Transcriptomic and physiological responses to fishmeal substitution with plant proteins in formulated feed in farmed Atlantic salmon (Salmo salar)

Background Aquaculture of piscivorous fish is in continual expansion resulting in a global requirement to reduce the dependence on wild caught fish for generation of fishmeal and fish oil. Plant proteins represent a suitable protein alternative to fish meal and are increasingly being used in fish feed. In this study, we examined the transcriptional response of Atlantic salmon (Salmo salar) to a high marine protein (MP) or low fishmeal, higher plant protein replacement diet (PP), formulated to the same nutritional specification within previously determined acceptable maximum levels of individual plant feed materials. Results After 77 days of feeding the fish in both groups doubled in weight, however neither growth performance, feed efficiency, condition factor nor organ indices were significantly different. Assessment of histopathological changes in the heart, intestine or liver did not reveal any negative effects of the PP diet. Transcriptomic analysis was performed in mid intestine, liver and skeletal muscle, using an Atlantic salmon oligonucleotide microarray (Salar_2, Agilent 4x44K). The dietary comparison revealed large alteration in gene expression in all the tissues studied between fish on the two diets. Gene ontology analysis showed, in the mid intestine of fish fed PP, higher expression of genes involved in enteritis, protein and energy metabolism, mitochondrial activity/kinases and transport, and a lower expression of genes involved in cell proliferation and apoptosis compared to fish fed MP. The liver of fish fed PP showed a lower expression of immune response genes but a higher expression of cell proliferation and apoptosis processes that may lead to cell reorganization in this tissue. The skeletal muscle of fish fed PP vs MP was characterized by a suppression of processes including immune response, energy and protein metabolism, cell proliferation and apoptosis which may reflect a more energy efficient tissue. Conclusions The PP diet resulted in significant effects on transcription in all the 3 tissues studied. Despite of these alterations, we demonstrated that high level of plant derived proteins in a salmon diet allowed fish to grow with equal efficiency as those on a high marine protein diet, and with no difference in biometric quality parameters.


Background
Aquaculture production has been the fastest growing animal food-producing sector globally for over half a century, with production growing at an average rate of 8.1% per year since 1961 [1], representing almost 50% of human consumed fish [1,2]. Additionally, the capture from wild fisheries has plateaued with approximately three quarters of wild fisheries fished to capacity, overfished or recovering [3]. Atlantic salmon (Salmo salar) production alone now reaches over 1.5 million tonnes per year [4]. Total production in 2008 is estimated at 1.4 million tonnes, representing a 6% increase on 2007 [5]. The positive growth trend of the industry is expected to continue, reflecting the rising demand for healthy human food products high in protein and marine oils.
Piscivorous fish, which include salmonids, require high protein diets and until recently this was almost exclusively derived from wild caught non food fish such as anchovies and sardines. It has become apparent that with increasing fish meal demand due to the expanding aquaculture industry the proportion of protein in diets cannot be sustained by use of fish meal alone. During 2003, 20% of total fish meal usage for aquaculture feed production was utilized by the salmon industry [6], but continued development of diets has reduced the inclusion of fish meal in salmon diets from 500 g kg -1 to 350 g kg -1 by replacement with alternative protein sources [7]. Plant proteins currently represent the only economic and sustainable protein alternative to fish meal and are increasingly being used in commercial fish feed, with the most common being soybean meal (SBM) or soybean protein concentrate (SPC) which have a high protein content and contain the majority of essential amino acids required by salmonids. All plants contain a number of anti nutritional factors (ANFs) as part of their inherent defence mechanisms, effected by lectins, saponins, phytic acid and proteinase inhibitors amongst others [8]. In early attempts to use plant proteins in fish diets ANFs were often co purified with the proteins, leading to metabolic dysfunction in fish liver [9], inflammation in the intestine [10,11], reduced protein deposition [12], and an impaired immune response [13]. Currently used plant replacement diets use a combination of different plant proteins, with SPC often the major component [14]. Processing of soy protein to SPC with alcohol-extraction removes the majority of ANFs facilitating use of these products in fish feeds at a high inclusion level without causing enteritis or other gross morphological changes [15]. Replacement of up to 50% of the fish meal with a mixture of plant proteins is possible in rainbow trout without affecting fish growth, and even complete replacement of fish meal with SPC has been reported [16,17]. Other successful plant proteins used regularly in aquaculture include wheat gluten at up to 40% of fish meal replacement in feed for salmon and trout (Hardy 1996), corn gluten and sunflower meal at up to 30% replacement, which show no adverse effects on salmonid growth performance [18][19][20]. Extensive studies investigating responses of Atlantic salmon to such diets indicate that a replacement of fish meal close to 100% can be used without negative effects on salmon growth providing the proteins are highly purified and the amino acid profile is well balanced [21]. High levels of pea protein concentrate do still cause enteropathy in salmon though [22].
In this paper we report a whole transcriptome based study on post-smolt Atlantic salmon fed on a low marine protein diet, where marine protein fish meal was partially replaced with a combination of plant derived proteins, compared to fish fed a diet high in fish meal. The high plant protein (PP) diet contained a balanced combination of soy protein concentrate, corn gluten and wheat gluten, and was supplemented with lysine and methionine to ensure the requirement for essential amino acids was met. The fish were fed for 77 days following which growth, feed efficiency and quality biometric parameters were measured and tissues were collected for histopathological changes and gene transcription analysis. The transcriptional response was examined using an Atlantic salmon"salar_2" microarray platform (Agilent) [23], enriched for genes involved in protein metabolism, lipid metabolism and immune function.

