High dietary arachidonic acid levels induce changes in complex lipids and immune-related eicosanoids and increase levels of oxidised metabolites in zebrafish (Danio rerio)

This study explores the effect of high dietary arachidonic acid (ARA) levels (high ARA) compared with low dietary ARA levels (control) on the general metabolism using zebrafish as the model organism. The fatty acid composition of today’s ‘modern diet’ tends towards higher n-6 PUFA levels in relation to n-3 PUFA. Low dietary n-3:n-6 PUFA ratio is a health concern, as n-6 PUFA give rise to eicosanoids and PG, which are traditionally considered pro-inflammatory, especially when derived from ARA. Juvenile zebrafish fed a high-ARA diet for 17 d had a lower whole-body n-3:n-6 PUFA ratio compared with zebrafish fed a low-ARA (control) diet (0·6 in the control group v. 0·2 in the high-ARA group). Metabolic profiling revealed altered levels of eicosanoids, PUFA, dicarboxylic acids and complex lipids such as glycerophospholipids and lysophospholipids as the most significant differences compared with the control group. ARA-derived hydroxylated eicosanoids, such as hydroxy-eicosatetraenoic acids, were elevated in response to high-ARA feed. In addition, increased levels of oxidised lipids and amino acids indicated an oxidised environment due to n-6 PUFA excess in the fish. To conclude, our results indicate that an ARA-enriched diet induces changes in complex lipids and immune-related eicosanoids and increases levels of oxidised lipids and amino acids, suggesting oxidative stress and lipid peroxidation.


Ethical considerations
The feeding experiment was approved by the Norwegian Animal Research Authority and was conducted according to current animal welfare regulations in Norway: FOR-1996-01-15-23. Facilities for zebrafish husbandry were optimally equipped to ensure refinement of breeding, accommodation and care. Handling and treatment of the fish ensured reduction of any possible pain, distress or lasting harm to the fish.

Formulation and preparation of diets
The feeding experiment and diets were made as previously described in detail (45) . In short, we fed zebrafish a control diet with low ARA levels (ARA 0·19 % of DM), or an experimental diet with high ARA levels (ARA 2·10 % of DM), referred to as high-ARA diet. The ARA levels were chosen on the basis of previous studies on fish to avoid deficient levels in the low-ARA group (22,46) . Protein, mineral and vitamin blends as well as oil composition are provided in Table 1. Initially, all feed ingredients were mixed (protein, mineral and vitamin blends, and oil including fish oil, rape seed oil, flax seed oil and Cargill's ARArich oil) with a solution of dissolved agar until a smooth texture was achieved. Astaxanthin was added to the agar solution before mixing with other feed ingredients. The feed paste was dried at 42°C for 72 h, ground and sieved into fractions of different feed-pellet sizes and stored at-20°C until feeding. The feeding regimen was as follows: <200 µm fed 27-43 DPF, 350 µm fed 44-57 DPF and 560 µm fed 58-90 DPF.

Experimental setup
Zebrafish AB strain (Danio rerio) were handled and fed as previously described (45) . In short, zebrafish embryos were collected randomly and incubated in Petri dishes. At 4 DPF they were transferred to beakers with sixty larvae (Fig. 1). Zebrafish were fed twice a day from 5 DPF with dry feed (Gemma micro ® ; Skretting) in addition to Artemia nauplii (Artemia; Silver Star) from 7 DPF until 27 DPF. At 15 DPF, larvae were randomly transferred into 3 litre tanks in a reverse osmosis water treatment system (Aquatic Habitats recirculation system). For each diet group we assigned ten sex-mixed 3 litre tanks containing sixty fish. Each tank represents one biological replicate. Both the control and the high-ARA diet were given from 27 DPF onwards, twice a day, until 90 DPF. Fish were fed ad libitum from 27-43 DPF, and thereafter from 44 DPF with a restrictive diet of 7 % of the tank total biomass (45) . Fish were kept under steadily monitored standard conditions with 28 ± 1°C, 14 h light-10 h dark period, conductivity of 500 µS and pH 7·5.

