Experimental evolution of evolutionary potential in fluctuating environments

Variation is the raw material for evolution. Evolutionary potential is determined by the amount of genetic variation, but evolution can also alter the visibility of genetic variation to natural selection. Fluctuating environments are suggested to maintain genetic variation but they can also affect environmental variance, and thus, the visibility of genetic variation to natural selection. However, experimental studies testing these ideas are relatively scarce. In order to determine differences in evolutionary potential we quantified variance attributable to population, genotype and environment for populations of the bacterium Serratia marcescens. These populations had been experimentally evolved in constant and two fluctuating environments. We found that strains that evolved in fluctuating environments exhibited larger environmental variation suggesting that adaptation to fluctuations has decreased the visibility of genetic variation to selection.


| INTRODUC TI ON
Variation is the raw material for selection, a driver of evolution and source for economic improvements in animal and plant breeding (Lynch & Walsh, 1998). High heritability, that is, a large proportion of additive genetic variation relative to total variation, guarantees fast evolutionary changes, but at the same time, such variation should be rapidly depleted when alleles are brought to fixation (Johnson & Barton, 2005). The maintenance of genetic variation in populations has long been an enigma, as empirical estimates show that there is more variation in populations that what is predicted by theory (Johnson & Barton, 2005). It has been suggested that fluctuating environments can maintain genetic variation (Bürger & Gimelfarb, 2002), but the evidence for this has been not been conclusive. Although high genetic variation is considered advantageous for species facing changing environments by enabling rapid response to altered selection pressure, in some cases slower evolutionary changes are more beneficial (Kawecki, 2000). Sometimes it could be better not to evolve at all. Lack of evolutionary change or a slower evolutionary response could be beneficial if the environment varies unpredictably, as a strong response to selection could be maladaptive in the next generation. This could lead to canalization and hiding genetic variation from selection (Kawecki, 2000). Unpredictable environments can also lead to the evolution of bet-hedging, where genotypes produce variable phenotypes that could secure their survival in uncertain conditions (Botero et al., 2015;Kronholm, 2022;Levins, 1968).
In essence, both cases are supposed to lead to lowered heritability. However, the hallmark of the latter is increased environmental variation. This could occur either through classical bet-hedging or if

Abstract
Variation is the raw material for evolution. Evolutionary potential is determined by the amount of genetic variation, but evolution can also alter the visibility of genetic variation to natural selection. Fluctuating environments are suggested to maintain genetic variation but they can also affect environmental variance, and thus, the visibility of genetic variation to natural selection. However, experimental studies testing these ideas are relatively scarce. In order to determine differences in evolutionary potential we quantified variance attributable to population, genotype and environment for populations of the bacterium Serratia marcescens. These populations had been experimentally evolved in constant and two fluctuating environments. We found that strains that evolved in fluctuating environments exhibited larger environmental variation suggesting that adaptation to fluctuations has decreased the visibility of genetic variation to selection.

K E Y W O R D S
bet-hedging, experimental evolution, fluctuating environments, genetic variation adaptation via inducible plasticity entails increased developmental instability, seen as random deviations in phenotype, as suggested by Tonsor et al. (2013). Inducible plasticity has been suggested to be theoretically very tractable adaptation to fluctuating environments (Botero et al., 2015;Kronholm, 2022;Levins, 1968). Whether evolution is slowed down due to low additive genetic variation, or due to high environmental variation in different environments and traits has gained interest from researchers over the years (Hoffmann & Merilä, 1999;Schou et al., 2020;Wilson et al., 2006).
We tested these ideas using strains of the bacterium Serratia marcescens resulting from a replicated (n = 10 populations in each treatment) evolution experiment, where we propagated populations for approximately 140 generations in stable (31°C) and in rapidly cycling thermal environments with similar mean temperature. We considered two kinds of thermal fluctuations: abrupt fluctuations generated by cycling between 24 and 38°C (45 min in each temperature), or smooth fluctuations generated by cycling sequentially from 24-31-38-31-24°C (30 min in each temperature). Thus, the two different fluctuations had different rates of temperature changes. After the experiment we measured the maximal growth rate and maximal population sizes of bacterial clones (12 per population) four times at 31°C in order to quantify, using quantitative genetic methods of clonal analysis (Lynch & Walsh, 1998), if the evolution in fluctuating environments led to increased environmental variability in strains than in those evolving in constant environments.

