Comparative transcriptomic and metabolomic analysis of fenofibrate and fish oil treatments in mice

Yingchang Lu, Mark V. Boekschoten, Suzan Wopereis, Michael Müller, Sander Kersten


Elevated circulating triglycerides, which are considered a risk factor for cardiovascular disease, can be targeted by treatment with fenofibrate or fish oil. To gain insight into underlying mechanisms, we carried out a comparative transcriptomics and metabolomics analysis of the effect of 2 wk treatment with fenofibrate and fish oil in mice. Plasma triglycerides were significantly decreased by fenofibrate (−49.1%) and fish oil (−21.8%), whereas plasma cholesterol was increased by fenofibrate (+29.9%) and decreased by fish oil (−32.8%). Levels of various phospholipid species were specifically decreased by fish oil, while levels of Krebs cycle intermediates were increased specifically by fenofibrate. Plasma levels of many amino acids were altered by fenofibrate and to a lesser extent by fish oil. Both fenofibrate and fish oil upregulated genes involved in fatty acid metabolism and downregulated genes involved in blood coagulation and fibrinolysis. Significant overlap in gene regulation by fenofibrate and fish oil was observed, reflecting their property as high or low affinity agonist for peroxisome proliferator-activated receptor-α, respectively. Fenofibrate specifically downregulated genes involved in complement cascade and inflammatory response. Fish oil specifically downregulated genes involved in cholesterol and fatty acid biosynthesis and upregulated genes involved in amino acid and arachidonic acid metabolism. Taken together, the data indicate that despite being similarly potent toward modulating plasma free fatty acids, cholesterol, and triglyceride levels, fish oil causes modest changes in gene expression likely via activation of multiple mechanistic pathways, whereas fenofibrate causes pronounced gene expression changes via a single pathway, reflecting the key difference between nutritional and pharmacological intervention.

  • peroxisome proliferator-activated receptor
  • liver
  • lipid metabolism
  • complement and coagulation cascades

one of the components of the metabolic syndrome and frequent complication of insulin resistance is hypertriglyceridemia (7, 14). Similar to elevated plasma cholesterol levels, hypertriglyceridemia is considered an independent risk factor for atherosclerosis (7, 14, 52). Elevation of circulating triglycerides (TG) in insulin resistance is related to the diverse actions of insulin on hepatic secretion of very low density lipoprotein particles (1).

Apart from weight loss and increased physical activity, both of which improve insulin sensitivity, a limited number of therapeutic options exist to lower circulating TG (15, 29, 56). A class of drugs that effectively lowers plasma TG are the fibrates. Reduction in plasma TG by fibrate treatment depends on baseline TG levels but can reach 40% (15). Fibrates also raise plasma levels of HDL-cholesterol, making it an attractive drug for the treatment of diabetic dyslipidemia, which is characterized by elevated TG and low HDL levels. Overall results of clinical trials, however, are mixed and not overwhelmingly in support of a broad application of fibrates to correct dyslipidemia and lower risk of coronary heart disease. The decrease in plasma TG by fibrates has been attributed to inhibition of synthesis and secretion of TG by the liver and stimulation of degradation of TG-rich lipoproteins (29, 56). Fibrates stimulate a large panel of genes involved in hepatic mitochondrial and peroxisomal fatty acid oxidation and lipoprotein metabolism by direct high-affinity binding and consequent activation of peroxisome proliferator-activated receptor alpha (PPARα) (13, 27, 29, 44, 56), thus serving as a direct pharmacological ligand of PPARα. PPARα is a transcription factor abundant in liver that mediates the adaptive response to fasting.

Besides via pharmacological approach, plasma TG can be effectively lowered via nutritional intervention in the form of fatty fish or fish oil. Fish oil, rich in eicosapentaenoic acid (EPA, C20:5n-3) and docosahexaenoic acid (DHA, C22:6n-3), has long been considered as potential treatment to lower risk of coronary heart disease (12, 18, 19, 38). Similar to fibrates, fatty acids present in fish oil induce hepatic expression of numerous genes via activation of PPARα (51). In addition, EPA and DHA suppress activity of the prolipogenic transcription factor sterol regulatory element binding protein-1 (SREBP-1) by inhibiting proteolytic processing of SREBP-1, a process required to generate the active mature SREBP-1 protein (24, 25). While fenofibrate and fish oil thus both lower plasma TG and can activate the same transcription factor, a comparative analysis of the effects of fenofibrate and fish oil at transcriptome and metabolome level has yet to be performed. Therefore, to gain further insight into mechanisms underlying the effects of fenofibrate and fish oil on cardiovascular risk factors and to investigate whether these mechanisms are shared between fenofibrate and fish oil, we performed hepatic transcriptional profiling and plasma metabolite profiling in mice treated with fenofibrate or fish oil for 2 wk. Inasmuch as fenofibrate and fish oil are already known to both stimulate PPARα-dependent gene regulation, the focus of the analysis was on genes and pathways uniquely regulated by either fenofibrate or fish oil.


