Increased duration of extreme thermal events negatively affects cold acclimation ability in a high-latitude, freshwater ectotherm (Ischnura elegans; Odonata: Coenagrionidae)

Instances of heat waves and cold snaps are becoming more frequent and of increasing duration worldwide. It is well established that short exposure to high or low-temperatures, such as during extreme weather events, often results in adaptive (acclimation/hardening) or maladaptive plastic changes in tolerance of organisms to subsequent thermal stressors. However, little information is available about how the duration of a prior stressful thermal event mediates future organismal thermal responses. Understanding durational effects of thermal conditioning can help predict ectothermic survival in response to novel extreme weather patterns. We assessed the effect of stressful temperature duration on tolerance to subsequent cold exposure in a widespread freshwater invertebrate species in Britain. Following a week-long acclimation period at 15°C, wild-caught blue-tailed damselfl y Ischnura elegans larvae were held at stressful thermal extreme (2°C or 30°C) temperatures for varying durations designed to mimic a range of extreme to plausible durations of heat waves or cold snaps in the wild (30 min, 2 h, or 24 h). After a period of re-equilibration to ambient temperatures (15°C), we then experimentally assessed CTmin, the temperature which renders an individual unresponsive, as an index of cold tolerance. Prior exposure to 2°C, simulating a cold snap, improved future cold tolerance, but only when individuals experienced very brief prior exposures to these conditions (30 min up to 2 h), and this benefi t was lost following 24 h prior exposure. Prior exposure to 30°C, simulating a heat wave, consistently worsened the subsequent cold tolerance of individuals, with the detrimental effect of prior heat exposure increasing linearly as a function of duration. The research indicates that cold snaps can provide benefi cial hardening effects against future cold exposures, but only when these (priming) extreme weather events are of very short duration (here, 30 min or 2 h). Longer durations of exposure to either extreme heat or cold weather events can reduce the ability of individuals to benefi cially react to subsequent cold stresses, and may have deleterious effects on future thermal tolerance. The results suggest that increasing durations of extreme temperature events will reduce cold hardening ability of freshwater invertebrates, and that the duration of extreme weather events, or durational changes in freshwater thermal regimes resulting from changes in snowmelt dynamics, is an important parameter to consider when studying organismal responses to climate change.


Duration of extreme weather events
Irregular bouts of temperature fl uctuation occur in marine, freshwater and terrestrial biomes alike (Parmesan, 2006). Global terrestrial climatic temperatures are increasing; as are instances of extreme spells of temperature fl uctuation known as cold snaps and heat waves (EASAC, 2013). An increasing area of the globe was affected by increases in extreme climactic events starting in the late 20th century (Frich et al., 2002), and these increasing durations of extreme weather events have been linked to increased concentrations of atmospheric CO 2 (Baker et al., 2018). Instances of heat waves will likely continue to become more posure times or when experienced out of context (Buckley & Huey, 2016).
Despite an appreciation that physiological mechanisms of thermal hardening, maladaptive plasticity, and acclimation processes differ from each other, and that therefore long vs. short exposures to thermal extremes may have different subsequent effects on thermal tolerance, little has been done to directly investigate how the duration of previous exposure to a stressful or priming thermal event affects the capacity of an organism to respond to a subsequent thermal challenge, after a period of recovery (Waagner et al., 2013;Alemu et al., 2017).

Aims and experimental design
The aim of this research was to understand how the duration of an extreme temperature bout affects the capacity of individuals to withstand a future bout of thermal stress, via either adaptive hardening or maladaptive physiological plasticity. We tested the hypothesis that as duration of exposure to cold temperatures increased, so too would the hardening response (i.e., ability to withstand a subsequent bout of cold). We further anticipated that short exposures to cold temperatures may produce different levels of future cold protection than longer durations of exposure, with longer durations of prior exposure potentially invoking more maladaptive responses to subsequent stressors than shorter-term prior exposures to the same priming temperature (Knapp et al., 2018). We furthermore questioned whether hardening responses could be achieved in as little as 30 min of exposure (Lee et al., 1987). Finally, we examined whether exposure to heat stress would provide a protective cold-tolerance response, and if this too depended on the duration of exposure.

