Background

The main driver for replacement of marine raw materials with alternative plant ingredients in fish feed is the ambition to maintain growth of the aquaculture industry and to secure flexibility regarding raw materials in feed production. However, in parallel with the decrease in fishmeal and increase in plant meals in fish feed, the prevalence of various intestinal disturbances has increased. Therefore, it is likely that some of the observed intestinal challenges may be due to deficient supply of nutrients, which are present at lower levels in plant ingredients than in fishmeal, but not corrected for due to lack of information on their essentiality and/or required level. The requirements for many nutrients have been defined for many species but all nutrient requirements are far from defined [1].

The present work addresses symptoms of a well-known intestinal disorder for which we suggest the term lipid malabsorption syndrome (LMS) and which since 2000 have been reported by salmon farmers to affect young as well as more mature fish [2,3,4]. The typical sign is increased lipid accumulation in the enterocytes giving the pyloric caeca a pale and foamy appearance on the macroscopic level. Similar signs have been reported also in other fish species fed diets high in plant meal [5,6,7,8] or high in plant oil [7, 9,10,11]. The apparent disturbance in lipid transport is also observed on the molecular level. Plant based diets may influence the expression of genes involved in lipid metabolism in a manner reflecting reduced lipid export from the enterocytes [8, 12,13,14,15,16,17,18,19,20]. However, the mechanisms underlying the excessive lipid accumulation are not yet fully clarified. Some studies seem to indicate that phospholipid synthesis, and in particular phosphatidylcholine, might be the bottleneck in lipid export from the enterocytes in fish showing such lipid accumulation [9, 10, 21,22,23,24,25,26,27]. Phosphatidylcholine, however, is not established as an essential nutrient for Atlantic salmon nor for any other fish species (NRC, 2011). For choline, on the other hand, of which about 95% is found in phosphatidylcholine [28, 29], a requirement is established for several fish [1]. Due to insufficient information, the question of whether choline is essential, and if so, the required amount, cannot currently be determined for Atlantic salmon.

In animals, including fish, choline is necessary for synthesis of phosphatidylcholine for use in lipid digestion and absorption, as a component in lipoproteins for lipid transport, in production of the neurotransmitter acetylcholine, and as a methyl donor in a wide range of methylation processes. Poor growth and low feed efficiency, fatty liver, high mortality, and anorexia are all reported effects of choline deficiency in the species for which we have documentation [1, 30,31,32]. Lipid accumulation in the intestinal mucosa is, however, not a common endpoint in studies of choline deficiency and requirement and has only been observed in an early study of Japanese eel (Anguilla japonica) as “white-grey colored intestines” [33]. This gut observation appears similar to that observed in Atlantic salmon with LMS. No studies have been conducted to define whether choline is essential for Atlantic salmon, or how much can be synthesized. Accordingly, a requirement is not established, and the question whether high plant diets might be deficient in phosphatidylcholine or choline, cannot be answered. Rainbow trout have been found to require choline at earlier life stages due to an inability to produce sufficient choline even with a high supply of methyl donors such as betaine and methionine [31]. On the other hand, channel catfish were able to produce sufficient choline, if the supply of methionine was high [1].

The level of fishmeal in today’s commercial salmon diets is in general low and decreases throughout the life cycle of the fish. A diet for salmon weighing 500 g or more typically contains between 5 and 10% fishmeal. Fishmeal would be the main contributor for choline in these diets, which means that supply of choline from the other main ingredients would be rather low. For example, the basal diet (LF) used in the present experiment was a commercially representative feed with 10% fishmeal which revealed a choline level of 944 mg/kg (Tables 1 and 2). With several recent reports from the salmon industry regarding LMS [4], investigation of the role of choline for LMS is needed. The present work therefore aimed to elucidate whether LMS is a result of insufficient choline supply and also addressed the role of choline in enterocyte lipid transport in post-smolt Atlantic salmon.

Table 1 Formulation and chemical composition of the experimental diets
Full size table
Table 2 Growth performance (Mean values with their standard errors)
Full size table

Results

Growth performance and nutrient digestibilities

Growth performance was significantly higher for fish fed the choline supplemented feed (LFC) compared to those fed the unsupplemented basal diet (LF, Table 2). Choline inclusion did not affect apparent digestibility (AD) significantly for any of the nutrients. The average AD (± SEM) for the two test diets was 96.1 (± 0.24) for crude lipid, 90.1 (± 0.19) for crude protein and 75.8 (± 0.65) for starch.

