Protein changes as robust signatures of fish chronic stress: a proteomics approach in fish welfare research

In the verge of getting a greater understanding of fish welfare, sensitive technological tools, such as proteomics, may assist the aquaculture industry as it allows an unbiased approach for the discovery of potential biomarkers for stress monitoring. Stress is characterized by a cascade of physiological responses that end-up inducing further changes at the whole-animal level that might either increase fitness or impair welfare. Monitorization of this dynamic process, up till now relies on indicators that are only a snapshot of the stress level experienced. Within this scope, using gilthead seabream ( Sparus aurata ) as a model, three chronic stress conditions, namely overcrowding, handling and hypoxia, were employed to evaluate the potential of the fish protein-based adaptations as reliable signatures of chronic stress, in contrast with the commonly used indicators of primary and secondary stress responses.


Abstract Background
In the verge of getting a greater understanding of fish welfare, sensitive technological tools, such as proteomics, may assist the aquaculture industry as it allows an unbiased approach for the discovery of potential biomarkers for stress monitoring. Stress is characterized by a cascade of physiological responses that end-up inducing further changes at the whole-animal level that might either increase fitness or impair welfare.
Monitorization of this dynamic process, up till now relies on indicators that are only a snapshot of the stress level experienced. Within this scope, using gilthead seabream ( Sparus aurata ) as a model, three chronic stress conditions, namely overcrowding, handling and hypoxia, were employed to evaluate the potential of the fish protein-based adaptations as reliable signatures of chronic stress, in contrast with the commonly used indicators of primary and secondary stress responses.

Results
A large spectrum of biological variation regarding cortisol and glucose levels was observed, which values rose higher in net handled fish. In this sense, a potential pattern of stressor-specificity was evidenced since the magnitude of response and tolerance varied markedly from a permanent (crowding) to a repetitive stressor (handling). Gelbased proteomics analysis of the plasma proteome also revealed that net handled fish had the highest number of proteins with significantly altered abundance, compared to the other trials, whereas mass spectrometric analysis, followed by gene ontology enrichment and protein-protein interaction analyses, characterized those as humoral components of the innate immune system and key elements on the response to stimulus.

Conclusions
Overall, this study represents the first screening of more reliable signatures of physiological adaptation to chronic stress in fish, allowing the future development of novel biomarker models to monitor fish welfare.

