Thermal constraints on exercise and metabolic performance do not explain the use of dormancy as an overwintering strategy in the cunner (Tautogolabrus adspersus)

The ability to cope with cold, and food-poor, winters is a key to persistence of animal populations at poleward latitudes (Marchand, 2014; Feary et al., 2014; Stuart-Smith et al., 2017). For ectotherms, frigid winter temperatures, in particular, can greatly constrain performance, activity and fitness (Ultsch, 1989; Shuter et al., 2012; Huey, 2010). Three main overwintering strategies have evolved to cope with winter and maintain a poleward range: (1) seasonal movement to warmer areas (cold avoidance), (2) maintenance of activity by compensation of physiology (i.e. improved cold performance) via cold acclimation and/or adaptation, and (3) an inactive, energy-saving state of dormancy (Marchand, 2014; Shuter et al., 2012; White et al., 2012). Winter dormancy in ectotherms is an inactive, sheltering, low metabolic rate and non-feeding state, generally analogous to mammalian hibernation (Gregory, 1982; Reeve et al., 2022; Ultsch, 1989). Among fishes, winter dormancy occurs in numerous temperate species but contrasts with the more common, and better studied, strategy of sustained winter activity supported by physiological thermal compensation (Amundsen and Knudsen, 2009; Block et al., 2020; Campbell et al., 2008; Crawshaw, 1984; Guderley, 1990; Reeve et al., 2022). Little is known about the driver(s) of winter dormancy in fishes or other vertebrate ectotherms – why do certain species use dormancy as a strategy to survive winter?

We propose the ‘cold performance limitation’ hypothesis (Fig. 1) whereby winter dormancy in fishes, and possibly other vertebrate ectotherms, is an adaptive behavioural response to severe, uncompensated constraints of cold on physiological performance that make it inefficient and/or risky to remain active in winter. In other words, winter dormancy is an alternative strategy to physiological compensation, allowing a species to tolerate its sluggish, inflexible performance in seasonal cold by pausing life activities and sheltering in an energy-savings mode. Essentially, winter dormancy could be a ‘behavioural filter’ (Garland and Losos, 1994) that compensates for poor cold performance. Dormancy theoretically could be advantageous by minimizing acclimation costs (Wilson and Franklin, 2002) and may be an acceptable trade-off for better performance at warm temperatures (Malusare et al., 2023). Indeed, we expect that dormant species, which tend to belong to ‘warm-adapted’ lineages (e.g. Labridae), have cold-constrained performance because they have warm-shifted performance curves, such that winter temperatures overlap the extreme cold end of their thermal window where performance nears zero (Fig. 1). The cold performance limitation hypothesis for winter dormancy, also alluded to by Shuter et al. (2012) and hinted at by reptile researchers for decades (Cowles, 1941; Gregory, 1982; Huey and Pianka, 1977; Huey and Stevenson, 1979; Huey et al., 2021), contrasts with (but is not exclusive of) the ‘energy availability’ hypothesis for mammalian hibernation whereby food scarcity drives dormancy (Humphries et al., 2003). Constraints of cold on performance can powerfully influence activity, habitat selection, predation risk and anti-predator behaviours among ectotherms (Bennett, 1990; Gunderson and Leal, 2016; Herrel et al., 2007; Huey, 1982; Huey and Stevenson, 1979; Kingsolver, 1983; Rand, 1964). However, the cold performance limitation hypothesis has not been investigated in ectotherms as an explanation for the widespread phenological event of winter dormancy.

Aerobic and exercise performance are ideal traits to investigate the cold performance limitation hypothesis because they are fitness-linked, temperature-dependent traits in ectotherms including fishes (Kieffer, 2010; Schulte, 2015). Sustained and burst swim performance, standard (SMR) and maximum metabolic rates (MMR), and aerobic scope (=MMR–SMR, the aerobic capacity for life activities such as swimming) all usually decrease with cooling below their optimal temperature (T_opt) in fish species; however, winter-active species such as salmonids show well-maintained swimming and aerobic scope when exposed to low temperatures that may be further ameliorated by cold acclimation (Guderley, 1990; Hvas et al., 2017, 2018; Johnston and Temple, 2002; Porter and Gamperl, 2023; Schurmann and Steffensen, 1997; Tirsgaard et al., 2015). Although variation in the cold sensitivity of metabolic performance is known among fishes (Guderley, 1990), temperature–performance relationships have rarely been interpreted in the context of species-specific overwintering strategies and have not often been studied at winter low temperatures (<5°C).

Here, we tested the cold performance limitation hypothesis for winter dormancy in fishes. We predicted that large and uncompensated decreases in physiological performance would occur at temperatures below the winter dormancy threshold temperature (WDTT: the temperature below which there are disproportionately great decreases in activity and feeding as the fish transitions into dormancy; Reeve et al., 2022) (Fig. 1). We studied cunner (Tautogolabrus adspersus), a temperate wrasse with a striking winter dormancy involving inactivity and fasting below their population-specific WDTT of 5–8°C (Bradbury and Green, 1997; Reeve et al., 2022; Speers-Roesch et al., 2018). Focusing on exercise and metabolic performance, we completed two experiments to assess acute and acclimated thermal performance in cunner from Nova Scotia (WDTT=7–8°C; Reeve et al., 2022): (1) swim speed performance (C-start burst escape response and swim speed during exhaustive chase), and (2) whole-animal exhaustive exercise endurance (time to exhaustion from chase), whole-animal aerobic performance (MMR and aerobic scope), exercise recovery [excess post-exercise O₂ consumption (EPOC) and its duration] and underlying muscle metabolic responses (aerobic and anaerobic enzyme activities, fuel metabolites). We used the thermal sensitivity quotient (Q₁₀), which describes the temperature dependency of a biological rate (Clarke, 2017), to evaluate whether thermal constraints on performance drive winter dormancy. Thermal sensitivity of metabolism and locomotor performance in ectotherms is normally associated with Q₁₀=1–3, whereas a higher Q₁₀ suggests a constraint (Clarke, 2017; Jørgensen et al., 2022; Peck, 2016; Seebacher et al., 2015). We predicted that greater decreases in performance, and consequently higher Q₁₀ (specifically, and conservatively, >3), would occur with cooling below the WDTT than above it (where conservatively, we expected Q₁₀ <3). We measured exercise recovery to test the prediction that potential thermal constraints on aerobic performance would translate into disproportionately greater EPOC (i.e. a greater cost for recovery) and/or longer recovery time in the cold. Finally, by comparing performance levels and Q₁₀ values of cunner acclimated to cooled temperatures with those of warm-acclimated (14°C) cunner acutely exposed to the same cooled temperatures, we evaluated whether compensation via plasticity (i.e. increased performance and reduced thermal sensitivity following acclimation) occurred below the WDTT; we expected to find no evidence of compensation (i.e. unchanged performance level and unchanged Q₁₀ between acute and acclimated exposures) (Fig. 1). Acute and acclimated thermal responses are rarely compared within the same study, despite the value for revealing plasticity (Havird et al., 2020). By studying dormant species such as cunner, we can provide insight into the origins of divergent thermal responses to cold among fishes.