Intestinal chyme dry matter and brush border leucine aminopeptidase

Choline supplementation tended to increase dry matter content of digesta along the intestine (Table 3). The increase was significant for the mid intestine (MI) and distal intestine (DI) sections of the intestine. The trend was clear also for proximal half of the pyloric intstine 1 (PI1) and PI2 (p = 0.062 and 0.086, respectively). Brush border membrane leucine aminopeptidase (LAP) activities for PI and DI are shown in Table 3. There were no significant differences in the enzyme activity between the two treatments either in PI or DI tissue.

Table 3 Intestinal dry matter and leucine aminopeptidase activity (LAP) (Mean values with their standard errors)
Full size table

Organosomatic indices, intestinal and liver lipid content and histology

Relative organ weights of the PI, MI, DI and liver (LI) are shown in Fig. 1. Choline supplementation reduced relative weights of PI, MI and LI significantly, but not of DI. Macroscopic observations revealed white and swollen pyloric caeca in most of the sampled individuals fed the LF diet, whereas this observation was not recorded for any of the fish fed the LFC diet (Fig. 2a). Accordingly, the histological examination showed a significantly higher degree of lipid droplet accumulation in the pyloric caeca in fish fed the LF diet compared to those fed the LFC diet (Fig. 2b and c, respectively, p <  0.001). The degree of vacuolation of the enterocytes was 0% in sampled fish fed the LFC diet compared to 100% in the LF fed group (Fig. 3). Choline supplementation significantly lowered triacylglycerol (TAG) concentration in the tissue of the PI (Fig. 4, p = 0.024). No significant differences due to supplementation were found for free fatty acids (FFA), monoacylglycerol (MAG), diacylglycerol (DAG) and phospholipid (PL). The histological examination of LI vacuolation did not indicate similar effects of choline supplementation as in the pyloric caeca. No significant differences in the degree of liver vacuolation was found between the two diets (p = 0.867). Likewise, calculation of absolute amount of liver lipid (g) did not reveal significant differences (p = 0.867) between LFC and LF fed fish, 0.33 (± 0.04) and 0.33 (± 0.03), respectively.

Fig. 1
figure 1

Organ somatic indices of the intestinal sections, pyloric intestine (PI), mid-intestine (MI), distal intestine (DI) and liver (LI). Values are means (PI n = 20 and MI, DI and LI n = 30) with standard errors represented by vertical bars. Significant differences (p <  0.05) between the LF and LFC group are indicated with *. The inclusion of choline resulted in a significant lower organ somatic index for PI, MI and LI (p <  0.05)

Full size image
Fig. 2
figure 2

a example of white pyloric caeca with grossly visible of lipid accumulation. Image credit: Vegard Denstadli. Histological appearance of pyloric caeca in fish fed (b) the low fishmeal diet, LF and (c) the choline supplemented diet, LFC. Scale bare = 100 μm

Full size image
Fig. 3
figure 3

Contingency charts of the pyloric intestine showing proportions of sampled individuals that scored vacuolation grade “normal”, “moderate” and “marked” (none scored “mild”). Fish fed the low fishmeal diet displayed hyper-vacuolated enterocytes. Choline inclusion resulted in normal epithelium. The differences between the diets were significant (p <  0.05; Chi-squared test)

Full size image
Fig. 4
figure 4

Distribution of the lipid classes; free fatty acids (FFA), monoacylglycerol (MAG), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) in pyloric caeca tissue. Values are means (n = 10) with standard errors represented by vertical bars. Significant differences (p <  0.05) between the LF and LFC group are indicated with *. The inclusion of choline resulted in a significant lower content of TAG (p <  0.05; T-test)

