Diurnal variation around an optimum and near‐critically high temperature does not alter the performance of an ectothermic aquatic grazer

Abstract The growing threat of global climate change has led to a profusion of studies examining the effects of warming on biota. Despite the potential importance of natural variability such as diurnal temperature fluctuations, most experimental studies on warming are conducted under stable temperatures. Here, we investigated whether the responses of an aquatic invertebrate grazer (Lymnaea stagnalis) to an increased average temperature differ when the thermal regime is either constant or fluctuates diurnally. Using thermal response curves for several life‐history and immune defense traits, we first identified the optimum and near‐critically high temperatures that Lymnaea potentially experience during summer heat waves. We then exposed individuals that originated from three different populations to these two temperatures under constant or fluctuating thermal conditions. After 7 days, we assessed growth, reproduction, and two immune parameters (phenoloxidase‐like activity and antibacterial activity of hemolymph) from each individual. Exposure to the near‐critically high temperature led to increased growth rates and decreased antibacterial activity of hemolymph compared to the optimum temperature, whilst temperature fluctuations had no effect on these traits. The results indicate that the temperature level per se, rather than the variability in temperature was the main driver altering trait responses in our study species. Forecasting responses in temperature‐related responses remains challenging, due to system‐specific properties that can include intraspecific variation. However, our study indicates that experiments examining the effects of warming using constant temperatures can give similar predictions as studies with fluctuating thermal dynamics, and may thus be useful indicators of responses in nature.


