All experimental protocols were approved by the Norwegian State Board of Biological Experiments with Living Animals (authorization number: 1296, 1588, 2285).

Eight-week-old male C57BL/6J mice (Taconic, Ry, Denmark) were housed on a 12 h light/dark cycle, with free access to food and tap water. Mice were divided into weight matched groups and fed either control (C; 2018 Teklad Global, Harlan Laboratories, The Netherlands), VHF, VHF containing farmed salmon fillet (VHF/S) or VHF containing farmed salmon fillet with reduced POP concentrations (VHF/S_-POPs) for 8 weeks. Level of protein in the purified diets was adjusted to an isonitrogenous basis at the expense of carbohydrates and corn oil was adjusted according to lipids present in salmon fillet so that total dietary fatty acid concentrations were similar in all purified diets (Table S1). Accordingly, experimental diets contained (in ∼percentage of calories): 16% protein (casein with no added amino acids), 72% fat (50% lard and 50% corn oil), and 12% carbohydrate (sucrose) for VHF; 16% protein (salmon), 72% fat (50% lard, 19% corn oil, and 31% salmon oil from salmon fillet), and 12% carbohydrate (sucrose) for both VHF/S and VHF/S_-POPs. The energy content of purified diets was measured in a bomb Parr calorimeter 6300 (Parr Instruments, Moilne, IL, USA) and was found to be isocaloric (∼26.5 kJ/g).

In another set of experiments, mice were fed either chow low-fat diet (C; D12450B Research Diet, Taconic, Gentofte, Denmark), WD, or WD containing farmed salmon fillet (WD/S) for 6 weeks. Purified diets contained 19% protein (casein with no added amino acids), 29% fat (corn oil), and 52% carbohydrate (88% sucrose and 12% cornstarch) for WD; 19% protein (salmon), 29% fat (salmon oil from salmon fillet), and 52% carbohydrate (88% sucrose and 12% cornstarch) for WD/S. Energy content of both WD and WD/S was ∼20.5 kJ/g.

All purified diets were supplemented with cellulose, choline bitartrate, AIN vitamin mixture 76, and AIN mineral mixture 76 as described previously [17]. Once made, experimental purified diets were screened for POP concentrations (Table S2).

Casein (no added amino acids), lard, cellulose, choline bitartrate, AIN vitamin mixture 76, and AIN mineral mixture 76 were purchased from MP Biomedicals (Illkrich, France), corn oil was from Landlord (Reitangruppen, Denmark), cornstarch was from Primeal (Peaugres, France), and sucrose was from Dan Sukker (Oslo, Norway).

Commercial farmed Atlantic salmon fillets were bought from an international fish retailer (Leroy Seafood Group, Bergen, Norway). To reduce the presence of POPs in farmed salmon fillet, we tailored Atlantic salmon by feeding them a standard fish feed in which the fish oil used was purified rather than crude; crude fish oil being the principal source of POPs in fish feed [18]. Through this approach, we managed to produce farmed salmon fillet with ∼50% lower POP concentrations compared with commercial farmed salmon fillet (Table S3). After 8 weeks, Atlantic salmon (1137±215 g, n = 36) were filleted. Both commercial and tailored farmed salmon fillets were then freeze dried and homogenised. Before being used for animal studies, homogenised salmon fillets were analyzed for crude protein level with a Leco FP-528 nitrogen analyzer (Leco Corp, St. Joseph, MI, USA), amino acid levels were assessed with Acquity Ultraperformance LC (Waters, Milford, Massachusetts, USA), lipid content and fatty acid composition were analysed as described previously [17], [19]. Characteristics of salmon fillets are shown in Table S3.

Concentrations of organochlorine pesticides, dioxins, furans, dioxin-like polychlorinated biphenyls (dl-PCBs), and 7 polychlorinated biphenyls (7PCBs; PCB28,52,101,138,153,180) were measured as described previously [17].

Overnight fasted animals were anaesthetized with isoflurane. Organs were dissected, weighed, dipped in liquid nitrogen, and stored at −80°C until further analysis. Other tissues were fixed for histological investigations.

Concentrations of TAG were determined in gastrocnemius muscle and liver (left lobe) samples of overnight-fasted mice as described previously [17].

Three weeks before the end of experiments, feces present in cages were collected over a period of 7 days. After collection, feces were dried, weighed, and total lipid was extracted from feces by homogenization in chloroform:methanol, (2∶1, v/v) with 19∶0 methyl ester as internal standard as reported previously [20].

