1 INTRODUCTION

The most rapidly growing animal food sector to support the growing human population is fish aquaculture, which will be expected to grow faster than the sector for the production of food animal-sources (Edwards et al., 2019). There is an increasing demand to expand the production with high stocking densities of fish coupled with high feeding through intensive and semi-intensive aquaculture practices. Despite that, increasing production of fish is often threatened by numerous diseases caused by a variety of pathogens (Lafferty et al., 2015). Furthermore, rigorous cultivation practices of aquatic organisms have increased the frequency and severity of diseases in fish (Bojko et al., 2020). In recirculating aquaculture systems (RAS), rearing fish at unsuitable densities causes water quality deterioration and clustering stress and may thus reduce immune competence and impair growth. Compared to their wild counterparts, farmed fish are more prone to some major stressors such as temperature, water quality, and pathogens (Kvamme et al., 2013). Of all the environmental factors affecting aquaculture exercises, dissolved oxygen (DO) is one of the most important and critical stressors that requiring continuous monitoring in the production system which can cause severe losses if not maintained appropriately.

As a gas, oxygen has a low solubility in water, which is the principal limiting factor of fish metabolism in nature. Furthermore, hypoxic (oxygen depleted) conditions can easily appear in production facilities with the intensification of aquaculture practices (Claireaux et al., 1995). Thus, by default, most aquaculture facilities are equipped with wide range of artificial aeration systems proceeding various configurations that aid in the mixing of atmospheric air, and thus let oxygen into the culture water. Therefore, oxygen diffuses into the water proportionate to its partial pressure (pO₂), which may or may not suffice the growing biomass. Thereby, most advanced aquaculture systems rely on the infusion of pure oxygen into the water to meet their intensive demand that supporting the exceedingly high-densities of organisms, as practiced in most RAS.

Meanwhile, changes in DO may affect the immune function of the fish. Indeed, there is much research demonstrating that low DO conditions in fish may led to change in the adaptive and innate immune responses in fish (Cecchini and Saroglia, 2002). Hypoxic conditions also increased the cumulative mortality due to shortened infection incubation period in yellow-tail jack challenged with Enterococcus seriolicida (Fukuda et al., 1997; Cuesta et al., 2003). Similarly, the percentage of mortality was found to increase significantly in Nile tilapia in an environment of low DO when challenged with Streptococcus agalactiae (Evans et al., 2003). However, the mechanism by which low DO reduces immune function has not been elucidated. Despite DO concentration being very important to fish immunity, the relationships and mechanisms between DO concentration and fish immunity are still unclear.

Although the aquaculture environment of fish is a complex system, the microbial communities have been observed to be associated significantly with various diseases. Currently, there is growing evidence on the associations between variations in the host health and the microbiota, as well as other neurological and physiological states. The nutrient rich environment of high-density aquaculture is advantageous for the proliferation of complex microbial symbiotic community. In microenvironments of fish mucosa, a complex and symbiotic community exists (Round and Mazmanian, 2010; Nutsch and Hsieh, 2012) and researchers have found that the interchanges between immune T-cell subsets can be affected by the environment and microbial community in the fish intestine (Omenetti and Pizarro, 2015). According to Kanther et al. (2011), germ-free transgenic zebrafish colonization with commensal microbiota induces transcriptional activation of nuclear factor 'kappa-light-chain-enhancer' of activated B-cells (NF-B) in dynamic spatial and temporal patterns. Galindo-Villegas et al. (2012) determined that commensal colonization primed neutrophils in newly hatched zebrafish and induced the expression of various genes encoding antiviral and proinflammatory mediators, increasing the resistance of fish larvae against infection by viruses. However, research on the composition of microbiota in the gut and fish immunity is still limited.

Atlantic salmon or Salmo salar is a major aquaculture species that is cultured commercially across the globe by different production systems, including sea cages (Norway and Chile), self-contained floating platforms (North Canada), and land-based RAS (Denmark) (Yang et al., 2018). In 2010, Atlantic salmon mariculture has been successfully introduced in China using RAS techniques because of being high economic and nutritional value. However, RAS-reared Atlantic salmons experience higher rates of illness due to bacterial infections when relative to fish grown in sea cages. Therefore, this indicated that Atlantic salmon reared in RAS might with lower immune capacity than salmon reared in sea cage. In this study, we analyzed the gastrointestinal microbial community of Atlantic salmon cultured under different DO concentrations to examine the relationship between the immune health and DO level in this species. The results obtained here will support to determine the optimal DO saturation levels required for intensive mariculture of S. salar in RAS and to devise an efficient rearing condition and also to disseminate their mariculture in China.