Background
Managing welfare and preserving the well-being of fish in captivity is of increasing importance, whether by productivity concerns or by sustainability issues (1). Welfare definition, apart from being complex and controversial, remains with no clear consent on how it should be defined or measured (2,3). Limitations like divergent coping mechanisms, the lack of complete knowledge regarding the nociceptive system of fish (e.g. emotionallike states; cognitive abilities, pain, suffering) (4-7) and the lack of reliable physiological indicators of fish welfare make its investigation even more challenging (8).
The perception of threatening situations (stressors) by the fish is proceeded by a physiological cascade of events in the animal's biological system (9). This adaptive mechanism, known as stress response, enables the fish to cope with the stressor.
However, when unpredictable and uncontrollable situations exceed the organism's natural regulatory capacity, consequent negative effects can impair animal's welfare (10,11).
Such effects are commonly used as late indicators of the welfare condition of the individuals (12). The physiological stress response starts with the immediate neuronal induction of the chromaffin tissue to secrete catecholamines into the bloodstream, followed by a slightly delayed endocrine induction of the interrenal cells to release a corticosteroid hormone (cortisol in teleosts) into circulation (13,14). These hormones lead to a series of secondary responses that involve alterations in the energy metabolism, respiratory and immune functions and osmotic regulation (15). The rapid mobilization of energy substrates such as glucose (the fuel needed for the coping mechanisms) is caused by the activation of glycogenolysis in the liver or muscle, and hepatic gluconeogenesis, by the catecholamines and cortisol, respectively (16,17). Stressful stimuli can also lead to strenuous exercise fuelled by anaerobic glycolysis in the muscle, generating lactate which is released into plasma (18,19). Finally, tertiary responses have to do with prolonged stressor exposure and are maladaptively reflected in the whole-animal performance, like perturbations at the reproduction, immunological, growth and behaviour levels (14).
There are multiple stressors inherent to daily routines in an aquaculture rearing facility, ranging, in duration and severity (20). Recurrent acute or chronic stress can lead to physiological reactions distinct from those characterizing the response to a single acute stress. The cortisol response, commonly along with glucose and lactate, are the most used physiological indicators of stress and welfare status in farmed fish (21). Nevertheless, several studies demonstrate already that these indicators can be unreliable in cases of chronic long-term stress, mostly due to: (i) high biological variability of responses; (ii) cortisol levels return to basal levels within minutes/hours after an acute stressor; (iii) in cases of repetitive or chronic stressors, fish can adapt, to certain extent, and the cortisol response is lower; and (iv) its secretion might also be affected by intrinsic and extrinsic factors (e.g. age, sexual maturity, social status, level of domestication, prior experience, nutritional status) (22)(23)(24)(25)(26)(27).
In this sense, behavioural observations and plasma physiological indicators need to be complemented to form a robust scientific welfare assessment, and more advanced and sensitive technologies are gaining popularity. Proteomics are promising alternatives for the discovery of candidate molecular markers that can indicate physiological alterations due to stress exposure (28). Despite the limitations to the use of these technologies in the aquaculture field (29), several studies prove already the huge potential of proteomics for the identification of these stress signatures (30)(31)(32)(33)(34).
Few data are available concerning long-term coping with a chronic stressor and the consistency of the range of the universally used indicators in these cases. Considering this 5 gap in research, we aim in the present study, to assess comparatively the fish stress response at different levels (i.e., plasma stress markers, changes in plasma proteins' abundance and muscle biochemistry). Using gilthead seabream (Sparus aurata) as a model, three chronic stress conditions were employed, and proteomics was used to benchmark potential signatures of stress adaptation in the plasma proteome since several proteins resultant from physiological events are released into circulation. Gilthead seabream was the chosen species in this study since it is one of the most important species in European aquaculture with high commercial value. Considering the large fish physiological repertoire regarding stress adaptation mechanisms, this work aims to pioneer a better understanding of the underlying molecular mechanisms behind the fish physiological adaptation to long-term stress. Additionally, it aims to bridge the gap between the scientific community and the industry by paving the way for the development of novel biomarkers to monitor fish welfare.

Fish general condition
Fish were monitored every day during the trials reaching the end of the experimental period with a 100% survival rate. Fish overall condition and growth performance were also monitored (Additional file 1), and initial (IBW) and final body weights (FBW) were recorded for each experiment. Average body weights showed reductions at the end of the net handling (NET) and hypoxia (HYP) trials, in all groups, including the control. However, there were no significant differences in final body weights between the control and any of the stressed groups (P > 0.05), suggesting that weight reductions were unrelated to the stressor.

Plasma stress markers analysis
Circulating cortisol, glucose and lactate levels were measured in gilthead seabream submitted to different chronic stressors and in control fish (Fig. 1)
The onset and resolution of rigor mortis (Fig. 2) showed significant differences between treatments in the NET and HYP trials, specifically at 8 HAD (P CTRL− NET4 < 0.001), and at 8 (P HYP30− HYP15 < 0.001) and 24 HAD (P HYP30− HYP15 = 0.020), respectively. In the OC trial, all fish reached averaged maximum rigor strength at 24 HAD. In the NET trial, averaged maximum rigor strength was reached at 48 HAD in CTRL and NET2, and at 24 HAD in NET4 group. In the HYP trial, all groups reached averaged maximum rigor strength at 48 HAD.

