Rosiglitazone and Fenofibrate Exacerbate Liver Steatosis in a Mouse

Jan 30, 2014 - The homeostatic model assessment index (HOMA-IR) was calculated as an ..... Our study is novel in that using an integrated transcriptom...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/jpr

Rosiglitazone and Fenofibrate Exacerbate Liver Steatosis in a Mouse Model of Obesity and Hyperlipidemia. A Transcriptomic and Metabolomic Study Anna Rull,†,‡ Benjamine Geeraert,†,§ Gerard Aragonès,‡ Raúl Beltrán-Debón,‡ Esther Rodríguez-Gallego,‡ Anabel García-Heredia,‡ Juan Pedro-Botet,¶ Jorge Joven,‡,∥ Paul Holvoet,§,∥ and Jordi Camps*,‡,∥ ‡

Unitat de Recerca Biomèdica (CRB-URB), Hospital Universitari de Sant Joan, Institut d’Investigació Sanitària Pere Virgili (IISPV), Universitat Rovira i Virgili, Reus, Spain § Atherosclerosis and Metabolism Unit, Department of Cardiovascular Sciences, Katholieke Universiteit Leuven, Leuven, Belgium ¶ Department of Internal Medicine, Hospital del Mar, Barcelona, Spain S Supporting Information *

ABSTRACT: Peroxisome proliferator-activated receptors (PPAR) play an important role in the regulation of lipid and glucose metabolism, inflammatory, and vascular responses. We show the effect of treatment with two PPAR agonists, fenofibrate (FF) and rosiglitazone (RSG), on ob/ob and LDLR-double deficient mice, by combined gene-expression and metabolomic analyses. Male mice were daily treated for 12 weeks with RSG (10 mg·kg1−·day−1 per os (p.o.), n = 8) and FF (50 mg·kg1−· day−1 p.o., n = 8). Twelve untreated ob/ob and LDLR-double deficient mice were used as controls. To integrate the transcriptomic and metabolomic results, we designed a hierarchical algorithm, based on the average linkage method in clustering. Data were also interpreted with the Ingenuity Pathway Analysis program. FF and RSG treatments significantly increased the hepatic triglyceride content in the liver when compared with the control group, and the treatments induced an increase in the number and size of hepatic lipid droplets. Both drugs simultaneously activate pro-steatotic and antisteatotic metabolic pathways with a well-ordered result of aggravation of the hepatic lipid accumulation. The present study is a cautionary note not only to researchers on the basic mechanism of the action of PPAR activators but also to the use of these compounds in clinical practice. KEYWORDS: fenofibrate, metabolomics, nonalcoholic fatty liver disease, rosiglitazone, steatosis, transcriptomics



INTRODUCTION

early detection of the disease and the assessment and monitoring of response to treatment thereof. In addition, pathogenic mechanisms are not completely understood.7−12 No proven treatment for the overall management of obesity and the associated metabolic syndrome has yet emerged other than the obvious weight loss through diet and exercise or bariatric surgery in morbidly obese patients. Insulin sensitizers, lipidlowering agents, antioxidants, cytoprotective agents, and novel drugs acting as selective and dual PPAR agonists are under intense and close examination.13−16 However, probably as the consequence of the lack of adequate animal models, data on the long-term safety and efficacy and their impact on histological outcomes of affected tissues are difficult to acquire in humans. We show here the effect of treatment with two PPAR agonists, fenofibrate and rosiglitazone, on our experimental model. These agonists are known to have distinct effects on the components of the metabolic syndrome and in several biochemical pathways in adipose and cardiac tissues.17,18 We reasoned that an approach combining gene-expression and

Peroxisome proliferator-activated receptors (PPAR) play an important role in the regulation of fatty acid oxidation, lipoprotein metabolism, inflammatory, and vascular responses.1,2 Pharmacological activation of PPAR represents a potential therapeutic approach. Adverse events are probably due to the fact that PPAR regulate many aspects of energy metabolism at the transcriptional level and to their pleiotropic and systemic effects, suggesting the need for further development of drugs with tissue-selective mechanism of action.1,3 To facilitate identification of obesity-associated diseases, we propose a mice model that combines obesity, hyperlipidemia, and insulin resistance.4,5 These mice spontaneously develop liver steatosis without the use of predisposing diets, representing a model in which the storage capacity of the adipose tissue is exceeded and ectopic lipid deposition begins in other organs. This is particularly important, because nonalcoholic fatty liver disease (NAFLD) is strongly linked to obesity and associated metabolic syndrome and cardiovascular disease in humans.6,7 The contribution of NAFLD to these derangements is generally neglected, probably because there is not a clinically efficient, sensitive diagnostic procedure for the © 2014 American Chemical Society

