Rapid adaptation to high temperatures in Chironomus riparius

Abstract Effects of seasonal or daily temperature variation on fitness and physiology of ectothermic organisms and their ways to cope with such variations have been widely studied. However, the way multivoltines organisms cope with temperature variations from one generation to the next is still not well understood. The aim of this study was to investigate whether the multivoltine midge Chironomus riparius Meigen (1803) responds mainly via acclimation as predicted by current theories or whether rapid genetic adaptation is involved. To investigate this issue, a common garden approach has been applied. A mix of larvae from five European populations was raised in the laboratory at three different pre‐exposure temperatures (PET): 14, 20, and 26°C. After three and five generations, respectively, larvae were exposed to three treatment temperatures (TT): 14, 20, and 26°C. Mortality was monitored for the first 48 hr and after emergence. After three generations, significant mortality rate differences depended on an interaction of PET and TT. This finding supports the hypothesis that chironomids respond rapidly to climatic variation via adaptive mechanisms and to a lesser extent via phenotypic plasticity. The result of the experiment indicates that three generations were sufficient to adapt to warm temperature, decreasing the mortality rate, highlighting the potential for chironomids to rapidly respond to seasonally changing conditions.

Even though the effects of short-term temperature variations on physiology and fitness have been well studied, the processes governing responses to short-term seasonal temperature variations on ectothermic organisms are still not fully understood (Paaijmans et al., 2013). This is in particular true for short-lived, multivoltine species whose successive generations are exposed to highly different temperature regimes throughout the year, at least in temperate regions with their pronounced seasonality.
To cope with such a broad temperature variation, likely exceeding their optimal temperature range at least in the extremes, multivoltine ectothermic organisms can respond in different ways, such as (a) behavioral changes, allowing organisms to escape or mitigate the environmental pressure (Hutchison & Maness, 1979;Lencioni, 2004); (b) phenotypic plasticity, resulting in changes in gene expression to alleviate abiotic stress (Johnston & Wilson, 2006); and/or (c) by genetic adaptation (Bergland, Behrman, O'Brien, Schmidt, & Petrov, 2014) to maximize their fitness. It is commonly assumed that (a) and (b) act rather on short timescales, while (c) was presumed to proceed over long timescales (Carroll, Hendry, Reznick, & Fox, 2007). However, it becomes increasingly clear that evolutionary adaptation can be very rapid (Messer & Petrov, 2013) and can even occur on seasonal timescales (Bergland et al., 2014).
Because both phenotypic plasticity and rapid adaptation are necessarily associated with changes in the phenotype, distinguishing whether a change in phenotype is genetically based or results only from phenotypic plasticity is difficult. It requires experimental testing over several generations in order to identify which of those mechanisms are involved. To this end, we performed a common garden experiment including different thermal environments with transplantation experiments after three and five generations of exposure to fixed temperature. This method allows us to compare the mortality of the individuals depending on which temperature the parents come from and the temperature experienced by the offspring (Gienapp, Teplitsky, Alho, Mills, & Merilä, 2008;Merilä & Hendry, 2014). This procedure also permits us to identify possible transgenerational effects by comparing the mortality at the temperature the parents were raised (Fox & Mousseau, 1998;Valtonen, Kangassalo, Pölkki, & Rantala, 2012), as well as possible hardening mechanism by recording the mortality 24 hr after hatching (Bowler, 2005;Sinclair & Roberts, 2005). If the response to different temperatures is governed by phenotypic plasticity, we expect the phenotype and the resulting fitness of each generation to depend only on the temperature experienced by the according generation, independent of the temperature experienced by previous generations. Thus, irrespective of generation, individuals should perform similarly well in each temperature (i.e., according to the respective reaction norm for this temperature). If, however, genetic adaption occurred, the phenotypic fitness response to different temperatures will rather depend on the temperature experienced by the previous generation(s), that is, organisms should perform better in temperatures their progenitors were exposed to. Consequently, we expect an interaction between temperatures experienced by previous generations and temperatures experienced by the test individuals. Therefore, this procedure also allows us to distinguish between adaptation and transgenerational effect such as parental effect (Burgess & Marshall, 2014). The aim of this study was to infer how an ectothermic multivoltine invertebrate, Chironomus riparius (MEIGEN 1803) (Supporting Information Figure S1), spending most of its life in an aquatic larval stage (Armitage, Cranston, & Pinder, 1995), copes with the short-term seasonal temperature variations over a few generations-via phenotypic plasticity or rapid adaptation. The species fares very well with experimental culture conditions over several generations (Downe & Caspary, 1973;Nowak, Jost, et al., 2007;OECD, 2004;2010) and is therefore well suited for evolutionary experiments. Moreover, the adaptive potential of this species to a large gradient of local climate conditions has already been shown (Nemec, Patel, Nowak, & Pfenninger, 2013;Waldvogel et al., 2018).
Wild populations of C. riparius harbors ample genetic variation and have a large effective population size (~10 6 , (Oppold & Pfenninger, 2017)). In addition, the high number of offspring per breeding pair likely render selection processes on quantitative traits very effective (Pfenninger, 2017). In this study, we hypothesize therefore that C. riparius can genetically adapt rapidly within a few generations to different temperature regimes, which would give first experimental evidence for rapid adaptation to seasonal temperature fluctuations. We decided to mix individuals from multiple populations in order to compensate for the expected loss of variation due to laboratory selection effect during the experiment  and to make sure that alleles involved in adaptation to the complete temperature gradient are present in the gene pool. This setup nevertheless is also raising the problem of possible assortative mating since individuals from different population could prefer individuals from the same population. This possible bias is regarded as negligible after three generations considering the size of the cage forcing reproduction between individuals from different populations. From this admixed base population, which was kept for one generation at 20°C, three subpopulations were created.

