Hemizygous male fat-1^+/− mice²² and their wild-type (WT) C57Bl/6 J littermates, were bred in-house at the Quebec Heart and Lung Institute animal facility (May 15, 2013 protocol). Seven-week-old mice were randomly assigned to five groups (n = 9–13/group) and housed individually in a controlled environment (12 h daylight cycles, 6 am–6 pm, 22°C), with food and water provided ad libitum. After 1 wk of acclimation on a chow diet (TD 2018), mice were fed either a low-fat [LF, TD 120651, Tecklad, 10% kcal fat (0.05% LA)–67% vegetable hydrogenated shortening, 33% corn oil, 7% kcal sucrose] or a HF diet [TD 93075, Tecklad, 55% kcal fat (3.9% LA)–95% vegetable hydrogenated shortening, 5% corn oil, 7% kcal sucrose]. Animals were gavaged daily during 12 wk with either ω-6 (safflower 0.21 mg/uL LA, President Choice) or ω-3 PUFA-rich oil (Webber Natural® capsules containing 900 mg ω-3 PUFA–600 mg EPA, 300 mg DHA) at 2.5μL/g of body weight. Doses were determined according to previous results and corresponded to ∼6% of the daily energy intake and equivalent to ∼4.6% of dietary lipids in the HF diet. Treatment groups were as follows: (1) for the Fat-1 part (genetic model): WT-LF, Fat-1-LF, WT-HF, and Fat-1-HF, all supplemented with ω-6 PUFA-rich oil; (2) for the fish oil supplementation part (dietary model): WT-LF and WT-HF supplemented with ω-6 PUFA-rich oil (LF-ω6 and HF-ω6) and WT-HF supplemented with fish oil (HF-ω3). WT-LF and WT-HF groups were used in both parts. In the dietary part, they were identified in the text as LF-ω6 and HF-ω6 and used as reference and control group, respectively, when compared with the HF-ω3 group. Energy efficiency was calculated as follows: final body weight (g)/total food intake (kcal)*100.

At weeks 10 and 12, mice were subjected to an intraperitoneal insulin tolerance test (ipITT, 0.65 U/kg) and an oral glucose tolerance test (oGTT, 1 g/kg), respectively.²⁶ After 12 wk, mice were euthanized by cardiac puncture following isoflurane anesthesia. Blood was collected in EDTA-coated tubes, centrifuged and the plasma fractions were collected and stored at −80°C. At sacrifice, tissues were collected, snap frozen in liquid nitrogen, and stored at −80°C. This study followed the Guide for the care and use of laboratory animals and all procedures had been previously approved by the Laval University Animal Ethics Committee.

Insulinemia was measured using the Ultrasentitive ELISA kit (Alpco 80-INSMSU-E01, Salem NH). Hepatic triglycerides (TG) were extracted in methanol/chloroform.²⁶ Plasma and hepatic TG were analyzed with Triglycerides Reagent kit (Randox Laboratories, UK). Cholesterol (Randox Laboratories, UK), nonesterified fatty acids (NEFA) (Thermo Scientific, USA) and leptin (Mouse/Rat Leptin Quantikineࣨ ELISA Kit, R&D Systems) were also measured in plasma samples. Alanine (ALT) and aspartate (AST) aminotransferases plasma levels were determined by biochemical analyses at the Quebec Heart and Lung Institute clinical biochemistry department. To determine cytokine concentrations in liver, proteins from 75 mg of tissue were extracted in PBS 1X buffer (Igepal 1% and proteinase inhibitors 1X). Proteins were measured (Pierce BCA Protein Bioassay kit) and standardized for the multiplex assay, which was performed according to the manufacturer’s instructions (Biorad). Analysis of FA profile in liver was achieved at CHU Sainte-Justine Research Center.²⁷

Plasma samples (40 μL) liver and muscle samples (5 to 10 mg) were extracted, and lipid mediators were measured using high-pressure liquid chromatography–tandem mass spectrometry (HPLC–MS/MS). The method can differentiate monoacylglycerol isomers at positions 1(3) and 2, but signals from both isomers of unsaturated FA were summed prior to analysis in order to account for their rapid interconversion. The following metabolites were quantitated: N-linoleoylethanolamine (LEA), N-arachidonoylethanolamine (AEA), N-docosahexaenoylethanolamine (DHEA), 2-linoleoyl-glycerol (2-LG), 2-arachidonoyl-glycerol (2-AG), 2-eicosapentaenoyl-glycerol (2-EPG), 2-docosapentaenoyl(n-3)-glycerol (2-DPG), and 2-docosahexaenoyl-glycerol (2-DHG).²¹

Lipid extraction and quantification process was adapted from Lamaziere et al.²⁸ and Shillito et al.²⁹ Briefly, total lipids were extracted from liver tissues by the method of Folch et al.³⁰ Individual phospholipid (PL) classes and their molecular species were quantified by a triple quadrupole Qtrap 6500, ABSciex (Les Ulis, France) in tandem with a Shimadzu Nexera XR liquid chromatography system (Shimadzu France, Marne la Vallee, France). A YMC-Pack PVA-Sil, bonded with a monomolecular polymer coating of vinyl alcohol (PVA) [Particle size: 5 μm, Pore size: 120 Å, Usable pH range: 2.0–9.5; (YMC, Japan)] was used to separate the different lipid classes. Concentrations of the PL compounds [phosphatidylethanolamine (PE), phosphatidylcholine (PC), and ceramide] were determined by comparing the peak area of each complex lipid with that of standards added with a known quantity.

