High fish density delays wound healing in Atlantic salmon (Salmo salar)

In this study, we look closer at how high fish densities influence wound repair mechanisms in post-smolt Atlantic salmon. The fish were wounded with a 5 mm skin punch biopsy needle and stocked at two different densities, a high fish density (100 kg/m3) treatment and a low fish density treatment (20 kg/m3) serving as the control. The healing wounds were followed for 57 days with samples taken 1, 3, 7, 14, 36, 43 and 57 days post wounding. The transcriptomic analysis suggests that high fish density enhance inflammation and represses cell proliferation, tissue secretion and collagen synthesis in the healing wounds. The histological analysis further showed delayed epidermal and dermal repair in the high fish density treatment compared to control. The overall wound contraction was also altered by the treatment. In conclusion, high fish density enhances immune responses and delay tissue repair, which ultimately results in delayed wound healing.


Results and Discussion
Cortisol levels. The HFD treatment did not affect mortality. The overall mortality rate was low, <1% in the control and <5% in the HFD treatment. Similarly, Calabrese et al. 2017 did not observe increased mortality in fish reared at high densities (25-125 kg/m 3 ). Significant differences in plasma cortisol levels (2-way ANOVA, p < 0.001) were found both between time points and treatment ( Fig. 1). However, the post-hoc analysis only showed significant differences between groups at 43 days post wounding (dpw). Similar observations were done in our previous experiment investigating animal welfare and the effect of five different fish densities (25,50,75, 100 and 125 kg/m 3 ) 4,5 . In this previous study, plasma cortisol levels peaked in the intermediate density treatment (75 kg/m 3 ) after four weeks, whereas plasma cortisol levels in the highest density treatment (125 kg/m 3 ) peaked after eight weeks. Other studies have also demonstrated that fish exposed to crowding have limited cortisol response. Basal levels of plasma cortisol in unstressed salmonid fish are normally in the range 0-5 ng/mL, while crowding resulted in an elevation of plasma cortisol to 10 mg/mL 21 . These results suggest that cortisol as a sole indicator of animal performance during long-term intensive rearing conditions may be misleading and other parameters should be included in order to assess animal welfare.
Wound contraction. To investigate how HFD affected the overall wound morphology and wound contraction, the length, width, total wound area and non-pigmented inner area of the wounds was measured. The results showed that HFD altered all measured parameters, except the total wound area (Figs 2 and 3). As a measure for wound morphology, the length and width ratios (l/w) of the wounds were calculated. Wounds from the control fish had a higher (l/w) ratio (p < 0.01) compared to the wounds of HFD treated fish from 36 dpw and onward. Thus, wounds from the HFD treatment were contracting in a more circular manner compared to control wounds. Differences was also observed between the inner non-pigmented wound area, being larger at the last three sampling points in the HFD treated samples. Fish weight and wound position did not have any significant impact on wound contraction (ANOVA, p-value > 0.05, data not presented).

Microarray.
To better understand the molecular processes behind the alterations in wound contraction, the transcriptomic response in the wounds was measured with a 15k oligonucleotide microarray. The effect of HFD was strongest at 3 and 7 dpw, with 254 differentially expressed genes (DEG) at 3 dpw, and 206 DEG at 7 dpw (Table 1). It should be mentioned that the general transcriptomic profile in the two treatments were similar. None of the DEG changed direction, up vs. down, because of HFD. The effect of HFD was therefore only on the magnitude of transcription. Clustering of DEG with known roles (N = 652) was performed for functional interrogation of transcriptomic differences between the two treatments (Fig. 4a). The majority of genes in the first cluster were downregulated by HFD the first two weeks of the experiment. Most of these genes were involved in secretion, DNA replication and immunity (acute phase, chemokines and immunoglobulins), genes encoding components of mucus and collagens were also found in this cluster (Fig. 4b). Genes in the second cluster were in general downregulated by HFD during the whole experimental period. Genes in this cluster were involved in secretion and exocytosis. Genes within the third cluster were in general enhanced by HFD. Most of these genes were involved in immune functions such as eicosanoids, lectins, proteases, cytokines and chemokines. Overall, the cluster analyzes indicate that HFD in general enhance the immune responses while tissue regeneration was repressed during the first two weeks after wounding.

