Skip to main content
NCBI home page
Evolution Letters logoLink to Evolution Letters
. 2023 Nov 10;8(2):243–252. doi: 10.1093/evlett/qrad055

Fluctuating selection facilitates the discovery of broadly effective but difficult to reach adaptive outcomes in yeast

Vincent J Fasanello 1, Ping Liu 2,3, Justin C Fay 4,5,✉, Carlos A Botero 6,7
PMCID: PMC10959490  PMID: 38525031

Abstract

Evolutionary compromises are thought to be common under fluctuating selection because the mutations that best enable adaptation to one environmental context can often be detrimental to others. Yet, prior experimental work has shown that generalists can sometimes perform as well as specialists in their own environments. Here we use a highly replicated evolutionary experiment (N = 448 asexual lineages of the brewer’s yeast) to show that even though fluctuation between two environmental conditions often induces evolutionary compromises (at least early on), it can also help reveal difficult to reach adaptive outcomes that ultimately improve performance in both environments. Specifically, we begin by showing that yeast adaptation to chemical stress can involve fitness trade-offs with stress-free environments and that, accordingly, lineages that are repeatedly exposed to occasional stress tend to respond by trading performance for breadth of adaptation. We then show that on rare occasions, fluctuating selection leads to the evolution of no-cost generalists that can even outcompete constant selection specialists in their own environments. We propose that the discovery of these broader and more effective adaptive outcomes under fluctuating selection could be partially facilitated by changes in the adaptive landscape that result from having to deal with fitness trade-offs across different environmental conditions. Overall, our findings indicate that reconciling the short- and long-term evolutionary consequences of fluctuating selection could significantly improve our understanding of the evolution of specialization and generalism.

Keywords: adaptation, experimental evolution, microbes, trade-offs

When does adaptation in fluctuating environments lead to specialist or generalist outcomes? We evolved yeast populations at varying levels of two different environmental stresses and measured their resulting changes in fitness. We find that there is often but not always a cost associated with generalist outcomes.

Introduction

The ability to persist under a wide range of environmental conditions is presumably advantageous because it could dramatically multiply a lineage’s opportunities on Earth. What then prevents the widespread evolution of generalists? The most common explanation for the relative rarity of generalists is that niche breadth tends to trade off with peak fitness, resulting in a cost to generalization (Sexton et al., 2017). However, empirical studies have often failed to detect such trade-offs (Bono et al., 2017, 2020; Remold, 2012), have found it difficult to distinguish them from conditional neutrality (Bono et al., 2017), or have failed to detect the actual performance dimensions that trade off against each other (Sexton et al., 2017). Alternatively, the evolution of generalism could also be constrained by comparatively faster rates of evolution toward specialization (Fry, 1996; Whitlock, 1996).

Experimental evolution has long been used to investigate the evolutionary process, particularly as it relates to exposure to heterogeneous conditions. While these experiments generally indicate that constant environments tend to promote the evolution of specialists and variable ones tend to promote generalization, there have also been reports of populations that respond to constant but not variable treatments (Bono et al., 2017; Kassen, 2002). Additionally, the grain of environmental variation, the type of variation (spatial or temporal), and the complexity and predictability of the environment have all been shown to impact rates of adaptation and adaptive outcomes (Boyer et al., 2021; Dhar et al., 2013; Graham et al., 2014; Karve et al., 2016, 2018; Kassen, 2002; Manenti et al., 2015; Serra et al., 2020; Sorensen et al., 2019; Tarazona et al., 2017; White et al., 2020). Thus, meaningful experimental investigations of the potential costs of generalism will require a proper standardization of ecological contexts. Comparing the evolution of identical microbial lineages under constant and fluctuating selection is a promising way to achieve such standardization.

In this study, we examine yeast evolution in constant and temporally variable environments. Specifically, we evolve 448 genetically barcoded Saccharomyces cerevisae lineages in 14 treatments that vary in the concentration and dynamics of two chemical stresses: salt and copper (Figure 1). We find that adaptation to constant environments results in fitness trade-offs for salt but not copper treatments and that most populations that evolve in fluctuating treatments with either salt or copper trade, as expected, perform for breadth of adaptation. Additionally, we find that such trade-offs are not universal, as some fluctuating selection populations in our experiment were able to improve performance in both environments and even evolved higher fitness than specialists in a constant environment.

Figure 1.

Figure 1.

Open in a new tab

Overview of treatments and experimental design. (A) Seven different constant or fluctuating environments used for both copper and salt experimental evolution treatments. (B) Experimental design used to measure changes in fitness of the evolved populations. Each ancestral population was established with equal abundance of two barcoded strains and evolved for 50 days. Ancestral and evolved populations were pooled into five groups and the fitness of each barcoded strain was measured relative to a common reference after 2 days of growth for each treatment, stress condition, and replicate.

Methods

Strains, media, and culture methods

Barcoded yeast strains were constructed using two isogenic haploid derivatives of a strain collected from an oak tree in Pennsylvania (YPS163) (Sniegowski et al., 2002): YJF153 (MATa, HO::dsdAMX4) and YJF154 (MATalpha, HO::dsdAMX4) (X. C. Li & Fay, 2017). One hundred and thirteen diploid strains (Supplementary Table S1) were constructed such that each contained a unique 20 bp barcode-sequence flanking KAN, inserted in the HO locus (Fasanello et al., 2020). A single barcoded strain (d1H10) was selected from this set to serve as a reference based on prior work showing no fitness differences among the barcoded strains (Fasanello et al., 2020). The remaining 112 barcoded strains were subjected to 50 days of experimental evolution (see Experimental design).

