Swimming energetics of Atlantic salmon in relation to extended fasting at different temperatures

Fish endures prolonged fasting periods during migration, while costs of swimming increase in warming environments, which restrains energy budgets. Here, critical and sustained swimming capacities along with metabolic rates and haematological parameters were assessed in Atlantic salmon before and after 4 weeks of fasting at midrange and elevated suboptimal temperatures.

deprivation without suffering detrimental effects owing to lower metabolic demands than in endotherms such as birds and mammals (McCue, 2010;Wang et al., 2006).
In nature, many species of fish will periodically encounter extended fasting periods because of seasonal fluctuations in food supplies in their habitat, migratory behaviours or during reproduction events (Green and Farwell, 1971;Miller et al., 2009;Van Ginneken et al., 2005). To cope during periods of food shortage, fish are able to downregulate metabolic rates to preserve resources (Fu et al., 2005;Hvas et al., 2020;Mehner and Wieser, 1994), which on the biochemical level is facilitated by beneficial changes in gene expressions, enzyme activities and mitochondrial functions in various organs (Bermejo-Nogales et al., 2015;Cassidy et al., 2016;Méndez and Wieser, 1993;Salin et al., 2018). Furthermore, following extended periods of food deprivation, such physiological adjustments can be reversed and stunted growth can be compensated as fish have flexible and indeterminate growth trajectories (Ali et al., 2003;Morgan and Metcalfe, 2001).
Although fish are well adapted to temporal variability in feeding opportunities, long-term starvation will eventually jeopardize survival as physiological functions become compromised. The impact of extended fasting periods will depend on species, initial body condition and various environmental factors. Water temperature is a particularly important environmental factor to consider, as increasing temperatures drastically elevates the metabolic demands in fish (Brett, 1971;Eliason and Farrell, 2016). Warming of aquatic habitats from anthropogenic climate change could therefore make species of fish more vulnerable to naturally occurring extended fasting periods as energy stores gets depleted sooner .
The Atlantic salmon (Salmo salar) is a relevant species to study with regards to the interactions between fasting and temperature on energetic costs and physiological capacities. It is an anadromous migratory species that endures prolonged fasting periods at various phases in its life cycle (Hendry and Beall, 2004;Kadri et al., 1995;Thorstad et al., 2012). Furthermore, it also encounters variable thermal environments ranging from 0-3 • C in its northern distribution (Lacroix, 2013;Reddin, 1985) to presumably above 20 • C occasionally during summer river migrations in its southern limits (Valiente et al., 2011). Although, for the majority of its life during the marine phase, it is found in waters below 10 • C (Jensen et al., 2014;Lacroix, 2013;Reddin, 1985). Growth of Atlantic salmon post-smolts is maximized at 13 • C (Handeland et al., 2003;Handeland et al., 2008) and is reduced above 18 • C owing to lower appetite and feed conversion efficiency (Hevrøy et al., 2015;Kullgren et al., 2013;Wade et al., 2019). Behaviorally, Atlantic salmon will actively avoid environments above 16 • C (Johansson et al., 2009;Lacroix, 2013) and long-term survival is not possible at chronic temperatures above 22 • C (Gamperl et al., 2020;. Populations of Atlantic salmon have long been declining as a consequence of habitat destruction and overfishing (Dadswell et al., 2021;Pardo et al., 2021;Parrish et al., 1998), with more recent threats coming from sea cage aquaculture that spreads parasitic copepodids onto wild seaward migrating fish (Johnsen et al., 2020;Krkošek et al., 2007). These parasites exert a substantial increased energetic burden on infected fish (Hvas and Bui, 2022), coinciding with observed reductions in body conditions of salmon returning from sea (Susdorf et al., 2018). Such energetic disadvantages will presumably be exacerbated by increasing temperatures and make completion of life cycles more difficult.
Energetics of fish are most often assessed from oxygen uptake rates (ṀO 2 ) as a proxy of the aerobic metabolic rate (Clark et al., 2013;Nelson, 2016). Using swim tunnel respirometers, theṀO 2 can be measured during swimming at defined activity levels together with maximum swimming capacities (Brett, 1964;Steffensen et al., 1984). FromṀO 2 at known swimming speeds the cost of transport (CoT) can be derived, and the optimal swimming speed (U opt ) can be defined as the speed that minimizes the CoT (Claireaux et al., 2006). For general cruising in search of food and migration specifically, fish should swim at their U opt to maximize energy efficiency, which also seems to be the case when monitoring fish movements in the wild (Drenner et al., 2012;Weihs, 1973).