Received: December 12, 2013 Published: January 30, 2014 1731

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

10% neutral-buffered formalin for histology. A portion of the caudate lobe of the liver was weighed, frozen in liquid nitrogen, homogenized in ice-cold PBS (5 mL PBS per 1 g of tissue) and centrifuged at 13 000g for 10 min at 4 °C for specific protein measurements. Other portions of the same lobe livers were snap frozen, stained with oil red O to visualize lipids, and their cholesterol and triglyceride content was quantified after extraction with isopropyl alcohol−hexane. For histopathological evaluation, a semiquantitative histological scoring system was used.22 For each variable and individual mouse, 30 fields from each of three different sections were analyzed. Immunhistochemical studies were performed using specific antibodies (Santa Cruz Biotechnology, Inc., Santa cruz, CA). Quantitative measurements were performed by using the image analysis software, AnaliSYS (Soft Imaging System, Münster, Germany), as described.23

metabolomic analyses might be a novel and useful tool to evaluate the effects of these compounds. We found that PPAR agonists improved insulin sensitivity and prevented loss of ventricular function in obese dyslipidemic mice, similarly to weight loss.19,20 Our data indicated that treatment with PPAR agonists improved insulin sensitivity by improving adipose tissue differentiation associated with improved glucose uptake and lipid metabolism. However, the effects of these treatments on liver metabolism are unknown. Therefore, the aim of our study was to investigate the effect of PPAR agonists administration on liver histology and metabolism in a model of obese and hyperlipidemic mice. Our results add a note of caution in the evaluation of such treatments and potential mechanisms of obesity-related metabolic disorders and NAFLD.



MATERIALS AND METHODS

Quantitative Real-Time PCR and Transcriptomic Profiling

Total RNA was isolated from liver tissue using the automated QIAcube system with spin-column kits from QIAgen (Izasa, Barcelona, Spain). TaqMan primers and probes were obtained from validated Assays-on-Demand products (Applied Biosystems, Foster City, CA) (Supplementary Table 1, Supporting Information) and used in real-time PCR (rt-PCR) amplifications on the 7900HT Fast Real-Time PCR system (Applied Biosystems). The level of RNA expression was calculated using the threshold cycle (Ct) value and normalized with the appropriate housekeeping gene according to the SABiosciences qRT-PCR data analysis software.24

Experimental Design

Experimental procedures in animals were performed in accordance with protocols approved by the Animal Care and Research Advisory Committees of KU Leuven. Double deficient ob/ob and low-density lipoprotein receptor (LDLR) mice were obtained at KU Leuven as previously described, on a C57BL6 background.4 Animals were housed at 22 °C on a fixed 12/12 h light−dark cycle and were fed regular chow diet throughout the experiment. Food and water were available ad libitum, and intake was similar in all treatment groups. Male mice were daily treated as described19 for 12 weeks starting at the age of 12 weeks, with the study compounds rosiglitazone (RSG, 10 mg·kg1−·day−1 per os (p.o.), n = 8) or fenofibrate (FF, 50 mg·kg1−·day−1 p.o., n = 8). Untreated, age-matched mice (control group) (n = 12) were used to assess the effects of treatment. Obese mice were compared with age-matched lean C57BL6 mice (wild-type) to evaluate the biochemical and liver histological alterations produced by the combination of both obesity, hyperlipidemia, and insulin resistance. The animals were allocated to experimental groups by computer-generated randomization schedules and investigators responsible for the assessment of outcomes had no knowledge of the experimental group to which the animals belonged. Treatments were prepared and labeled by an independent investigator according to the randomization schedule to ensure allocation concealment. There were no animals excluded from analysis.

Metabolomic Profiling

Portions of the left and right lobes of the liver (n = 6 for each group) were sent to Metabolon, Inc. (Research Triangle Park, Durham, NC), extracted, and prepared for analysis using described methods.25 Small molecule metabolites from an equivalent amount of liver tissue were extracted with methanol, and the resulting extract was divided into equal fractions for analysis by ultra-high-performance liquid chromatographytandem mass spectrometry (UPLC-MS/MS; separately under positive mode and negative mode) and gas chromatography− mass spectrometry (GC-MS). Metabolites were identified by comparison of ion data to a reference library of chemical standards (∼ 2800) entries that included retention time, mass (m/z), and MS or MS/MS spectra. Statistical Analyses

Biochemical Measurements

Data were initially analyzed using ANOVA (single-factor or two factors) when necessary. Differences between any two groups were assessed with the Mann−Whitney U-test. Spearman correlation coefficients were used to evaluate the degree of association between variables. The SPSS/PC + 18.0 (SPSS, Chicago, IL) was employed for these purposes. Areas under the curve were calculated using the Graph Pad Prism version 5 program. Comparisons in metabolomic profile were performed with Welch’s t-tests and/or Wilcoxon’s rank sum tests as well as ANOVA for repeated measures. To correct for multiple testing we used the false discovery rate estimated using the q-value as described.26 The level of significance was set at p < 0.05.