| MATERIAL S AND ME THODS
Two replicates from each subpopulation were raised at three different temperatures: 14, 20, and 26°C with light-dark rhythm of 16:8 hr at 60% humidity, following the OECD guideline 219 (OECD, 2004;Oppold et al., 2016). These temperatures were oriented at the maximum, average, and minimum mean temperature recorded during the meteorological summer (26.9°, 20.9°, and 15.3°C) for the five populations of origin (Waldvogel et al., 2018). Those values were adjusted to obtain constant 6°C differences among experimental treatments. Field measured temperatures have been used to avoid reaching critical temperatures that are not commonly experienced in the wild, avoiding possible bias on the recorded mortality by artificially causing bottlenecks in the population (Supporting Information Table S1).
Larvae were raised in medium constituted of purified water with sea salt (Tropic Marin) adjusted to a conductivity of 520-540 µS/cm and pH 8. The bottom of the glass bowl (20 cm diameter) was covered with washed sand. Populations were raised at these temperatures for three and five generations to simulate the possible seasonal exposure range (i.e., number of generations which could be expected during meteorological summer); this phase will hereafter be referred to as "Pre-Exposure Temperature" (PET).

| Survival (mortality) test
Survival tests were common garden experiments performed in the third and fifth generation, that is, individuals were exposed to PET for three and five generations, respectively. These two different time  Individuals coming from the same egg clutch will be referred to as families. Eighteen larvae from each family of each replicate were raised at three different experimental temperatures called "Treatment Temperature" (TT) for the analysis (18 larvae × 5 families × 3 temperatures × 2 population replicates) (Figure 2). Each larva was individually raised in six-well plates (Ø3.5 × 2 cm) filled with 6 ml of medium for 48 hr without feeding to avoid altering the medium. Larval mortality rates were measured first after 24 and then 48 hr. After 48 hr, the surviving larvae were pooled by families in glass bowls (Ø20 × 10 cm) with sediment and medium and reared until emergence. The mortality of pupation and emergence was calculated as the number of individuals not emerged per family, for each combination of PET and TT, when adults were removed from bowls. During this stage, larvae were fed daily with dried fish food (0.5 mg/individual of grounded TetraMin ® flakes) and the water level adjusted daily with deionized water in order to conserve the physicochemical parameters.

