitness distribution, and proportion of adaptive variation (α) in the wild, SSF, and LSF groups, respectively. (A) Shows the unfolded SFS spectra of nSNPs, indels, and SVs (including INS/DEL) in wild, SSF strains, and LSF strains. (B) Displays the inferred fitness effects distribution (Nes) of nSNPs, indels, and SVs in wild, SSF strains, and LSF strains. (C) Proportion of adaptive variation (α) in the wild, SSF, and LSF strains. Error bars represent mean ± 95% CI. Dots are outliers. *** p = 0.001 (t-test).

Figure 4

Comparison of genetic loads in S. cerevisiae. (A) Comparison of additive, heterozygous, and recessive (the number of homozygous SVs per S. cerevisiae) loads in wild-type, SSF, and LSF isolates. The middle bars represent the median, while the bottom and top of each box represent the 25th and 75th percentiles, respectively, and the whiskers extend to 1.5 times the interquartile range. Dots are outliers. (B) Correlation between levels of genomic heterozygosity and its harmfulness (estimated by SIFT v2.1). A smaller SIFT score indicates that the mutation is more likely to be deleterious. (C) Heterozygosity levels for different types of genic variants in the wild, LSF, and SSF groups. Error bars represent mean ± SD. *** p = 0.001 (t-test).

Figure 5

Analysis of genomic variation patterns relative to recombination based on all variant information. (A) Number of dSNPs in each window for the three groups (wild, SSF, and LSF). (B) Number of SVs in each window for the three groups. The red line represents the regression line, and Pearson’s correlation coefficient and significance p-value are annotated in each graph. The recombination rate (rho) and number of variants (isolating + fixed) were measured in 50 kb windows.

Figure 6

Comparison between selection scan regions (SS) and non-SS regions (NSS) based on the entire variant dataset. (A) Manhattan plot of F_ST values, based on 10 kb window variants is presented for comparisons between wild and SSF strains, as well as between wild and LSF strains. (B) Ratios of dSNPs, indel, and SVs to sSNPs in SS and non-SS regions of SSF and LSF strains. Error bars represent mean ± SD. *** p = 0.001 (t-test).