Juvenile cunner [Tautogolabrus adspersus (Walbaum 1792)] were obtained from the Cooke Aquaculture captive breeding program at the Huntsman Marine Science Centre in St Andrews, New Brunswick, in September 2017 and June 2018 (F1 offspring of wild-caught parents from St Mary's Bay, Nova Scotia; see Table S1 for fish body masses and lengths for each experiment). Fish were transferred to 165 litre fiberglass circular tanks supplied with aerated, filtered seawater from a recirculating seawater system (14±0.5°C) at the University of New Brunswick Saint John. Fish were fed every second day to satiation with commercial pellets (Gemma Diamond 1.8 mm; Skretting, St Andrews, NB, Canada). The fish were kept on a winter photoperiod typical of Atlantic Canada (10 h:14 h light:dark), which included a simulated sunrise and sunset (30 min each) with sunrise/sunset lamps (WADEO, Brooklyn, NY, USA) to minimize disturbances from sudden light changes. This winter photoperiod also was used in all experiments. Cunner remain active and feed under a winter photoperiod, so long as the temperature is above their WDTT of 7–8°C (Reeve et al., 2022). Cunner were acclimated to these holding conditions for at least 6 weeks before experiments began.

We ran two experiments, outlined below, using a common temperature exposure protocol for measuring thermal performance which involved acutely challenging (3°C h⁻¹) 14°C-acclimated cunner to a broad range of temperatures (26–2°C in 3°C increments, with a different group of 14°C-acclimated fish acutely exposed to each temperature) as well as acclimation (5–8 weeks, after a gradual 0.5°C day⁻¹ cooling) of cunner to a range of temperatures relevant to the summer to winter transition (14–2°C in 3°C increments, with a different group of fish acclimated to each temperature) (see ‘Experiment 1’ for more information, and ‘Experiment 1’, ‘Experiment 2A’ and ‘Experiment 2B’ for experiment-specific details). The slower acclimation cooling rate of 0.5°C day⁻¹ and up to 8 weeks of acclimation better approximates seasonal cooling while keeping acclimation duration logistically feasible (Huey and Buckley, 2022). Temperatures of 14°C and 2°C are common summer and winter ocean temperatures, respectively, encountered by cunner across their range. The upper and lower thermal limits of cunner are roughly 28°C (Kelly et al., 2014) and 0°C (Green and Farwell, 1971). Thus, our acute temperature exposure range (26–2°C) captures nearly the entire thermal tolerance range of cunner, and our lowest acute/acclimated temperature of 2°C falls at the extreme cold end of their thermal performance curve. A broader range of temperatures was used for acute exposures to characterize the acute thermal performance curve for each performance metric including T_opt and thermal sensitivity over a broader range of warm, active temperatures (see ‘Thermal sensitivity analysis’ section); the acclimation exposures, which are more space- and time-intensive, focused on the temperatures that equally bracket the WDTT (±6°C around the WDTT of approximately 8°C; Reeve et al., 2022). Although a fully factorial design involving acute exposures of each acclimation group to all other test temperatures provides the most robust assessment of thermal plasticity (Havird et al., 2020), this is logistically challenging with relatively large fish such as cunner. Thus, we used our simplified multi-temperature design involving comparisons of 14, 11, 8, 5 or 2°C-acclimated fish with 14°C-acclimated fish acutely exposed to the same acclimation temperatures; the difference in the measured variable between the acute and acclimated exposures at the same temperature inferred the extent of plasticity.

Cunner were transferred from holding tanks to dedicated experimental aquarium systems for temperature acclimation prior to physiological performance measurement. Each acclimation system corresponded to a specific acclimation temperature and consisted of a recirculating seawater system that supplied four glass aquaria per system (or six aquaria at 14°C to provide room for the additional fish required for acute temperature exposures). Each aquarium contained 60 litres of seawater and three PVC pipe shelters (each 10×2.5 cm, length×diameter). Fish were equally distributed among aquaria (14 fish per tank initially); later, for acute or acclimated performance measurements, fish were taken from randomly selected aquaria from the 14°C-acclimated group (for measurement of acutely exposed fish) or from each acclimation temperature (for measurement of acclimated fish). The seawater within each system was temperature controlled to ±0.5°C with a 1/3 HP Arctica chiller (JBJ Chillers, St Charles, MO, USA) and water quality was maintained via active biomedia, mechanical filtration, constant aeration and occasional seawater replacement (Arctica water chillers also were used for all experiments described below). All acclimation systems were housed in the same wet lab room. Fish were fed every second day to satiation on commercial pellets (but rarely ate below ∼8°C) and fasted for 48 h before measurements of performance. All experimental procedures were approved by the animal care committee of the University of New Brunswick in compliance with Canadian Council of Animal Care guidelines.

Swim speed performance was assessed using two metrics: (1) instantaneous maximum burst velocity [U_max; body lengths (BL) s⁻¹], during the anaerobic C-start escape response (Domenici and Blake, 1997), and (2) annular chase swimming speed (U_chase; BL s⁻¹), which includes aerobic and, especially, anaerobic swimming and was achieved using a variation of the standardized chase method where fish were chased in laps around a circular raceway until exhaustion (Norin and Clark, 2016). We measured U_chase, as opposed to U_crit (maximum sustained swimming speed), because it allows assessment of swim velocity performance in species that swim poorly in swim tunnels (Norin and Clark, 2016). During preliminary trials with cunner, we found it difficult to encourage them to swim continuously in a tunnel, similar to other temperate wrasse species (Yuen et al., 2019).

Different groups of cunner were acclimated to either 14, 11, 8, 5 or 2°C for 5–6 weeks (n=12 per temperature; see Table S1 for body masses and lengths), and another group of cunner was acclimated to 14°C and then acutely exposed to either 26, 23, 20, 17, 14 (as a sham control for the acute exposure procedure), 11, 8, 5 or 2°C (n=12 per temperature; see Table S1 for body masses and lengths). Each acclimation system (described above) was cooled at a rate of 0.5°C day⁻¹ until the acclimation temperature was reached (excluding 14°C), with cooling start dates staggered so all acclimation groups reached their acclimation temperature on the same date. Acute trial fish were taken from the 14°C acclimation system and cooled or warmed to their test temperature at 3°C h⁻¹ (over a fixed 4-h period, e.g. 8°C-exposed fish were held at 14°C for 2 h and then cooled over 2 h; 14°C-exposed fish were held at 14°C for 4 h, etc.) in a pre-test holding system adjacent to the test arenas (see below); the fish then remained at their test temperature for 1 h before the swim test (total of 5 h before the test). Acclimated fish were also placed in the same pre-test holding system for a 5-h period at their respective acclimation temperature prior to swim tests. The pre-test holding system consisted of a 75 litre glass aquarium (where fish were held) that drained into a sump (55×42×20 cm, diameter×height×water depth) containing a submersible pump that transferred aerated temperature-controlled seawater back to the glass aquarium.