2 MATERIAL AND METHOD 2.1 Conditions for recirculating aquaculture systems

All the experiments were conducted in fifteen 2 000-L circular tanks (2-m diameter and height of 1.1 m) (divided into three groups), provided with recirculated filtered natural seawater. Each day, 1 200 L of water was replaced with an equal volume of new water in the tanks at a flow rate of 2 500 L/h. The rate of water exchange was thus 10% of the system volume per day, with 10 days of retention time. Water temperature in the tanks was maintained at 16.0±0.7 ℃ and salinity at 28.0±0.5, with all the values measured at the tank inlets. Illumination was continuous through one fluorescent tube light in each tank at a surface light intensity of 300±50 μmol/(m²·s). Each tank was fitted with a high purity liquid oxygen supply system and oxygen levels were kept stable by a pressure regulating valve.

The experimental Atlantic salmon fish procured from the Shandong Oriental Ocean Sci-Tech Co., Ltd. (Shandong, China) were stocked at an initial density of 14.87±0.56 kg/m³ and final 38±4 kg/m³ fish/tank (about 1.5 kg/fish). Commercial dry pellets containing 22% crude lipid and 42% crude protein (Han Ye Science & Technology Co., Ltd., Beijing, China) were fed thrice at 06:00, 14:00, and 22:00 daily during the acclimatization and experimentation periods. The RAS conditions were described in our previous paper (Sun et al., 2016; Wang et al., 2019). After monitoring the uneaten feeds and consumption in the tanks, the daily feeding rate was adjusted. Fish were acclimatized for 10 days to the RAS conditions, prior to experiment. After acclimated, about 349 fish were randomly assigned to three groups and exposed to the desired DO saturation levels: H: 200% DO; CK: 100% DO; and L: 60% DO. Treatment-specific DO saturation levels were gradually achieved in three days starting from 100% DO to reach the desired saturation levels. This was deemed necessary to prevent the experimental fish from being exposed to acute oxidative stress. Experimental trials were conducted for 60 days. The fish were fasted for one day before sample collection. Moreover, all experiments on live animals were as per guidelines approved internationally for the use of animals in research (National Research Council, 1996).

2.2 Atlantic salmon culture and intestinal tissue collection

At the end of six weeks of Atlantic salmon cultivation in three different DO saturation levels, 15 fishes from each group were collected randomly from the tanks and euthanized by 300-mg/L tricaine methanesulfonate (MS 222). After euthanization, their individual body weight and body length were measured. Subsequently, the intestinal tissue samples were aseptically rushed by phosphate-buffered saline (PBS) and transferred immediately into a chilled 2-mL centrifuge tube and stored at -80 ℃, until DNA extraction. Meanwhile, 4-mL blood of the caudal vein of the sampling fish and their plasma were also kept at -80 ℃.

2.3 DNA extraction and polymerase chain reaction (PCR)

Extraction of DNA from microbes was performed using the HiPure Soil DNA Kits (or HiPure Stool DNA Kits; Magen, Guangzhou, China) as per provided protocols. From the ribosomal RNA gene, 16S rDNA target (Table 1) region was amplified by PCR (94 ℃ for 2 min, then 30 cycles for 10 s at 98 ℃, 30 s at 62 ℃ (except for 16S V4: 55 ℃ for 30 s), and 30 s at 68 ℃, and a final extension for 5 min at 68 ℃) using primers listed in Table 1. PCRs were carried out in triplicate in a 50-μL volume using 100 ng of template DNA, 5 μL of dNTPs (2 mmol/L), 5 μL of 10× KOD Buffer, 1.5 μL of each primer (10 μmol/L), 3 μL of 25-mmol/L MgSO₄, and 1 μL of KOD Polymerase. Related PCR reagents were procured from TOYOBO (Japan). Other primers are listed in Table 1.

2.4 16S rRNA gene pyrosequencing and data processing

Adapters or low-quality reads in raw data affect the assembly and subsequent analysis. Thus, for clean, high quality reads, raw reads were filtered further as per the rules of FASTP (version 0.18.0) (Chen et al., 2018):

1) Removal of reads with > 10% of unknown nucleotides (N);

2) Removal of reads with < 50% of bases having quality (Q-value) > 20.