| INTRODUC TI ON
Temperature is a fundamental driver of many natural processes ranging from metabolism to reproduction and behavior, making the study of organismal performance across thermal ranges an important foundation in our understanding of biology (e.g., John-Alder, Morin, & Sharon, 1988;Knies, Izem, Supler, Kingsolver, & Burch, 2006;Smith, 1973;Vaughn, 1953). The growing threat of global warming has led to a profusion of studies in the past two decades examining the effects of elevated temperature on biota. Because temperature can be difficult to manipulate and control under natural conditions, field studies that have modified thermal regimes have often had to accept considerable ambient variation and limited replication (Hillebrand, Soininen, & Snoeijs, 2010;Hood et al., 2018;Nelson et al., 2017).
Consequently, these challenges mean that most experimental studies assessing the effects of warming are conducted in highly controlled laboratory or outdoor mesocosm settings. Such experiments can be extremely effective by directly testing responses to warming treatments that, for example, simulate future climate scenarios using expected mean or extreme temperatures (IPCC, 2014), but often ignore the natural variability experienced by real ecosystems (Thompson, Beardall, Beringer, Grace, & Sardina, 2013).
In particular, seasonal or regional trends in temperature are often accompanied by diurnal fluctuations (Bozinovic et al., 2011). Therefore, the difference between the daily maximum and the daily minimum can be large although the magnitude of this difference can vary among locations (Figure 1). Despite such diurnal temperature variation, a great majority of laboratory experiments assess thermal responses of organisms using treatments with a constant temperature. This is because it is easier to expose organisms to constant rather than to fluctuating temperatures, although such a contrivance may affect the conclusions made. Thus, it is imperative to estimate how well constant temperature experiments capture the variation in responses compared to studies with fluctuating temperature.
For example, temperature variation can hasten the development of aquatic insects (e.g., Gresens, 1997;Huffaker, 1944;Sweeney & Schnack, 1977). Colinet, Sinclair, Vernon, and Renault (2015) reviewed the few published studies on this topic and found that a fluctuating temperature close to the thermal optimum may improve the organismal performance in insects compared with individuals exposed to constant temperature. Further, a fluctuating temperature close to the thermal extremes may either negatively impact insects due to cumulative damage during exposure or bring thermal refuge from harmfully low or high temperatures (Colinet et al., 2015).
Fluctuating thermal dynamics may also yield contrasting responses to high and low temperatures in same organism, for example, simultaneously promoting tolerance to high temperature and reduced tolerance to cold (Salachan & Sørensen, 2017).
Earlier research that compares responses of organisms to constant and fluctuating temperature is largely biased toward terrestrial insects as model organisms (Kutcherov & Lopatina, 2018). Whilst few studies have shown how responses to fluctuating versus constant temperatures may vary between different genders (Fischer, Kölzow, Höltje, & Karl, 2011) and life stages (Salachan & Sørensen, 2017), other studies have observed similar thermal responses across populations (e.g., Fragata et al., 2016;Manenti, Sørensen, & Loeschcke, 2017). Also cryptic phenotypes and their underlying genetic variation, that normally have little or no effect on phenotypic variation in a population, but under extreme conditions can generate heritable traits, may determine a population's ability to adapt to changes in the thermal environment (Paaby & Rockman, 2014 Here, we compared responses to a constant and diurnally fluctuating (±3°C) thermal regime at both "optimum" and "high" (close to the critically high temperature, Figure 2) mean temperatures using the great pond snail, Lymnaea stagnalis, as a model organism F I G U R E 1 Temperature fluctuation recorded in three natural ponds in Zürich (47°24′10N, 8°35′50E; 47°23′56N, 8°32′57E; and 47°23′58N, 8°32′56E, respectively), Switzerland over 22 days during the 2015 central European heat wave. The dashed lines indicate the maximum and minimum air temperature measured in Zürich (extracted from MeteoSchweiz) ( Figure 3). We compared the responses of snails that originated from three different populations within the same region. This increased the potential for historical contingencies generating population variability, thus enabling a more general view of thermal responses (constant vs. fluctuating) within this species. L. stagnalis is a hermaphroditic pulmonate gastropod (thus, no gender-based differences), with a wide distribution in stagnant or slowly flowing water bodies in the Northern Hemisphere. It is extensively used as a model organism to investigate the effects of warming (e.g., Leicht, Jokela, & Seppälä, 2013;Salo, Räsänen, Stamm, Burdon, & Seppälä, 2018;Salo, Stamm, Burdon, Räsänen, & Seppälä, 2017;Seppälä & Jokela, 2011) and pollutants (e.g., Coutellec & Lagadic, 2006;Nyman, Schirmer, & Ashauer, 2014;Salo et al., 2018;Salo et al., 2017) on organisms, as well as host-parasite interactions (e.g., Karvonen, Savolainen, Seppälä, & Valtonen, 2006;Leicht & Seppälä, 2014) and immunology (e.g., Dikkeboom, Knaap, Meuleman, & Sminia, 1985;. In this species, exposure to high temperatures increases several organismal process rates (Salo et al., 2018), which leads to increased growth rate and reproductive output with a temporal threshold (1 week), after which the reproductive rate is reduced . High temperatures also reduce snail immune defense Salo et al., 2017;Seppälä & Jokela, 2011), which increases their susceptibility to trematode parasites (Leicht & Seppälä, 2014). However, these earlier studies on the effects of warming have all used experimental treatments with constant temperatures. Following Colinet et al. (2015), we made two main predictions regarding snail performance under constant or fluctuating temperatures. Firstly, we expected that temperature fluctuations mimicking natural diurnal patterns around the optimum temperature would be either neutral or beneficial for snails (i.e., no impact or increased growth rate, reproductive output and immune defense compared to constant temperature conditions at the same average temperature). Secondly, we expected that the same fluctuations close to the critically high temperature would be harmful (i.e., reduced trait values compared to conditions with constant temperature at the same average temperature). Better understanding of how this realistic source of uncertainty affects organismal responses to temperature helps address the potential risks in conclusions based on studies using constant temperature treatments.

| MATERIAL AND ME THODS
The study was carried out in two phases. In the first phase, we determined the thermal performance curves ranging from medium to high temperature for four traits (reproduction, growth, and two immune parameters) to select "optimum" and "close to critically high" average temperatures (Figure 2) to be used in the second phase of the study. In the second phase, we then assessed whether snails' responses to these two average temperatures depended on whether the temperature was constant or fluctuating (±3°C).