Yeast were cultured in complete medium (CM; 20 g/l dextrose, 1.7 g/l yeast nitrogen base without amino acid and ammonium sulfate, 5.0 g/l ammonium sulfate, 1.3 g/l dropout mix complete without yeast nitrogen base) with or without additional chemical stress in 96-deep-well plates (2.2-ml poly-propylene plates, square well, v-conical bottoms; Abgene AB-0932) covered with rayon acrylate breathable membranes (Thermo Scientific, 1,257,605). Growth plates were incubated at 30 °C for 24 hr inside an incubator (VWR, Forced Air Incubator, basic, 120v, 7 cu. ft.), with agitation using a horizontal electromagnetic microplate shaker (Union Scientific LLC, 9779-TC). Saturated (stationary phase) 24-hr culture was diluted (1:1,000) into 0.5-ml fresh medium at the same time each day to initialize the next round of growth.

Experimental design

The experimental design included a 50-day experimental evolution with subsequent fitness quantification of ancestral (Day 0) and evolved (Day 50) yeast via competition-based fitness assay (Figure 1).

Experimental evolution

One hundred and twelve barcoded yeast strains were divided evenly among seven treatment variables for chemical stress concentration and temporal dynamics. With 16 replicate populations per treatment and two barcodes per population, there were 448 barcoded lineages in 224 isolated populations. Constant chemical stress treatments were evolved for 50 days in CM plus chemical stress at 0% (EH0, read as: Evolutionary History 0%), 40% (EH40), or 80% (EH80) of the lethal limit for unevolved yeast strains in our library. The ancestral lethal limit for salt was 20 g/l NaCl and for copper was 8 μm CuSO4. For these constant environments, chemical stress concentration did not change from transfer-to-transfer. Fluctuating treatments were evolved for 50 days in chemical stress that alternated daily between two concentrations: 0%–40% (EH0_40), 20%–60% (EH20_60), 40%–80% (EH40_80), or 0%–80% (EH0_80) of the ancestral limit. This design was copied to create four microplates, which were evolved in parallel for 50 days: two were exposed to salt stress and two were exposed to copper stress. Stress concentrations were selected such that they were comparable between chemical stressors and such that the 80% stress treatment reduced growth but did not result in extinction (from transfer-to-transfer) for an average ancestral strain. Samples were collected from the initial mixtures (starting material for plate copies, Day 0) and from the final overnight cultures (on Day 50). These samples served as the starting material for the Day-0 and Day-50 fitness assays, respectively.

Fitness assays

Sequencing-based competition assays, hereafter fitness assays, were subsequently conducted on Day-0 (ancestral) and Day-50 (evolved) yeast to assess fitness relative to the reference strain. Yeast lines from Day 0 and Day 50 of the experimental evolution were revived from stocks and mixed, separately, in equal proportions to create five pools. A single pool was created from each [evolutionary microplate] × [day] for a total of one Day-0 sample (the template for the evolution) and four Day-50 samples (two salts evolved and two copper evolved). Because barcodes were unique within but not between each microplate, four evolved pools were generated to ensure each pool had unique barcodes. The reference strain was then spiked into each pool at a high proportion (~70%). Pools were diluted into fresh medium and cultured for two rounds of growth to allow competition to occur. Fitness assays were conducted in CM with and without additional chemical stress. Day-0 and Day-50 fitness assays were run in quadruplicate and initial measures (barcode starting proportions) for each were taken in quintuplicate. Samples were collected from the initial mixtures (fitness assay starting material) and from the final cultures (ca. 20 generations later). From these data, the fitness of each barcoded line prior to evolution (Day-0 assays) and after evolution (Day-50 assays) was quantified and the resulting values were used to assess change in fitness for each line in each environment relative to the reference (see Fitness calculations).

Library construction and sequencing

DNA was isolated using a ZR Fungal/Bacterial DNA Kit (Zymo Research D6005) in individual 2.0-ml screw-cap tubes following the manufacturer’s instructions. Physical cell disruption by bead-beating was conducted in a mixer mill (Retsch, MM 300) at 30 Hz (1,800/min) for 10 min (1-min on, 1-min off, times 10 cycles). MoBY barcodes were then amplified with forward/reverse Ion Torrent adapters containing a 9–12 bp index for multiplex sequencing (Supplementary Table S2). PCR products for library construction were generated using 25 cycles and were subsequently quantified with a Qubit 3.0 Flourometer (ThermoFisher Scientific, Q33216) using the high-sensitivity assay kit (ThermoFisher Scientific, Q32851). Products were combined at equimolar concentrations and purified using a Zymo DNA Clean and Concentrator kit (Zymo Research D4014) to create a single multiplexed library for sequencing. Additional control samples were included in the library to track barcode cross-contamination as well as any contamination that may have occurred during sample processing. An aliquot of the library was sequenced using an Ion Torrent sequencer (Ion Proton System, Ion Torrent) at the Genomics Core Facility at Saint Louis University with a customized parameter to assess polyclonality after 31 bp (the start of the forward Ion Torrent adapter index sequence). A second aliquot was sequenced to augment read depth following a preliminary assessment of data quality.

Sequence data processing and calculations

Sequence data set

Sequence data in FASTQ format were parsed and demultiplexed using custom scripts in R. 96,807,316 reads were retained for analysis that perfectly matched a forward adapter index (9–12 bp), a reverse adapter index (9–12 bp), and a MoBY genetic barcode (20 bp) included in the full experimental design. Of these, 52,853,350 (54.6%) mapped to nonreference barcodes. The average number of reads per nonreference barcode per sample was 5,655 and the median value was 3,442 (Supplementary Figure S1). To avoid noise due to low counts entries with ≤20 reads were treated as missing data and removed prior to read summary reporting and downstream analyses.