Feeding trial
During the 77 day feeding trial fish were fed either a high marine protein (MP) diet or a plant protein replacement (PP) diet (Table 1), and both groups doubled in weight over this period. No significant differences were seen between the groups in growth performance, feeding efficiency, condition factor or organ indexes ( Table 2). There were also no significant mortalities in any of the tanks.

Histology
A number of histological changes were found, however different types of anomalies were observed in tissues of individuals from both treatment groups and such changes were observed in low frequency indicating that the fish were in a good condition (Table 3). There was no significant dietary effect for any histological change except the frequency of perivascular and peri-biliar duct infiltrations in the liver, where 21% of fish fed the MP diet exhibited this change compared to 3.8% of fish fed the PP diet (Table 3). A number of fish from both diets showed a mild fatty change in the liver, but this was not restricted to a particular dietary group (Table 3). Equally, some anomalies in heart were observed at a low level in both dietary treatments. For the intestine, no changes were observed in pyloric caeca or mid intestine, whereas for the distal intestine an inflammatory infiltration in the submucosa and lamina propria was observed but at a low level in both diets, with no significant difference in occurrence between dietary groups (Table 3).

Transcriptomic responses
Fish fed the PP diet showed changes in transcriptome in all 3 tissues examined compared to fish fed the marine protein rich diet MP. In total 8,151 oligo features were found to be expressed at different levels as a result of the dietary manipulation. For subsequent analysis only significant (p < 0.05) features with a fold difference of greater than 2 were analysed further (1961 mRNAs). The intestine exhibited the greatest differential gene expression with 1,236 genes significantly different in expression level between fish fed the PP diet and MP diet. Of these 615 genes were higher in expression and 621 lower in PP fed fish. In skeletal muscle 505 genes had an expression higher than two fold with 132 higher in PP fed fish and 373 lower and in liver 220 genes were found modified with 161 higher and 59 genes lower in PP fed fish vs MP fed fish (Table 4).
There were no genes expressed at different levels in common between all the tissues examined indicating there was not a systematic response to the diets. However there was a co regulation of features between the mid intestine and skeletal muscle with 9 genes higher in both tissues of PP fed fish vs MP fed fish and 10 lower in both. Similarly in the liver and intestine 6 genes were found to be significantly different in common, with 5 genes higher in expression and 1 gene lower in the PP fed fish. To further investigate the biological significance of the differently expressed gene sets we used gene ontology analysis to help identify biological processes that were significantly different in tissues following the feeding trial.
Annotation of the microarray allowed 77% of the features to be allocated to a functional protein and 55% of these proteins were assigned a gene ontology (GO) identifier. This allowed statistical analysis for enrichment for GO biological processes to help interpret the changes in the transcriptome following the feeding trial ( Figure 1) to gain a holistic view of which biological processes were significantly modified following feeding the PP diet.

Genes expressed at different levels in mid intestine
The intestine is the major tissue to come into direct contact with any feed components and for this reason may be very sensitive to dietary changes. This is shown here by the intestine having the greatest transcriptional response in terms of the number of genes expressed at different levels and the magnitude of different expression of these genes. Several biological processes were significantly different in the PP diet fed fish, related to intestinal functions including immune and stress related processes, protein metabolism, energy and mitochondrial activity, lipid metabolism and transport (Table 4).

Immune and stress response
Genes encoding proteins related to both innate and acquired immune function were expressed at different  Growth performance and somatic indexes. SGR represents the specific growth rate of fish, FCR is the feed conversion ratio, SFR is the specific feeding rate, CSI is the cardiac somatic indice, HSI is the hepatic somatic indice and ISI is the intestinal somatic indice. levels by the dietary treatments. For fish fed PP innate immune parameters were found to be both higher and lower in expression indicating a dynamic response. Modulators of the immune response were higher in PP fed fish, such as MyD88, a key transcription factor associated with induction of an inflammatory response and P105 subunit of an inhibitory protein that sequesters NF-κB in the cytoplasm. Another transcription factor, interferon regulatory factor (IRF) 1 was also higher, and potentially increased expression of interferon responsive genes including Mx, Vig-2, interferon induced protein 35 and virus induced TRIM protein. Genes encoding cytokine receptors (IL-17R) and regulators of cytokine function including the suppressor of cytokine signalling (SOCS)-7) were also higher. A number of genes related to innate immune responses were also lower, and included IL-17D, MCSFR and chemokines such as CCL3, CXC13 and IL-8 (CXCL8). Other innate immune serum proteins were also lower in expression including serum amyloid A, and a number of lectins. Genes related to proteins in the acquired immune response were mostly higher in expression in fish fed the PP diet compared to the MP diet. For example, T cell receptor chains (α and γ) and their signalling subunits CD3 epsilon and gamma/delta were higher as were genes involved with antigen presentation such as β2 microglobulin and MHC class II). Related to antibody production by B cells, a B-cell linker protein that regulates B-cell function and development was higher in PP fed fish as was a B cell enhancing factor. Regarding stress related genes, PP fed fish showed an elevated expression of glutathione-S-transferase and a thioredoxin interacting protein, involved in reducing oxidative stress, while heat shock proteins (HSP) β-7, β-11, 30 and 70a were all suppressed.

Cell proliferation and apoptosis
The intestinal transcription of genes encoding proteins related to cell proliferation (tumor necrosis alpha-induced protein 2, tetraspanins-14 and −16 and melanoma-derived growth regulatory protein) and the cell cycle (DNA topoisomerase 2-alpha, replication factor C subunit 3, FGFR3 protein, cyclin A1) were found to be higher in PP fed fish; whilst transcription of caspases 3, 8 and 14,which are directly involved in apoptosis, were found to be lower in expression relative to fish fed MP.

