Gene Loss Predictably Drives Evolutionary Adaptation

Abstract

Loss of gene function is common throughout evolution, even though it often leads to reduced fitness. In this study, we systematically evaluated how an organism adapts after deleting genes that are important for growth under oxidative stress. By evolving, sequencing, and phenotyping over 200 yeast lineages, we found that gene loss can enhance an organism's capacity to evolve and adapt. Although gene loss often led to an immediate decrease in fitness, many mutants rapidly acquired suppressor mutations that restored fitness. Depending on the strain's genotype, some ultimately even attained higher fitness levels than similarly adapted wild-type cells. Further, cells with deletions in different modules of the genetic network followed distinct and predictable mutational trajectories. Finally, losing highly connected genes increased evolvability by facilitating the emergence of a more diverse array of phenotypes after adaptation. Together, our findings show that loss of specific parts of a genetic network can facilitate adaptation by opening alternative evolutionary paths.

Keywords: adaptation; evolvability; experimental evolution; fitness landscape; genetic network; oxidative stress.

Fig. 1.

A comprehensive genetic screen identifies cellular processes important for resistance to paraquat stress. (A) Schematic overview of the experiment. All the deletion strains from the haploid deletion collection were pooled, and the pool was grown with and without paraquat to determine the relative fitness of each individual mutant under paraquat stress. (B) Interaction network with the genes that, when deleted, increase sensitivity to paraquat (log FC < −3.5). Three of the most enriched cellular processes are highlighted in gray (chromatin organization), gold (cellular response to oxidative stress), and blue (vesicle-mediated transport). The size of the nodes represents the sensitivity to paraquat as determined in the genome-wide screen. The thickness of the edges represents the confidence score associated with the interaction as determined by STRING. Nodes with red borders represent genes for which deletion mutants were made and which were selected for experimental evolution. (C) Growth rates and final OD₆₀₀ values after 120 h of deletion strains selected as starting strains for experimental evolution on YP 2% (w/v) glycerol with 0.125 mM paraquat. Error bars represent SEM of four replicate measurements. Strains are colored based on the cellular process they belong to: gray for chromatin organization, gold for cellular response to oxidative stress, and blue for vesicle-mediated transport.

Fig. 2.

Deletion strains greatly improve growth during adaptive evolution. (A) Schematic overview of the evolution experiment and gene network with all selected genes. Strains with deletions in different modules of the genetic network underlying resistance to oxidative stress are evolved in quadruplicate under paraquat stress for 150 generations. Genes in the network are colored corresponding to the indicated cellular processes they belong to. The color scheme is maintained throughout the figure. (B) Growth rates of one fit clone from each evolved population on rich medium with 2% (w/v) glycerol and 0.125 mM paraquat. Each point represents the mean of four replicate measurements. The deletion strains are sorted according to their average growth rate after evolution. The red dotted line represents the average growth rate of the wild type before evolution. (C) Improvement in growth rate of one fit clone for each evolved population on rich medium with glycerol and paraquat. (D) Result of a multivariate linear model analysis of the growth rate after evolution in function of various biological parameters. Shades of blue correspond with significance level. (E) Estimated regression coefficient per network module of the deleted gene with their 95% confidence intervals (multivariate linear regression).

Fig. 3.

Evolution is convergent at the level of modules and individual genes. (A) Number of SNVs and indels per sequenced sample. (B) Number of times a particular gene is hit across the sequenced samples, compared with what would be expected by chance. (C) Overview of the cellular processes (GO terms) that are mutated more often than expected by chance across all samples (supplementary table 7, Supplementary Material online), and two of the most frequently duplicated chromosomes. The size of the circles represents the fraction of independently evolved clones with a mutation in the indicated gene or pathway. (D) Specificity of each mutated process for the original genetic background and its influence on growth rate. The left side of the figure shows how often each mutated process (rows) is found within the original genetic backgrounds (columns), the latter of which are represented by genetic modules of the deleted genes. P-values represent enrichment scores and were calculated using a competitive gene set overrepresentation test as calculated by the camera function of the edgeR package. The right side of the figure shows the effect of having a mutation in one of these processes on the growth rate of the evolved strains. The average growth rates are calculated across all evolved lineages. **P < 0.01, ***P < 0.001, ****P < 0.0001 (Student’s t-test).

Fig. 4.

Validation of module-specific adaptation. (A) Mutations in ERAD-M, SPS, or NRT1, projected onto the original gene deletion network with genes important for resistance to paraquat stress. The intensity of the colors corresponds to the relative frequency of individually evolved clones with a mutation in the indicated complex, pathway or gene. (B) Growth curves of double deletion mutants (ancestral deletion and adaptive deletion) on YP 2% (w/v) glycerol with 0.125 mM paraquat. Each row contains one of the original genetic backgrounds that were used during experimental evolution, and each column contains deletions in genes that were indicated to be adaptive. Dark blue curves represent the growth of the ancestor strain without any other additional deletions (first column), orange curves correspond to the growth of the double deletion strain indicated by the row and column headings, and gray curves represent growth curves of all other strains within the same row. Error bars represent the standard deviation of four replicates. (C) Heat map with the increase in growth rate for each combination of deletions. (D) Heat map with the increase in final OD₆₀₀ for each combination of deletions.

Fig. 5.

Adaptation after losing genes with more interaction partners results in a higher phenotypic variability between independent lineages. (A) Heat map with the fitness increase (%) of the evolved strains (columns) across 22 different phenotypic conditions (rows). Red hues indicate a fitness increase as compared with the corresponding strain before evolution and blue hues indicate a decrease in fitness. (B) Phenotypic distance between replicate lineages in function of the number of genetic interactions of the deleted gene. A high value for the phenotypic distance corresponds with more variation between the phenotypic profiles of the replicate lineages of a particular deletion strain. *P < 0.05, ***P < 0.001 (Student’s t-test). (C) Mutational distance between replicate lineages of the wild-type strains compared with strains lacking genes from the HIR complex. The mutational profile is defined by the genes and GO categories that were mutated after evolution. A high value for the mutational distance corresponds with more variation between the mutational profiles of the replicate lineages of a particular deletion strain. ***P < 0.001 (Student’s t-test). (D) Phenotypic distance between replicate lineages in function of phenotypic potential, as determined in Levy and Siegal (2008). The phenotypic potential represents the overall phenotypic variation when a gene is deleted. Lines represent the best linear fit with its 95% confidence interval.