Limited variability in upper thermal tolerance among pure and hybrid populations of a cold-water fish

To forecast the impact of climate warming on cold-water fishes, thermal tolerance trials were conducted on six populations of brook trout from Cape Race, Newfoundland, Canada. Of these, three were outcrossed to assess the effect of hybridization. Although differences were found, there was little variation among populations in how they responded to thermal stress.


Introduction
Human-induced climate change may be the single greatest threat to global biodiversity (Sala et al., 2000). Climate change can interact with habitat fragmentation by creating physical (i.e. drought) or physiological (i.e. temperature) barriers (Hughes, 2000;Walther et al., 2002;Pearson and Dawson, 2003;Travis 2003), which limit the potential for independently 1 mobile organisms to shift habitats. As fragments become smaller, so too do the populations within them, resulting in a loss of genetic diversity via an increased likelihood of inbreeding, genetic drift and reduced gene flow (Young et al., 1996;Keller and Largiader, 2003;Andersen et al., 2004;Ezard and Travis, 2006). In general, this process reduces adaptive potential, the ability of a population to tolerate environmental change, and further decreases population size. This study was designed to investigate the ability of isolated and differentsized populations to deal with climate warming, and the extent to which hybridization might enhance this response.
By increasing genetic variability within a population upon which natural selection can act, hybridization may improve population responses to climate warming, such as upper thermal tolerance (Stockwell et al., 2003;Pickup et al., 2012). Nevertheless, a number of factors influence the outcomes of hybridization, which may also decrease or have no effect on average fitness within populations (Edmands, 2007;Fraser et al., 2008). In part, such outcomes depend on population size; larger populations are expected to provide more genetic material than small ones, and small populations are therefore expected to benefit greatly from hybridization due to having lower genetic variability and possibly reduced phenotypic plasticity prior to hybridization (Lande, 1988;Ellstrand and Elam, 1993;Frankham, 1996;; but see . Therefore, when studying population responses to climate warming, it is important to consider both the relative benefits of population size and hybridization concurrently. A practical metric for assessing the thermal tolerance of individuals from different, fragmented populations of a species is the critical thermal maximum (CT max ; but see Elliott, 1981), defined as the temperature at which an organism can no longer maintain coordinated movement or equilibrium control (Becker and Genoway, 1979). In nature, a loss of equilibrium affects an organism's ability to forage or avoid predation, which may ultimately affect individual fitness. As acclimation temperature (T a ) has been found to be correlated positively with CT max (Cox et al., 1974;Zhang and Kieffer, 2014;McDonnell and Chapman, 2015, although see Galbreath et al., 2004;Recsetar et al., 2012), the ancestral history and origin of a population are thought to be linked to an organism's ability to tolerate temperature increases (Stockwell et al., 2003;McDermid et al., 2012) and, as such, stream temperature regimes may result in population-specific thermal tolerance. A relatively new metric to assess thermal tolerance, agitation temperature, described by McDonnell and Chapman (2015) as the temperature at which a fish first begins to exhibit refugia-seeking behaviour (circling of the chamber, seeking refuge in substrate), may also provide insight to how quickly individuals can sense and attempt to react to environmental change. In addition, the difference between these two traits (CT max -agitation window) may represent a fitness metric yet unexplored in thermal tolerance literature.
Salmonids are a socioeconomically important family of cold-water fishes having traditional and commercial value. Although they are rich in populations and occupy a diverse range of habitats, habitat fragmentation has depleted their numbers, and their viability is of growing concern as climate change warms northern regions (Walther et al., 2002;Alley et al., 2003;Hinzman et al., 2005). Recently, studies on the ability of salmonids to tolerate climate warming have been variable, with some finding evidence of population-level variation in thermal physiology (Eliason et al., 2011) and others finding little or none (Elliott and Klemetsen, 2002;Kelly et al., 2014). In particular, the brook trout (Salvelinus fontinalis) is an extremely diverse (Angers et al., 1995; and highly plastic (Hutchings, 1996;Imre et al., 2002) stenotherm, inhabiting a thermal window of 1-22°C (Xu et al., 2010) to maintain both an internal body temperature below 20°C (Scott and Crossman, 1973) and physiological pathways affecting individual growth, reproductive timing, foraging and predator avoidance (De Staso and Rahel, 1994;Magoulick and Wilzbach, 1998). With such a low thermal window, cold water species such as brook trout might be strongly affected by climate change, as global temperature is expected to increase by 0.7-7.4°C over the course of the 21st century (Rouse et al., 1997;Heino et al., 2009). Additionally, northern brook trout populations may be at an adaptive disadvantage due to being genetically depauperate as a result of isolation in glacial refugia and historical bottlenecks (Bernatchez and Wilson, 1998). Few studies have investigated the thermal performance of brook trout, with many among them having looked only at: (i) the effects of ploidy, heating rate, or interspecific differences with other salmonids; (ii) thermal tolerance at a static upper thermal limit; or (iii) comparing few populations with long histories of hatchery manipulation (McCauley, 1958;Benfey et al., 1997;Galbreath et al., 2004;McDermid et al., 2012;Stitt et al., 2014). More research is needed in order to prepare for, and adequately address, the effects of climate change on this socioeconomically important species at the intraspecific scale, taking into account population size and hybridization.
Our study used six fragmented, genetically distinct populations of brook trout occupying streams on Cape Race (CR), Newfoundland, Canada (Fig. 1), to explore effects of hybridization and population size on upper thermal tolerance. Fragmentation of CR streams occurred as a result of the late-Wisconsinan glaciation (10-12 000 years before present; Danzmann et al., 1998), and these populations have been studied extensively (Hutchings, 1991;Fraser et al., 2014;. Besides having a common ancestry, CR trout populations have a number of additional attributes for such a study. First, the small size of CR streams (ranging in length from 0.27 to 8.10 km) allows for thorough sampling and accurate estimates of population size. and N b the effective number of individuals breeding in one spawning season; an analogue of effective population size that is positively related to genetic diversity of a population; Bernos and Fraser, 2016). Third, consistent with theory, small CR populations have less neutral genetic variation than large CR populations , yet genetic variation underlying quantitative traits does not vary with population size ; such apparent discrepancies make this population system an intriguing one for investigating what genetic metrics best predict responses to environmental change. Finally, CR streams vary in their thermal regimes (Supplementary material Fig. S1), with those inhabited by the two smallest populations in the present study having the coldest overall mean monthly temperatures (Table 1).
The upper thermal tolerance of pure and hybrid individuals were compared in terms of CT max , agitation temperature and a new metric coined the 'CT max -agitation window', i.e. the difference between an individual's CT max and agitation temperature. A smaller window signifies that an individual can continue to carry out normal behaviours for a longer period before demonstrating avoidance behaviour in increasing temperatures, whereas a large window indicates that the individual displays avoidance behaviour earlier and, as such, regular behaviours are disrupted sooner. We hypothesized that large populations would have higher thermal tolerance (i.e. higher CT max , higher agitation temperature and smaller CT max -agitation window) due to more genetic variation, as greater genetic variation may increase the likelihood of more thermally tolerant individuals, and small populations would have limited thermal tolerance as a result of both lower genetic diversity and colder thermal regimes in the wild (Table 1 and Supplementary material Fig.  S1). We also hypothesized that the magnitude of population size difference would affect that relationship; specifically, hybridizing small populations with a large one would disproportionally benefit the small populations, as they might have reduced fitness because of inbreeding depression.