| MATERIAL S AND ME THODS
The strains used in this study were obtained from a previous study (Ketola & Saarinen, 2015). Briefly: an evolution experiment using the bacterium Serratia marcescens strain DB 11 was initiated from an overnight culture of single clone in DM25 medium. Experimental microcosms were wells of Bioscreen C 100 well plates. The ancestor clone was seeded to three plates, with 10 populations on each plate, that were assigned to three different thermal conditions: The experiment was continued for 27 renewals, that is, 54 days and approximately 140 generations.
After the evolution experiment, we extracted 12 clones from each population. These clones were frozen 1:1 in 80% glycerol in a randomized order in four bioscreen 100 well plates and stored at −80°C. Growth measurements were initiated by cryo-replicating frozen clones to fresh media (Duetz et al., 2000). After 24 h the clones were replicated to new plates containing DM25 and the plates were placed in Bioscreen spectrophotometers at 31°C for 3-4 days, that is, until growth in all wells had stopped. We repeated growth measurements 4 times at 31°C. Maximal growth rate and yield were obtained from growth data (measured every 5 min, 420-580 nm) using a custom R script described previously (Ketola et al., 2013). In short, ln-transformed growth data were analysed in a 25 time point sliding window excluding data from the first five timepoints (due to vapour).
To find the maximal growth rate, we fitted linear regression to the data of each 25 timepoint sliding window. The maximum slope of the regression equals the maximal growth rate. Maximum biomass yield was obtained by finding the window with the largest mean optical density over the 25 timepoints.
Maximal growth rate and yield were subjected to statistical analyses. Determination of the instability of the genotypes across repeated measurements at a constant 31°C required analysing the evolutionary treatments separately. Prior to analysis, wells, where no growth had occurred, were removed from the data. This left n = 479, n = 480 and n = 480 observations for the constant, smooth and abrupt treatments, respectively. Traits and inoculum size were standardized to a mean of zero. This transformation was made for each of the evolutionary treatments separately to allow testing differences in variance without an effect of trait mean on the variance. We extracted variance components arising from population, clone and replicate using a Bayesian multilevel linear model. The model was: where α is the intercept, α p is the population effect, α c is the clone effect, β m is the effect of inoculum size, σ p , σ c , and σ e are the standard deviations for population, clone and environmental effects respectively.
The priors were weakly informative, with a normal distribution with a mean of zero and standard deviation of five for the intercept and inoculum size effect, and an exponential prior with rate 1 for population, environmental and clonal standard deviations. The model was fit using Hamiltonian Monte Carlo implemented via the Stan language and the R package brms (Bürkner, 2017). (1) (1) Variation stemming from the clone effect quantifies the amount of genetic variation within the populations (Lynch & Walsh, 1998) and population effect variation quantifies variation across the populations indicating the propensity of populations to reach different evolutionary outcomes within evolutionary treatments or uniformity of selection pressures across replicated populations (Ketola et al., 2013;Travisano et al., 1995).

| RE SULTS
Total variation in both traits (maximal growth rate and yield) was comparable between the evolutionary treatments, as the highest posterior density intervals for the difference in total variances between treatments were not different from zero (Figure 1).
For maximal growth rate, we did not observe differences between the clonal, or population variances (Figure 2a). Clonal variance was slightly higher in the constant treatment but the 95% highest posterior density interval for differences overlapped with zero ( Figure 2a). The fluctuating treatments had somewhat higher environmental variances than the constant treatment, but only the difference between abrupt and constant treatment was different from zero ( Figure 2a).
In contrast to maximal growth rate, results for yield showed larger differences and higher precision of the estimates (Figure 2b). There were no differences in clonal or population variance (Figure 2b).
However, the two fluctuating treatments had higher environmental variance than the constant treatment ( Figure 2b). Environmental variance was 1.9 [1.5, 2.3] times higher in the smooth treatment than in the constant treatment, and 2.2 [1.8, 2.7] times higher in the abrupt treatment than in the constant treatment.
Replicates of the ancestral clone were stored in random locations in the cryoplates, which allowed the estimation of environmental variance also for the ancestor (n = 63

