Introduction

Good welfare is crucial in the development of a sustainable Atlantic salmon (Salmo salar) farming industry. Stress is reportedly known to negatively affect fish health and welfare and is gaining increased attention in aquaculture research and management (Sneddon et al. 2016). Perceived stress may translate into a range of whole-body responses such as impaired growth, reproduction, immune capacity and survival (Madaro et al. 2020). It is generally accepted that several husbandry-related operations or events may be considered stressful by reared fish, and one such event is sea transfer. Today, the majority of post-smolt rearing in Norway takes place in open sea cages (Directorate of Fisheries 2020). The process of relocating live salmon from hatchery to sea is an inherently stressful event including handling, crowding, pumping, transport as well as adjusting to a new environment (Norwegian Scientific Committee for Food Safety 2008). Indeed, the transition involves moving from controlled conditions with few environmental stressors to facing multifactorial environmental stress and a high degree of unpredictability. Mortality is often relatively high during the first 3 months after sea transfer (Nilsen et al. 2020), which could potentially be attributed to downstream effects of a sustained stress response. Meanwhile, there is little information available focused on fish stress in a commercial setting. The majority of stress experiments are performed in tank systems under controlled circumstances and are meant to simulate conditions and events in fish cages, when in reality these conditions cannot easily be reproduced. One study by Iversen et al. (2005) measured plasma cortisol levels in Atlantic salmon in commercial well boat transports and effects on survival after transfer to sea. To our knowledge, no efforts have been made towards evaluating the effects of sea transfer on stress-related gene expression in commercially reared Atlantic salmon.

In the present case study, we followed a batch of farmed fish from land-based breeding tanks through transportation to sea site and a period thereafter. We applied real time quantitative reverse transcription PCR (RT-qPCR) on gill tissue sampled at four different time points to assess a stress-related gene panel of interest. Gills were chosen as they are sensitive and multifunctional organs in immediate contact with the external surroundings. They are known to be affected by several types of stressors, both chronic and acute, and are also involved in the acclimation to a new environment (Takei & Hwang 2016). A panel of 12 target genes, presented in Table 1, were picked for analysis and included (in alphabetical order): alox5, cyp1α, hif1α, il4/13a, muc2, muc5, muc18, nrf2, pcna, phb, p38 and tnfα. Their corresponding gene products can be connected to acute and long-term physiological stress responses and belong to pathways involved in cell proliferation, metabolism and apoptosis, inflammation and immune functions, as well as oxidative/cellular stress.

Table 1 Gene products and main cellular functions of target genes
Full size table

Target genes were selected based on their well-known roles in stress as well as findings in Marcos-López et al. (2018). While the above-mentioned article focused on one of the most common stressors in aquaculture, namely pathogen infection (amoebic gill disease), we wished to explore how the biomarkers behaved under other stress exposures i.e. sea transfer. Moreover, recent research has directed attention towards the roles of mucins in stress (Benktander et al. 2021; Sveen et al. 2017).

Through evaluating the relative levels of target mRNAs we aim to discuss how commercial fish are influenced by the event of sea transfer and, in particular, the transition from a highly controlled setting at land to an unpredictable environment with potential multifactorial stress and major management operations at sea (e.g. net pen cleaning). We aim to explore the potential utility of selected markers as stress indicators and provide insight into the daily lives of salmon at a Norwegian farming facility to contribute to the ongoing understanding of fish welfare and stress.

