In the present study, saponins negatively affected the intestine and caused lower growth performance, but only in combination with PPC. Histology of the distal intestine revealed enteritis resembling the changes associated with soy enteropathy [12, 13], which was also confirmed with pronounced transcriptome changes. Moreover, microarray results were consistent with our previous report on soya induced gene expression changes in the distal intestine of Atlantic salmon . Similar effects of SBM on expression levels of genes related to lipid, iron and xenobiotic metabolism have also been observed in the distal intestine of Atlantic halibut . Interestingly, inflammation was not induced in the latter case, suggesting that suppression of some metabolic processes observed in the present work could occur independent of the activation of immune responses.
In previous studies, feeding 18% field peas  or 20% PPC  did not produce histomorphological changes, while high dietary PPC levels (35%) produced enteritis and adverse effects on growth performance and nutrient digestibility . Peas have been reported to contain between 0.7 and 2.5 g kg−1 saponin [26–28]. However, the PPC used in the current work, produced by air classification, may have contained higher levels compared to unprocessed peas due to changes during processing [29–32]. The difference in distal intestinal morphology in fish fed the PPC + S diet could therefore be due to a dosage effect of saponins as suggested previously . None of the other protein sources were expected to contain appreciable amounts of saponins [6, 33–36]. Therefore, the basal saponin level in the different diets could explain the differences in the results. However, the level of saponin supplementation was equivalent to the amount found in a diet containing approximately 40% SBM, which consistently causes distal intestine inflammation. Moreover, similar histology was reported in fish fed either a fishmeal based control diet or the same diet supplemented with 2.6 g kg−1 of a 65% soyasaponin concentrate , an amount equivalent to a diet containing approximately 30% defatted SBM, a level which also consistently produces distal intestine inflammation. This indicates that some other component or contributing factor(s) is (are) necessary to induce an inflammatory response. The necessary factor may involve interactions with other ANFs present in peas, such as protease inhibitors, phytic acid, oligosaccharides, lectins, tannins, and/or dietary antigens [37, 38], or changes in the intestinal microbiota .
The inflammatory response was characterized by marked involvement of genes regulating T-cell functions, in line with the high T-cell reactivity seen during the development of soy enteropathy . Increase in the expression of the T-cell receptor (TCR, 5.8-fold), responsible for the primary costimulatory signal for T-cell activation, CD86 (1.9-fold) that provides a secondary positive signal, as well as the binding partner of CD86, T-cell inhibitory CTLA4 (2-fold) , suggested the need to tightly regulate T-cell-mediated processes. IL-18-dependent polarisation of Th responses towards the Th1 and Th17 lineages is consistent with the increased level of IL-22 observed in the chronically inflamed intestine . Preferential expression of pro-inflammatory Th responses can promote a wide range of pathological responses in the intestine, mediated either by T-cells or by excessive innate immune activation .
Activation of TNFalpha dependent responses including induction of NFkB and the respiratory burst are typical for myeloid cells (neutrophils recruited from circulation and / or resident macrophages). This was further supported by the substantial remodelling of ECM, as evidenced by a number of significantly affected genes encoding proteins involved in ECM deposition and degradation. The high destructive power of the effectors explains severe tissue damages. In the healthy intestine, basal immune activation maintains barrier function and commensal microflora composition; however, excessive and uncontrolled inflammation likely represents a central contributor to the pathophysiology of the feed-induced distal enteritis in salmon. Microarray data revealed 11-fold induction of annexin-A1, which together with the induction of phospholipase D2, involved in processing of the annexin-1 receptor (formyl peptide receptor), suggested promotion of the pathway that can result in the inhibition of the transendothelial migration of neutrophils . Suppression of the recruitment of leukocytes into the mucosa by reducing leukocyte-endothelial adhesive interactions could be an attempt to abrogate exaggerated immune responses. However, annexins may also inhibit biosynthesis of eicosanoids and therefore potentially reduce production of prostaglandin E2 (PGE2) that has a crucial role in multiple gastrointestinal defences . The context of immune activation is crucial; although immune effector cells play essential roles in protective immunity against harmful luminal agents, similar effector functions seemed to be engaged during inappropriate inflammatory responses against dietary antigens. Although previous work from our group did not find a protective effect of oxytetracycline against enteritis in salmon , the need to engage in handling of commensal flora that breached damaged mucosal barrier should not be ruled out. Further investigation of the involvement of other, oxytetracycline resistant bacteria in the soy induced inflammatory response is warranted.
Activation of phagocytes such as neutrophils and macrophages was further indicated by the regulation of several components of the respiratory burst complex. These findings are important because neutralization of ROS represents an important defence against self-inflicted damage. Decreased mRNA levels of several glutathione-s-transferases and the key antioxidant enzymes catalase and glutathione peroxidase indicated that animals could be vulnerable to oxidative stress. Similar observations have been made in salmon and rainbow trout hepatic transcriptome after restricted feeding, SBM feeding and infection [22, 46, 47]. The coordinated decrease of genes encoding iron and heme proteins may have also influenced redox status. Excessive amounts of toxic metals may be prevented from entering the body by retention in the gut tissue bound to specific proteins such as metallothionein (MT) and ferritin . Decreased MT and ferritin levels may thus have resulted in increased susceptibility to metal toxicity. Interaction between the regulation of inflammation and biotransformation of toxic compounds has been observed under various conditions [49, 50]. Up-regulation of the NFkB pathway may be necessary to activate protection against cellular stress, indicated by the expression pattern of several heat shock proteins and ubiquitin (up 10-fold). However, NFkB suppresses AhR, which co-ordinates transcription of genes involved in xenobiotic metabolism . The down-regulation of biodegradation in the intestine observed in this study may possibly increase vulnerability of the intestinal tissue and consequent hepatic loading.