2.4.1 Assembly of the reads

FLASH (version 1.2.11) was used to merge paired end clean reads as raw tags (Magoč and Salzberg, 2011) with at least a 10-bp overlap and 2% mismatch error rates.

2.4.2 Filtering of raw tags

QIIME pipeline v 1.9.1 (Caporaso et al., 2010) was used to filter noisy sequences of raw tags under specified conditions (Bokulich et al., 2013) to acquire the clean tags of high quality. The conditions for filtering were:

1) From the first low quality base site, raw tags were broken, when the base count in the continuous low-quality value (default quality threshold 3) matched the set length (default length=3),

2) Then, those tags with continuous high-quality base length were filtered having < 75% of the tag length.

For pyrosequencing, 454 GS FLX+platform (Roche, Basel, Switzerland) was used at Guangzhou Gene de novo Biopharm Technology (China) as per provided instructions. Then, data for standard flow gram file were obtained and Mothur 1.31.1 was used to extract sequence information. Low quality sequences were removed from the raw sequences by processing: (1) allowing two mismatches to the sequencing primers and one mismatch to the barcode; (2) eliminating sequences with any ambiguous base-length > 8 homopolymers and < 300 bp; (3) eliminating sequences with the average quality score within any 50-bp windows of < 25. Then, using Chop, sequences were shortened from 520 bp. Some sequences harboring the reverse primer were removed using the Seqs command. Clustering of fragments, alignment of sequences, and pre-cluster analysis were conducted as per Mothur 1.31.1 standard operation procedure. Non-bacterial and chimera sequences were detected and eliminated using, respectively, the classify.seqs and uchime commands, implemented in Mothur 1.31.1. Classification of operational taxonomic units (OTUs) were done at a dissimilarity level of 0.03. After sampling 4 000 sequences, rarefaction analysis was performed. The raw reads of bacterial communities in Atlantic salmon intestine were submitted to the sequence Read Archive of the National Centre for Biotechnology Information (https://www.ncbi.nlm.nih.gov/sra/PRJNA656663).

2.5 Analysis of OTUs

Clustering of effective tags into OTUs of ≥ 97% similarity was performed using UPARSE pipeline v 9.2.64 (Edgar, 2013). Then, within each cluster, tag sequence having the highest abundance was chosen as the representative sequence. Venn analysis between groups was carried out in R project VennDiagram package v 1.6.16 (Chen and Boutros, 2011) and to identify unique and common OTUs, upset plot was performed in R project UpSetR package v 1.1.3 (Conway et al., 2017).

2.6 Analysis of community composition

Classification of representative sequences into organisms was done using a naive Bayesian model and RDP classifier v 2.2 (Wang et al., 2007) based on SILVA database (version 132) with 0.8 as the confidence threshold value. For species abundance, heatmap was plotted using pheatmap package v 1.0.12 in R project (Kolde and Kolde, 2015), and the map was built using the first 18 phyla. Correlation coefficient networks were generated using an online, dynamic interactive real-time platform for data analysis Omicsmart (http://www.omicsmart.com), or igraph package v 1.1.2 in R project.

2.7 Analysis of alpha diversity

Simpson, Chao1, and other indexes for alpha diversity were calculated in QIIME v 1.9.1. Plotting of rank abundance curves and rarefaction curves for OTU were done in R project ggplot2 package v 2.2.1. Calculation of between groups Alpha index comparison was done by Welch's t-test and Wilcoxon rank test in R project Vegan package v 2.5.3. Computation of among groups comparison for Alpha index was done by Tukey's HSD test as well as Kruskal-Wallis H test in R project Vegan package v 2.5.3.

2.8 Analysis of Beta diversity

Alignment of sequences was done using Muscle v 3.8.31 (Edgar, 2004), then unweighted and weighted unifrac distance matrix were obtained using GuniFrac package v 1.0 in R project. Jaccard and Bray-Curtis distance matrix were estimated in R project Vegan package v 2.5.3. Multivariate statistical techniques such as PCoA (principal coordinates analysis) of (Un) weighted Jaccard, unifrac, and Bray-Curtis distances were obtained in R project Vegan package v 2.5.3 and plotted in R project ggplot2 package v 2.2.1. Statistical analysis of Wilcoxon rank test, Welch's t-test, Kruskal-Wallis H test, Tukey's HSD test, Anosim test, and Adonis (also called Permanova) was performed in R project Vegan package v 2.5.3.