Plasma proteomics analysis
A comparative analysis of the gilthead seabream plasma proteins between the control and the stress treatments detected, within the pH range of 4-7 and a molecular mass range of 11-114 kDa, 681, 752 and 681 protein spots for the OC, NET and HYP trials, respectively, by the gel image analysis software. After statistical analysis, 19, 360 and 34 protein spots within the OC, NET and HYP trials, respectively, were found to present significantly differential abundance (significance threshold at P < 0.05) between experimental conditions. From these, 7, 171 and 12 were manually excised from the 2D gels for MALDI-TOF/TOF MS analysis. No proteins were identified with significance for the OC trial. For the NET and HYP trials, 107 and 2 differential protein spots, respectively, were successfully identified by a combination of PMF and MS/MS search, with significant scores (protein score > 76, total ion score > 60, P < 0.05). Among the spots identified from the NET trial, 13 showed more than one significant protein identification (202, 326, 521, 559, 586, 604, 677, 877, 950, 959, 990, 996 and 1157), indicating that multiple proteins migrated to the same spots on the gel. The identified proteins are listed in an additional file (see additional file 2). A representative 2D-gel of the gilthead seabream plasma proteome is shown in Fig. 3.
Considering the number of proteins identified in each trial, only the proteins from the NET trial were considered for further statistical and bioinformatics analyses. At a log-fold change cut-off of ± 1.0 (P < 0.05) (Fig. 4-A), a total of 56 identified proteins spots (corresponding to 20 single entries) were considered significant, from which 19 were up- as haptoglobin, were found to be up-regulated. Fibrinogen alpha-chain was identified in two spots (521 and 544) and were both up-regulated. Alpha-1-antitrypsin homolog, apolipoprotein B-100, beta-actin, calcium/calmodulin-dependent protein kinase type II, leucine-rich alpha-2-glycoprotein, fetuin-B-like, hemopexin-like, hyaluronic acid-binding protein 2 and pentraxin were identified in a single protein spot each.
Hierarchical clustering (HCA) and principal component (PCA) analyses were performed for the identified 107 proteins spots with differential relative abundance across NET groups to check how well the samples grouped based on the expression patterns of the protein spots. Both the HCA (Fig. 4-C) and the PCA (Fig. 4-B) showed two main clusters belonging to the control and NET4 samples, while 2 biological samples belonging to the NET2 group clustered together with the control samples and 1 with the NET4 samples. The clustering of the spots revealed likewise two main groups, one displaying a pattern of higher and the other of lower relative abundance in stressed fish, when compared to control. The 107 differential protein spots were centralized into two principal components (PC), PC1 and PC2, which represented the maximum variation (65.6%) and the next highest variation and specific topological parameters were selected to demonstrate the importance and distribution of the nodes in the network: a darker colour intensity of the nodes indicates higher degree, while the size was estimated using the variation in protein abundance (fold-change). For every single entry, one protein spot was chosen as the most representative of each protein (Table 1), based mainly on the protein score and experimental molecular weight and pI close to the theoretical ones. From these 18 spots, 11 were down-and 7 were up-regulated, however, these differences in abundance were mostly significant (log-fold change > 1.0 or < -1.0, q-value < 0.05) for the NET4 treatment (only 2 were exclusively significant for the NET2 treatment and 2 were significant for both treatments). Thus, the fold-change of these 18 spots between NET4 and CTRL groups was used to estimate the size of the nodes on the PPI network, which ranged from − 4.04 to + 2.78. SERPINC1 (antithrombin-III), TFA (transferrin) and FGA (fibrinogen alpha-chain)  Table 1 String annotations and fold-changes of the proteins in the PPI network. Bold lettering in the "FC" column indicates significant fold-changes (> 1.0 and < -1.0). List is given in ascending order of spot number.