When assessing maximum swimming capacities, protocols most commonly focus on the highest attainable speed by incrementally increasing water currents until fatigue is reached. These are known as critical swim speed (U crit ) tests, which involves utilization of both aerobic and anaerobic metabolism (Farrell, 2007;Wilson and Egginton, 1994). Less commonly, protocols may also focus on strictly aerobic fueled metabolism in sustained swimming trials where fish are tested at a constant speed for several hours (Beamish, 1978;Cotterell and Wardle, 2004;Hvas et al., 2021a). In the case of fed Atlantic salmon within midrange temperatures, the transition from sustained aerobic swimming to a mixture of aerobic and anaerobic swimming is at 80-85% of the U crit , while the U opt for minimum CoT tends to be 60-65% of the U crit (Beddow and McKinley, 1999;Hvas et al., 2021a;. Together, the U crit , the U opt and sustained swimming capacity provides valuable insights into the athleticism and energetics of fish that are reflective of environmental adaptations (Beamish, 1978;Claireaux et al., 2006;Plaut, 2001).
The physiological responses to extended fasting periods in Atlantic salmon have so far mainly been studied at the lower and mid ranges of their thermal niche. For instance, at 12 • C Atlantic salmon gradually downregulate resting metabolic rates over a 4-week fasting period as an adaptive response to save energy, while still maintaining the ability to adequately respond and recover from acute stress (Hvas et al., 2020(Hvas et al., , 2021b (Einen et al., 1998;Lie and Huse, 1992).
Prolonged fasting may influence aerobic and anaerobic swimming properties differently. Fast white muscle fibres tend to be more affected than slow red muscle fibres in a range of fish species investigated with regards to fibre size, protein synthesis, glycogen levels and capillary supply, and effects are larger on glycolytic than on oxidative functionality regardless of fibre type (Beardall and Johnston, 1983;Johnston and Goldspink, 1973;Loughna and Goldspink, 1984;Martínez et al., 2003;Scarabello et al., 1991). As such, food deprivation should first reduce high-speed short-term swimming capacity (e.g. U crit ), while sustained aerobic swimming capacities will remain preserved for longer periods. In Atlantic cod (Gadus morhua) fasted for 12 weeks, the U crit was indeed reduced due to impairment of anaerobic work (Martínez et al., 2004). Although the U crit in Atlantic salmon remained unaffected after 4 weeks of fasting at 12 • C (Hvas et al., 2021b), similar resilience to food deprivation is presumably not possible in warmer, less optimal environments. Moreover, it is unknown to what extent Atlantic salmon are able to downregulate metabolic demands at higher temperatures during fasting periods and how sustained swimming, CoT and U opt are affected. Food-deprived fish may therefore have less capacity for long-term high-intensity swimming within their normal aerobic limit when substrate depletion is imminent.
The purpose of the present study was to compare the impact of a 4-week fasting period at a normal midrange (9 • C) and an elevated suboptimal (18 • C) temperature with regards to swimming energetics and capacities in Atlantic salmon. Fed and fasted fish at both temperatures were subjected to a typical U crit trial as well as a 4-hour sustained swimming trial at 80% of the U crit . It was hypothesized that fooddeprived Atlantic salmon would be more impaired at the higher temperature because accelerated metabolic rates would deplete resources more quickly. Specifically, it was predicted that fasting at 18 • C would reduce U crit and that fish would struggle to complete the sustained swimming trial, while fasted fish at 9 • C would maintain their swimming capacities as well as benefitting from an improved CoT. Finally, analyses of haematological parameters in fatigued fish were done to assess potential differences in osmotic, endocrine and metabolic disturbances, for instance owing to changes in the functionality of anaerobic muscle fibres in fasted fish.

Fish husbandry
Atlantic salmon post-smolts produced on site at the Matre Research Station, Institute of Marine Research, Norway, were allocated in six indoor circular holding tanks (diameter, 3 m; volume, 5.3 m 3 ). The holding tanks were supplied with aerated, filtered and UV-C-treated full strength seawater of 34 ppt with a continuous inflow of 120 l min −1 , which ensured constant oxygen levels above 85% saturation and prevented accumulation of waste products. Three tanks were maintained at 9 • C and the other three tanks at 18 • C. The temperature was controlled by automatic mixing of ambient and heated water supplies in large header tanks prior to reaching the holding tanks. The fish were subjected to a simulated natural photoperiod and fed commercial feed in excess each day via automated feeders (pellet size, 4.5 mm; Skretting, Norway). Before starting the experimental trials, the fish had been acclimating in these conditions for approximately 1 month.
This study was made between September and December 2021 following ethical approval by The Norwegian Food Safety Authorities under permit number 254966 for use of animals in scientific research.