Animals and experimental design.

We housed 12-wk-old male C57BL/6 mice (Charles River, L'Arbresle Cedex, France) two per cage in a light-and-temperature controlled facility (lights on 6:30 to 18:30, 21°C) and acclimated them for 3 wk. The mice were randomized by weight-matching into three groups (n = 10 in control group and n = 12 in fenofibrate or fish oil intervention group) and fed a research diet (no. D10012M, Research Diets, NJ; Supplementary Table S1) supplemented with sunflower oil (containing 81.3% oleic acid, 7% energy intake) in control group, sunflower oil (7% energy intake) and fenofibrate (0.03% wt/wt) in fenofibrate group, and fish oil (Marinol C-38 fish oil: 23.1% EPA and 21.1% DHA, 7% energy intake) in fish oil group for 2 wk.1 Mice received fresh diet every 3rd day, and food consumption rate and body weight gain were monitored. At the end of treatment, mice were fasted from 7:00 to 13:00 with drinking water available and were subsequently killed by cervical dislocation under isoflurane anesthesia. Blood was collected via orbital puncture. Livers were dissected, directly frozen in liquid nitrogen and stored at −80°C until further analysis. Blood was centrifuged (4,000 g for 10 min at 4°C), and plasma was stored at −80°C. The animal experiments were approved by the animal ethics committee of Wageningen University.

Affymetrix GeneChip microarray analysis.

Total RNA was prepared from mouse livers using TRIzol reagent (Invitrogen, Breda, The Netherlands), treated with DNase, and purified on columns using the RNeasy Mini Kit (Qiagen, Venlo, The Netherlands) following the supplier's protocol. RNA concentrations were measured by absorbency at 260 nm, and the quality and integrity were verified with the RNA 6000 Nano assay on the Agilent 2100 Bioanalyzer (Agilent Technologies, Amsterdam, the Netherlands) according to the manufacturer's instructions. Microarray analysis was performed on individual mouse livers. Five micrograms of RNA were labeled using the Affymetrix One-Cycle Target Labeling Assay kit (Affymetrix, Santa Clara, CA). Hybridization, washing, and scanning of Affymetrix Mouse Genome 430 2.0 Arrays were done according to standard Affymetrix protocols. Scans of the Affymetrix arrays were processed using packages from the Bioconductor project (16). Raw signal intensities were normalized by using the GCRMA algorithm (69). Probe-sets were defined according to Dai et al. (10) using remapped chip definition file (CDF) version 11.0.2 based on the Entrez gene database. The Affymetrix Mouse Genome 430 2.0 Arrays target 16,331 unique genes based on this CDF. Genes were filtered on expression value >20 in five samples, resulting in a set of 7,400 expressed genes. The Bioconductor R package Linear models for microarray data (LIMMA) was used to identify differentially expressed genes. To balance between unspecific responses from the two treatments and relative weak transcriptional effects by fish oil, the genes that met the cut-off of mean absolute fold change >1.2 and false discovery rate corrected q-value <0.05 (55) were considered to be significantly regulated. Among the significantly regulated genes, only those that are associated with a canonical pathway in the Ingenuity Pathway Knowledge Base were considered for Ingenuity Pathway Analysis (IPA; Ingenuity Systems, Redwood City, CA). In addition, all genes represented on the array were also considered for the unbiased Gene Set Enrichment Analysis (GSEA) (58). This analysis was run using 1,000 permutations per gene set. All microarray data are MIAME compliant and have been submitted to the Gene Expression Omnibus (accession number GSE 32706).

Lipidomic and metabolic profiling.