Discussion
The linkage between the underlying mechanisms behind the fish physiological adaptation to long-term stressors and reliable indicators able to measure unambiguously the stress condition of the fish, is a longstanding problematic in fish welfare research. Within this context, high-throughput technologies like proteomics are powerful techniques for the discovery of new potential biomarkers and lever the aquaculture industry with new approaches to its sustainability (28). To the best of our knowledge, few studies have used similar approaches to assess the potential of the fish protein-based adaptations as reliable signatures of stress, in contrast with the common and untrustworthy stress indicators i.e., cortisol, glucose and lactate. In this study we have shown that farmed gilthead seabream exposed to different chronic stressors exhibited a panoply of physiological responses, therefore, supporting the accepted idea that the commonly used indicators are far from the robust to assess the level of fish stress. Fish respond to acute and chronic challenges by the activation of a coordinated series of neuroendocrine pathways, in an attempt to regain homeostasis, but when this stability is surpassed, detrimental effects at the wholeanimal level are visible (14). These responses or effects are commonly used as indicators of compromised welfare and are rather late indicators and usually a point of no return. By this reason, cortisol has been the most commonly used physiological indicator of the primary response to stress (21). However, there is a shred of evidence indicating that this corticosteroid is not a reliable biomarker of long-term stress (27,(35)(36)(37), and our results are in agreement with this allegation. Gilthead seabream exposed to high stocking densities (45 kg/m 3 ) during the 54 days of the OC trial showed to reconfigure the cortisol secretion by expressing a tendency decreased response of this metabolite as compared to unstressed fish. Such consequence is suggestive of habituation to the stressor whereas a potential negative feedback has occurred when unstressed fish become stressed (i.e., consequence of the sampling on fish that did not have to adjust physiologically to previous environmental challenges). The same outcome was observed in juvenile gilthead 13 seabream confined for 14 days at 26 kg/m 3 (38) and in meagre cultured at different stocking densities for 40 days (39). In the NET trial, contrarily, plasma cortisol levels were significantly higher in handled fish, suggesting that gilthead seabream was not able to adapt to the handling stressor and it might indicate a mechanism of sensitization of the HPI axis. The fact that it was not constant (like overcrowding) but repeated instead, summed to its unpredictability and severity, could have prevented the possibility of accommodation to stress. Within this same context, under hypoxia conditions, was observed a chronic down-regulation of the cortisol response and here, either habituation or exhaustion might have occurred by overstimulation of the HPI axis. Exhaustion is however speculative since differences in behaviour were not monitored despite the reduction of activity seen alongside with the reduction of oxygen. Other explanation could be the fact that the decrease in food intake by prolonged hypoxia, reduces energy needs and thus oxygen demand, therefore lowering cortisol levels. Overall, the aforementioned observations suggest that the cortisol response and the capacity of adaptation were modulated by the nature of the stressor.
A prolonged and constant stressor might, therefore, induce a different physiological reaction compared to repeated acute stress. Studies report contradictory results regarding whether fish adapt similarly to prolonged chronic stressors and repeated acute stressors, but factors like species, age, sex and individual coping mechanisms seem to be ubiquitous and impact their adaptive processes (24,40,41). However, the process of habituation, characterized by a progressively weaker stress response after a prolonged stimulus, was already suggested and demonstrated in other studies as a result of the acclimation of the hypothalamus-pituitary-interrenal (HPI) axis (37,42), a mechanism that is not completely understood so far. High individual variability was also found in every trial, for both stressed and undisturbed fish, most likely due to individual differences in the stress response related to intrinsic factors of the animal, namely, their coping styles, the cognitive perception of the stressor and the value given to such challenge (38,43,44).
Additionally, values registered for control fish, in every trial, are higher than the reference values reported in the literature for this species (45). These discrepancies can have several causes and that is why cortisol should be used with caution when evaluating the magnitude of the stress response. The difficulty of measuring the resting levels of this hormone might also be one of these causes, since the lack of proper planning when sampling cortisol, or the manipulation needed to net and anaesthetize the fish, can result in high "control" cortisol levels that do not correspond to the "genuine" basal levels i.e., the non-manipulated fish levels. Also, it is well established that following the perception of an acute stressor, the levels of circulating stress markers increase within the first minutes or hours of stress response, returning to basal levels while time elapses, usually within 24 h (17,46,47).
Secondary physiological responses are characterized by an increase in glucose and lactate levels in blood plasma to satisfy the increased energy expenditure. Glucose release is primarily mediated by catecholamine-induced glycogenolysis and maintained by cortisolinduced hepatic gluconeogenesis (13). Thus, changes in glucose usually follow similar trends than cortisol after the stressor (14), which is corroborated by the levels of plasma glucose registered for all the three trials (Fig. 1). In this study, glucose levels, besides following the same trend as cortisol levels, are, in general, below the basal values for this species (45). This could be related to the fish's inability to maintain the same levels of glucose in the blood due to the high demand for glucose mobilization to other tissues. The decrease of plasma glucose levels in OC is consonant with the decrease in the cortisol levels, supporting the hypothesis of habituation, or by exhaustion of the endocrine system to the chronic stressor and depletion of energy reserves (27). The significant increases in the plasma glucose levels of stressed fish from NET and HYP trials are consistent with previous studies, which show that glucose rises during air exposure or low oxygen levels, due to stimulation of muscle glycogenolysis and hepatic gluconeogenesis, where glucose is synthesized to maintain the energetic substrates' demand (48). Similarly to cortisol, glucose and lactate circulating levels also return to basal levels within hours poststressor, which also make of these metabolites weak reliable markers in case of prolonged stressors (49,50). Additionally, studies also demonstrate that glucose variations in the blood are not only hormonal-induced due to stressful practices. Factors like variations in the water temperature and pH, anaesthesia, diet composition or fasting can also affect plasma glucose levels (51,52).
When tissue oxygen demands are exceeded, in the case of exhaustive exercise, air exposure or hypoxia, for instance, the response to a stressor can become anaerobic and trigger instant glycolysis in order to produce ATP, from stored glycogen, to meet cellular requirements, which consequently leads to lactate accumulation in the muscle (19,53). In this study, changes in circulating lactate are not in agreement with cortisol and glucose variations. Statistically significant differences in the lactate levels were only observed in the OC trial. In this case, if the cortisol response is indeed lower due to HPI-axis acclimation, as suggested before, the lactate recycling rate in the hepatic glycogenolysis is reduced, explaining the significant plasma lactate increase in stressed fish.
Additionally, previous studies show that during hypoxia and intense swimming activity, fish produces lactate in the muscle at a higher rate than it can be processed by other tissues (53).
