Differential and shared effects of eicosapentaenoic acid and docosahexaenoic acid on serum metabolome in subjects with chronic inflammation

The omega-3 fatty acids eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) affect cell function and metabolism, but the differential effects of EPA and DHA are not known. In a randomized, controlled, double-blind, crossover study, we assessed the effects of 10-week supplementation with EPA-only and DHA-only (3 g/d), relative to a 4-week lead-in phase of high oleic acid sunflower oil (3 g/day, defined as baseline), on fasting serum metabolites in 21 subjects (9 men and 12 post-menopausal women) with chronic inflammation and some characteristics of metabolic syndrome. Relative to baseline, EPA significantly lowered the tricarboxylic acid (TCA) cycle intermediates fumarate and α-ketoglutarate and increased glucuronate, UDP-glucuronate, and non-esterified DHA. DHA significantly lowered the TCA cycle intermediates pyruvate, citrate, isocitrate, fumarate, α-ketoglutarate, and malate, and increased succinate and glucuronate. Pathway analysis showed that both EPA and DHA significantly affected the TCA cycle, the interconversion of pentose and glucuronate, and alanine, and aspartate and glutamate pathways (FDR < 0.05) and that DHA had a significantly greater effect on the TCA cycle than EPA. Our results indicate that EPA and DHA exhibit both common and differential effects on cell metabolism in subjects with chronic inflammation and some key aspects of metabolic syndrome.

. Characteristics of subjects at screening visit. Values are reported as mean ± SD. *P < 0.05, women versus men. BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TG, triglyceride.
All (n = 21) Male (n = 9) Female (n = 12)  Table 2. Serum concentrations of biochemical parameters at baseline and after EPA and DHA supplementation 1 . 1 Values are reported as unadjusted mean ± SD. To correct for skewed distribution, logtransformation was applied before analysis. 2 Comparison of changes from baseline to EPA or DHA was assessed using a liner mixed-effects model.FDR-adjusted P values; *FDR < 0.05. 3 Pairwise comparison between EPA and DHA (changes form baseline) were conducted using the lsmeansLT function from the lmerTest package for the linear mixed effect model. 4 Creatinine (p = 0.0231) and uric acid (p = 0.04) showed a significant sequence effect; therefore, only data from the first period was used to run ANCOVA analysis, with baseline value as a covariate. SGPT, serum glutamic-pyruvic transaminase; SGOT, serum glutamic-oxaloacetic transaminase; LDH, lactate dehydrogenase; ALP, alkaline phosphatase; BUN, blood urea nitrogen; NEFA, nonesterified fatty acids. www.nature.com/scientificreports/ Effects of EPA and DHA on serum primary metabolites. The PLS-DA plot of metabolites at baseline and after EPA and DHA supplementation is shown in Fig. 1. Relative to baseline, EPA supplementation significantly altered 11 known and 15 unknown serum primary metabolites (Table 3). EPA significantly increased nonesterified DHA and the glucose metabolites UDP-glucuronic acid and glucuronic acid. EPA lowered the tricarboxylic acid (TCA) cycle intermediate metabolites α-ketoglutarate and fumarate, the saturated fatty acids stearic acid and palmitic acid, and the leucine metabolite ketoisocaproic acid, in addition to parabanic acid (Table 3 and Fig. 2). The change in fumarate was inversely associated with the change in p-tolyl-glucuronide (Fig. 3). Relative to baseline, DHA significantly affected 19 known and 22 unknown metabolites. DHA increased glucuronic acid, the TCA cycle metabolite succinic acid, and deoxytetronic acid and reduced the amino acids alanine, valine and cysteine, the TCA intermediate metabolites pyruvate, α-ketoglutarate, citric acid, isocitrate, fumarate and malate, and the saturated fatty acids myristic acid, palmitic acid and stearic acid, in addition to phenylethylamine (Table 3 and Fig. 2). During DHA supplementation, the reduction in citrate was significantly associated with the reduction in myristic, palmitic and heptadecanoic acids and isocitric acid (Fig. 3). Nonesterified DHA concentrations did not change significantly after DHA supplementation (FDR = 0.07).