2.9 Environmental factor analysis

Canonical correspondence analysis (CCA), redundancy analysis (RDA), mantel test, envfit test, and variation partition analysis (VPA) were carried out in R project Vegan package v 2.5.3 to assess the influence of environmental factors on community composition. Pearson correlation coefficient between species and environmental factors was calculated in R project psych package v 1.8.4 (Hornik, 2012). Network and heatmap of correlation coefficient were obtained using Omicsmart (http://www.omicsmart.com), an online dynamic and interactive real-time platform for analysis of data.

2.10 Western blot

Whole blood samples were used to isolate protein and separated by sodium dodecyl sulfate (SDS) polyacrylamide gel (10%) electrophoresis (Solarbio, Beijing, China), and then transferred by electroblotting to membranes of polyvinylidene difluoride (Millipore, Burlington, MA, USA). These membranes were blocked in PBS with 0.1% Tween (PBST) containing bovine serum albumin (10%) at 25 ℃ for 2 h. Then, incubation of membranes was done for 2 h in tris-buffered saline with 0.1% Tween and specific primary antibodies, such as rabbit anti-interleukin-1β (IL-1β) (1꞉1 000), rabbit anti-nucleotide-binding oligomerization domain like protein 2 (NOD2) (1꞉1 000), and rabbit anti-toll-like receptors 4 (TLR4) (1꞉800). After three washes (10 min each) in TBST, the membranes were incubated for 1 h at 37 ℃ with goat anti-rabbit IgG conjugated with horseradish peroxidase (HRP) and diluted to 1꞉800 (Beyotime, Beijing, China). The membranes were then washed thrice in TBST and processed for detection of proteins using an enhanced chemiluminescence system. The loading control was actin antibody. All experiments were conducted thrice. Protein expressions were estimated as the average of per band area densities per group.

2.11 Statistical analysis

Oxygen was measured daily throughout the experiment, while pH was measured thrice a week. Data were analyzed using analysis of variance (one-way). Data are presented as mean±standard error of the mean (SEM).

3 RESULT 3.1 Water quality

Dissolved oxygen saturation levels were stable at each tank and in agreement with the experimental design. Carbon dioxide levels in each tank showed a positive correlation with DO saturation, and significant difference among treatments in terms of pH (Table 2).

3.2 Microbial biomass and community structure

Although the intestinal bacterial communities of a wide range of fish have been studied, little is known on the relationship between these communities and DO levels. In the present study, 807 888 high-quality sequences with 300−490 bp were retrieved for 15 Atlantic salmon intestinal samples. The OTUs at the 97% similarity level totalized 530–2 201 across all treatments. Alpha diversity index values were 5.9, 6.2, and 6.9 in L, H, and CK groups, respectively, suggesting that bacterial diversity was not high and not severely affected by DO levels (Fig. 1).

Most Atlantic salmon intestinal bacterial community pyrosequence reads were assigned to the Proteobacteria (45%), Firmicutes (20%), and Bacteroidetes (14%) (Fig. 2). These phyla dominated the intestinal bacterial community across all DO treatments but there was a variation in relative abundances and lineages of bacteria at the genus level.

A Venn diagram was prepared to identify the Atlantic salmon core intestinal bacteria under the different DO treatments. The 1 158 OTUs shared among all Atlantic salmon intestinal samples accounted for 48.9%, 55.6%, and 48.7% in the L, H, and CK groups, respectively (Fig. 3). The H and CK groups shared another 246 OTUs (11.8% and 10.3% of the sequences in the H and CK groups, respectively) while the L and CK groups shared an additional 239 OTUs (10.1% and 10.1% of the sequences in the L and CK groups, respectively); groups H and L shared 193 OTUs (9.3% and 8.1% of the sequences in the H and L groups, respectively). A further 484 OTUs (23.3% of the sequences in the H group), 779 OTUs (32.9% of the sequences in the L group), and 734 OTUs (30.9% of the sequences in the CK group) were distinct to the H, L, and CK groups, respectively. These findings suggest that Atlantic salmon exposure to various DO levels caused the establishment of distinct populations of microbiota.

The principal coordinate analysis (PCoA) performed to discriminate the bacterial community compositions of Atlantic salmon reared at different DO levels based on OTUs distributions revealed clusters based on the DO level of the culturing water (Fig. 4). Bacterial communities were clustered into three groups along the first PCoA axis, which accounted for 40.49% of the total variation; the second PCoA axis, only 9.57% of the total variation in bacterial communities was explained.