| Phase 1: thermal performance curves
Typically, thermal performance curves show the highest performance (signifying optimum temperature) at an intermediate temperature and reduced performance at temperatures below and above F I G U R E 2 Conceptual thermal performance curve for a hypothetical trait across temperatures ranging from critical minimum (CT min ) to optimum (T opt ) and to critical maximum (CT max ). The black box indicates the temperature range targeted in the first experiment to define the optimum temperature (T opt ) and a temperature between the optimum and critical maximum (T high )  Figure 2). As we were interested in the effects of warming, we examined snail phenotype at temperatures ranging from mid to high ( Figure 2): 15, 18, 21, 24, 27, 30, and 33°C. This range covers the optimum temperature for the growth of juvenile L. stagnalis snails (Vaughn, 1953) and the typical temperatures in ponds in the study region (Zürich, Switzerland; T. Salo unpublished data). To estimate a general population-independent thermal response curve, we used a laboratory stock population created by mating individuals that originated from seven different donor populations in Northern Switzerland (Langeloh, Behrmann-Godel, & Seppälä, 2017). A total of 140 adult individuals from the fifth generation of this "mixed" population (shell length: 28.7 ± 0.3 mm, average ± SE) were placed in 200 ml glass containers filled with aged tap water. The snails were allowed to acclimate for 18 days to 15°C in a climate chamber, after which the temperature was increased by placing the containers holding the snail into water baths with the target water temperature (20 snails per temperature). The target temperatures in the water baths were reached by using EHEIM aquarium heaters (100 W) and circulation pumps to distribute the temperature evenly. Water in the experimental containers was changed every second day, and all individuals were fed ad libitum with fresh lettuce.
As traits may differ in their thermal responses, we assessed multiple traits to generalize across single trait thermal performance curves and to gain more reliable estimates for "optimum" and "high" temperatures. Survival was inspected every second day and dead individuals were removed from the containers. After 10 days, growth, reproduction, and two immune parameters were assessed for each surviving individual (see the Section 2.3).

| Phase 2: fluctuating versus constant optimum and high temperatures
Two target average temperatures, "optimum" (21°C) and "high" (28°C), were selected upon estimating the above-mentioned thermal performance curves. The temperature with the highest snail performance was selected as "optimum" and a temperature close but still below the critically high temperature was selected as "high" (see The target temperatures were reached by placing aquarium heaters (EHEIM, 100-300 W) together with a water pump to a water bath.
Fluctuations were obtained by switching one heater per bath on and off at specific times in the morning and in the afternoon, respectively, allowing the water to heat and cool as planned. An additional heater was installed in each water bath to ensure a stable minimum target temperature. The temperature in each water bath was quantified every 10 min using Onset HOBO temperature loggers. The final temperature profiles in the different treatments were 21.1 ± 1.4°C, 21.6 ± 2.8°C, 27.9 ± 0.9°C, and 28.3 ± 3.5°C, respectively ( Figure 4). (47°34′59N, 8°51′52E), and Zürichberg (47°23′32N, 8°33′52E). All three ponds were shallow (depth < 2 m), and the surface area ranged between ca 9-50 m 2 . These three locations were also included in the seven donor locations for the "mixed population" that was used in the first phase of the study (see above). Although Figure 1 rep-

| Response variables
The growth of individuals was quantified by measuring their shell length at the beginning and the end of the experiment to the closest 0.1 mm. These measurements were then used to calculate the specific growth rate according to Equation (1), where S 1 and S 2 are lengths at the beginning and at the end of the experiment, respectively, and t is the time between the measurements in days (Seppälä, Karvonen, Kuosa, Haataja, & Jokela, 2013).
To measure snail reproductive output, all egg clutches oviposited by the snails were collected and counted. To do this egg clutches were photographed on a light table, and the number of embryos in each image was then counted by using cell counter function in Image-J (Salo et al., 2017). The total number of embryos produced during the treatments (excluding acclimation period) was used as a measure for reproductive output.
The immune defense parameters phenoloxidase (PO)-like activity and antibacterial activity were measured using hemolymph collected from each snail at the end of the experiment. In invertebrates, including molluscs, phenoloxidase enzymes form part of the defenses against eukaryotic pathogens (Cerenius & Söderhäll, 2004), while humoral antibacterial enzymes are used against microbial infections (Imler & Bulet, 2005;Leicht et al., 2013). Each snail was blot dried and its foot was gently tapped using a pipette tip until it retreated into the shell simultaneously releasing hemolymph through the hemal pore (Sminia, 1981). Ten µl of hemolymph was mixed with 100 µl of PBS buffer and 100 µl pure hemolymph were collected for PO-like and antibacterial activity analyses, respectively. The samples were frozen in liquid nitrogen, stored at −80°C and analyzed using l-Dopa and Escherichia coli assays for PO-like activity and antibacterial activity according to Leicht et al. (2013). Briefly, the enzyme PO oxidizes l-Dopa causing an increase in optical density of the solution, while antibacterial enzymes in the hemolymph destroy lyophilized E. coli cells that decrease optical density. The changes in optical density were measured spectrophotometrically (SpectraMax 190, Molecular Devices).