Full size image

Intestinal gene expression

Figure 5 illustrates the molecular regulations of the studied genes involved in synthesis of phosphatidylcholine, cholesterol, and lipids, as well as intestinal lipid transport, lipoprotein assembly and secretion. Table 4 presents the results of the effect of choline supplementation on intestinal gene expression. Expression of genes encoding three enzymes involved in the pathway of phosphatidylcholine biosynthesis was analysed. The expression of the pcyt1a gene was significantly down-regulated whereas the effect for chk showed a trend towards down-regulation (p = 0.068). No significant effect was observed on expression of pemt. Genes involved in cholesterol (CH) transport were also significantly affected in fish fed the choline enriched diet. Niemann-Pick C1 like 1 (npc1l1) and abcg5 were up-regulated. Expression of fabp2 homologs, encoding fatty acid transporters, and the transcription factors pparα and pparγ were significantly enhanced. Also mgat2a, responsible for TAG re-esterification, was significantly up-regulated. A similar up-regulation was seen for both apoAI and apoAIV, involved in lipoprotein assembly. The general marker for lipid load of non-adipogenic cells, adipophilin/ perilipin 2 (plin2) was down-regulated. The taurine transporter slc6a6 was up-regulated in the choline treated fish.

Fig. 5
figure 5

Overview of genes involved in lipid digestion and absorption in the intestine of Atlantic salmon and studied in the present study. Arrows indicates steps in the pathways. Studied genes are italicized. Green color indicates genes which were significantly down-regulated and red color indicate up-regulated genes. No color represents genes not significantly affected. Dietary choline (CL) is synthesized by choline kinase (chk) to phosphocholine (P-CL) and after an intermediate step not studied here, choline-phosphate cytidylyltransferase (pcyt1a) to phosphatidylcholine (PC). PC could also be synthesized from endogenous phosphatidylethanolamine (PE) by phosphatidylethanolamine N-methyltransferase (pemt). PC is an important element in the membrane portion of lipoproteins preventing triacylglycerol (TAG) from leaking out. Cholesterol (CH) is transported from the lumen and over the membrane by Niemann-Pick C1-Like1 (npc1l1). Acyl-CoA cholesterol acyltransferase (acat) located in ER, facilitates the esterification of CH to cholesterol esters (CE). ATP-binding cassette G5 (abcg5) returns some of the free cholesterol back to the gut for reuse. Some of the free cholesterol is also shuttled to the basolateral membrane for biogenesis of high-density lipoprotein (HDL) mediated by ATP-binding cassette A1 (abca1). Fatty acids (FA) are transported from the gut lumen over the brush border membrane and into the epithelial cell by cd36 (cluster of differentiation 36). The fatty acid-binding protein 2 (fabp2) shuttles the fatty acids within the epithelial cell and the fatty acid transport protein (fatp) further to the smooth endoplasmic reticulum (ER). Monoacylglycerol (MAG) is esterified by monoacylglycerol acyltransferase (mgat2a), located in ER, to diacylglycerol (DAG) which is further transformed into triacylglycerol (TAG), a step not studied here. Microsomal triglyceride transfer protein (mtp) further facilitates the transport of TAG by assisting in the assembly of the lipoprotein. The three apolipoproteins apoB48, apoAI and apoAIV are important elements for successful production and secretion of the lipoprotein. The formation of lipoproteins is again an essential step for export of lipid to the general circulation and to other organs such as the liver. Excess lipid is stored as lipid droplets in the enterocytes. The lipid droplet structure and formation are regulated by the amphiphilic structural protein, adipophilin/perilipin 2 (plin2)

Full size image
Table 4 Gene expression profiling of pyloric caeca samples by qPCR
Full size table

Hepatic gene expression

In the microarray analysis (fold change > 1.6, p <  0.05), 168 entities were found to be differentially expressed between the two diet groups (Additional file 2). The differentially expressed genes appeared to be distributed among many functional classes, and a search for enriched GO and KEGG terms provided little meaningful information (data not shown). Among the highest responding transcripts, two innate immunity-related lectins (nattectin, c-type mbl-2 protein) were markedly induced by the choline treatment. In contrast, rxr, pmm and sod3 were down-regulated by choline supplementation. Perilipin 2 (plin2) showed up-regulation in the liver in contrast to the down-regulation found in PI. To further verify the microarray data, rxr, pmm, sod3 and plin2 were quantified by qPCR (Table 5). In accordance with microarray data, pmm and plin2 were down- and up-regulated, respectively, whereas no differences for rxr and sod3 were observed with qPCR. The lipoprotein and sterol associated transcripts measured in pyloric caeca were also quantified in liver using qPCR (Table 5). In accordance with the microarray data, we observed no significant changes for any of these transcripts. Altogether, microarray and qPCR data were closely correlated (Person’s correlation coefficient: 0.74, p = 0.0003).