EDTA-blood was obtained for all analyses. Total cholesterol and triglycerides were measured with standard enzymatic assays (Boehringer, Mannheim, Germany), glucose with a glucometer (Menarini Diagnostics, Zaventem, Belgium), and insulin with a mouse ELISA (Mercodia, Oxon, U.K.). The homeostatic model assessment index (HOMA-IR) was calculated as an estimate of insulin resistance.21 Glucose tolerance was determined by the intraperitoneal (i.p.) glucose tolerance test (IPGTT). Glucose was measured in samples obtained by tail bleeding before and 15, 30, 60, 120, and 240 min after i.p. glucose administration (20% glucose solution; 2 g·kg1−). Adiponectin, interleukin-6 (IL6), and tumor necrosis factor-alpha (TNFα) were measured with specific mouse ELISA (R&D Systems, Uppsala, Sweden).

Functional Association Analyses

To interpret the results of the transcriptomic and metabolomic analyses, we designed a hierarchical algorithm, based on the average linkage method in clustering.27 This algorithm was used to arrange genes and metabolites according to their correlation coefficients. Results were expressed as a matrix of gene

Assessment of Liver Morphology and Immunohistochemical Analyses

Livers were perfused, removed, prepared in portions, and either flash-frozen and stored at −80 °C until used or fixed for 24 h in 1732

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

expression data (heat map) where rows represent the genes analyzed and columns represented the groups studied. The clusters were designed with the Java Treeview software.28,29 Data were also interpreted with the Ingenuity Pathway Analysis (IPA; Ingenuity Systems, Inc., Redwood City, CA; www. Ingenuity.com). Differentially expressed genes and metabolites with known gene symbol were submitted to IPA to interpret network functions, canonical signaling pathways, and toxicity functions associated to each treatment compared to the control group. Biologically relevant pathways were constructed using the IPA application.



RESULTS

Biochemical and Liver Histological Changes in Obese, Hyperlipemic, and Insulin-Resistant Mice Compared to Wild-Type Mice

At 24 weeks of age, control animals (untreated ob/ob and LDLR-double deficient mice) revealed significant changes in body weight, lipid and glucose metabolism compared to wildtype mice. There was a 2.6-fold increase on average in body weight that was accompanied by an increase in fasting serum glucose, serum cholesterol, and triglycerides. These changes were also paralleled by a significant increase in the titer of autoantibodies against malondialdehyde-modified low density lipoproteins (Anti MDA-LDL; an estimate of lipid peroxidation) compared to wild-type mice (Table 1). Histological Table 1. Body Weight and Biochemical Variables in Untreated ob/ob and LDLR-Double Deficient Mice (Control Group) and Wild-Type C57BL6 Mice wild-type body weight (g) glucose (mmol/L) AUC of IPGTTa cholesterol (mmol/L) triglycerides (mmol/L) anti MDA-modified LDLb

24.5 4.3 41.5 1.1 0.3 2.2

± ± ± ± ± ±

1.3 0.3 2.8 0.1 0.0 0.5

ob/ob LDLR-double deficient 63.7 7.2 84.1 12.1 2.4 9.4

± ± ± ± ± ±

Figure 1. Hepatic steatosis and the number and size of lipid droplet content in untreated ob/ob and LDLR-double deficient mice (control group) and wild-type C57BL6 mice.

1.4* 0.46* 7.7* 0.6* 0.2* 0.5*

Treatment with FF and RSG Exacerbates Spontaneous Liver Steatosis

FF and RSG treatments significantly increased the hepatic triglycerides content when compared with the control group. In addition, RSG but not FF increased the hepatic cholesterol content (Figure 2A,B). Histological sections of the liver reflected these biochemical changes (Figure 2C−F). Thus, FF- and RSG-treated mice presented an increase in the number and size of lipid droplets that was more evident in animals receiving RSG treatment (Figure 2G,H). Macrophage infiltration in the liver was measured by the presence of F4/80 antigen positively stained cells (Figure 2I−L). Both treatments significantly decreased the amount of macrophage content in the liver compared to control group, but no differences were observed among treatments.