The liquid chromatography-mass spectrometry (LC-MS) methods for measuring plasma lipids and nonesterified free fatty acids (NEFAs) and the gas chromatography (GC)-MS method for measuring a broad range of metabolites were identical to the methods reported by Wopereis et al. (68). The samples were analyzed in randomized order. Data for each sample were corrected for the recovery of the internal standard for injection. The performance of the methods was carefully monitored by using multiple internal standards (5–10 depending on the method, including analogs and 2H- and 13C-labeled metabolites) as described previously (68). Furthermore, a quality control sample prepared by pooling of plasma from all samples was analyzed after every 10th study sample. Batches were only accepted if the relative standard deviation (RSD) of the peak area ratio for all internal standards was <20%. Metabolites were only accepted if the RSD was <20%, unless large differences between treatment groups were observed. Batch to batch differences in data were removed by synchronizing medians of quality control samples per batch. Metabolites were annotated by using an in-house metabolite database containing retention time information, MS spectra (electron impact ionization for GC-MS data), MS/MS spectra (LC-MS), and accurate mass data (LC-MS) of reference substances. The confidence of identification was 100% unless indicated otherwise. Accurate MS and MS/MS data of reference substances and metabolites in the study samples were acquired by using Thermo LTQ-FT and Thermo LTQ-Orbitrap instruments (Thermo Fisher Scientific, Waltham, MA). Finally, the LC-MS NEFA dataset contained 22 free fatty acids; the LC-MS lipid dataset contained 184 lipid metabolites; and the GC-MS dataset contained 137 metabolites. Also, 41 different metabolite ratios and sums were calculated. Detailed information on these metabolites could be acquired upon request.

Statistical analysis.

Results are reported as means ± SE. The comparison of different groups was carried out using ANOVA and unpaired two-tailed Student's t-test. The Kruskal-Wallis test or Mann-Whitney U-test was used if groups did not show equal variance. Spearman rank correlation was used to correlate hepatic gene expression signals with plasma metabolite levels. Differences were considered statistically significant when P < 0.05.


Changes in liver and body weight and selected plasma metabolites.

Mice received fenofibrate (0.03% wt/wt) or fish oil (3% wt/wt) in their feed for 2 wk. Neither fenofibrate nor fish oil influenced food intake (Table 1). Fish oil but not fenofibrate significantly increased bodyweight (+6.2% vs. +2.7%), while fenofibrate but not fish oil increased liver weight (+1.8% vs. −0.1%, respectively). Fenofibrate raised plasma total cholesterol (+29.9%), whereas fish oil had the opposite effect (−32.8%). Both fenofibrate and fish oil reduced plasma triglycerides although the effect of fenofibrate was more pronounced (−49.1% vs. −21.8%). Consistent with stimulation of hepatic fatty acid oxidation, fenofibrate markedly raised plasma ketone bodies (+316%). A similar effect was observed for ribose. Interestingly, fish oil significantly reduced plasma levels of phospholipids (−42.3% for combined phosphatidylcholine and phosphatidylethanolamine), lysophospholipids (−28.0% for combined lysophosphatidylcholine and lysophosphatidylethanolamine), and sphingomyelins (−19.4%), which was not observed for fenofibrate.

View this table:
Table 1.

General characteristics and selected plasma metabolite levels

Fenofibrate reduced plasma total NEFA levels (−18.8%), with its effect equally distributed among the various fatty acid classes (SFA, MUFA, PUFA). Fish oil lowered total plasma NEFA to a similar extent as fenofibrate but its effect across the various fatty acid classes was different. Fish oil decreased levels of all n-6 and n-9 long-chain fatty acids and, as expected, increased plasma levels of n-3 PUFA levels (Table 2). The stronger reduction in MUFA by fish oil compared with fenofibrate was mainly attributable to reduced levels of C18:1 fatty acids and presumably due to lower content of oleic acid in the feed of fish oil treated mice (Table 2 and Supplementary Table S1).

View this table:
Table 2.

Plasma nonesterified fatty acid levels

Fenofibrate and fish oil treatments lead to PPARα activation.

