Climate change- induced deprivation of dietary essential fatty acids can reduce growth and mitochondrial efficiency of wild juvenile salmon

1. Omega- 3 long- chain polyunsaturated fatty acids ( n − 3 LC- PUFA) are essential micronutrients for optimal functioning of cellular metabolism and for somatic growth of all vertebrates including fishes. In addition, n − 3 LC- PUFA could also play a key role in response of fishes and other ectothermic vertebrates to chang ing temperatures. 2. An important, but largely overlooked, consequence of climate change is the re duced availability of dietary n − 3 LC- PUFA in aquatic food webs. Changes in avail ability of dietary n − 3 LC- PUFA have recently been proposed as a major driver of novel adaptations and diversification of consumers. Yet, there is only limited

These factors increase the influx of terrestrial matter that is deficient in n − 3 LC-PUFA (Brett et al., 2017;Tiwari et al., 2017) and hamper the production of n − 3 LC-PUFA by primary producers in the stream (Hixson & Arts, 2016;Taipale et al., 2018). Therefore, climate change will not only affect the thermal environment (Pinsky et al., 2019), but also decrease the availability of dietary n − 3 LC-PUFA for consumers in small head water streams, including juvenile Atlantic salmon Salmo salar. Understanding the impacts of an n − 3 LC-PUFA-deprived diet on the performance and thermal response of stream-dwelling fishes can help us better predict their resilience to changing climate.
It has been shown that dietary n − 3 LC-PUFA, especially DHA, influence various ecologically important traits of fishes such as somatic growth (Tocher, 2010) and development of brain and cognitive skills (Pilecky et al., 2021). However, this evidence is largely based on domesticated fishes and laboratory models, whose genotype and phenotype substantially differ from their wild counterparts (Albert et al., 2012;Johnsson et al., 2014). A handful of studies on fish of wild origin is available from marine environments (e.g. Gourtay et al., 2020), where access to dietary n − 3 LC-PUFA is rarely a limiting factor for consumers (Colombo et al., 2017). The tissue content of n − 3 LC-PUFA determines numerous cellular processes in ectotherms, including functioning of mitochondrial metabolism (Martin et al., 2013;Salin et al., 2021), which limits the performance of organs and the whole organism (Dawson, Alza, et al., 2020;Salin et al., 2015), including their thermal performance (Salin et al., 2015). Therefore, the link between the diet quality, fatty acid composition of the tissue and mitochondrial metabolism provides a promising, but rarely explored venue for a more mechanistic understanding of the diet quality impact on performance of cells and whole-animal and potential feedbacks to the entire ecosystems under changing climate.
Aquatic prey of juvenile fishes is rich in EPA and alpha-linoleic acid (ALA; 18:3 n − 3) that are produced by periphyton at the base of the stream food web, but terrestrial resources contain only ALA and are almost deprived of EPA (Twining et al., 2019). The conversion of EPA to DHA has been shown to be more effective and contributes 10 times more to the building of neuronal structures than the less efficient conversion of ALA to DHA (Pilecky et al., 2021). Juvenile salmon and other small stream fishes have only limited direct access to dietary DHA, but they can occasionally acquire it, for example, by consumption of eggs (Näslund et al., 2015). The elevated metabolic costs (i.e. the additional energy spend) for internal synthesis of DHA from ALA, for example, during periods of floods characterized by low availability of n − 3 LC-PUFA micronutrients (Argerich et al., 2016;Arts et al., 2009), may lead to a reduction of somatic growth (Tocher, 2010). Yet, under scarcity of dietary n − 3 LC-PUFA this costly physiological adaptation can be important to maintain proper functioning and growth of brain (Ishizaki et al., 2005;Lund et al., 2014). Here, we posit that dietary n − 3 LC-PUFA deprivation caused by climate change as a result of reduced primary production and elevated flux of terrestrial subsidies should force fish into a trade-off between maintenance of n − 3 LC-PUFA tissue content and somatic growth. Reduced content of n − 3 LC-PUFA in brain and muscle should then have negative consequences for their thermal tolerance, mitochondrial efficiency, aerobic metabolism, and cognition.
To test our predictions, we exposed juvenile salmon (offspring of wild parents) to two experimental diets, one with a content of optimal n − 3 LC-PUFA conducive for their proper development (hereafter control diet) and the other with reduced n − 3 LC-PUFA content simulating the decreased availability of these micronutrients in degraded aquatic food web (hereafter n − 3 LC-PUFA-deprived diet).
The treatment lasted 8 weeks which is a time period comparable to pulses of n − 3 LC-PUFA poor terrestrial resources at headwater streams during seasonal floods (Argerich et al., 2016). Our experimental design allowed us to test whether an n − 3 LC-PUFA-deprived diet: (a) leads to a reduction of n − 3 LC-PUFA and particularly DHA in brain and muscle tissues; (b) decreases efficiency of mitochondrial metabolism and increases maximum metabolic rate (MMR) particularly at elevated temperature; (c) reduces brain size and deteriorates learning ability particularly at elevated temperature and (d) has a negative effect on somatic growth of wild fish.