Protein metabolism
Feeding the PP diet the expression of genes involved in protein metabolism was significantly different compared to MP. For example, protein synthesis related genes including elongation factor, translation initiation factors 2 and 4e and ribosomal protein S6 kinase b were higher, while a number of 60 S ribosomal protein encoding genes were lower relative to MP fed fish. Protein degradation encoding genes were more highly expressed as shown by lysosomal proteases, cathepsins A, C, and Z, ubiquitin proteasome route (UbP) of proteolysis, and proteasome subunits α5, α6, α7 and β7. Genes involved in targeting of proteins for degradation in the UbP were also higher including ubiquitin conjugating enzyme E2 and ubiquitin E3 ligases. Genes encoding the structural proteins troponin, tubulin, titin, actin and keratin were all found to have a higher expression level whereas a number of myosin mRNAs were lower in expression.

Lipid metabolism
A higher transcript level of genes involved in lipid transport and metabolism was observed, with higher transcription of fatty acid-binding protein, apolipopro-tein A-I, apolipoprotein B and delta-6 fatty acyl desaturase, delta-5 fatty acyl desaturase, PUFA elongase 5A and glycolipid transfer protein, respectively.

Energy and mitochondrial activity/kinases
There was an induction of many genes related to energy metabolism in pathways such as glycolysis (glyceraldehyde-3 phosphate dehydrogenase), the tricarboxylic acid cycle (succinate dehydrogenase) and gluconeogenesis (fructose 1-6-bisphosphate) which were expressed at a higher level in PP fed fish relative to the MP fed fish.

Cellular transport
Transport of solutes into and from a cell is an energy demanding process. Genes involved in membrane transport were higher in expression in fish fed the PP diet, and included a number of solute carrier family members such as SCF22 (organic cation transporter), SCF 27 (fatty acid transporter), SCF 25 (mitochondrialphosphate carrier), SCF 39 (zinc transporter), SCF 16 (monocarboxylic acid transporter), and SCF 31 (copper transporter). Two genes encoding haemoglobin subunits (hemoglobin subunit alpha-4 and hemoglobin subunit beta-1) and beta  globin, however, were found to have a lower relative expression level.

Genes expressed at different levels in liver
Several biological processes were significantly different in liver that can be related to immune parameters, stress responses, and protein and lipid metabolism, with key genes shown in Table 5.

Immune and stress response
Genes encoding proteins involved in innate immunity such as complement c1q, C-type MBL-2 and a dendritic cell specific lectin CD209f which binds mannose carbohydrate molecules, and hepcidin a major liver associated antimicrobial peptide were lower in PP fed fish. A limited number of genes related to adaptive immunity were also found to be expressed at different levels, with MHC class I and SMAD 3 lower in expression level, whereas MyD88 was higher in fish fed the PP diet relative to MP fed fish. Moreover, genes involved in the oxidative response (MPV17 protein, amine oxidase and heat shock cognate 70 kDa protein) were also higher in fish fed the PP diet.

Cell proliferation and apoptosis
The PP diet stimulated/modified hepatic cell turnover as indicated by higher expression of serine protease HTRA1 and annexin A3 and the stimulation of apoptotic processes through genes involved in TGF beta pathways (cell death activator CIDE-3 and angiopoietin-related protein 4).

Protein metabolism
Hepatic protein metabolism was clearly stimulated in PP fed fish since all genes significantly different in expression were higher in these fish, including ribosomal protein S18a, ribosomal protein l39 and peptidyl-prolyl cis-trans isomerise. Similarly genes involved in protein degradation were more highly expressed as seen with the non-lysosomal protein calpain and the proteasome subunit β3.

Genes expressed at different levels in skeletal muscle
The GO analysis showed a number of processes to be significantly different in skeletal muscle that can be related to protein metabolism, immune function and energy metabolism (Table 6).

Immune and stress response
Regarding the innate immune status, PP fed fish showed lower expression level of inflammatory related genes such as platelet-activating factor receptor, SAMHD1, tumor necrosis alpha-induced protein 2, and the antimicrobial peptide beta defensin 1. Likewise, fish fed the PP diet showed a suppression of genes involved in the adaptive immune response, i.e. TCR β, MHC class I, SMAD 3, CD80 (a T cell activator). Interferon γ induced proteins γ-ip (CXCL10), Gig2-like and both interferoninduced protein 35 and IRF2-binding protein were also lower in fish fed the PP diet. In contrast, Mx and Vig-2 which are responsive to type I interferon were higher expressed in PP compared to MP. Lastly, the stress induced genes HSP 30 and 47 were lower in expression in PP fed fish relative to MP fed fish.

Cell proliferation and apoptosis
Genes involved in cell proliferation were lower expressed in the PP fed fish, including ankyrin repeat domain-containing protein 54, nucleostemin and growth hormone-inducible transmembrane protein. Several genes involved in apoptosis were found to be higher such as cell death inducing protein and bh3 interacting domain death agonist although in contrast whereas both caspase 8 and 14 showed lower transcript levels in the PP fed fish.

Protein metabolism
Genes involved in protein synthesis such as several 60 S and 40 S ribosomal proteins were lower in expression in skeletal muscle of PP fed fish. A number of enzyme genes that modify essential amino acids (eg serine/threonine-protein kinase PINK1, serine/threonine-protein kinase 35, receptor-interacting serine/threonine-protein kinase 4 amongst others) were also lower in expression. Likewise, a major lysosomal peptidase, cathepsin D was lower in expression in PP fed fish, however, two components of the UbP were higher expressed in PP (ubiquitin-like protein precursor and proteasome subunit beta type-9). Genes encoding proteins involved in muscle structure and physiology (acta1 protein, actin-binding homolog 1a and tropomyosin-1 alpha chain) were also more highly expressed in skeletal muscle of fish fed the PP diet.