Study system
Zygoptera ("damselfl ies") are a suborder of Odonata which are known to have substantial phenotypic and ecological diversity, making them excellent study subjects for linking ecology, genomic and evolutionary research (Córdoba-Aguilar, 2008;Bybee et al., 2016). Blue-tailed damselfl ies (Ischnura elegans Vander Linden, 1820) are ubiquitous across Eurasia and are characteristic of mesotrophic to eutrophic permanent wetlands (Bouton et al., 2011). Climate plays a strong role in the distribution of I. elegans (Lancaster et al., 2015;Sánchez-Guillén et al., 2016;Fitt & Lancaster, 2017), and this species has been undergoing a rapid range shift in response to climate change (Hickling et al., 2006).
The recent range shift has affected changes in thermal tolerance of this species, with the result that cold tolerance cold snap conditions will become more severe (Kodra et al., 2011). In Scotland, where this study was conducted, cold spells have consistently increased in duration over the past 60 years, from an average of 2.1-2.2 days per year in 1961-1990to 3.4-4.6 days per year in 2008(Kendon et al., 2018. Changes in snowpack and runoff dynamics under climate change can also lead to changes in the durations of extreme heat and cold environmental temperature exposures to freshwater ecosystems at high latitudes and elevations. Thus, it is now timely to understand how changing durations of climate extremes affect organismal responses, as a high priority for climate change ecologists. To explain how species will persist despite global warming, it is important to understand the responses of wild individuals, populations, and communities to the different aspects of these temperature perturbations (Deutsch et al., 2008;Gilman et al., 2010;Sgro et al., 2016).
To survive under variable thermal conditions typical of most temperate climate regimes, ectotherms often exhibit adaptive plasticity which allows them to anticipate changing thermal conditions within their lifespan (Sgro et al., 2016;Abram et al., 2017;Manenti et al., 2017). There are two distinct mechanisms of adaptive thermal plasticity in response to cold weather events that afford an organism protection in subsequent otherwise lethal conditions. A rapid cold hardening effect (RCH) is a phenomenon whereby increased survival probability at an otherwise lethal cold temperature follows a short period of exposure to extremely low temperatures (Lee et al., 1987;Chanthy et al., 2012;Sgro et al., 2016). In contrast, cold acclimation is a process believed to be distinct from RCH, and instead results from a longer-term prior exposure to moderately stressful environments which improves survival and performance over longer time frames of days or even weeks (Colinet & Hoffmann, 2012;Nyamukondiwa et al., 2018).
A number of molecular mechanisms have been implicated in acclimation and cold hardening processes, including production of heat shock proteins (Sørensen et al., 2003;Stoks & De Block, 2011), mucins, epigenetic regulation, detoxifi cation and proteins involved in ion and water homeostasis Enriquez & Colinet, 2019), and other transcriptome-and metabolome-wide changes (MacMillan et al., 2016) which are often highly labile within species .
Although the mechanisms which drive long-term acclimation vs. rapid cold hardening are not completely understood, studies have shown that these mechanisms may be physiologically independent of one another and may even be antagonistic (Rajamohan & Sinclair, 2009). Moreover, exposure to one extreme temperature (heat or cold) may also confer protective benefi ts or maladaptive consequences following exposure to the other extreme (Burton et al., 1988, Gotcha et al., 2018. Finally, exposure to stressful temperatures, especially over long durations, may weaken the ability of individuals to respond adaptively, resulting in maladaptive effects of prior thermal exposure at long ex-phenotypes have improved and gained in adaptive plasticity in cold stress towards the range edge (Lancaster et al., 2015(Lancaster et al., , 2017aWood et al., 2019), and changes in HSP allele frequencies have also shifted towards the range edge in response to temperature gradients (Dudaniec et al., 2018). Understanding how this well-studied species responds to changes in the duration of stressful thermal priming events, in terms of adaptive or maladaptive adjustments to its acclimation capacity, will help develop spatiotemporal estimations for shifts in thermal tolerance and distributional shifts in this and other species (Sánchez-Guillén et al., 2016;Fitt et al., 2019), when taking changes in the duration of different extreme weather events into account.

Experimental protocol
Larval I. elegans were collected from a series of ponds at Midmar Trout Fishery, Aberdeenshire, UK (57°9´27.23˝N, 2°30´10.63˝W) during April-June 2018 using fi ne mesh nets and transported to the laboratory in individual containers fi lled with pond water. Once returned to the laboratory, each damselfl y was identifi ed to species using (Cham, 2009). Head width in millimetres (precision to the nearest 0.01 mm) was measured using a Yenway dissecting microscope (YenCam5/ISH500, China) and Yencam associated image software, following Fitt & Lancaster (2017), as head width is a good proxy for body size in this species. Head width was not positively correlated with capture date (cor = -0.04, Kendall τ = -0.06, z = -1.26, P = 0.21). Each larva was given a unique identifi cation number. Damselfl ies were then acclimated at 15°C for 7-9 days. During this period the larvae were fed daily ad libitum with Artemia spp. and Tubifex tubifex. Because damselfl y larvae are cannibalistic, they were kept individually in plastic containers (diameter of 6.5 cm and a height of 3 cm) with a piece of artifi cial pondweed as a perch.