3.3 Relationships between Atlantic salmon intestinal bacterial communities and DO levels

Although the different DO levels had no significant effects at the level of family, some differences were found at the bacterial level of genus. In the group L, Acinetobacter baumannii was the most represented OTU (42%) while it only accounted for 29% and 10% of the bacteria in H and CK groups, respectively. No other bacterial genus showed significant differences across treatments (Supplementary Fig.S1). Thus, the different DO levels tested here might have no significant effects on bacterial communities, except for A. baumannii, which might be the OTU most related to DO challenges.

The RDA used to investigate the association between the bacterial community in the intestine of Atlantic salmon and environmental factors including pH, DO, NO₂, and chemical oxygen demand (COD) revealed that DO, NO₂, and pH significantly contributed to the bacterial community-environment relationship (Fig. 5). Acinetobacter, Sphingomonas, and Klebsiella correlated positively with DO, Aliivibrio, Planctomyces, and Blastopirellula were positively correlated with pH, and Stheptomyles, Bacteroides, and Ralstonia were positively correlated with NO₂ (Fig. 5). These results demonstrated that DO and pH might be the major determinants shaping Atlantic salmon intestinal bacterial communities.

3.4 Haematological and biochemical immunity responses to DO levels

To determine whether DO level could influence the immunity of Atlantic salmon reared in RAS, serum biochemical parameters were assessed. Firstly, the leucocyte number was detected, the results showed that leucocyte number was higher at significant levels in fish of the L group relative to the fish of the CK and H groups, ranging from 70.10×10⁹ to 155.71×10⁹ cells /mL(Fig. 6a). Then the plasma cortisol were detected in this experiment. The results showed that plasma cortisol of the L group was higher at significant levels compared with that of the other two groups ranging from 14.32 ng/mL to 36.03 ng/mL. However, when the fish supplied with high concentration of DO have no significantly influence to immunity response (Fig. 6b).

3.5 Immune-related gene expression analysis

Acinetobacter baumannii was the main species affected by the different DO levels in Atlantic salmon reared in RAS. However, A. baumannii is an emerging opportunistic pathogen, which can induce innate immunity response by the recruitment of neutrophils and macrophages (Van Faassen et al., 2007). Here, the expressions of TLR4, NOD2, and IL-1β were assessed by western blotting. The results showed that low DO levels (60%) (L group) could induce inflammatory response in Atlantic salmon. The IL-1β were increased significantly in Atlantic salmon serum. Proteins related to the A. baumannii-induced pathway, such as NOD2 and TLR4, were also increased significantly in the L group. The expressions of NOD2 and TLR4 in the H and CK groups were less than that in the L group (Fig. 7).

4 DISCUSSION

Although animals in aquatic environment are in continuous contact with the surrounding water, their intestinal microbiota composition may differ from microbial community composition in the water in which they thrive (Meziti et al., 2012). DO is an important environmental factor for fish growth and health. Welker et al. (2007) reported that bactericidal activity, hemolytic complement, and antibody response were lower in channel catfish in the presence of low DO and induced immunosuppressive effects were also detected. Evans et al. (2003) reported a significant increase in mortality rates in Nile tilapia exposed to S. agalactiae following in the presence of low DO. Further studies have also shown that low DO can modulate both innate and adaptive immune responses in fish (Cecchini and Saroglia, 2002; Cuesta et al., 2003). However, the mechanism by which low DO reduces fish immunity is still unclear.

Based on human studies, microbiota is known to regulate nearly each aspect of the physiology of the host, such as the immune response. According to the seminal study of response of zebrafish intestinal microbiota (Kelly and Salinas, 2017), most of the microbial mechanisms controlling the immune system in mammals may be conserved in teleost (Rawls et al., 2006). Sound information on microbiota structure, composition, and stability of fish reared in RAS will provide knowledge for understanding the association between microbiota, environment, and host. Bacterial community optimization would also facilitate optimal water quality and the potential for fish immune modulation in RAS. Thus, in this study, the effects of different levels of DO were assessed on the intestinal microbial community and immune status of Atlantic salmon reared in RAS.

Firstly, our findings indicated that DO levels exert significantly influence on intestinal bacterial composition of Atlantic salmon. The functions of microbial consortia in the ecosystem are wide and important for aquatic animals' growth and disease suppression. In the present study, distinct intestinal microbiota structures were observed in Atlantic salmon reared in RAS under different DO levels. No significant differences in bacterial phyla diversity and frequency distribution were found between the different treatment groups. In all groups, Bacteroidetes, Proteobacteria, Actinobacteria, Firmicutes, and Verrucomicrobia were the five dominant OTUs. However, some differences in bacterial species were found between treatments.