Lipidomics: Tissues samples (25 mg) were homogenized using a bullet blender (Next Advance, NY, USA) in 300 μL of a chloroform: methanol solution (2:1) containing 0.005% of butylated hydroxytoluene (BHT), 100 mg of stainless steel 0.9–2 mm beads, and 50 μL of internal standard (Resolvin D1-d₅, RvD1-d₅; 1 ng/mL; Resolvin D2-d₅, RvD2-d₅; 5 ng/mL; leukotriene B₄-d₄; LTB₄-d₄, 5 ng/mL; DHA-d₅, 200 ng/mL; EPA-d₅, 200 ng/mL; 9S-hydroxyocatdecadienoic acid-d, 9S-HODE-d₄, 1 ng/mL; Cayman chemicals, MI, USA). Homogenates were centrifuged at 4713xg rpm for 10 min; supernatants were evaporated under nitrogen to 100 μL and diluted in 3 mL of water 0.1% formic acid before performing solid phase extraction (SPE) using Strata-X 60 mg columns (Phenomenex, Torrance, CA, USA) preconditioned with methanol and water 0.1% formic acid. Solid phase extraction columns were then washed 2 times with water (2 mL) and water: methanol solution (80:20) 0.1% formic acid (2 mL). Analytes were eluted using 2 mL methyl formate. Elutes were evaporated under nitrogen, reconstituted in 100 μL of water: methanol (40:60) and filtered with microspin 0.2 μmol/L filter column (Canadian Life Science, Dorval, Canada). The same procedure was also applied to analytical standards initially diluted in PBS. Fifteen microliter was then injected into the chromatographic system consisting of an ultra-high-pressure liquid chromatography (UHPLC) instrument (Shimadzu Scientific Instruments, Columbia, MD, USA). The chromatographic separation was achieved with a C₁₈ column from Agilent (150 × 2.1 mm Poroshell 120 EC; 2.7 μm particles; Santa Clara, CA, USA) at 40°C, and the following conditions: solvent A = ammonium formate in water (5 mmol/L) 0.01% acetic acid and solvent B = methanol 0.01% acetic acid. Separation was performed at a flow rate of 0.3 mL/min as follows: 60% B and linear gradient to 70.8% B over the next 12 min, B was increase to 80% over the next 7 min and increase to 85% over 5 min. Column was then flushed at 100% B during 4 min, and return to initial conditions over the next 7 min. All analytes were quantified by tandem mass spectrometry (MS/MS) using an API6500 instrument (Applied Biosystems, Concord, ON, Canada). The temperature was set at 550°C and entrance potential (EP) at −8 V. MS/MS parameters [ion transition (m/z); declustering potential (volt); and collision energy (volt)] were as follows: DHA (327.2→229.2; −80; −20); EPA (301.1→203.0; −80; −20); 17hydroxy-DHA (17S-HDHA: 343.1→201.1; −65; −20); 18R/S-HEPE (317.1→215.0; −80; −21) PDX (359.1→206.0; −60; −21); PD1 (359.1→206.0; −60; −21); RvD1 (375.1→140.9; −50; −20); 17R-RvD1 (375.1→140.9; −50; −20); RvD2 (375.1→175.0; −70; −31); DHA-d₅ (332.2→230.2; −80; −20); EPA-d₅ (306.1→208.0; −80; −20); 9S-HODE-d₄ (299.3→171.9; −80; −26); LTB4-d₅ (335.1→195.0; −75; −22); RvD1-d₅ (380.1→140.9; −50; −20); and RvD2-d₄ (380.1→175.0; −70; −31). Under these conditions, the limits of quantification were 1 ng for DHA and EPA; 10 pg for RvD2; 5 pg for 17S-HDHA, 18R/S-HEPE, RvD1, and 17R-RvD1; 1 pg for PDX and PD1.²⁴,³¹,³² Where a valid peak was detected but below the level of quantification (BLOQ) the method of utilizing level of quantification/2 (LOQ/2)³³ was employed.

