Evolutionary rescue by compensatory mutations is constrained by genomic and environmental backgrounds

Abstract

Since deleterious mutations may be rescued by secondary mutations during evolution, compensatory evolution could identify genetic solutions leading to therapeutic targets. Here, we tested this hypothesis and examined whether these solutions would be universal or would need to be adapted to one's genetic and environmental make-ups. We performed experimental evolutionary rescue in a yeast disease model for the Wiskott-Aldrich syndrome in two genetic backgrounds and carbon sources. We found that multiple aspects of the evolutionary rescue outcome depend on the genotype, the environment, or a combination thereof. Specifically, the compensatory mutation rate and type, the molecular rescue mechanism, the genetic target, and the associated fitness cost varied across contexts. The course of compensatory evolution is therefore highly contingent on the initial conditions in which the deleterious mutation occurs. In addition, these results reveal biologically favored therapeutic targets for the Wiskott-Aldrich syndrome, including the target of an unrelated clinically approved drug. Our results experimentally illustrate the importance of epistasis and environmental evolutionary constraints that shape the adaptive landscape and evolutionary rate of molecular networks.

Keywords: Wiskott–Aldrich syndrome; aneuploidy; epistasis; experimental evolution; genotype‐by‐environment interaction.

Figure EV1

Sequence alignment of WASP and its homologs The conserved W64 residue is shown in bold. Sequences were obtained from the Ensembl database and aligned with Geneious 6.1.6.

Figure EV2

The thermosensitive phenotype of las17-41 on glucose and galactose synthetic media in BY and RM genetic background is recessive, showing a loss-of-function mutation The las17-41 allele causes a comparable thermosensitive phenotype in a haploid BY and RM genetic background on both carbon sources. The mechanism causing thermosensitivity is not a gain of function, because LAS17/las17-41 and LAS17/las17Δ hemizygous strains have the same growth phenotype as their wild type in both backgrounds. Spot assay shown after 3 days of growth.

Figure 1

Context-dependent las17-41 compensatory evolution

1. Compensatory mutation rate of the thermosensitive las17-41 phenotype in four contexts: RM and BY genetic backgrounds and two carbon sources, glucose (Glu) and galactose (Gal) (left axis, n = 3, 8, 4, and 4 for BY gal, BY glu, RM gal, and RM glu, respectively). Bars = average, error bars = SD. Overlaid dots show the number of unique rescue mutation (right axis) found in each context. Mutations at a shared locus and nucleotide position were considered equivalent.

2. The average growth of evolved strains sharing compensatory mutation types at 22°C (top panel) and 37°C (bottom panel) in four culture media is shown relative to their respective progenitor. The las17 strains carry R41W, R41L, or R41Q functional reversions. WT = wild-type strain. Growth phenotypes were measured on synthetic complete (SC) and rich media (YP), n = number of strains next to labels, each inferred from eight replicates. One cap1 mutant was excluded because of mitochondrial loss and one mutant because it contained two rescue mutations (act1 and cap2).

3. Proportion of compensatory mutations identified in the four experimental contexts. Only two genes were identified in all four contexts, indicating that at the target level, up to 80% of the mutations could be context dependent.

4. Proportion of each rescue mutation type encountered in each experiment, n = same as in C, SNV = single nucleotide variant.

Figure EV3

Predicted functional effects of point mutations Functional effect scores were computed with SNAP2 (Yachdav et al, 2014), a neural network-based method that uses in silico derived protein information to make predictions regarding functionality of individual substitutions. A score > 50 indicates a strong signal for effect, −50 < score < 50 indicates a weak signal, and a score < −50, a strong signal for neutral/no effect. Each dot corresponds to a possible substitution for each position of a protein. Larger dots were substitutions observed in the experiment. Dots are colored according to the experimental context in which they were observed. Black dots indicate the pre-existing SNP between the two ancestral genomes.

Figure 2

Ploidy variation among strains

1. Six strains had unequal read coverage across chromosomes, with a coverage of approximately 1.5× along up to six entire chromosomes. This factor is consistent with trisomia in diploid strains. The heat map shows the normalized average coverage across chromosomes for each sequenced mutant strains. ChXII is not shown because of a constant bias in its coverage estimate caused by the multiple copies of rRNA genes it encodes.

2. Ploidy measurements of all the sequenced strains using DNA content measurement by flow cytometry confirmed trisomia. Haploid and diploid assignments were determined by k-means clustering on test and control strains. Trisomic assignment is based on the unequal coverage values in (A). A diploid control strain of the RM genetic background is also included. Some strains for which a coding mutation was identified were also diploid. Because the frequency of the variant was in all cases near 100% (Table EV1), meaning that the diploid mutants are homozygous for the rescue mutation, it can be assumed that diploidization occurred after the mutation.