| Statistics
Statistical analyses were performed using R (Version 3. F I G U R E 2 Schema of the experimental design used to record mortality rate depending on temperature using a common garden approach In case of significant interactions of two or more factors, each instance of the interaction was analyzed separately. In order to investigate significant differences between the TTs, ANOVA followed by Tukey post hoc tests was used for data following the assumptions of normality and homoscedasticity. Kruskal-Wallis tests followed by Dunn post hoc tests were used if one or both of the previous assumptions were rejected.

| Larval mortality
Larval mortality was significantly affected by pre-exposure temperature (PET) and test temperature (TT), with a significant interaction between these two factors. However, larval mortality did not differ significantly between the third and the fifth generations or between replicates. Based on this result, both generations and replicates were grouped for further statistical analyses (Table 1), and data for each generation can be found in the Supporting Information. Mean larval mortality increased with the increase in delta temperature between pre-exposure and experimental temperature (ΔT°= TT-PET) (Kruskal-Wallis 2 4,175 = 57.253, p-value = 1.1 × 10 −11 ) (Figure 3).
The pupae mortality was not significantly different between the any of the TTs (ANOVA F 2,177 = 1.023, p-value = 0.36) or PETs With generation, PET, and TT as fixed factor and clutches (replicate per PET) as random factor. The significance level of the p-value is displayed as the number of stars (***p-values < 0.001).