Sampling and growth measures
Fish were deprived of food 18 h before sampling. In all, 44 and 91 DPF zebrafish were anaesthetised with 0·05 % tricaine methanesulfonate (MS-222; Metacain) before weighing, standard length measuring and sampling. Anaesthetised whole fish were snap-frozen with liquid N 2 and stored at -80°C for fatty acid and metabolic profiling.

Fatty acid analysis in feed and fish
Fatty acid composition of the dried diets and 44 DPF whole fish was determined on isolated fatty acid fractions as previously described by Jordal et al. (47) , modified by Lie & Lambertsen (48) using GLC. Quantification of fatty acids was done using 19 : 0 as the accredited internal standard and integration of peak areas was done using Dionex Chromeleon (version 7.1.3.2425). Fatty acid quantification was done in whole-44-DPF-fish homogenates, where three parallels of twenty pooled individuals were made by combining two tanks.

Metabolic profiling
Targeted metabolic analysis involving high-throughput characterisation of all detectable metabolites was performed by Metabolon ® , Inc. in order to examine the effect of higher ARA levels on the general metabolism in 44 DPF zebrafish. Six parallels consisting of forty pooled individuals in each parallel were sent for analysis. Targeted metabolic profiling measures defined groups of characterised metabolites in a quantitative manner using internal standards. All applied methods used Waters ACQUITY Ultra HPLC and Thermo Scientific Q-Exactive high-resolution/accurate MS interfaced with a heated electrospray ionisation source and Orbitrap mass analyser operated at 35 000 mass resolution. Sample preparation, extraction and metabolite identification was as described by Skjaerven et al. (45) . Compounds were identified by comparison with library entries of purified standards by Metabolon ® . Values of detected compounds of known identity were normalised to the Bradford protein concentration, log-transformed and described as intensity scale.

Statistical analysis
Data visualisation and statistical significance testing of weight, length and fatty acid analysis data were performed using GraphPad Prism version 6.00. Weight and length data were analysed using a non-parametric test (Mann-Whitney test) as none of the data showed Gaussian distribution, except weight data at 91 DPF, which was analysed by an unpaired, parametric t test with Welch's correction. Fatty acid levels are presented as mean values and standard deviations and tested with an unpaired t test to reveal significant differences between the feed groups. Statistical significance was generally accepted at P < 0·05. For metabolic profiling, each value got rescaled to set the median = 1, and got described as scaled intensity in tables and figures. Missing values in one sample were assumed to be below the detection limit and were imputed with the minimum value from other samples for subsequent statistical analysis. Welch's two-sample t test (ArrayStudio; Omicsoft) was used on log-transformed data to identify significant (P < 0·05) metabolites with pairwise comparison between the two feed groups. Calculation of the false discovery rate (q value) took into account multiple comparisons that occur in metabolic-based studies by using a cut-off point (q ≤ 0·05) for indication of high confidence in a result. Relative fold changes, termed as mean ratios (MR), were calculated from high-ARA to control group using group averages of the scaled intensity values. Scaled data are presented in the online Supplementary Cytoscape Plugin was used to calculate sub-pathway enrichment. Pathway enrichment scores determine the number of statistically significant regulated compounds (k) relative to all detected compounds (m) in a pathway, compared with the total number of significant regulated compounds (n) relative to all detected compounds (N) in the analysis: (k/m)/(n/N).

Results
High dietary arachidonic acid affected weight but not length Dietary high ARA levels had a slight effect only at 44 DPF, where the high-ARA group was significantly lighter (P = 0·04) compared with the control group (Table 2). At this stage, both feed groups show a large weight variation. At 91 DPF we observed no differences in weight and length ( Table 2).

