The trial was conducted at the Aquaculture Research Station, Tromsø (Norway) using a flow-through rearing system. The fish were treated according to Norwegian legislation and approved by the Food and Safety Authority under the ID project number 4986: “The use of functional feed to improve the overall performance of Atlantic Salmon smolt.” A batch of 1476 Atlantic salmon parr (IPN sensitive, AkvaGen strain, Kyrksæterøra, Norway) were reared at 6 h light/18 h darkness and 8°C (winter stimuli). The fish were sedated with Benzoak (benzocaine; 0.1 mg/ml), pan-jet marked with Alcalian blue subcutanously in the abdomen and allocated into six 500 L tanks (246 fish per tank, average weight 35 ± 3.4 g). Fish were acclimated for 1 week on a commercial diet and changing lighting conditions to a 24 h light regime and 12°C (summer stimuli). The oxygen saturation was maintained above 85% throughout the experiment. The feeding trial lasted 12 weeks involving 6 weeks in freshwater during smoltification (24 h light, 12°C) and 6 weeks in seawater (24 h light, 12°C). The smoltification process was followed by 24 h seawater (35‰) challenge tests (Blackburn, 1987) to confirm that the fish were all optimally smoltified by the time of SWT.

For analysis of the distal intestinal microbiota, the digesta from three fish per tank from each sampling were pooled (n = 3), whereas microbiota from the mucosa was analyzed from two individual fish per tank (n = 6).

Distal intestine from five fish during the seawater stage, i.e., 2 from two tanks and 1 from the third tank (n = 5) were sampled for gene expression analysis. Each target gene was normalized using the geometric average expression of two reference genes (elongation factor 1 and beta-actin). The primer sequences of the genes evaluated in this study are listed in Table 2.

Histology was performed on fixed distal intestinal sections from three fish per tank from each sampling point (n = 9). Samples were dehydrated in graded ethanol concentrations prior to embedding in paraffin wax. For each specimen, multiple transverse sections (5 μm) were stained with haematoxylin and eosin (HE) as well as periodic acid–Schiff (PAS). Images from histology were taken from each DI section using a light microscope and analyzed with the Image J version 1.36 (National Institutes of Health, United States). The mucosa fold length average was measured in at least 15 well-oriented folds per section stained with HE. Fold length was only measured in primary folds with at a minimal length of 200 μm, complex folds were not taken into account. Goblet cells were counted in PAS stained sections and counted across a distance of 200 μm in at least fivefold per section and then averaged. The perimeter ratio (PR) of each intestinal section [arbitrary units (AUs)] was measured using the external perimeter (EP) and lumen perimeter (LP) and calculated by the formula: PR = IP/EP.

An IPNV challenge was conducted using a virulent strain of IPNV [S-IPNV-TT96 isolate, serotype Sp (Johansen and Sommer, 2001)]. One passage away from primary isolation was used for the challenge. The virus isolate was propagated and quantified in Chinook salmon embryo-214 (CHSE-214) cells (Lannan et al., 1984) as described previously (Johansen and Sommer, 2001). Virus susceptibility was tested on the same batch of Atlantic salmon prior to challenge at SWT (start of week 7). After 6 weeks in FW, 2 × 30 fish (pan-jet marked to identify the tank of origin) from each experimental tank were transferred to SW in 2 tanks at the challenge unit. Three days post SWT, the fish were challenged through co-habitation by adding 20% IPNV infected cohorts (injected with 0.2 ml IPNV, dose 2,6 × 10⁶ infectious units per fish) to each tank. The temperature during the challenge trial was 10°C. Throughout the challenge, both tanks were fed the control SW diet. The tanks were monitored for 48 days following the challenge, with daily registrations of moribund/dead fish. The IPN status of dead fish was verified using an IPNV rapid co-agglutination (Hall et al., 1994) test developed by the National Veterinary Institute in Norway (Taksdal and Thorud, 1999). Verification of viral infection was also conducted by registrations of macroscopic signs of disease and isolation of IPNV from dead fish by titration of head kidney homogenates onto CHSE-214 cells in 96-well plates.