To gain insight into the unique pathways associated with the changes in primary metabolites induced by EPA and DHA supplementation, pathway analysis was performed. Compared with baseline, EPA supplementation significantly affected six pathways including: ascorbate and aldarate metabolism; pentose and glucuronate interconversions; TCA cycle; alanine, aspartate and glutamate metabolism; tyrosine metabolism; and amino sugar and nucleotide sugar metabolism (Table 4). DHA supplementation was significantly associated with four pathways, including: ascorbate and aldarate metabolism; TCA cycle; pentose and glucuronate interconversions; and alanine, aspartate and glutamate metabolism ( Table 4).
The TCA cycle was differentially affected by EPA and DHA supplementation, with a greater reduction of citrate by DHA and of fumarate by EPA (Table 4, Fig. 4).
There was no significant pathway associated with the changes in biogenic amines during EPA and DHA supplementation.

Discussion
At baseline, participating subjects had low-grade inflammation and some characteristics of metabolic syndrome. They presented with high AA and low EPA and DHA in plasma phospholipids, in agreement with their inflammatory status 12 . Supplementation with both EPA and DHA led to significant changes in the concentrations of several serum metabolites that are tightly associated with various biological and metabolic pathways, with DHA affecting a greater number of metabolites than EPA.
In the context of no significant change in total serum non-esterified FA (NEFA) by EPA and DHA, both n-3 FA lowered the non-esterified saturated myristic, palmitic and stearic acids but only EPA supplementation significantly increased non-esterified DHA. In the fasting state, serum NEFA are mostly derived from adipocyte TG lipolysis 13 . In TG, DHA is preferentially found at the sn-2 position, while saturated fatty acids can be found randomly at any position. Adipocyte TG lipase (ATGL), which catalyzes the first step of adipocyte lipolysis, preferentially cleaves fatty acids at the sn-2 position of TG 14 and its activity has been shown to be regulated by n-3 FA 15 . The increase in serum non-esterified DHA observed after EPA supplementation, but not after DHA Table 3. Peak height values of serum primary metabolites at baseline and after EPA and DHA supplementation (alphabetic order) 1 . 1 Values are unadjusted and reported as median (25th, 75th). To correct for skewed distribution, generalized log-transformation and auto-scaling were applied before analysis. 2 Comparison of changes from baseline to EPA or DHA was assessed using a liner mixed-effects model. FDRadjusted P value: * < 0.05, ** < 0.01, *** < 0.001. 3 Pairwise comparison between EPA and DHA (changes form baseline) were conducted using the lsmeansLT function from the lmerTest package for the linear mixed effect model. 4 Fumaric acid showed significant sequence effect (p = 0.0073). When there was a sequence effect, data from the first period is extracted to run ANCOVA analysis, with baseline value and age as covariates. EPA supplementation reduced level of fumaric acid (p = 0.0107) when sequence effect was taken into account. www.nature.com/scientificreports/ supplementation, may therefore signal a greater activation of ATGL by EPA. Alternatively, other enzymes such as phospholipase may also contribute to this effect. N-3 FA are activators of peroxisome proliferator-activated receptor γ (PPARγ), a transcriptional factor upregulating key genes involved in adipocyte lipogenesis and TG storage 16 . The reduction in non-esterified saturated FA after EPA and DHA supplementation, even in the context of no significant change in total NEFA, may in part be explained by adipocyte PPARy activation 17,18 . In addition, since after DHA supplementation the reduction in myristic, palmitic and heptadecanoic acids was associated with the reduction in the TCA metabolite citrate, improved metabolic health due to greater mitochondrial efficiency and adipose tissue insulin sensitivity may explain our findings.
