Polygenic cis-regulatory adaptation in the evolution of yeast pathogenicity

Abstract

The acquisition of new genes, via horizontal transfer or gene duplication/diversification, has been the dominant mechanism thus far implicated in the evolution of microbial pathogenicity. In contrast, the role of many other modes of evolution--such as changes in gene expression regulation-remains unknown. A transition to a pathogenic lifestyle has recently taken place in some lineages of the budding yeast Saccharomyces cerevisiae. Here we identify a module of physically interacting proteins involved in endocytosis that has experienced selective sweeps for multiple cis-regulatory mutations that down-regulate gene expression levels in a pathogenic yeast. To test if these adaptations affect virulence, we created a panel of single-allele knockout strains whose hemizygous state mimics the genes' adaptive down-regulations, and measured their virulence in a mammalian host. Despite having no growth advantage in standard laboratory conditions, nearly all of the strains were more virulent than their wild-type progenitor, suggesting that these adaptations likely played a role in the evolution of pathogenicity. Furthermore, genetic variants at these loci were associated with clinical origin across 88 diverse yeast strains, suggesting the adaptations may have contributed to the virulence of a wide range of clinical isolates. We also detected pleiotropic effects of these adaptations on a wide range of morphological traits, which appear to have been mitigated by compensatory mutations at other loci. These results suggest that cis-regulatory adaptation can occur at the level of physically interacting modules and that one such polygenic adaptation led to increased virulence during the evolution of a pathogenic yeast.

Figure 1.

Illustration of approach for finding cis-regulatory selection. Each yeast cell represents a hybrid between two diverged strains. Four genes from a single gene set are shown, with two alleles per gene (red and green). The number of curved lines next to each allele represents that allele's expression level. (A) All four genes are expressed equally from each allele, so there is no allele-specific expression (ASE) or evidence for selection. (B) Each gene has a down-regulating cis-regulatory mutation (blue x), but these act on alleles from each parent with equal frequency, so there is still no evidence for selection. (C) All four genes have ASE in the same direction (down-regulating, red), a pattern indicating lineage-specific selection on cis-regulation (though more than four genes are required to achieve statistical significance).

Figure 2.

Population-genetic analysis. In all panels, patterns of genetic variation at the eight confirmed ASE genes are compared with the variation at all 971 genes showing ASE in the same direction. Variants are compared in a 1-kb moving window (500-bp step size), with all genes aligned by their 5′ ends and oriented in the same direction; the average ORF length is shown in red, and approximate promoter length is shown in green. Comparisons of either genetic diversity θ (A,B) or Tajima's D (C,D) are shown, with P-values computed using the nonparametric Wilcoxon test. The test is one-sided so that low P-values specifically indicate lower θ or D values among the eight ASE genes—as expected from a selective sweep—compared with the 971 control genes. Comparisons use either the 21 strains most closely related to Y (A,C), or one of three sets of 21 control strains (B,D): those most closely related to S (blue), those of intermediate distance to Y (red), or those of greatest distance to Y (green).

Figure 3.

Fitness of reciprocal hemizygous (RH) strains. (A) Fitness in vivo. RH strains and the wild-type Y/S hybrid were competed for 5 d in immunocompromised mice. Twenty out of 21 strains with fitness values significantly different from the wild type were fitter than the wild type (P = 10⁻⁵), despite showing no measurable growth differences at 30°C in vitro (Supplemental Figs. 1, 2). All but the two SLA2 RH strains differ significantly from the wild type. Error bars, ±1 SE. (B) High-temperature (37°C) growth in vitro predicts fitness in vivo. Each strain's relative fitness rank is plotted both in vivo and in vitro; a significant (P = 0.002) correlation is observed. (C) Fitness in vitro. Y-alleles yield higher fitness than S-alleles at 37°C in vitro for 11 of 11 RH strain pairs. Genes are ordered by decreasing significance of the Y-allele advantage. Error bars, ±1 SE.

Figure 4.

Pleiotropic effects on morphology. (A,B) Representative micrographs from a hemizygous deletion pair, LSB3 Δy/S (A), and LSB3 Y/Δs (B). Cells are stained with FITC- concanavalin A (green), Alexa Fluor 594 phalloidin (red), and DAPI (blue). (C) Barplot of a single morphological trait, cell size of budded yeast with two nuclei (C101_C), for all RH strain pairs. Error bars, ±1 SE. The work by Ohya et al. (2005) contains a full list of phenotypes and their descriptions. (D) Heatmap of hierarchically clustered nonredundant morphological phenotypes for all RH pairs. Phenotypes are the top principal components of 220 quantitative morphological traits of unbudded (A), budded with one nucleus (A1B), and budded with two nuclei (C) cells. The three leftmost phenotypes primarily correspond to cell size of yeast from each of the three cell types (A, A1B, and C).