Samples (25 mg) were homogenized using a bullet blender (Next Advance, NY, USA) and 100 mg of stainless-steel beads (0.9–2 mm) in 500 μL of a methanol solution (0.1% formic acid). Fifty microliters of internal standards (mix of CDCA-d4, DCA-d4, CA-d4, LCA-d4, and GCA-d4; C/D/N Isotopes Montréal, Canada) were added. Homogenates were centrifuged at 5000 g for 5 min. Supernatants were then evaporated under nitrogen and suspended in 1 mL of a water-0.1% formic acid solution, before performing SPE using a preconditioned (methanol and water-0.1% formic acid) Strata-X 60 mg 96-wells plate (Phenomenex, Torrance, CA, USA). Solid phase extraction columns were washed with water (2 mL) and a water: methanol solution 80:20 (v/v) containing 0.1% formic acid (2 mL). Analytes were then eluted using 2 mL methanol. Eluates were evaporated under nitrogen and reconstituted in 100 μL water: methanol 50:50 (v/v) prior injection to the LC–MS system. One microliter of each sample or analytical standard was then injected into the chromatographic system consisting of a Nexera UHPLC instrument (Shimadzu Scientific Instruments, Columbia, MD, USA). The chromatographic separation was achieved with a C18 column from Agilent (150 × 2.1 mm Poroshell 120 EC-C18; 2.7 μm particles; Santa Clara, CA, USA) at 37°C, and the following mobile phases: solvent A = ammonium acetate in water (5 mm) and solvent B = acetonitrile. Separation was performed at a flow rate of 0.3 mL/min using the following sequence: 67% A:23% B as initial conditions, then a linear gradient to 25% B over the next 10 min, followed by an increase of B to 35% in 20 min and to 60% in 10 min. Column was then flush at 95% B over the next 12 min and back to initial conditions for 8 min. All analytes were quantified by tandem mass spectrometry (MS/MS) using an API6500 instrument (Applied Biosystems, Concord, ON, Canada). The temperature was set at 500°C.

Protocol for short-chain fatty acid (SCFA) extraction in the caecum was adapted from Garcia-Villalba et al.³⁴ Briefly, 100 mg of caecal content were accurately weighted and homogenized in 1 mL of phosphoric acid 0.5% containing 4-Methylvaleric acid (diluted in ethyl acetate and used as intern standard). After 10 min of centrifugation, 17 949 × g at 4°C, supernatants were transferred in a new 1.5 mL conic tube with 1:1 (v/v) of ethyl acetate and vortexed during 2 min. Two aliquots of volume 200 µL each were stored at –20°C until analysis by gas chromatography-flame ionization detector (GC-FID).

For structural analysis of liver, paraffin-embedded tissues were sectioned to a thickness of 4–5 mm and fixed to glass slides. Once deparaffinized, slides were stained with H&E.

Total RNA, from 25 mg liver tissue, was homogenized in TRIzol reagent (Thermo Fisher Scientific) using a VWR beadmill and purified using the Direct-zol RNA MiniPrep kit (Zymo Research). The concentration and purity (A260/280) of the RNA samples were determined on a BioTech BioDrop. One microgram total RNA was reversed transcribed using the high-capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Primers (IDT) were optimized for efficiency and suitability as reference genes (Actb and Ppia) and are listed in Table S5. Real-time qPCR was performed using a CFX 364 (BioRad) and results were analyzed using CFX Maestro software (BioRad).

Bacterial DNA was extracted from fresh feces. Briefly, feces (30–50 mg) were homogenized in a lysis buffer containing lysozyme and incubated at 37°C, 30 min. Then, proteinase K and SDS 10% were added to the mix, and incubated at 60°C, 30 min. After 2 extractions with phenol:chloroform:isoamylalcohol [25:24:1 (v/v/v)] and phenol:chloroform [25:24 (v/v)], DNA was precipitated by addition of 100% EtOH in combination with 3 M sodium acetate at −20°C for 1 h. Finally, after centrifugation, precipitated DNA was re-suspended in Tris buffer (10 mm, pH8), and purified with Blood and Tissue Kit (Qiagen), according to the manufacturer’s instructions. DNA yield was assessed using a NanoDrop ND-1000 spectrophotometer (Thermo Scientific). Extracted DNA was stored at −80ºC until further use.

16S rRNA amplification of the V3–V4 region followed by sequencing were performed at the IBIS (Institut de Biologie Intégrative et des Systèmes—Laval University) platform.³⁵ Already-demultiplexed reads were then analyzed using the QIIME software package (version 1.9.1). Paired-end sequences were merged with at least a 50-bp overlap. Downstream sequence analysis involves removing low-quality reads (Phred score ≤ 25; presence of ambiguous bases «N») and sequences with minimum and maximum lengths ≤ 400 pb and ≥ 460 pb, respectively. Forward and reverse primers were trimmed from the filtered reads, followed by detection and removal of chimeric sequence artefacts, which was performed using UCHIME.³⁶ The resulting sequences were clustered into OTUs (Operational taxonomic units) using an open-reference methodology performed with USEARCH 61 version 6.1.544.³⁷ Representative OTU sequences were assigned taxonomy against the Greengenes reference database (August 2013 release)³⁸ using the naive Bayesian RDP-classifier.³⁹ Low confidence OTUs (singletons and OTUs with a number of sequences < 0.005% of total number of sequences) were discarded.⁴⁰ Rarefaction, to a subsampling depth of 5887 reads per sample (determined by the minimum number of sequences in a sample from a single time point) was performed on all samples of the dataset to normalize sampling effort. The RDP classifier against the RDP database (version September 30, 2016)⁴¹ was used to further classified OTUs that were unassigned against Greengenes at the genus level.

