Regulation of metabolism by dietary carbohydrates in two lines of rainbow trout divergently selected for muscle fat content

The ability of fish to use dietary carbohydrates varies greatly among and within species in relation to their feeding habits. Carnivorous fishes are generally considered to be poor utilisers of dietary carbohydrates (Wilson, 1994; Moon, 2001; Hemre et al., 2002). For example, the rainbow trout (Oncorhynchus mykiss) exhibits persistent hyperglycaemia following ingestion of digestible carbohydrates or injection of glucose (Bergot, 1979; Wilson, 1994; Moon, 2001), without any clear increase in the provision of net energy (Hilton et al., 1987). Various assumptions were put forth to explain this poor utilisation, out of which insulin deficiency, lack of insulin-dependent glucose transporters and lack of an inducible hepatic glucokinase were proved to be false (Mommsen and Plisetskaya, 1991; Panserat et al., 2000a; Planas et al., 2000); low potency of glucose as an insulin secretagogue compared with amino acids, relatively low number of insulin receptors, lack of inhibition of endogenous glucose production and poor hepatic lipogenesis from glucose were proved to be true (Mommsen and Plisetskaya, 1991; Hemre and Kahrs, 1997; Navarro et al., 1999; Panserat et al., 2000b; Enes et al., 2009); reduced peripheral utilisation of glucose and inhibition of insulin secretion by other hormones requires more elaboration (Moon, 2001). Further, the presence of most of the key enzymes involved in carbohydrate metabolic pathways in fish indicate that the poor utilisation may be due to an aberrant hormonal and nutritional regulation caused by evolutionary adaptation (Cowey and Walton, 1989; Enes et al., 2009; Polakof et al., 2011a). For example, a carbohydrate-rich diet induces the hepatic glycolytic enzymes in rainbow trout very much as in mammals, but it does not affect the activity or mRNA levels of key enzymes of gluconeogenesis in complete contrast (Panserat et al., 2009). However, there are also reports of a better glycaemic profile in rainbow trout, which is genetically selected (Skiba-Cassy et al., 2009) or treated with anti-diabetic drug, linking enhanced de novo lipogenesis to better glucose homeostasis (Panserat et al., 2009; Polakof et al., 2011b).

From a practical perspective of salmonid aquaculture, the diminishing supply of fish meal and fish oil (the protein and energy source of choice in feeds, respectively) together with their escalating cost and demand emphasise the critical need to substitute them with ingredients of plant origin (naturally rich in carbohydrates) for the sustainability and expansion of production (Naylor, 2000; Tacon and Metian, 2008; Stone, 2003). Therefore, identifying and selecting new genotypes of fish with an improved capability to use the absorbed carbohydrate will be of economical advantage. Though some early studies do not indicate a promising prospect (Edwards et al., 1977; Austreng et al., 1977; Refstie and Austreng, 1981), others have revealed strains of salmonids exhibiting differences in the utilisation of dietary carbohydrates (Mazur et al., 1992; Blanc, 2002). Moreover, the large genetic variation for production traits in the farmed population of salmonids with medium to high heritabilities makes the fish family ideal for genetic selection (Gjedrem, 2000; Henryon et al., 2002).

Recently, two experimental lines of rainbow trout have been developed through divergent selection for low or high muscle fat content (Quillet et al., 2005). Metabolic studies suggest that the two lines differ in utilisation of energy sources when fed diets differing in lipid content but with a constant level (9–12%) of carbohydrates. Reduced hepatic fatty acid oxidation and enhanced glycolysis was observed in liver and muscle of the fat line (F) fish as compared with the lean line (L) fish (Kolditz et al., 2008a). The F line was also found to have lower plasma glucose levels 24 h after the last meal than the L line when fed a standard diet containing 10% starch (Skiba-Cassy et al., 2009). Based on these observations, we hypothesised that the F line has a higher ability to metabolise dietary carbohydrates than the L line. Our objective was to describe the metabolic response of both trout lines to diets with a high level of carbohydrates (17.1%) or without carbohydrates (<1%). We analysed the effects of genetic selection and dietary carbohydrate supply on growth, feed utilisation, body composition and post-prandial plasma metabolite levels. At the tissue level, we examined the signalling of intracellular energy sensors (AMPK), activation of insulin and nutrient pathway (Akt, TOR-S6) and gene expression of key enzymes involved in glycolysis, gluconeogenesis, lipogenesis, β-oxidation and fatty acid bioconversion in liver, which is the centre of intermediary metabolism. We also analysed gene expression in other tissues implicated in glucose utilisation, i.e. white adipose tissue and muscle.