Inflammation.
As suggested by the cluster analysis, a wide range of immune genes were up-regulated in the HFD group, including eicosanoids, lectins, proteases, cytokines and chemokines (Fig. 5). Most of the inflammatory genes showed a transient response to HFD, with enhanced transcription levels at 3 and/or 7 dpw. This included multiple genes involved in the metabolic pathway of leukotriene B4 (Cytosolic phospholipase A2, Prostaglandin G/H synthase, arachidonate 5 lipoxygenase, lipoxygenase 3, leukotriene A4 hydrolase, cytochrome P450 4F3, leukotriene B4 receptor 1). Leukotriene B4 is known to be a major product of activated neutrophils and macrophages, with the ability to recruit and activate a range of immune effector cells 22 . Transcription of several proteases was enhanced by HFD at multiple time points. The Matrix metalloproteinases 9 and 13 (mmp9 and 13) showed a four-fold higher difference during the first two weeks of the experiment (Fig. 5). These proteases were the immune genes that were longest and strongest induced by the HFD treatment. Matrix metalloproteinases are secreted both by keratocytes and macrophages, and they are essential components of several wound healing processes (Schultz et al., 2005). Since both mmp9, mmp13 and leukotriene b4 are produced by leukocytes, these results may indicate stronger recruitment of inflammatory cells in the HFD wounds. In this context, it is relevant to mention that the transition from inflammation to tissue regeneration is dependent on matrix metalloproteinases activity [23][24][25] . Enhanced proteolytic activity, in particular mmp9 and mmp13, is reported as a key factor causing chronic and delayed wound healing in mammals 26,27 . Tissue repair. Transcription of genes involved in tissue repair was temporarily repressed by HFD. Several genes involved in DNA replication were repressed at 3 dpw (Fig. 5), including proliferating cell nuclear antigen (PCNA). At 1, 3 and 7 dpw, several collagens were down-regulated by HFD and this effect was further enhanced at 14 dpw with downregulation of multiple collagens, growth factors and mitochondrial genes. This transcriptional response shows several similarities with our previous study where a panel of collagen genes and growth factors were down-regulated in the skin of Atlantic salmon with cortisol injections 28 . Several genes involved in secretory functions and mucus responses were also dampened by HFD. The genes being strongest down-regulated by HFD was several transcripts of zymogen granule membrane protein 16. In mammals, this protein is found in mucous-secreting cells of the digestive system 29 , and therefore likely to be involved in mucous secretion in fish skin. Also glycosyltransferases, which are involved in glycosylation of mucins and two giant mucus proteins were down-regulated at 7 dpw in the HFD treatment. Further, our results indicate that transcription of muc5ac.1 was affected by treatment (p < 0.05), with lower transcription in the HFD treated fish at six out of seven time points (Fig. 6). Transcription of muc5b and muc5ac.2 changed during the healing process, but transcription was not affected by treatment. These results are contradictive to our previous findings where high biomass led to increased mucin transcription in the skin 4,30 . However, acute handling stress had the opposite effect reducing mucin transcription 30 . Synthesis and secretion of large amounts of high molecular weight proteins with heavy glycosylation represents a significant metabolic commitment of the cell 31,32 . Hence, the observed reduction in transcripts related to mucus production in the HFD treatment may be an allostatic response to a challenging environment.
Histology. In concordance with the wound measurements and the transcriptional results, histological analysis of tissue samples further confirmed that HFD resulted in delayed epidermal repair, scale mineralization and formation of dermis.  Epidermal repair. We observed a clear effect of the treatment on the epidermal layer at 3 dpw (Fig. 7).
Severe epidermal spacing was observed in five out of six HFD samples, compared to one in five control samples (Supplementary File 1). The epidermal spaces were caused by keratocytes with extended pseudopods, resulting in extracellular spaces in the epidermis ( Fig. 7d-f). Immunohistochemistry with PCNA further showed that cell proliferation at 3 dpw was mainly located in the epidermal layer ( Fig. 7g,h). The observations also suggests less proliferative activity in the epidermis in the HFD treatment, a finding also supported by the transcriptional results (Fig. 5). The observed epithelial spacing at 3 dpw could be a side effect of reduced epidermal cell proliferation. Another factor, which may have contributed to the reduced epidermal repair, could be enhanced protease activity in the wounds. In murine models, excessive protease activity provoked type-IV collagen degradation and resulted in delayed epithelial migration 26 . Similar observations with epidermal spacing have also been reported in Atlantic salmon with 5 mm punch biopsy wounds reared at low temperatures (4 °C) 16 . In order to further investigate the relationship between the environment and the morphology of the keratocytes, a cell culture experiment with primary keratocytes was performed. The stress factor in this experiment were low (4 °C) and high (16 °C) temperatures. The results showed that keratocytes reared at both high and low temperatures had longer pseudopods compared to the control cells (12 °C) (Fig. 5g-i). This morphology was similar to the epidermal cells at 3 dpw in the HFD treated fish (Fig. 5df). It is known that the mitotic rate of the keratocytes is temperature dependent 11 , however it is unclear if the observed epidermal spacing emerged due to reduced cell proliferation in the wounds or as a response to the environment.