Materials and methods

Fish rearing conditions and transport

Samples were collected during commercial salmon farming activity in spring 2020; before sea transfer (B), during sea transfer (S), 3 weeks (S3) and 6 weeks after sea transfer (S6). The experiment was initiated as samples from Atlantic salmon post-smolts (strain Rauma, SalMar ASA, Norway) were obtained from a commercial hatchery (Hjelvik Settefisk AS, Vågstranda, Norway). Fifteen fish were collected by sweep netting at each time point i.e. n = 60 in total. Fish were regarded as healthy by farmers i.e. screened for pathogens and healthy-appearing according to common operational welfare indicators (Noble et al. 2018). In the hatchery, fish were kept in five 40-m3 tanks with a stocking density of approximately 87 kg m−3 and reared under a 24-h light and continuous feeding regime with water temperature at 6.5 °C. Oxygen saturation varied between 85 and 105% and water flow at 1–1.5 fish lengths per second. At the time of the first sampling the fish had undergone smoltification (smolt status confirmed by ATPase activity measurements by Sintef Norlab, Mo i Rana, Norway) and had been kept on 50/50 seawater and freshwater (salinity 16‰) for the preceding 2 weeks to prepare for transfer to sea. Weight of sampled B fish was 102.8 ± 33.8 g (mean ± SD). Fish were starved 4 days prior to sea transfer, before loaded onto a wellboat by pumping and transported to a closely situated sea cage (Prophylaxia AS, Vikebukt, Norway). Seawater for transport (6.4 °C) was collected from this site. The transfer process from start to finish took approximately 6 h, including a transportation time of only 15–20 min from land to sea site. Group S were collected in the process of pumping fish out into the sea cage. Weight of sampled S fish was 85.4 ± 17.8 g. Fish were again collected 3 weeks and 6 weeks after sea transfer in association with the fish farmers’ routinely health check, in which no lice or apparent signs of disease were observed. Water temperature and fish weight for sampling S3 was recorded to 9.4 °C and 88.5 ± 17.8 g, respectively, and 12.8 °C and 89.8 ± 21.3 g for S6. All samplings were conducted from the same production unit/pen (approximately 190,000 individuals) to follow the same batch of fish throughout the experiment. Average weight estimates reported from staff were 95.1 g upon transfer (timepoint S), 130.3 g at the end of the month (18 days after transfer), and 288.7 g (48 days after transfer). Fish weights are discussed in “Limitations and future perspectives.” Accumulated losses for the first months after sea transfer (May–August) were 0.59%. We emphasize that environmental conditions and potential stressors in sea phase cannot be controlled and similarly reproduced, as each sea site and season is unique. As we wished to gain insight into the lives of salmon under regular commercial activity, fish farmers on site were asked to proceed with their normal routines throughout the duration of the experiment and to keep note of potentially relevant events in terms of stress during this period. It was reported that net pen cleaning had been performed 3 days before the last sampling (S6). An overview of samplings and potential stressors is illustrated in Fig. 1.

Fig. 1
figure 1

Timeline of events and context for samplings of commercial fish. B: Before sea transfer, S: during sea transfer, S3: 3 weeks after sea transfer and S6: 6 weeks after sea transfer

Full size image

Tissue sampling

Fish handling procedures were performed in compliance with the guidelines of Directive 2010/63/EU as well as Norwegian legislation. All fish were euthanized in a bath of lethal anaesthetic dose (Benzoak vet., 50 mg/l, ACD Pharmaceuticals AS). Tissue sampling from start to finish took approximately 5 min per fish. Fork length and weight were recorded (supplementary Table S1). Tissue (~ 50 mg) was collected from the left gill, second branchial arch using sterilized dissecting scissors and forceps. Samples were placed in individual tubes containing 1 ml RNAlater™ (cat. no. AM7021, Invitrogen) and immediately placed on ice. Samples were stored at –80 °C prior to subsequent analysis.

RNA extraction

Tissue samples (20–25 mg) were lysed and homogenized using a 600 μl RLT buffer (cat.no. 74101, Qiagen) and TissueLyser II (Qiagen) with 5-mm stainless steel beads in 2 × 2 min lysis at 25 Hz. Total RNA was extracted using RNeasy mini kit (cat. no. 74101, Qiagen), according to manufacturers’ protocol. On-column DNase digestion was performed with RNase-Free DNase I Set (cat. no. E1091, OMEGA Bio-tek). RNA integrity was confirmed by Qubit 4.0 Fluorometer (Invitrogen) and Qubit RNA IQ Assay Kit (cat. no. Q33221, Invitrogen). Recorded RNA integrity numbers (RIN) were consistently high, ranging from 9.2 to 10.0, implicating large/intact RNA in samples.

