Molecular specificity, convergence and constraint shape adaptive evolution in nutrient-poor environments

Abstract

One of the central goals of evolutionary biology is to explain and predict the molecular basis of adaptive evolution. We studied the evolution of genetic networks in Saccharomyces cerevisiae (budding yeast) populations propagated for more than 200 generations in different nitrogen-limiting conditions. We find that rapid adaptive evolution in nitrogen-poor environments is dominated by the de novo generation and selection of copy number variants (CNVs), a large fraction of which contain genes encoding specific nitrogen transporters including PUT4, DUR3 and DAL4. The large fitness increases associated with these alleles limits the genetic heterogeneity of adapting populations even in environments with multiple nitrogen sources. Complete identification of acquired point mutations, in individual lineages and entire populations, identified heterogeneity at the level of genetic loci but common themes at the level of functional modules, including genes controlling phosphatidylinositol-3-phosphate metabolism and vacuole biogenesis. Adaptive strategies shared with other nutrient-limited environments point to selection of genetic variation in the TORC1 and Ras/PKA signaling pathways as a general mechanism underlying improved growth in nutrient-limited environments. Within a single population we observed the repeated independent selection of a multi-locus genotype, comprised of the functionally related genes GAT1, MEP2 and LST4. By studying the fitness of individual alleles, and their combination, as well as the evolutionary history of the evolving population, we find that the order in which these mutations are acquired is constrained by epistasis. The identification of repeatedly selected variation at functionally related loci that interact epistatically suggests that gene network polymorphisms (GNPs) may be a frequent outcome of adaptive evolution. Our results provide insight into the mechanistic basis by which cells adapt to nutrient-limited environments and suggest that knowledge of the selective environment and the regulatory mechanisms important for growth and survival in that environment greatly increase the predictability of adaptive evolution.

Figure 1. Increased fitness in nutrient-limited environments is associated with amplification of specific permease genes.

(A) Fitness increases for clones recovered from each selection are typically >10%. Haploid (1N) and diploid (2N) ancestral strains were also tested in ammonium-limited chemostats but did not show fitness differences. (Amm : ammonium, Arg : arginine, Gln : glutamine, Pro : proline, Glu : glutamate, Urea : urea, Alla : allantoin, Anc : ancestor). (B) DNA copy number was estimated using aCGH. Each black point represents a measurement from a unique probe on the microarray from analysis of population DNA samples. We detected CNVs containing genes with clear connections to nitrogen import at high frequencies in populations (red lines) and clones (blue lines). Retrotransposon (Ty) sequences were frequently found at the boundary regions of CNVs.

Figure 2. Overview of the classes of mutations identified in lineages adapted to nitrogen-limited conditions.

(A) In total, 117 mutational events (Table S1) were identified in 18 sequenced clones resulting in sequence (red) and structural (blue) variation. (B) The number and type of mutations acquired in each individual clones genotyped using aCGH and whole genome sequencing. Non-synonymous SNPs and CNVs are found in most clones.

Figure 3. Adaptive mutations occur in functionally related loci.

(A) A small number of loci are mutated in multiple nitrogen-limited environments. Some loci found to be mutated in nitrogen-limiting conditions have also been reported as associated with adaptive evolution in glucose-limited environments (*Wenger, J. et al , **Kvitek, D.J. et al , ***Gresham, D. et al [9]). The color of edges represents the type of allele and the width of the edge represents the frequency of the allele in the population. (B) GO term enrichment analysis of mutated loci within clones and populations, analyzed at different allele frequency thresholds, identified in nitrogen-limited environments shows enrichment for specific cellular functions.

Figure 4. Functional effects of adaptive mutations in a gene network polymorphism.

(A) NCR genes are altered in expression in clones recovered from ammonium-limited conditions. Only genes having at least one observation with log₂ ratio >|1.5| were included (29/38 NCR genes [71]). Genes and samples are hierarchically clustered using centered correlation and complete linkage. (B) Three independently acquired GAT1 mutations found in a single ammonium-limitation adapted population are clustered in the zinc finger DNA binding domain of the encoded protein. The wild type GAT1 protein sequence was queried using the Protein Model Portal database . (C) Two different point mutations in MEP2 found in ammonium-limitation adapted clones change the identical codon within a putative trans-membrane domain. Domain information was obtained from SGD database (http://www.yeastgenome.org/). (D) GAT1 and LST4 likely regulate the production and delivery of MEP2 to the plasma membrane at the transcriptional and post-translational level, respectively.

Figure 5. Recurrent selection and evolutionary dynamics of a GNP.

(A) Estimated genotype dynamics during adaptive evolution. The time of introduction of each new mutation (dotted circles) is estimated on the basis of detecting an allele frequency of at least 5% in the population. Some mutations were clustered based on their similarity in the dynamics (see Figure S11). The temporal order of mutations that occurred in rapid succession (white arrows) was determined on the basis of their allele frequencies in the final evolved population estimated using deep sequencing data (Figure S7). (B) Fitness estimates of 8 backcrossed strains, representing all possible combinations of alleles that comprise the GNP, from clone 3 isolated from the ammonium-limitation selection were determined by direct competition with either the ancestral or the gat1-2/lst4-2/mep2-2 genotypes. Error bars are 95% CI of the regression coefficient. (C) Fitness landscape reconstruction based on the fitness estimates for the 8 genotypes. The selection coefficient values of each strain are represented as color intensity. The width of each edge is proportional to the difference in fitness between two genotypes that edges connect. A solid line indicates a favored path whereas a dashed line indicates a disfavored path. Selection favors thicker, solid lines in the evolutionary trajectory.