Post-mortem muscle pH and rigor mortis have been used as tissue indicators of antemortem stress in numerous fish species (54)(55)(56). Stress at slaughter is the main contributor leading to an acceleration of the adenosine triphosphate (ATP) degradation rate before and after cessation of the blood circulation and oxygen supply, consequently influencing the post-mortem muscle pH and the onset and release of rigor mortis (57). The depletion of ATP reserves stimulates the anaerobic glycolysis in the muscle in order to maintain the energy expenditure. This process results in the accumulation of lactic acid, generating H + ions and consequently lowering muscle pH (57). The magnitude and rate of this pH fall depend on the fish's energy reserves prior to death, while these energy reserves are influenced by the intensity and duration of the pre-mortem stress. To our knowledge, no studies were performed in this species regarding the effects of long-term chronic stressors on the evolution of post-mortem biochemical processes. Results from this study (Fig. 2) followed the same pH trends as previous studies on gilthead seabream (58,59), however, comparing with the existent studies on pre-slaughter stress (55,60,61), muscle pH values immediately after death are below the ones found in this study, suggesting that slaughter stress was low in our fish. Poli et al 2005 state that in cases of exposure to a chronic stressor for a long time before death, the lactic acid produced can be gradually cleared from the muscle, but simultaneously the energy sources will likewise become gradually exhausted. Hence, when the fish is killed, muscle pH remains higher due to an early end of post-mortem anaerobic glycolysis caused by energy source scarcity.
This might explain the significant differences found in the HYP trial, where the highest pH values were observed in the highly stressed fish (HYP15), suggesting that these fish had lower energy reserves. Nevertheless, pH values registered after the 24 HAD, in every treatment, are in agreement with the reported by previous studies in this species at the same sampling times (58,62).
The onset of rigor mortis occurs with ATP depletion. When ATP reaches low levels, actin and myosin in the muscle bind together forming the actomyosin complex and causing stiffness of the fish body (63). Rigor mortis is inextricably correlated with muscle ATP and pH decline. A strong relationship between low muscle pH immediately after death, resultant of increased muscular activity at slaughter, and a rapid onset of the rigor state was demonstrated in a range of fish species (57,60). In this study, the evolution of rigor mortis (Fig. 2) was similar between treatments and significant differences were only found in the NET and HYP trials at 8, and at 8 and 24 HAD, respectively. A delayed onset was observed, starting between 2 and 6 HAD in every trial and reaching the maximum rigor index between 24 and 48 HAD. This delay is consonant with the high muscle pH registered at the time of death, suggesting low energetic reserves. A delayed rigor onset in rested and anaesthetized fish was also shown in previous studies (60). Measuring the glycogen and ATP content in the fish muscle and liver would be a complementary assessment to infer about the energetic reserves a corroborate our hypothesis.
Plasma proteins were evaluated in this study since blood plasma is a very informative biological fluid as it acts as a mirror of the physiological condition of the organism. Also, it is well known that different kinds of stressors result in a variety of immune changes, and proteins with key roles in immunological systems are the main components of blood plasma (42,64). Stress and stress-related hormones are recognized as modulators of the fish immune system (65), however, reactions depend on the intensity and duration of the stressor. The innate immune system is a fundamental defence mechanism in fish for protection against potentially harmful situations (66). The acute phase response is part of this system and it is mainly regulated by cytokines and glucocorticoids (67). This response is characterized by the release of acute-phase proteins (APP), by the hepatocytes, into circulation (68). APP can be classified as "positive" or "negative" depending on their plasma concentration increases or decreases during an inflammatory event (69). The response profile of our fish demonstrated the same tendency of protein changes.
This proteomics-based study of protein changes in the plasma after exposure of farmed gilthead seabream to multiple chronic stressors indicated that the fish's immune system was affected mainly by net handling and hypoxia stressors. Nevertheless, net handling was shown to be the most impacting. The levels of 20 different plasma proteins (distributed by 56 significantly differential spots), all related with immunological processes, were shown to be modulated by repetitive net handling, comparing to 2 proteins modulated by hypoxia. As mentioned, the same proteins were often detected from different spots on the 2D gels. Such a phenomenon can be due to existent isoforms or caused by adaptive changes of the proteome in an attempt to maintain cellular homeostasis under stress. This adaptation, in addition to protein abundance, may involve changes at the protein degradation level, localization, function and activity -all of which can be modulated by post-translational modifications (PTMs) (70). PTMs can regulate fundamental biochemical processes and be more energetically efficient than altering protein abundance, constituting potential interesting signatures of stress. Studies on PTMs in fish are still scarce.
The protein changes detected (listed in additional file 2), along with the network and GO enrichment analyses (Fig. 5) performed, confirmed the involvement of several components of the innate immune system in the physiological adaptation to these stressors. Proteins considered to be "positive" APP were likewise shown to be increased in abundance in the plasma of fish stressed by net handling (fibrinogen alpha-chain, complement component C3, haptoglobin, complement factor B, warm-temperature acclimation 65 kDa protein, alpha-1-antitrypsin), while proteins considered as "negative" were decreased (transferrin, inter-alpha-trypsin inhibitor, apolipoprotein A-I) (71). A diverse number of proteins involved in the APR was also previously found to be modulated in chronically stressed gilthead seabream (72).
Apolipoprotein A-I (Apo-AI) was only modulated by net handling stress and 17 proteoforms were identified in the plasma proteome map, being mostly decreased in abundance. Apo-AI is the main protein constituent of the high-density lipoprotein (HDL), playing a role in lipid metabolism and participating in the reverse transport of cholesterol from tissues to the liver (73,74). Apo-AI was also found to be decreased in abundance in crowded Atlantic salmon (75). In cod (Gadus morhua) it acted as a negative regulator of the complement system (76). Other two apolipoproteins were also found to be down-regulated in the plasma of fish from NET2 and NET4 groups (Apolipoprotein Eb and apolipoprotein B-100).
The complement system is an essential part of the innate immune system which can be activated through three pathways: the classical, alternative and lectin pathways (77). Fish display a plethora of complement components, mainly complement component C3 (C3), which may present around five proteoforms in a single species (78). C3 is one of the most abundant proteins in the plasma and plays a central role in the innate immune system, supporting the activation of all three pathways (77). In this study, C3, identified in 5 proteoforms, and complement factor B (Bf), identified in 4, were found to be increased in abundance by net handling. Contrarily, C3 was down-regulated in fish exposed to low oxygen levels. Bf also plays a role in complement activation by acting as the catalytic subunit of C3 convertase, an enzyme responsible for the proteolytic cleavage of C3, in the classical and alternative pathways (77).
Several metal-binding proteins, existent in the plasma of vertebrates, can chelate iron, zinc and copper, which are essential elements for the virulence of bacteria (79). Alpha-2macroglobulin (A2M) is a multifunctional protein (80) found to be down-regulated in the plasma of fish submitted to handling stress. It is mostly known to act as a broad range serine proteinase inhibitor and to bind metal ions (79). Contrarily, haptoglobin, which is also responsible for the sequestration of iron by binding to hemoglobin, was found to be increased in the plasma of handled fish. Similarly, warm-temperature acclimation-related 65 kDa protein (Wap65), which is involved in the scavenging of free heme (81), was increased in abundance by net handling and hypoxia stressors. Wap65 in fish is the homologue of mammalian hemopexin (82) and in most teleosts presents two proteoforms (83). In this study, two spots were also matched to this protein suggesting the presence of these two proteoforms. Transferrin (Tf) decreased in abundance in the plasma of fish stressed by net handling. Tf is a plasma protein also capable of binding iron and an important constituent of the iron homeostasis (33).
In fish, antiproteases are important participants of the non-specific humoral immune defence mechanism (71). A2M is an important factor in this mechanism. Alpha-1antitrypsin is a serine protease inhibitor, up-regulated in net-handled fish, which is responsible to negatively regulate blood clotting molecules to prevent thrombosis (84).
Inter-alpha-trypsin inhibitor H3 is also a serine protease inhibitor, which was found to be down-regulated in the plasma of fish from NET groups. Same protein changes were verified for fetuin-B, a cysteine proteinase inhibitor recently described in teleosts (85).
Fibrinogen alpha-chain, a beta-globulin involved in blood clotting, an integral part of innate immunity (84), was found to be up-regulated in the plasma of fish belonging to NET groups.