The study was performed with two lines of rainbow trout (Oncorhynchus mykiss, Walbaum 1792), namely L (lean line) and F (fat line), obtained after four generations of divergent selection for high or low muscle fat content using a non-destructive method (Distell Fish Fatmeter, Fauldhouse, West Lothian, UK) as detailed by Quillet et al. (Quillet et al., 2005). Fish were reared in the INRA experimental facilities at Donzacq (Landes, France) at a constant water temperature of 17.5±0.5°C, under natural photoperiod during the months of July–September. They were fed a standard trout commercial diet (T-3P classic, Skretting, Fontaine-les-Vervins, France) during the acclimatisation period. Two experimental diets, namely 0C (without carbohydrate) and 20C (with carbohydrate), were prepared in our own facilities (INRA, Donzacq, Landes, France) as extruded pellets. Gelatinised starch was included as the carbohydrate source. The 0C diet contained <1% starch, whereas the 20C diet contained 17.1% starch (Table 1). The increase in dietary carbohydrate content was accompanied by a decrease in the proportion of protein in the 20C diet.

Fish of each line were distributed into six tanks of 35 animals (mean body mass 27±0.5 g). Triplicate groups of each genotype were fed either the 0C or the 20C diet for 10 weeks, two times daily ad libitum. The fish were bulk weighed every 3 weeks and counted to calculate the mean body mass. Feed intake was recorded and feed efficiency was calculated. At the end of the trial, nine fish per group (three per tank) were randomly sampled at 2, 8 and 24 h after the meal. Trout were killed by a sharp blow to the head and individually weighed. Blood was removed from the caudal vein into heparinised syringes and centrifuged (3000 g, 5 min); the recovered plasma was immediately frozen and kept at –20°C until analysis. The gut content of each fish was systematically checked to confirm that the fish sampled had effectively consumed the diet. Liver, perivisceral white adipose tissue and a sample of dorso-ventral white muscle were dissected, weighed, immediately frozen in liquid nitrogen and kept at –80°C pending analysis. The mass of liver was used to calculate the hepatosomatic index [HSI (%)=100×(liver mass/body mass)]. Six fish per genotype at the beginning of the feeding trial and three fish per tank at the end were killed and then frozen for analysis of whole-body composition. They were pooled per tank, ground and freeze-dried before being analysed. The experiments were carried out in accordance with the Guidelines of the National Legislation on Animal Care of the French Ministry of Research (decree no. 2001-464, 29 May 2001) and were approved by the Ethics Committee of INRA (according to INRA 2002-36, 14 April 2002).

Growth performance and nutrient utilisation traits were calculated for each tank with 35 fish, which represented one experimental unit. Final body mass was calculated as the final biomass divided by the number of fish in each tank at the end of the feeding trial. Daily growth coefficient was calculated as 100×(mean final body mass^1/3 – mean initial body mass^1/3)/day. Daily feed intake was calculated as the total amount of ingested food (kg) divided by the mean biomass over the trial [(initial biomass + final biomass)/2, expressed in kg wet mass (WM)] and the number of days. Feed efficiency was estimated as the gain in total biomass [(final biomass – initial biomass) (kg WM)] divided by the amount of ingested dry matter (kg DM). Protein efficiency ratio was the gain in total biomass divided by the amount of ingested crude protein (kg). Nitrogen and fat gain were calculated as (final carcass nutrient content – initial carcass nutrient content) divided by the mean biomass over the trial [(initial biomass + final biomass)/2 (kg WM)] and the number of days, where nutrient refers to nitrogen and fat. Protein and fat retention were calculated as [100×(final body mass × final carcass nutrient content) – (initial body mass × initial carcass nutrient content)]/nutrient intake, where nutrient refers to protein and lipid.

The chemical composition of the diets and the freeze-dried samples of whole body was analysed using the following procedures: dry matter after drying at 105°C for 24 h, fat by petroleum ether extraction (Soxtherm, Gerhardt, Konigswinter, Germany), protein content (nitrogen content × 6.25) by the Kjeldahl method after acid digestion, gross energy in an adiabatic bomb calorimeter (IKA, Heitersheim Grißheimer, Germany) and ash content by incinerating the samples in a muffle furnace at 600°C for 6 h. Total starch content in the diets was estimated by the enzymatic method (InVivo Labs, Vannes, France). Total lipid content in the muscle was measured according to the method of Folch et al. (Folch et al., 1957), with dichloromethane instead of chloroform as the solvent. Tissue glycogen levels were determined using the amyloglucosidase method (Keppler and Decker, 1974). Plasma glucose (Glucose RTU, bioMérieux, Marcy l'Etoile, France), triglycerides (PAP 150, bioMérieux) and free fatty acid (NEFA C kit, Wako Chemicals, Neuss, Germany) levels were determined using commercial kits adapted to a microplate format, according to the recommendations of the manufacturer. Total plasma free amino acid levels were determined by the ninhydrin reaction according to the method of Moore (Moore, 1968), with glycine as standard.