High dietary arachidonic acid affected fatty acid profiles
Feeding of experimental diets for 17 d changed the fatty acid profiles in 44 DPF zebrafish. The ratios of n-3:n-6 PUFA with control and high-ARA feed were 0·6 and 0·2, respectively (Table 3). GLC analysis revealed six times higher ARA concentrations in the high-ARA group compared with the control group, accompanied by significantly elevated levels of its elongated and desaturated products adrenate (22 : 4n-6) and n-6 DPA (22 : 5n-6). Fatty acid levels are presented in Table 3 (full list of analysed fatty acids is given in the online Supplementary Table S2 (feed) and Table S3 (fish)). Oleic acid (18 : 1n-9) and linoleic acid (18 : 2n-6) were the most abundant fatty acids in both feed groups. Feeding the high-ARA diet, which consisted of relatively lower amounts of rape-seed and flax-seed oil, compared with the control feed (control 6·8 %; high ARA 2·4 %), resulted in lower levels of α-linolenic acid Statistically different mean values between the control and the high-ARA group were determined using unpaired t test using GraphPad Prism. * P <0·05, ** P <0·01, *** P <0·001. † Mean is calculated for three biological replicates consisting of twenty pooled 44 DPF zebrafish/replicate. ‡ Data are expressed as the mean of two technical replicates.
(ALA; 18 : 3n-3) in the high-ARA group compared with the control group. Although EPA and DHA (22 : 6n-3) levels were balanced in the two feeds, there was a difference in EPA and DHA levels between the control and the high-ARA group. High dietary ARA levels contributed to an altered n-3:n-6 PUFA ratio of 0·6 in the control group compared with 0·2 in the high-ARA group, indicating higher n-6 PUFA levels and lower n-3 PUFA levels in the high-ARA group.
General functional characterisation of detected metabolites from metabolic profiling In total, 153 out of 566 detectable compounds from metabolic profiling were statistically different between the feed groups, using Welch's two-sample t test (P < 0·05; q < 0·05; online Supplementary Table S1). Principal component analysis shows a clear separation of the samples according to the feed groups (online Supplementary Fig. S1). The majority of metabolites affected by ARA belong to the lipid and amino acid main pathway. The lipid main pathway shows the highest count (65 %) for statistically significant metabolites among all significantly different biochemicals ( Fig. 2(a)). The high-ARA group yielded a shift in lipid metabolism, especially in PUFA, complex lipids and ARA-derived eicosanoids. Moreover, a total of six sub-pathways related to lipid metabolism were enriched in the high-ARA group (Fig. 2(b)). The entire set of enriched sub-pathways is given in the online Supplementary  Table S4.
High dietary arachidonic acid affected redox environment, vitamins, cofactors, carbohydrate and energy metabolism ARA-derived endocannabinoids and monoacylglycerols like 2-arachidonoyl-glycerol and its inactive analogue 1-arachidonoylglycerol showed statistically higher levels in the high-ARA group.  Moreover, N-stearoyl-taurine and N-palmitoyl-taurine, known as aminoacyl-endocannabinoids, had higher levels in response to higher n-6 PUFA levels as well. 7-hydroxy-cholesterol as well as oxidised products of amino acids and peptides like methionine-sulfoxide and N-acetyl-methionine-sulfoxide (oxidised products of methionine), cysteine s-sulfate (oxidised product of cysteine), cysteine-glutathione-disulfide and 4hydroxy-nonenal (4-HNE)-glutathione (oxidised products of glutathione) were elevated in the high-ARA group. Cystine, the disulfide form of cysteine, was increased in the high-ARA group compared with the control group. Cysteine for glutathione synthesis was significantly decreased in the high-ARA group, whereas both GSH and GSSG decreased slightly (not significant; 0·05 < P < 0·1) in the high-ARA group. In addition, ascorbate (ascorbic acid) and its oxidised derivatives like threonate and oxalate were increased in the high-ARA group. Urate levels were three times decreased and carnosine levels were 2·5 times decreased in the high-ARA group. δ-tocopherol levels were decreased, whereas α-/β-/γ-tocopherol levels were unaffected in the high-ARA group. Concurrently, pyridoxate and pyridoxamine showed lower levels in the high-ARA group, whereas pyridoxal and pyridoxamine phosphate were not different. Central metabolites related to glycolysis and gluconeogenesis (glucose-6-phosphate) and the pentose phosphate way (ribose-5-phosphate) showed lower levels in the high-ARA group. Concerning the TCA cycle, α-ketoglutarate and succinylcarnitine were significantly decreased, and malate was slightly (not significant; P < 0·1) decreased in the high-ARA group.