Conclusions
In summary, the results suggest that the physiological changes were higher in fish exposed to repeated handling and that mild and permanent stressors may allow the fish to refine their physiological processes to adapt to the challenges. The mechanism of habituation, along with the return to basal levels after a few hours and all the factors presented affecting the release of cortisol, glucose and lactate, demonstrate that these indicators may not be the most robust in case of chronic stress monitoring. On the other hand, plasma proteomics allowed to detect, in stressed fish, a cohesive network of protein changes associated with essential immunological pathways, which will be useful in understanding the biological processes behind protein-based stress adaptation in fish and may, therefore, represent the first screening for potential biomarker candidates of chronic stress in gilthead seabream. Nevertheless, the energetic cost associated with mounting the stress response (allostatic load) during long-term consequences of repeated or prolonged exposures to stress, still need to be answered in the context of other necessary life functions, namely, growth, development, disease resistance, behaviour, and reproduction. The fish were then individually weighed and distributed among conical fiberglass tanks (500 L), according to the density requirements of each trial. The tanks were supplied with natural flow-through seawater from Ria Formosa, and kept under natural temperature (13.4 ± 2.2 °C) and photoperiod, salinity at 34.7 ± 0.8‰, and artificial aeration (dissolved oxygen above 5 mg.L -1). Fish were fed by hand once a day, with a diet manufactured by AquaSoja Portugal, following the species' nutritional requirements.