Swim tunnel setup
Critical and sustained swim trials were performed using a large Brett-type swim tunnel respirometer that has been described in detail previously Remen et al., 2016). The key dimensions of this system were a 248cm-long cylindrical swim section with an internal diameter of 36 cm and a total water volume of 1905 l. A camera was fixated behind the rear grid downstream of the swim section so that the fish being tested could be observed remotely. An oxygen sensor (RINKO ARO-FT, JFE Advanced, Japan) was attached next to the camera and was setup via computer software to log oxygen concentrations in 2-second intervals (MiniSoft SD200W, SAIV Environmental Sensors & Systems, Norway). In the downstream end of the swim section the top lid could be removed for access when fish had to be transferred into or from the tunnel. To minimize turbulence and achieve approximately laminar flow conditions, a resting chamber followed by a flow straightener with honeycombshaped cells (5 mm in diameter) and a reduction cone was situated upstream of the swim section. Water current speeds were generated with a motor-driven propeller (Flygt 4630, 11 • propeller blade, Xylem Water Solutions, Norway) opposite of the swim section, and a flow metre (Höntzsch Flow Measuring Technology, Germany) was used to confirm that desired current speeds were obtained with a given motor output. A large pipe connected to the same header tanks as used for the holding tanks supplied water into the tunnel via an inlet behind the propeller section. Here, a steady open flow through the tunnel ensured stable temperatures and normoxic conditions. However, for measurements ofṀO 2 this water supply could be closed off periodically followed by subsequent rapid flushing to re-establish oxygen levels.

Experimental design and protocols
In the first part of the experiment, U crit and sustained swim trials were performed on fed fish acclimated to either 9 • C or 18 • C. After completion of these trials, feeding was stopped, In the second part of the experiment, after the 4-week fasting period, U crit and sustained swim trials were performed again at both temperatures.
Prior to each U crit trial, eight fish were netted from an appropriate holding tank in the afternoon and quickly transferred to the swim tunnel that was located in the same room. The fish were then allowed to acclimate inside the swim tunnel overnight at a low current speed of 15 cm s −1 . The following morning the U crit trial started and consisted of a stepwise increase in current speed of 15 cm s −1 every 30 minutes. At high speeds the fish would eventually struggle to maintain swimming positions and fall back on the rear grid. Fatigue was then defined as when a fish was unable to resume swimming even after tactile encouragement by the experimenter's hand. At this point, time was noted, and the fish was removed and euthanized with a blow to the head. Immediately after, a blood sample was drawn via caudal puncture with a heparinized syringe and stored on ice momentarily until further processing. The weight and fork length was then recorded. The U crit trial was continued until all eight fish had reached fatigue.
To measureṀO 2 in response to increasing swimming speeds in the U crit trials, the water inflow was closed off for 20 minutes during each speed interval and flushed for the remaining 10 minutes, which ensured oxygen saturations above 85% throughout the trial. Closed measurement periods began 5 minutes into a new speed increment to provide some time for the fish to adjust swimming efforts and respiratory requirements, and thus ended 5 minutes prior to the onset of the next speed increment. Once the first fish had been removed due to fatigue,ṀO 2 was no longer monitored on the remaining fish. AllṀO 2 measurements thereby represented an average value of eight fish at defined activity levels . Background respiration was not corrected as it was found to be undetectable at different temperatures in this large setup . Additionally, the tail beat frequency was measured in four random fish at each current speed by using camera observations to count the time required to perform 50 tail beats. Three replicate U crit trials were performed at each temperature on both fed and fasted fish meaning that a total of 24 fish were tested per U crit treatment group.
The sustained swim trials were performed after the associated U crit trials had been completed at a specific temperature and feeding status on novel fish. The U crit trials needed to be conducted prior to sustained swimming trials in order to determine the sustained swimming test speed, which was defined as 80% of the mean U crit from the same cohort treatment. Similar to the U crit trials, eight fish were moved to the tunnel in the afternoon to acclimate overnight at 15 cm s −1 . The following morning the sustained swim trial started by increasing current speeds every 5 minutes by 15 cm s −1 until 80% U crit had been reached. This speed was then maintained for 4 hours. If a fish became fatigued, it was removed and time was noted, while fish that did not become fatigued after 4 hours were noted to have completed the test. The fish were then euthanized whereafter weight and fork length were recorded. During the sustained swim trials,ṀO 2 was measured in 30-minute intervals using 20-minute closed periods followed by 10 minutes of flushing. Furthermore, the tail beat frequency was measured every 30 minutes on four random fish, as in the U crit trials. Three replicate sustained swim trials were performed at each temperature on both fed and fasted fish.
After the U crit trials, blood sampled from fatigued fish were centrifuged in Eppendorf tubes at 6000 g for 5 minutes. The plasma supernatant was then transferred to new Eppendorf tubes and stored at −80 • C until later analyses. Once all swim trials had been completed, plasma osmolality was measured with freeze point determination in 20 μl subsamples using a Fiske 210 Micro-Sample Osmometer (Advanced Instruments). The concentration of plasma cortisol was measured in 20 μl subsamples using an ELISA assay kit (standard range: 20-800 ng ml −1 ; IBL International GmbH). Finally, the concentration of plasma lactate, Na + and Cl − were measured in 65 μl subsamples with an ABL90 FLEX blood gas analyzer (Radiometer).