Contamination rate

No instances of culture contamination were observed. Barcode cross-contamination rate was also measured and is defined as the total number of counts mapping to barcodes included in the full experimental design (library) but not expected to be present in that particular sample (given pair of forward/reverse Ion Torrent adapter indices) (Fasanello et al., 2020). The rate of barcode cross-contamination was tracked using samples seeded with a single pair of barcoded strains, such that the subsequent presence of other barcodes in these wells could be identified and quantified by sequencing. Barcode cross-contamination was low overall (0.269% ± 0.619%) and exhibited minor variation among sample sets (Day-0, 0.01% ± 0.007%; Day-50 salt, 0.030% ± 0.012%; Day-50 copper, 0.765% ± 0.990%).

Fitness calculations

The Malthusian fitness of focal barcoded line i, relative to the reference (d1H10), at experimental evolution generation gn, mi gn, was measured as,

mign=(lnCi48~-hoursCi0-hours−lnR48-hoursR0-hours)/20 (1)

where Ci and R refer to barcode counts for the focal barcode and reference barcode at fitness assay time 0-hours (initial mixtures) and time 48-hours (final overnight cultures), and 20 is the number of generations over 48 hours (two overnight cultures at 9.97 generations each—calculated from the number of doublings based on optical density data) (Chevin, 2011). We use the standard equation m = ln(w) to convert Malthusian fitness values to Wrightian fitness values (Orr, 2009; Wu et al., 2013). Hereafter, fitness, denoted by a w, will refer to Wrightian fitness. The change in fitness of line i between Day 0 and Day 50, Δwi, was therefore computed as,

Δwi=wiDay~-50−wiDay~-0 (2)

where wiDay~-0 and wiDay~-50 are the line’s fitness relative to the reference at Day-0 (before the experimental evolution) and Day-50 (end of the experimental evolution) as measured from Equation 1. We thus assume no frequency-dependent selection and that fitness relative to the reference is not influenced by other strains in the pool.

Statistical analysis

Analysis and visualization tools

R version 4.0.2 was used for all calculations, analyses, and figure generation (Core R Team, 2020). Data processing uses base-R functionality supplemented with methods from the plyr package (Wickham, 2009). Statistical models with linear mixed effects utilize the lme4 (Bates et al., 2014) and lmerTest packages (Kuznetsova et al., 2017). Power analyses were conducted using the pwr package (Champely et al., 2017). Figures and tables were generated with ggplot2 (Wickham, 2009) and sjPlot (Lüdecke et al., 2023); multipanel figures were built using methods from grid (Core R Team, 2020), gridExtra (Auguie & Antonov, 2017), and cowPlot (Wilke, 2020).

Power analysis

Power analyses were conducted using root mean squared error (RMSE) and population standard deviation (PSD). RMSE for fitness change was 2.419 and PSD was 0.009. Consequently, we have 80% power to discern fitness differences of 0.306% between treatments and 80% power to identify fitness deviations from zero of 0.634% (Supplementary Figure S2B). We have 80% power to detect a fitness change (increase or decrease) of 2.163% for any individual barcode (Supplementary Figure S2A).

Fitness change in 50 days of experimental evolution

Using the 80% power cutoff, we called individual barcoded yeast lines that increased (≥2.163%) or decreased (≤2.163%) in fitness over the 50-day experimental evolution. The effect of evolutionary treatment on fitness change was assessed using linear mixed-effects models with change in fitness as the response variable and treatment as the predictor variable. A random effect of line ID was placed on the model intercept (fitness change ~ treatment + (1|line ID) + 0; family = Gaussian). Separate models were run for each dataset by environment (salt lines in CM, salt lines in CM + NaCl, copper lines in CM, copper lines in CM + CuSO4). This set of linear mixed effects models was repeated without the intercept term to assess fitness differences among treatments (rather than versus zero fitness change).

Results

Fitness gains in 500 generations of experimental evolution

We expected yeast lineages to adapt to the chemical stresses to which they were exposed. Accordingly, lineages that experienced less chemical stress during experimental evolution tended to exhibit greater fitness gains in the 0% stress environment (Figure 2A and B), whereas those that experienced higher stress concentrations during evolution tended to exhibit greater fitness gains in media with 80% chemical stress (Figure 2C and D).

Figure 2.

Figure 2.

Open in a new tab

Fitness change in 0%, 80% chemical stress after 500-generation experimental evolution. (A) Fitness change in 0% stress for lineages from the salt dataset. (B) Fitness change in 0% stress for the copper data set. (C) Fitness change in 80% salt stress. (D) Fitness change in 80% copper stress. Lower-triangle insets illustrate fitness differences among treatments; significant associations have beta-values, nonsignificant associations in gray. Asterisks denote treatments significantly different from 0 (*p < .05, **p < .01, ***p < .001). Black open circles are treatment median fitness. Black closed circles are treatment mean fitness with standard error bars depicted.

Adaptation to a particular environmental condition is likely to impact fitness in other environmental conditions. These effects can be complementary, neutral, or antagonistic. We observed all three cases in our data. In the salt data set, adaptation to 0% chemical stress was associated with negative fitness change in 80% stress (Figure 2A and C—EH0) and adaptation to 80% stress was associated with negative fitness change in the 0% stress environment (Figure 2A and C—EH80, EH40_80, and EH0_80). In the copper data set, adaptation to 0% chemical stress frequently resulted in a fitness increase in 80% stress (Figure 2B and D—EH0), while adaptation to 80% stress had neither positive nor negative effects on fitness in the 0% chemical stress environment (Figure 2A and C—EH80, EH40_80, and EH0_80).