Application of thermal stress
Each damselfl y was allocated a treatment of either 2°C (nonlethal stressful cold temperature typically experienced by wild populations during cold snaps that can occur under winter and early spring conditions; Krokowski, 2007), 15°C (average spring pond temperature, assessed using HOBO onset pendant thermal data loggers; Bourne, MA, USA) or 30°C (temperature equal to that a wild population might experience in heat wave conditions; Jackson et al., 2018) for a duration of zero hours (i.e., controls retained at 15°C), 30 min, 2 h, or 24 h. Size distributions of larvae were kept consistent across treatments (Table 1). After being held at their designated acclimation temperature for the designated duration, individuals were then returned to 15°C for a minimum of 2 h and a maximum of 3.5 h prior to assessing CTmin (see below). This allowed individuals from both the 0.5 h exposure time and the 2 h exposure times to re-acclimate before commencing thermal trials. There was no difference in resulting thermal tolerance between the 0.5 h and 2.0 h exposure durations (Table S1, Fig.  S1), confi rming that variation in the reacclimation time did not affect results. This 2 h re-acclimation at 15°C ensures that we are testing the effect of prior thermal exposure on acclimation status, as decoupled from effects of starting conditions, which can confound estimates of the acclimation response (Terblanche et al., 2007). While a rapid transition from 15°C to 2°C is unlikely to occur in freshwater bodies in nature within 30 min, this extreme example was used to reveal short term, measurable changes in CTmin in response to a brief cold snap.

Thermal tolerance
Knockdown temperature was chosen as the metric to establish approximate critical thermal minima (CTmin). CTmin is the temperature at which an individual cannot be deemed as functionally responsive and would therefore be unable to react to threats such as predation or make use of resources to feed (Terblanche et al., 2007). Cold tolerance trials were carried out on individuals using a temperature-controlled water bath (Grant-TX150, Cambridge, UK). Each larva was placed into an individually-labelled 50 ml test tube with 27 ml of sieved pond water (to remove substrate) and suspended in the water bath, which was itself fi lled with a mixture of water and propylene glycol to prevent freezing. A maximum of 25 and minimum of 18 larvae were tested for CTmin per trial. The water bath program commenced at 15°C and reduced temperature at a rate of 0.1°C per minute. An external temperature probe, which was connected to the water bath thermostat, was inserted into a control 50 ml test tube to ensure that the cooling within the individual test tubes proceeded at the programmed rate. Once the bath temperature reached 11°C the larvae were observed constantly, and individuals were gently prodded every minute once activity ceased, to assess knock-down temperature. This was characterised when larvae exhibited loss of muscle coordination and could no longer respond to prodding. Once this unresponsive state was reached, the temperature (CTmin) was recorded and the test-tube containing the individual was then removed from the water bath. A total of 10 cold tolerance trials were run on 222 individual larvae between the 25 th of April and the 16 th of June.

Statistical methods
All linear regression analyses were performed using linear models in R version 3.4 (R Core Team, 2017). CTmin was used as the response variable throughout, with fi xed, explanatory effects of head width (continuous), temperature (factor)*duration of exposure (continuous) or temperature*duration^2 (modelled as an orthogonal polynomial using the poly(x,2) function in R (Kennedy Jr & Gentle, 1980;Zuur et al., 2010), and the day of year (continuous) on which the damselfl ies were captured (to account for seasonal effects on cold tolerance). For statistical analysis, individuals retained at 15°C were assigned a duration of 0 min for the test temperature (2°C or 30°C) run on the same day (2°C and 30°C treatments were run on different days, always accompanied by controls). To identify the best model, the corrected Akaike Information Criterion was used to compare the full model to reduced models (Burnham & Anderson, 2002;Mazerolle, 2016;Fitt & Lancaster, 2017). We log(n+1) transformed the duration values to conform to the assumptions of the parametric test. Due to the inability of individuals to survive below 0 degrees, CTmin values were modelled using a Gamma general linear model with an identity link function. Variance infl ation factors were calculated using the vif() function in the car package for R (Fox & Weisberg, 2018), and were all found to be < 2. For plotting, signifi cant marginal quadratic or linear effects of duration were depicted separately for 2°C or 30°C prior treatment temperature. We further modelled duration of exposure as a factor, and supplementary results are presented in the S.I.

Effect of acclimation temperature on subsequent thermal tolerance
CTmin varied between 1.5 and 9.1°C across treatments (Table 1). CTmin was best explained by head width, day of year, temperature, duration of exposure, and an interactive effect of temperature × duration and temperature × duration 2 (Tables 2, 3). In general, CTmin decreased (improved) with both size and day of year, refl ecting a combination of ontogenetic and, potentially, seasonal environmental changes. Increasing duration of exposure at either of the stressful temperatures had signifi cant effects on CTmin. Exposure to 2°C had benefi cial effects on thermal tolerance at short timescales of prior exposure (30 min or 2 h), but these disappeared at longer prior exposure times (24 h) (Fig. 1A, C). In contrast, exposure to 30°C always worsened subsequent responses to cold (Fig. 1B, D). Outputs of comparable models in which duration was modelled as a factor are qualitatively similar (Table S1, Fig. S1).