Energy metabolism
Genes related to energy metabolism were transcriptionally modified with peroxisomal membrane protein 11 C and cytochrome P450 being higher in expression whereas cytochrome c oxidases (COX2, 8 and 18) were found to be lower in PP fed fish compared to MP fed fish. In addition genes involved in glycolysis (phosphoglycerate kinase 1 and L-lactate dehydrogenase B chain) and gluconeogenesis (malate dehydrogenase 1) were found to be expressed lower in PP fed fish.

Co-regulated genes
Only a small number of genes were co regulated between tissues following the feeding trial. No genes were significantly higher or lower regulated in all the tissues in this experiment. The greatest co regulation was observed between mid intestine and skeletal muscle, with 9 genes significantly higher and 10 genes significantly lower in PP compared to MP. Genes that were more highly expressed in mid intestine and liver of fish fed the PP diet compared to the MP diet play a role in energy metabolism, such as List of selected mRNAs found to be higher or slower in expression following the feeding trial in mid intestine (a), liver (b) or skeletal muscle (c) and grouped according to functional classes (shown in bold). The selection was based on manual assignment of function and genes with greatest fold differences in expression are presented, the genes that are lower in expression are denoted by (−) value. The genes shown were significant at p < 0.05 following t-tests and greater than 2fold change. 1 Indicates the unique code for the feature on the microarray, 2 Accession number of the cDNA sequence. 3 Fold-change for genes higher or lower expressed in fish fed PP relative to fish fed MP diet. 4 Identity of the probe target as determined by BLASTX and BLASTN searches. cytochrome P450, a membrane-associated protein located in the inner membrane of mitochondria and peroxisomal membrane protein 11 C that induces peroxisomal proliferation proteins involved in the pentose phosphate pathway. Genes that showed lower expression in mid intestine and liver have a role in protein synthesis (40 S ribosomal protein S10) and in the immune response, such as SAMHD1 a nuclease involved in innate immune responses by acting as a negative regulator of the cellintrinsic antiviral response. Five genes were higher expressed in both intestine and liver of fish fed the PP diet, including immune related genes, such as CD209f and MyD88 and only 1 gene (complement c1q) showed lower expression in both tissues. Skeletal muscle and liver did not have any higher expressed gene in common and only 3 genes (membrane-bound O-acyltransferase domaincontaining protein 7, SMAD 3 and the non annotated gene (NP_998306) were significantly lower between these two tissues ( Figure 2; full list of genes expressed at different levels is given in supplementary Table 1).

Confirmation of expression by real time PCR
Real time PCR analysis was performed on a number of genes for each tissue to confirm microarray data ( Table 7). The real time PCR expression data was normalized with HPRT1 as the expression of this housekeeping gene was not found modulated by microarray analysis. For all genes the expression pattern showed the same trend between microarray and real time PCR analysis ( Figure 3).

Discussion
With the continual expansion of aquaculture of piscivorous fish there is a global requirement to reduce the dependence on wild caught fish for the generation of fishmeal and fish oil. There are a number of studies that have explored the transcriptional response to plant and vegetable oil replacements in fish diets, often when 100% replacement experimental diets are used results in reduced growth [24,25]. When both proteins and oils were replaced [24] in rainbow trout microarray analysis of liver indicated processes such as protein metabolism and cell cycle being altered, however the reasons for these changes could either have been combined effect of lowered essential fatty acids or changes in amino acid profile. There is also growing body of work demonstrating the genotype diet interaction in rainbow trout [25][26][27] and Atlantic salmon [28], with the latter concluding transcription of key metabolic regulators that respond to plant based feeds depend on the fish's genotype. Commercial salmon feed formulations partially replace fish meal with plant derived proteins, with varying effects on fish physiology and performance including fish growth and food conversion [24,29]. This paper examined the physiological and performance effects with transcriptional responses in mid intestine, liver and skeletal muscle in salmon to a post smolt feed containing high levels of plant derived proteins with low fishmeal content compared to a low plant protein-high fishmeal diet. The tissues examined play different physiological roles in the fish all of which may be affected by dietary changes as reflected in the different response profiles seen in these tissues. In common with a number of other transcriptomic studies on nutritional effects (where extreme dietary manipulations are avoided) the impact of the different diets appears quite subtle suggesting limited changes to physiological and metabolic pathways [30]. Although, in general, the scale of gene response is quite low, distinct changes in all tissues examined do indicate that there are responses occurring as a result of the dietary components. When interpreting the findings it is important to take a comparative holistic approach whereby apparent down regulation of genes in fish fed one diet may in fact be the result of up regulation in fish fed the other dietary treatment and vice versa. In this paper the expression is analysed relative to the marine profile (MP) diet.