Protein extraction and western blotting were performed on liver samples of fish sampled 2 h after the meal. Six individual samples per experimental condition were used as biological replicates. Frozen livers (200 mg) were homogenised on ice with an Ultraturrax homogeniser in 2 ml of buffer containing 150 mmol l^–1 NaCl, 10 mmol l^–1 Tris, 1 mmol l^–1 EGTA, 1 mmol l^–1 EDTA (pH 7.4), 100 mmol l^–1 sodium fluoride, 4 mmol l^–1 sodium pyrophosphate, 2 mmol l^–1 sodium orthovanadate, 1% Triton X-100, 0.5% Nonidet P-40-IGEPAL and a protease inhibitor cocktail (Roche, Basel, Switzerland). Homogenates were centrifuged at 1000 g for 15 min at 4°C and supernatants were again centrifuged at 20,000 g for 30 min. The resulting supernatants were aliquoted and stored at –80°C. Protein concentrations were determined using the Bio-Rad protein assay kit (Bio-Rad Laboratories, Munich, Germany) with bovine serum albumin as standard. Liver protein lysates (10 μg of protein for Akt and S6, and 20 μg for AMPK) were subjected to SDS-PAGE and western blotting using the appropriate antibody. Anti-phospho Akt (Ser473), anti-carboxyl terminal Akt, anti-phospho-S6 (Ser235/236), anti-S6 ribosomal protein, anti-phospho-AMPK (Thr172) and anti-AMPK antibodies were used (Cell Signaling Technology, Saint Quentin Yvelines, France). These antibodies have been shown to cross-react successfully with rainbow trout proteins of interest (Skiba-Cassy et al., 2009; Polakof et al., 2011c). After washing, membranes were incubated with an IRDye infrared secondary antibody (LI-COR Biosciences, Lincoln, NE, USA). Bands were visualised by infrared fluorescence using the Odyssey imaging system (LI-COR Biosciences) and quantified by Odyssey Infrared Imaging System software (version 1.2, LI-COR Biosciences).

Analyses of mRNA levels were performed on samples from the liver, white muscle and perivisceral white adipose tissue of fish sampled 8 h after the last meal, the time interval corresponding to the post-prandial peak of nutrient absorption in juvenile rainbow trout reared at 17°C. Six individual samples per experimental condition were used as biological replicates. Total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's recommendations, quantified by spectrophotometry (absorbance at 260 nm) and integrity was controlled using the Agilent 2100 bioanalyzer (Agilent Technologies, Kista, Sweden). One microgram of the resulting total RNA was reverse transcribed into cDNA using the SuperScript III RNaseH-reverse transcriptase kit (Invitrogen) and random primers (Promega, Charbonniéres, France) according to the instructions of each manufacturer. Target gene expression levels were determined by quantitative real-time PCR as previously described, using specific primers (Kolditz et al., 2008a; Skiba-Cassy et al., 2009; Panserat et al., 2009; Polakof et al., 2009), except for Δ6 desaturase (forward: AGGGTGCCTCTGCTAACTGG; reverse: TGGTGTTGGTGATGGTAGGG; GenBank AF301910) and Δ9 desaturase (forward: GCCGTCCGAGGGTTCTTCTT; reverse: CTCTCCCCACAGGCACCAAG; GenBank FP323026, Sigenae clone tcba0028.c.12 5). The transcripts analysed were glucokinase (GK; EC 2.7.1.2), hexokinase 1 (HK1; EC 2.7.1.1), 6-phosphofructo-1-kinase (6PFK; EC 2.7.1.11) and pyruvate kinase (PK; EC 2.7.1.40) for glycolysis; glucose-6-phosphatase (G6Pase; EC 3.1.3.9), fructose 1,6-bisphosphatase (FBPase; EC 3.1.3.11) and phosphoenolpyruvate carboxykinase (PEPCK; EC 4.1.1.32) for gluconeogenesis; glucose 6-phosphate dehydrogenase (G6PD; EC 1.1.1.49), ATP citrate lyase (ACLY; EC 2.3.3.8), fatty acid synthase (FAS; EC 2.3.1.85) and the transcription factor sterol regulatory element binding protein 1-like (SREBP-1c) for lipogenesis; Δ6 desaturase (D6D; EC 1.14.19.3) and Δ9 desaturase (D9D; EC 1.14.19.1) for fatty acid bioconversion; and carnitine palmitoyl transferase 1 (CPT1; EC 2.3.1.21) and 3-hydroxyacyl-CoA dehydrogenase (HOAD; EC 1.1.1.35) for fatty acid oxidation. When different isoforms of a gene were known in rainbow trout (as for G6Pase and CPT1), gene expression analysis was performed on each isoform. The PCR run included quadruplicates for each sample (duplicates of reverse transcription and PCR amplification, respectively) and also negative controls (reverse transcriptase and RNA free samples). Elongation factor-1 alpha (EF1α) was employed as a non-regulated reference gene (Olsvik et al., 2005) and it was found to be stably expressed in this study. Relative quantification of target gene expression was performed using the mathematical model described by Pfaffl (Pfaffl, 2001) with the Relative Expression Software Tool (REST).