Since fenofibrate and EPA/DHA are known ligands of PPARα, we determined to what extent PPARα was activated by the two treatments. Fenofibrate and fish oil caused upregulation of many genes involved in PPARα-dependent pathways such as hepatic fatty acid uptake (Fig. 1A), mitochondrial fatty acid β-oxidation (Fig. 1B), peroxisomal fatty acid β-oxidation (Fig. 1C), microsomal fatty acid ω-hydroxylation (Fig. 1D), and ketogenesis (Fig. 1E), including many classical PPARα target genes (50). Consistent with fenofibrate being a higher affinity PPARα agonist compared with EPA/DHA, more pronounced inductions were observed in fenofibrate treated mice. These data demonstrate that fenofibrate and fish oil treatment led to activation of PPARα, in line with published data (36).

Fig. 1.

Heat map showing the parallel induction of specific proliferator-activated receptor (PPAR)α-dependent gene sets in liver of fenofibrate- and fish oil-treated mice. A: genes involved in fatty acid uptake, activation, and binding. B: genes involved in mitochondrial fatty acid β-oxidation/oxidative phosphorylation. C: genes involved in peroxisomal fatty acid β-oxidation. D: genes involved in microsomal fatty acid ω-hydroxylation. E: genes involved in ketogenesis. The expression levels in the control (Con) mice were set at 1 (black) and expression levels in fenofibrate (Fen)- and fish oil (Fis)-treated groups were calculated relative to the control group. The definition of the gene sets were based on Ref. 50.

Comparative analysis of gene regulation by fenofibrate and fish oil.

Dendrogram of hierarchical clustering of the various mice based on the complete liver transcriptome showed that fenofibrate-treated mice formed a highly distinct group from fish oil-treated and control mice, illustrating the much more pronounced effects of fenofibrate on hepatic gene expression compared with fish oil (Fig. 2). Furthermore, since the three groups clustered separately, the analysis indicated that the variability in gene expression within the three groups was much less compared with the variability between the groups. To compare the whole hepatic genome effects of fenofibrate and fish oil, we carried out scatter plot analysis (Fig. 3). The similarity in gene regulation between fenofibrate- and fish oil-treated mice was relatively small and was mainly observed with respect to upregulation of gene expression. Fold inductions of gene expression were generally higher in fenofibrate- compared with fish oil-treated mice. Interestingly, some genes strongly upregulated by fenofibrate were downregulated by fish oil, including Ly6d, Cidec, Pdk4, Defb1, Ucp2, Cidea, and Pltp. Other genes were upregulated by fish oil but not by fenofibrate, including Mt2, Derl3, and Agxt2l1, or even strongly downregulated by fenofibrate, such as Clec2h, Aox3, Cyp2c37, and Hsd3b5 (Fig. 3). As fenofibrate and fish oil both activate PPARα, differential hepatic gene regulation by fish oil compared with fenofibrate suggest another type of regulatory mechanism by fish oil treatment not involving PPARα.

Fig. 2.

Hierarchical clustering of microarray data from mice treated with fenofibrate, fish oil, or receiving control treatment. Dendrogram shows separate clustering of mice treated with fenofibrate compared with fish oil- or control-treated mice, illustrating the much more pronounced effects of fenofibrate on hepatic gene expression compared with fish oil.

Fig. 3.

Similarities and differences in gene expression in livers between fenofibrate- and fish oil-treated mice. Scatter plot shows fold-change in gene expression after treatment with fenofibrate (x-axis) plotted against fold-change in gene expression after treatment with fish oil (y-axis). Selected genes that were upregulated or downregulated specifically by fenofibrate or fish oil are indicated.

Pathway analysis of microarray data.

To gain insight into pathways uniquely regulated by fenofibrate and fish oil, GSEA was performed. Partial overlap in the upregulated gene sets was observed between fenofibrate and fish oil, mostly covering various aspects of fatty acid metabolism (Fig. 4). Many gene sets were specifically induced by fenofibrate, including electron transport chain and TCA cycle (Fig. 4A). Also, a small number of gene sets was upregulated by fenofibrate but decreased by fish oil, including lipogenesis, glycolysis, and gluconeogenesis, and the pentose phosphate pathway. Interestingly, autophagy, ABC transporters, and arachidonic acid metabolism were specifically induced by fish oil (Fig. 4B). In terms of downregulation, blood clotting cascades, complement cascades, and antigen processing and presentation were commonly regulated by fenofibrate and fish oil. Gene sets that were exclusively downregulated by fish oil included lipoprotein metabolism, cholesterol synthesis and esterification, and prostaglandin synthesis regulation. Taken together, several pathways specifically regulated by either fenofibrate or fish oil could be identified. Expectedly, most of the common regulation by fenofibrate and fish oil relates to PPARα-dependent gene sets connected to fatty acid catabolism.