Levels of fecal bioactive flagellin and lipopolysaccharide (LPS) were quantified⁴² using human embryonic kidney (HEK)-Blue-mTLR5 and HEK-Blue-mTLR4 cells, respectively (Invivogen, San Diego, CA, USA). Fecal material was resuspended in PBS to a final concentration of 100 mg/mL and homogenized for 10 s using a Mini-Beadbeater-24 without the addition of beads to avoid bacteria disruption. Samples were then centrifuged at 8000 g for 2 min, and the resulting supernatant was serially diluted and applied on mammalian cells. Purified Escherichia coli flagellin and LPS (Sigma-Aldrich) were used for standard curve determination using HEK-Blue-mTLR5 and HEK-Blue-mTLR4 cells, respectively. After 24 h of stimulation, the cell culture supernatant was applied to QUANTI-Blue medium (Invivogen) and the alkaline phosphatase activity was measured at 620 nm after 30 min.

HepG2 cells were cultured in DMEM (Multicell) supplemented with 10% fetal bovine serum (FBS) at 37°C in 5% CO. Cultures were passaged with 0.05% Trypsin-EDTA at 80% confluence.

For determination of ω3-PUFA and derived endocannabinoid concentrations cells were plated into 6-well plates at 800 000 cells per well and left to adhere for 24 h. Cells were rinsed with HBSS, and treatment conditions applied: DMEM + 10% FBS (Control), DMEM + 10% FBS + 25 µm ω-3 PUFA mix (2:1 EPA:DHA) or 100 µm ω-3 PUFA mix (2:1 EPA:DHA). After 24 h, media and cells were collected and mixed with an equal volume of methanol with 0.01% acetic acid. A Bligh and Dyer for LC–MS method was used to extract lipids.

For lipid accumulation, experiments cells were plated into 48-well plates at 100 000 cells per well and left to adhere for 24 h. Cells were rinsed with HBSS and treatment conditions applied: DMEM + 10% FBS (untreated reference), DMEM + 10% FBS + 200 µm OA in DMSO (Control); the OA media was then supplemented with one: 100 µm ω3-PUFA mix (2:1 EPA:DHA), 50 nm DHEA or 50 nm EPEA and left for 24 h. Lipid accumulation was determined using AdipoRed following the manufacturer’s instructions. Briefly, cells were rinsed with PBS, then 0.2 mL PBS + 6 µL AdipoRed was added to each well. The plate was mixed and incubated for 10 min. Fluorescence was determined using a BioTek Cyt5 plate reader with excitation of 485 nm and emission at 572 nm. Lipid accumulation experiments were repeated four times in triplicate.

Data are expressed as mean ± SEM. For the Genetic model, statistical analysis was performed using a two-way ANOVA or a mixed model for repeated measurement analyses followed by Tukey’s post-hoc test. Data were log-transformed when violating the ANOVA postulates. Main effects (cut-off value at P = .10) for diet (D), gene (G), or time (T) factors and corresponding interaction effects were indicated under the title of each graph. For the Dietary model, a one-way ANOVA was performed followed by Dunnett’s post-hoc test, considering the HF-ω6 group as control. The corresponding nonparametric Kruskal–Wallis test followed by Dunn’s post-hoc test was used when data violated the ANOVA postulates. LF-ω6 group was considered as a reference for the HF-ω6 group and represented by a dotted line in bar graphs. For cell AdipoRed assay, oleic acid (OA) group was compared to Control through a two-tailed t-test, while a one-way ANOVA followed by Dunnett’s post-hoc test was performed to compare ω-3 PUFA and derivative treatments to OA. For endocannabinoid dosage in cells, a one-way ANOVA was performed. Significant differences (detected by post-hoc tests when P < .05) were recorded as follows: *P < .05, **P < .01, ***P < .001; with an exception for cell results where Control vs OA: ^###P < .001. Multiple factor analysis (MFA) was made using MFA function of the “FactoMineR” package and performed using R studio software. Permutational multivariate analysis of variance (PERMANOVA) have been made using Adonis function of the package “vegan” with 100 000 permutations. Multiple factor analysis plots were made with fviz_mfa() of the factoextra package and “ggplot2” package.

We first analyzed obesity-related phenotypic parameters and found a global genotype effect on body weight between Fat-1 mice and their WT littermates, leading to a lower final body weight with no effect on food intake. However, it did not reach significance in the Fat-1-HF mice compared with WT mice (P = .09) (Figure 1A–C). We also observed a genotype effect on energy efficiency that was decreased in Fat-1 mice (Figure 1D). The daily energy excretion, the VAT mass, and the plasma leptin levels were only affected by diet in the genetic model (Figure 1E–G).

Conversely, ω-3 PUFA supplementation prevented body weight gain induced by the HF diet, and it was statistically different from week 6 (Figure 1H). At week 12, HF-ω3 mice displayed a lower body weight gain (P < .05), a similar food intake, and a decreased energy efficiency (P < .01) compared with HF-ω6 (Figure 1I–K). Daily energy excretion was not affected by ω3-rich oil gavage; the decrease of VAT accumulation did not reach significance (P = .09) but circulating leptin levels were reduced (P < .05) in HF-ω3 compared with HF-ω6 mice (Figure 1L–N).