| Statistical analyses
The thermal response curves were analyzed by fitting the data with a polynomial regression (y = ax 2 + bx + c) for each examined trait (in Euclidean distance. All analyses were run 9,999 times using type I SS.  Figure 5b).

| Phase 2: fluctuating versus constant optimum and high temperatures
The multivariate analysis (PERMANOVA) showed that the average temperature, not temperature fluctuations or their interaction, affected snail performance (Table 2). There was, however, an interactive effect of population and temperature fluctuation on the multivariate phenotype (Table 2). When the responses of different traits were analyzed separately ( Figure 6), growth rate increased and antibacterial activity decreased at high temperature (Table 3, Figure 6b,h). Whether individuals were exposed to constant or fluctuating temperature had no impact on the responses (Figure 6b,h).
Populations differed in their growth rates at the tested average temperature levels (Table 3, Figure 6a). Further, the PO-like activity was interactively dependent on population and fluctuating temperature (Table 3, Figure 6c). Reproductive output and PO-like activity increased with size of individuals (Table 3).

| D ISCUSS I ON
Responses of organisms to changing temperature conditions can be alleviated by behavioral and physiological modifications of the phenotype (i.e., plastic responses). These responses are, however, typically examined in experiments where organisms are exposed to constant temperatures without mimicking natural thermal regimes (e.g., diurnal variation). The use of such artificial conditions has been criticized, because it potentially underestimates the effects of warming on organisms (Paaijmans et al., 2013). To compare how responses to constant versus fluctuating temperatures differ in an aquatic grazer species, we exposed three populations of the snail L. stagnalis to "optimum" and "high" average temperatures both Our results show that diurnal temperature variation across all populations had no significant independent effects on organismal traits. This contrasted with the results of Bozinovic et al. (2011) showing that thermal variation enhanced the rate of population growth at a low average temperature but depressed this rate at a high average temperature. Theoretically, whether or not fluctuating and constant temperatures with the same average generate different or similar responses in organisms depends on (a) the shape of the species' thermal response curve, (b) at which average temperature the fluctuations take place as well as (c) the magnitude of temperature fluctuation (Colinet et al., 2015). Other factors that might further influence responses include the duration and frequency (i.e., predictability) of temperature fluctuations (Manenti, Sørensen, Moghadam, & Loeschcke, 2014).
Thermal response curves for the studied traits indicate that our study organism has a wide tolerance for temperatures on the warmer side of the temperature optimum. This generalist-like functioning over a large range of temperatures could potentially explain the lack of responses to fluctuating temperatures in comparison to constant temperatures. Such a response could be expected in our study organism, given its wide distribution and preference for standing water habitats. In contrast, aquatic ectotherms with a more asymmetric temperature envelope could be more susceptible to thermal variation. For example, semi-voltine aquatic insects that emerge in spring (e.g., some Plecoptera and Ephemeroptera) have short-life cycles which exposes nymphs to lower and less variable temperatures than other longer-lived species, resulting in a more asymmetric thermal niche with a lower critical thermal maximum (Ernst, Beitinger, & Stewart, 1984). Nonetheless, the daily maximum temperature in our treatment with high fluctuating temperature (28 ± 3°C) did get close to the upper thermal limit of our study organism (100% mortality at 33°C in the thermal performance curve experiment).
Whether temperature fluctuations do or do not differ from exposure to constant temperature with the same mean temperature can depend on whether the fluctuating temperature reaches over linear, convex or concave part of a thermal response curve (Colinet et al., 2015). At temperatures with linear change, fluctuations should not affect the response, while during concave or convex thermal response may increase and decrease the response in relation to constant temperature, respectively (Paaijmans et al., 2013). Hence, the consistently high performance independent of fluctuations at high temperature could indicate somewhat linear response to high temperatures in our model organism. Furthermore, in a study involving a mosquito model, thermal variation strongly reduced the performance at near-critically high temperatures (Paaijmans et al., 2013).