Statistical comparisons in all the analyses were conducted between the control and probiotic groups at the same sampling point, except in the LEfSe analysis where the effect of the SWT was evaluated comparing control groups between freshwater and seawater. To investigate experimental group differences between bacterial communities (beta diversity), the software package PRIMER-E v.6 PERMANOVA+ was used (PRIMER-E, Plymouth, United Kingdom) (Clarke and Gorley, 2006). Beta diversity was calculated in QIIME based on weighted and unweighted UniFrac metrics. Dissimilarity matrixes from UniFrac were imported to PRIMER-E to evaluate significant differences between groups by permutational multivariate analysis of variance (PERMANOVA). To assess alpha and beta diversity indexes of mucosa samples, a linear mixed model was applied using water habitat and diet as fixed factors, and tank as random factor. For digesta samples, a regular linear model (two-way ANOVA with water habitat and diet as fixed factors) was applied since samples were pooled per tank prior analysis.

Differences in the relative abundance of OTUs between groups were analyzed with LEfSe (Segata et al., 2011), available at http://huttenhower.sph.harvard.edu/galaxy/ using the default parameters. This tool first identifies significant differences among experimental groups and then evaluates whether these differences are consistent with other features such as the phylogenetic affiliation of the OTUs. LEfSe implements different statistics test involving firstly, a non-parametric factorial Kruskal–Wallis rank sum test; secondly, a pairwise test using Wilcoxon sum-rank test; and finally, linear discriminant analysis (LDA) to estimate the effect size of each differentially abundant OTU. To detect the impact of SWT on the potential differences of OTUs with LEfSe microbiota data from the control group was used.

Survival curves from the in vivo viral challenge were calculated by the Kaplan–Meier method and compared by log rank (Mantel–Cox) pairwise comparison testing for equality of survival pattern between groups using pooled replicates. Statistical analyses were performed using SPSS version 11.0 (SPSS, Inc., Chicago, IL, United States) with significance determined at P < 0.05.

The results of all other analysis (i.e., alpha diversity, gene expression, and histological evaluation) were analyzed using JMP Pro 14.3 (SAS Institute, Inc., Cary, NC, United States). The data was checked for normality (Shapiro–Wilk test) and homogeneity of variance (Tap et al., 2009). Data fulfilling parametric test assumptions was analyzed either by t-test or two-ways ANOVA (followed by Tukey HSD). Data that did not fulfill parametric test assumptions were log transformed to achieve normality or otherwise analyzed with a non-parametric test such as the Mann–Whitney U-test. All data are presented as mean ± standard deviation (SD), and significance was accepted at P < 0.05.

Growth performance was comparable between groups in both freshwater and seawater in terms of SGR and TGC (Table 3). The performance was judged in accordance with expectations for fish at this age and this type of diets.

The expression of a panel of immune, stress, and apoptosis related genes in the DI was measured during FW and SW stages to evaluate the effect of the probiotic diet compared to the control diet (Figure 7). Genes related to antiviral protection were modulated in FW and SW by the probiotic diet; mx1 and tlr3 levels were lower in fish fed the probiotic diet in FW, but higher in the SW in comparison with fish fed control diet. Further, pcna and tnfa were significantly higher in the probiotic group than in the control group in seawater. A higher expression of il-1b was seen in the fish fed the probiotic diet in FW while hsp70 was not affected by the probiotic treatment regardless of the water habitat.

The parameters evaluated by light microscopy were not significantly different between the control fish and the probiotic fish (Table 6). The histological evaluation of distal intestinal morphology in control and probiotic groups during both FW and SW did not show any signs compatible with an active inflammatory response. The histological structure was characterized by a finger-like mucosal fold architecture, covered with an aligned epithelium of a single layer of enterocytes with supranuclear vacuoles in the apical zone, a thin lamina propria and low abundance of intraepithelial leukocytes suggesting a good health status of the fish.

The first IPNV related mortalities were registered in fish from the control group at 19 and 20 days post-challenge (in the two replicate challenge tanks), while the first mortality in the probiotic group was registered at 21 and 25 days post-challenge (Figure 8). Average cumulative mortality at termination (day 49) was 26.8 ± 3.2% for the control group and 21.7 ± 2.9% for the probiotic group. These differences in mortality were not statistically significant.