Data analyses
The U crit was calculated, according to Brett (1964), as follows: where U f is the highest completed swimming speed, U i is the increment between speeds (15 cm s −1 ), t f is time endured at the last speed before reaching fatigue and t i is the time interval between speeds (30 minutes). Since the cross-sectional area of the swim section being large relative to the size of the fish, solid blocking effects were not corrected for in the reported U crit values, as this effect would be minimal (Bell and Terhune, 1970;Plaut, 2001).
TheṀO 2 was calculated for each measurement period in the critical and sustained swim trials from the decrease in dissolved oxygen over time as follows: where O 2 / t is the change in oxygen over time; V sys is the volume of the swim tunnel system; V b is the volume of the fish, assuming a density of 1 kg l −1 ; and M b is the biomass of the fish. The CoT at specific swimming speeds could then be derived and was expressed as mg O 2 kg −1 km −1 . In addition, a quadratic regression was fitted to the U-shaped relationship between CoT and swimming speed in the U crit trials, where the minimum of this function was defined as the U opt for minimum CoT.  Figure 1: Size parameters. Statistical differences between temperatures regardless of feeding status are indicated with letters, and statistical differences between fed and fasted fish within a specific temperature are indicated with asterisks (two-way ANOVA with Hold-Sidak post-hoc analyses P < 0.05). N = 48 and data are mean ± s.e.m.
To assess the impact of fasting on body morphology, the condition factor of all fish was calculated as 100 (weight (g)/length(cm) 3 ) (Ricker, 1975).
To test for statistical differences between feeding status and acclimation temperature on U crit , U opt, size parameters and haematological parameters, a two-way ANOVA with the Hold-Sidak method for multiple comparison procedures was used after confirming normality and equal variance of the data with the Shapiro-Wilk test and the Brown-Forsythe test, respectively (SigmaPlot 14.5). To adhere to these test assumptions, it was necessary to perform a log transformation of the data for weight, osmolality, Na + and Cl − . TheṀO 2 , CoT and tail beat frequency data were similarly analyzed with a two-way ANOVA to assess effects of treatment (feeding status plus temperature) at specific swimming speeds or time points in the U crit and sustained swim trials, respectively. A Pvalue below 0.05 was considered significant in the statistical tests and data are reported as the mean ± s.e.m.

Results
The 4-week fasting period did not result in any mortalities at either 9 • C or 18 • C, nor were any apparent changes in behaviours or incidences of injuries observed on the fish within the holding tanks.
Fish acclimated to 18 • C were significantly heavier and longer than at 9 • C regardless of feeding status (twoway ANOVA, DF = 191, P < 0.001 for both parameters). Condition factor was similar in fed fish across temperature (P = 0.491), but lower in fasted fish at 18 • C compared with fasted fish at 9 • C (P < 0.001). The fasting period reduced the weight significantly within both temperatures (9 • C, P = 0.017; 18 • C, P = 0.009) from 450 ± 12 to 417 ± 10 g at 9 • C and from 506 ± 10 to 464 ± 7 g at 18 • C, which corresponded to an average percentage weight loss of 7.3% and 8.3%, respectively (Fig. 1A). Meanwhile length increased significantly during fasting within temperature (P < 0.001 in both) from 32.4 ± 0.2 to 33.5 ± 0.2 cm at 9 • C and from 33.7 ± 0.2 to 35.3 ± 0.2 cm at 18 • C (Fig. 1B). These changes in weight and length resulted in significantly lower condition factors in fasted fish (P < 0.001 within both temperatures), from 1.31 ± 0.01 to 1.10 ± 0.01 at 9 • C and from 1.32 ± 0.01 to 1.05 ± 0.01 at 18 • C (Fig. 1C). The lowest measured condition factor on an individual fasted fish was 0.97 at 18 • C.
The U crit was significantly higher at 18 • C compared with at 9 • C across feeding status both when expressed as cm s −1 and as body length s −1 (two-way ANOVA, DF = 95, P < 0.001) ( Fig. 2A,B). However, the U crit was unaffected by fasting period within either temperature when expressed in cm s −1 (9 • C, P = 0.125; 18 • C, P = 0.058) as well as in body length s −1 (9 • C, P = 0.199; 18 • C, P = 0.197) ( Fig. 2A,B). Although not significantly different, fasted fish had a slightly higher U crit when expressed in cm s −1 and a slightly lower U crit when expressed in body length s −1 , which can be ascribed to the increased fork lengths in fasted fish.
The sustained swim trials were based on the 80% U crit of fish from the same temperature and feeding status. The specific sustained test speeds used were 82.1 cm s −1 (2.56 body lengths s −1 ) on fed fish at 9 • C, 83.9 cm s −1 (2.50 body lengths s −1 ) on fasted fish at 9 • C, 89.5 cm s −1 (2.67 body lengths s −1 ) on fed fish at 18 • C and 91.7 cm s −1 (2.61 body lengths s −1 ) on fasted fish at 18 • C. All fed and fasted fish tested at either temperature managed to complete 4 hours of swimming at 80% U crit without becoming fatigued.