Fig. 4.

Overrepresented pathways in the liver after fenofibrate and fish oil treatments identified by Gene Set Enrichment Analysis. A: ranking based on the normalized enrichment score (NES) of pathways regulated by fenofibrate (black bars), with pathways regulated by fish oil shown in parallel (gray bars). Pathways with FDR q-value < 0.2 are shown. The NES reflects the degree to which a gene set in certain pathway is overrepresented at the top (upregulated) or bottom (downregulated) of the ranked gene list and is corrected for gene set size. Sources of the gene sets consist of BioCarta, GenMAPP, KEGG, Sigma-Aldrich pathway, and Signal Transduction Knowledge Environment. B: ranking based on the normalized enrichment score of pathways regulated by fish oil (gray bars), with pathways regulated by fenofibrate shown in parallel (black bars). For additional information, see supplementary note.

One of the gene sets specifically induced by fenofibrate corresponded to TCA cycle. In agreement with this finding, fenofibrate but not fish oil significantly raised plasma levels of several TCA cycle intermediates, including isocitrate, α-ketoglutarate, succinate, fumarate, and malate (Table 3). In accordance with altered hepatic amino acid metabolic pathways upon fenofibrate and fish oil treatment, significant changes in plasma amino acids levels were observed in both treatment groups (Supplementary Table S3). Most assayed amino acids were increased by fenofibrate, including glutamate, glycine, isoleucine, leucine, phenylalanine, serine, tryptophan, tyrosine, and valine, whereas only glycine was increased by fish oil (Supplementary Table S3). These data correspond well with downregulation of genes involved in urea cycle and metabolism of amino groups in fenofibrate treated mice (Fig. 4A) (Supplementary Table S2).

View this table:
Table 3.

Plasma metabolite (organic acids) involved in TCA cycle

Genes specifically regulated by fenofibrate or fish oil treatment.

Venn diagrams were created to identify genes that were specifically regulated by fenofibrate or fish oil treatment (Fig. 5, A and B). Most genes upregulated or downregulated by fish oil were similarly regulated by fenofibrate. To better characterize fenofibrate- and fish oil-specific gene regulation, we compiled the top 40 of genes specifically upregulated (Fig. 5, C and D) and downregulated (Fig. 5, E and F) by fenofibrate and fish oil, respectively, and ranked according to fold change. Changes in gene expression by fish oil and fenofibrate are shown in parallel.

Fig. 5.

Nonoverlapping genes between fenofibrate and fish oil treatment. Venn diagrams show the overlap in significantly upregulated (A) and downregulated (B) genes after treatment with fenofibrate or fish oil for 2 wk. Genes were included if mean fold-change (MFC) by fenofibrate or fish oil exceeded the value of 1.2 and Bayesian corrected q was <0.05. C: top 40 genes most strongly upregulated (based on MFC) by fenofibrate but not fish oil. Known PPARα target genes upregulated by fish oil by >1.2-fold yet for which the effect failed to meet statistical significance due to large variation were removed to exclusively highlight fenofibrate-specific genes (Mogat1, Acot3, Cd36, Cpt1b, Slc16a13, Acot4, Crat, and Slc27a4). D: top 40 genes most strongly upregulated by fish oil but not fenofibrate. E: top 40 genes most strongly downregulated by fenofibrate but not fish oil. F: top 40 genes most strongly downregulated by fish oil but not fenofibrate. Expression levels in the control mice (Con) were set at 1 (black), and expression levels in 2 treated groups were calculated relative to the control group. The list was sorted based on fold-change of fenofibrate (Fen: C and E)- or fish oil (Fish: D and F)-treated mice.