Procuration of brook trout
From 13 to 26 October 2014, gametes were collected from six CR populations: Cripple Cove (CC), Freshwater (FW), Ouananiche Beck (OB), Still There By Chance (STBC), Whale Cove (WC) and Watern Cove (WN). For larger streams, individuals were collected from previously documented spawning sites (four to six per stream) and from areas observed to have obvious redd formations and large brook trout aggregates (Wood et al., 2014). For smaller streams, wherein fish densities were lower in the spawning grounds, individuals were collected throughout the entire stream.
Potential spawning individuals were collected via electrofishing surveys and checked for 'readiness'; a release of sperm for males, and an elongated cloaca/soft belly for females. Readiness was assessed in the days leading up to the expected date of gamete collection, and ready fish were held for 24-72 h in flow-through cages before collection.
Gamete collection took place between 19.00 and 01.00 h. Sperm was collected in 1.5 ml microcentrifuge tubes, whereas eggs were collected in 60 ml opaque plastic containers. Gametes were kept on ice and insulated so as not to freeze, and transported to St John's, Newfoundland, immediately after collection. They were then flown directly to Montreal, and crossed within 15 h from the beginning of gamete collection. Crosses were conducted to produce pure population offspring as well as full-reciprocal F1 hybrids (Supplementary material Table S1). Families were incubated separately within mesh-bottom containers 5.2 cm in diameter placed randomly with respect to population within a single 1000 litre recirculating tank and maintained at 7.0 ± 0.3°C (mean ± SD). Eggs were left mostly undisturbed until the eyed stage, except to remove fungal eggs, to reduce potential mortality following fertilization, at which point dead individuals were counted and removed daily. Dissolved oxygen and pH did not differ in different tank locations and were consistently maintained throughout the experiment.
After reaching yolk absorption, separate brook trout families were kept in flow-through bins within two larger, identical, 3000 litre tanks prior to thermal tolerance trials. The water temperature was maintained between 15.5 and 16.5°C (±0.2°C, SD), and multiple air stones ensured dissolved oxygen saturation. pH was 7.5 (±0.2, SD) across all tanks, and artificial light was set at a natural daylight cycle (corresponding to St John's, Newfoundland). Tanks were cleaned daily, fish were fed ad libitum two times daily, feeding time was constant, and all fish were kept in the same thermal conditions from fertilization to the end of the thermal tolerance experiments. At the time of the experiment, fish were~2-4 months post-yolk absorption.