RT-qPCR

First strand cDNA synthesis was carried out using the qScript cDNA Synthesis Kit (cat. no. 95047, QuantaBio) according to manufacturer’s protocol, using 10 μl RNA (amount range from ca. 100–500 ng) in a final reaction volume of 20 μl. A 2720 Thermo Cycler (Applied Biosystems) was applied for cDNA synthesis. Quantification of cDNA was carried out using Qubit Fluorometer 4.0 and Qubit 1X dsDNA HS Assay Kit (cat no. Q33231, Invitrogen). Primer sequences for qPCR are presented in Table 2. Additional primer information is given in supplementary Table S2. qPCR was performed for twelve genes of interest, using β-actin (actb) and elongation factor 1 alpha (ef1α) as internal reference genes. Main functions of genes are as follows: oxidative stress (cyp1α, hif1α and nrf2), cell proliferation, signalling and apoptosis (pcna, phb and p38), immune genes and inflammation (alox5, il4/13a and tnfα) and mucins (muc2, muc5 and muc18). Primer sequences for pcna, phb, nrf2, alox5, il4/13a, muc5, muc18 and actb were obtained from Marcos-López et al. (2018), muc2 from Sveen et al. (2017), cyp1α from Sanden and Olsvik (2009) and hif1α from Olsvik et al. (2013). Primers for p38, ef1α and tnfα were designed using NCBI primer designing tool. All primers are prepared by Invitrogen.

Table 2 Primers for RT-qPCR for selected genes
Full size table

The efficiency of each primer assay (included in Table S2) was calculated by analysing a tenfold dilution series in duplicates. qPCR reactions were prepared using PerfeCTa SYBR Green SuperMix, Low ROX kit (cat. no. 95056, QuantaBio), according to manufacturers’ protocol. All samples were run in duplicates. Each PCR-well contained 7.5 μl SYBR Green SuperMix, 1 + 1 μl primer forward and reverse (see supplementary Table S2 for specific primer concentrations) and 400 pg cDNA; and RNase free water to a total volume of 15 μl. RT-qPCR was carried out using the AriaMx Real-time PCR System (Agilent Technologies). Reference gene stability for our conditions was validated using BestKeeper software tool (Pfaffl et al. 2004). No template controls (NTC) were confirmed negative (no Ct) as well as between-plate variation (< 10%). The following thermal profile for qPCR was used: initial denaturation step of 95 °C for 3 min, followed by 3-step cycling for a total of 38 cycles of denaturation for 15 s at 95 °C, annealing for 20 s at 58 °C and extension for 30 s at 72 °C. The specificity of qPCR reactions was verified by melting curve analysis (4-step continuous: 15 s at 95 °C, 1 min at 60 °C, 30 s at 95 °C, 15 s at 60 °C). Standard deviation was calculated for technical duplicates, and those with SD > 0.5 for Ct or Tm were considered unreliable and rerun.

Data analysis and statistics

Raw gene expression data were exported from the Agilent AriaMx 1.8 software and processed in Microsoft Excel (2008) (supplementary Table S4). As we are interested in relative changes of target gene expression levels between different time points and not absolute numbers of expressed molecules, the 2−ΔΔCt method is convenient, as described by Livak and Schmittgen (2001). An extension of the method to include multiple reference genes is outlined in Riedel et al. (2014), and in present calculations two reference genes actb and ef1a were used as endogenous controls for normalization. Relative quantification of gene expression was made with two different perspectives. Firstly, we aimed to assess the change in mRNA levels before and after transfer to sea, for which the applied calibrator group was pre-stressed post-smolts in hatchery i.e. group B. We also wanted to assess the change in mRNA levels after the event of net pen cleaning. Here, we used group S3 as the calibrator, as we assume these fish to be acclimatized 3 weeks after transfer and have a normalized gene expression for sea phase. Reference genes remained stable in all groups and did not appear to be influenced by external factors. Resulting figures depict changes in gene expression on the level of − ∆∆Ct, which provides readily interpretable information in terms of relative up and downregulation. Statistical analyses were performed in IBM SPSS Statistics 27 on the level of ∆Ct. Outliers were identified and excluded based on Grubb’s test (p < 0.05). Shapiro–Wilk test was conducted to evaluate normality. Significance of results was assessed with independent two-sample t-test (p < 0.05). Where data sets failed the normality test, non-parametric Mann Whitney U-test was performed (p < 0.05). The results for significance tests are included in supplementary Table S3.

Results and discussion

Two events: sea transfer and net cleaning

In the present work, we have evaluated transcriptional levels of a stress-related gene panel in the gill tissue of healthy commercial Atlantic salmon. We aimed to explore how gene expressions are affected as fish are moved from land-based tanks with constant conditions to sea cages where environmental factors may vary on a day-to-day basis. It is important to emphasize that the inability to control environmental conditions is an inherent feature of sea phase and a central aspect in the current study. Still, by evaluating relative expression (compared with control groups) through multiple timepoints, we are able to provide information on molecular responses towards stress-related events and a glimpse into the lives of commercially reared salmon.