The TCA cycle, operating in the mitochondrial matrix, plays a critical role in the integration of several anabolic and catabolic pathways 19 . It serves as the primary source for the production of ATP, mostly through its linkage to oxidative phosphorylation 20,21 . In mice, excessive caloric intake or feeding a high-fat diet was associated with elevated TCA cycle activity, paralleled by mitochondrial dysfunction and increased oxidative stress and inflammation 22 . Our data show that 10-week supplementation with both EPA and DHA resulted in the reduction of two key TCA cycle intermediates, α-ketoglutarate and fumarate, and that DHA supplementation led to a  www.nature.com/scientificreports/ further significant reduction of four TCA cycle intermediates, including pyruvate, citrate, isocitrate and malate. Considering that our subjects were overweight or obese and displayed several characteristics of metabolic syndrome, the reduction in TCA metabolites during n-3 FA supplementation may indicate a greater efficiency of the TCA cycle, possibly due to improvements in mitochondrial function or insulin sensitivity. DHA is incorporated into the inner and outer mitochondrial membranes and thus may affect mitochondrial membrane organization and functions, including membrane potential, respiration and reactive oxygen species (ROS) production 23 . DHA also decreases membrane viscosity and potentially alter lipid-lipid and lipid-protein interactions 23,24 . The effect of DHA on membrane viscosity is greater than that of EPA 25 . Furthermore, as PPAR activators, EPA and DHA may regulate mitochondrial function through regulation of mitochondrial gene expression: PPARγ activation increases mitochondrial DNA copy number, mitochondrial biogenesis and the expression of genes in the fatty acid β-oxidation pathway 26 . The greater reduction in citrate by DHA than EPA is in accordance with a previous animal study which showed that DHA supplementation, rather than EPA, significantly lowered hepatic citrate compared to a control Western diet 27 .
In contrast to the other TCA cycle metabolites, succinate was increased after DHA, but not EPA, supplementation. Recent evidence indicates that succinate is not only a metabolite in the TCA cycle but also a gut microbiota metabolite 28 . Succinate is involved in multiple pathways and has been linked to both health and a range of pathologic conditions 28 . Through binding to a plasma membrane G-protein coupled receptor on macrophages 29 , succinate has been shown to initiate the process of active resolution of acute inflammation in the context of obesity 30 . Whether the upregulation of succinate by DHA is caused by succinate-producing bacteria or changes in the host metabolism is not known.
New roles of TCA cycle intermediates in cellular signaling and activation of immune cells have been documented 31,32 . Acetyl-CoA, isocitrate and citrate can alter both innate and adaptive immune responses and www.nature.com/scientificreports/ stem cell function, and α-ketoglutarate was shown to repress pro-inflammatory responses while improving the anti-inflammatory profile in macrophages 32 . In contrast, succinate and fumarate may work as oncometabolites that promote tumorigenesis through the suppression of histone and DNA demethylation 32 . Given that our data show significant changes in these metabolites, more studies are needed to further elucidate their systemic effects on non-metabolic signaling and their association with diseases. Glucuronate was significantly affected by EPA and DHA supplementation in relation to both the ascorbate and aldarate metabolism and the pentose and glucuronate interconversion pathway. EPA demonstrated a higher impact by increasing both glucuronate and UDP-glucuronate, while DHA only raised glucuronate. These metabolites are generated from glucose through a series of enzymatic reactions 33 . Conjugation with glucuronic acid is the most common phase-II reaction, with UDP-glucuronosyltransferases catalyzing the formation of glucuronides from a large variety of xenobiotics and endogenous substrates like bilirubin, making them more hydrophilic and easily excreted by the kidneys 34 . At low doses, DHA has been shown to reduce the activity of the UDP-glucuronosyltransferase A1 in newborn mice, leading to increased blood bilirubin levels 35 . At higher doses, however, DHA rather activated the enzyme, possibly through PPAR induction 35 . In our study, bilirubin levels were not affected by either DHA or EPA supplementation. Therefore, the increase in UDP-glucuronate and glucuronate levels by n-3 FA may be due to increased synthesis from glucose.
In the subset of 10 participants who had serum biogenic amines assessed, several metabolites were differently affected by EPA and DHA supplementation. EPA significantly reduced alanine levels, possibly due to reduced muscle protein breakdown. Creatinine levels, a marker of kidney function, was also reduced, but reductions in serum creatinine may also indicate reduced protein breakdown.
DHA supplementation changed the concentration of a greater number of metabolites than EPA. Specifically, DHA reduced plasma levels of short-chain acylcarnitines, including acetylcarnitine, propionylcarnitine and hydroxyisovalerylcarnitine. Patients with insulin resistance and type 2 diabetes tend to present with elevated levels of these acylcarnitines, which are considered a marker of inefficient β-oxidation and of mitochondrial dysfunction 36 . Reduction in acetylcarnitine following n-3 FA supplementation has been previously reported 37 . Studies have demonstrated that acetyl-CoA accumulates in conditions of TCA cycle overwork, leading to increased formation of acetylcarnitine 38 . In addition, propionylcarnitine and hydroxyisovalerylcarnitine are metabolites of amino acid metabolism and their reduction may indicate reduced protein breakdown. Taken together, our findings suggest improved β-oxidation, and mitochondrial function, and reduced protein www.nature.com/scientificreports/ breakdown by DHA. On the other hand, the finding of increased octanoylcreatinine by DHA remains unexplained in the context of the reduction of the other acylcarnitines. DHA supplementation was also associated with increased levels of glycine and serine. In the serine synthesis pathway, glucose is converted to 3P-pyruvate, and, with glutamate, 3P-pyruvate is converted to serine and then glycine 39 . In endothelial cells, this pathway leads to the synthesis of heme in the mitochondria and results in improved mitochondrial function and reduced oxidative stress 39 .