Table 5 Gene expression profiling of liver samples by qPCR
Full size table

Blood plasma endpoints

Most of both the TAG and cholesterol in plasma was present in the high-density lipoprotein (HDL) fraction independent of treatment and the distribution of TAG and cholesterol among the lipoproteins were similar. Choline supplementation significantly decreased the plasma level of TAG reflecting reductions in HDL and low-densitylipoprotein (LDL) (Table 6). The opposite effect was seen on plasma cholesterol reflecting cholesterol increase in all the lipoprotein fractions. Plasma lathosterol, indicative of the rate of cholesterol synthesis, increased upon choline supplementation. The level of 7α-hydroxycholesterol, a metabolite in cholesterol catabolism and conversion to bile acids, also increased, whereas C4 (7α-Hydroxy-4-cholesten-3-one), a later metabolite in the cholesterol catabolism, was not significantly affected. Plasma levels of other catabolic products of cholesterol, i.e. the oxysterols 7β- hydroxycholesterol, 7β-keto-hydroxycholesterol, 24-hydroxycholesterol and 27-hydroxycholesterol were increased by dietary choline supplementation.

Table 6 Blood plasma variables
Full size table

Discussion

In brief, the present study revealed that choline supplementation to a plant based diet, 4.3 g/kg, improved growth by 18%, without effects on macronutrient digestibilities or other observed indicators of digestive function. The relative weight of the pyloric caeca decreased by 40% - reflecting a reduction in TAG and was shown histologically as elimination of enterocyte hyper-vacuolation. On the molecular level the supplementation caused down-regulation of genes involved in the CDP-choline pathway in which phosphatidylcholine is synthesized from free choline and a phosphorylated diglyceride (chk and pcyt1a), but had no significant effect on expression of pemt involved in synthesis of phosphatidylcholine from phosphatidylethanolamine via the PEMT pathway, the second pathway for phosphatidylcholine synthesis. Choline supplementation up-regulated two genes involved in cholesterol transport (abcg5 and npc1l1), as well as genes involved in lipid metabolism and transport (mgat2a and fabp2), and lipoprotein formation (apoA1 and apoAIV). The reduced intracellular lipid level was reflected in marked suppression of the lipid droplet marker plin2.

The aim of the present study was to elucidate if choline deficiency is a key contributor for LMS, and whether dietary supplementation with choline might prevent development of LMS. Our results clearly affirm these hypotheses. In this respect, our results are in line with the observations of lipid accumulation in Japanese eel fed choline deficient diets [33]. Our observations also highlight the importance of choline in lipid turnover in post-smolt Atlantic salmon, and supply information relevant for later developmental stages.

Choline effects on performance

Choline supplementation of the feed for Atlantic salmon of the size used in the present study, start weight 362 g and final weight 740 g, was found to have a great improving effect on SGR, by 18%. Similar improvements have been observed at juveniles stages in Atlantic salmon as well as in other species [Histology

Pyloric caeca and liver samples were processed at the Norwegian University of Life Sciences (NMBU) using standard histological techniques: dehydration in ethanol, clearing in xylene, and embedding in paraffin before sectioning (5 μm). Hematoxylin and eosin were used for tissue staining. The samples were evaluated for enterocyte vacuolation blinded in a randomized order using a light microscope. Vacuolation was assessed based on appearance of lipid-like vacuoles, swelling and irregularity of the cells, and condensation of the nuclei. Vacuolation was assessed semi-quantitatively as the proportion of total tissue affected: normal (≤ 10%), mild (10–25%), moderate (25–50%) or marked (≥ 50%) and presented as percentage of vacuolated enterocytes (Fig. 6).

Fig. 6
figure 6

Severity of vacuolation (steatosis) of the pyloric caeca tissue, representative for (a) marked (b) moderate (c) mild and (d) normal. Scale bar = 100 μm

Full size image

RNA extraction

Total RNA was extracted from pyloric caeca samples (~ 30 mg) using a Ultraturrax homogenizer, TRIzol® reagent (Invitrogen, ThermoFisher Scientific) and chloroform according to the manufacturer’s protocol. Obtained RNA was DNase treated (TURBO™, Ambion, ThermoFisher Scientific) and purified with PureLink RNA mini kit (Invitrogen, ThermoFisher Scientific). Total RNA from liver samples (~ 30 mg) were also extracted using Trizol® /chloroform whereas the homogenization was carried out using a FastPrep-24 (MP Biomedicals) before the samples were purified with PureLink RNA mini kit including an on-column DNase treatment according to the manufacturer’s protocol. The integrity of the RNA from pyloric caeca samples were assessed by gel electrophoresis, and in addition selected samples were verified with a 2100 Bioanalyzer using a RNA Nano Chip (Agilent Technologies). All liver samples were evaluated by Bioanalyzer. RIN values for both pyloric caeca and liver samples were all > 8. RNA purity and concentrations were measured using the NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies). Total RNA was stored at − 80 °C until use.