The use of molecular methods such as high-throughput sequencing has rapidly expanded our knowledge of the bacterial communities in the fish intestine (Zhou et al., 2014; Tarnecki et al., 2017). In the present study, the main phyla observed in the DI of Atlantic salmon during both freshwater and seawater stages were Firmicutes, Proteobacteria, Fusobacteria, and Actinobacteria. Previous studies investigating the intestinal microbiota of Atlantic salmon under farming conditions have also found these phyla as normal residents of the intestine (Zarkasi et al., 2014; Gajardo et al., 2016, 2017; Schmidt et al., 2016; Dehler et al., 2017a; Lokesh et al., 2018; Rudi et al., 2018). Despite heterogeneous experimental conditions and approaches among studies, the phyla Firmicutes and Proteobacteria were consistently reported as dominant bacteria in the intestine of Atlantic salmon. Although some of the phyla were consistently observed in the DI of fish throughout this study, the SWT had a major role in modulating the abundance of specific bacterial communities (significant effect on four phyla: Firmicutes, Fusobacteria, Proteobacteria, and Cyanobacteria) at the digesta and mucosa level of the DI as evidenced by the LEfSe and beta diversity results. However, our results on the digesta microbiota must be cautiously interpreted since the pooling of 3 fish per sample may have impacted the possibilities to detect differences.

Even though several observations from the study of Dehler et al. (2017b) were based on data from digesta, our results are in agreement with their findings on the important role of SWT on modulating the intestinal microbiota of Atlantic salmon. Another important finding detected in both studies is the significant decrease in the relative abundance of several LAB genera after SWT. On the other hand, some important differences should be noted between the two studies, which could suggest the importance of considering both digesta- and mucosa-associated microbiota in such studies. In particular, we detected a strong increase in the anaerobic bacteria Fusobacterium and Cetobacterium in the mucosa, suggesting a potential strong contribution of Fusobacteria in Atlantic salmon during SW stages which was not observed in the study of Dehler et al. (2017b). Fusobacteria is often mentioned as part of oral and intestinal microbiota in humans (Chen and Jiang, 2015; D’Argenio and Salvatore, 2015). Some reports associate the presence of Fusobacterium spp. with different human pathologies (Kostic et al., 2012; Han, 2015). However, in fish, its clinical relevance remains poorly described. Cetobacterium has often been reported to be part of the gut microbiota of several fish species and also reported as a species with the potential to produce vitamin B12. Furthermore, Brugman et al. (2009) reported that a vancomycin treatment in zebrafish with enterocolitis increased the abundance of Cetobacterium somerae, which was associated with a reduction of inflammation.

We did not detect a strong contribution of Mycoplasma spp. (phylum Tenericutes) in the digesta of SW fish, as opposed to several other studies (Holben et al., 2002; Abid et al., 2013; Llewellyn et al., 2015; Jin et al., 2019) that have detected this genus as a substantial contributor of Atlantic salmon microbiota in SW and even as a putative member of the core microbiota (Dehler et al., 2017b). The reasons behind this result are not clear but the reported impact of environmental factors, dietary regimes, sampling strategy, and host genetic variation on the intestinal microbial communities of fish may explain the differences observed between studies (Tarnecki et al., 2017).

To the authors’ knowledge, our study is the first to report the effect of SWT on the mucosa-associated microbiota of Atlantic salmon. In both fish and mammals, it has been recognized that the digesta and mucosa compartments harbor substantially different microorganisms (Eckburg et al., 2005; Looft et al., 2014; Gajardo et al., 2016, 2017; Lyons et al., 2017). Thus, these two different microbial communities may have different roles in the intestine and may differently impact the health of the host. In fish, most of the studies investigating factors that modulate the gut microbiota have so far focused on the so-called allochthonous microbiota, which is associated with fecal or digesta samples. However, in mammals, several authors have speculated that the mucosa-associated microbiota could have a central role in modulating the intestinal physiology of the host. This hypothesis is based on the fact that the mucosa-associated bacteria could interact with the host epithelium both directly and indirectly, while those in the digesta only interact with the host indirectly (as reviewed by Van den Abbeele et al., 2011). The latter would suggest that the intestinal microbiota shift caused by SWT could have large implications for the gut health of salmon following SWT, hence supporting the particular interest to further study the impact of certain taxa associated with the mucosa, such as members of Fusobacteria, on salmon health. Prior to and during SWT commercially raised Atlantic salmon undergoes important physiological changes (i.e., smoltification) and is exposed to changes in salinity and to dietary modifications implemented to meet the nutritional requirements of the new smolt stage. Thus, this study did not aim to identify the weight of every single factor separately but highlight the extent of the impact of all the factors together on the intestinal microbiota under conditions that resemble the current commercial practices of Atlantic salmon production.