Upper thermal tolerance trials
Subjecting an organism to a linear increase in temperature, the onset of spasms and loss of equilibrium are used as markers for CT max , with loss of equilibrium being the most commonly used (Lutterschmidt and Hutchison, 1997). Since its introduction (Cowles and Bogert, 1944), CT max studies have evolved to account for a number of factors influencing CT max results. A low rate of temperature increase (e.g. 0.02°C/min) allows organisms to acclimate to rising temperatures, whereas a high rate of temperature increase (e.g. 1°C/min) results in core body temperature lag, skewing CT max results (Cox et al., 1974;Becker and Genoway, 1979;Galbreath et al., 2004).
Experimental fish were starved for 24 h prior to trials, and trials were performed at the same time daily to ensure similar metabolic rates (Clark et al., 2013). Fish were given 1 h of acclimation time after being moved to a rectangular experimental test tank (60 cm × 32 cm × 30 cm; length × width × height) to reduce stress associated with handling and to acclimate to minute changes in water temperature. Significant efforts were made to standardize starting water temperature, which ranged from 16.20 to 17.97°C over 34 trials. Each trial consisted of two pure trout from the same family and two maternal hybrids, with a total of n = 122 trout tested across 61 families from six populations ( Table 2). Within the larger rectangular test tank, four smaller, tapered circular flow-through chambers (14 cm top diameter × 10 cm bottom diameter × 11.5 cm deep) were used to hold each experimental fish. Rock substrate was provided in each chamber to act as potential fish refuge. During the trial, individuals were subjected to a constant (0.3°C/min) increase in water temperature that was controlled, monitored and recorded by a temperature-control unit and software (TMP-REG, AutoResp; Loligo Systems; McDonnell and Chapman, 2015).

4
Agitation temperatures were recorded for each fish as the point where an obvious shift in behaviour first occurred. For brook trout in this study, fish generally remained relatively still as temperatures increased until a point (agitation temperature), after which they began to circle the chamber hurriedly or sought refuge in substrate, or a combination of both behaviours. This agitation temperature, along with temperature at CT max , were both confirmed after each experiment using time-stamped video footage taken via a mounted webcam. Immediately after loss of equilibrium, fish were removed and placed in an aerated recovery chamber until regaining equilibrium and normal opercular movement; total length (in millimetres) was then recorded.

Statistical analysis
All analyses were conducted using R (R Core Team, 2016), and all packages were retrieved from its open-source directory. Linear mixed models were used to determine whether CT max , agitation temperature or CT max -agitation window differed across populations and between pure fish and their corresponding hybrids. Two models were run for each hybrid comparison and for the pure data alone. Model type 'A' initially included length and cross-type as independent fixed effects, whereas model 'B' included length, population size and the percentage coefficient of variation for summer temperature (summer CV) in each stream and for all years of data available. Hybrid comparison 'B' models did not include population size due to smaller data sets. All models were run using mean values of each of the three thermal tolerance traits, as well as using the family-level (within-population) mean variance for each trait. Factor significance was determined by reverse model selection using the R package pbkrtest to compare complex models with less complex ones using F-tests (Halekoh and Højsgaard, 2014). Mother identity was included as a random effect to account for maternal effects on thermal tolerance and to incorporate the replication of trials (two trials per family). For each model, data were normally distributed (variances logged), and pairwise P-values were calculated and corrected for false discovery rates (Benjamini and Hochberg, 1995) using the R package lsmeans (Lenth, 2015).