Each fish went through two swimming speed tests beginning with the C-start test (to measure U_max) followed by the annular chase test (to measure U_chase). C-start trials were carried out in a large fibreglass tank (30×85 cm, depth×diameter) filled to 15 cm depth with aerated temperature-controlled seawater (the C-start arena). The seawater in the arena drained to an insulated sump and seawater was recirculated back to the arena via a chiller by a submersible sump pump. The pump had a bypass back to the sump to ensure water flow into the arena was low and seawater inflow to the arena was removed during the trial (i.e. static system during C-start attempts). A high-speed digital video camera (240 frames s⁻¹, Olympus Tough TG-5; Olympus, Shinjuku, Tokyo, Japan) was mounted 115 cm above the setup by attaching a flexible tripod to a PVC pipe frame. A white corrugated flat plastic sheet marked with a line of known distance (3 cm; for distance calibration) was fitted to the bottom of the circular arena to provide the fish with good contrast. A PVC pipe (75×3.8 cm, length×diameter) was fastened vertically to the inside of the tank, into which a weighted Falcon tube (38 g) could be dropped into the tank water to provide a mechano-acoustic stimulus to startle the fish and elicit a C-start response. A 65 cm long string was tied to the Falcon tube to ensure only the tube tip entered the water and which allowed inconspicuous retrieval of the tube before the next trial. This entire system was enclosed by black plastic bags to minimize outside disturbances and eliminate glare on the water from overhead lights.

Prior to beginning the C-start test, individual fish were transferred via net immersed in seawater from the pre-experiment holding system to the C-start arena and allowed 30 min to adjust to their surroundings. Then, once the fish was within ∼40 cm (radius of tank) of the PVC pipe, video recording commenced followed by dropping of the weighted Falcon tube (a mechano-acoustic stimulus). An attempt was deemed successful if the fish elicited a C-start response and a total of three attempts using the weighted tube drop were completed (regardless of a successful C-start) with 5 min rest periods between each (Domenici and Blake, 1997; Marras et al., 2011). During preliminary tests at 2°C, some fish did not react to the tube drop during the initial three attempts; therefore, an additional two attempts were employed on all fish at all temperatures whereby 20 cm long forceps were used to pinch the fish's caudal fin quickly and gently. At all temperatures, this procedure elicited a C-start, though some fish would react and burst before contact; in either scenario, any identified C-starts were analyzed. Subsequent statistical analysis showed no significant effect of stimulus type on the C-start variables of U_max or maximum acceleration (A_max) (P=0.184 and 0.655, respectively; ANOVA). Pre-simulation variables including distance to the arena wall, distance to the stimulus, and angle of the fish relative to the stimulus are provided in Table S2. Distance to stimulus or wall were not significantly different (P=0.152 and 0.159, respectively; ANOVA) between acute and acclimated treatments, and thus were unlikely to influence our evaluation of acclimation effects. Distance to stimulus varied significantly (P<0.05) across temperatures owing to the increased need to use forceps (which approached the fish more closely) to stimulate fish at the coldest temperatures. Distance to wall also varied significantly across temperatures, owing to higher values at the coldest temperature; however, successful C-starts occurred in the direction away from the wall, so it is unlikely to have impeded the performance. Furthermore, if any effect occurred, a greater distance to wall would likely result in higher U_max or A_max, yet the coldest fish still showed the lowest burst performance; thus, at worst, we conservatively estimated the cold constraint. Finally, stimulus angle varied significantly between treatments as well as across temperatures, again because of increased forceps use at the coldest temperatures, especially in acutely cooled fish; forceps were directed at the tail (180 deg stimulus angle) so resulted in higher values for stimulus angle. Because forceps stimulation was necessary at colder temperatures, the effect of their use on pre-stimulation variables was unavoidable, but is unlikely to have had meaningful effects on our measurements of U_max and A_max for the reasons above and also because U_max or A_max were not affected by stimulus type and C-starts are stereotyped, neuromuscular reflexes once triggered.

Following the last C-start attempt, the fish was transferred via net in seawater to the adjacent annular swim arena and allowed a 30 min recovery period before measuring U_chase. The annular swim arena consisted of a round bucket (55×42 cm, diameter×height) containing aerated, temperature-controlled seawater, with a smaller white bucket (24×21.5 cm, diameter×height; weighted with seawater) centered within it; this created a circular loop between the walls of the two buckets where the fish could swim (dimensions of the swimming zone: 15.5×176×20 cm, width×circumference×water depth). The submersible pump recirculating seawater to a chiller was removed following the recovery period so that during the test the fish could swim around the circular loop without obstacles. A manually operated PVC pipe (80×1.9 cm, length×diameter) was used to frequently make contact with the fish's caudal fin to elicit bursts and sustained swimming until the fish no longer responded to five contacts, which we defined as exhaustion. Each test was recorded using a digital video camera (Olympus Tough TG-5; Olympus, Shinjuku, Tokyo, Japan) mounted above using a tripod, and the videos were analyzed later for total laps swam over the time from start of chase to time to exhaustion.

The C-start U_max and A_max for each fish was measured from the high-speed videos of C-start responses using the automated tracking software Tracktor (Sridhar et al., 2019; https://github.com/vivekhsridhar/tracktor). The video analysis software Tracker (v5.1.5; https://physlets.org/tracker/) was used for videos where poor contrast made it impossible for Tracktor to detect the fish. Comparisons of the output from the same videos analyzed with both software confirmed that measured parameters were comparable in both cases. Pixel-to-distance calibrations were done in each software using the 3 cm line drawn on bottom of arena. To measure U_max and A_max, a common time period was analyzed for each fish (75 ms) following initiation of the C-start; this corresponded to the average time for each fish between initiation of stage 1 of the C-start response until the completion of stage 2 (Domenici and Blake, 1997). In Tracker, we used point mass analysis with the center of the fish's head between the eyes set as the reference point. Measurement of U_max and A_max was carried out for each C-start obtained from the 5 attempts per fish, and the highest value of U_max and the associated A_max were recorded (i.e. both values were from the same attempt). U_max was measured in cm s⁻¹ and standardized to the individual's body length (and reported as BL s⁻¹) and A_max was measured in cm s⁻² and reported as m s⁻².

We also measured C-start responsiveness for each fish, which was calculated as the percentage of total attempts that elicited a C-start response to the weighted tube drop or approach of the forceps during the pinch attempt (i.e. prior to direct tactile contact). By only considering C-starts elicited prior to contact with the fish to be successful responses, our measure of responsiveness reflects visual and/or motion sensing rather than tactile responsiveness. In fact, all fish that were contacted with the forceps responded with a C-start, indicating the robustness of tactile stimuli in triggering a C-start regardless of temperature exposure.

Annular chase swimming speed (U_chase) was defined as the total number of laps swam per unit time over the entire chase duration (i.e. until time of exhaustion) standardized to body length (BL s⁻¹).