Salinity is a well-known factor affecting microbiota by limiting or promoting the establishment of specific bacterial communities in given environments (Lozupone and Knight, 2007; Canfora et al., 2014). The effect of salinity and SWT on bacterial communities associated with fish has been previously studied (Schmidt et al., 2015; Lokesh and Kiron, 2016; Lokesh et al., 2018; Rudi et al., 2018; Jin et al., 2019). A study conducted by Zhang et al. (2016) even demonstrated that the bacterial community associated with the fish intestine also respond to salinity changes. Nevertheless, the changes in intestinal microbiota during SWT is most likely due to the combination of several factors that will lead to a competitive niche appropriation by certain microbial species hence reorienting the bacterial composition in the gut. In particular, the host as “habitat filter” influenced by the shift in osmoregulatory functions of the gut and the altered immune response after SWT (Johansson et al., 2016; Karlsen et al., 2018). Based on our results we propose using SWT experiments, coupled with meta-transcriptomic and metabolomics analysis, to assess the functional role of the intestinal mucosa-associated microbiota in salmon during FW to SW transition.

The use of probiotics in aquaculture has been implemented as a regular practice to improve the health and performance of fish under stressful farm conditions (Lauzon et al., 2014; Pérez-Sánchez et al., 2014; Hoseinifar et al., 2018, 2019). The beneficial effects of some probiotics in humans and animals are well-documented, but more research is necessary to ascertain their modes of action. In aquaculture, this limitation is also evident. Some of the proposed possible modes of action of the probiotics are suggested to be mediated by the modulation of the microbiota (Merrifield and Carnevali, 2014; Wang et al., 2015). The present study investigated the effect of the only dietary probiotic currently registered for use in aquaculture in the European Union, namely P. acidilactici MA18/5M. The scientific literature about the use of this specific probiotic strain in aquatic animals is continuously increasing (Castex et al., 2008; Ferguson et al., 2010; Merrifield et al., 2011; Lamari et al., 2013, 2016; Standen et al., 2013). Currently, two published studies reported the role of this probiotic in the intestine of Atlantic salmon during the seawater stage (Abid et al., 2013; Vasanth et al., 2015). Both studies were able to identify a positive effect of this bacterium in the intestine of Atlantic salmon; however, only the study from Abid et al. (2013) investigated the effect on the intestinal microbiota. In the present study, the genus Pediococcus was detected by high-throughput sequencing in digesta and mucosa samples from both groups, supporting the fact that Pediococcus is a ubiquitous genus encountered in the intestinal microbiota of salmonids. Indeed this genus has been previously reported as a normal inhabitant of the gut microbiota of Atlantic salmon (Merrifield et al., 2014; Gajardo et al., 2016; Dehler et al., 2017b) and rainbow trout (Araújo et al., 2016). In our study, our sequencing strategy did not allow to differentiate autochtonous Pediococcus species/strains from the one used in the probiotic diet. Nevertheless, the Pediococcus genus was detected at a higher level in samples from probiotic fed fish compared to the control group. Focusing on the probiotic group, overall the abundance of this genus was low in both water habitats as previously established by our group (unpublished data), with the higher relative abundance in the digesta of FW fish and the lowest abundance observed in the mucosa samples during the seawater stage. Lower abundance of P. acidilactici in mucosa compared with digesta in Atlantic salmon is in line with the study from Abid et al. (2013). Other studies using P. acidilactici as a probiotic in fish have demonstrated that this bacteria was able to survive in the intestine of freshwater (Ferguson et al., 2010; Merrifield et al., 2011) and seawater fish (Villamil et al., 2010; Lamari et al., 2013). Remarkably, despite this lower abundance, the strongest modulation of the microbiota in fish fed the probiotic was observed during the seawater stage in the intestinal mucosa. In fact, the probiotic effect on the mucosa-associated microbiota was an important factor driving the differences between dietary groups as shown by beta diversity results. This finding could support the hypothesis of a strong contribution of a probiotic induced “host effect” driving the different clustering of the microbiota composition. However, the significant interaction with water habitat revealed by the PERMANOVA analysis from UniFrac suggests that the probiotic effect is dependent on the water habitat. The most striking changes detected in the probiotic group at the mucosa level consist in the modulation of the relative abundance (↑ increase or ↓ decrease) of taxa belonging to class Fusobacteriia (↓), orders Clostridiales (↓), Actinomycetales (↓), Pasteurellales (↓), and family Bradyrhizobiaceae (↑) and Streptococcacea (↓). Interestingly, apart from Bradyrhizobiaceae that were more predominant in the mucosa, the microbiota of probiotic fed fish seems to display an “antagonistic like pattern” among several taxa. Some of these taxa (Streptococcacea and Pasteurellales for instance) are known as pathogenic or opportunistic bacteria in fish. Therefore, a reduction in their abundance at the mucosa level can be seen as a positive sign, but the absence of information about the other taxa affected by the probiotic (Actinomycetales and Fusobacteriacea in particular) does not allow us to draw further conclusions. Information about the function of these bacteria in the intestine and the role in the adaptation of Atlantic salmon to seawater require further investigation. Finally, most of these changes were only observed in the mucosa, highlighting the importance to investigate such samples when assessing the effect of probiotics in fish.