Histological changes
The occurrence of histological changes was very similar for fish fed either diet, with low frequencies observed for the majority of parameters measured. The main histological observation was a mild fatty change in liver, which is considered normal under intensive rearing conditions. This anomaly appeared to have no clinical effect, and was not prevalent for either dietary group. Although there was a higher incidence of peri-biliar duct infiltration in liver of fish fed the MP diet no other histological changes were observed to indicate an inflammatory response. This reduced infiltration in the liver by the biliary ducts may be related to the altered immune gene expression in the liver. In mammals viral infections can result in biliary atresia [31] and specifically can be related to immune related injury to bile ducts following infiltration of CD4 + T and production of interferon γ [32]. At this stage we are unable to tell if the reduction in the structures is a direct result of the nutritional components or due to the altered hepatic (See figure on previous page.) Figure 2 Gene ontology (biological processes) found to be significantly enriched following the feeding experiment in mid intestine (a), liver (b) and skeletal muscle (c). Only GO categories for which >3 genes were represented are included.
immune gene expression in between the two diets. The distal intestine is often the target organ for anti-nutriental factors, in particular those present in SBM and pea protein concentrate [22], inducing histological changes including shortening of simple and complex mucosal folds with widening of central stroma, inflammatory cell infiltration in the submucosa and lamina propria with a mixed leukocyte population [33,34]. In this study both dietary groups presented fish with distal intestine anomalies but with low frequencies, and no significant difference between diets. These results together with those reported by Sanden et al. (2005) [35] indicate soybean products, including the alcohol soluble fraction, may be used within formulation constraints without inflicting gross histopathological anomalies in the intestine of Atlantic salmon.

Transcriptome changes in the mid intestine
The fish intestine has multiple functions which will be the first to respond to changes in nutritional intake; particularly digestion and absorption of nutrients and immune responsiveness to ingested pathogens, antigens and new antigens generated by the gut flora via gut associated lymphoid tissue [36]. The intestine contains three distinct regions: the proximal intestine containing pyloric caeca, the mid and distal intestine. Nutrient absorption occurs in all regions via enterocytes, by passive and facilitated diffusion and active transport [37], with the highest rates of uptake in the proximal section [37,38]. The intestine of piscivorous fish can be particularly sensitive to plant derived ANFs and non-starch polysaccharides (NSP) in feed, resulting in altered permeability in trout feed 44% 44% SBM [7,39] and enteritis in the distal intestine of salmon given a high dietary inclusion of SBM [33]. The inflammation/enteritis may be similar to a hypersensitivity reaction [33,40]. Often these effects are temporary and quickly disappear when the intestinal tract is no longer exposed to the ANFs [11,33,41]. Processing of plant products for fish feed is under continual improvement and some current plant derived protein concentrates have very low contaminating factors or botanical impurities. In addition knowledge of ANF containing plant feed materials has increased to the point where commercial feed formulations permit plant derived proteins at acceptable inclusion levels where no negative health effects or impacts on growth and performance occur. This was confirmed in the current study by the histology assessment where no gross morphological changes associated with plant ANFs were observed. In addition there was no difference in growth or feed utilisation efficiency during the feeding period, where a doubling of weight was achieved. However there were more subtle changes to biological processes that were not apparent during the classical physiological  Figure 3 Quantitative real-time PCR confirmation of genes expressed at different levels in mid intestine (a), liver (b) and skeletal muscle (c) of fish fed the PP diet compared to fish fed the MP diet for 6 genes identified by microarray analysis. Bars represent mean ± standard error of five fish, asterisks indicate significant (p < 0.05) differences. The genes chosen for the intestine were: Fatty acid-binding protein (Fabp2), Caspase-14 precursor, TCR alpha, IgM heavy chain, Tropomyosin-1 alpha chain (Tpm1) and Replication factor C subunit 3 (RFC3). The genes selected for the liver were: Serine protease HTRA1, TGF beta receptor, 28 S Ribosomal protein S18a (MRPS18A), T-cell immunoglobulin and mucin domain (Timd2), hepcidin I and TRC alpha. The gene chosen for the skeletal muscle were: cell death inducing protein (CDIP), acta1 protein, Transforming growth factor beta-1 (TGF beta 1), MHC class Ib antigen (RT1-CE5), 60Sribosomal protein L6 (60 S rib prot) and Heat shock protein 30. evaluation but were detected by global transcriptomic analysis. In this trial the mid intestine showed the greatest transcriptome response of the tissues studied, reflecting the sensitivity of the intestinal cells to dietary factors. Processes modified in the intestine were related to immune parameters, cell proliferation, apoptosis, protein metabolism, energy metabolism, transport and lipid metabolism. Fish fed the PP diet showed higher expression of genes involved in inflammation suggesting a possible dysfunction in immune regulation. Our findings support previous studies on gut intraepithelial and systemic T cells in fish which showed rainbow trout IELs are rich in T cells [10,42]. Additionally the expression of TSC22D3, a regulator of T cell receptor mediated cell death, was found higher in PP, this protein may be induced by glucocorticoids [43,44] activated by components in the PP diet. Together these results support previous reports that trace levels of substances with allergenic properties may cause expression of genes indicative of a hypersensitivity reaction in the intestine [45].
Interestingly genes involved in the inflammatory response were both higher and lower expressed in PP compared to MP. In particular genes involved in NF-κB pathway were induced such as the signalling adaptor molecule MyD88 and the inhibitory proteins of the IkB family, NF-kB1 p105 which sequesters NFκB in an inactive form in the cytoplasm [46]; inhibition of this pathway results in the production of proinflammatory cytokines [47]. MyD88 is also part of the signalling pathway that induces type I IFNs through the interaction of the MyD88-TRAF6-IRF7 complex [48]. IRF1 [49] had higher expression in fish fed the PP diet, this transcription factor may have activated interferon responsive genes in PP fed fish including Mx [50], virus induced gene (vig) -2 [51] and a virus induced TRIM protein [52]. PP fed fish also had higher expression of (SOCS)-7, which functions to reduce inflammation [53], potentially counteracting the expression of genes related to the inflammatory response in the intestine. Other genes involved in the innate immune response were expressed at lower level in fish fed PP. IL-8 is a chemokine that attracts neutrophils to a site of inflammation [54], whereas IL-17D coordinates the clearance of extracellular bacteria and contributes to the pathology of many autoimmune and allergic conditions in Atlantic salmon [55]. The lower expression of these genes in fish fed the PP diet may indicate there was not a proinflammatory response to the PP diet compared to MP diet.
Several antioxidant genes were expressed higher in PP fed fish mid intestine indicating protection against oxidative damage. The free radicals could be either endogenously produced by immune cells or present in the diet. Alternatively a potential lower concentration of antioxidants in the PP feed, due to the lower fishmeal content, may require the antioxidant system within the fish to be increased accordingly. The overall low expression of heat shock protein mRNAs, suggests a limited stress response in the intestine. Additionally, two genes directly involved in apoptosis process (caspase-3 and14) were expressed at higher level in PP potentially indicating increased apoptotic activity of mid intestinal cells in these fish compared to MP fed fish.
The intestine has an extremely high rate of cellular turnover and hence generally high levels of protein synthesis and protein degradation. Protein metabolism genes relating to both synthesis and degradation were found generally to be higher in PP fed fish suggesting an increase in intestinal protein turnover. Genes related to both transcription and translations were at a higher level such as translation initiation factors, elongation factors and the ribosomal protein S6 kinase. Interestingly, a number of mRNAs encoding ribosomal proteins were at a lower expression in PP. This may relate to the stability of the mRNAs or multiple use of the ribosomal subunits during translation. In parallel to general increase in synthesis genes related to protein degradation were also at a higher level in PP such as cathepsins and components of the ubiqutin proteasome pathway [56,57]. The higher protein turnover is likely to be energy demanding and this is related to an increase in genes encoding proteins involved in oxidative energy metabolism. Together these changes in expression suggest modulation in control of protein turnover in the intestine with components of both synthesis and degradation being altered which may reflect an increased activity of the intestine in fish fed the PP compared to the MP diet.
Cellular membrane transport related genes were higher expressed in PP, which could suggest that salmonids are able to adaptively modulate the densities of transporters to match changes in diet composition. Lipid metabolism and transport were also affected even though the PP diet contained the same fish oil content as the MP diet. mRNAs encoding two apolipoproteins were higher in expression in PP fed fish reflecting a greater mobilization and transport of cholesterol and fatty acids in the intestine, possibly an adaptive response to the lower dietary cholesterol content in PP compared to MP diet. Fatty acid metabolism genes were higher expressed in the PP diet including both FAD5 and 6, a PUFA elongase and other genes related to cholesterol metabolism. FADs are critical enzymes in the biosynthesis of long-chain highly unsaturated fatty acids (HUFA) from shorter chain PUFA [58,59]. These lipid metabolism differences between PP and MP are of interest as the intestine is often over looked regarding these processes and cholesterol, even if the essential fatty acids are present at required levels, other factors including cholesterol may change, revealing the complex nature of the early digestion and modifications of nutrients in the mid intestine.