Microarrays

A two-colour microarray design was used for liver transcriptome profiling. Samples from five fish per treatment were labeled with fluorescent Cy3 and hybridized against a common reference sample (pool of 10 individual fish fed a fishmeal-based diet) labeled with fluorescent Cy5. Nofima’s Atlantic salmon 15 k oligonucleotide microarray SIQ-6 (GEO Omnibus GPL16555) was manufactured by Agilent Technologies (Santa Clara, CA USA). Reagents and equipment were from the same source unless indicated otherwise. RNA amplification and labelling were performed with a Two-Colour Quick Amp Labelling Kit and Gene Expression Hybridization kit was used for fragmentation of labelled RNA. A total of 200 ng RNA was used as input for each reaction. After hybridization in an oven over night (17 h, 65 °C, 10 rpm rotation speed), arrays were washed with Gene Expression Wash Buffers 1 and 2 and scanned with a GenePix 4100A (Molecular Devices, Sunnyvale, CA, USA). GenePix Pro 6.0 was used for spot to grid alignment, assessment of spot quality, feature extraction and quantification. STARS were used to carry out the subsequent bioinformatics data analysis [71]. Low quality spots were flagged by GenePix and filtrated away before Lowess normalization of log2-expression ratios (ER) was performed. Genes that passes quality control in at least four samples per group were included in subsequent analyses. The differentially expressed genes (DEG) were selected by the following criteria: fold difference > 1.6 and p <  0.05 (T-test). Enrichment of GO and KEGG terms in the list of DEG was assessed with Yates’ corrected chi-square using all probes that passed quality control as reference. Enriched terms corresponding to at least five differentially expressed genes were selected.

Quantitative real-time PCR (qPCR)

Quantification of hepatic gene expression by qPCR was conducted to validate the microarray results and to examine particular genes of interest in detail. qPCR was also used for quantification of genes related to lipid and sterol metabolism and transport in pyloric caeca. Assays were carried out in accordance to the MIQE standards [72]. First strand cDNA synthesis was carried out using four fish from each tank giving a total of eight fish per treatment, and Superscript III in 20 μL reactions (Invitrogen) with total RNA (0.8 μg) and oligo (dT)20 primers were used. Negative controls were performed in parallel by omitting RNA or enzyme. Obtained cDNA was diluted 1:10 before use and stored at − 20 °C. Quantitative PCR primers were obtained from literature or designed using Primer3web version 4.0.0 (http://bioinfo.ut.ee/primer3/). Detailed information of the primers is shown in Additional file 1. PCR reaction efficiency (E) for each gene assay was determined separately for both pyloric caeca and liver using 2-fold serial dilutions of randomly pooled cDNA. A LightCycler 480 (Roche Diagnostics) was used for DNA amplification and analysis of the expression of individual gene targets. Each 10 μl DNA amplification reaction contained 2 μl PCR-graded water, 2 μl of 1:10 diluted complementary DNA template, 5 μl of LightCycler 480 SYBR Green I Master (Roche Diagnostics) and 0.5 μl of each forward and reverse primer. Each sample was assayed in duplicate in addition to a no template control. The three-step qPCR program included an enzyme activation step at 95 °C for 5 min followed by 40 or 45 cycles (depending on the individual gene tested) of 95 °C (10 s), 58, 60 or 63 °C (10 s depending on the individual gene tested) and 72 °C (15 s). Quantification cycle (Cq) values were calculated using the second derivative method. The PCR products were evaluated by analysis of melting curve and by agarose gel electrophoresis to confirm amplification specificity. All primer pairs gave a single band pattern on the gel for the expected amplicon of interest in all reactions. For target gene normalization, actb, ef1a, gapdh, rnapolII and rps20 were evaluated for use as reference genes by ranking relative expression levels according to their stability, as described previously [73]. For liver samples, rnapolII was used as normalization factor, whereas gapdh was used for pyloric caeca. Relative expression of target genes was calculated using the Δ ΔCT method [74].