| Experimental fish origin and rearing
Experimental fish were offspring of wild parents (14 families originating from unique crossing of 14 females and 14 males) collected during the spawning migration between November and December 2018 in a permanent trap at Loch na Croic on the river Black Water, which is situated within the catchment of River Conon in Northern K E Y W O R D S aerobic metabolism, animal behaviour, animal cognition, ATP production, ecosystem functioning, omega-3 fatty acids Scotland, UK. Atlantic salmon and resident brown trout Salmo trutta are the dominant fish species in the system (Auer et al., 2018).
Fertilized eggs developed until hatching at the Contin hatchery, which is supplied by water from the Black Water River. Juvenile salmon were transported post-hatching on 9 April 2019 to the aquarium facility at the University of Glasgow. From the onset of external feeding until the start of the experiment individuals were kept in flow-through stream system supplied with UV-treated, recirculating freshwater kept at 13℃. They were fed daily until apparent satiation with a mixture of frozen bloodworms and commercial pellet food with high content of fish oil (EWOS).
On 19 November 2019 (i.e. 7 months post-hatching), 80 individuals of similar size were anaesthetized in a benzocaine solution and measured for fork length (FL) to the nearest millimetre and body mass (bM) to the nearest 0.01 g (mean ± SD, FL = 55 ± 5 mm, bM = 1.92 ± 0.61 g). Fish were then tagged with Visible Implant Elastomer (VIE; Northwest Marine Technology Inc.) for individual identification and distributed among ten rearing tanks (32 L, 40 × 40 × 30 cm, 8 individuals per tank). Tanks were evenly designated to the two feeding treatments with five replicated rearing tanks and 40 individuals in each treatment. One tank from each diet treatment (16 individuals in total) was used to test design of learning test and these individuals were removed from the study. During the experiment salmon were fed on isonitrogenous and isocaloric fish pellet feeds (GARANT TM , Austria) formulated to provide the same amount of energy, lipid and protein sufficient to meet somatic requirements for salmonids (Murray et al., 2014;Tocher, 2015) but differing in their n − 3 LC-PUFA concentration (ESM 1). We evaluated total fatty acid profiles and mitochondrial metabolism in salmon muscle and brain, as well as maximum metabolic rate of individuals (MMR), their somatic growth rate, and learning ability. All traits were measured on all individuals available with exception for measurements of mitochondrial metabolism, which were based on a subset of nine individuals randomly selected from each treatment group (Table 1). We measured MMR, learning ability and mitochondrial metabolism at low (13℃) and elevated (18℃) temperatures to evaluate the response of these traits to acute warming. These temperatures are near below and above the physiological optimum reported for Atlantic salmon (Elliott & Hurley, 1997). Fish were exposed to a progressive increase of temperature over the course of 24 hr and then remained at 18℃1 for another 48 hr until the test. Following the test, the temperature of the rearing tank was progressively cooled over 24 hr back to 13℃. Three-day exposure to temperature increase by 6℃ is ecologically relevant as it can occur in salmon streams and Atlantic salmon are able to physiologically cope with the temperature change of a similar intensity (Gallant et al., 2017).