Previous studies in salmonids have implied the presence of a so called core intestinal microbiota based on next generation sequencing of digesta or mucosa samples (Wong et al., 2013; Gajardo et al., 2016, 2017; Dehler et al., 2017a; Rudi et al., 2018). In the present study, we investigated the core microbiota in the intestinal mucosa and despite the large effect of probiotic administration and SWT, a “core like” mucosal microbiota with 11 shared OTUs was identified across all samples regardless of the water habitat and diet. This set of “resilient” microorganisms account for at least 24% of the relative abundance of the bacterial sequences. This suggests that the OTUs identified as members of this core microbiota may play an important role in the DI of Atlantic salmon, for instance by fulfilling central functions in the gut. The relatively high number of OTUs belonging to the phylum Proteobacteria, as part of the core microbiota of Atlantic salmon is in agreement with previous studies (Gajardo et al., 2016, 2017; Dehler et al., 2017b), whereas the absence of Tenericutes and Bacteroidetes or presence below detection thresholds must be noted. Seawater transfer had a significant effect on the core microbiota member Lactobacillus which was significantly lower in the control group in SW. It is important to highlight that Lactobacillus has previously been identified as a member of the core microbiota in SWT-related studies (Dehler et al., 2017b; Rudi et al., 2018), this finding deserves further investigation to clarify the role of this well-adapted LAB in the intestine of Atlantic salmon. Fish fed the probiotic diet had the strongest effect on the abundance of different taxa of the core microbiota, particularly in the genus Bradyrhizobium which had a relatively low abundance in control and probiotic group during FW 1.73 ± 1.69 and 1.93 ± 1.91% respectively but was significantly more abundant in the probiotic group in SW (36.89 ± 12.76%) compared to the control group (1.38 ± 2.04%).

Probiotic administration influenced the expression of some of the genes investigated in this study, however, the type of response tended to differ according to the water habitat. Activation of tnf-α and il-1b in the intestine are commonly associated with stimulation of the immune response. In the present study, fish fed the probiotic diet had a significantly higher response in il-1b and tnf-α compared to fish fed the control diet. However, this response was only observed in FW. The activation of genes encoding pro-inflammatory cytokines in fish after supplementation with P. acidilactici has also been reported previously in Atlantic salmon (Abid et al., 2013; Vasanth et al., 2015) and tilapia (Standen et al., 2013). These authors suggested that activation of pro-inflammatory cytokines after supplementation with P. acidilactici may indicate a potential immuno-stimulatory response that could be beneficial to fight an eventual pathogen aggression.