Mucus response.
A clear difference in the mucus response between control and HFD treatment was observed at 7 dpw. At this time point there were less mucus and mucous cells on HFD samples (Fig. 8). The average mucus score of the wounds from HFD treatment was 1.7 (SEM ± 0.34) while the average score of the control samples were 3 (SEM ± 0.31) (Supplementary File 1). Combined with the transcriptional results with down-regulation of mucins, zymogens, transferases and giant mucus proteins (Figs 5 and 6), these findings strongly suggest that HFD dampens the mucus response at 7 dpw. The poor mucus response may be an indirect effect of the delayed organization of the epidermal layer at 3 dpw (Fig. 7). The rapid re-epithelialization process, followed by formation of a neo-epidermis and a mucus plug is believed to be essential in order to protect the healing wound 16,33 . Here we show that increased epidermal spacing and reduced mucus production is an early consequence of HFD, which in turn may affect the ability of the wound to withstand secondary infections.
Scale formation. At 14 dpw formation of new scales at the wound margins was observed in all the analyzed samples (Fig. 9a,d). In the HFD treatment, none of the scales contained mineralized matrix. In contrast, three out of five control samples contained scales with mineralized matrix. This trend continued at 36 dpw with weaker staining, on both right and left scale, in the HFD treated fish (Fig. 9b,c,e,f). As fish scales consist of an upper mineralized layer of hydroxyapatite and a lower fibrous layer of un-mineralized matrix and collagen fibers 34 , delays in collagen transcription as shown by the array results (Fig. 5), may cause a delay in scale formation and mineralization. In the European eel (Anguilla anguilla), vertebral bone demineralization has been observed after chronic cortisol treatment 35 . Further, Atlantic salmon vertebra is less mineralized when exposed to increased temperature stress, such as 16 °C compared to 12 °C 36 . The same has also been shown for salmon mesenchymal stem cells differentiating to bone cells in vitro 37 . Overall, HFD may have a negative effect on scale development and mineralization, which may impair the mechanical barrier of the fish.
Fibrous repair and pigmentation. Fibrous repair and restoration of skin pigmentation was delayed by HFD (Fig. 10). In normal fish skin, the pigment cells are localized below the epidermal and dermal layer (Fig. 10). At 36 dpw, four out of six control samples and only one out of six HFD samples had a layer of melanocytes below the epidermal layer (Fig. 10b,g). At 43 dpw, five out of six control samples had pigment cells organized in two layers, under the epidermal and dermal layer, while none of the HFD samples had this organization (Fig. 10c,h). These data support our findings that wounds from HFD treated fish retains a bigger non-pigmented area in the wound center (Fig. 3). Furthermore, the dermal layer looked more organized in the control samples, thus the pigment cells appear to follow the formation of connective tissue. At 57 dpw all samples had melanocytes organized beneath the epidermal and dermal layer, suggesting that tissue repair in the HFD treated fish was catching up on the control (Fig. 10d,i). The transcriptomic results also support this finding, with higher collagen transcription in wounds from HFD treated fish at 43 dpw (Fig. 5).
Cortisol treatment in Atlantic salmon and zebrafish (Danio rerio) 10,12 , and chronic stress situations in mice and humans is associated with reduce dermal repair and delayed wound contraction [38][39][40][41][42][43] . Given these results presented in this article, we believe that chronic stress and associated physiological responses is the best explanation for the delayed wound healing in the HFD treatment.