Small pieces of epididymal or liver (left lobe) fat were fixed overnight at 4°C by immersion in phosphate buffered 4% paraformaldehyde. Next, samples were dehydrated and embedded in paraffin. Hematoxylin and eosin (H&E) staining was performed on 5 µm thick sections to assess morphology. Other 5 µm sections were deparafinized, rehydrated and endogenous peroxide was inactivated by 3% hydrogen peroxide. To reduce non-specific staining the sections were incubated in heat-inactivated normal goat serum (10%, 10 min). Sections were then incubated overnight at 4°C with rat anti-mouse F4/80 (1∶500; AbD Serotec, Oxford, England), and then incubated ith rat anti-goat-IgG (1∶250; AbD Serotec, Oxford, England) conjugated with horseradish peroxidase for 2 h. Specific binding was visualised using diaminobenzidine.

We collected blood from the tail vein of conscious animals one week before the end of feeding trial. Plasma insulin concentrations were assessed by ELISA (Mouse Insulin ultrasensitive ELISA, DRG, Marburg, Germany). Blood glucose levels were measured with a glucometer (Ascensia Contour, Bayer Healthcare, Oslo, Norway).

Insulin and glucose tolerance tests were performed in fed and 5 h fasted mice, respectively. At week 7 and 5 for VHF and WD trials, respectively, mice were injected i.p. with human insulin (Actrapid, Novo Nordisk, Bagsværg, Denmark; 1.0 U/kg body weight) or glucose (1.0 mg/g body weight), and blood glucose was measured prior to injection and at 15, 30, 60, and 90 min after injection.

One week before the end of feeding trial, 5 h fasted mice were injected i.p. with 2 mg glucose/g body weight and plasma insulin measured at 0 and 15 min post-injection.

In vivo insulin stimulation was performed as described previously [21], with minor modifications. Overnight fasted animals were injected i.p. with saline or human insulin (20 U/kg body weight), and 10 min later, whole gastrocnemius muscles (both white and red parts) were harvested, frozen in liquid nitrogen, and stored at −80°C. Frozen muscle samples were weighed, homogenised and Akt/phospho-Akt (Ser⁴⁷³) was measured by immunoblotting as described previously [22]. Each sample was run at least three times. Antibody binding was detected by enhanced chemiluminescence (ECL; Amersham Pharmacia, Biotech, Buckinghamshire, UK). For semi-quantification of immunoblots, chemiluminescence imaging was assessed by G:Box system with Gene Snap software, and images were analyzed by Gene Tool software (Syngene, Cambridge, England). Primary and secondary antibodies were from Cell Signaling (Beverly, MA, USA).

Glucose uptake was assessed in intact soleus muscle, a small thin slow-twitch muscle, incubated without or with insulin (60 nmol/l) as described previously [22].

3T3-L1 preadipocytes were grown to confluence in standard high glucose Dulbecco's modified Eagle's medium (DMEM; PAA, Les Mureaux, France) supplemented with 10% bovine serum (PAA) at 37°C in a humidified atmosphere of 5% CO₂. Two days after reaching confluence, differentiation was initiated by incubation in DMEM supplemented with 10% fetal bovine serum, FBS (PAA), 10 µg/ml insulin (Sigma-Aldrich, Saint-Quentin Fallavier, France), 50 nM cortisone (Sigma-Aldrich) to avoid maximal differentiation [23], and 0.5 mM isobutylmethylxanthine (Sigma Aldrich). After two days, the medium was removed and the cells were cultured for three additional days in DMEM containing 10% FBS and insulin supplemented or not with different concentrations of pp′-dichlorodiphenyldichloroethylene (pp′-DDE; Sigma Aldrich). The medium was then removed and the cells were maintained in 10% FBS medium alone without pp′-DDE for two days. Lipid accumulation was quantified using Oil Red O staining. Oil red O was dissolved in isopropanol overnight at a concentration of 0.35% followed by filtration, dilution in water to a final concentration of 0.2%. Differentiated adipocytes were washed with PBS and fixed in 10% formalin for 60 min. After washing, cells were stained with Oil Red O for 10 min at room temperature. Following washes, cells were dried and Oil Red O was eluted by 100% isopropanol and absorbance was measured at 500 nM.