Enzyme activities were measured on liver samples of fish sampled 24 h after the meal. Six individual samples per experimental condition were used as biological replicates. Liver samples used to measure the lipogenic enzyme activities were homogenised by ultrasonic disruption in four volumes of ice-cold buffer (0.02 mol l^–1 Tris–HCl, 0.25 mol l^–1 sucrose, 2 mmol l^–1 EDTA, 0.1 mol l^–1 NaF, 0.5 mmol phenylmethyl sulphonyl fluoride, 0.01 mol l^–1 b-mercaptoethanol, pH 7.4), centrifuged at 24,000 g at 4°C for 20 min and the supernatant were used immediately for enzyme assays in pre-established conditions. G6PD (final substrate concentration 0.5 mmol l^–1 glucose-6-phosphate) and FAS (final substrate concentration 50 μmol l^–1 malonyl-CoA) were assessed following the method described in Figueiredo-Silva et al. (Figueiredo-Silva et al., 2010) adapted to trout tissues. PK enzyme activity was measured as described by Kirchner et al. (Kirchner et al., 2003). Enzyme reaction rates were determined by the increase or decrease in absorbance of NAD(P)H at 340 nm using a microplate spectrophotometer. Enzyme activity units (IU), defined as micromoles of substrate converted to product, per minute, at 37°C, were expressed per gram of liver. All of the measurements for each individual sample were performed in duplicate.

The results are presented as means ± s.d. The effects of diets, lines and the diet × line interaction on the different parameters were tested using StatView (version 5; SAS Institute, Cary, NC, USA) by means of a two-way ANOVA with diet and line as independent variables. Post hoc comparisons were made using a Student–Newman–Keuls multiple comparison test, and differences were considered statistically significant at P<0.05. When interactions were significant, means were compared using one-way ANOVA (P<0.05).

Feed intake was similar irrespective of line or diet (Table 2). At the end of the feeding trial, trout from the lean line (L) had a significantly higher body mass (P=0.005) than fish from the fat line (F), irrespective of diet. Feed efficiency was thus better in L line than in F line fish (P=0.03). Dietary treatments did not affect growth rate or final body mass. But as expected, the protein efficiency ratio was higher with the 20C diet (P<0.0001), because of the protein-sparing effect of the carbohydrates in this diet. The L line exhibited better protein efficiency (P=0.02) than the F line. The 20C diet improved the protein (P<0.0001) and fat retention (P=0.03) in both the lines. Daily nitrogen and fat gains/retention were not significantly different between the two trout lines. No significant line × diet interaction was recorded for any of these parameters.

Dietary treatments did not modify the whole-body composition of fish from both lines, but an effect of genotype was found (Table 3). F line fish had higher whole-body lipid content (P=0.046) and reciprocally lower moisture content (P=0.046) than the L line fish. The protein content was similar in both lines. As expected, the target of genetic selection, the muscle lipid content was significantly higher in the F line fish than in the L line fish (P=0.0004), with no significant dietary influence. Both trout lines showed higher HSI when fed the 20C diet (P<0.0001). The measure of liver glycogen levels (Table 3) confirmed the increased storage of glucose as glycogen when fed the 20C diet (P<0.0001). The F line fish had higher liver glycogen levels (P=0.03) than the L line fish, irrespective of the diet. The glycogen levels in muscle remained unaffected by the diets or genetic selection. In perivisceral adipose tissue, glycogen content was negligible (data not shown).

The results of post-prandial plasma metabolites measured at 2, 8 and 24 h after the meal are presented in Table 4. Plasma glucose levels were higher in fish of both lines fed the 20C diet at 2 and 8 h after the meal (P<0.0001 and P=0.001, respectively) compared with fish fed the 0C diet. However, the level of post-prandial hyperglycaemia induced by dietary carbohydrates was not very high (<8 mmol l^–1) in either trout line. Particularly in the F line fish fed the 20C diet, a statistically significant post-prandial hyperglycaemia was not observed but the plasma glucose levels were relatively higher at all time intervals. No significant effect of genetic selection was found for the plasma glucose level at any of the post-prandial times analysed. Plasma free fatty acid levels were not significantly different with respect to line or diet and all the treatments exhibited a similar post-prandial pattern with a peak at 8 h. The F line fish showed a significantly higher level of plasma triglycerides than the L line fish, irrespective of diet and time. The highest plasma triglyceride levels were recorded at 8 h after the meal, and were nearly twofold higher in the F line than in the L line (P=0.0002). Fish fed the 0C diet showed significantly higher plasma free amino acid levels than fish fed the 20C diet at 8 h, the point of post-prandial peak (P=0.02). The F line fish exhibited higher plasma free amino acids than the L line at 24 h (P=0.03).