Conclusion
The results presented in this article show that HFD interferes with the wound healing capacity in Atlantic salmon. At the transcriptional level, the HFD treatment enhances inflammatory reactions in the wound while repressing tissue repair (cell proliferation, tissue secretion, and collagen production) (Fig. 11). The observed alterations in gene transcripts had a lag time, manifesting themselves at later time points on the morphological appearance of the wounds. These morphological differences included poor epidermal organization, delayed scale mineralization, delayed the formation of fibrous tissue and altered wound contraction. Combined, our findings suggest that HFD interferes with the wound healing capacity of the fish resulting in delayed epidermal and dermal repair.

Materials and Methods
Fish stock, rearing conditions and sampling procedure. This study was carried out at the Industrial Laboratory (ILAB, Bergen Norway) between November 4 th 2014 to January 30 th 2015. Smolts (mean size 80 g) were distributed randomly in two 500 L tanks. The fish were stocked at 20 kg/m 3 (N = 125) in the control treatment and at 100 kg/m 3 (N = 625) in the HFD treatment. From the 4 th to the 6 th of November the fresh water in each tank was gradually replaced with seawater. The specific water flow was adjusted to 5 L/min in the control tank and 25 L/min in the HFD tank, corresponding to 0,5 L/kg/min at the start of the experiment. Water velocity (10 cm/sec) in each tank was kept stable and equal by adjusting the angle on the inlet water pipe. The oxygen level in the outlet water was kept higher than 80% saturation by automatic oxygenation of the water in the header tanks (Oxyguard Commander). Both temperature (ranging from 9.4-10 °C) and oxygen saturation were measured once daily (YSI 550, Xylem Inc., Yellow Springs, USA). Following transfer to full strength seawater the fish were exposed to a 12 hours light and 12 hours dark regime. The fish were fed a commercial dry diet (EWOS, size 2-3 mm, Oslo, Norway) in 10% excess throughout the study. The main experimental period lasted from 28 th of November 2014 until 30 th of January 2015. The biomass was not adjusted in order to avoid adding additional stressors to the experiment. As a result the fish growth were greater than the biomass sampled in the HFD treatment, causing the biomass to increase over the time course of the study. The fish density in the HFD group ranged from 116 to 146 kg/m 3 , with a mean fish density of 126 kg/m 3 . The fish growth in the control treatment did not compensate for the biomass lost at sampling, thus the fish density was reduced over time, ranging from 6 to 22 kg/m 3 with a mean fish density of 14 kg/m 3 . Four weeks after transfer to seawater three biopsies (N = 90 per treatment), were excised with a 5 mm biopsy needle (Integra TM Miltex TM ) as described by several authors 16,44 . Prior to wounding the fish were fully anesthetized with (MS22, Sigma-Aldrich). Skin samples were taken at 1,3,7,14,36,43 and 57 dpw (N = 12 for time point and treatment). Fish for sampling were killed with an overdose of anesthetic (MS-222). Individual fish were weighed (g) and their length measured (cm). Blood was sampled with a heparinized syringe (Omnifix-F) from the caudal blood vessels and centrifuged (10 min at 4 °C and 4000 rpm). Plasma was stored at −80 °C for further analysis. Skin samples were collected from a standardized 1 cm 2 area around each wound. Samples for gene transcriptional analyses were snap frozen in liquid nitrogen and transferred to −80 °C for storage. The skin samples were fixed in 4% Paraformaldehyde solution (Electron microscopy science) overnight and then washed in 1 x PBS (Sigma Aldrich), before stepwise dehydration to 70% ethanol and transferred to −20 °C for storage.
In vitro-study, primary cell culture. In order to investigate the effect of temperature stress on keratocyte morphology, an in-vitro study with primary cell cultures were established. Keratocytes were cultured from scale explants, a method modified from previous work 45, 46 . In brief, Atlantic salmon (N = 9, weight ~ 500 g), were killed . Each well contained three scales, and for each fish three plates were used. Each plate were cultured at one out of three different rearing temperatures: control (12 °C), low temperature (4 °C) and high temperature (16 °C). The temperatures were chosen based on optimum (12-14 °C) rearing temperature for Atlantic salmon as both lower and higher temperatures are associated with reduced fish growth 47,48 . After four days the cells were microscopically analyzed (Leica).