Total RNA was prepared from epididymal fat, and real-time PCR assays were performed as described previously (17). Briefly, first-strand complementary DNA's (cDNAs) were first synthesized from 250 ng of total RNA in the presence of 100 units of Superscript II (Invitrogen, Eragny, France) using a mixture of both random hexamers and oligo (dt) primers. The real-time PCR was performed with a Rotorgene Q (Qiagen, Courtaboeuf, France) in a final volume of 20 µl containing 5 µl of a 60-fold dilution of the RT reaction medium, and 15 µl of reaction buffer from the ABsolute QPCR SYBR Green Mix (Thermo Scientific). For quantification, a standard curve was systematically generated with 6 different amounts (150 to 30,000 molecules/tubes) of purified target cDNA cloned in the pGEM plasma (Promega, Charbonnières, France). Each assay was performed in duplicate. Data were normalized using TATA Box Binding protein (TBP) mRNA levels measured in each sample by real-time PCR as an internal standard and then expressed as arbitrary unit (AU). Primer sequences are available upon request (hubert.vidal@univ-lyon1.fr).

Results are expressed as mean ± SE. Differences between groups were examined for statistical significance using analysis of variance (ANOVA) with the least-square difference (LSD) post hoc test. Repeated measures ANOVA was used on data obtained from tolerance tests. When appropriate, statistical significance was determined using Student's t test. Differences were considered significant at p<0.05.

VHF feeding was reported to develop a diabetic phenotype before the onset of obesity in mice [24]. We took advantage of this dietary model and investigated the impacts of farmed salmon fillet intake in VHF-fed mice. After 8 weeks, mice fed VHF/S gained about two times more weight than other animals (Figure 1A) despite similar energy intake (Figure 1B). Interestingly, VHF/S-fed mice were characterized by enhanced fat absorption as demonstrated by reduced fat excretion in feces (Figure S1), a mechanism that likely participated to increase body weight gain of these animals. In association with their increased body weight, animals fed VHF/S exhibited increased visceral fat (Figure 1C), which was associated with a prominent increased of adipocyte size in epididymal fat pad (Figure 1D). Furthermore, the expression of Mac2-a, a galactose-binding lectin expressed by activated macrophages, was increased by about 13-fold in epididymal fat of animals fed VHF/S compared with VHF (Table 1), thereby suggesting macrophage infiltration in adipose tissue of animals exposed to farmed salmon fillet. Because activated macrophages may release inflammatory molecules, we next measured expression of IL-6, iNOS, and TNFα. In white adipose tissue of mice fed VHF/S, mRNA levels of TNFα and iNOS were up-regulated by about 5- and 3-fold, respectively, compared with VHF-fed mice whereas IL-6 expression was unchanged (Table 1). In concert, these results indicated that farmed salmon feeding robustly increases body weight and fat mass, and induces chronic-low grade inflammation.

We next examined glucose homeostasis of animals by challenging them to a glucose tolerance test. In comparison with C-fed mice, mice fed VHF had higher blood glucose levels at all time points following the glucose load (Figure 1E), thereby confirming the diabetic phenotype of VHF-fed mice despite the absence of obesity. Importantly, dysregulation of whole-body glucose homeostasis was worsen in mice fed VHF/S (Figure 1E), and was associated with increased insulin production in response to glucose challenge (Figure 1F). VHF/S-fed mice also exhibited a reduced response to glucose clearance following insulin load relative to VHF- and C-fed mice (Figure 1G); all features demonstrating a state of insulin resistance. Accordingly, mice fed VHF/S had increased blood glucose (Figure 1H) and plasma insulin (Figure 1I) relative to other animals. Because skeletal muscle is a predominant peripheral site of insulin-dependent glucose disposal [25], we further investigated insulin-stimulated glucose uptake in soleus muscles of animals. In the absence of insulin, basal muscle glucose uptake was similar in all experimental groups (Figure 1J). However, the ability of insulin to stimulate glucose uptake was reduced in muscles of mice fed VHF (-22%) and VHF/S (-47%) compared with C-fed mice (Figure 1J). In accordance with these results, we found that insulin-stimulated phosphorylation of Akt, a key insulin signaling protein, was decreased in skeletal muscles of both VHF and VHF/S-fed mice relative to animals fed control diet (Figure 1K). Loss of insulin action in skeletal muscles was associated with enhanced muscle TAG accumulation (Figure 1L), which is consistent with previous report [26]. Furthermore, the development of hepatic steatosis, which is a severe manifestation of metabolic syndrome, was also exacerbated in mice fed farmed salmon fillet (Figure S2). By using F4/80 staining, no mature macrophage infiltration was found in livers of VHF/S-fed mice (data not shown). Altogether, these data demonstrated that intake of farmed salmon fillet caused visceral obesity and accelerated the development of insulin resistance induced by VHF feeding.