Data on the mRNA levels of the glycolytic enzymes analysed are illustrated in Fig. 1. In the liver, GK mRNA levels were induced several hundred-fold in fish in both the lines when fed the 20C diet (P<0.0001); this induction was also dependent on genotype, with the F line showing a higher expression than the L line, as indicated by the significant interaction between diet and line (P=0.04). Expression levels of 6PFK mRNA remained unchanged between the lines whatever the diet. PK mRNA levels were uncharacteristically reduced in fish fed the 20C diet compared with those fed the 0C diet (P<0.0001), and F line fish showed a higher PK expression than L line fish (P=0.04). In muscle and adipose tissue, the mRNA levels of the glycolytic genes analysed were similar between lines and remained unchanged with dietary treatment, with one exception in adipose tissue. An interaction between diet and genotype was observed for PK (P=0.01), where the L line exhibited decreased mRNA levels compared with the F line when fed the 20C diet. The mRNA levels of the hepatic gluconeogenic enzymes analysed were found to be differentially regulated by dietary carbohydrates (Fig. 2). The 20C diet upregulated the expression of G6Pase-2 (P=0.04), whereas it significantly downregulated the expression of FBPase (P<0.0001). Trout from the F line exhibited significantly higher expression levels of FBPase and PEPCK mRNA than the L line (P<0.0001 for both). In the liver, mRNA expression levels of genes encoding enzymes involved in NADPH synthesis and lipogenesis (G6PD, ACLY and FAS) were significantly higher in the F line fish than in the L line fish (Fig. 3). The 20C diet enhanced the transcript levels of these lipogenic enzymes in the trout lines, particularly in the F line. A significant interaction between diet and line was observed for G6PD mRNA levels (P=0.04), with a strong induction in the F line when fed the 20C diet. No significant difference between the lines was observed for the transcription factor SREBP1c. Regardless of diet and line, no changes in relative transcript levels were observed for all the lipogenic enzymes examined in the adipose tissue except for the transcription factor SREBP1c, the expression of which was significantly higher in the F line than in the L line (P=0.01). Further concerning fatty acid bioconversion in the liver (Fig. 3), the F line demonstrated a significantly higher mRNA expression of D6D than the L line (P<0.0001) irrespective of diet. In the case of D9D, mRNA expression levels were considerably enhanced by the 20C diet in both lines (P=0.0003), but the level of induction varied significantly depending on the genotype, leading to a significant interaction (P=0.008). As shown in Fig. 4, dietary treatment induced no variation in the mRNA levels of candidate enzymes linked to fatty acid oxidation in all the three tissues analysed. But an effect of genetic selection was observed for the mRNA levels of the two hepatic β-oxidation enzymes investigated. The L line showed a significantly higher level of expression of both A and B isoforms of CPT1 than the F line (P=0.0002 and 0.0003, respectively), whereas an inverse line effect was observed for HOAD (P=0.0009).

To substantiate the results from gene expression analysis, the activities of two lipogenic enzymes, FAS and G6PD, were determined 24 h after the last meal (Fig. 5). The activity of FAS and G6PD were significantly enhanced by dietary carbohydrates in both fish lines (P=0.01 and P<0.0001, respectively), but no significant difference was observed between the lines. The PK activity was assessed because of its atypically decreased mRNA levels in both fish lines when fed the 20C diet in this study. In contrast to the mRNA levels, the activity of PK was not found to be different between the dietary treatments or the lines.

To better understand the gene expression regulation in liver (in the two other tissues the gene expression was almost constant), we investigated the effect of diet with or without carbohydrates on the activation of different signalling molecules involved in the insulin pathway (Akt), the nutrient sensing pathway (TOR-S6) and cellular energy homeostasis (AMPK) in the two lines. Fig. 6 illustrates the results of western blot analyses of phosphorylated (P-) and total forms of Akt, S6 and AMPK obtained in F and L lines fed either the 0C or the 20C diet. Analysis of the AMPK ratio data indicated that the phosphorylation levels were significantly reduced in fish fed the 20C diet compared with fish fed the 0C diet in both lines (P<0.0001), but were not different between the lines. In contrast, S6 ribosomal protein ratio data showed that the 20C diet increased the activation levels compared with the 0C diet (P<0.03), with no effect of genetic selection. Akt protein phosphorylation was not influenced by diet; however, it was slightly modified by genetic selection (P<0.04).

The hypothesis of the present study was that the F line has a higher ability to metabolise dietary carbohydrates than the L line, as previously suggested by Kolditz et al. (Kolditz et al., 2008a) and Skiba-Cassy et al. (Skiba-Cassy et al., 2009), who worked with these trout lines fed diets containing constant level of carbohydrates (∼10%). Our objective was thus to verify this hypothesis by analysing the metabolic response of the two trout lines to diets with or without carbohydrates. The effect of diet, genotype and their interaction are discussed separately below.