Results
Mean CT max was significantly different between five out of fifteen pure population comparisons (overall effect: d.f. = 32.96, F = 5.44, P < 0.001). Plots of 95% confidence intervals for mean CT max by population (Fig. 2) Table 3). Mean agitation temperature and CT max -agitation window did not differ between all other pure populations, and the variance of all traits did not differ between pure populations (Fig. 2). Summer CV (thermal variation) had a significant effect on only mean CT max (d.f. = 38.42, F = 18.10, P < 0.001) and mean CT max -agitation window (d.f. = 38.13, F = 5.60, P = 0.02), with larger variance resulting in lower thermal tolerance (i.e. lower CT max and larger CT max -agitation window; Fig. 3). Length and population size had no effect on mean and variance values of thermal tolerance for pure populations (Table 3A) and were therefore not included in the final models.
Length had a significant effect on thermal performance (Table 3B) in some models for pure vs. hybrid comparisons; however, this effect was not specific to one trait, nor the trait's measure (i.e. mean or variance). Of all traits, only one comparison of thermal tolerance (mean CT max , FW vs. FW-STBC hybrid) was found to show a weak significant difference (d.f. = 22.15, F = 3.37, P = 0.02). However, mean CT max differed by a maximum of 0.3°C between different hybrid cross-types (Fig. 4, Supplementary material Fig. S2 and Table 3B). No effect of summer CV was found for any pure vs. hybrid comparisons.
Supplementary visual analyses were conducted to determine whether any relationship existed between the dependent variables CT max , agitation temperature and CT max -agitation window and the year-long variation in temperature, mean maximal and minimal summer temperatures in each stream across all years for which data were collected. Of these, none was found to have an obvious relationship with any of the dependent variables (Supplementary material Fig. S3).