Chemical analyses

Diets and faecal samples were analysed for dry matter (after heating at 105 °C for 16–18 h), ash (combusted at 550 °C to constant weight), crude protein (by the semi-micro-Kjeldahl method, Kjeltec-Auto System, Tecator, Höganäs, Sweden), lipid (diethylether extraction in a Fosstec analyzer (Tecator) after HCL-hydrolysis), starch (measured as glucose after hydrolysis by alpha-amylase (Novo Nordisk A/S, Bagsvaerd, Denmark) and amylo-glucosidase (Bohringer Mannheim GmbH, Mannheim, Germany), followed by glucose determination by the “Glut-Dh method” (Merck Darmstadt, Germany)), gross energy (using the Parr 1271 Bomb calorimeter, Parr, Moline, IL, USA) and yttrium (by inductivity coupled plasma (ICP) mass-spectroscopy as described by Refstie et al. [75]. The plasma variables; free (non-esterified) fatty acids, cholesterol and total triacylglycerides were analysed according to standard procedures at the Central Laboratory of the Norwegian University of Life Sciences (NMBU). Lipoprotein profile analyses (HDL, LDL and VLDL) in plasma were carried out by size exclusion chromatography and measurements of cholesterol and triglycerides on-line using microliter sample volumes as described by Parini et al. [76]. Isotope dilution mass spectrometry as described by Lund et al. [77] was used for analyzing lathosterol. 7α-hydroxy-4-cholesten-3-one (C4) was analyzed by isotope dilution and combined HPLC-MS as described by Lövgren-Sandblom et al. [78]. Plasma levels of oxysterols, sitosterol and camposterol were analyzed by isotope dilution and combined GC-MS after hydrolysis as described by Dzeletovic et al. [79] for the first mentioned and by Acimovic et al. [80] for the last two mentioned. The lipid classes free fatty acids (FFA), monoacylglycerol (MAG), diacylglycerol (DAG), triacylglycerol (TAG) and phospholipid (PL) in the pyloric caeca were extracted using the Folch procedure [81], then analysed using HPTLC Silica gel 60 F plates. DigiStore 2: documentation was used for visual documentation and the integration program WinCats was further used for calculating the amount of the lipid classes.

Enzyme analyses

Brush-border membrane enzyme activity were analysed by measuring the activity of the enzyme leucine aminopeptidase (LAP; EC 3.4.11.1) in intestinal tissue homogenates. The homogenates were prepared from tissue thawed on ice-cold tris-mannitol buffer (1:20 w/v) containing the serine proteinase inhibitor 4-[2-Aminoethyl] benzensulfonylfluoride HCL (Pefabloc® SC; Pentapharm Limited). LAP activity was then determined colorimetrically with a kit (Sigma procedure no. 251) using L-leucine-β-napthylamide as substrate.

Calculations

Growth of the fish was calculated as specific growth rate (percent growth per day): SGR = ((ln FBWg / ln IBWg) / D) X 100. IBW and FBW are the initial and final body weight (tank means) and D is number of feeding days. Organ somatic index was calculated as percentages of the weight of the organ in relation to body weight. Apparent digestibilities (AD) of main nutrients was estimated by using Y2O3 [82] as an inert marker and calculated as: ADn = 100 – (100 X (Mfeed/Mfaeces) X (Nfeed/Nfaeces)), where M represents the percentage of the inert marker in feed and faeces and N represents the percentage of a nutrient in feed and faeces.

Statistical analysis

The diets in the present study were part of a larger trial. To obtain the best estimate of variance of tank means (SEM), results from all treatments were included. The other results of the experiment are published elsewhere [66, 67]. Tank was the experimental unit for all responses except for the histological observations for which the individual fish were the unit. Statistical analyses were performed using SAS (SAS Institute Inc., Cary, NC, USA). Data was analysed using the General Linear Model procedure with diets and tanks as class variables. Specific differences were evaluated by Duncan’s test. The level of significance was set to P <  0.05, and P-values between 0.05 and 0.1 were considered as indications of effects and mentioned as trends. All data are means ± SEM. A Chi-squared test was used for analyzing histology data.