A relatively large proportion of top 40 genes specifically upregulated by fenofibrate were known PPARα target genes (50). In contrast to the classical PPARα target genes involved in fatty acid catabolism shown in Fig. 1, these genes, which included Cidec, Pdk4, Ucp2, Cidea, Pltp, Abcd2, Me1, and Fabp4, were downregulated by fish oil (Fig. 5C). Among the top 40 genes specifically downregulated by fenofibrate treatment, several genes are involved in complement cascade (C8b, C6, and C9), coagulation cascade (Serpine2 and F11) and general inflammatory regulation and response (Orm2, Bcl6, Saa1, Saa2, Ifit1, Il6ra, Il1r1, and Irf5) (Fig. 5E), which is consistent with the strongly downregulated acute phase response signaling (IPA, data not shown), complement cascade and coagulation cascade in fenofibrate treated mice (Fig. 4A and Supplementary Table S2). Also, a number of fenofibrate-specific genes were related to steroid metabolism, which fits with reduction in androgen and estrogen metabolism as indicated by GSEA (Fig. 4A). With respect to fish oil treatment, several of the top 40 genes specifically upregulated by fish oil are involved in amino acid metabolism (Agxt2l1, Aox3, Clpx, Cyp7b1, and Aadat) and arachidonic acid metabolism (Cyp2c37, Cyp2c44, and Cyp2j9) (Fig. 5D). Most of them were downregulated by fenofibrate, suggesting that the underlying mechanism of regulation is fish oil specific and not PPARα related. Among the top 40 genes specifically downregulated by fish oil, a marked enrichment of SREBP target genes was apparent (21, 43) (Fig. 5F and Supplementary Table S2), which is consistent with the observed downregulation in both cholesterol and fatty acid biosynthetic pathways (Fig. 4B). Strikingly, expression of several SREBP1 targets involved in lipogenesis (Fasn, Elovl6, Fads2, and Fads1) was upregulated by fenofibrate, but downregulated by fish oil (Fig. 5F). The hepatic down- and upregulation of SREBP1 target stearoyl-CoA desaturase (Scd1, Supplementary Table S2) by fish oil and fenofibrate was substantiated by corresponding changes in plasma Δ9-desaturation indexes (stearoyl-CoA desaturase activity) (Table 4).

View this table:
Table 4.

Liver Δ9-desaturation indexes (Stearoyl-CoA desaturase activity)

The precise molecular mechanism behind the plasma TG lowering effect of fish oil remains controversial. To gain insight into potential mechanisms, we determined which genes showed the highest correlation with plasma TG levels in the combined control and fish oil-treated mice (Fig. 6A). Remarkably, many genes showing a highly significant negative correlation with plasma TG were PPARα target genes involved in fatty acid metabolism. In comparison, lipogenic and cholesterol biosynthetic genes, despite being strongly downregulated by fish oil, showed weaker correlation with plasma TG. Similar observations were made for correlations of gene expressions with plasma cholesterol levels and phosphatidylcholine (Fig. 6A). IPA on genes showing the highest positive or negative correlation with plasma TG, cholesterol, and phosphatidylcholine identified “fatty acid metabolism,” a PPARα regulated pathway, as the most significant pathway (data not shown). Pathways related to fatty acid or cholesterol biosynthesis were much less significant. Specific examples of genes showing a highly significant correlation with plasma TG, cholesterol, or lysophosphatidylcholine are shown in Fig. 6B. These results may imply that the effects of fish oil on various plasma lipids may occur via changes in fatty acid oxidation via activation of PPARα, and to a lesser extent via suppression of fatty acid and cholesterol biosynthesis, which are under control of SREBP.

Fig. 6.

Correlation analysis between plasma metabolites and hepatic gene expression signals in control and fish oil-treated mice. Spearman rank correlation analyses were conducted between plasma metabolites and hepatic gene expression signals for those genes that were statistically significant regulated after fish oil treatment (q < 0.05). A: the respective top 20 positive (pos, r > 0.77) and negative (neg, r < −0.81) correlated genes with plasma triglyceride (TG), cholesterol, lysophosphatidylcholine (LPC), and phosphatidylcholine levels. The genes are either PPARα regulated and involved in fatty acid metabolism (neg side, in circle), or SREBP regulated and involved in lipogenesis and cholesterol biosynthesis (pos side, in square). B: scatter-plots between plasma TG, cholesterol, and lysophosphatidylcholine (LPC) levels and hepatic expression level of top correlated genes. AU, arbitrary units.