DISCUSSION
The duration and frequency of thermal extreme events are increasing in response to global warming. Although effects of frequency of previous heat or cold exposures on subsequent thermal tolerance has been often characterised (Abram et al., 2017), the effect of duration of exposure to thermal extreme events on hardening processes is less well understood. Here we provide novel data indicating that the duration of a single extreme cold challenge event can differentially impact organisms' capacity for response to future thermal challenges. Specifi cally, we found that short-term (~30 min -2 h) exposure to challenging cold conditions (2°C) can produce a benefi cial response, which then protects individuals against future cold weather challenges. However, we fi nd that exposure to extreme 'cold snaps' over longer periods of time (at some point > 2 h and < 24 h) rapidly decreases the adaptive benefi t of such exposures (Figs 1A, 1C, S1). These results are in accord with a Rapid Cold Hardening (RCH) effect, which requires a short duration of recent exposure in order to be effective. This implies that, for species which depend on rapid cold hardening to protect against damage from a future thermal stressor, increased duration of the initial hardening event will have deleterious consequences.
All exposure times to an extreme heat challenge (30°C) induced deleterious responses to a future cold challenge (Fig. 1B, D). These results fail to support the hypothesis that rapid hardening responses refl ect a 'generalised stress response' that is transferrable to alternative thermal regimes (e.g., Burton et al., 1988). Instead, our data support the idea that stressful exposure to one extreme temperature may have detrimental effects on the ability to respond to the alternate extreme.
Our fi nding in this regard is similar to recent fi ndings that increased duration of heat stress has a negative impact on future cold tolerance in the soil arthropod Orchesella cincta (Alemu et al., 2017). However, unlike in our study, Alemu et al. report a lack of detrimental effects of increasing cold stress duration on future cold tolerance, which may refl ect differences among the studies in the degree of cold stress applied (4°C in the previous study, vs. 2°C in the present study) or differences in thermal capacity between freshwater and terrestrial environments. More work is needed to further investigate how the durational effects of thermal stress vary according to the degree of thermal stress applied (i.e., to examine temperature × duration interactions over a wider range of temperatures). Moreover, work on more species is needed, as previous studies suggest that species vary in the degree to which cold acclimation or hardening is induced by heat stress or vice versa (Burton et al., 1988;Hemmati et al., 2014;Bar-Ziv & Scharf, 2018;Leonard & Lancaster, in press). Combined with these earlier studies, our results highlight a clear knowledge gap in understanding how interactions between intensity and duration of thermal stress can be generally predicted to affect hardening capacity. A better understanding of these effects furthermore depends on the underlying mechanisms that drive hardening and longerterm acclimation processes under different durations of heat and cold stress, as the trade-off dynamics of these alternative mechanisms are not currently widely understood.
Although not a focus of our study, our results also indicated that cold tolerance improved as body size (measured as head width) increased. In a previous Ischnura elegans thermal tolerance study, which was carried out according to similar thermal ramping protocols on larvae from the same study region, but which did not include a prior heator-cold hardening component, we found that there was no relationship between body size and thermal tolerance (Wood et al., 2019). The discrepancy between these studies suggests that body size (or developmental stage) may affect acclimation capacity, rather than directly affecting absolute thermal tolerance. Generally there is little evidence to support an improved thermal tolerance with age (Bowler & Terblanche, 2008), although some studies have shown that a rapid hardening effect declines with age (Czajka & Lee, 1990). More work is necessary to determine how the impacts of the duration of heat or cold stress on subsequent cold tolerance differ across life stages, and how this is carried on into adult fi tness in important aquatic species such as damselfl ies; previous work suggests that acclimation at different life stages can have different fi tness costs and benefi ts (Ma et al., 2004;Mutamiswa et al., 2018).
Freshwater systems are particularly vulnerable to climate change and other anthropogenic changes in thermal regime (Woodward et al., 2010), including rapid changes in snowpack and runoff dynamics. Moreover, freshwater species are often represented by small metapopulations of individuals which have limited dispersal mechanisms and are therefore more sensitive to environmental thermal change (Hering et al., 2009). Therefore, investigating the implications of interspecifi c or inter-populational differences in effects of durational RCH responses could yield interesting results with implications for predicting future freshwater (or other) community compositions under climate change (Loeschcke & Hoffmann, 2007;Stoks et al., 2012;Lancaster et al., 2017b). Better understanding of interspecifi c and interpopulational variation in durational effects of heat and cold stress on acclimation capacity of aquatic invertebrates at different life stages can contribute to better understanding of how changing durations of extreme weather events under climate change (IPCC, 2001) might impact freshwater communities and ecosystems.