We next determined the impact of endogenous production vs ω-3 PUFA supplementation on glucose homeostasis. During the ipITT, the improvement of glucose response failed to reach significance in both Fat-1 groups compared with their WT littermates (genetic effect, P = .11) (Figure 2A). However, the corresponding AUC decreased in Fat-1-HF mice compared with their WT controls (P < .05) (Figure 2B). At 12 wk, the oGTT demonstrated that glucose tolerance was improved in Fat-1 mice (genetic effect, P = .02), regardless of diet (Figure 2C). The corresponding AUC were reduced in Fat-1-HF mice compared with WT-HF mice (P < .05) and associated with a lower fasting glycaemia (P = .06) (Figure 2D and E). Despite improved glucose response, the reduction of insulin secretion following the glucose bolus (GSIS) failed to reach significance (P = .07), the HOMA-IR index decreased significantly in Fat-1-HF compared with WT-HF mice (P < .05) (Figure 2F–I). On the contrary, ω-3 PUFA supplementation did not improve glucose homeostasis as shown by lack of change in fasting glycaemia, insulin sensitivity, glucose tolerance, GSIS, and HOMA-IR (Figure S1A–I).

Once ingested, food compounds directly interact with the gut microbiota. We thus tested whether increasing ω-3 PUFA bioavailability through digestive or genetic pathways could prevent gut microbiota alterations induced by HF diet consumption. Here, HF diet induced an expected shift in the gut microbiota populations compared with LF diet-fed animals (Figure 3A; at time 0, all mice were under regular chow diet, Figure S2A), which was characterized by a significant increase in relative abundance in Parabacteroides, Oscillospira,Dorea, Oscillibacter, Akkermansia, Ruminococcus, Streptococcus, Lactococcus, and some genera from the Clostridiales order, accompanied by an important depletion in Allobaculum (belonging to the Erysipelotrichales order) and Bifidobacterium (Figure S2B). Interestingly, compared with control, both models of ω-3 PUFA enrichment improved the gut microbiota composition by increasing specific bacterial groups that belong to the Erysipelotrichales and Turicibacterales orders, and decreasing the proportion of Clostridiales and Bacteroidales (Figure 3A). Both models of HF-fed mice cluster separately from the corresponding LF diet-fed groups on a PCoA; the HF-Fat-1 and HF-ω3, are intermediary to these clusters (Figure 3B and C).

Analysis performed at family and genus levels revealed that ω-3 PUFA supplementation had a higher impact on gut microbiota composition than the genetic fat-1 model. Focusing on similarities, results demonstrated a decreased relative abundance of Parabacteroides family in both models (Figure 3D and E). The relative abundance of a genus from the S24-7 family (recently renamed as Muribaculaceae family)⁴³ and Oscillospira were depleted in Fat-1-HF mice compared with WT-HF control (Figure 3D). Besides, ω-3 PUFA supplementation induced a reduction of Lachnospiraceae_g, Ruminococcaceae_g, Dorea, Proteus, and Coprococcus relative abundance and an increase in the Turicibacter and Allobaculum genera and (Figure 3E), the latter one being of great interest. Indeed, the major effect of both models was the similar increase of the Erysipelotrichaceae family, mainly represented by the Allobaculum genus (around 90% of the family), which was found to be significantly increased in the dietary ω-3 PUFA model (P < .05) but not in Fat-1-HF mice (P = .05; Figure 3D and E).

Fecal levels of bioactive LPS and flagellin were measured after 8 wk of treatment via cells engineered to express TLR4 and TLR5, respectively (Figure 3F–I). Results showed that endogenous ω-3 PUFA production had no significant impact on the concentration of these proinflammatory molecules (Figure 3F–G). In stark contrast, exogenous ω-3 PUFA administration through dietary exposure clearly impacts microbiota function, as revealed by the highly significant decreased fecal levels of LPS and flagellin (Figure 3H–I). Further analysis demonstrated that the endogenous production of ω-3 PUFA did not affect fecal lipocalin 2 (Lcn2), caecum weight, or caecal SCFA (Figure S3C–K), whereas caecum of HF-ω3 mice were heavier than HF-ω6 mice, with higher concentrations of isovalerate and isobutyrate (Figure S2L–T).

Knowing that the gut microbiota is closely linked to the liver through the portal vein and the importance of the gut-liver axis in metabolic health, we next investigated the impact of the 2 models on both circulating and hepatic lipid profiles. First, the determination of plasma lipid profiles revealed that Fat-1-HF mice displayed a lower cholesterol concentration compared with WT-HF mice (P < .05) (Figure 4A) but not circulating TG (Figure 4B), or NEFA (Figure S3A). Hepatic transaminases, nonexclusive indicators of liver injury, revealed a diet effect with ALT, but not AST, levels being significantly increased in Fat-1-HF compared with their WT littermates (Figure S3B–C). In the Fat-1 model, we found no genetic or HF diet-induced difference in EPA levels (Figure 4E). However, there was both a HF diet and Fat-1 genetic effect, which increased the DHA and docosapentaenoic acid n-3 (DPA) levels (Figure 4F–G). The HF diet effect increased the ω-6 PUFA linoleic acid (LA) (P = .05), while Fat-1 genetic effect decreased its concentrations. The genetic effect decreased AA (genetic effect, P = .01) compared with WT controls (Figure 4H–I). The detailed FA profile (Tables S1) exhibited a significant reduction of total ω-6 PUFA (P < .05) associated with the increase of total ω-3 PUFA (P < .05) in Fat-1-HF mice compared with WT-HF mice resulting in a reduced ω-6/ω-3 ratio (P < .001), which was also observed in the Fat-1-LF group (P < .001) (Figure 4J). However, at the tissue level, increased hepatic production of ω-3 PUFA in fat-1^+/− mice did not prevent hepatic lipid accumulation (Figure 4K), as reported previously.²³