The diet with carbohydrate did not adversely impact feed intake, feed efficiency or growth of the trout lines because the inclusion level of gelatinised starch was within the acceptable limits and the energy from glucose was utilised as efficiently as that of protein (Kaushik et al., 1989; Kim and Kaushik, 1992; Wilson, 1994). The higher protein efficiency ratio and protein retention associated with dietary carbohydrate intake indicate a protein-sparing effect, as was expected (Kim and Kaushik, 1992; Hemre et al., 2002). Dietary carbohydrate intake resulted in larger liver and increased glycogen content, which is also well documented (Wilson, 1994; Hemre et al., 2002). Moreover, as expected post-prandial hyperglycaemia accompanied the intake of dietary carbohydrates, even though the level of induction was relatively moderate compared with previous studies using similar levels of carbohydrate intake (Panserat et al., 2000a; Bergot, 1979; Kaushik and Oliva Teles, 1985). In the F line, long-term carbohydrate intake resulted in an overall higher glycaemic profile that was not modulated post-prandially after a single meal. We studied the regulation of metabolism by dietary carbohydrates at the signalling, molecular and enzymatic levels in target tissues involved in glucose utilisation, i.e. liver, muscle and adipose tissue. However, we caution that in the present study it was not possible to reject completely any effect of dietary protein levels, as fish fed 20C diet ingested less protein. The absolute difference in carbohydrate intake (<1% versus 17.1%) between the two dietary groups, but not for protein (always higher than the requirement; 59% versus 45%), enables us to consider that the majority of the effects on intermediary metabolism are linked to the carbohydrate intake.

We analysed specific signalling pathways involved in nutrient (TOR-S6) and energy (AMPK) sensing in liver, the main tissue involved in nutrient homeostasis. We observed that carbohydrate intake decreases the phosphorylation of AMPK in rainbow trout just like in mammals (Assifi et al., 2005), which in turn enhances the activation of S6 mitochondrial protein (Wullschleger et al., 2006; Shaw, 2009). AMPK is a metabolic master switch by which cells sense and decode changes in cellular energy status, whereas S6 protein is a major indicator of the TOR signalling cascade involved in nutrient sensing and cell growth. In mammals, AMPK is known to exert a negative control over the TOR signalling (Rutter et al., 2003; Wullschleger et al., 2006; Shaw, 2009). Existence of these two proteins with a functional regulatory mechanism has already been reported in rainbow trout muscle and liver (Seiliez et al., 2008; Lansard et al., 2009; Polakof et al., 2011c). However, our results strongly suggest, for the first time in fish, that dietary carbohydrates contribute to a specific hepatic cell signalling response, as reflected by the decrease of AMPK phosphorylation as well as the increase of the S6 phosphorylation, when fish were fed with carbohydrates.

We analysed the glucose metabolism in liver, muscle and fat tissue. In liver, we reaffirm the recognised on–off regulation of GK in response to dietary carbohydrates at the transcriptional level (Panserat et al., 2000a; Panserat et al., 2009; Seiliez et al., 2011). The absence of molecular regulation of endogenous glucose production by dietary carbohydrates is also reiterated from the mRNA levels of the rate-limiting gluconeogenic enzymes, G6Pase and PEPCK (Panserat et al., 2000b; Panserat et al., 2001a; Kirchner et al., 2008). The synchronised higher expression of G6Pase 2 and GK suggests the possibility of a futile cycle following dietary carbohydrate intake; further investigations are required to validate it. Moreover, the existence of a paradoxical inhibition of hepatic PK by dietary carbohydrate at the mRNA level could also play a negative role in glucose utilisation in liver (Pilkis and Granner, 1992; Panserat et al., 2001b). This inhibition could result from an interaction between levels of dietary carbohydrates and polyunsaturated fatty acids, as reported in rats (Yamada and Noguchi, 1999), but this hypothesis requires further confirmation. However, contrary to previous reports in trout (Panserat et al., 2001b; Kirchner et al., 2008), FBPase expression was downregulated by dietary carbohydrates, similar to mammals and a few other fish species (Pilkis and Granner, 1992; Panserat et al., 2002). Though the exact reason behind this downregulation remains elusive, this could explain the moderate hyperglycaemia observed, as the rate of glucose production from substrates entering the pathway at the triose phosphate level would be limited. In the white muscle, the absence of dietary effect in the expression of glycolytic enzymes confirms the low response of this tissue to a high influx of glucose, as shown previously (West et al., 1994; Hemre et al., 2002; Panserat et al., 2009). A similar absence of differences in the scarcely studied adipose tissue reinforces the theory that glucose utilisation is poorly regulated in the peripheral tissues of carnivorous fish (Moon, 2001).