We used the standardized chase method followed by respirometry (Norin and Clark, 2016) to assess: (1) exhaustive exercise endurance, quantified as duration to exhaustion from chase, and (2) aerobic performance, quantified by measuring oxygen consumption rate immediately after chase (i.e. MMR) until recovery back to resting levels (i.e. SMR), allowing estimation of aerobic scope as well as EPOC. We assessed the thermal sensitivity of exhaustive exercise endurance and aerobic performance in cunner following the same acute and acclimation protocols described for experiment 1 but using different fish and with an acclimation time of 5–7 weeks (n=11–16; see Table S1 for body masses and lengths). In brief, individual fish were transferred via net immersed in seawater from their respective thermal acclimation tank into the pre-test holding system, where they were either (1) cooled or warmed (3°C h⁻¹, using 14°C-acclimated fish) to their test temperature between 26 and 2°C at which they remained for 1 h prior to testing (acutely exposed fish), or (2) kept at their acclimation temperature (14, 11, 8, 5 or 2°C) for 5 h prior to testing (acclimated fish). Then, each fish was transferred via net in seawater to the chase arena: a round bucket (55×42×20 cm, diameter×height×water depth) filled with aerated seawater at the same test temperature. After a 15 min adjustment period in the arena, the fish was manually chased by hand until exhaustion, which was defined as no longer bursting away after five gentle caudal fin pinch attempts, and the time to exhaustion was recorded.

Aerobic performance was determined by estimating SMR and MMR via measurement of oxygen consumption rate (Ṁ_O2; mg O₂ h⁻¹ kg⁻¹) using automated intermittent-closed optical respirometry. Following exhaustion from chase, fish were immediately transferred by hand from the chase bucket into a sealed respirometer (∼160 ml volume, glass container with plastic lid fitted with O-ring) submerged within a larger clear plastic water bath (70×41×16 cm, length×width×depth). Each respirometry trial consisted of four fish run simultaneously in individual respirometers. The water bath was supplied with temperature-controlled (at respective test temperature) seawater from an insulated sump via a submersible pump. Each respirometer was fitted with an oxygen probe to measure the within-chamber temperature-compensated oxygen level using a 4-channel FireSting (PyroScience, Aachen, Germany). The respirometer water was mixed with stir bars separated from the fish by plastic mesh and driven by a 6-channel magnetic stir plate placed underneath the water bath. The flush cycles were controlled by a smaller Eheim 600 pump powered through a commercially available programmable digital timer set to temperature-specific open–close intervals. The flush period was always 3 min (except 2 min at 26°C), which ensured full re-oxygenation of the respirometer water, and the closed period was modified depending on experimental water temperature because the slope of oxygen decline (metabolic rate) was lower at cooler temperatures (5 min for 26 and 23°C, 6 min for 20°C, 7 min for 17°C, 9 min for 14 and 11°C, 17 min for 8°C, 27 min for 5°C, and 42 min for 2°C). Fish were allowed to recover from the chase exercise in the respirometers overnight and were removed the following morning (∼14–18 h total recorded Ṁ_O2), allowing estimation of MMR, SMR and EPOC. Background respiration rates in the respirometers were calculated after removing each fish and recording oxygen change over three closed cycles. Once background measurements were recorded, the system was cleaned with dilute bleach and rinsed twice with freshwater before re-use.

MMR was calculated using the maximum slope of oxygen decline in a 60-s period following exhaustive exercise. This maximum slope was identified by an iterative approach in which 60-s slopes were taken sequentially in 30 s increments from the moment exercised fish were placed in the respirometer (i.e. 0–60, 30–90, 60–120 s, etc.) over the first 15 min of closed periods; the highest slope always occurred within the first 4.5 min post-exercise (with 94% of MMR occurring in the first 2.5 min). Subsequent to the closed period in which MMR occurred, Ṁ_O2 values were calculated using the entire slope of oxygen decline during each entire closed cycle (excluding the first 120 s and last 60 s to ensure equilibration following flush and avoid sampling the beginning of the flush period) until the fish was removed from the respirometer. SMR was calculated as the average of the lowest 10% slopes of oxygen decline, excluding any outliers (outside 2 standard deviations of the average of the lowest 10% slopes). In the absence of simultaneous measurement of activity, the average of a lowest subset percentage of Ṁ_O2 values is a common method to estimate SMR in fishes (Chabot et al., 2016). Absolute aerobic scope (AAS) was calculated for each fish by subtracting their SMR from their MMR.

To calculate EPOC and recovery time, Ṁ_O2 of each fish following its MMR was more finely calculated by taking the slopes of oxygen decline over sequential 60-s periods within each closed period (excluding the first 120 s and last 60 s of a cycle) until 4 h post-MMR; then the oxygen decline was measured over a 3 min period every 40–50 min (depending on the temperature) until the fish was removed from the respirometer the following morning. During the 4-h post-MMR period, we filtered the data to remove Ṁ_O2 values at time points where Ṁ_O2 values were not available for all fish (at certain time points, some fish were in a flush period and thus without a measured Ṁ_O2, because fish were chased and introduced to their respirometers over time within a given closed cycle). Thus, a Ṁ_O2 recovery curve was generated for each fish. Recovery time was defined as the time post-MMR at which the first of three consecutive Ṁ_O2 values were equal to SMR±10%. Total EPOC for each fish was determined by integrating the area between the recovery curve and SMR, i.e. from the time of MMR to the recovery time (Lee et al., 2003).

Additional cunner acclimated for 7–8 weeks to 14, 8 and 2°C during experiment 2A were exhaustively chased following the same protocol described in experiment 2A, but instead of being transferred to a respirometer, were killed via cervical dislocation (n=10 per temperature; see Table S1 for body masses and lengths) for measurement of energy metabolite contents in white muscle. Following euthanasia and body size measurement, a white muscle steak (∼1 cm thick, anterior of the caudal peduncle) was flash-frozen in liquid nitrogen. Additionally, non-exercised fish that had been either acutely cooled or acclimated for 7–8 weeks to 14, 8 and 2°C were sampled in the same manner to serve as resting controls for metabolite measurements. Measurements of metabolic enzyme activities were also made on chased fish from each acclimation temperature. Muscle was stored in a −80°C freezer, then ground under liquid nitrogen with a mortar and pestle and stored at −80°C until enzyme and metabolite assays. All chemicals used in the enzyme and metabolite assays were purchased from Sigma-Aldrich (St Louis, MO, USA).

Frozen ground white muscle (47–60 mg) was homogenized in ice-cold homogenization buffer (50 mmol l⁻¹ imidazole, 1 mmol l⁻¹ disodium EDTA, 0.1% Triton X, pH 7.4) using a polytron-type homogenizer (BioSpec Products, Bartlesville, OK, USA) with three 5-s bursts at the highest speed setting. We added 15 volumes of buffer when samples weighed <25 mg and 10 volumes of buffer when samples weighed >25 mg. Homogenates were spun in a microcentrifuge at 1000 g for 5 min at 4°C. The supernatant was diluted and used to measure the activities of lactate dehydrogenase (LDH), citrate synthase (CS) and creatine kinase (CK). We measured LDH activity to assess anaerobic glycolytic capacity, CS activity as an indicator of aerobic capacity, and CK activity given its importance for rapid substrate-level generation of ATP during burst exercise (Guderley et al., 2001).