In contrast, the supplementation of ω-3 PUFA significantly decreased plasma TG (P < .05) and cholesterol (P < .001) (Figure 4L, M), while it tended to decrease plasma NEFA (P = .08), ALT (P = .06), and AST (P = .07) (Figure S3D–F). In the liver, ω-3 PUFA supplementation completely inhibited ectopic fat accumulation compared with HF-ω6 control, decreasing liver weight (P < .01) and drastically reducing (−62%) hepatic TG (P < .001) (Figure 4N–O). In the dietary model, we found a higher concentration of hepatic ω-3 EPA, DHA, and DPA and a decrease in ω6 AA and LA in HF-ω3 compared with HF-ω6-treated mice (P < .001 for all these FA) (Figure 4P–T). Detailed lipid profile analysis (Table S1) revealed a decreased total ω-6 PUFA level (P < .001) and an increased total ω-3 PUFA level (P < .01), resulting in an ω-6/ω-3 ratio 12 times lower in HF-ω3 mice than in HF-ω6 mice (P < .01) (Figure 4U).

Due to their high PUFA content (at the sn-2 position), plasmalogens are considered key PUFA storage sites. Ether lipids are highly abundant molecules that account for approximately 20% of the total phospholipid content in mammalian cells.⁴⁴ We thus determined the hepatic levels of several markers of interest: phosphatidylcholine (PC) species, phosphatidylethanolamine (PE) species, total PC/total PE ratio, (Tables S2 and S3, Figure S4), and total ceramides (Figures S5 and S6). The hepatic levels of ether-linked PC and PE were higher after ω-3 PUFA tissue enrichment in both models (Figure S4), with a specific ω-3 PUFA increase in long-chain PE and PC 40:6 and 40:7 species (Tables S2 and S3). The ratio of total PC/PE (Figure S4C and F) a specific marker reported as inversely correlated with hepatic steatosis development⁴⁵ was also increased after both endogenous production and exogenous supplementation of ω-3 PUFA. These PL PUFAs can be further metabolized into potent second messenger molecules, such as protectins and resolvins, to induce anti-inflammatory or proresolving effects. We restricted our measurements to those from our previous fat-1 animal study²³ and for which we have developed protocols.²⁴,³¹ Liver tissue was analyzed for DHA-derived 17-HDHA, Resolvin D1, Resolvin D2, protectin D1, and DX along with the EPA-derived 18-HEPE. In the genetic model, both 17-HDHA and 18-HEPE were readily detected and quantified. Endogenous production of ω-3 PUFA tended to enhance the levels of these mediators in the liver of Fat-1-LF-fed mice but not in Fat-1-HF-fed animals (Figure S4D and E), while the downstream resolvins and protectin D1 were not detectable in WT animals. Protectin DX was detectable in Fat-1-LF-fed but not Fat-1-HF-fed mice but at a concentration BLOQ. Exogenous supplementation was more effective at augmenting the concentration of 17-HDHA and 18-HEPE mediators (Figure S4I and J). Similar to the genetic model, dietary supplementation did not lead to detectable levels of resolvins or protectin-D1 but did result in detectable levels of protectin, DX which were significantly increased compared to WT-HF (Figure S4K).

Ceramides, which are biologically active lipids, link nutrient excess to mitochondrial dysfunction and oxidative stress contributing to a proinflammatory insulin resistant environment and the development of steatosis.⁴⁶,⁴⁷ Here, we found that hepatic total ceramide levels were significantly reduced in the group of mice supplemented with ω-3 PUFA-rich oil only compared with the HF control group (Figure S6A), with no effect in fat-1 transgenic mice (Figure S5A). Numerous animal studies have reported that ω-3 PUFAs counter HF diet induced hepatic ceramide lipotoxicity thereby reducing inflammation, specifically TNF-α levels.⁴⁸ Regarding the hepatic inflammation status, the endogenous production of ω-3 PUFA in tissues led to a global genotype effect on hepatic inflammation. Indeed, IL-2, IL-10 and, surprisingly, TNF-α levels were decreased in both LF- and HF diet-fed Fat-1 mice (Figure S5B). On the contrary, in the dietary model, ω3-PUFA supplementation also induced a significant decrease of hepatic IL-2, IL-3, and IL-10 levels compared with control (Figure S6B), but while we observed a significant lower ceramide level, unfortunately this did not translate into significantly lower TNF-α levels (Figure S6B).