TheṀO 2 increased with swimming speed and differed between all treatment groups in the U crit trials (two-way ANOVA, DF = 59, P < 0.001), where fish at 18 • C had a higheṙ MO 2 than at 9 • C and fasted fish had a lowerṀO 2 within both temperatures (Fig. 3A). Specific differences between fed and fasted fish inṀO 2 were found at 30, 45 and 60 cm s −1 at both temperatures whileṀO 2 also differed at 75 cm s −1 within 9 • C. At the highest test speeds, as the maximum aerobic metabolic rate was approached, theṀO 2 became similar between fed and fasted fish at both temperatures (Fig. 3A).
In the sustained swim trials, theṀO 2 differed between treatment groups (two-way ANOVA, DF = 95, P < 0.001).  Fish at 18 • C had a higherṀO 2 than fish at 9 • C (P < 0.001), and within temperature fasted fish had a lowerṀO 2 at 9 • C (P < 0.001), but not at 18 • C (P = 0.247) (Fig. 3B). Significant differences at specific time points between fed and fasted fish within temperature were found at 30 minutes for 18 • C and at 90, 120, 150, 180 and 210 minutes for 9 • C (Fig. 3B). Thė MO 2 was generally constant over time within each treatment group during the 4-hour sustained swim test. One exception to this was in the fasted 18 • C group where the 30-minute time point was higher than all other points, while the 60-minute time point differed from the 120-, 150-and 180-minute time points (Fig. 3B).
The CoT derived from the U crit trials also differed between treatment groups and swimming speeds (two-way ANOVA, DF 59, P < 0.001) (Fig. 3C). Specifically, fish at 9 • C had an overall lower CoT and fasted fish had a lower CoT within temperature. The CoT derived from the sustained swimming trials differed between groups in a similar manner to theirṀO 2 response (two-way ANOVA, DF 59, P < 0.001) (Fig. 3D). Notably, CoT was lower at 9 • C while fasted fish had lower CoT within 9 • C (P < 0.001), but not within 18 • C (P = 0.256) (Fig. 3D).
Tail beat frequency increased approximately linearly with swimming speed in the U crit trials and remained constant over time during the 4-hour sustained swim trials within all treatment groups (Fig. 3E,F). Significant differences in tail beat frequency between treatments were found in the U crit trials (two-way ANOVA, DF = 239, P < 0.001), where the most notable effect was that fed fish at 18 • C had higher tail beat frequencies than fasted counterparts at 30, 45, 60, 75 and 90 cm s −1 (Fig. 3E). A similar tendency for higher tail beat frequencies in fed fish within 18 • C was also observed in the sustained swim trials (two-way ANOVA, DF = 383, P < 0.001) (Fig. 3F).
The U opt expressed in body lengths s −1 , as a percentage of the U crit , and as cm s −1 versus CoT are shown in Fig. 4. Fasted fish had a significantly lower U opt (body length s −1 ) than fed fish within both temperatures (twoway ANOVA, DF = 11, P < 0.001). Fed fish showed similar U opt across temperatures (P = 0.802) whereas fasted fish were different and had a lower U opt (body lengths s −1 ) at 9 • C (P = 0.032) (Fig. 4A). Similarly, fasted fish had a lower U opt (%U crit ) within temperature (two-way ANOVA, DF = 11; 9 • C, P < 0.001; 18 • C, P = 0.002); however, both fed and fasted fish showed similar U opt (%U crit ) across temperatures (Fed, P = 0.068; Fasted, P = 0.0646) (Fig. 4B). The U opt (cm s −1 ) showed the same patterns in statistical differences as the U opt (body lengths s −1 ) (Fig. 4C). The minimum CoT differed in all comparisons between the four treatment groups with the fasted 9 • C group being lowest and the 18 • C fed group being highest (two-way ANOVA, DF = 11, P < 0.05) (Fig. 4C).
The haematological parameters measured in fatigued fish after the U crit trials are shown in Fig. 5. Plasma cortisol was higher at 18 • C (two-way ANOVA, DF = 95, P < 0.001) and was unaffected by feeding status within each temperature (P = 0.863) (Fig. 5A). Significant differences in plasma lactate were only observed within 9 • C where fed fish had higher lactate concentrations (P = 0.017) (Fig. 5B).

Weight loss and body condition
It was hypothesized that extended fasting at 18 • C would assert a greater negative impact on body condition in Atlantic salmon owing to accelerated metabolic rates draining resources more quickly compared with at 9 • C. Indeed, 4 weeks of fasting did cause a higher average weight loss and a greater reduction in condition factor at 18 • C, coinciding with a higherṀO 2 at similar activity levels.