| Maximum metabolic rate and somatic growth rate
Maximum metabolic rate of individuals was estimated from the rate of oxygen uptake after exhaustive exercise using intermittent flow respirometry . The metabolic assays were conducted between 6 and 15 January 2020, after 7 weeks of exposure to the dietary treatments. MMR of each fish was measured twice, once at the acclimation temperature, that is, 13℃ and once at elevated temperature, that is, 18℃. Assays took place in 16 glass chambers (0.10204 L) in recirculation loops using a peristaltic pump (also used to achieve good water mixing in the chambers) and gas-tight PVC tubing (volume of the loop between 8 to 15 ml) submerged into a thermoregulated water bath (92 L, 80 × 40 × 29 cm). Fish from two rearing tanks were thus measured simultaneously. Bacterial oxygen consumption was kept at a minimum by using a UV filter sterilizer and it was evaluated daily before and after the fish were placed in the respirometry chambers. One chamber was also kept empty during the measurements to estimate the evolution of the bacterial consumption over time. Measured oxygen uptake was then adjusted by a linear increase in bacterial metabolism over the time of the assay.
The whole system was also fully bleached between each trial. Flush pumps, connected to a timer, flushed oxygenated water through the chambers for 2 min with 8-min intervals between flushes, during which the oxygen uptake of individuals was measured. Oxygen concentration within each glass chamber was recorded every 2 s by a fibre optic sensor inserted into probe holders inside the loop and using a four-channel FireSting O2system (PyroScience GmbH). The level of oxygen within the chambers never dropped under 60% air saturation. Oxygen probes were calibrated on a daily basis. Prior to TA B L E 1 Overview of chronology of the measurements and sample size of the experimental fish no response to taping of the tail fin) for 3 min in a circular tank containing aerated freshwater . MMR was calculated using rolling regression slopes of 2 min every 2 s, after a wait period of 20 s. The slope was multiplied by the volume of the respirometry chamber and tubing after deduction of the fish volume (assuming a fish density of 1 g/ml) was used to calculate the maximum oxygen uptake (mg O 2 /hr). After respirometry, all individuals were measured for FL to the nearest millimetre and bM to the nearest 0.01 g. Three individuals died before the respirometry (1 and 2 from the control and n − 3 LC-PUFA-deprived diet treatment respectively). Due to a technical issue with the respirometry setup, data from 16 individuals (eight from each diet treatment group) were recorded at 13℃, but not at 18℃.
Specific growth rate of individuals was calculated following equation from (Brett & Groves, 1979) using initial and final FL as measured at tagging and final tissue sampling respectively, and time interval between the two measurements in days.

| Learning tests
Focal fish were trained within their rearing tank to associate a visual stimulus with a food reward, which was then used to determine the success probability of approaching the rewarded stimulus in a maze test (see ESM 4 for details of fish training and experimental setup). Fish were scored individually in three 10-min consecutive trials with 20 min acclimation period in between the trials during each scoring day. In the first two trials in each scoring day, the salmon were presented with two stimuli-red circle (i.e. rewarded stimulus) and blue circle (i.e. unrewarded stimulus).
To test the effect of egocentric spatial learning (i.e. left-right decision), the salmon were presented with two blue circles in the third trial (Rodriguez et al., 1994). In the third trial, individuals were rewarded only if they approached the stimulus that was on the side where the red circle had been presented in the previous two tri-

| Fatty acids profile analysis
Samples of brain and white muscle tissue were collected for fatty acid analysis between 14 January and 5 February 2020 immediately after individuals were killed with an overdose of benzocaine. The be noted that majority of functionally important fatty acids in the brain is from polar lipids (Ebm et al., 2021) which also applies to DHA in the muscle tissues (ESM 2). We calculated descriptive statistics for the main groups of fatty acids (Table 2), but for further analysis we used only the 15 most dominant fatty acids with the average relative concentration higher than 1%. We also calculated bioaccumulation factors (Kainz et al., 2006) of ALA, EPA and DHA as: The bioaccumulation factor is an indicator of the internal synthesis