Prior to conducting the analyses, we were expecting to see a change in gene expression in group S following sea transfer (handling, pumping, transport etc.) initiating an acute stress response, as well as the potential salinity stress of transfer to seawater. We were further expecting transcriptional levels to tend back towards pre-transfer levels after acclimation to sea phase i.e. S3 and S6 groups. Meanwhile, cleaning of pens was performed shortly before time point S6. Regular in situ net pen cleaning is performed to prevent overgrowth and negative effects of biofouling i.e. unwanted accumulation of living organisms on submerged artificial surfaces (Blöcher 2013). The cleaning process will release loose fouling particles from the net and contaminate the water, and the debris remains in and around the cage for a certain period of time, depending on current direction and intensity. As gills are sensitive towards waterborne pollutants, this may trigger a physiological stress response and influence gene expression. In addition, the presence of cleaning equipment itself generated turbidity or noise may also evoke a stress response (Comas et al. 2021). A study performed by Østevik et al. (2021) implies that the negative impacts of in situ net cleaning on gill health are relatively small and short lived (no significant difference 8 days after cleaning). However, the potential added stress of different management-related operations should not be neglected in terms of challenges that farmed salmon need to overcome. We therefore separate between two potentially stressful events when assessing the results: sea transfer and net cleaning. The results from RT-qPCR analyses are presented in Fig. 2 (sea transfer) and Fig. 3 (net cleaning).

Fig. 2
figure 2

Relative changes in mRNA levels in the event of transfer from land to sea. A Genes related to oxidative stress, apoptosis, cell signaling and proliferation: cyp1α, hif1α, nrf2, pcna, phb and p38. B Genes related to inflammation and immune functions and mucins: alox5, il4/13a, tnfα, muc2, muc5 and muc18. Fish were sampled during (S), 3 (S3) and 6 weeks after (S6) sea transfer and compared with calibrator group B (before sea transfer, baseline 0). Bar graphs represent arithmetic mean; error bars indicate the standard deviation. Significant alterations in gene regulation are marked with asterisk (*) (p < 0.05) or (**) (p < 0.01)

Full size image
Fig. 3
figure 3

Relative changes in mRNA levels in the event of net pen cleaning for full target gene panel: cyp1α, hif1α, nrf2, pcna, phb, p38, alox5, il4/13a, tnfα, muc2, muc5 and muc18. Fish sampled 6 weeks after sea transfer (S6), compared to calibrator group S3 (acclimatized to sea for 3 weeks, baseline 0). Bar graphs represent arithmetic mean; error bars indicate standard deviation. Significant alterations in gene regulation are marked with asterisk (*) (p < 0.05) or (**) (p < 0.01)

Full size image

Based on the selected target genes, the results imply that the process of relocation itself (S) was not particularly stressful for the fish. There is always the possibility, however, that the gene panel does not capture the stress of the fish in this particular setting, as the physiological stress response is dependent on context i.e. type of stressor as well as magnitude and duration of stressor exposure. It should also be noted that in this particular case wellboat transportation was brief i.e. 20 min from hatchery to sea site. Water quality (e.g. dissolved oxygen and pH) will not drop considerably in this time, or will there be a build-up of CO2 and ammonia (Erikson et al. 2022; Nomura et al. 2009). A distinction has been suggested of dividing transport times into two categories, namely short transport: < 8 h and long transport: > 8 h (Sampaio & Freire 2016). In this perspective, transport in the present study and the accompanying stress levels may not be representative for the typical sea transfer process in commercial aquaculture. Meanwhile, it appears that transition to sea phase has a more sustained effect on gene expression. Commercial salmon production normally takes place in two distinct environments; first, in strictly controlled settings in land-based breeding tanks. This typically includes a certain biomass under a constant artificial light regime and set parameters for water flow, dissolved oxygen and salinity etc. This is in great contrast to conditions in sea, where fish now need to cope with day/night cycles, varying turbidity and biomass distribution in the water column, variable weather and current, passing leisure traffic, other sea animals inside and outside of sea cages as well as water quality (temperature, salinity, dissolved oxygen levels and possible pollutants). In short, the transition to a new environment can be stressful and requires adapting to unfamiliar and uncontrolled conditions (Noble et al. 2018), and this potential multifactorial stress may impact gene expression levels. In the present results, we observe a clear change in expression of genes related to different metabolic pathways upon transfer to sea phase (Fig. 2A). In general, most genes involved in oxidative stress, apoptosis, cell signalling and proliferation show an apparent drop in expression levels, whereas mucins and immune-related genes appear non-affected. The pattern could be attributed to stress, however, considering the low mortality rate (0.59% first 3 months) and an apparent absence of lice and disease symptoms it is more likely stable adjustments in cellular processes as a result of changed surroundings. The same markers (cyp1α, hif1α, nrf2, pcna and phb) are not affected by the event of net cleaning (Fig. 3) when sea phase acclimatized fish are used for relative comparison. Conversely, we see a powerful upregulation of mucins and immunological markers (alox5, tnfα, muc2, muc5 and muc18) 3 days after this event, but no notable difference in the event of sea transfer (Fig. 2B). The results suggest the latter genes to be relevant markers in assessing stress impact of net cleaning.