The mucosal epithelial barrier of the alimentary tract is continuously exposed to noxious and immunogenic substances, including pathogens, dietary antigens and toxins. Decreased numbers of goblet cells, reduced mucus secretion and abnormalities of its composition are well described in a number of intestinal disorders [52–54]. In the present work, increased mucus production, as indicated by increased number of goblet cells, likely increased barrier properties and provided a degree of protection. In accordance, microarray data suggested augmented production of mucin glycoproteins. Notably AGA, which targets mucin glycoproteins for degradation, was down-regulated while B3GN7, which modifies glycoproteins, was 4-fold up-regulated. Another crucial aspect of the mucosal barrier is cell-cell adherence. Paracellular passage of luminal content is restricted by TJs that seal the most apical space between intestinal epithelial cells. Knudsen and co-workers suggested that soyasaponins might increase transepithelial inflow of dietary antigens and microflora through increased TJ permeability in salmon distal intestine . Although molecular components of the TJ complex and their individual contributions to barrier function within the intestinal epithelium of Atlantic salmon have not been studied extensively, up-regulation of the major protein of mammalian TJs occludin could be interpreted as a sign of extensive junctional reorganization during assembly of new junctions and/or an attempt to increase TJ strength. Multiple TJ components interact with the actin cytoskeleton through binding to PDZ domains found on cytoplasmic adapter proteins, a number of which was induced in the current work. Furthermore, activation of actinomyosin contractility that leads to increased paracellular permeability was suggested by up-regulation of several myosins and RhoGTPase signaling (RhoG and RhoGEF) and suppression of the inhibitory myosin phosphatase-Rho interacting protein (see  for review on factors involved in regulation of TJ functions). The microarray data set was enriched for genes involved in arginine and proline metabolism (Table 1). Polyamines produced from arginine have previously been shown to be essential both in early mucosal restitution by cell migration and in regeneration by proliferation . Intracellular polyamine levels are tightly regulated by the activity of ODC and ARG, which were both induced by saponins. In addition to being responsible for the generation of polyamines, ODC and ARG have been shown to be protective in a mouse model of colitis by competitive inhibition of NO production .
Collectively, gene expression data support the proposal that continuous cell renewal and an increased need to replace lost cells may come at the expense of proper differentiation of intestinal cells. This could partly explain the observed profile of genes involved in digestive processes and is in line with previously made observations that SBM-induced enteritis affects differentiation of epithelial cells in salmon . The general down-regulation of digestive proteases was consistent with our study on SBM inclusion in salmon diets  and the reduced brush border enzyme activities observed in salmon fed plant-based diets such as SBM and PPC [13, 14]. Dys-regulation of proteolytic actions has also been described for inflammatory bowel diseases (IBD), which shows similarities to fish enteropathy . Pancreatic secretory trypsin inhibitors were found to be markedly reduced in the colon of patients with IBD , whereas MMP levels were elevated . In the present study, decreased faecal dry matter in fish fed the PPC + S diet suggested that these fish had diarrhea , similar to what has been reported in salmon with SBM-induced enteritis . In accordance, microarray data indicated disruption of water and solute absorption. The observed down-regulation of aquaporins is in accordance with several mammalian IBD studies [61–63], and the role of Aqp8 (down 14-fold) as a key water channel in the intestinal tract of salmonids . Another interesting finding was increased levels of guanylin, which may decrease intestinal fluid absorption, increase chloride secretion and cause diarrhea in mammalian models . It is known that SBM causes increased permeability of the distal intestinal epithelium , which likely disrupts water and ion balance. Furthermore, the observed decreased expression of nutrient transporters in the current work is in line with decreased carrier-mediated transport after SBM feeding .
As presented elsewhere , apparent digestibility of cysteine decreased when saponins were added to the PPC containing diet, indicating reduction in sulfur containing amino acids. Cysteine is a precursor for taurine, and decreased cysteine uptake together with a marked effect on cysteine deoxygenase 1 (CDO1, 46-fold down-regulated) as well as down-regulation of taurine transporter mRNA levels may have consequences for taurine biosynthesis and subsequently conjugated bile salt levels. Additionally, reduced cysteine uptake may negatively affect synthesis of the highly cysteine-rich MTs . The observed transcriptional effects on MT and taurine metabolism as well as the reduced bile salt levels may therefore have resulted from reduced cysteine uptake when saponins were added to the PPC diet. In general, SBM inclusion in fish feed has been associated with decreased lipid digestibility, reduced bile salt levels and hypocholesterolemia [13, 67–74]. In the present study, similar negative effects of saponins on lipid and fatty acid digestibility were observed (data presented elsewhere ), and microarray analyses revealed dramatic suppression of lipid and steroid metabolism. This could result both from activation of immunity and limited absorption of nutrients as a consequence of dys-regulated cellular differentiation and loss of function, as previously mentioned. Our previous studies with the 1.8 k cDNA microarray revealed similar changes in the liver of salmon infected with ISA virus  and the same tendency was caused with restricted feeding in salmon and rainbow trout [22, 47] and by SBM inclusion in feed for salmon [22, 76] and halibut . It remains unknown whether suppression of lipid metabolism in the intestine could affect the condition of fish or biosynthesis of steroid hormones that require cholesterol. However, down-regulation of cholesterol and sulfate metabolism likely impaired production of bile. In accordance, saponins reduced bile salt concentration in fish fed PPC by 60% in the pyloric intestine and 56% in the mid intestine . Given the key role of bile in lipid digestion and absorption, this could partly explain the observed decrease in lipid digestibility. Additionally, bile salts have several signaling properties regulating metabolic, detoxifying, antibacterial and immunomodulatory actions , which may have been compromised and could account for some of the observed transcriptional responses.