We measured maximal enzyme activities at a common temperature (25°C) with a SpectraMax 190 microplate spectrophotometer (Molecular Devices, Sunnyvale, CA, USA) by following the oxidation or reduction of pyridine dinucleotides at 340 nm [millimolar coefficient 6.22 (mmol l⁻¹)⁻¹ cm⁻¹] (LDH and CK) or the appearance of 5-thio-2-nitrobenzoic acid resulting from the reaction of liberated free CoA with 5,5′-dithiobis (2-nitrobenzoic acid) (DTNB) at 412 nm [millimolar coefficient 13.6 (mmol l⁻¹)⁻¹ cm⁻¹] (CS) over a 10-min reaction period. The assay conditions were as follows (in mmol l⁻¹): LDH=50 imidazole, pH 7.4, 0.2 NADH and 1 pyruvate (omitted for control); CK=50 imidazole, pH 7.4, 1 ADP, 10 AMP, 0.3 NADP, 4 d-glucose, 5 MgCl₂, 2 U ml⁻¹ hexokinase, 2 U ml⁻¹ glucose-6-phosphate dehydrogenase and 25 creatine phosphate (omitted for control); and CS=50 imidazole, pH 7.4, 0.3 acetyl-CoA, 0.1 DTNB and 0.5 oxaloacetate (omitted for control). Substrates were at saturating levels and optical pathlength was calculated for a final microplate well volume of 200 µl. Enzyme activity was measured in duplicate and run simultaneously with a single control reaction with omitted substrate to correct for non-specific background reactions. Preliminary measurements confirmed the reaction rates to be linear with time and proportional to the amount of homogenate added.

Enzyme activities were calculated as nmol substrate converted to product per minute per milligram tissue protein (nmol min⁻¹ mg⁻¹ protein) as well as µmol substrate converted to product per minute per gram wet weight (µmol min⁻¹ g⁻¹ ww). Enzyme activities per gram wet weight indicate the physiologically relevant enzymatic capacity of the tissue, while enzyme activities per milligram protein allow us to correct for potential influences of variable tissue protein content. Protein content of the enzyme supernatant was measured using the Bradford assay.

Approximately 90 mg of frozen ground muscle was homogenized in 10 volumes of 8% perchloric acid using three 5-s bursts of a polytron-type homogenizer (BioSpec Products) at its highest setting. Neutralized extracts were thawed on ice and assayed for ATP, creatine phosphate, lactate and glycogen according to well-established protocols (Bergmeyer, 1983).

We used R for all analyses (version 4.0.2, www.r-project.org). Assumption of normality was tested by examining normality plots and homogeneity of variances were tested using Levene's test of equality of variances. Statistical significance was accepted at P<0.05 and all results values are presented as means±s.e.m., unless otherwise stated.

The effects of temperature (14, 11, 8, 5 or 2°C), acclimation status (acutely exposed or acclimated) and their interaction on each performance metric were tested using a two-way ANOVA with Tukey HSD post hoc tests. These two-way ANOVAs of acute and acclimated performance between 14 and 2°C evaluated the constraints of cold on performance and tested our prediction that dormant cunner would lack compensation via plasticity (i.e. no difference between the acute and acclimated performance level at a given temperature). Additional acute temperature exposures above 14°C were also carried out; the effect of acute exposure to temperature (26, 23, 20, 17, 14, 11, 8, 5 or 2°C) on each performance metric was separately assessed using a one-way ANOVA. These one-way ANOVAs of acute thermal performance were used to characterize the performance of cunner over a broader temperature range, including approximation of optimal temperatures. The effect of acute and acclimated temperature interval (acute 14–20°C, acute 14–8°C, acute 8–2°C, acclimated 14–8°C and acclimated 8–2°C) on the mean Q₁₀ of all performance rate metrics and the mean percent change of all performance metrics was determined using individual one-way ANOVAs and Tukey HSD post hoc tests. The relationship between exhaustive exercise effort and total EPOC was determined using linear regressions for both acclimated and acute groups from 14 to 2°C, as that was the temperature range shared between acclimated and acute exposures. The effect of acclimation temperature (14, 8 and 2°C) on LDH, CS and CK activity in white muscle of cunner was tested using one-way ANOVAs and Tukey HSD post hoc tests. The effects of temperature (14, 8 and 2°C), exercise treatment (acute resting, acute post-exercise, acclimated resting and acclimated post-exercise), and their interaction on white muscle metabolite contents were tested using individual two-way ANOVAs and Tukey HSD post hoc tests for each parameter.

The responses of the measured variables to acute or acclimated temperature exposures are shown in Figs 2–7; the corresponding thermal sensitivities of performance above and below the WDTT are shown in Fig. 8.

U_max and U_chase significantly increased with acute warming of 14°C-acclimated cunner, and a plateau of maximum performance was reached from 17 to 23°C (i.e. a broad T_opt) (P<0.0001; Fig. 2). Cooling from 14 to 2°C caused significant reductions in swim speeds of both acutely exposed and acclimated cunner (P<0.0001; Fig. 2); chase duration and distance swam during the annular swim showed a similar response (Table S3). U_chase was similar between acute and acclimated fish at every temperature (Fig. 2D). U_max was greater on average in acclimated cunner compared with acutely cooled cunner at the same temperature, but this difference was only significant at 2°C, where acclimated U_max was nearly twice as high as acute U_max (Fig. 2A). Acutely exposed and acclimated cunner had similar A_max and responsiveness at all temperatures from 14 to 2°C (Fig. 2B,C). A_max decreased modestly with cooling and was similarly temperature dependent above and below the WDTT for both acutely cooled and acclimated fish (P<0.0001; Figs 2B and 8B; also see Fig. S1). U_max and U_chase following acute temperature change were more thermally sensitive below the WDTT (8–2°C; Q_{10,acute8–2°C}=2.5 and 3.0, respectively) than above (Q_{10,acute14–8°C}=1.6 and 1.1, respectively), but partial compensation accompanied 2°C acclimation (Q_{10,acclimated8–2°C}=1.5 and 2.3 for U_max and U_chase, respectively) (Fig. 8B, Table S4; also see Fig. S1). In contrast, responsiveness plummeted from ∼80% to ∼20% successful C-starts between 8°C and 2°C, with little improvement following acclimation (i.e. −86.4% versus +0.0% acute change or −67.3% versus −16.1% acclimated change for 8–2°C versus 14–8°C, respectively) (P<0.0001; Figs 2C and 8D, Table S4).