Experimental design
The study was performed in three separate trials: (1) Overcrowding (OC), (2) Net Handling (NET) and (3) Hypoxia (HYP), due to logistic issues. Each trial followed a 2-week acclimation period and the initial rearing density was established at 10 kg/m 3 (except in the experimental groups of high stocking densities). In the OC trial, during the 54 days of experiment, fish (initial body weight (IBW) = 372.33 ± 6.55 g) were stressed using different high stocking densities, by increasing the number of fish in the tanks. Three different experimental groups were tested in triplicate: Control -10 kg/m 3 (OC CTRL ), medium density -30 kg/m 3 (OC 30 ), high density -45 kg/m 3 (OC 45 ). The NET trial lasted for 45 days and the fish (IBW = 375.69 ± 11.88 g) were stressed by 1-min air exposure, using nets designed to fit inside the tanks and to be lifted to perform the stressful event.
The experimental groups were established, in triplicate, as follows: Control -undisturbed fish (the net was also placed in the tanks but not lifted) -(NET CTRL ), fish air-exposed twice a week (NET 2x ) and fish air-exposed four-times a week (NET 4x ). In the HYP trial, fish (IBW

Consent for publication
Not applicable.

Availability of data and materials
The authors declare that all relevant data supporting the findings of this study are available within the article (and its additional files).

Figure 2
Post-mortem changes in muscle pH and rigor mortis of gilthead seabream (Sparus aurata) submitted to different chronic stressors (A -overcrowding, B -net handling, C -hypoxia), in two intensities, and unstressed fish (control), stored in ice for 72 h. Data points are the mean ± S.D. of n = 9 for each sampling time.
Means labelled * are different at P < 0.05.

Figure 3
Representative pattern of gilthead seabream (Sparus aurata) blood plasma on a 12.5% polyacrylamide 2D gel. Black circles represent the 107 proteins identified by MALDI-TOF/TOF MS with significant differences in abundance in NET groups and black squares the 2 proteins with significant differences in abundance in HYP groups (P < 0.05).
43  A -Protein-protein interaction network generated with 18 differential proteins identified in the plasma of fish from NET trial. Nodes represent proteins and edges the functional associations between them. STRING annotations are described in Table 1