To confirm the ability of farmed salmon fillet to induce insulin resistance and visceral obesity in a more conventional experimental diet inducing both insulin resistance and obesity, we challenged animals with high-fat/high-carbohydrate diet; WD. After 6 weeks, mice fed WD/S had higher body weight gain than mice fed WD and C (Figure 2A) despite similar energy intake (Figure 2B). As observed under VHF feeding, the percentage of fat excreted in feces was largely reduced in mice fed WD/S relative to WD (Figure S3), thereby suggesting increased intestinal fat absorption in WD/S-fed mice. Consistent with enhanced body weight, mice fed WD/S were characterized by an overgrowth of adipose tissue, including epididymal, retroperitoneal and inguinal fat pad (Figure 2C), and was associated with adipocyte hypertrophy and macrophage infiltration (Figure 2D). Next, we assessed glucose tolerance and insulin sensitivity of animals. In fasted state, there were no differences in blood glucose (Figure 2E) and plasma insulin concentrations (Figure 2F). However, in fed condition, a mild increase of blood glucose (Figure 2E) and dramatic increase in plasma insulin levels (Figure 2F) was found in mice challenged with WD/S, suggesting a whole-body insulin resistance state. In accordance, both glucose and insulin tolerance tests revealed that animals fed WD/S had impaired glucose tolerance (Figure 2G) and systemic insulin resistance (Figure 2H). Furthermore, insulin-stimulated glucose uptake in skeletal muscles was significantly reduced in WD/S-fed mice (Figure 2I), and correlated with increased muscle TAG storage (Figure 2J). Elevated TAG concentrations in liver of animals fed WD/S were also found (Figure S2). Taken together, these results demonstrated that mice fed WD/S, similarly to animals fed VHF/S, developed serious disorders linked to type 2 diabetes and obesity.

We have previously reported that POPs may counteract the benefits of salmon oil and impair insulin action in both in vivo and in vitro models [17]. To determine whether the presence of POPs modulates the outcomes of farmed salmon fillet intake, we compared the metabolic profile of mice challenged with VHF/S to those fed VHF/S_-POPs (Table S1 and S2). In epididymal fat, mice fed VHF/S_-POPs had about 20% and 50% lower concentrations of 7PCBs and dichlorodiphenyltrichloroethanes (DDTs), respectively, than animals fed VHF/S (Figure 3A), thereby indicating that mice fed VHF/S_-POPs had reduced body burdens of POPs. Interestingly, mice fed VHF/S_-POPs had reduced body weight gain (Figure 3B) and visceral fat compared with animals fed VHF/S (Figure 3C). This decrease in adipocity was correlated with reduced adipocyte size and macrophage infiltration as revealed by H&E and F4/80 staining (Figure 3D), as well as reduced Mac2-a expression (Figure 3E). In addition, mRNA levels of TNFα, but not IL-6 and iNOS, in white adipose tissue of mice fed VHF/S_-POPs were significantly down-regulated compared with VHF/S-fed mice (Figure 3E). There were no significant differences in energy intake and fat absorption between groups (data not shown).

Further, we investigated whether mice fed VHF/S_-POPs had comparable glucose intolerance and insulin resistance to animals fed VHF/S. In fasted state, blood glucose tended to decrease in VHF/S_-POPs relative to VHF/S-fed animals (7.1±0.6 vs. 8.3±0.6 mmol/l, respectively), whereas plasma insulin concentrations were significantly reduced in mice challenged with VHF/S_-POPs compared with VHF/S (137.5±17.4 vs. 175.7±10.44 pmol/l, respectively). Following glucose injection, blood glucose response was similar between mice fed VHF/S and VHF/S_-POPs (Figure 3F) whereas glucose-stimulated insulin production was reduced in mice fed VHF/S_-POPs (Figure 3G). Furthermore, mice fed VHF/S_-POPs had a tendency to better whole-body insulin sensitivity following insulin load (Figure 3H). To investigate this further, we measured ex vivo muscle glucose uptake. Consistent with enhanced insulin action, animals consuming VHF/S_-POPs had improved insulin-stimulated glucose uptake in soleus (Figure 3I), which was accompanied by reduced muscle TAG accumulation (Figure 3J), and tended to have less hepatic TAG concentration (Figure S2). Overall, these results highlight that POPs may affect the metabolic impacts induced by farmed salmon fillet intake. To further confirm a causal role of POPs in metabolic disorders, especially in adipose tissue related disorders, we investigated the ability of pp′-DDE, the DDT congener most present in salmon fillet, to stimulate lipid accumulation in 3T3-L1 cells. In this murine cell line, exposure to pp′-DDE was found to enhance lipid accumulation compared with vehicle-treated cells (Figure 3K). This finding highlights the ability of this organochlorine pesticide to regulate adipose tissue biology, and extends our previous observations that exposure to DDT mixture containing pp′-DDE impairs insulin action and glucose uptake in 3T3-L1 cells [17].