Using a combination of transcriptional and plasma metabolite profiling, we carried out a comprehensive comparison of the effects of fenofibrate and fish oil treatment in mice. Due to the treatment duration, some of the observed changes in gene expression changes will be secondary to metabolic perturbations elicited by the treatment, although a major portion of gene expression changes likely reflects direct regulation. Both fenofibrate and fish oil induced numerous genes involved in hepatic fatty acid catabolism and other PPARα-dependent pathways. Fenofibrate consistently caused higher fold-inductions, in agreement with fenofibrate being a better PPARα agonist compared with EPA and DHA (51). In contrast to fish oil, fenofibrate caused hepatomegaly and raised plasma ketone bodies, which are known PPARα-dependent effects. The data thus indicate that compared with fenofibrate, fish oil treatment leads to modest activation of PPARα in liver. However, while effects of fenofibrate are almost entirely mediated by PPARα, fish oil additionally acts upon other regulatory pathways to exert multiple effects, reflecting a property characteristic of nutrients.

Consumption of fish oil lowers circulating TG in humans (19), which could be reproduced in mice. Fish oil fatty acids may stimulate postendoplasmic reticulum (post-ER) presecretory proteolysis of ApoB via a mechanism dependent on fatty acid peroxidation, leading to reduced TG secretion by hepatocytes (47). Alternatively, lowering of circulating TG may occur via the observed downregulation of fatty acid synthesis. In addition, a number of other mechanisms has been proposed, including activation of PPARγ in adipose tissue (12, 18). Conflicting data exist on whether lowering of plasma TG and cholesterol by fish oil is dependent on PPARα (11, 59). We observed pronounced enrichment for PPARα targets and genes involved in fatty acid metabolism among genes showing the most significant negative correlation with plasma TG, cholesterol, and phospholipids. In comparison, lipogenic and cholesterol biosynthetic genes showed good correlation with plasma phospholipids but weaker correlation with plasma TG and cholesterol. Our results may suggest that the effect of fish oil on plasma TG and cholesterol primarily occurs via activation of PPARα, whereas the effect on plasma phospholipids seems to rely proportionally more on suppression of SREBP-dependent regulation of lipogenesis and cholesterol metabolism.

In contrast to earlier studies in rats but in line with the effect in humans (13), fenofibrate increased plasma cholesterol, which in mice is almost exclusively carried in HDL. The reason for the discrepancy with previous rat studies is unclear (42, 57). Besides via changes in apoAI expression, which was slightly but significantly reduced by both fenofibrate and fish oil, the HDL-raising effect of fenofibrate in humans may be mediated via changes in Pltp expression and activity (33, 66). Alternatively, fenofibrate may raise plasma cholesterol levels in mice by downregulating Scarb1 (SR-BI), which was confirmed in our study (35).

In line with numerous papers, fish oil downregulated expression of numerous SREBP target genes involved in fatty acid and cholesterol synthesis (24, 25, 59, 60). Although SREBP1 and SREBP2 have both been suggested to be inhibited by PUFAs, data implicating SREBP1 in downregulation of gene expression by PUFAs are much more plentiful. Recently, the target of PUFAs was identified as Ubxd8, an ER membrane-bound protein that facilitates the degradation of Insig-1, thereby promoting proteolytic processing and activation of SREBP-1 (30). It was shown that PUFAs inhibit the activity of Ubxd8 (30). Hence, downregulation of hepatic fatty acid and cholesterol biosynthetic genes by fish oil may occur via inhibition of Ubxd8.

While several classical PPARα targets involved in fatty acid oxidation were upregulated by both fenofibrate and fish oil, other established PPARα targets were induced only by fenofibrate, including Cidec, Cidea, Pdk4, Ucp2, Pltp, Abcd2, Me1, and Fabp4. With the exception of Abcd2, none of these genes are involved in fatty acid oxidation but instead participate in other (lipid) metabolic pathways. Interestingly, malic enzyme (Me1) is controlled by both PPARα and SREBP, as are Ucp2 (37, 50), Fads1, Fads2, Scd1, Acsl3, Acsl4, and Acsl5 (21, 50), Abcd2 (50, 67), and Fabp4 (26, 50). Similarly, the PPARα targets Pltp and lipid droplet-associated protein Cidea were shown to be regulated by SREBP1 (43, 64). Thus, it is possible that the above gene set is induced by fish oil via PPARα but that the effect is counterbalanced by suppression of SREBP-mediated transcriptional regulation. These data indicate that responsiveness to synthetic PPARα agonist may not always properly predict regulation by dietary PPARα agonists, as the latter act via multiple mechanistic pathways that may converge on a single gene.

