Incipient balancing selection through adaptive loss of aquaporins in natural Saccharomyces cerevisiae populations

Abstract

A major goal in evolutionary biology is to understand how adaptive evolution has influenced natural variation, but identifying loci subject to positive selection has been a challenge. Here we present the adaptive loss of a pair of paralogous genes in specific Saccharomyces cerevisiae subpopulations. We mapped natural variation in freeze-thaw tolerance to two water transporters, AQY1 and AQY2, previously implicated in freeze-thaw survival. However, whereas freeze-thaw-tolerant strains harbor functional aquaporin genes, the set of sensitive strains lost aquaporin function at least 6 independent times. Several genomic signatures at AQY1 and/or AQY2 reveal low variation surrounding these loci within strains of the same haplotype, but high variation between strain groups. This is consistent with recent adaptive loss of aquaporins in subgroups of strains, leading to incipient balancing selection. We show that, although aquaporins are critical for surviving freeze-thaw stress, loss of both genes provides a major fitness advantage on high-sugar substrates common to many strains' natural niche. Strikingly, strains with non-functional alleles have also lost the ancestral requirement for aquaporins during spore formation. Thus, the antagonistic effect of aquaporin function-providing an advantage in freeze-thaw tolerance but a fitness defect for growth in high-sugar environments-contributes to the maintenance of both functional and nonfunctional alleles in S. cerevisiae. This work also shows that gene loss through multiple missense and nonsense mutations, hallmarks of pseudogenization presumed to emerge after loss of constraint, can arise through positive selection.

Figure 1. Yeast aquaporins underlie freeze-thaw tolerance in yeast.

(A) Genome scan of YPS163 x S288c identified a major QTL on the left arm of Chromosome 12. (B) A second scan in which the first QTL was treated as a fixed term identified a second QTL on the right arm of Chromosome 16. LOD scores at which FDR = 0.01 are indicated by horizontal red lines. Additional peaks in Figure 1B were not significant when both QTL on Chromosomes 12 and 16 were held fixed, suggesting they may be false positives. (C) Effect plot from QTL mapping. (D) Freeze-thaw tolerance was measured in parental strains, YPS163 (‘YPS’) mutants, and S288c-derivative BY4741 (‘S’) harboring empty vector or a plasmid-borne copy of AQY1 or AQY2 from YPS163, as described in Methods. The average and standard deviation of biological quadruplicates is shown.

Figure 2. Multiple independent losses of AQY2 and AQY1 in diverse S. cerevisiae isolates.

Bayesian trees of the recapitulated AQY proteins (where gaps were treated as missing data before translation) for Aqy2 (left) and Aqy1 (right). Trees were generated using Mr. Bayes 3.1 and a mixed amino acid replacement model with invgamma rates. All nodes displayed posterior probabilities >0.92. Strains with full-length proteins are nearly identical and do not resolve in the tree. Each star represents the appearance of a different deletion, including the Malaysian G528 (grey), 11-bp (yellow), and Asian G25 (blue) deletions in AQY2, and the amino-terminal Malaysian (green) and A881 (orange) deletions in AQY1. Appearance of the aqy1-V121M polymorphism is highlighted with a pink circle. Strains are color coded according to their niche as shown in the key. Freeze-thaw tolerance scores are listed to the right of strains shown in the Aqy1 tree (with +++ for tolerance and–for sensitivity, see Methods for details).

Figure 3. Skewed patterns of variation surrounding the AQY2 locus.

The average number of pairwise nucleotide differences per 1,000 bp sliding window of step size of 100 bp was plotted on Chromosome 12. Variation within (blue curve) and between (red curve) groups of (A) 4 strains harboring the Asian G25 AQY2 deletion, (B) 16 strains with the 11-bp deletion allele, and (C) 6 strains containing the full-length AQY2. Horizontal blue lines represent the genome-wide average of pair-wise variation within each group, and vertical lines represent manually defined breakpoints in trends. Regions identified with skewed between-group minus within-group variation (see Methods) are highlighted in blue. (D) A plot of F_ST based on the three groupings in (a–c); horizontal grey line represents the genome-wide average. Gene positions are shown above plots as black boxes, with AQY2 highlighted in yellow.

Figure 4. Strains lacking AQY genes show a fitness advantage under high osmolarity.

Cells were grown on solid agar plates containing 1.5 M sorbitol for 2 days and the number of colony-forming units (CFU) was compared to a no-stress control plate. (A) CFU for YPS163 and YPS163 aqy1Δ aqy2Δ grown on rich medium plus 1.5 M sorbitol; (B) S288c-derivative BY4741 harboring the empty vector or a plasmid expressing the YPS163 allele of AQY1 or AQY2, grown on selective medium with 1.5 M sorbitol. The average and standard deviation of biological triplicates is shown.

Figure 5. AQY function is required for sporulation in YPS163.

(A) Denoted strains were sporulated as described in Methods for 2 days, and the number of events with 4, 3, 2 spores or ‘other’ (representing unsporulated or unscorable cells) was recorded. Each plot represents the average and standard deviation of biological triplicate. Results from S. paradoxus strain NRRL Y-17217 are shown here. (B) Vineyard strain M22, sake strain K1, SK1, and S228c derivative BY4743 (BY) harboring empty vector or pYPS_AQY1 plasmid were sporulated as described for 2–3 d before scoring.