However, the difference in weight loss was surprisingly minor between the two temperatures being 7.3% and 8.3% at 9 • C and 18 • C, respectively. Meanwhile, usingṀO 2 obtained at the lowest test speed in the U crit swim trials as an indicator of routine metabolic costs irrespective of feeding status, fish at 18 • C had a 57% higherṀO 2 than fish at 9 • C. Factorial changes in functional rates in 10 • C intervals are often expressed with the temperature quotient (Q 10 ), and the increase inṀO 2 here corresponded to a Q 10 of 1.7. Depending on acclimation history and section of the thermal niche assessed, resting metabolic rates in fish species generally increases with a Q 10 of 1-3 (Lefevre, 2016), while the resting metabolic rates of thermally acclimated Atlantic salmon previously was found to increase with a Q 10 of 2.1 between 8 • C and 18 • C . As such, if resting and routine energetic costs approximately doubles with a 10 • C temperature increase, then weight loss over an extended fasting period should theoretically also double. However, similar proportional effects were clearly not found in the present study (i.e. 57% higher costs, but only a 1% difference in weight loss). This discrepancy may be explained by theṀO 2 only being an indirect measurement of aerobic metabolic rates (Nelson, 2016), and that the relationship between oxygen uptake and energy production can vary between environmental conditions (Salin et al., 2015). Specifically, mitochondrial functioning becomes less efficient at higher temperatures due to impaired phosphorylation caused by increased proton leakage and decreased activity of complex I (Michaelsen et al., 2021;Roussel and Voituron, 2020).  Figure 4: The optimal swimming speed (U opt) for minimum CoT between groups expressed as body lengths s −1 (A), a percentage of the critical swimming speed (U crit) (B) and in cm s −1 versus the CoT (C). Statistical differences in (A) and (B) between fed and fasted fish across temperature are indicated with letters, and statistical differences between 9 • C and 18 • C within fed or fasted fish are indicated with asterisks. Vertical differences in U opt in (C) are indicated with letters, while numbers indicate horizontal differences in CoT (two-way ANOVA with Hold-Sidak post-hoc analyses, P < 0.05). N = 3 and data are mean ± s.e.m. Figure 5: Haematological parameters. Blood samplings were done on exhausted fish immediately after U crit measurements. Statistical differences between temperatures across feeding status are indicated with letters, and statistical differences between fed and fasted fish within a specific temperature are indicated with asterisks (two-way ANOVA with Hold-Sidak post-hoc analyses, P < 0.05). N = 24 and data are mean ± s.e.m. acclimated to 20 • C compared with 12 • C acclimation (Gerber et al., 2021). Consequently, if less ATP is produced per oxygen at elevated temperatures, a higherṀO 2 is required to support similar metabolic demands as in lower temperatures. In this way temperature can cause a non-proportional relationship betweenṀO 2 and energy production in fish and other ectotherms (Salin et al., 2015), and thus help explain a lower weight loss than otherwise expected from the proportionally greater increase inṀO 2 at 18 • C. Assuming thatṀO 2 and energy production were proportional across temperatures, a weight loss of approximately 12% instead of 8.3% would then have been expected at 18 • C in the present study.
The reported weight loss of food-deprived Atlantic salmon in previous studies have generally been lower than the present one, likely owing to having used larger size classes, as mass specific energetic requirements decreases with size (Killen et al., 2006;Oldham et al., 2019), combined with lower temperatures. For instance, 8-9 weeks of fasting reduced weights by 6-7% Reimers et al., 1993) and 11-12 weeks of fasting reduced weights by 10-11% (Einen et al., 1998;Lie and Huse, 1992), where these studies used fish ranging from 1.2 to 5 kg. previously been shown in food-deprived Atlantic salmon and indicates that vertebrae and muscle development may work independently of each other (Einen et al., 1998;Hvas et al., 2021b. Furthermore, continued length increases could prepare fish for accelerated muscle growth when feed becomes available again and thereby help facilitate compensatory growth (Ali et al., 2003).
Exacerbated by continued length growth, the condition factor dropped substantially during fasting from 1.32 across temperature down to 1.10 and 1.05 at 9 • C and 18 • C, respectively. However, a state of severe starvation and malnourishment in Atlantic salmon is generally first associated with condition factors below 0.9 (Noble et al., 2018). The fish in the present study could therefore have endured a longer fasting period before entering a stage of protein catabolization, which signifies severe starvation (Wang et al., 2006). As such, considering the modest changes in weights and the resulting intermediate condition factors, 4 weeks of fasting was well within the tolerance limit of Atlantic salmon, even at a suboptimal elevated temperature of 18 • C.

Swimming capacity
The U crit is highest in the upper ranges of the thermal niches in Atlantic salmon and other fish species studied, coinciding with higher maximum metabolic rates (Brett, 1964;Claireaux et al., 2006;Yuen et al., 2019). A higher U crit at 18 • C compared with at 9 • C was therefore to be expected in the present study.