Data abound indicating that fish oil fatty acid and fibrates stimulate hepatic fatty acid uptake and catabolism (5), which likely explains the reduced plasma NEFA levels in fenofibrate and fish oil treated mice. Although fish oil caused much weaker induction of genes involved in fatty acid catabolic pathways compared with fenofibrate, the reduction in plasma NEFA was identical between the treatments. Fish oil may reduce plasma NEFA levels and increase weight gain by attenuating fatty acid release from adipose tissues (18, 23, 45, 48), perhaps via activation of PPARγ. EPA and DHA present in fish oil are endogenous PPARγ ligands (9, 12, 61), and fish oil has been shown to upregulate PPARγ and its responsive genes in epididymal adipose tissue (8, 9, 17, 22, 39, 61). Fish oil may thus mimic the stimulatory effect of synthetic PPARγ agonists on fatty acid trapping (9, 12, 61), and weight gain (3). Since we did not collect adipose tissue, it is impossible to determine whether fish oil induced PPARγ target genes in adipose tissue and whether it increased adipose tissue mass.

Remarkably, pathways relevant to blood coagulation and fibrinolysis were strongly downregulated by fenofibrate and fish oil. Individual genes within these pathways were consistently downregulated by fenofibrate and fish oil, although the effect of fenofibrate was much more pronounced (Supplementary Table S2). Suppression of hepatic expression of these genes upon pharmacological PPARα activation has been reported in rats and monkeys (6, 28) and was shown to be PPARα dependent for fibrinogen α-, β-, and γ-chain (28). In humans, fibrates reduce plasma fibrinogen levels by 12–25% (65, 70). Recently, several genes in this pathway were found to be downregulated in human hepatocytes by synthetic PPARα agonist (49).

A weak hypocoagulant effect of fish oil has been observed in studies in humans and is likely mediated by suppression of clotting factors. Fish oil was shown to decrease hepatic transcription of kallikrein B, fibrinogen β-chain, antithrombin III, and protein C genes (60, 63) and lower blood fibrinogen, factors II, V, VII, and X, antithrombin III, and protein C levels in rodents (2, 32, 40, 41, 63). Reduced activity of factor V, VII, VIII, IX, X, XI, and XII and protein C (20, 34) and reduced plasma levels of fibrinogen, factor V, VII, and X, protein C, antithrombin III, plasminogen activator inhibitor, and α2 antiplasmin were also observed in human fish oil intervention studies (4, 20, 34, 46, 54, 62). The striking parallel downregulation of blood coagulation and fibrinolysis pathways by fenofibrate and fish oil suggest they occur via a common mediator, e.g., PPARα.

Interestingly, consistent with previous data showing enhanced TCA cycle flux (53) and increased TCA cycle enzymes (31) by synthetic PPARα ligand, plasma levels of all TCA cycle intermediates was induced by fenofibrate, perhaps as a result of enhanced amino acid degradation by fenofibrate, as revealed by GSEA.

In conclusion, the transcriptomic and metabolomic effects of fish oil and fenofibrate reflect the highly specific activity of fenofibrate toward PPARα, whereas fish oil engages additional regulatory pathways to impact numerous biological processes. Our data provide better insight into how fish oil modulates circulating levels of lipids and other metabolites in humans. Moreover, the data may provide clues toward additional potential health benefits of fish oil.


This study was financed by the Netherlands Nutrigenomics Centre and the Netherlands Heart Foundation (2006B195).


No conflicts of interest, financial or otherwise, are declared by the author(s).


Author contributions: Y.L., M.V.B., S.W., and S.K. analyzed data; Y.L., M.V.B., S.W., M.M., and S.K. interpreted results of experiments; Y.L., M.V.B., and S.K. prepared figures; Y.L. and S.K. drafted manuscript; Y.L., M.V.B., S.W., M.M., and S.K. edited and revised manuscript; M.V.B., M.M., and S.K. conception and design of research; M.V.B. and S.W. performed experiments; M.M. and S.K. approved final version of manuscript.


We thank Rob Vreeken of the Netherlands Metabolomics Centre for the lipidomics measurements and Ivana Bobeldijk for GC-MS and free fatty acids measurements. Coauthors are members of NuGO association.


  • § M. Müller and S. Kersten are cosenior authors.

  • 1 The online version of this article contains supplemental material.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
View Abstract