Discussion
Metabolic profiling of zebrafish fed high-ARA levels revealed changes in complex lipids, fatty acid metabolism and immune-related eicosanoids, and suggests a challenged redox environment (Fig. 5).
ω-oxidation may become more important (50) . Little research has been done on the biological significance of fatty acid ω-oxidation as described by Miura (51) . Intermediates formed by ω-oxidation, such as dicarboxylic acids, accumulated in the high-ARA group, suggesting an overwhelmed β-oxidation. Furthermore, increased ketogenesis (3-hydroxybutyrate) and decreased carnitine and acetylcarnitine levels, as observed in the high-ARA group, are indicative of a challenged β-oxidation as well.
High dietary ARA levels affected fatty acid metabolism by favouring n-6 PUFA, but our results also highlight changes in complex lipid profiles (glycerophospholipids, plasmalogens and lysophospholipids). These changes are particularly characterised by increased levels of arachidonoyl-containing complex lipids following the high dietary ARA intake and suggesting subsequent incorporation of dietary fatty acids into complex lipids (52) . Changing the fatty acid composition of phospholipids can impact membrane and immune cell function in the fish. Composition of phospholipids can affect their affinity as substrates for enzymes which generate signalling molecules, and could contribute to the alteration of immune cell responsiveness as suggested by Calder & Grimble (53) .
Eicosanoids ARA can be metabolised by a variety of enzymes resulting in a complex mixture of biologically active derivatives (eicosanoids) with distinct functions (24) . The properties and precise roles of those eicosanoids are not fully understood in mammals, and even less is known for fish. In the present study, high dietary ARA gave rise to intermediate lipoxygenase products such as 5-HETE, 5-KETE and 12-HETE in the fish. 5-HETE can reversibly be converted to 5-KETE through an oxidation reaction, and oxidative stress could increase the conversion to 5-KETE (24,54) . ARA-derived eicosanoids (HETE) show a diversity in biological functions under different physiological and pathophysiological conditions (24) . As for PG and leukotrienes, HETE has also been described in context of inflammatory responses (15) . Several studies have shown that a disproportionally high intake of n-6 PUFA can promote inflammation, resulting in pathophysiological effects in humans (13,14,24) . However, it is suggested that n-6 PUFA and their metabolites are involved in both pro-and anti-inflammatory signalling pathways in mammals (15,16) . In the present study, abundant ARA could have contributed to the promotion of metabolically triggered inflammation, as suggested by others (55)(56)(57) . Particularly, altered fatty acid profiles can change eicosanoid production and can thereby impact physiological functions by altering the range of inflammatory and immune cell responses (53,58) .