| Mitochondrial metabolism
Mitochondrial metabolism from the white muscle and brain of juvenile salmon was measured in 2 ml of respiration solution using a high-resolution respirometer (Oxygraph-2k with O2k-Fluorescence module; Oroboros Instruments, Innsbruck, Austria) at 13 and 18℃ under continuous stirring. Samples were prepared as described in Salin et al. (2016) using the shredded tissue technique. White muscle tissue was dissected, removing skin and bone tissue, from an area ~1 cm above the lateral line under the dorsal fin. Brain was bisected and one half of the brain was used for mitochondrial analysis and other for analysis of fatty acids. Muscle (40 mg wet mass) or brain tissue (10 mg wet mass) were added to the chamber and allowed to sit for 5 min. Respiration rate was measured as the rate of decline in O 2 concentration. Mitochondrial respiration protocols to determine LEAK (Ln) and respiratory control ratio (RCR) were adapted from Dawson, Millet, et al. (2020), for details see ESM 3.  (Kotrschal et al., 2013). Repeatability of MMR (adjusted for temperature, diet treatment, and body mass) and success in the learning test (adjusted for the trial and diet treatment) was quantified using the intra-class correlation coefficient extracted from linear mixed models with individual identity as a random factor (Nakagawa & Schielzeth, 2010). We used simple linear models and did not evaluate repeatability across the measurements of mitochondrial metabolism, because of the low sample size which prevented us to calculate sufficiently robust estimate of inter-individual variance (Dingemanse & Dochtermann, 2013). Non-significant interactions were removed from the models. Significance of the final models was evaluated using ANOVA tables using Type II sums of squares. Variables were log-transformed when needed to approach the normal distribution of residuals. Differences among groups were analysed using Tukey's HSD post-hoc test.

| RE SULTS
Dietary treatment influenced the concentration of fatty acids in muscle (MANOVA F 15,35 = 68.33, p < 0.001, Figure 1a, Table 3) and brain (MANOVA F 15,16 = 11.44, p < 0.001, Figure 1a, Table 3) tissue of juvenile salmon. We also found an effect of body mass on fatty acid profile of the tissues, but this covariable was significant only in the model for muscle (MANOVA F 1,49 = 3.61, p = 0.001) and not the brain (MANOVA F 15,16 = 1.96, p = 0.100) tissue. Dietary treatment had no effect on amount of total lipids in fish tissues, but it has significantly affected concentration of SFA, MUFA, n − 3 PUFA, n − 6 PUFA and n − 3:n − 6 ratio in muscle and brain of juvenile salmon (Table 3). We found that fish did not bioaccumulate ALA and EPA as concentration of these fatty acids was lower in the tissues of the fish then in their diet. In contrast, fish in both TA B L E 2 Summary of final models reported in Results. Models for muscle and brain tissue had identical structure. 15FAs -% of 15 dominant fatty acids, diet-dietary treatment, bM-body mass at the time of measurement, bM i -initial body mass, T-temperature, F b DHAbioaccumulation factor of DHA, position-position of the rewarded stimulus, trial-trial number, socRank-social rank of individual within the rearing tank, resBrain-residual brain mass, ID-individual identification treatments bioaccumulated large amount of DHA and the bioaccumulation factor for both muscle and brain were significantly higher in salmon fed n − 3 LC-PUFA-deprived diet then in the control group (Table 3) There was no effect of diet treatment on residual MMR, that is, after correcting for body mass (χ 2 = 0.02, p = 0.903), but residual MMR increased in both treatments with increasing temperature (χ 2 = 6.00, p = 0.014; Figure 3). Inter-individual differences in residual Muscle Brain the first testing day was not affected by the left or right position of rewarded stimulus, that is, lateralization did not affect success rate in the learning test (χ 2 = 1.97, p = 0.160).