The observed findings imply that the fish are essentially different organisms at land and in sea in terms of gene expression. When performing RT-qPCR on samples collected in situ, fish acclimatized to sea phase should be used for normalization when studying events at sea and vice versa for assessing events in land-based tank systems. This may be particularly true for gill tissue due to direct contact with external surroundings that can influence transcript levels.

Metabolic gene markers

P38 mitogen-activated protein kinase (MAPK) is a key signal transduction mediator that can be activated in response to a range of stressful stimuli e.g. exposure to environmental stress and is linked to numerous biological processes including inflammation, cell proliferation and apoptosis (Zarubin & Han 2005). In current results, p38 displays a marked change in expression both during transport (Fig. 2A) and 3 days after net cleaning (Fig. 3). p38 thus stands out as a potential biomarker for future studies when examining acute as well as long-term stress. By comparison, a commonly used indicator for stress — cortisol — is depleted and similar to control 3 days after an event of acute stress (Laberge et al. 2019). In mammals P38 belongs to a family consisting of four known isoforms α, β, γ and δ. In the present study, we are looking at the salmonid counterpart to P38β, which in humans is ubiquitously expressed in all tissues (Han et al. 2020) and which has been previously explored in Atlantic salmon (Holen et al. 2011; Marcos-López et al. 2018). Cell studies in mammals have suggested the P38β isoform to have anti-apoptotic properties (Otterbein & Zuckerbraun 2005). One alternative explanation for observed results might be decreased apoptosis during sea transfer and after net cleaning, where the onset of a stress response could lead to arrest of this energy-demanding process. It has been shown that mammalian cells can induce processes to inhibit apoptosis during environmental stress exposure, for instance hypoxia (Takahashi et al. 2013). cyp1a encodes cytochrome P450 1A, a hemoprotein belonging to a superfamily of enzymes involved in oxidative metabolism. CYP1A in fish is known to metabolize a variety of endogenous and exogenous compounds (Uno et al. 2012) and has been suggested a biomarker for aquatic pollution (Goksøyr 1995). A study by Leguen et al. (2010) found cyp1a gene expression in rainbow trout gills elevated during salinity challenge, implying that CYP1A may be involved in the acclimation of rainbow trout to seawater. In the present study, fish collected in sampling B had recently completed smoltification and were kept on 50/50 seawater and freshwater to prepare for sea transfer. It is possible that cyp1a was highly expressed at time point B, giving the impression of downregulation in later samplings (Fig. 2A). On a general note, the fact that fish were exposed to different salinities at different stages of the experiment is not optimal, and although selected target genes are not primarily involved in osmoregulation, this should be taken into consideration when interpreting results. Meanwhile, there is also a suite of other parameters that differ between commercial land- and sea-phase, and it is exactly the effect of these environmental conditions collectively that we have an interest in studying.