Genetic correlation in fitness

Selection under uniform environmental conditions favors individuals whose fitness is highest in that environment and should therefore result in the evolution of specialists (Kassen, 2002). Consequently, cross-environment genetic correlation in fitness should evolve to become negative if adaptive mutations perform better in their evolutionary environment than in other environments. Positive genetic correlations in fitness arise when mutations perform equally or better in an alternative environment than lineages that actually evolved in such an environment (Kassen, 2002). We tested for cross-environment genetic correlation in fitness by comparing the fitness of strains evolved under different experimental treatments in media with 0%, 40%, and 80% stress.

As in Kassen (2002), we identified cross-environment genetic correlations in fitness as the slope of the line connecting the fitness of the constant treatments for that pair of environmental conditions (Figure 3). Despite noteworthy differences in patterns of adaptation in the salt and copper data sets (Figure 2), patterns of cross-environment genetic correlation were qualitatively similar under both chemical stressors. Specifically, we observed negative cross-environment genetic correlations in fitness between 0% and 40% chemical stress (Figure 3A) and between 0% and 80% chemical stress for both stress treatments (Figure 3B). The slope of this relationship becomes more negative with increasing environmental dissimilarities (i.e., compare Figure 3A and B). In contrast, we found a positive cross-environment genetic correlation in fitness between the 40% stress and 80% stress environments for both salt and copper treatments (Figure 3C).

Figure 3.

Figure 3.

Open in a new tab

Cross-environment genetic correlation in fitness. Fitness change for each treatment is shown by the fitness change in the lower stress environment on the x-axis and fitness change in the higher stress environment on the y-axis with respect to the ancestor. Negative slopes indicate the presence of a negative cross-environment genetic correlation in fitness change. Positive slopes indicate the presence of a positive correlation. (A) EH40, EH0 in 0%, 40% stress; (B) EH0, EH80 in 0%, 80% stress; (C) EH40, EH80 in 40%, 80% stress. Salt and copper data are indicated by color and treatment indicated by endcaps. Dotted lines included for visual comparison of slopes.

Costs of adaptation in constant conditions

Negative cross-environment genetic correlation in fitness can evolve due to fitness trade-offs, in which adaptation to one environment has a fitness cost in others. Alternatively, negative cross-environment genetic correlation in fitness can arise in the absence of strict costs if direct responses to selection are larger than correlated responses in other environments, i.e., if each strain is more closely adapted to its home environment than to other conditions (Reboud & Bell, 1997).

To evaluate the prevalence of costs of adaptation in our data, we assessed whether strains that evolved under constant conditions exhibited fitness gains in their respective home environments and losses in other environments. When examining extremes at the treatment level, that is, treatments EH0 and EH80 under 0% and 80% chemical stress, we observed costs in the salt data set but not in the copper dataset (Figure 3). Results for individual lineages replicated these findings, indicating that costs were relatively common with salt (Figure 4A and C; Supplementary Table S3), but rare with copper (Figure 4D and F; Supplementary Table S3). Our results for the EH0 and EH80 treatments in 40% stress were intermediate (Figure 4A, C, D, and F; Supplementary Table S3). In fact, the majority of lineages from both the salt and copper EH40 treatments did not exhibit evidence of costs at 0% or 80% stress (Figure 4B and E; Supplementary Table S3).

Figure 4.

Figure 4.

Open in a new tab

Fitness relationships for individual barcoded lineages. Fitness change in CM plus 0%, 40%, and 80% chemical stress for treatments exposed to constant chemical stress at 0% (EH0), 40% (EH40), and 80% (EH80) the ancestral lethal limit. Each line depicts the fitness change in three assay environments for each barcoded yeast population. Top row shows data from the salt dataset: (A) EH0, (B) EH40, (C) EH80. Bottom row shows data from the copper data set: (D) EH0, (E) EH40, (F) EH80. One barcoded strain with low fitness and missing data removed from (A) and one strain with low fitness removed from (C) for visualization purposes only. Gray dotted lines indicate 2.16% change in fitness.

Costs of adaptation in fluctuating conditions

It is also possible that the cost of generalization lies not in the existence of evolutionary trade-offs between environments but rather in a slower rate of adaptation (Kassen, 2002; Whitlock, 1996). Under that scenario, generalists could theoretically improve fitness in both environments if given enough time. To evaluate whether trade-offs exist in this system, we assayed the fitness of lines selected in constant (EH0, EH80) and fluctuating conditions (EH0_40, EH20_60, EH0_80, EH40_80) in environments with 0% and 80% stress. Under a model with equal rates of evolution, the absence of a trade-off between breadth of adaptation and performance could be inferred from the observation that populations evolved under fluctuating selection lie on the line connecting the performance of populations evolved under the EH0 and EH80 treatments. Alternatively, if the breadth of adaptation increases at the expense of performance, then lineages evolved under fluctuating selection should fall below this line.