| D ISCUSS I ON
Our results show that dietary deprivation of n − 3 LC-PUFA induced by experimental treatment simulating subsidies from a degraded aquatic food web caused, in comparison to a control diet, a change in the fatty acid composition and, importantly, a decrease of DHA content in the muscle and brain of juvenile salmon. DHA is a key micronutrient for optimal functioning of cellular membranes and vital tissues of vertebrates (Pilecky et al., 2021). The reduced tissue content of DHA coincided with reduced mitochondrial efficiency of ATP production in muscle, but it did not affect maximum metabolic rate.
Intriguingly, the functioning of mitochondrial in brain and learning ability of the fish were unaffected, despite the substantial decrease of DHA content in the brain. Our findings indicate that juvenile salmon have a capacity to mitigate at least some potentially negative effects of temporarily reduced dietary intake of n − 3 LC-PUFA on functioning of brain. However, the DHA bioaccumulation factor in the salmon tissue was negatively correlated with the growth rate of individuals, indicating that individual adaptation to dietary deprivation of n − 3 LC-PUFA comes at an energetic cost related to internal synthesis or retention of DHA, which translates to reduced somatic growth.
The DHA content in the experimental diets was likely much lower than that individuals would need for proper long-term functioning of their tissues, but this is also typical for the aquatic and terrestrial prey of stream fishes (Twining et al., 2019). The bioaccumulation factor of DHA in both treatments was much higher than one and about five times higher in salmon fed n-3 LC-PUFA-deprived diet than in the control group. In contrast, differences in bioaccumulation factor of EPA and ALA between the diet treatments were not significant and for both molecules had value lower than one. This suggests that dietary EPA and ALA were not retained in the salmon tissues, but instead likely used for the synthesis of DHA by fish in both dietary treatments (Kainz et al., 2006). Besides internal synthesis of DHA (Murray et al., 2014;Taipale et al., 2018) individuals also could attain high DHA via dietary retention before the experiment and through the maternal provisioning of eggs from which they hatched (Fuiman, 2018). Despite a higher bioaccumulation factor, the DHA content was lower in both muscle and brain tissues n − 3 of the salmon fed on the n − 3 LC-PUFA-deprived diet. Feeding on n − 3 LC-PUFA-deprived diet was associated with slower body growth, and body growth in both dietary treatments decreased with an increasing DHA bioaccumulation factor. This finding corresponds to previous studies on domesticated fishes (Lazzarotto et al., 2015;Murray et al., 2014;Taipale et al., 2018) and demonstrates that maintaining high DHA contents in tissues despite limited dietary supply is costly also for wild fish. Our results thus suggest that the physiological adaptation to mitigate dietary shortage of n − 3 LC-PUFA TA B L E 3 Differences in the lipid composition of muscle and brain tissue of the juvenile salmon and their experimental diet in n − 3 LC-PUFA-deprived (−n − 3) and control (+n − 3) treatment, F b -bioaccumulation factor. Descriptive statistics report mean ± SD and the p-value of statistical differences is based on models listed in Table 2 can be limited even in consumers such as juvenile salmonids, which often prey on resources with relatively low n − 3 LC-PUFA, for example, terrestrial invertebrates (Evangelista et al., 2014;Syrjänen et al., 2011).
Omega-3 LC-PUFA-deprived diet lowered mitochondrial efficiency of ATP production in muscle tissue as indicated by lower RCR, which was likely caused by the increment of mitochondrial membrane LEAK. However, the effect of diet on mitochondrial in muscle did not translate to differences in MMR or its response to acute warming.
This was unexpected because muscle tissue represents the majority of fish biomass, and thus muscle oxygen consumption should relate to overall oxygen uptake of the fish, especially during or after strenuous physical activity (Norin & Malte, 2012). However, the association of oxygen uptake by muscle mitochondrial and whole organism is also driven by other factors such as mitochondrial production of harmful reactive oxygen species (Salin et al., 2015). Previous studies provide equivocal results about the importance of dietary n − 3 LC-PUFA for mitochondrial functioning. For example, mitochondrial LEAK in muscle of domesticated trout have been reported to either increase (Martin et al., 2013), but also decrease (Guderley et al., 2008) under n − 3 PUFA-deprived diet. While the mechanism behind different resilience of the mitochondrial metabolism remains to be explored, it could explain why we observed no effect of n − 3 LC-PUFA deprivation on mitochondrial metabolism in the brain, despite significant decrease of DHA in its cellular membranes.
The functional resilience of the brain to n − 3 LC-PUFA-deprived diet was also apparent at the level of cognition, as we found no effect of dietary treatment on learning. This is in contrast with some previous studies, which showed for example that n − 3 LC-PUFAdeprived diet reduced responsiveness to visual stimulus in southern flounder Paralichthys lethostigma (Oberg & Fuiman, 2015) or deteriorate antipredator behaviour in pike perch Sander lucioperka (Lund et al., 2014). The discrepancy could be caused by relatively short dietary deprivation of n − 3 LC-PUFA in our study. Nonetheless, other dimensions of behaviour that we did not examine may still be affected and so more study is warranted given the obvious differences in tissue concentrations. Individual differences in success probability in the learning test were highly repeatable across the nine trials and F I G U R E 3 Residual maximum metabolic rate (MMR) (i.e. measured as residuals from linear regression between MMR and body mass of individuals at the time of respirometry). Large circles and whiskers indicate mean ± 95% CI of each group (green: n − 3 LC-PUFA-deprived diet, red: control diet) ( ) F I G U R E 2 Respiratory control ratio RCR in muscle (a) and brain (b), and mitochondrial LEAK in muscle (c) and brain (d) brain positively associated to the residual brain mass. Intra-specific variation in brain size have been shown to be positively related to neuronal abundance (Marhounová et al., 2019) and learning skills in fishes (Kotrschal et al., 2013). While the brain mass did not differ between the treatment groups, the link between brain size and success rate indicates that the learning test was cognitively challenging and the lack of the effect of dietary treatments is evidence that the experimental diet had no influence on learning ability in this study.
In this unique laboratory experiment, we linked biochemistry and physiology at the cellular level to physiological and cognitive processes at the individual level to provide one of the first evidences