Transcriptome changes in the liver
The liver receives nutrients and compounds from the intestine and needs to respond to any substances that may have detrimental effects on the fish.
The PP fed fish showed a significant difference to fish fed MP for genes related to immune function with a lower expression of innate factors such as lectins and hepcidin. Acquired immune system components were also found at lower level including T cell receptors, MHC I and II and components of the TGF-β pathway. TGF-β has an important role in the maintenance of T-cell [65] and B-cell homeostasis [66] by regulating cell proliferation process and apoptosis in these cells. This result, with the higher expression of genes involved in cell death such as CIDE-3 and angiopoietin-related protein 4, indicates that apoptosis may be a mechanism induced by the PP diet salmon liver. This is not surprising as apoptosis plays a central role in the differentiation and maintenance of the liver [67]. A balancing effect on the apoptotic TGF-β pathway is seen in the induction of several genes encoding proteins related to cell proliferation (such as HTRA1 serine protease and annexin A3). In particular, HTRA1 serine protease inhibits signalling mediated by TGF-β family proteins [68], playing an important role in contrast to cell death, whereas annexin A3 has a role in the signalling cascade during liver regeneration [69]. Other researchers have found the immune system to be altered following vegetable oil replacement in salmon [28] and in sea bass [70].
Genes involved in oxidative stress response (MPV17 protein, amine oxidase and HSP 70 kDa protein) were higher in liver of fish fed PP compared to MP. This is interesting to note as increases in antioxidant genes were also noted in a salmon diets that had marine oil replaced with vegetable oils [28] in the liver. In rainbow trout HSP expression in liver was increased following SBM rich replacement diets [9,62,63,71], the induction of these genes may indicate a diet-induced stress response in fish fed the PP diet. Moreover, during general high protein turnover to deal with misfolded proteins [72] as may be the case of fish fed PP diet.
It is interesting to note that few lipid related metabolic genes were found significantly different in the liver. Vigilin, a protein implicated in both biosynthesis and metabolism of lipids and steroids, facilitates removal of excess cholesterol from cells [73] and secondly apolipoprotein A IV which facilitates transport of cholesterol to the liver were both expressed at higher level in fish fed PP compared to MP. The high expression of apolipoproteins in fish fed diets containing high levels of plant proteins has previously been observed [9,74] and is most likely indicates reduced cellular cholesterol in fish fed the PP diet due to decreased dietary cholesterol, associated with low fishmeal content, and/or in response to trace levels of phytoestrogens [75] and phorbol esters [76] co purified with the plant proteins.