Cortisol analysis.
A direct enzyme immunoassay (EIA) was used to measure plasma cortisol 49    Three independent researchers analyzed the samples blind for mucus, epithelial spacing and scale mineralization, all scores may be found in Supplementary File 1. The amount of mucus present on samples were scored in a 10X magnification area on a scale from 0-4. Score values were defined as 0 (no mucous cells), 1 (less than 15 mucous cells), 2 (more than 15 mucous cells partly forming a continuous layer), 3 (one continuous layer of mucous cells), 4 (two continuous layers of mucous cells). Epidermal spacing were scored on a scale from 0-3 in a 5X magnification area, with the following scoring values: 0 (normal epidermis with no-spacing between keratocytes), 1 (little epidermal spacing), 2 (occurrence of epidermal spacing) and 3 (severe epidermal spacing). Since quantification of alizarin in sections is complicated, scoring of alizarin stained scales was defined as those with the strongest staining gained the maximum score "high", those with less staining were classified as "low".  RNA extraction. Frozen skin sections with wounds were cut in half by a diagonal section, and repeated, and approximately one-fourth of the wound were transferred directly to 1 mL TRIzol (Thermo Fisher Scientific  and muc5b and the housekeeping genes elf1a and etif3 as describe by 30 . Two reference genes were evaluated for stability using the web-based comprehensive tool RefFinder which integrates the computational programs geNorm, Normfinder, BestKeeper and the comparative delta-Ct method 51 . The etif3 and elf1a obtained similar stability score, while the mean value of the two genes were most stable and was therefore used in the normalization procedure. Five replicates per group and time-point were included in analyses. Relative expression ratios to mean expression levels in intact skin (N = 10, equal amounts of control and HFD) were calculated and the data was log 2 transformed before data analysis and plotting.
Statistics. Data analyses were performed in R (version 3.3.1, www.r-project.org). Data series were tested for normal distribution (Shapiro-Wilk normality test, R-function shapiro.test()). If the test was passed (p-value > 0.05), data were analyzed by ANOVA (R function aov()) and in case significant differences were found (p-value < 0.05), a Tukey post-hoc test was calculated (R function TukeyHSD()). If the normality test was not passed, Kruskal-Wallis rank tests (R function kruskal.test()) were used. For evaluations of microarray results, the differentially expressed genes (HFD-control) were selected by the following criteria: log 2 -Expression Ratio >0.8| (1.75-fold) and p < 0.05. A complete gene list of DEG, gene identifier and their respective STARS category 52 , can be found in Supplementary File 1. The Euclidean distances were calculated, and the complete linkage clustering was drawn as a heat map. The dendrogram was pruned in order to identify 3 clearly-defined sub-clusters. For each cluster one-tailed Fisher tests for significant over-representation of functional categories (STARS) were run. Filtering, statistical analyses and plotting of results were performed in R.
Animal statement. This study was approved by the local responsible laboratory animal science specialist under the surveillance of the Norwegian Animal Research Authority (NARA) and registered by the national ethics committee (the Norwegian Food Safety Authority, ID7058). The methods were carried out in accordance with the relevant guidelines and regulations.

Data Availability
Data were submitted to Gene Expression Omnibus (GSE122142).