| DISCUSS ION
After experimental evolution, during which Serratia marcescens DB 11 was exposed to constant and fluctuating thermal environments, we found changes in evolutionary potential. In comparison to evolving in a constant environment, in fluctuating environments the environmental variation was higher, suggesting that fluctuating environment hindered the ability of S. marcescens to evolve. Total variation was comparable between the experimental strains, indicating that the observed changes were not due to changes in total variation (Figure 1).
Environmental fluctuations have been suggested to maintain a higher amount of genetic variation. However, the evidence is mixed (Kassen, 2002). In our experiment genetic variation between strains did not differ in the different evolutionary treatments. In principle, higher variation in fluctuating environments could be due to maintained polymorphism of temperature specialists (Levins, 1968).
Bet-hedging is known also in bacteria, for example, in propensity to alter between morphotypes, and it plays important role in bacterial fitness and growth characteristics (Beaumont et al., 2009).
Morphotype variation and associated growth variation are also important in natural conditions (Kunttu et al., 2009;Pulkkinen et al., 2022;Sundberg et al., 2014). Therefore, the repeatability of growth patterns, quantified as an environmental variation, can also be indicative of bet-hedging or developmental instability and could be followed when more easily recordable traits, such as morphotypes, do not exist. We observed that environmental variation was higher in strains that evolved in fluctuating environments than in strains that evolved in a constant environment. Interestingly, the ancestor had even higher environmental variation, which indicates that laboratory conditions had generally decreased the environmental variation, especially so in constant temperature. This finding could be indicative of diversity and complexity of the environment that this environmentally growing opportunistic pathogen (Flyg et al., 1980) might often encounter (see below). As measurements were conducted using plates with a balanced number of clones from all treatments in randomized order, treated similarly and measured simultaneously, we can safely exclude technical effects as an explanation.
In addition to bet-hedging, our result could be explained by the evolution of phenotypic plasticity and instability of genotypes of highly plastic individuals (Tonsor et al., 2013). Improved tolerance of thermal transitions could be obtained by inducible phenotypic plasticity but, in order to mount a plastic response at the correct time, individuals require a cue about the upcoming environment. Picking up the cue and responding strongly to it is obviously beneficial if, indeed, the environments fluctuate, and the cue is reliable. However, F I G U R E 1 Comparison of total variation difference between evolved strains for biomass yield (K) and maximal growth rate (r). Difference environments also contain, 'noise', that could act as a false cue for a response, unavoidably resulting in a suboptimal phenotype also in apparently constant environments. This instability can also effectively act against the evolution of inducible phenotypic plasticity (Tonsor et al., 2013).
Previous experimental evolution work conducted at very rapidly fluctuating environments suggests that tolerance to fast fluctuations might be due to the evolution of inducible plasticity (Ketola et al., 2004;Ketola & Saarinen, 2015;Saarinen et al., 2018). Our findings here support the earlier work and suggest potential consequences: higher environmental sensitivity, that is, environmental variation. However, regardless of the exact mechanism (bet-hedging or a side effect of phenotypic plasticity), our data show important an result: that evolution alters evolutionary potential by altering visibility of genetic variation and the population's potential for responding to new conditions.

ACK N O WLE D G E M ENTS
We thank Kati Saarinen for conducting the measurements, Kalevi Viipale for insightful discussions and Academy of Finland for funding (TK: 278751, IK: 321584).

CO N FLI C T O F I NTE R E S T S TATE M E NT
Authors declare no conflict of interest.

PE E R R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1111/ jeb.14178.