Costs are universal at the treatment level

All treatments with evolutionary histories in fluctuating environments (EH0_40, EH20_60, EH0_80, and EH40_80) generally traded performance for breadth of adaptation in the copper data, that is, the mean values for all treatments fell significantly below the no-cost line at a detection limit of 0.634% fitness change (Figure 5B and D; Supplementary Table S4, Methods). Although treatments EH0_40, EH20_60, and EH0_80 also exhibited this trade-off in the salt dataset, results were nonsignificant for the EH40_80 treatment (Figure 5A and C; Supplementary Table S5). The magnitude of realized cost (distance below the no-cost line) was negatively associated with the amount of chemical stress experienced during evolution in the salt data set (Figure 5A; Supplementary Table S5). The opposite pattern exists in the copper data, where cost was positively associated with the amount of chemical stress experienced (Figure 5B; Supplementary Table S4). Costs were also observed for the EH0_40 treatment relative to the constant EH0 and EH40 treatments for both chemicals (Supplementary Figure S3).

Figure 5.

Figure 5.

Open in a new tab

Lineages with evolutionary histories in fluctuating chemical stress environments do not always trade depth for breadth of adaptation. (A, B) Black circles are the treatment mean fitness for the constant chemical stress lineages EH0 (bottom, right) and EH80 (top, left). Orthogonal distance below- and to the left of the dot-dash line connecting the EH0 and EH80 indicates a fitness cost in the form of a trade-off of depth for breadth of adaptation. Lineages on the line pay no cost of generalization. Lineages orthogonally above- or to the right of the dot-dash line enjoy additional fitness benefits in the form of depth and breadth of adaptation. Gray dotted lines indicate 2.16% fitness difference from the no-cost dash dot-dash. (A, B) Cost of generalism for all fluctuating chemical stress treatments (EH0_40, EH20_60, EH0_80, EH40_80) relative to the constant 0% (EH0) and 80% chemical stress (EH80) treatments. (C, D) Corresponds to (A, B) data; density plots depict cost/benefit as negative/positive values based on the distance from the dot-dash line. (A, C) Data for the salt data set; (B, D) copper data set.

A diversity of outcomes exists within treatments

Coarse, treatment-level, results may suggest the general existence of a trade-off between breadth of adaptation and performance but fail to capture the full range of adaptive outcomes that are observable within treatments. For example, examination of fitness changes for individual lineages revealed a rich diversity of phenotypes within each treatment and uncovered broad overlap in fitness phenotypes among treatments (Figure 5, small open circles). In the salt data, 58/105 (55%) of populations across the EH0_40, EH20_60, EH0_80, and EH40_80 treatments exhibited significant costs at a detection limit of 2.163% fitness change (Methods). In the copper data set, the number was 46/88 (82%). In both data sets, we also found a subset of replicates that exhibit no evidence of a trade-off given that their performance was not significantly different from the no-cost line connecting the EH0 and EH80 treatments (i.e., 30/105 or 29% of replicates in the salt data set and 34/88 or 39% of replicates in the copper dataset). Furthermore, we also found that every treatment contained at least some lineages that fell significantly above the no-cost line, indicating that some lineages may have developed adaptations that allowed them to generally increase performance across the range of fluctuating conditions (i.e., 17/105 or 16% of replicates in the salt data set and 8/88 or 9% of replicates in the copper data set, Supplementary Table S6).

Discussion

Generalists are often depicted as “jacks of all trades [and] masters of none” (MacArthur, 1984) because exploiting a range of environmental conditions can often involve compromise. We examined the potential for antagonistic pleiotropy in yeast by exposing a variety of lineages to constant selection from either salt or copper and found that stress-adapted populations exhibit clear fitness trade-offs only in the salt context. Although the negative fitness correlations reported in Figure 3 may be caused by antagonistic pleiotropy, they could also arise from (cost-free) conditional mutations that are beneficial in one environment but neutral (Bono et al., 2017) or less effective (Reboud & Bell, 1997) in other environments. Thus, although uncertainty in the actual mechanism remains, our findings confirm that specialization to one context can impact performance in others.

The notion that effective exploitation of a given environmental condition can compromise performance in other contexts is further supported by our finding that lineages exposed to fluctuating selection evolved to be less fit in complete media and under chemical stress than lineages exclusively evolved under either of those conditions (Figure 5). The fact that such compromises are evident even under copper stress (a treatment that showed no evidence of fitness trade-offs) demonstrates that the costs of generalism in yeast are not exclusively related to antagonistic pleiotropy (or conditional performance). A likely alternative is that fluctuating selection lineages were unable to match the performance of constant selection lineages in our experiment because they exhibit slower rates of evolution (Fry, 1996; Kassen, 2002; Whitlock, 1996) and had simply not yet fully adapted. Regardless of the mechanism, our findings demonstrate that initial adaptation to fluctuating selection in yeast involves adaptive compromises that generally reduce performance across the range of environmental conditions that these organisms experience (a likely version of bet-hedging; Childs et al., 2010).

Our results are consistent with prior studies of adaptation in fluctuating environments, but the large number of replicates revealed both common patterns as well as rare trajectories. A prior review of 40 studies concluded that the evolution of trade-offs can be predicted by their ecological and evolutionary history and found that the absence of trade-offs is associated with temporal rather than spatial heterogeneity and with shorter rather than longer periods of adaptation (Bono et al., 2017). Thus, trade-offs may not initially be present, but as populations evolve closer to a Pareto front that constrains multi-environment optimization, antagonistic pleiotropy could become more important (Joshi & Thompson, 1995; Satterwhite & Cooper, 2015; Schick et al., 2015; Y. Li et al., 2019). Our findings relate to the initial stages of adaptation to fluctuating selection because our experiment was relatively short (500 generations). However, the large number of replicate lineages per treatment (up to 32) enabled us to not only capture costly outcomes but also rare cost-free trajectories (Bono et al., 2017; Remold, 2012). The rarity of no-cost generalists could be related to the fact that mutations that confer advantages in both environments are either rare themselves or have smaller individual effects (Orr, 1998). In either case, the competitive superiority of these mutants in both chemical contexts of our experiment suggests that no-cost generalists may be relatively common in natural populations of the brewer’s yeast, where both the genetic diversity and the time frame of evolution are much larger than in our experimental setup.