Interestingly, the U crit was unaffected by the 4-week fasting period within both acclimation temperatures. In a separate study Atlantic salmon also maintained their full U crit over a 4-week fasting period at 12 • C (Hvas et al., 2021b), and a similar result was therefore expected at 9 • C owing to lower temperatures having a lesser impact on metabolic demands. However, an unaffected U crit at 18 • C was surprising as this elevated temperature represents suboptimal conditions for growth and behavioural preferences in this species (e.g. Johansson et al., 2009;Kullgren et al., 2013). In a study on food-deprived cod the U crit was reduced in the food-deprived treatment, but their condition factor had decreased from 1.0 to 0.5 (Martínez et al., 2004). While different species have different morphometrics, a halving of condition factor is a dramatic change compared with only a 20% reduction at 18 • C in the present study. Hence, a more modest impact on body condition likely explains why U crit remained unaffected in Atlantic salmon.
All fish tested managed to complete 4 hours of sustained swimming at 80% U crit without becoming fatigued. The limit of sustained swimming in fed Atlantic salmon have consistently been found to be 80-85% U crit at midrange temperatures of 12-13 • C where speeds above 85% U crit necessitates recruitment of anaerobic white muscle fibres and eventually causes fatigue (Beddow and McKinley, 1999;Hvas et al., 2021a;. However, whether lower or higher temperatures would alter this threshold for aerobic swimming was previously unknown. It has been shown in other fish species that white muscles are recruited at lower swimming speeds in lower temperatures, which causes a reduction in sustained swimming capacity (Rome et al., 1990;Sisson and Sidell, 1987;Taylor et al., 1996). Similarly, in Atlantic salmon swimming at 3 • C and 8 • C, the transition to partially anaerobic burst and glide swimming occurs at lower speeds than at higher temperatures . However, in the present study, aerobic swimming limits were not yet impaired at 9 • C.
The aetiology of fatigue can either be a failure to supply sufficient metabolites at the required rates (e.g. U crit tests) or the depletion of metabolite supplies (e.g. fixed long-term velocity tests) (Jones, 1982). The latter being analogous to collapsing marathon runners 'hitting the wall' despite working within their normal aerobic limit. It was therefore hypothesized that fasted fish at 18 • C eventually would fatigue in the sustained swim trials as food deprivation results in a greater exhaustion of resources at this temperature. However, as discussed above, the fasting regime at 18 • C only caused a moderate reduction in body condition meaning that fish still were able to fuel long-term high intensity aerobic swimming. The 80% U crit threshold for sustained swimming in Atlantic salmon can therefore now also be applied to a wider thermal interval and to fish that have experienced extended fasting periods.
TheṀO 2 remained stable within each treatment group during 4 hours of sustained swimming, providing evidence for that swimming efforts indeed remained aerobic. If this was not the case, anaerobic metabolism would cause lactate to accumulate in muscles and blood since lactate clearance is slow in salmonids (Hvas et al., 2021b;Schulte et al., 1992). Furthermore, only 3 ATP are produced when glycogen is converted to lactate while 5 ATP are required to convert lactate back to glycogen (Burgetz et al., 1998). Hence, if a significant anaerobic component was present, the increasing need to metabolize lactate at a cost deficit while swimming would drive up theṀO 2 until maximum rates were reached whereafter fatigue would be imminent.
Exhaustive exercise stress in salmonids causes substantial disturbances in osmotic, endocrine, and metabolic balances (Kieffer, 2000;Wood, 1991). The haematological parameters measured in the present study after the U crit tests were therefore characteristic of fatigued Atlantic salmon in seawater (Hvas et al., 2018(Hvas et al., , 2020. However, since the U crit was unaffected by food deprivation, only modest variations were found between treatment groups. Fish at 18 • C were swimming for longer and at higher final speeds while achieving a substantially higherṀO 2 , which facilitated a greater increase in plasma osmolality, Cl − and Na + compared with at 9 • C owing to the osmorespiratory tradeoff of the fish gill in hyperosmotic environments (Hvas et al., 2018;Sardella and Brauner, 2007). Interestingly, fasted fish at 18 • C showed a greater ionic disturbance, which may be a subtle indicator that athletic performance was starting to become affected, as this would have some implications for repeat-swimming capabilities since recovery of ion homeostasis is particularly slow in seawater adapted Atlantic salmon (Hvas et al., 2018(Hvas et al., , 2020. The higher cortisol values at 18 • C were likely caused by a combination of increased swimming efforts and temperature, as the cortisol response to a fixed stressor also was elevated at 17 • C compared with 9 • C in a previous study on Atlantic salmon (Madaro et al., 2018). Plasma lactate was unaffected by temperature, which suggests that similar anaerobic swimming efforts were made prior to fatigue meaning that the improved U crit at 18 • C was aerobically driven. However, fasted fish within 9 • C had lower lactate levels, which could hint at lower functionality of white muscle fibres as shown for other food-deprived fish species (Beardall and Johnston, 1983;Martínez et al., 2003). Although, considering that resting lactate levels should be <1 mM (e.g. Wood, 1991), and all treatment groups had greatly elevated values of >7 mM, this indicates preserved high capacities for anaerobic work following 4 weeks of fasting in Atlantic salmon.