There is a strong relationship between glucose and lipid metabolism in animals, which can impact glucose use (Randle, 1998). Indeed, excess carbohydrate intake can be converted to fatty acids through de novo lipogenesis (Towle et al., 1997; Dentin et al., 2006). Lipogenesis from dietary carbohydrate has been previously demonstrated in rainbow trout by elevated lipogenic enzyme activities and lipid retention ratios above 100% (Kaushik et al., 1989; Brauge et al., 1995; Alvarez et al., 2000). We found that dietary carbohydrate intake enhanced mRNA levels and activity of hepatic G6PD, representing a potential higher production of NADPH, the cytosolic reducing equivalent required for lipogenesis, as shown previously in different fish (Lin et al., 1977b; Hilton and Atkinson, 1982; Hemre et al., 2002; Panserat et al., 2009). Further, even though the transcript abundance of SREBP1c, a transcription factor involved in regulation of FAS gene expression (Eberle et al., 2004; Ferre and Foufelle, 2007), did not increase with carbohydrate intake, we show evidence for increased hepatic lipogenic capacities through the enhanced activity of FAS and the elevated mRNA levels of ACLY, the main lipogenic enzyme diverting glycolytic carbon flux into lipid biosynthesis (Lin et al., 1977a; Lin et al., 1977b). Another interesting observation is that D9D transcript levels are also elevated by feeding dietary carbohydrates, suggesting a higher synthesis of monounsaturated fatty acids following the carbohydrate-induced lipogenesis, as in mammals (Tocher et al., 1998; Ntambi et al., 2004). Looking further along the fatty acid bioconversion pathway, we did not find any effect of dietary carbohydrate on the expression of another desaturase, the D6D, contrary to Seiliez et al. (Seiliez et al., 2001). The high level of fish oil and fish meal in the diets probably supplied enough of the essential fatty acids (Vagner and Santigosa, 2010). In the adipose tissue, we found no stimulatory effect of dietary carbohydrates on the expression of lipogenic enzymes, as has been reported in catfish (Likimani and Wilson, 1982), confirming the lack of response in salmonids to dietary manipulations (Lin et al., 1977a; Lin et al., 1977b). There was also no dietary impact on regulation of fatty acid catabolism, as reflected by the constant levels of CPT1 and HOAD mRNAs. Overall, the enhanced hepatic lipogenic pathway may play a pivotal role in storing excess glucose, ultimately leading to an improved glucose homeostasis in this glucose-intolerant species, as previously suggested using a specific anti-diabetic drug (Panserat et al., 2009; Polakof et al., 2011b).

We confirm the better growth potential of the L line compared with the F line, which has been reported from the third generation onwards, and it is not linked to a difference in the voluntary feed intake. Rather, it is associated with an improved feed efficiency and protein utilisation in the L line (Quillet et al., 2007; Kolditz et al., 2008a; Skiba-Cassy et al., 2009). The slightly higher whole-body lipid content in the F line can be attributed to the increased divergence in muscle lipid content with the progress of selection (Kolditz et al., 2008a; Skiba-Cassy et al., 2009). Moreover, the nutrient gain and retention estimates for protein and lipid do not reflect any preferential utilisation of energy substrates in the lines. Finally, and contrary to data from a previous study (Skiba-Cassy et al., 2009), the F and L lines did not show any difference in the post-prandial regulation of glycaemia, but the measures were not made in the same preceding conditions (fasted/refed). However, the analyses of the molecular actors in glucose and lipid metabolism clearly show significant differences between the two lines.

Studies at the transcriptional level suggest a higher glycolysis in the liver of F line as indicated by the elevated mRNA levels of the enzymes GK and PK (Kolditz et al., 2008a). The higher liver glycogen content in the F line suggests better glucose storage. However, we found that control over the hepatic endogenous glucose production was weaker in the F line than in the L line irrespective of the dietary carbohydrate levels, as indicated by the higher mRNA abundance of the gluconeogenic enzymes PEPCK and FBPase. This indicates that the potential export of glucose synthesised de novo through unregulated gluconeogenesis may counter-balance the uptake of glucose from the bloodstream for glycolytic purposes in the F line, which may explain why we did not find any difference in the plasma glucose levels between the two lines. We must note here that the cell signalling analysis in liver cannot explain the differences in glucose metabolism between the two lines: there was a slightly lower activation of Akt phosphorylation in the liver of the F line (which show paradoxically higher insulino-dependent metabolic pathways–glycolysis and glycogen) and no difference in S6 and AMPK phosphorylation, in contrast to our previous study (Skiba-Cassy et al., 2009). In the muscle, we did not observe an enhanced glycolysis in the F line, which was suggested as a possible genetic predisposition by Kolditz et al. (Kolditz et al., 2008a). This inconsistency may be due to differences in the generation of selection investigated (fourth versus third), rearing temperature (17°C versus 11°C), diet composition or post-prandial time of analysis (8 h versus 24 h). Our hypothesis of a better utilisation of glucose in the peripheral tissues of the F line was therefore not validated.