Trade-offs may also depend on the strength of selection and range of environmental fluctuations (Bennett et al., 1992; Boyer et al., 2021; Cisneros-Mayoral et al., 2022; Jerison et al., 2020; Morley & Turner, 2017; Travisano et al., 1995). Although we did not measure the strength of selection, we expect it to increase with stress and found larger responses and more frequent costs with higher concentrations of salt (Figures 2 and 4). The range of salt fluctuations showed mixed results. Negative fitness costs were more common in the more dissimilar fluctuating environment (EH0_80) relative to the more similar environments (EH0_40 and EH40_80) (Supplementary Table S3), consistent with EH_40 showing few costs in either 0%, 40%, or 80% stress (Figure 4). However, fluctuating treatments showed little or no differences in their average deviation from the no-cost line (Figure 5, Supplementary Tables S4 and S5) and no-cost generalist lineages were found in all of the fluctuating treatments. Together, our results provide further evidence that strong selection in a homogeneous environment can limit a population’s adaptive potential (Abdul-Rahman et al., 2021).

In conclusion, our findings demonstrate that initial adaptation to various regimes of fluctuating selection can have diverse evolutionary outcomes, as can be seen in the compiled literature (Bono et al., 2017). More importantly, in the context of oscillating conditions, these diverse outcomes include the evolution of generalists that can even outcompete specialists in their respective home environments. This latter finding is particularly significant because it suggests that fluctuating selection can facilitate the discovery of broadly adaptive traits that are effectively inaccessible under constant selection. One potential way in which these discoveries may be facilitated is by modifying the adaptive landscape in a way that more readily reveals an evolutionary path to those more favorable trait combinations. For example, geometric mean fitness under fluctuating selection could smooth or perhaps even remove fitness peaks or valleys that in constant selection backgrounds drive populations away from these broadly favorable trait combinations. Thus, our findings suggest overall that a better reconciliation of the short- and long-term evolutionary consequences of fluctuating selection may significantly improve our understanding of the evolution of specialization and generalism.

Supplementary Material

qrad055_suppl_Supplementary_Figures
qrad055_suppl_supplementary_figures.pdf (87KB, pdf)
qrad055_suppl_Supplementary_Tables
qrad055_suppl_supplementary_tables.xlsx (18.5KB, xlsx)

Contributor Information

Vincent J Fasanello, Department of Biology, Washington University, St. Louis, MO, United States.

Ping Liu, Department of Biology, Washington University, St. Louis, MO, United States; Department of Genetics, Washington University, St. Louis, MO, United States.

Justin C Fay, Department of Genetics, Washington University, St. Louis, MO, United States; Department of Biology, University of Rochester, Rochester, NY, United States.

Carlos A Botero, Department of Biology, Washington University, St. Louis, MO, United States; Department of Integrative Biology, University of Texas at Austin, Austin, TX, United States.

Data and code availability

Data formatted for analysis and custom R scripts utilized for all data processing, statistical analyses, and figure generation are available from GitHub (https://github.com/VinceFasanello/FS_Code_Supplement and https://zenodo.org/doi/10.5281/zenodo.10014955). Instructions to reproduce the analyses and to confirm the results presented in this article are provided. Yeast lines are available from the Fay Lab.

Author contributions

Conceptualization and writing: V.J.F., J.C.F., C.A.B. Supervision and project administration: J.C.F. and C.A.B. Methodology: V.J.F. and P.L. Software, validation, formal analysis, investigation, data curation, visualization: V.J.F.

Funding

This work was supported by a National Institutes of Health grant (GM080669) to J.C.F., a National Science Foundation Graduate Research Fellowship Program (DGE1745038) to V.J.F., and funds from Washington University to C.A.B.

Conflict of interest: The authors declare no conflict of interest.