Exhaustive exercise endurance (i.e. exhaustive chase duration; Fig. 3) and aerobic performance metrics (Fig. 4) significantly decreased with cooling below their T_opt range in both acclimated and acutely exposed cunner (P<0.0001). Based on the plateaus of maximum performance of acutely exposed cunner, the T_opt range was approximately ∼11–17°C for exhaustive exercise endurance and 20–23°C for aerobic performance. At acute exposure temperatures warmer than T_opt, chase duration and aerobic performance decreased. SMR consistently decreased with temperature (Q₁₀ of 1.8–3.1) and was similar between acclimated and acute fish at each temperature (Fig. 4A, Table S4). The percent change in exhaustive chase duration and the thermal sensitivities of MMR and AAS in cunner acutely exposed to temperature change were greater below the WDTT than above the WDTT (Q_{10,acute8–2°C}=3.8–4.0 versus Q_{10,acute14–8°C}=1.9 and Q_{10,acute14–20°C} =1.4–1.5) (Figs 8B,D, Table S4; also see Fig. S1). Above the WDTT, acclimated and acutely cooled fish had similar temperature dependency for exhaustive chase duration, MMR and AAS (Q_{10,acclimated14–8°C}=1.4–1.5) (Fig. 8B,D, Table S4). At 2°C, but at no warmer acclimation temperature, exhaustive chase duration, MMR and AAS of acclimated fish was significantly higher compared with performance of acutely cooled fish (P<0.001; Figs 3 and 4). This occurred despite the MMR and AAS of acclimated cunner at 11 and 14°C happening to be significantly lower than acute cunner (P<0.01); furthermore, at 5 and 8°C, MMR and AAS were not different between acclimated and acute fish, and chase duration was not different between acute and acclimated fish at any temperature except 2°C (P>0.05; Figs 3 and 4). Below the WDTT (8–2°C), acclimated cunner had a much smaller constraint on exhaustive chase duration (Fig. 8D, Table S4) and markedly lower Q₁₀ values (similar to those above the WDTT) for MMR and AAS compared with acutely cooled cunner (Q_{10,acclimated8–2°C}=1.9–2.0 versus Q_{10,acute8–2°C}=3.8–4.0) (Fig. 8B, Table S4).

Following exhaustive chase, the metabolic rates of fish acclimated or acutely exposed to each temperature were elevated before gradually recovering to SMR±10% by 3.9–7.9 h post-MMR (Fig. 5A–D). Total EPOC decreased with cooling below 11–14°C (while being unusually elevated at the two warmest acute temperature exposures), but recovery time was relatively similar across temperatures; both were statistically similar between acclimated and acutely exposed cunner at all temperatures (Fig. 5C,D). Exhaustive exercise effort (exhaustive chase time×U_chase) was significantly positively correlated with total EPOC in both acutely cooled and acclimated cunner across their common test temperatures of 14–2°C; the slopes differed significantly from one another (Fig. 6). The thermal sensitivity of total EPOC above and below the WDTT showed a similar pattern to that described above for MMR and AAS (Fig. 8B, Table S4; also see Fig. S1).

LDH and CK enzyme activities were not significantly affected by acclimation temperature, regardless of whether expressed as a function of protein content or tissue capacity (Fig. 7A,D and C,F). CS enzyme activity also was unaffected by temperature, although in 2°C-acclimated fish it was 75–79% higher on average compared with 14°C-acclimated fish (Fig. 7B,E).

Exhaustive chase durations of fish used for metabolite measurements in experiment 2B (Fig. S2A) were similar to those for experiment 2A (Fig. 3), including similar effects of acute and acclimated temperature exposures (specifically, significant compensation at 2°C). Temperature and the treatment×temperature interaction significantly affected glycogen content (P<0.0001) (Fig. S2B). Resting glycogen was significantly lower in cunner acclimated to 2°C compared with 8 or 14°C (P=0.01 or 0.03, respectively). Glycogen levels were similar pre- and post-exercise at all acute or acclimated temperatures, but showed a decreasing trend following exercise at 14°C, likely explaining the temperature×treatment interaction. Lactate content was significantly affected by temperature, treatment and their interaction (P<0.0001, 0.0001 and 0.001, respectively) (Fig. S2C). Resting lactate levels were similar across temperature and treatment groups. In acutely cooled cunner, mean lactate contents were higher post-exercise compared with rest at every temperature, but only significantly at 14°C (P=0.002). In acclimated cunner, exhaustive exercise increased lactate content at 14 and 8°C, but not 2°C (P<0.0001 and P=0.048 for 14 and 8°C, respectively). Exhaustive exercise significantly affected creatine phosphate (CrP) stores (P<0.0001) (Fig. S2D). CrP levels were always lower following exercise, although only significantly in 2°C-acclimated cunner (P=0.013). Resting and post-exercise CrP contents were unaffected by temperature or acute/acclimated exposure.

Our comprehensive analysis of thermal exercise and metabolic performance of cunner did not support the cold performance limitation hypothesis for winter dormancy. Instead, we revealed for the first time that partial thermal compensation of performance via acclimation occurs even in a winter-dormant fish. As is typical for ectotherms (Johnston and Temple, 2002; Schulte, 2015), the burst escape, exhaustive exercise and aerobic performance of cunner declined with cooling below their respective thermal optima (∼17–23°C for cunner based on our acute exposures). Acute exposure or acclimation to temperatures above the WDTT of 7–8°C (Reeve et al., 2022) was associated with comparable, typical Q₁₀ values for teleost performance rates of 1–3 (Q_{10,acute14–8°C}=1.1–1.9, mean=1.5; Q_{10,acute14–20°C}=0.9–2.1, mean=1.5; Q_{10,acclimated14–8°C} =1.2–2.9, mean=1.7) (Fig. 8A,B) (Clarke, 2017; Jørgensen et al., 2022; Seebacher et al., 2015). Below the WDTT, the acute thermal sensitivities of performance were significantly higher (Q_{10,acute8–2°C} range=1.5–4.9, mean=3.3), indicating a thermal constraint from rapid cooling consistent with the cold performance limitation hypothesis. However, contrary to our hypothesis, performance never neared zero and actually was enhanced following multi-week acclimation at 2°C, abolishing the acute constraint and resulting in lowered Q₁₀ values not significantly different from the typical Q₁₀ values across active temperatures above the WDTT (Q_{10,acclimated8–2°C}=1.4–3.0, mean=2.0) (Fig. 8A). Analysis of the percent decreases of all performance metrics supported the Q₁₀ analysis, with a greater acute constraint below the WDTT (8–2°C) that was blunted following 2°C-acclimation (Fig. 8C,D). Thus, our findings indicate partial thermal compensation and relatively unconstrained performance in cunner when acclimated to a typical, dormant winter temperature of 2°C. Because acclimation has a prolonged timescale more relevant to the gradual seasonal cooling associated with winter onset for wild cunner (Costa et al., 2013), our results provide strong evidence against the cold performance limitation hypothesis. Overall, winter dormancy in a model species, the cunner, does not appear to operate as a behavioural filter (Garland and Losos, 1994) to mitigate thermal constraints on metabolic or exercise physiology.