While both models exhibited a reduction in the ω-6/ω-3 ratio, ω-3 PUFA supplementation led to a greater improvement of hepatic steatosis and inflammatory lipid mediator profile, compared with endogenous production in Fat-1 mice. Interestingly, only diet (LF or HF) modified the hepatic bile acid profile, whereas no effect of either endogenous or exogenous ω-3 PUFA was observed (Table S4).

Endocannabinoids and their related bioactive lipids are ultimately synthesized from the corresponding PUFA-containing phospholipids (and in the case of 2-MAGs, also from PUFA-containing-lysophospholipids) and their levels can be influenced by dietary FA.⁷,⁸ Somewhat surprisingly lysophospholipid (LP)-EPA was not detectable and/or quantifiable in either WT or fat-1 mice regardless of diet. As expected both DHA-LP and DPA-LP (Figure 5A and B) were significantly increased by endogenous production of ω-3 PUFA. Increased ω-3 PUFA tissue concentrations did not result in reduced ω-6 related AA or LA concentrations (Figure 5C and D).

Thus, we investigated the impact of ω-3 PUFA supplementation or endogenous production on hepatic endocannabinoids and related NAEs and 2-MAGs and their FA precursors. The EPA-derived lipid mediators N-eicosapentaenoyl-ethanolamine (EPEA) was not detected in the liver and while 2-eicosapentaenoyl-glycerol (2-EPG) levels were not different between groups (Figure 5E). This is consistent with the reported lack of detection of EPA. However, and as expected, Fat-1 mice displayed higher levels of 2-MAG derivatives of DHA and DPA (2-DPG and 2-AG, P < .001 and P < .01, respectively), but unexpectedly, also those of AA and LA (2-AG and 2-LG, P < .05 and P < .01, respectively) compared with their WT counterparts, regardless of diet treatments (Figure 5F–I). Fat-1-LF mice showed an increase in hepatic N-docosahexaenoyl-ethanolamine (DHEA) concentration compared with WT-LF mice (P < .001), but this effect was abrogated with the HF diet (Figure 5J). However, the NAE derivatives of the ω-6 PUFA, AEA and LEA, were not modified by either diet or genetic model (Figure 5K and L). Without significant impact on hepatic palmitic acid concentration (P = .07), palmitic acid-containing MAGs and NAEs were surprisingly increased in transgenic mice where the effect on N-palmitoyl-ethanolamine (PEA) was observed in Fat-1-LF mice (P < .001) but not in their HF-fed counterparts (Figure S7A–C). We also observed a decrease of OA in transgenic animals and a counterintuitive increase of 2-oleoyl-glycerol (2-OG) in Fat-1-HF mice (P < .01) as well as an increase of N-oleoyl-ethanolamine (OEA) in both groups of Fat-1 mice (genetic effect, P = .02) (Figure S7D-F). HF-ω3 mice also displayed higher levels of the corresponding 2-MAGs containing ω-3 PUFA—ie, 2-EPG, 2-DHG and 2-DPG (P < .001)—while no difference was observed in the levels of AA and LA-containing 2-MAGs (Figure 5R–V).

Unlike the Fat-1-HF, the HF-ω3 had a significant increase in EPA-LP levels (Figure 5M) as well as significantly higher levels of DHA-LP and DPA-LP (Figure 5N and O) like the genetic model. The elevated LP levels were reflected in the ω3 2-MAG and NAE concentrations. Both ω6 PUFA, AA-LP, and LA-LP, were significantly increased with ω3 supplementation (Figure 5P and Q) but this did not result in an increase in related 2-MAG molecules (Figure 5U and V). Contrary to the genetic model, NAE levels were consistent with those of 2-MAGs as we reported an increase of EPEA and DHEA (P < .001), and a decrease in AEA and LEA (P < .001) in HF-ω3 mice compared with HF-ω6 mice (Figure 5W–Z). Lower hepatic oleic and palmitic acid levels were observed in HF-ω3 compared with HF-ω6 mice, but no effect was detected in 2-MAGs and NAEs containing these FA (Figure S7G–M). Thus, the hepatic concentrations of ω-3 2-MAGs and NAEs similarly increased in both models, with a greater effect in the dietary model. However, the two models differentially affected the hepatic concentrations of non-ω3 2-MAGs and NAEs, since PEA and/or OEA and/or 2-OG were sensitive only to the endogenous production of ω-3 PUFA, while ω-6 NAEs (but not 2-MAGs) were only decreased by ω-3 PUFA administration.

The muscle endocannabinoidome profile was quite similar to that of liver, but with some exceptions (Figures S7 and S8). In the gastrocnemius, we detected a genotype effect and an increase of 2-EPG and DHEA concentrations in Fat-1 mice (P < .001) (Figure S8A–E). The genotype effect observed in the liver for 2-LG (P = .03) and 2-AG (P = .002) was only driven by the WT-LF group in the muscle (Figure S8F and H), and we did not find any increase in 2-OG and OEA levels in Fat-1-HF mice in this tissue (Figure S8L and M). In the dietary model, ω-3 PUFA administration decreased gastrocnemius concentration of 2-LG (P < .01) and 2-AG (P < .001) (Figure S8S and U).