References

  1. Abdul-Rahman, F., Tranchina, D., & Gresham, D. (2021). Fluctuating environments maintain genetic diversity through neutral fitness effects and balancing selection. Molecular Biology and Evolution, 38(10), 4362–4375. 10.1093/molbev/msab173 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Auguie, B., & Antonov, A. (2017). gridExtra: Miscellaneous Functions for “Grid” Graphics (2.3) [Computer software]. https://cran.r-project.org/web/packages/gridExtra/index.html [Google Scholar]
  3. Bates, D., Mächler, M., Bolker, B., & Walker, S. (2014). Fitting linear mixed-effects models using lme4. arXiv. 10.48550/arXiv.1406.5823 [DOI]
  4. Bennett, A. F., Lenski, R. E., & Mittler, J. E. (1992). Evolutionary adaptation to temperature. I. Fitness responses of Escherichia coli to changes in its thermal environment. Evolution, 46(1), 16–30. 10.1111/j.1558-5646.1992.tb01981.x [DOI] [PubMed] [Google Scholar]
  5. Bono, L. M., Draghi, J. A., & Turner, P. E. (2020). Evolvability costs of niche expansion. Trends in genetics, 36(1), 14–23. 10.1016/j.tig.2019.10.003 [DOI] [PubMed] [Google Scholar]
  6. Bono, L. M., SmithJr, L. B., Pfennig, D. W., & Burch, C. L. (2017). The emergence of performance trade-offs during local adaptation: Insights from experimental evolution. Molecular Ecology, 26(7), 1720–1733. 10.1111/mec.13979 [DOI] [PubMed] [Google Scholar]
  7. Boyer, S., Hérissant, L., & Sherlock, G. (2021). Adaptation is influenced by the complexity of environmental change during evolution in a dynamic environment. PLoS Genetics, 17(1), e1009314. 10.1371/journal.pgen.1009314 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Champely, S., Ekstrom, C., Dalgaard, P., Gill, J., Weibelzahl, S., Anandkumar, A., Ford, C., Volcic, R., & De Rosario, H. (2017). pwr: Basic functions for power analysis (1.3) [Computer software]. https://CRAN.R-project.org/package=pwr
  9. Chevin, L. -M. (2011). On measuring selection in experimental evolution. Biology Letters, 7(2), 210–213. 10.1098/rsbl.2010.0580 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Childs, D. Z., Metcalf, C. J. E., & Rees, M. (2010). Evolutionary bet-hedging in the real world: Empirical evidence and challenges revealed by plants. Proceedings of the Royal Sciences B: Biological Sciences, 277(1697), 3055–3064. 10.1098/rspb.2010.0707 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Cisneros-Mayoral, S., Graña-Miraglia, L., Pérez-Morales, D., Peña-Miller, R., & Fuentes-Hernández, A. (2022). Evolutionary history and strength of selection determine the rate of antibiotic resistance adaptation. Molecular Biology and Evolution, 39(9), msac185. 10.1093/molbev/msac185 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Core R Team. (2020). R: A language and environment for statistical computing. http://www.r-project.org [Google Scholar]
  13. Dhar, R., Sägesser, R., Weikert, C., & Wagner, A. (2013). Yeast adapts to a changing stressful environment by evolving cross-protection and anticipatory gene regulation. Molecular Biology and Evolution, 30(3), 573–588. 10.1093/molbev/mss253 [DOI] [PubMed] [Google Scholar]
  14. Fasanello, V. J., Liu, P., Botero, C. A., & Fay, J. C. (2020). High-throughput analysis of adaptation using barcoded strains of Saccharomyces cerevisiae. PeerJ, 8, e10118. 10.7717/peerj.10118 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Fry, J. D. (1996). The evolution of host specialization: Are trade-offs overrated? The American Naturalist, 148, S84–S107. 10.1086/285904 [DOI] [Google Scholar]
  16. Graham, J. K., Smith, M. L., & Simons, A. M. (2014). Experimental evolution of bet hedging under manipulated environmental uncertainty in Neurospora crassa. Proceedings of the Royal Sciences B: Biological Sciences, 281(1787), 20140706. 10.1098/rspb.2014.0706 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Jerison, E. R., Nguyen Ba, A. N., Desai, M. M., & Kryazhimskiy, S. (2020). Chance and necessity in the pleiotropic consequences of adaptation for budding yeast. Nature Ecology & Evolution, 4(4), 601–611. 10.1038/s41559-020-1128-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Joshi, A., & Thompson, J. N. (1995). Trade-offs and the evolution of host specialization. Evolutionary Ecology, 9(1), 82–92. 10.1007/bf01237699 [DOI] [Google Scholar]
  19. Karve, S. M., Bhave, D., & Dey, S. (2018). Extent of adaptation is not limited by unpredictability of the environment in laboratory populations of Escherichia coli. Journal of Evolutionary Biology, 31(9), 1420–1426. 10.1111/jeb.13338 [DOI] [PubMed] [Google Scholar]
  20. Karve, S. M., Tiwary, K., Selveshwari, S., & Dey, S. (2016). Environmental fluctuations do not select for increased variation or population-based resistance in Escherichia coli. Journal of Biosciences, 41(1), 39–49. 10.1007/s12038-016-9592-2 [DOI] [PubMed] [Google Scholar]
  21. Kassen, R. (2002). The experimental evolution of specialists, generalists, and the maintenance of diversity. Journal of Evolutionary Biology, 15(2), 173–190. 10.1046/j.1420-9101.2002.00377.x [DOI] [Google Scholar]
  22. Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest Package: Tests in linear mixed effects models. Journal of Statistical Software, 82(13), 1–26. 10.18637/jss.v082.i13 [DOI] [Google Scholar]
  23. Li, X. C., & Fay, J. C. (2017). Cis-regulatory divergence in gene expression between two thermally divergent yeast species. Genome Biology and Evolution, 9(5), 1120–1129. 10.1093/gbe/evx072 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Li, Y., Petrov, D. A., & Sherlock, G. (2019). Single nucleotide mapping of trait space reveals Pareto fronts that constrain adaptation. Nature Ecology & Evolution, 3(11), 1539–1551. 10.1038/s41559-019-0993-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Lüdecke, D., Bartel, A., Schwemmer, C., Powell, C., Djalovski, A., & Titz, J. (2023). sjPlot: Data Visualization for Statistics in Social Science (2.8.14) [Computer software]. https://cran.r-project.org/web/packages/sjPlot/index.html