(c)
Control Omega-3 LCPUFA deprived smoltification (Bell et al., 1997) and migration to the n − 3 LC-PUFA rich marine feeding grounds (Colombo et al., 2017). However, the freshwater juvenile life stage is a major selection bottleneck (Elliott, 1994) and diet quality induces a strong selection pressure on the physiological phenotype of juvenile salmonids (Auer et al., 2018). It is thus possible that the long-term reduction of dietary n − 3 LC-PUFA availability in aquatic ecosystems caused by homeoviscous adaptation of primary producers to increasing water temperature, for example, diatoms in periphyton on stream bottom (Hixson & Arts, 2016;Taipale et al., 2018) will eventually cause a broader range of negative impacts on physiology, behaviour, and fitness than what we observed in this short (i.e. 8 weeks) feeding experiment. It has been shown that ~16℃ is a breakpoint for decrease of n − 3 LC-PUFA primary production in aquatic ecosystems (Hixson & Arts, 2016). This means that increasing temperature in temperate regions-where Atlantic salmon and other salmonid fishes are distributed-will reduce primary production of n − 3 LC-PUFA mainly from spring to autumn but in lower latitudes also during the winter (IPCC, 2013). Our study was conducted in winter and temperatures used (i.e. 13 and 18℃) correspond to the winter conditions at the southern edge of the species distribution (Jonsson & Jonsson, 2011; O'Briain, 2019), but salmon populations from higher latitudes will likely experience these temperatures mainly from spring to autumn.
Prey consumption and growth rate of Atlantic salmon is not strongly affected by the local adaptation to the thermal conditions in the stream of origin and both peak at temperatures between 16 and 21℃ (Jonsson et al., 2001). Our results suggest that juvenile salmon will, at these temperatures, likely experience lower mitochondrial efficiency and slower growth rate caused by reduced availability of dietary n − 3 LC-PUFA. Mitochondrial efficiency (Salin et al., 2015) and growth rate (Morgan & Metcalfe, 2001) are suggested to be related to fitness of wild fishes, thus our findings highlight the need for further studies on ecological significance of dietary essential fatty acids for populations of wild fishes.

ACK N OWLED G EM ENTS
Authors are grateful to Neil Evans, Magnus Lovén Wallerius, Samuel-Karl Kämmer, and Katharina Winter for their assistance with collection and processing of samples. are.13061684 and in Supporting Information .