Cell proliferation marker pcna also shows significant downregulation after transfer to sea. Proliferating cell nuclear antigen (PCNA) is a highly conserved DNA sliding clamp involved in synthesis and repair of DNA and cellular responses to replication stress (Strzalka & Ziemienowicz 2010). As the relative expression of pcna is low after sea transfer, this may imply that cell proliferation is reduced at these time points and/or is substantial at time point B. In freshwater larvae, pcna mRNA levels were found to be reduced by salinity challenge (Martins et al., 2014), potentially implying salinity as a relevant factor for present observations. A recent study by Tang et al. (2022) also found pcna suppressed in post-smolt Atlantic salmon brain after transfer to sea cages. Nuclear factor erythroid 2-related factor (NRF2) is a marker of oxidative stress and acts as a transcription factor that binds to antioxidant response element (ARE) and promotes transcription of antioxidant genes (Mukaigasa et al. 2012). This pathway has been demonstrated in coho salmon (Oncorhynchus kisutch) and suggests the function of NRF2 to be conserved in salmonids (Ramsden & Gallagher 2016). In the present results, nrf2 expression is elevated (not significant) in S, which may reflect a low degree of acute oxidative stress in gill tissue during sea transfer. Hypoxia inducible factor 1 subunit alpha (HIF1A) is an oxygen-sensitive transcription factor functioning as a master regulator of gene expression in response to hypoxia (Ke & Costa 2006). Both acute and chronic hypoxia can distinctly affect mRNA levels of hif1α, which is a frequently used fish biomarker of hypoxia exposure (Emam et al. 2022; Olsvik et al. 2013; Rissanen et al. 2006; Østevik et al. 2022). The lack of a relative change in hif1α expression in the observed results implies normoxia both under and after sea transfer. Downregulation of phb is observed in groups S3 and S6. Prohibitin (PHB) is well-known for its protective role during oxidative stress and is further suggested to be involved in cell proliferation and survival (Mishra et al. 2010; Theiss et al. 2007). Meanwhile, little is known about the roles of PHB in the gill tissue of salmonids. Knockdown studies in mammalian tissues suggest that subcellular localization of PHB is essential for its function and whether it holds anti- or pro-apoptotic properties (Peng et al. 2015). The significance of the present results are difficult to interpret, and gill tissue-specific protein studies on PHB could be of relevance for future research.

Mucins and immunological gene markers

Biofouling communities on salmon cages have been demonstrated to act as reservoirs for parasitic agents (Tan et al. 2002). The present results (Fig. 3) implies that the presence of fouling particles in the water column mounts inflammatory responses and increased mucus production in salmon gills. tnfα and alox5 mRNA levels are still significantly upregulated 3 days after the cleaning of sea cages, in addition to mucins 2, 5 and 18. Tumor necrosis factor-alpha (TNF-α) is a pro-inflammatory cytokine, and arachidonate 5-lipoxygenase (ALOX5) is an inflammation marker that stimulates the transformation of free arachidonic acid (AA) to pro-inflammatory leukotrienes (Werz et al. 2002). Increased gill tnfα expression in the presence of environmental contaminants has been observed in other fish species e.g. common carp (Cyprinus carpio L.) (Özdemir et al. 2018). In mammals, both muc2 and muc5 has been shown to be stimulated by TNF-α (Ahn et al. 2005; Smirnova et al. 2000), and this correlation can also be seen in the present results. Mucins are a family of highly glycosylated proteins found in mucus in the epithelial surfaces of mucosal tissues (skin, gills and gastrointestinal tract) (Dash et al. 2018). Fish mucus aims to provide a stable physical and chemical barrier as a first line of defence against invading pathogens, and mucins with its gel-forming properties are a crucial component. There are two structurally distinct families of mucins; large secreted gel-forming mucins (e.g. MUC2 and MUC5) and membrane-bound forms (e.g. MUC18) (Pérez-Sánchez et al. 2013). Research has shown that different types of stress exposure lead to increased mucus secretion and number of mucus-producing cells to enhance prevention of pathogen entry (Benktander et al. 2021; Karlsen et al. 2018; Vatsos et al. 2010). Little knowledge currently exists about the mucin-encoding genes in fish. A study by Sveen et al. (2017) found gill muc2 expression upregulated with handling stress, and Marcos-López et al. (2018) found muc5 gill mRNA was significantly elevated during infection with Neoparamoeba peruans, implying a key role for this mucin in amoebic gill disease (AGD). Meanwhile, the latter study also saw a strong downregulation of muc18 in AGD-infected fish, in contrast to significant elevation in the present results. MUC18 in humans is involved in cell–cell interactions and cell migration and has been linked to tumour progression (Bai et al. 2015) and is also suggested to have pro-inflammatory functions in airway epithelial cells (Berman et al. 2014). A transcriptional tissue screening in gilthead sea bream found muc18 the most abundant mucin in skin, gills and stomach, implying a central role in these tissues (Pérez-Sánchez et al. 2013). The expression levels and functions of muc18 in Atlantic salmon tissues remain largely unexplored and should be further investigated.