  26. MacArthur, R. H. (1984). Geographical ecology: Patterns in the distribution of species. Princeton University Press, Princeton, New Jersey. [Google Scholar]
  27. Manenti, T., Loeschcke, V., Moghadam, N. N., & Sørensen, J. G. (2015). Phenotypic plasticity is not affected by experimental evolution in constant, predictable or unpredictable fluctuating thermal environments. Journal of Evolutionary Biology, 28(11), 2078–2087. 10.1111/jeb.12735 [DOI] [PubMed] [Google Scholar]
  28. Morley, V. J., & Turner, P. E. (2017). Dynamics of molecular evolution in RNA virus populations depend on sudden versus gradual environmental change. Evolution, 71(4), 872–883. 10.1111/evo.13193 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Orr, H. A. (1998). The population genetics of adaptation: The distribution of factors fixed during adaptive evolution. Evolution, 52(4), 935–949. 10.1111/j.1558-5646.1998.tb01823.x [DOI] [PubMed] [Google Scholar]
  30. Orr, H. A. (2009). Fitness and its role in evolutionary genetics. Nature Reviews. Genetics, 10(8), 531–539. 10.1038/nrg2603 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Reboud, X., & Bell, G. (1997). Experimental evolution in Chlamydomonas III Evolution of specialist and generalist types in environments that vary in space and time. Heredity, 78(5), 507–514. 10.1038/sj.hdy.6881210 [DOI] [Google Scholar]
  32. Remold, S. (2012). Understanding specialism when the Jack of all trades can be the master of all. Proceedings of the Royal Sciences B: Biological Sciences, 279(1749), 4861–4869. 10.1098/rspb.2012.1990 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Satterwhite, R. S., & Cooper, T. F. (2015). Constraints on adaptation of Escherichia coli to mixed-resource environments increase over time. Evolution, 69(8), 2067–2078. 10.1111/evo.12710 [DOI] [PubMed] [Google Scholar]
  34. Schick, A., Bailey, S. F., & Kassen, R. (2015). Evolution of fitness trade-offs in locally adapted populations of Pseudomonas fluorescens. The American Naturalist, 186(Suppl 1), S48–S59. 10.1086/682932 [DOI] [PubMed] [Google Scholar]
  35. Serra, M., Carmona, M. J., García-Roger, E. M., & Ortells, R. (2020). Surviving uncertainty: Biodiversity, adaptation, and environmental fluctuation in rotifers. Metode Science Studies Journal, 0(10), 183–191. 10.7203/metode.10.14398 [DOI] [Google Scholar]
  36. Sexton, J. P., Montiel, J., Shay, J. E., Stephens, M. R., & Slatyer, R. A. (2017). Evolution of ecological niche breadth. Annual Review of Ecology, Evolution, and Systematics, 48(1), 183–206. 10.1146/annurev-ecolsys-110316-023003 [DOI] [Google Scholar]
  37. Sniegowski, P. D., Dombrowski, P. G., & Fingerman, E. (2002). Saccharomyces cerevisiae and Saccharomyces paradoxus coexist in a natural woodland site in North America and display different levels of reproductive isolation from European conspecifics. FEMS Yeast Research, 1(4), 299–306. 10.1111/j.1567-1364.2002.tb00048.x [DOI] [PubMed] [Google Scholar]
  38. Sorensen, J. W., Dunivin, T. K., Tobin, T. C., & Shade, A. (2019). Ecological selection for small microbial genomes along a temperate-to-thermal soil gradient. Nature Microbiology, 4(1), 55–61. 10.1038/s41564-018-0276-6 [DOI] [PubMed] [Google Scholar]
  39. Tarazona, E., García-Roger, E. M., & Carmona, M. J. (2017). Experimental evolution of bet hedging in rotifer diapause traits as a response to environmental unpredictability. Oikos, 126(8), 1162–1172. 10.1111/oik.04186 [DOI] [Google Scholar]
  40. Travisano, M., Vasi, F., & Lenski, R. E. (1995). Long-term experimental evolution in Escherichia coli III variation among replicate populations in correlated responses to novel environments. Evolution, 49(1), 189–200. 10.1111/j.1558-5646.1995.tb05970.x [DOI] [PubMed] [Google Scholar]
  41. White, P. S., Choi, A., Pandey, R., Menezes, A., Penley, M., Gibson, A. K., de Roode, J., & Morran, L. (2020). Host heterogeneity mitigates virulence evolution. Biology Letters, 16(1), 0744. 10.1098/rsbl.2019.0744 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Whitlock, M. C. (1996). The red queen beats the jack-of-all-trades: The limitations on the evolution of phenotypic plasticity and niche breadth. The American Naturalist, 148, S65–S77. 10.1086/285902 [DOI] [Google Scholar]
  43. Wickham, H. (2009). ggplot2: Elegant Graphics for Data Analysis. Springer. 10.1007/978-0-387-98141-3 [DOI] [Google Scholar]
  44. Wilke, C. O. (2020). cowplot: Streamlined Plot Theme and Plot Annotations for “ggplot2” (1.1.1) [Computer software]. https://cran.r-project.org/web/packages/cowplot/index.html
  45. Wu, B., Gokhale, C. S., van Veelen, M., Wang, L., & Traulsen, A. (2013). Interpretations arising from Wrightian and Malthusian fitness under strong frequency dependent selection. Ecology and Evolution, 3(5), 1276–1280. 10.1002/ece3.500 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

qrad055_suppl_Supplementary_Figures
qrad055_suppl_supplementary_figures.pdf (87KB, pdf)
qrad055_suppl_Supplementary_Tables
qrad055_suppl_supplementary_tables.xlsx (18.5KB, xlsx)

Data Availability Statement

Data formatted for analysis and custom R scripts utilized for all data processing, statistical analyses, and figure generation are available from GitHub (https://github.com/VinceFasanello/FS_Code_Supplement and https://zenodo.org/doi/10.5281/zenodo.10014955). Instructions to reproduce the analyses and to confirm the results presented in this article are provided. Yeast lines are available from the Fay Lab.

Articles from Evolution Letters are provided here courtesy of Oxford University Press

RESOURCES