Research on the association between DO levels and bacterial communities of fish intestine is limited. Based on the present study, Atlantic salmon might adapt to changes in DO levels, if these are sufficient or high, as no significant effects were detected on intestinal bacterial communities indicating fish would require more DO. When the DO level is low, fish might decrease their consumption of oxygen by for example reducing movement and metabolic rate. To detect whether DO is a key participant in modulating the intestinal bacterial community of Atlantic salmon, an RDA was conducted and it revealed significantly different abundances for some bacterial species. The most abundant OTUs were associated with A. baumannii in the intestine of RAS-reared Atlantic salmon, especially in those belonging to the low DO (L) group. Thus, A. baumannii might play an important role in reducing fish immunity under low DO levels.

The Gram-negative coccobacillus, A. baumannii is ubiquitous and has the capacity to thrive for long periods in the environment, such as water. A. baumannii infection in humans can cause bacteraemia, pneumonia, skin, bloodstream, meningitis, and infection in the urinary tract (Cerqueira and Peleg, 2011). In host defence against bacterial infections, innate immune responses have key roles. Inflammatory response is characterized by the recruitment of neutrophils and macrophages to the site of the infection (Qiu et al., 2012). In this study, we firstly detected the serum biochemical parameters related to immunity and found that leucocytes and plasma cortisol were increased in the low DO group. The IL-1β was also increased in plasma. These findings indicated that low DO level might induce inflammatory response, probably by increasing the abundance of A. baumannii. Some researchers have found that many signaling and innate immune components are activated in response to A. baumannii infection (Pires and Parker, 2019). Immune stimulation may therefore play a major role in RAS-grown Atlantic salmon to control disease.

In the present study, the expressions of some potential immune-related proteins were assessed. Because it responds to the lipopolysaccharides on the outer membrane of Gram-negative bacteria, TLR4 has been extensively studied regarding Gram-negative bacterial infections (Lu et al., 2008). In addition, TLR4 is an important molecule for hosts challenged with A. baumannii as its deletion in mice resulted in enhanced bacterial burden and dissemination compared to those in wild-type mice (Knapp et al., 2006). In this study, TLR4 increased significantly in the low DO group, indicating that low DO levels might induce immune response, mostly owing to the increase of A. baumannii in the intestine.

To confirm the pathway of A. baumannii-induced inflammation, NOD2 expression was also assessed in our study. This protein has been shown to be important at the early stage of infection, as an increased bacterial burden appeared at 4- and 12-h post-infection in Nod^-/- mice in comparison to those in wild type mice (Kale et al., 2017). Knocking down NOD2 in mice also increased the level of neutrophil infiltration and IL-1β (Wu et al., 2003). In the present study, the increased IL-1β levels might have been caused by lacking of NOD2 expression, as this is a key factor in A. baumannii infection. In conclusion, RAS is a highly-efficient fish farming technique, useful for producing high-value species near niche markets outside their natural environment. Low DO levels may lead to immune response in Atlantic salmon reared in RAS, and this immune response might be caused by the high abundance of A. baumannii in the intestine of fish subject to such levels. High abundance of A. baumannii might induce immune response via the TLR4/NOD2 signaling pathway. In this paper, we found the main microflora in Atlantic salmon reared in RAS, and A. baumannii might induce a series of immune response in Atlantic salmon. However, RAS is a closed system, the microflora is easily influenced by a lot of factors such as environment, biofilter, and management et al. So, it might need more work to get more details to this research.

5 CONCLUSION

In this paper, we found that lower DO level might induce stronger immune response in Atlantic salmon. In addition, we found that the abundance of A. baumannii in Atlantic salmon intestine might be involved in the intensity of immune. Our results showed that low DO level could increase the abundance of A. baumannii. The classification and distribution of intestinal bacterial communities vary different significantly. Furthermore, the white blood cell, the level of cortisol, the expression of IL-1β, TLR4, and NOD2 in low DO group are significantly higher than these in other two groups. In brief, we analyzed the effect of DO on the intestinal flora of Atlantic salmon and investigated the mechanism of effect of DO on immune response.

6 DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Electronic supplementary material

Supplementary material (Supplementary Fig.S1) is available in the online version of this article at https://doi.org/10.1007/s00343-021-1336-y.