Transcriptome changes in the skeletal muscle
Genes related to processes such as protein metabolism, energy metabolism, cell proliferation, apoptosis and immune function were all significantly different in PP fed fish compared to MP. Transcripts related to protein metabolism such as ribosomal protein mRNAs, transcription and translation initiation factors were generally lower in PP fed fish compared to MP fed fish indicating lower protein synthesis. In parallel, protein degradation related genes were also less, for example the ubiquitin proteasome pathway and lysosomal peptidase proteins, together these would suggest a lower protein turnover activity in the muscle tissue in PP group. Both protein synthesis and degradation are highly energy demanding processes [77] and the indication of lower protein turnover, may suggest reduced energy wastage [61,78,79]. This idea is strengthened by lower expression of COX2 and COX8 and other genes encoding proteins involved in glycolysis and gluconeogenesis in fish fed PP. Genes involved in cell proliferation were also expressed at lower levels in PP fed fish indicating further energy saving. Together these changes in biological processes may indicate efficient metabolic activity following feeding of the plant protein enriched diet.
Biological tissues with high metabolic rate require mechanisms to deal with free oxygen radicals, on the other hand those tissues where the metabolic rate is reduced, for example when protein turnover is decreased a reduction in oxidative stress response may be observed as was observed in this study. Additionally a number of HSPs 30, 47 and a heat shock transcription factor 1a were all at lower levels in PP reflecting the reduced protein turnover and requirement of stabilising many newly translated proteins. Genes involved in cell proliferation and related to cell death including two caspases (caspase-8 and 14) were also lower expressed in fish fed PP compared to the MP diet. The induction of bh3 interacting domain death agonist, a pro-apoptotic member of the Bcl-2 protein family [80] and the suppression of caspases is in accordance with apoptosis of skeletal muscle in mammals where the Bcl2/bax system was found crucial for muscle apoptosis, whilst the caspase activity appeared inhibited [81]. Differential expression of a number of immune related genes, particularly a decrease in interferon responsive genes including γ-ip [82], Gig2-like [83] and interferoninduced protein 35 [84] were also expressed at lower level in PP. Other pro-inflammatory agents including platelet-activating factor receptor was also at a lower level PP, which regulates several pro-inflammatory functions such as chemokine and eicosanoid receptors [85]. An antimicrobial peptide beta defensin 1which is a central component of the non-specific defences [86] was also found lower in PP. Relating to the adaptive immune factors, TCRβ, MHCI and CD80 were all at lower levels in PP. Although we have observed differences in genes related to immune function in skeletal muscle of fish fed PP, the low level of inflammation and the subsequent immune response observed in the intestine did not cause a large immune shift in the muscle tissue. Instead, the lower expression of such genes, may allow fish fed PP to spend less energy resources on immune function for use in growth [61].

Conclusions
In conclusion, the present study confirm that a moderate level of plant protein derived proteins in a salmon diet allowed fish to grow with equal efficiency as those on a high marine protein diet, and with no difference in biometric quality parameters. The PP diet formulated with higher levels of soy protein concentrate, corn gluten, sunflower meal and wheat gluten resulted in significant effects on transcription in the mid intestine, liver, and skeletal muscle. The PP diet induced tissue specific changes in gene expression, with the mid intestine showing activation of the adaptive immune response indicating potential for hypersensitivity and an increase in protein turnover, although no difference in histopathological changes were observed in the proximate, mid or distal intestine. In liver cell proliferation and apoptosis indicate cellular reorganization and the general suppression of processes such as immune response. In contrast skeletal muscle tissue showed reduced protein metabolism and decrease in immune gene expression suggesting less energy expenditure in this tissue. The presence of only few genes in common between tissues may be due to the relatively mild changes that occurred and the complex nature of studying gene expression between tissue types on conservative dietary changes. These results improve the understanding of mechanisms and pathways activated by fishmeal replacement; in particular substitution by plant derived proteins and suggests that such diets can function well in Atlantic salmon aquaculture, hence to some extent reducing the burden on wild caught fish for fish meal. Additionally these results can assist in selection of molecular biomarkers useful for the development of new alternative feeds in salmonid aquaculture.

Fish husbandry and sampling
One hundred and five juvenile mixed sex Atlantic salmon of approximately 100 g were maintained in 6 replicate 1 m 3 tanks at SGS Chile Ltd., Puerto Mont, Chile, at 10.1°C and 27.6% salinity. Fish were fed a plant protein diet (PP) or a high marine protein diet (MP), both of which were formulated to the same digestible protein and energy content within formulation constraints of a commercial feed specification (CPK 100, 3 mm, 24/44 lipid: protein, BioMar S.A., Chile) ( Table 1). The feeding trial was conducted in triplicate tanks per dietary treatment and lasted for 77 days. At the end of the feeding period all fish in each tank were bulk weighed and 35 fish from each tank were killed by percussive stunning for individual biometric measurements of round weight, fork length, gutted weight, and weight of liver, heart or intestine for calculation of specific growth rate (SGR), feed conversion ratio (FCR), specific feeding rate (SFR), condition factor, and somatic indices of hepatic (HSI), cardiac (CSI) or intestinal (ISI) organs. Liver, heart and intestine was excised from distinct anatomical regions of the same individuals and fixed in 10% buffered formalin for histological assessment. All fish were sampled 30 minutes following the final meal to ensure there were no differences in postprandial gene expression. An additional 12 fish from each tank were sampled for mid intestine, skeletal muscle and liver and immediately stabilised in RNAlater (Ambion) at 4°C overnight then stored at −20°C until RNA extraction for gene expression analysis.