Figure 3. Network of protein-coding rescue genes

Empty circles regroup targets that were found in a particular context. Nodes are colored by their function and edges represent protein–protein interactions reported in Biogrid 3.3.122 (Chatr-Aryamontri et al, 2015). Protein-coding compensatory genes are strongly associated with Las17 at the network level as they show 15-fold enrichment in protein–protein interactions among themselves (Fisher's exact test, P-value < 1e-16). Ten of these proteins physically interact with Las17, a significant enrichment (Fisher's exact test, P-value = 3e-14), suggesting that interacting proteins are prime candidates for rescue mutations. Interestingly, in case where an interaction interface with Las17 has been identified, the mutations found do not coincide with the interacting residues (Chereau et al, ; Ti et al, 2011), suggesting more complex compensatory mechanisms than direct physical interactions with the altered interface of Las17. These interactors also have a smaller shortest path to one another than expected by randomly sampling the Las17 interactome (left-sided P-value = 0.03), showing that this subset is intimately connected to the essential function of LAS17. In agreement with this observation, these 13 genes are enriched in GO terms related to Las17 functions, for example, actin filament polymerization (GO:0030041, Holm–Bonferroni corrected P-value = 3e-14).

Figure 4

Variation in intra- and intergenic compensatory gain of function across genotypes and environmental conditions

1. Bars show the fraction of colonies transformed with a mutant allele showing more growth at 37°C than 95% of the control, that is, the originating wild-type allele in BY and RM las17-41 thermosensitive strains on glucose and galactose. The x-axis indicates the amino acid changes for each allele. A low fraction of compensated transformants could reflect that the addition of the mutant allele increases the compensatory mutation rate only or that it has a partially recessive effect. Because the wild-type copy is also present in the transformed haploid strains, a recessive gain of function would not be revealed by this experiment.

2. Increasing the gene dosage of actin by adding a wild-type actin allele from BY or RM (x-axis) can compensate las17-41 thermosensitivity in the RM background on glucose. Bars show the fraction of colonies that grow more than 95% of control colonies at 37°C. Control colonies are BY and RM las17-41 transformed with an empty plasmid (pRS316), which reflects the background growth owed to compensatory mutants arising during the experiment.

Figure 5

A target-oriented network of genetic remedies to WAS disease The inner circle depicts the WAS disease-associated gene network known in human. The blue nodes expanding this network represent human orthologs of las17-41 compensatory rescue genes (corresponding yeast genes are in parenthesis). HumanNet-based connections between WAS disease genes and compensatory orthologs are highlighted in black. Black outlined nodes are new candidates for human Wiskott–Aldrich syndrome pathway genes. Dashed outlines indicate the orthologous ARP2/3 complex and Ca²⁺/calcineurin pathway, which appear as context-dependent functional target in yeast. The network was generated by MORPHIN (Hwang et al, 2014) and visualized with Cytoscape (Su et al, 2014). Source data are available online for this figure.

Figure EV4

Compensation by loss of function Spot dilution assay shown after 3 days of growth on SC glucose or galactose media for whole gene deletions of genes identified by genome sequencing in the thermosensitive strains. The results confirm that the loss of function of certain genes can counterweight las17-41 thermosensitivity, in a context-specific manner. The case of twf1Δ in particular is restricted to the RM genetic background as previously observed among the sequenced mutants. The deletion of bsp1 follows the same pattern, but spontaneous mutants can be observed in the BY background on glucose. It is noteworthy that the only BY bsp1 mutant recovered in the experiment was also diploid; hence, bsp1 mutations may be compensatory in this background on glucose only in combination with another mutational event.

Figure 6. Pharmacological inhibition of calcineurin by cyclosporin A compensates the las17-41 thermosensitivity at 37°C in a genetic background and carbon-source specific manner

Optimal growth is observed with 10 μg/ml of cyclosporin A (FigEV5) indicating that a calcineurin-independent cytotoxic effect of cyclosporin A is likely at play at higher concentrations (Singh-Babak et al, 2012), although this would be las17-41-specific since the wild-type strain exhibits normal growth. Dissection of heterozygous LAS17 deletion strains revealed that in glucose, LAS17 is essential in RM, but thermosensitive in BY as shown by the spot assays. In galactose, LAS17 is also essential in the BY background. Cyclosporin A does not compensate thermosensitivity of BY las17Δ at 37°C, suggesting that its effect is specific to this disease mimicking mutation in LAS17. Spot assays shown after 3 days of growth.

Figure EV5

Pharmacological compensation of las17-41 thermosensitivity by cyclosporin A is optimal at 10 μg/ml.