| D ISCUSS I ON
The aim of this study was to investigate the possibility of rapid adaptation of C. riparius to seasonal temperature changes. Experimental populations, composed of individuals from five natural populations across a wide distribution range, were reared under three different temperature regimes and mortality rates tested after the third and fifth generations, at the same three temperatures, respectively.
The pupation mortality did not display any significant differences between treatments while having an important mortality which is consistent with pupation being the most critical life stage (Oliver, 1971). The insensitivity of pupal mortality to external factors suggests that rather internal factors and processes are responsible Apparently, the small number of generations for pre-exposure did not elicit a benign effect on the fitness trait. Negative pleiotropic interactions to heat resistance or transgenerational epigenetic effect (Heard & Martienssen, 2014;Whittle, Otto, Johnston, & Krochko, 2009) could have caused the heightened mortality at high temperatures. However, since temperature conditions experienced by the next generation are difficult or even impossible to predict for a multivoltine species, such a canalization should be selected against if under genetic control. Nevertheless, it is possible that this difference was induced by the experimental protocol, as the experiment did not include feeding during the 48H of treatment. This could have been detrimental for cold-raised larvae, because while their developmental rate was increased with high temperature, they lacked the necessary energy. This therefore brings the light on the fact that the higher temperature treatment led to an increase in development speed. This means that larvae subjected to 26°C treatment theoretically developed more during 48 hr than at 14°C and this, without food intake.
The intermediate temperature regime applied for three and five generations did also not result in a relative or absolute decrease in mortality when the offspring was exposed to this temperature and even an increase compared to 14°C regarding the linearity of the results. This is probably because 20°C can be considered as the most benign temperature for the species (Supporting Information Table   S1), Therefore it did not apply a large enough selection pressure to overcome the genetic drift in the relatively small experimental populations. The mortality level of 30%-40% observed at this temperature is a known baseline for the species, at least in experimental conditions (Nemec, Heß, Nowak, & Pfenninger, 2012;Nowak, Jost, et al., 2007;Vogt, Pupp, et al., 2007;Waldvogel et al., 2018). The slightly, albeit marginally nonsignificant increase in mortality at 26°C is potentially due to a stronger selection pressure at this temperature.
The pre-exposure to 26°C yielded surprising results, because it reduced mortality significantly and substantially down to about 12% in all treatment temperatures despite the difference in developmental speed between the treatment temperatures ( Figure 5 right panel) contradicting the baseline mortality of warm region population from Waldvogel et al. (2018). The fact that there is no variation among the treatments, that is, a lack of interaction between the genotypes with the respective environment argues for a deterministic, that is, genetic effect. It is, however, unclear why mortality is not generally so low in C. riparius, if it can demonstrably be swiftly achieved by selection. Trade-offs with other unmeasured, costly fitness traits appear a reasonable explanation.
Another indication for selective processes was the obvious difference in variance within treatments: In PET 14° and PET 20°, the variance of mortality is high, indicating that a lot of variation is present within replicates, while it is strongly reduced for PET 26°, which can be expected under selection of resilient phenotypes (Bettencourt, Kim, Hoffmann, Feder, & Noor, 2002;Cavicchi, Guerra, La Torre, & Huey, 1995;Debat & Le Rouzic, 2018). We thus hypothesize that this adaptation is based on standing variation given moderate number of individuals, the low mutation rate of our species (Oppold & Pfenninger, 2017), the identical response of all replicates, and the short pre-exposure time (five generations) which would not be long enough to see de novo mutations rising to appreciable frequencies in the population. Our results are consistent with studies in Drosophila melanogaster. Populations reared at 28°C survived better when compared to populations reared at lower temperature (Cavicchi et al., 1995). The same laboratory populations were later found to have fixed alleles of the heat shock Hsp70 gene, but also a lower level of inducible heat shock proteins at 28°C than at lower TTs (Bettencourt et al., 2002). This lower level of inducible heat shock proteins was interpreted as lower capability to respond via phenotypical plasticity to temperature increase (Bettencourt, Feder, & Cavicchi, 1999); following the hypothesis that in this case, T opt has shifted to 28°C. A similar pattern was also  (Eriksson, Ryden, & Jonsson, 1989;Tee et al., 1985;Tee, Cranston, & Kay, 1987) and are able to spread diseases as such as cholera (Broza & Halpern, 2001;Halpern, 2011).
Our results suggest that (a) rapid adaptation to different temperature regimes is possible in C. riparius within a few generations and (b) therefore the response to seasonal changes in the temperature regime of natural populations may be at least in part be driven by adaptation and not only by phenotypic plasticity. However, our experiments have been conducted on a mixture of populations from across Europe. This mixture perhaps resulted in an artificial increase in genetic diversity composed of warm and cold temperature adapted individuals. Therefore, even if it was shown that many of these haplotypes may occur together in natural populations because of high gene flow (Waldvogel et al., 2018), the adaptation noticed during our experiment may not be as fast under natural conditions.
Still, previous studies indicated that field populations have a high genetic variability (Waldvogel et al., 2018). On the other hand, mixing of different populations may have led to outbreeding depression by induction of intrinsic genetic incompatibilities which, in turn, could have slowed or masked adaptation to the external selection regime (Lynch, 1991). Such outbreeding depression has been observed in crossings of different C. riparius populations . In our experiment, we only studied mortality as response to temperature; this may have biased our results by overseeing some responses that could have change the overall interpretation of the adaptation mechanisms. Therefore, to strengthen the inference of rapid adaptation over seasonal timescales further, it needs to be shown that similar phenotypic changes also occur in natural populations and that these changes are linked or at least associated with respective genomic allele frequency changes.

| CON CLUS ION
In this study, we showed that C. riparius lowered mortality due to high-temperature exposure by rapid adaptation rather than phenotypic plasticity. Our results thus indicate that C. riparius is able to react to seasonally varying temperature regimes by rapid adaption within a few generations, modifying the survival of the organism not only for warm but also for colder temperatures. This study adds to the currently accumulating literature of rapid adaptation in multivoltine species.