Histology assessment
Fixed tissues were submitted to dehydration process following paraffin embedding (Aquagestión, Puerto Montt, Chile). Sections (5 μm) were stained with hematoxylin and eosin (H & E) and examined under a light microscope. Micrographs were examined "blind" by the same experienced pathologist. Liver sections were evaluated for lipid degeneration level (fatty change) and the integrity of the whole organ. Heart tissue was evaluated for the presence of inflammatory infiltrates, myodegeneration and other possible abnormalities. Intestinal morphology was evaluated according to the following criteria: a) widening and shortening of intestinal folds, 2) loss of supranuclear vacuolization in enterocytes in the intestinal epithelium, 3) widening of central lamina propria within the intestinal folds, and 4) infiltration of (mixed) leukocytes in the lamina propria and submucosa [33]. The occurrence of histological changes for each dietary treatment were analysed by Kruskal-Wallis non-parametric Anova with InfoStat v 2009 (University of Córdoba, Argentina), with p < 0.05 considered significant.

RNA isolation
RNA was extracted from 100 mg of tissue by homogenisation in 1 ml TRIZol (Invitrogen) using tungsten carbide beads (3 mm, Qiagen) and shaking (300 times per min) following the manufacturer's instructions. The RNA pellet was washed in 500 μl 80% ethanol, air dried and resuspended in RNase free H 2 O. The concentration was determined by spectrophotometry (Nanodrop ND1000, LabTech) and the integrity of the RNA was determined by electrophoresis (Agilent Bioanalyser 2100). The RNA was then stored at −80°until required.

Microarray analysis
Microarray platform: Microarray experiments were performed using a custom-designed, Agilent-based microarray platform with 4 × 44 K probes per slide (Salar_2; Agilent Design ID:025520). The array contained primarily an Atlantic salmon resource with 34,441 features from Atlantic salmon coding sequences but additionally a further 9,111 features from rainbow trout (Oncorhynchus mykiss) sequences -the latter being selected when no homologues appeared to be available within Salmo salar datasets. Full details of the microarray platform and design are shown in Tacchi et al. 2011 [23].

Hybridization and analysis
For microarray analysis, 4 pools of RNA were produced for each tissue from fish fed PP and MP diets. Each RNA pool was an equimolar RNA mix from four different fish chosen randomly from each group. The microarray hybridisation was performed using a reference design, using a reference RNA sample, which comprised an equimolar mix of RNA extracted from all individual fish and tissue samples. Each experimental sample (labelled with Cy3 TM ) was hybridised against this reference sample (labelled with Cy5 TM ) in a 2-colour experiment. mRNA amplification, labelling and hybridization was performed as follows: mRNA was amplified using a MessageAMP TM aRNA Amplification Kit (Ambion). Briefly, 2 μg total RNA was reverse transcribed and the cDNA was used as a template for in vitro transcription in the presence of amino allyl modified dUTP, which allowed the generation of amplified antisense RNA (aRNA). For labelling, aRNA (3 μg) was denatured at 70°C for 2 min in a volume of 10 μl to which 3 μl of 0.5 M NaHCO 3 and 2 μl Cy dye (dye Cy3 TM or Cy5 TM mono-reactive dye pack, Amersham) was added. Incorporation of dyes was performed for 1 h in the dark, and after excess label was removed using a DyeEx TM 2.0 spin kit (QIAGEN) the amount incorporated was checked with a Nanodrop ND1000 (LabTech) spectrophotometer. Prior to hybridisation, 825 ng of each labelled template was fragmented in the presence of 11 μl of 10X blocking agent, 2.2 μl of 25X Fragmentation buffer (Agilent), and made up to a final volume of 20 μl with nuclease-free dH 2 O. The solution was then incubated in the dark at 60°C for 30 min, after which 57 μl of 2X GEx Hybridisation buffer (Agilent) was added to each sample and 103 μl of each hybridisation solution was dispensed on the Agilent 4x44K Atlantic salmon "Salmo salar2" oligo array (Agilent array design, 025520, Array express platform A-MEXP-1940). The hybridisations were performed in a Microarray Hybridisation Oven (Agilent) overnight (18 h) at 65°C. Following hybridisation, the slides were washed in Gene Expression Wash buffers 1 and 2 (Agilent) following the manufacturer's instructions. The slides were then scanned using a GenePix personal 4100A Scanner (Axon Instruments) at a resolution of 5 μm and saved as *. TIF files. Images were extracted and initial analysis was performed by Feature extraction v9.5.3 (Agilent) performing background correction of feature intensities (within the software). A Lowess normalisation of background corrected data was next conducted and all intensity values <1.0 were set to 1.0. Statistical analysis of the arrays was performed using Genespring GX analysis platform (version 9.5; Agilent Technologies). Quality control of the data was performed within Genespring and included removal of saturated probe features, non-uniform features, population outliers and those features showing intensities not significantly different from background in the Cy3 or Cy5 channels. After these relatively stringent procedures, 20,095 of the original 43,730 array features were maintained for subsequent analyses. The experimental hybridisations are at European Bioinformatics Institute archived under accession number E-MEXP: E-TABM-1207.
Significant differential expression between fish fed the PP diet and fish fed the control diet was established by t-test analysis (p < 0.05). Further filtering on fold change was conducted, and only transcripts showing more than two-fold change in expression were further characterised.

Analysis of gene ontology
Enrichment for gene ontology (GO) biological processes was performed on all cDNA features that had GO identifiers associated using the GOEAST program [87]. Fisher's exact test was performed within the GOEAST program to determine if GO identifiers occurred more often in a group than would appear by chance. For GO analysis only biological process GO identifiers were considered that occurred more than 3 times.