Oxidative and antioxidative response
An inflammatory environment is often associated with antioxidative events as a consequence of a changing oxidised environment. In the high-ARA group, we observed distinct changes in the redox environment ( Fig. 6) compared with the control group. Oxidised products of lipids (7-hydroxycholesterol), amino acids and peptides (cis-4-decenoylcarnitine, methionine-sulfoxide, N-acetylmethionine-sulfoxide, cysteine-S-sulfate, cysteine-glutathione-disulfide, 4-HNE-glutathione) were increased. Especially, increasing levels of cystine and decreasing levels of cysteine reflect changes in the oxidationreduction state in the high-ARA group. Furthermore, increased methionine-sulfoxide levels resulting from methionine oxidation can have profound functional consequences for target proteins, especially when signalling protein residues are affected (63) . These results indicate that increasing dietary levels of n-6 PUFA resulted in changes in oxidising conditions in the fish. Oxidative stress can potentiate the possibility of systemic inflammation (7) , which in turn affects the susceptibility for inflammation-underlying diseases (5,6) .
Formation of these oxidised products triggered an antioxidant demand to prevent increasing oxidation as indicated in the high-ARA group. Interestingly, some metabolites with known antioxidative properties such as glutathione, carnosine, δ-tocopherol and urate (73) were decreased, whereas ascorbate (74) was increased in the high-ARA group. Zebrafish, like humans, depend on dietary uptake of ascorbate (75,76) , which suggests an increased uptake from the feed rather than an increased endogenous synthesis of ascorbate in the high-ARA group. At the same time, decreased glutathione levels are consistent with an increasing demand for 4-HNE detoxification through glutathione. In cell-culture studies, increased conversion of ARA into HETE showed an increase in oxygen free radicals accompanied by glutathione depletion (GSH) that was leading to cellular damage (77) . Taken together, increasing n-6 PUFA availability in high-ARA fish led to an antioxidative response due to an enrichment of several oxidative products originating from enzymatic and non-enzymatic oxidation, indicating increasing lipid-peroxidation.

Zebrafish growth
We observed a slight weight difference in juvenile zebrafish (only 44 DPF) between the feed groups. Developmental processes during larval and juvenile stages are especially sensitive to a dietary imbalance (78) . Higher susceptibility to dietary imbalances during larval and early juvenile stages might have contributed to the growth effect we observed in 44 DPF fish that disappeared later at the adult stage. Although we do not know the mechanisms behind the growth recovery, compensatory growth has been demonstrated in fish following food deprivation (79) . Lie et al. (23) and de Vrieze et al. (22) observed no growth effect on cod larvae and zebrafish, respectively, in response to high dietary ARA levels. Boglino et al. (46) showed that both too-high and too-low n-6 PUFA levels did affect the growth of Senegalese sole larvae. Meinelt et al. (42) showed a positive correlation of higher dietary n-6 PUFA levels with growth in zebrafish, just as other animal and human studies show an association between higher n-6 PUFA intake and weight gain (46,80) . Different nutritional composition of the diets, like single fatty acid balance, magnitude of the dietary n-3:n-6 PUFA ratio, minerals and vitamins, might explain the difference in growth.

Conclusions
We have shown that high dietary ARA levels dramatically affect n-3:n-6 PUFA profiles, especially ARA-derived eicosanoids, which can greatly impact physiologic outcomes in the fish. Lipid peroxidation and an oxidised and pro-inflammatory environment, as implicated by our results, may result from both high n-6 PUFA availability and a shift in ARA-derived pro-and anti-inflammatory eicosanoids in zebrafish. The effect was characterised not only by a general change in lipid profiles and eicosanoids, but also by changed metabolites, indicating lipid peroxidation, oxidation of amino acids and changes in antioxidant status. However, the link between a low dietary n-3:n-6 PUFA ratio, elevated eicosanoid and endocannabinoid levels, and the regulation of the redox and immune system needs to be further studied, which is required to elucidate the underlying mechanisms. To our knowledge, the present study is the first to use metabolomics to reveal the metabolic fingerprint of high dietary ARA levels in teleosts. Previous studies using metabolic profiling have focused on the involvement of ARA pathways in vascular endothelial cells and CVD (81)(82)(83) . We find that zebrafish can be a useful vertebrate model to study the impact of nutrients on the manifoldness of the metabolic fingerprint. ; , Statistically significant (P < 0·05) lower and higher metabolite levels in the high-ARA group compared with the control group. ; , Lower and higher metabolites levels, which narrowly missed the statistical cut-off point for significance (0·05 < P <0·1) in the high-ARA group.
Our results from juvenile zebrafish highlight the metabolic fingerprint shaped by a specific diet.