Finally, we measured the hepatic expression of Cnr1, coding for the cannabinoid receptor type-1 (CB₁), which is involved in the regulation of feeding, energy expenditure and reward system, and hepatic lipogenesis and insulin sensitivity, and is activated by AEA and 2-AG, but we did not detect a significant difference in either genetic Fat-1 mice or the dietary model, indicating that in the current dietary and genetic conditions endocannabinoid signaling is driven by ligand concentration and not receptor quantity (Figure S7N–O).

To understand better, the role of ω-3 PUFA-containing NAEs in liver steatosis prevention in the dietary model, we further performed in vitro experiments. We first observed that HepG2 cells exposed to OA during 24 h significantly increased lipid accumulation compared to control (Figure 6A). While co-treatment of OA-treated hepatic cells with the combination of the two ω-3 PUFA significantly decreased lipid accumulation, diminished lipid accumulation was significantly greater with either EPEA or DHEA single co-administration, with the level of lipid accumulation similar to that of control cells (no OA) for DHEA (Figure 6A). This finding is in agreement with the observed reduced lipid accumulation in the dietary model of ω-3 PUFA treatment, where DHEA and EPEA were significantly increased and hepatic TG was significantly decreased (Figure 4V), whereas in the genetic model, there was no significant increase in either EPEA or DHEA and no significant improvement in hepatic TG (Figure 4H). Therefore, this finding suggests that the two NAEs are very likely responsible for the lipid-lowering effects of exogenous ω3-PUFA in the liver. Accordingly, when hepatocytes were treated with ω-3 PUFA under the same conditions used to induce lipid accumulation, a trend towards increased levels of DHA and, particularly, EPA metabolites, including EPEA, was observed (Figure 6B–L).

To better understand the role of ω-3 PUFA in metabolism, we next performed an MFA. Quantitative variables (n = 165) were grouped into 11 factors (Figure 7A, Table S5) representing various metabolic pathways or biological features (eg, gut microbiota populations, hepatic FA, body composition) allowing us to determine each factor’s contribution to the observed differences between ω-3 PUFA enrichment models and to evaluate overall changes between mice. Mice tended to differ based on their genetic background (WT vs Fat-1, P = .08, Figure 7B) and were significantly differentiated by dietary fat content (high vs low, P = .002, Figure 7C) or the presence/absence of ω-3 PUFA, regardless of dietary fat content (P = .002, Figure 7D). Clusters corresponding to each HF diet treatment group position at similar levels on the dimension 2 axis (Figure 8E), while the LF diet mice were at a similar level as the WT-HF ω3 mice on dimension 1 axis. The mice fed a LF diet (WT or Fat1) had a similar profile for the variables contributing to dimension-1 axis along with the HF-ω3 mice. More specifically, the genera Allobaculum (from the factor Microbiota Figure 7A) was associated with the LF- diet (dimension-2, Figure 7G) and HF-ω3 (dimension-1, Figure 7F) mice while the liver FA species were associated with the WT-HF-ω6 mice (Figure 7H).

The hepatic concentration of the proinflammatory cytokine TNF-α strongly differentiated the three HF diet-fed groups along dimension 2 (Figure 7G and H) and to a lesser degree dimension 1 (Figure 7F). When analyzed within each treatment model the reduction in hepatic TNF-α was not statistically significant (Figures S5 and S6) but was consistently reduced by the presence of ω-3 PUFA either genetically or through dietary means. Other liver cytokines (IL-1β and IL-6, Figures S5 and S6) contributed to differentiating the groups but to a much lesser extent.

The strongest individual variable that contributed to the differences along axis 1 (ω-3 vs ω-6 PUFA, Figure 7E) is total hepatic FA concentration, specifically hepatic LA and palmitic acid (LIV_FA, LIV_LA, and Palmitic) and plasma leptin (Figure 7F). This was supported by the FA profiles (Table S1) and reduced body weight gain and increased plasma leptin (Figure 1B, I, and N) for the LF and ω-3 PUFA groups (Fat-1 or WT-ω3). The factors Sat (total hepatic saturated FA content), Unsat (total hepatic monounsaturated FA content) and NAFLD cluster along dimension 1 and while there is a less pronounced difference between the dietary and endogenous ω-3 models, the MFA in conjunction with the individual FA analysis (Table S1) provides insight into the observed difference in hepatic steatosis between these two groups (Figure 4F and T). From individual hepatic FA concentrations, we observed that dietary ω-3 PUFA significantly reduced total hepatic saturated FA content and specifically palmitic, eicosanoic, and stearic but there was no genetic effect (Fat-1, P = .07). Vaccenic acid, OA, and overall monounsaturated FA were reduced with dietary ω-3 PUFA, while the Fat-1 genotype did not have this effect (P = .05) and finally while the ω-6/ω-3 PUFA ratio (a variable within the NAFLD factor) was significantly lower in both ω-3 PUFA models, dietary ω-3 PUFA had a greater effect (Table S1). Therefore, our results are in line with our hypothesis that the supplementation of ω-3 PUFA has a more profound influence on the hepatic FA status, with the subsequent increased production of EPEA and DHEA at the expense of endocannabinoids possibly resulting in a lower hepatic fat accumulation.