Thus, the similar responses to constant and varying temperature at the high temperature treatment used in our study suggest that our study species uses the thermal refuge from high temperatures (i.e., cooling periods) for recovery in order to preserve individual performance. High temperatures are thus likely to require more cumulative exposure (i.e., longer duration) to adversely impact our study species.
The discrepancies between responses to fluctuating and constant temperatures have been suggested to depend on how temperature-sensitive the respective trait is (Colinet et al., 2015). While we assessed several traits, only reproductive output showed a slight (nonsignificant) tendency toward our hypothesis of temperature fluctuations improving reproductive output at optimum temperature but decreasing the output when exposed to high temperature.
This could reflect either that (a) reproduction does not gain from the thermal refuge to the same degree as other traits, or that (b) the fluctuating temperature provides enough thermal refuge to prevent terminal investment (Williams, 1966).
Just as acclimation can increase an individual's thermal tolerance (Colinet et al., 2015;Marshall, Brahim, Mustapha, Dong, & Sinclair, 2018), natural selection should alter thermal responses in populations over contemporary evolutionary timescales (Merilä & Hendry, 2014). For instance, Leicht, Seppälä, and Seppälä (2017) observed high family-level trait variation in L. stagnalis at different temperatures suggesting some evolutionary potential in response to warming in our model organism. However, the evidence still remains weak, with numerous examples contradicting these predictions (e.g., Fragata et al., 2016;van Heerwaarden, Lee, Overgaard, & Sgrò, 2014;Manenti et al., 2017 organismal responses to environmental heterogeneity (Manenti et al., 2017). Some potential factors explaining this discrepancy may be the levels of standing genetic variation and presence of cryptic phenotypes in populations (Salinas et al., 2019), and in aquatic environments, rising or extreme temperatures may also be exacerbated by an associated decline in oxygen availability, thus strengthening the selection pressure on aquatic ectotherms (Hoffmann, Chown, & Clusella-Trullas, 2013). In our experiment, the differences between populations explained more of the variation in trait responses compared to whether organisms were to faster growth rates in butterflies and while fluctuating and constant temperatures could result in differing average growth rates, the direction of change compared to the low-temperature treatment was always independent of the temperature variation. Bernhardt, Sunday, Thompson, and O'Connor (2018) further showed that while thermal performance curves of phytoplankton differ depending on whether they experienced constant or fluctuating temperatures, the former can reliably be used to predict the latter. As long as studies concentrate on relative changes rather than mean trait values per se, and consider that decreased environmental variability may yield more conservative results, constant temperature studies may still give a good estimation for temperature-induced responses (Fischer et al., 2011).
In conclusion, our study suggests that exposure to fluctuating temperatures does not necessarily yield altered responses when compared to constant temperature treatments. Whilst extrapolating results from temperature experiments conducted in benign laboratory conditions should be done with careful consideration, constant temperatures are a valid proxy when estimating responses to warming.

ACK N OWLED G M ENTS
Laura Langeloh and Tamara Schlegel are thanked for their help with the experiments and the laboratory analyses. Christian Stamm and Katja Räsänen gave valuable comments on the experimental design. Anja Taddei took the photo of the model organism ( Figure 3). We thank anonymous reviewers for their helpful comments on the manuscript. The study was funded by Eawag discretionary funds.

CO N FLI C T O F I NTE R E S T S
We declare we have no competing interests.

AUTH O R CO NTR I B UTI O N S
TS, TK, FJB, and OS designed the study; TS and TK conducted the experiments and laboratory analyses; TS analyzed the data and TS, FJB, and OS all contributed to the writing.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data are available in Dryad Digital Repository https ://doi.