Interleukin (IL)-4 and IL-13 are two closely related cytokines with similar functions, and in species three paralogues of the IL-4 and IL-13 cytokines have been reported; il4/13a, il4/ nd il4/13b2 (Sequeida et al. 2020). IL-4/13 are known to inhibit anti-inflammatory properties in mammals and play central roles in T helper 2 (Th2) cell responses (Huang et al. 2015). Similar functions for IL-4/13 have been suggested in gills of Atlantic salmon and rainbow trout (Oncorhynchus mykiss) (Takizawa et al. 2011). Correspondingly, in the present results il4/13a expression is low in S6 where expression of pro-inflammatory mediators is high. Since IL-13 has been found to induce muc5ac overexpression in human airway epithelial cells (Yu et al. 2010), it has been suggested that increased il4/13a levels may also stimulate muc5 production in fish (Marcos-López et al. 2018); however, this pattern is not detected in the current results (Fig. 2B).

Limitations and future perspectives

We emphasize the importance of performing research in field, as it provides insights into the actual lives of salmon in high-intensity aquaculture conditions. There are, however, some challenges to conducting experiments on fish in a large-scale farming level. As described in “Fish rearing conditions and transport”, fish farmers’ estimates for average fish weights in sea phase (based on sophisticated growth models) differed from the weights of our sampled fish, which were substantially lower. It is a known issue that sampling of Atlantic salmon from sea cages is size-biased and not necessarily representative for the population, whereas small (or large) individuals may be overrepresented by chance (Nilsson & Folkedal 2019). There may also be other factors to consider. As previously discussed, adapting to a new life at sea can be stressful which may itself negatively influence growth, and the transition has been known to make some individuals stop feeding or switch to a zooplankton diet for a period of time (Noble et al. 2018). We observed that after sea transfer the presence of “loser fish” appeared in the collected material (supplementary Table S1). Fish are typically defined as losers when their condition factor (K), a well-established indicator of the nutritional status of fish, is lower than 0.9 (Noble et al. 2018). So-called losers are emaciated fish with stunted growth and generally poor appearance and performance and are most likely moribound. They tend to swim slowly near the net at the surface and are therefore more likely to be captured by netting, typically resulting in overestimation of their abundance (Folkedal et al. 2016). This bias is well-known among farmers, and emaciated fish are often discarded from samples, for instance during lice counts, as they are not representative for the population. We wished to include them in our material, as they are relevant in terms of welfare assessment, and we observed no deviations in gene expression levels. Similarly, behavioral patterns could also be involved, as caged salmon is demonstrated to display size-dependent swimming depths and smaller fish tend to swim shallower in the water column (Folkedal et al. 2012). These factors could contribute to a lower weight of sampled material, compared to recorded estimates by staff on site. Nevertheless, the cage as a whole shows a solid increase in biomass, implying a normal healthy population.

While mRNA expression levels serve as a sensitive snapshot of the physiological status of a tissue, some caution should be made when interpreting the results. mRNA expression changes may be temporary depending on a suite of post-transcriptional modifications (Maier et al. 2009). To determine long-standing effects of stress exposure, future studies should aim to address accompanying changes in protein level. Up- and downstream mediators of presented stress-related gene products could be interesting for future exploration. The responses elicited by sea transfer or other operations, e.g. sea lice treatments, could also be compared with those in fish suffering from serious infection or disease, to assess the relative severity of different management practices in terms of stress exposure and potential negative effects on welfare and health.

Conclusion

Overall, the current study provides insight into the physiological stress response of commercial Atlantic salmon towards transfer from land to sea and the event of net pen cleaning. Transition to sea phase involved substantial changes in gill tissue gene expression and implies stable cellular adaptations to a new environment. Cleaning of sea cages appears to induce inflammatory responses and increased mucin production. P38 emerges a potentially interesting stress indicator for acute as well as long-term stress but requires further exploration. With reference to observed results, we advise that fish farmers carefully plan operational events to avoid accumulation of stress. Lastly, we emphasize the importance of performing research on commercially reared fish with numerous uncontrollable environmental variables.