Values are shown as means ± SEM aAUC of IPGTT = area under the curve of the intraperitoneal glucose tolerance test. bAnti MDAmodified LDL: titer of autoantibodies against malondialdehydemodified LDL. *p < 0.05.

analysis of the liver also revealed that the ob/ob and LDLRdouble deficient mice presented an increase in hepatic steatosis and the number and size of lipid droplet content (Figure 1). Effects of FF and RSG Treatment on Body Weight Increase, Insulin Resistance, and Selected Laboratory Variables

FF and RSG significantly reduced body weight gain at the end of the 12 week treatment, but no effect was observed on food uptake. RSG treatment decreased blood glucose, insulin, and HOMA-IR and resulted in an improved glucose tolerance (IPGTT). FF treatment had no significant effect on serum glucose levels, but there was a trend to an improvement in IPGTT, insulin, and HOMA-IR. RSG treatment significantly increased serum cholesterol concentrations, and FF increased serum triglicerydes. Anti MDA-LDL titer was significantly decreased at the end of the 12 week FF and RSG treatments. Serum adiponectin was lowered by FF, but it was notably increased by RSG. No effect was observed on serum IL-6 and TNFα concentrations in any of the two treated groups compared to the untreated group. Results are shown in Table 2.

Trancriptomic Effects of FF and RSG on Selected Genes in the Liver

To define the genomic basis involved in the aggravation of spontaneous liver steatosis caused by FF and RSG treatments, we examined the expression of selected genes involved in glucose and lipid metabolism, inflammation, and energy metabolism (Supplementary Table 2,Supporting Information). Among the 46 genes studied, only 26 showed statistically significant changes in treated groups compared to the untreated group (Figure 3A,B). To examine to what extent FF and RSG treatment influenced the same genes, we created Venn diagrams showing the overlap in genes significantly altered 1733

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Table 2. Fenofibrate and Rosiglitazone Effects on Body Weight and Blood Variables in ob/ob and LDLR-Double Deficient Mice at the End of 12 Weeks of Treatment control group body weight (g) glucose (mmol/L) insulin (mU/L) HOMA-IR AUC of IPGTTa adiponectin (g/mL) total cholesterol (mmol/L) triglycerides (mmol/L) anti MDA-LDLb IL-6 (pg/mL) TNFα (pg/mL)

63.6 7.2 1974 604.2 84064 2.9 12.1 2.3 9.4 34.4 22.7

± ± ± ± ± ± ± ± ± ± ±

fenofibrate (FF)

1.4 1.6 235 99.7 7706 0.5 0.5 0.2 0.5 16.5 3.5

58.3 7.2 1511 499.8 74240 1.1 14.2 2.5 3.7 24.6 33.5

± ± ± ± ± ± ± ± ± ± ±

1.1* 0.5 200 89.9 3889 0.1* 1.1 0.3* 0.3* 4.8 1.4

rosiglitazone (RSG) 54.9 4.9 1079 243.0 49500 13.7 16.5 2.2 3.4 18.2 37.2

± ± ± ± ± ± ± ± ± ± ±

0.9* 0.2*,† 345* 75.8*,† 1676*,† 1.7*,† 1.3* 0.4 0.5* 3.5 10.9

Values are shown as means ± SEM. aAUC of IPGTT = area under the curve of the intraperitoneal glucose tolerance test. bAnti MDA-LDL: titer of autoantibodies against malondialdehyde-modified LDL. cSMC = smooth muscle cells. *p < 0.05, with respect to the control group. †p < 0.05, with respect to FF-treated mice.

Figure 2. Treatment with fenofibrate (FF) or rosiglitazone (RSG) impairs liver steatosis but ameliorates macrophage infiltration in ob/ob and LDLRdouble deficient mice. Hepatic accumulation of triglycerides (A) and cholesterol (B) was evaluated after lipid extraction. The degree of steatosis (C) was established by histological evaluation of hematoxylin and eosin-stained liver sections (representative examples in panels D−F). FF and RSG treatment increase the size of lipid droplets (G) and the number of macrovesicular steatosis (H). Macrophage accumulation was determined by image analysis of F4/80-stained liver sections (I). Histology confirmed a decrease in macrophage infiltration in FF- and RSG-treated mice (representative examples in panels J−L). *p < 0.05 with respect to the control group; ≠ p < 0.05 with respect to FF-treated mice.

group (Figure 3D). The gene of the antioxidant enzyme paraoxonase-1 (Pon1) was down-regulated by both treatments compared to the untreated group, an effect not observed in Pon2 and Pon3 expression. Peroxysomal β-oxidation (Acox1),

upon both treatments (Figure 3C). As expected, PPARγ gene expression was significantly up-regulated by RSG treatment, and both RSG and FF treatments significantly increased PPARα and PPARδ gene expressions compared to the control 1734