The C-start burst escape response (e.g. U_max) and exhaustive exercise performance (e.g. U_chase and exhaustive chase duration) are critical for predator evasion and foraging success in fishes (Marras et al., 2011; Videler and Wardle, 1991). In cunner, U_chase and exhaustive chase duration were constrained (Q₁₀>3) by acute cooling below the WDTT, but were strongly compensated following 2°C acclimation and otherwise had Q₁₀ typical for swimming (e.g. U_crit) and metabolic-linked performance traits in fishes (i.e. 1–3). U_chase and exhaustive chase duration have been rarely measured in fishes, limiting interspecific comparison of their thermal sensitivities. U_max and A_max in cunner had low thermal sensitivity (Q₁₀=1.3–1.6) in acclimated animals or following acute temperature change above the WDTT, which is typical for acclimated or acutely exposed fishes, including in tropical, winter-active and winter-lethargic species, as well as a previous study on cunner (Beddow et al., 1995; Claireaux et al., 2006; Johnson et al., 1996; Moran et al., 2019; O'Steen and Bennett, 2003; Temple and Johnston, 1998). Acute cooling of cunner below the WDTT was associated with a greater thermal constraint on U_max (Q_{10,acute8–2°C}=2.5); similarly, Q₁₀ values >2 have been observed in tropical damselfishes acclimated to stressful temperatures cooler than their latitudinal range limit and in cold temperate sculpins acutely exposed to an unusually frigid 0.8°C temperature (Djurichkovic et al., 2019; Temple and Johnston, 1998). Thus, extreme cold exposure can constrain C-start performance in fishes regardless of overwintering strategy. Importantly, compensation of cunner U_max enabled robust cold performance following winter-relevant 2°C acclimation. Compensatory thermal plasticity of the C-start escape response is common among cold-active fishes (Johnston and Temple, 2002), but here we show that this extends to a winter-dormant fish, alongside compensation of exhaustive exercise performance.

The C-start also involves important non-locomotor components (e.g. responsiveness to a predator) (Marras et al., 2011). Unlike the C-start locomotor components (e.g. U_max), responsiveness to the C-start stimulus in cunner was greatly constrained by cold and was uncompensated following acclimation. Extreme cold can cause impaired sensorimotor function and lethargy in fishes, but usually only as they reach their lower critical limit or following large acute cooling (Friedlander et al., 1976; Preuss and Faber, 2003; Reid et al., 2022). C-start latency (time between stimulus and C-start) has been found to increase with cooling in fishes (Morgan et al., 2022; Preuss and Faber, 2003), but little is known about thermal effects on responsiveness to stimuli. This is important because not responding to a stimulus can have greater repercussions than a small millisecond change in the latency of a successful C-start. Two previous studies provide comparisons to our results for cunner. First, Johnson and Bennett (1995) reported that only 30% of 35°C-acclimated goldfish responded to C-start stimulus after acute 10°C exposure. Second, following acute exposure of 15°C-acclimated sculpins to 0.8°C, Temple and Johnston (1998) reported that C-start responsiveness of the Arctic-ranging Myoxocephalus scorpius was unaffected, whereas only 40% of the temperate Taurulus bubalis responded; following 5°C acclimation, however, all T. bubalis responded. Thus, cunner are conspicuous in the constraint that winter cold has on their C-start responsiveness, which could elevate predation risk if not for their inactive, sheltering dormancy. Interestingly, all cunner C-started in response to direct touch, even at 2°C, hinting that cold specifically constrains visual or mechanoacoustic senses. Unlike the other escape response metrics, the great cold sensitivity and limited compensation of responsiveness, and underlying sensory pathways, could be an alternative proximate explanation for winter dormancy in fishes.

MMR and AAS of cunner were constrained by acute cooling below the WDTT (Q₁₀=3.8–4.0 from 8 to 2°C versus 1.4–1.9 above 8°C) but were strongly compensated following cold acclimation (Q_{10,acclimated8–2°C}=1.9–2.0), contradicting our hypothesis. In fact, even at dormant temperatures, the acclimated thermal sensitivity of aerobic performance in cunner was similar to that of winter-active fishes, in which acclimated Q₁₀ values fall within the typical range of 1–3 across species-specific summer and winter temperatures (Brett, 1964; Hvas et al., 2018; Mackey et al., 2021; Porter and Gamperl, 2023; Schurmann and Steffensen, 1997; Wolfe et al., 2020). Furthermore, the absolute levels of acclimated MMR and AAS were robust in cunner even at 2°C, being similar to mass- and temperature-corrected rates measured at 2–5°C in non-dormant species with comparable athleticism (Atlantic cod, Gadus morhua; Atlantic killifish, Fundulus heteroclitus) (Claireaux et al., 2000; Healy and Schulte, 2012; Reeve et al., 2022; Schurmann and Steffensen, 1997; Tirsgaard et al., 2015). Little is known about thermal dependency of aerobic performance in winter-dormant species; however, the winter-lethargic Atlantic killifish, like cunner, exhibited constrained aerobic performance with acute cooling and partial compensation following cold acclimation (thus, lowering thermal sensitivity; AAS Q_{10,acute15–10°C}=3.5 and Q_{10,acute10–5°C}=4.4 versus Q_{10,acclimated15–10°C}=2.1 and Q_{10,acclimated10–5°C}=2.3) (Healy and Schulte, 2012). Although MMR and AAS at 11°C and 14°C in cunner were greater in acutely exposed fish compared with acclimated fish, likely a result of interindividual variability similarly observed in another study with a comparable design (Healy and Schulte, 2012), the acclimated fish still had greater MMR and AAS at 2°C, highlighting the compensatory cold acclimation. Overall, contrary to our hypothesis, cold sensitivity of aerobic performance is similar between winter-active and winter-dormant or -lethargic fish, because of a shared ability for partial compensation via plasticity.

SMR Q₁₀ in cunner was ∼2–3, confirming that energy savings in winter-dormant fishes primarily arise from inactivity and passive thermodynamic slowing of SMR, and not from metabolic rate depression (Crawshaw, 1984; Reeve et al., 2022; Speers-Roesch et al., 2018).