Discussion
The aim of this study was to investigate how varyingly-sized and genetically distinct populations of a cold-water species  respond to acute thermal warming, and how hybrids created with these populations respond in comparison. Our study on brook trout found significant differences in mean CT max between some pure populations and significant effects of summer CV on mean CT max and mean CT max -agitation window. Interestingly, populations that had greater variability in summer temperatures exhibited lower fitness, i.e. lower CT max and larger CT max -agitation windows. However, no population differences in agitation temperature or CT maxagitation window were seen (means or variances), no heterosis was observed, nor was there any effect of population size on any of the traits (means or variances). Although some  (Eliason et al., 2011;Gradil, 2015), including brook trout (McDermid et al., 2012;Stitt et al., 2014), others have found no differences between populations (Elliott and Klemetsen, 2002), differences between hybrid and pure crosses (Fields et al., 1987), evidence for heterosis in thermal tolerance of copepods (Willett, 2012) and evidence for increased survival of heterozygotes at near-lethal temperatures in Eastern mosquitofish (Gambusia holbrooki; Meffe et al., 1995). To our knowledge, agitation temperature has been assessed in fishes only once before (McDonnell and Chapman, 2015) but was studied in relationship to sex and acclimation temperature.
Our study examined six genetically distinct populations of brook trout that have been isolated, without gene flow or human disturbances, for potentially 12 000 years (Danzmann et al., 1998). Previous studies have found that although these CR brook trout populations differ nearly 50-fold in census size N and 10-fold in effective number of breeders N b (Bernos and Fraser, 2016), there is no evidence for differences in (i) quantitative genetic variation and trait differentiation in relationship to population size, nor (ii) phenotypic plasticity in relationship to population size . Therefore, our study provides further evidence that population size may not be tightly related to the ability of a population to respond to environmental change, and that thermal tolerance, in particular (physiologically and behaviourally), may be highly conserved even in such a plastic species as S. fontinalis. A key factor supporting this is that some of our populations fall below what many deem a minimal viable population (MVP) size for long-term persistence, which is hypothesized to range from N = 4100 to 7300 (Table 1; Traill et al., 2007). It might then be expected that populations exceeding this size would show greater thermal tolerance, but we found little supporting evidence for this. Thus, while the second largest population (WN: mean N b = 178.59 and mean N = 2836.00) had the highest CT max and one of the two smallest populations had the lowest (STBC: mean N b = 27.65 and mean N = 916.87), the largest population (FW: mean N b = 200.05 and mean N = 5118.30) also had the second lowest CT max . Another expected outcome might be that hybridizing between populations above and below the minimal viable population threshold would benefit smaller populations disproportionally. In our hybridized crosses, one large population (FW, see above) was hybridized with two smaller populations (STBC, see above; WC: mean N b = 31.36 and mean N = 783.09; Bernos and Fraser, 2016) with little effect on any thermal tolerance trait, whether measured by mean or variance.
Owing to the considerable number of populations we compared, hybridizing between populations vastly different in population size and annual thermal regimes (Supplementary material Fig. S1) and assessing both mean values and variance in each trait, our results suggest that thermal tolerance (in terms of temperature tolerance and behavioural responses to temperature increases) seems to be highly conserved in S. fontinalis and remains relatively unchanged across isolated populations and in their hybrid offspring. Even the significant mean differences found between our populations support this theory, as their magnitude is likely not to be biologically meaningful (0.68°C; see Fields et al., 1987 for a similar discussion), nor their variances smaller in smaller populations. Allowing our populations to acclimate at a temperature regularly experienced during summer months in the wild (16°C; Table 1 and Supplementary material Fig. S1), measuring CT max in a fluctuating thermal environment (Ketola and Saarinen, 2015), increasing temperature at a rate previously found to be optimal for fish studies on thermal tolerance (0.3°C/min; Becker and Genoway, 1979) and measuring both CT max and a behavioural metric of agitation temperature have provided new evidence for less variability in thermal tolerance than previously thought. Reasons for this may be the scale at which other studies were conducted (see McDermid et al., 2012;Stitt et al., 2014) as well as the historical genetic or environmentally driven similarities in the populations being assessed. In our case, although populations experience  (Table 1 and Supplementary material Fig. S1) and have been isolated for thousands of years, the relative geographical proximity of the populations may have resulted in similar environmental pressures shaping their ability to cope with climate warming. Nonetheless, we would expect smaller populations to have reduced variance in thermal tolerance and therefore reduced adaptive potential; however, this was not the case. It is possible that at larger scales, betweenpopulation and pure vs. hybrid differences in thermal tolerance may have been larger. Additionally, similar thermal performance Figure 4: Means and 95% confidence intervals for critical thermal maximum (CT max ), agitation temperature and CT max -agitation window (mean response values and mean variance) for one pure-hybrid comparison. Freshwater (FW; large population) was crossed with Still There By Chance (STBC; small population). A statistically significant difference in mean CT max was found between FW and H1 (d.f. = 22.15, F = 3.37, P = 0.02). H1 and H2 represent reciprocal F1 hybrid crosses. 10 may have to do with the highly conserved natures of heat shock proteins (for example, see Molina et al., 2000;Basu et al., 2002;Gradil, 2015) as well as similar haematocrit and peak heart rate values across populations (Gradil, 2015). As these proteins have increased expression in thermal crises, their naturally high level of conservation may be correlated with a highly conserved CT max . A next research step could be to acclimate these trout populations to different temperatures and then measure CT max , to determine whether acclimation potential is reduced with population size.
Our study is one of only a few to have examined upper thermal tolerance in a large number of populations of cold-water fishes (see also Eliason et al., 2011;McDermid et al., 2012;Stitt et al., 2014) and is the first to assess the CT max -agitation window. Additionally, we have accounted for genetic population size and family-level variation and tested for the effects of population mixing, which are factors that may affect the degree of tolerance. We found little population differentiation in upper thermal tolerance and no indication that population size or hybridization (enhancing genetic variability) affects thermal tolerance. It is, however, difficult to disentangle the effects of temperature regime from population size, as the two show a weak positive correlation. We have therefore highlighted the potential for a highly plastic and divergent species to have lower than expected resilience in the face of climate warming; large or mixed populations are not necessarily conferred any greater resilience to climate warming than small, isolated populations, nor do they provide increased resilience to small populations via hybridization. Although we cannot completely disentangle the relative roles of historical genetic vs. environmentally driven similarities in the populations being assessed, our results are a cause for concern for the general conservation of this and related cold-water species as the climate warms.

Supplementary material
Supplementary material is available at Conservation Physiology online.