In conclusion, we observed significant and shared effects of EPA and DHA on metabolites related to lipid and adipocyte metabolism, TCA cycle efficiency, mitochondrial function and detoxification processes in subjects with chronic inflammation and some key aspects of metabolic syndrome. However, EPA and DHA also exhibited differential effects on the serum metabolome as evidenced by a greater number of metabolites affected by DHA than EPA and a significantly greater effect on selected metabolites by DHA or EPA.
Strengths of our study include the randomized, controlled, crossover study design, the high compliance of participants, and the assessment of the metabolic effects of EPA and DHA effects in subjects with chronic inflammation, a population at higher risk of cardiometabolic diseases. Limitations are the small sample size and the lack of a healthy control group.

Methods
Study design. The details of the study design and related primary outcomes have been published previously 11 . The study had a randomized, double-blind, controlled, crossover design and consisted of an initial 4-week lead-in phase during which the subjects were given 3 g/day of high-oleic acid sunflower oil as a control supplement, followed by randomization to a 10-week supplementation with either EPA or DHA (3 g/day) and crossover after a 10-week washout phase. The random allocation sequence was generated by REDCap. Subjects, study investigators and laboratory personnel were blind to treatment. The end of the 4-week lead-in phase was defined as baseline. Ethyl ester forms of EPA and DHA were provided in capsules containing 750 mg/capsule with a purity > 97% (Prevention Pharmaceuticals Inc, CT). Each participant was instructed to take a total of four capsules daily, two with the morning meal, and two with the evening meal. Subjects' compliance was assessed by the number of returned capsules at the end of each phase and was greater than 80% in all participants. All research was performed in accordance with the Declaration of Helsinki.

Study subjects. Subjects were studied at the Jean Mayer USDA Human Nutrition Research Center on
Aging at Tufts University. Between April 2016 and November 2017, 24 participants were enrolled and 21subjects (9 men and 12 postmenopausal women) completed the study. Three subjects dropped out of the study during the placebo lead-in phase. Subjects, aged 50-75 years and meeting the following inclusion criteria, completed the study: (1) serum C-reactive protein (CRP) concentrations ≥ 2 gµ/mL; (2) fasting plasma TG concentrations between 90 and 500 mg/dL; and (3) at least one of the criteria for the modified definition of metabolic syndrome including abdominal obesity (waist circumference ≥ 102 cm for men and ≥ 89 cm for women), hypertension (blood pressure ≥ 130/80 mmHg or use of antihypertensive medications), and fasting blood glucose concentrations 100-125 mg/dL 11 . Subjects were excluded if they had consumed either fish more than 2 times/week or supplements containing fish oil or EPA/DHA during the 6 months prior to the trial; were allergic to sardines and/ or sunflower oil; had regularly used anti-inflammatory medications such as NSAID or were under anticoagulant therapy; had other known diseases that might alter their metabolic status such as kidney or liver diseases; were alcohol users (> 7 drinks/week) or smokers. Subjects were instructed to follow a low-saturated fat diet (25-35% of calories from total fat, < 7% from saturated fat, and cholesterol < 200 mg/day) and keep their physical activity levels unchanged throughout the study. Subjects were also asked not to consume more than 2 fish meals/week and to abstain from taking fish oil supplements during the study. To enhance adherence to dietary and lifestyle instructions, consultations with a registered dietitian were carried out at baseline and at each study visit. The study protocol was approved by the Tufts Health Sciences Institutional Review Board and is registered at clinicialtrials.gov (NCT02670382, 01/02/2016). All study participants provided informed consent.