Swimming energetics and U opt
While the capacities for both aerobic and anaerobic swimming were preserved following extended fasting, some temperature-dependent changes in swimming energetics were observed. At low and intermediate swimming speeds from 30 to 60 cm s −1 , fasted fish within both temperatures had a loweṙ MO 2 . At 9 • C this corresponded to an average reduction of 43.1 mg O 2 kg −1 h −1 or 23.8% across speeds, and at 18 • C a comparable absolute reduction of 41.4 mg O 2 kg −1 h −1 was found, but a lower percentage reduction of 15.6% caused by higher basal costs. Similarly, the resting metabolic rate of Atlantic salmon after 3-4 weeks of fasting at 12 • C was reduced by 22% (Hvas et al., 2020), while other fish species also downregulate metabolic rates in response to food deprivation (Fu et al., 2005;Mehner and Wieser, 1994).
These reductions inṀO 2 in fasted fish were unlikely to be an artefact of elevatedṀO 2 in the fed control fish caused by specific dynamic action effects, as these fish had been fasting overnight for a minimum of 20 hours prior to measurements, which should have allowed for completion of most digestive processes (Handeland et al., 2008;Storebakken et al., 1999). Hence, lowerṀO 2 in food-deprived fish should represent real adaptive responses to minimize energetic requirements that may involve a range of biochemical adjustments in gene expressions, decreased enzyme activities, reduced protein syntheses and improved efficiency of mitochondrial functioning in appropriate organs (Bermejo-Nogales et al., 2015;Cassidy et al., 2016;Méndez and Wieser, 1993;Salin et al., 2018).
Interestingly, the ability to save energy during food deprivation was less pronounced at higher activity levels where thė MO 2 between fed and fasted fish converged, and this pattern occurred earlier at 18 • C. As such, the ability to preserve energy appeared to be less efficient when fasting at higher temperatures in Atlantic salmon. In addition, the similarṀO 2 between fed and fasted fish at high activity levels suggests that maximum aerobic metabolic rats were unaffected, in agreement with an unaffected U crit . This observation is advantageous for food-deprived Atlantic salmon, as they can use less resources during resting and routine conditions while still being able to perform at their highest athletic ability, if needed.
The reducedṀO 2 in fasted fish resulted in lower CoT and U opt at both temperatures, and these effects were greater at 9 • C. This means that fasted Atlantic salmon in colder waters, in theory, will swim slower and at lower costs when foraging or migrating. For instance, based on the U opt of the present study, a hypothetical 1000 km migration would increase from 16.3 to 19.1 days between fed and fasted fish at 9 • C, and similarly increase from 15.7 to 17.2 days at 18 • C. While using longer time to migrate a given distance, the overall energetic requirement will still be reduced in food-deprived Atlantic salmon, with 17.8% and 11.3% lower total costs at 9 • C and 18 • C, respectively. Moreover, regardless of feeding status, cruising at U opt was 27.5% cheaper at 9 • C compared with at 18 • C. The maximum distance that can be covered by food restricted migratory Atlantic salmon will consequently be lower in warming marine environments.

Conclusion
The Atlantic salmon is an environmentally flexible species that is able to handle and adapt to many of the diverse challenges it may encounter during its life cycle. This study exemplifies this flexibility as extended fasting was successfully endured with moderate effects on weight and body condition because of the ability to beneficially downregulate metabolic requirements, even at an elevated suboptimal temperature. Furthermore, despite of 4 weeks of food deprivation, Atlantic salmon maintained their full capacity for both aerobic and anaerobic swimming across temperatures as shown by a preserved U crit and completion of the sustained swim trials.
However, some ecological implications may be considered with regards to swimming energetics in a warming climate in conjunction with other stressors. Elevated temperatures reduced the fasting-induced energetic savings, while inherently also driving up basal maintenance costs and thus draining resources more quickly. This could restrain migratory limits and have implications for survival, both during initial seaward travel in newly smoltified fish and upriver return in spawning adults. Specifically, seaward-bound juvenile Atlantic salmon are under substantial predation pressure owing to their small sizes and are also at high risk of parasite infections dispersed from aquaculture farms when travelling through the fjords and coastal zones (Johnsen et al., 2020;Thorstad et al., 2012). Rapid growth at sea is therefore crucial to ensure survival, but this task could become more difficult in warmer marine environments, particularly if fish also have to cope with the energetic burden of parasites (Hvas and Bui, 2022;Susdorf et al., 2018).