Previous studies have investigated the expression of hsp70 and pcna as markers for intestinal stress and cell proliferation in Atlantic salmon after adverse intestinal conditions (Olsvik et al., 2007; Sanden and Olsvik, 2009; Krogdahl et al., 2015). In this study, the low differences in the expression of both hsp70 and pcna between the control and probiotic groups, together with a normal histological morphology of the intestine suggest that neither fish fed the control diet nor fish fed the probiotic diet were undergoing an inflammatory response in the intestine.

A decrease in the expression of several immune genes involved in the antiviral response has previously been reported in Atlantic salmon during the transition from freshwater to seawater (Johansson et al., 2016; Karlsen et al., 2018). In this study, the dietary supplementation with P. acidilactici MA18/5M modulated the intestinal antiviral response. This modulation was dependent on the water habitat as both investigated genes, namely tlr3 and mx1, were significantly lower compared to the control the day before SWT, whereas a significant increase was observed after 5 weeks in seawater. This response to the probiotic has already been observed in SW (Abid et al., 2013) but not in FW. This may suggest that mechanisms associated with the water habitat are influencing the antiviral response and the response of the fish to the probiotic. One possible explanation for the different response of the intestine to probiotic supplementation could be related to the profound changes in the microbiota composition that take place during SWT. It is well-known that the intestine plays a major role in the adaptation to the new marine habitat. It is important to highlight that the strength of the modulation of mx1 was higher during the seawater stage (five times increased compared to the control group) than in the freshwater stage (2.2 times decreased compared to the control group). Studies have demonstrated that mx1 is expressed at high levels after the stimulation of tlr3 agonist in head kidney leukocytes suggesting that these two genes are connected during the antiviral response (Arnemo et al., 2014). These results are in agreement with Abid et al. (2013) who demonstrated up-regulation of tlr3, tnf-α, and mx-1 in the distal and proximal intestine of Atlantic salmon in SW stage (post-smolts) under a dietary regimen supplemented with P. acidilactici MA18/5M and short-chain fructooligosaccharides (scFOS) as synbiotic additives. The main function of tlr3 in innate immunity is to act as a sensor of viral RNA. Moreover, some authors suggest that the activation of tlr3 could trigger the expression of genes encoding cytokines and proteins responsible for modulating the immune response against viral infections (Rodriguez et al., 2005). Even though, tlr3 expression in healthy Atlantic salmon has been detected in several tissues, the gut and spleen showed the highest level of expression (Arnemo et al., 2014). Due to the important role of these genes in antiviral response, their increase in expression in seawater may suggest a potential protective role of the probiotic in an eventual viral infection. In order to evaluate further the potential antiviral effect of P. acidilactici MA18/5M, an experimental challenge with IPN virus was conducted 3 days after SWT. The average accumulated mortality was lower in the fish previously fed the probiotic diet (during FW phase) compared to fish fed the control diet (22 and 27%, respectively), even though this difference was not statistically significant. The mentioned finding may suggest a potential role of the probiotics in improving anti-viral response. It is worth noting that the fish in the challenge tanks were fed the same control diet throughout the challenge. This means that the “probiotic” fish only had a history of being fed probiotics in FW. The results from the feeding trial part of this study, indicating a larger modulatory effect of the probiotic during SW, suggest that the variable results obtained in the challenge could be due to the lack of probiotics feeding during this experimental stage. Previous studies have looked at the resilience of P. acidilactici following a switch to a non-supplemented feed, showing that the probiotic bacteria is not detectable after 3 days in rainbow trout (Abid, 2014). This means the probiotic bacteria and its effect could have been present at least early in the challenge trial. Interestingly, the onset of mortalities started later for the “probiotic” fed fish relative to the control in both challenge tanks, and the probiotic group from the tank with the lowest cumulative mortality had a 6 day delay in this onset. As no samples for microbiota composition was taken from this part of the trial we cannot assess whether the probiotic was maintained in the gut of these fish longer than in the other tank. Ideally, the number of fish per tank and the number of replicate tanks should have been higher, in order to be able to detect potential differences. Additional future experiments should also allow for the continued feeding of the probiotic supplemented diet during the viral challenge. As such, further studies are necessary to confirm whether the induction of the antiviral response seen in this study as well as in Abid et al. (2013) results in a significant improvement in survival after viral infection.