Whole genome, whole population sequencing reveals that loss of signaling networks is the major adaptive strategy in a constant environment

Abstract

Molecular signaling networks are ubiquitous across life and likely evolved to allow organisms to sense and respond to environmental change in dynamic environments. Few examples exist regarding the dispensability of signaling networks, and it remains unclear whether they are an essential feature of a highly adapted biological system. Here, we show that signaling network function carries a fitness cost in yeast evolving in a constant environment. We performed whole-genome, whole-population Illumina sequencing on replicate evolution experiments and find the major theme of adaptive evolution in a constant environment is the disruption of signaling networks responsible for regulating the response to environmental perturbations. Over half of all identified mutations occurred in three major signaling networks that regulate growth control: glucose signaling, Ras/cAMP/PKA and HOG. This results in a loss of environmental sensitivity that is reproducible across experiments. However, adaptive clones show reduced viability under starvation conditions, demonstrating an evolutionary tradeoff. These mutations are beneficial in an environment with a constant and predictable nutrient supply, likely because they result in constitutive growth, but reduce fitness in an environment where nutrient supply is not constant. Our results are a clear example of the myopic nature of evolution: a loss of environmental sensitivity in a constant environment is adaptive in the short term, but maladaptive should the environment change.

Figure 1. Histogram of maximum allele frequencies reached of all mutations discovered across the three experiments.

Figure 2. Clonal interference plays a prominent role in the dynamics of adaptation.

Each point represents a mutation identified in one of the three experiments. Mutations that are at a lower allele frequency in the final time point than at an earlier time point (i.e. below the y = x line) have decreased due to interference from a competing lineage. Clonal interference affects 63% of all mutations, while 36% of mutations are driven to extinction due to clonal interference.

Figure 3. Dynamics and linkage of mutations above 10% allele frequency in (A) E1, (B) E2 and (C) E3.

Mutations and frequencies were discovered from population sequencing, and linkage with other mutations and the fluorescent marker was determined by genotyping clones. HXT6/7 frequency data in E3 are from . The red, yellow and green indicate the frequencies of the fluorescent makers (compare to Figure 1 of [12]) with lineages within those differently marked subpopulations originating from within them.

Figure 4. Identity, severity and function of recurrently mutated genes across all experiments, grouped by pathway.

Only genes with two or more identified mutations are included; bars are colored according to the predicted severity of each mutation on protein function.

Figure 5. Enrichment of mutation categories relative to the expectation.

Expectation was calculated empirically assuming random mutation across the genome, and significance of enrichment was determined using a chi-squared test. A) Coding versus non-coding mutations. B) Mutations that are predicted to be disruptive of protein function versus mutations predicted to not affect protein function. C) Enrichments within coding mutations.

Figure 6. Reduction in cell viability as a function of time in (A) clones isolated from the chemostat experiments and (B) strains carrying a single mutation from E3.

The deeper the blue color, the more significant the reduction in cell viability compared to a wild-type strain. Multiple independent clones with the same known genotype are indicated.

Figure 7. A model for adaptive strategy in the constant, glucose-limited environment of the chemostat.

The accumulation of beneficial mutations disruptive of signaling networks leads to the decoupling of sensing from response and the loss of environmental sensitivity. Loss of control of cAMP/PKA pathway function eliminates some of the normal checks required to pass START A, likely to a shortened G1 and constitutive cell division. Likewise, loss of repressors of glucose transporter transcription leads to their constitutive activation, likely enabling the cell to sequester more glucose, leading to increased growth and division.