The partial cold compensation of whole-animal exercise and metabolic performance was not associated with compensatory increases of white muscle metabolic enzyme activities. However, neither was there a decrease in enzyme activities, consistent with the non-plastic SMR during dormancy. White muscle CK activity was similar between warm- and cold-acclimated cunner, as found in the putatively winter-lethargic three-spined stickleback (Guderley et al., 2001; Ressel et al., 2022), whereas winter-active species (e.g. chain pickerel) have shown compensation (Kleckner and Sidell, 1985). LDH activity tends to decrease or remain unaffected in white muscle of cold-acclimated fishes (Guderley and Gawlicka, 1992; Guderley et al., 2001; Kleckner and Sidell, 1985), similar to our findings in cunner (perhaps the high activity of white muscle LDH makes compensation unnecessary). Other metabolic adjustments must underlie the partial cold compensation of cunner exhaustive exercise endurance. The unapparent cold compensation of white muscle CS activity in cunner contrasts with findings for winter-active (e.g. rainbow trout) and -lethargic (e.g. smallmouth bass) fishes (Battersby and Moyes, 1998; Guderley and Gawlicka, 1992; Guderley et al., 2001; Kleckner and Sidell, 1985; Kolok, 1991; Orczewska et al., 2010). However, whereas dormant cunner are inactive and non-feeding, winter-lethargic species can maintain low levels of activity and feeding (Reeve et al., 2022; Shuter et al., 2012), potentially explaining their enzymatic compensation. Nonetheless, the response of muscle CS activity to cold acclimation can vary even within a winter-lethargic species, the Atlantic killifish, including increases or no effect (Dhillon and Schulte, 2011; Healy et al., 2017). Thus, cold-induced CS responses may be under a dynamic range of influences, as suggested for cytochrome c oxidase (Bremer and Moyes, 2011). Perhaps, the compensation of aerobic performance in cold-acclimated cunner relies more on oxygen supply (e.g. cardiac) adjustments (Keen et al., 2017).

Total EPOC was constrained by acute cooling below the WDTT (Q₁₀=4.9) whereas temperature change above the WDTT or cold acclimation were associated with Q₁₀ values (0.9–3.0) typical for fishes (e.g. 1.7–2.4 in salmonids), indicating a partial thermal compensation of EPOC consistent with the partial compensation of the exhaustive chase duration that preceded EPOC measurement (Brett, 1964; Lee et al., 2003). The strong relationship between total EPOC and exhaustive exercise effort (=U_chase×exhaustive chase duration) in acute and acclimated animals suggests that the thermal sensitivity of EPOC in cunner is determined largely by the level of exhaustive exercise achievable at given temperatures; in other words, lowered EPOC with cooling follows from the temperature sensitivity of exhaustive exercise performance rather than that of EPOC-specific processes. This is further supported by the slopes of effort versus EPOC being significantly different between acute and acclimated cunner, probably, in part, because of the enhanced MMR and AAS seen specifically following 2°C acclimation, which would allow increased exercise effort while avoiding a concomitant increase in EPOC (seen as a rightward shift of the 2°C value following acclimation in Fig. 6). This scenario is also consistent with the lack of lactate accumulation at 2°C compared with increases at other temperatures, which would reduce EPOC. Overall, it is parsimonious that the compensatory increase in exhaustive chase duration and effort in 2°C-acclimated cunner – which was not associated with markedly greater EPOC or lactate accumulation, nor with markedly enhanced use of CrP or glycogen compared with acutely exposed 2°C fish – primarily resulted from the enhanced aerobic performance following acclimation. Improved oxygen supply could prolong exhaustive exercise by slowing CrP depletion and lactate accumulation (meaning that other factors, such as ion imbalance, may be the primary limit on exhaustive exercise in cold, dormant cunner) (Kieffer, 2010). Finally, the comparable recovery time across temperatures likely results from exercise effort and oxygen uptake responding similarly to temperature change (e.g. both were higher at warmer temperatures). Overall, our findings suggest that, even in extreme cold, EPOC in cunner is not, in and of itself, constrained; EPOC simply matches the achievable exhaustive exercise effort and associated cost of recovery, similar to other fishes (Lee et al., 2003; Svendsen et al., 2010).

Contrary to the cold performance limitation hypothesis, winter-relevant cold acclimation (2°C) in the winter-dormant cunner resulted in partial thermal compensation of multiple key metrics of whole-animal metabolic and exercise performance. Consequently, and further opposing our hypothesis, the absolute level of acclimated performance was robust in the cold, while thermal sensitivities for acclimated performance rates were similar across warm active and cold dormant temperatures, and typical for ectotherms (Q₁₀=1–3). Compensatory thermal acclimation of performance is well known among winter-active fish species (Beddow et al., 1995; Guderley, 1990; Hvas et al., 2017; Johnston and Temple, 2002), but now appears as a conserved response among temperate fishes regardless of overwintering strategy, mirroring its surprising retention also in ‘stenothermic’ polar fish (Seebacher et al., 2005). For dormant fishes, thermal compensation could facilitate escape if uncovered by a winter-active predator. Regardless, cunner remain dormant despite strong performance following cold acclimation, strengthening our conclusion that thermal constraints on metabolic and exercise performance cannot explain the origin of dormant, and perhaps lethargic (Reeve et al., 2022), overwintering strategies among fishes. Incidentally, among ectothermic tetrapods, poor metabolic and exercise performance in the cold remains a possible explanation for winter dormancy, considering their well-recognized lethargy at low temperatures (Gregory, 1982); empirical support is lacking, though. Alternative drivers of winter dormancy in fishes should be considered, such as environmental food scarcity or the marked cold sensitivity of sensory responsiveness we found in cunner. Even if constraints underlying winter dormancy are identified, dormancy remains an innovation (sensu Miller et al., 2023), allowing certain fishes to range poleward to exploit the food-rich, warm summers while tolerating the harsh winters with minimal effort. Indeed, winter dormancy would bypass the constraint that thermal dependence of activity is thought to impose on ectotherm cold range limits (Buckley et al., 2012; Stuart-Smith et al., 2017).

Interestingly, thermal compensation in cunner only became apparent at the coldest acclimation temperature of 2°C. In other words, only at 2°C was performance increased and Q₁₀ reduced significantly in acclimated fish relative to acutely exposed fish; this was weak or unapparent at other cold temperatures (5°C, 8°C and 11°C). Similarly, Moran et al. (2020) found that isolated muscle performance in cunner was uncompensated following 5°C acclimation (colder temperatures were not examined), and various compensatory responses underlying cold acclimation in fishes can have distinct onset temperatures (Guderley, 1990). To explain our novel observation, and building upon Guderley (1990), we propose the ‘wait-for-winter’ hypothesis: compensatory cold acclimation responses only intensify once temperature becomes low enough to (1) reliably signal the arrival of a consistent frigid winter environment and/or (2) substantially constrain performance in the absence of compensation (e.g. our 2°C acute exposures). Because acclimation can be energetically costly and temperature fluctuates widely during seasonal cooling, a ‘wait-for-winter’ strategy would optimize the cost of environment–phenotype matching by initiating plasticity only when cue reliability was high (Wilson and Franklin, 2002). ‘Wait-for-winter’ thermal plasticity could be common across temperate fishes with varied overwintering strategies, because the winter environment brings ubiquitous thermal and energetic challenges.

We thank Cooke Aquaculture for donating cunner, Geoff McBriarty and Kelly Cummings-Martell for fish husbandry advice, Sam Savoy and Annie Trembley for fish care, Mia Francis for analytical assistance, Tommy Norin for technical advice, Matt Gilbert for useful discussions, and two anonymous reviewers for helpful comments.