Blood sample collection. Twelve-hour fasting venous blood samples were collected into silica-coated tubes (Becton Dickinson, NJ) at baseline and the end of each supplementation phase. After being kept at room temperature for 20 min to allow coagulation to occur, blood was centrifuged (1,000 g for 15 min at 4 °C) and serum was aliquoted and immediately stored at − 80 °C for further analysis.
All methods used in the current study for the assessment of biochemical parameters and metabolites have been previously described and have been carried out in accordance with relevant guidelines and regulations. The effects of EPA and DHA on the serum metabolome was intended as an exploratory analysis.
Measurement of biochemical parameters. Serum concentrations of TG, glucose, albumin, and other blood chemistry analytes were assessed at the screening visit and at the end of each phase using commercial reagents on an AU480 Chemistry Analyzer (Beckman Coulter Inc, CA) 11 .
Metabolic profiling. Serum primary metabolites. Primary metabolites, including a panel of carbohydrates and sugar phosphates, amino acids, hydroxyl acids, free fatty acids, purines, pyrimidines, aromatics, and exposome-derived chemicals, were measured in fasting serum samples, obtained at baseline and at the end of each supplementation phase from all 21 subjects, by gas chromatography-time of flight mass spectrometry (GC-TOF MS) by West Coast Metabolomics 40,41 . Briefly, serum samples were extracted and derivatized using acetonitrile:isopropanol:water (3∶3∶2; v/v/v), centrifuged and dried. A set of 13 C8-C30 fatty acid methyl ester internal standards were added, and samples were derivatized by 10 µL methoxyamine hydrochloride in pyridine  41 , combined with the validity of chromatogram (< 10 peaks with intensity > 10^7 counts s-1), detection of unbiased retention index marker (MS similarity > 800 with high m/z marker ions) and retention index calculation by 5 th order polynomial regression, as previously described 40

Data analysis
Results are presented as mean and standard deviation (SD) or median (25 th and 75 th percentile). Metabolite data were subjected to generalized log transformation and auto-scaling before statistical analyses due to skewness. Statistical analyses, including partial least square discriminant analysis (PLS-DA), were conducted in R version 4.0.0 (R core team, 2020). Comparisons of the change in serum metabolites from baseline to EPA versus DHA supplementation were evaluated using a linear mixed-effects model, by the lmer function from the lme4 package. The model accounted for the random effect of subjects and the fixed effects of treatment, period, and sequence. Baseline values and age were included in the model as covariates. The effect of EPA or DHA supplementation versus baseline was assessed by using the lsmeansLT function from the lmerTest package, which allows for comparisons of least square means between baseline and each supplemental phase in the linear mixed model. When there was a significant sequence effect, data from only the first supplementation period was selected to perform ANCOVA for the comparisons between EPA and DHA and paired t-tests for the comparisons between baseline versus each treatment, with the inclusion of baseline concentration as a covariate. No sex differences were observed in the changes of metabolites following EPA and DHA supplementation, therefore men and women were combined in our analyses. To correct for multiple testing, false discovery rate (FDR) was used to determine adjusted p-values. Statistical significance was defined as FDR < 0.05. MetaboAnalyst 4.0 (https:// www. metab oanal yst. ca), a web-based platform for comprehensive analysis of quantitative metabolomic data, was used to perform pathway analysis and visualize the biological pathways associated with significantly affected metabolites. The pathway analysis algorithms are based on the hypergeometric test for over-representation analysis, and relative-betweenness centrality for pathway topology analysis. Hypergeometric test compares the number of significant metabolites within a specific pathway with the expected value 43 . Results included the impact value of the metabolic pathway based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) for Homo sapiens database that were most relevant to the selected metabolites 44 . Pathways with an impact score > 0 and FDR < 0.05 were considered statistically significant.
To identify metabolites with similar or opposite changes after EPA and DHA supplementation, Pearson's correlation coefficient analysis was conducted. Significantly changed metabolites from the linear-mixed model were chosen and the coefficients were calculated across metabolite pairs. In addition, the change in concentration of each metabolite was correlated with the change in plasma phospholipid EPA or DHA mol percentage. The correlation coefficients (r) of metabolites within each feature, which reflect the strength of their relationships, were ranked from high to low using the pattern searching function. Heatmaps were created to provide an overview and visualization of the correlation matrix. An absolute value of r that is greater than 0.6 indicated a moderate to strong linear relationship, and P < 0.05 was considered statically significant.