Regarding the lipid metabolism, we confirm the higher hepatic lipogenic potential in the F line (FAS, ACLY and G6PD) as previously shown (Skiba-Cassy et al., 2009). This is also reflected by the elevated plasma triglyceride levels in the F line. In contrast, in Atlantic salmon similarly selected for muscle adiposity, the fat genotype exhibits lower plasma triglyceride levels reportedly because of an increased lipid uptake by peripheral tissues (Morais et al., 2011b), but the lipogenic capacity of the genotypes were in agreement with our results. In the adipose tissue, though we found an abundance of SREBP1c in the F line, it was not accompanied by higher expression of lipogenic enzymes. The trout adipose tissue was recently proved to have lipogenic ability with possible involvement in glucose homeostasis (Polakof et al., 2011d). Overall, we confirm that the F line is characterised by higher lipogenic capacities than the L line, suggesting a better ability to store excess glucose as triglycerides. Moreover, the lower abundance of the hepatic transcripts of CPT1 a and b isoforms (catalysing the first limiting step for lipid catabolism) in the F line points towards a lower utilisation of fatty acids for energy supply, confirming earlier observations (Kolditz et al., 2008a). These data suggest a possibility for glucose to be used as energy source linked to a decrease in lipid oxidation in liver. Regarding fatty acid bioconversion, we found that the F line showed an enhanced ability through the higher D6D expression, although an opposite case of genotypic difference was reported in similarly selected Atlantic salmon fed a diet rich in vegetable oil (Morais et al., 2011a). The potentially decreased mitochondrial fatty acid oxidation coupled with increased lipogenesis and fatty acid bioconversion in the liver may define the fat muscle phenotype of the F line.

The F and L trout lines were recognised as unique biological material to study diet × genotype interactions focusing on the metabolic utilisation of carbohydrates, based on previous results (Kolditz et al., 2008a; Skiba-Cassy et al., 2009). However, we found only limited interactions. Growth, feed efficiency and nutrient retention did not show any interaction between diet and line, as was previously observed in different rainbow trout families (Edwards et al., 1977; Austreng et al., 1977; Refstie and Austreng, 1981). But the possible genetic determinism of adaptability to dietary carbohydrates in salmonids has been suggested (Mazur et al., 1992; Blanc, 2002). Mazur et al. (Mazur et al., 1992) reported a reduced post-prandial hyperglycaemic response in a specific strain of chinook salmon fed high levels of carbohydrate, whereas we failed to find such a response in the F line, in contrast to Skiba-Cassy et al. (Skiba-Cassy et al., 2009). In hepatic glucose metabolism, we show a diet × line interaction for the glycolytic enzyme GK and the NADPH-producing enzyme G6PD. Dietary carbohydrates enhanced their expression more specifically in the F line. In an earlier study using the same genotypes, with diets varying in their lipid content, these two genes exhibited a diet × line interaction (Kolditz et al., 2008b). This identifies them as potentially differentiated by genetic selection, responding also to dietary manipulations. Additionally, we report here a disordinal interaction for the hepatic D9D and adipose tissue PK expression, enhanced only in the F line when fed dietary carbohydrates. Though the few interactions endorse our hypothesis that the F line has a higher capacity for using high levels of dietary carbohydrates, they are insufficient to clearly establish this to be the case.

We combined genetic selection and nutritional strategies to investigate the genotype-specific metabolic response to dietary carbohydrates in trout. The enhanced hepatic lipogenic potential and the higher liver glycogen content in the F line illustrate its superior ability to store excess glucose compared with the L line. However, this does not result in an improved regulation of post-prandial glycaemia in the F line, possibly because of unregulated gluconeogenic enzyme transcripts. Because all our major results were obtained in the liver and not in the supposed major glucose utilisation tissues, i.e. muscle and white adipose tissue, our study evidences their lack of response to dietary manipulations with carbohydrates. The poor uptake and utilisation of glucose in these peripheral tissues may be one of the key factors underlying the impaired glucose homeostasis in rainbow trout. Moreover, the diet and genotype interactions that we observed in glucose metabolism (even being few) suggest that it is possible to obtain a trout genotype with an improved ability for lipid biosynthesis in the presence of dietary carbohydrates.

 
• 0C

  diet without carbohydrates

 
• 20C

  diet with carbohydrates

 
• 6PFK

  6-phosphofructo-1-kinase

 
• ACLY

  adenosine triphosphate citrate lyase

 
• Akt

  protein kinase B

 
• AMPK

  5′ adenosine monophosphate activated protein kinase

 
• D6D

  Δ6 desaturase

 
• D9D

  Δ9 desaturase

 
• F

  fat line

 
• FAS

  fatty acid synthase

 
• FBPase

  fructose 1,6-bisphosphatase

 
• G6Pase

  glucose 6-phosphatase

 
• G6PD

  glucose 6-phosphate dehydrogenase

 
• GK

  glucokinase

 
• HK1

  hexokinase 1

 
• HSI

  hepato-somatic index

 
• L

  lean line

 
• PEPCK

  phosphoenolpyruvate carboxykinase

 
• PK

  pyruvate kinase

 
• SREBP1c

  sterol regulatory element binding protein 1c

 
• TOR

  target of rapamycin

We thank M. J. Borthaire, E. Plagnes-Juan, C. Figueiredo-Silva, K. Dias, L. Larroquet, A. Surget and M. Larquier for technical assistance in the laboratory. We also acknowledge the technical staff of the INRA experimental fish farm at Donzacq (Y. Hontang, F. Sandres and F. Terrier) for conducting the feeding trial and the technical staff of the INRA experimental fish farm at Lees Athas (P. Maunas) for conserving the trout lines.