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Figure 3. Effects of fenofibrate (FF) or rosiglitazone (RSG) treatment on specific genes involved in several pathways related to glucose and lipid metabolism, inflammation, and energy metabolism. Panels (A) and (B) are Volcano plots showing genomic data with statistically significant changes in FF- or RSG-treated animals compared to the control group. (C) Venn diagram showing the overlap in genes significantly altered in FF- and RSGtreated mice compared to the control group. (D) Effect of FF and RSG treatment on hepatic PPAR genes compared to control group. (E) Expression of selected genes involved in lipid transport (Fabp4), glycolysis (Slc2a2), and immune response (Emr1, Xbp1).

Differential Metabolite Expression Patterns in the Liver Suggest Diverse Effects of PPAR Activation in the Course of Liver Steatosis

lipid storage and transport (Adf p, Slc27a1), fatty acid metabolism (Cpt1a, Foxo1, and Scd1), insulin signaling (Irs1, Irs2), and immune response (mTOR, Tnfaip3) were upregulated by FF and RSG treatment. In addition to these changes shared by both RSG and FF, RSG treatment also up-regulated the expression of genes involved in triglyceride synthesis (Dgat2), gluconeogenesis (Pck1), and glycolysis (Slc2a2). Of note, RSG treatment significantly induced Fabp4 gene expression (a lipid transporter), which tended to be down-regulated by FF. Genomic analysis confirmed that both treatments down-regulated macrophage gene expression in the liver (Emr1). FF also differed from RSG in the expression of Catalase and Ppargc1α, both significantly up-regulated by FF. Summarizing the differences between RSG and FF, we observed that RSG increased lipid and glucose transport in the liver by upregulating the hepatic expression of Fabp4 and Slc2a2 and altered the immune and oxidative response, by down-regulating the expression of Emr1 and Pon1 and up-regulating the expression of Xbp1 (Figure 3E).

Our metabolomic approach clearly shows changes in relevant biochemical pathways that differ following treatment with FF or RSG (Figure 3 and Supplementary Table 3, Supporting Information). Hepatic glucose levels were significantly reduced following both treatments; 2-hydroxybutyrate levels were also decreased, more markedly in RSG-treated mice. These results are consistent with a reduction in hepatic insulin resistance. Moreover, sorbitol and fructose levels were decreased by both treatments, indicating reduced aldose reductase activity and the consequent prevention of intracellular osmotic imbalance and probable cellular damage. Glycogen (by increased formation or decreased breakdown) was increased in both treatment groups as indicated by the reduction in mannose and maltose-related metabolites. RSG-treatment decreased the activity of the pentose phosphate shunt (PPS) indicating an increased potential for elevated oxidative stress (decreasing levels of NADPH) with respect to FF-treated mice. Consistent with these changes, opthalmalate, an analog of glutathione was increased significantly in response to RSG indicating an early, 1735

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Figure 4. Differential metabolite expression patterns in fenofibrate (FF)- or rosiglitazone (RSG)-treated mice and in the control group in a number of relevant biochemical pathways: carbohydrate and glutathione metabolism (A), amino acid metabolism (B), and lipid metabolism (C).

tricarboxylic acid cycle for energy production, increased only in response to FF, which suggests a selective decrease in their breakdown and/or utilization. We also observed a significant rise in the levels of methionine, a glucogenic amino acid, and of the aromatic amino acids phenylalanine, tryptophan, and tyrosine, which can be either glucogenic or ketogenic in their degradation. This may reflect the potential capacity of FF, more than RSG, to lower hepatic gluconeogenesis and reduce the

sensitive indicator of glutathione demand and a marker of elevated oxidative stress (Figure 4A). Various aspects of amino acid metabolism were observed to be differentially modulated between both treatments (Figure 4B). Histidine levels showed a significant rise in response to FF and RSG with respect to controls indicating better glucose utilization, consistent with the observed decreases in hepatic glucose. The branched-chain amino acids (BCAA) isoleucine, leucine, and valine, which can form ketoacids able to enter the 1736

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Figure 5. (A) Hierarchical cluster analyses of both transcriptomic and metabolomic profiles. The red or green colors of increasing intensity indicate up-regulation or down-regulation, respectively. The dendrogram generated on the left of the cluster indicates the relatedness of the genes. Selected clusters (box) with the most distinct transcriptomic and metabolomic expression patterns were analyzed by IPA, showing changes highly associated to the branched-chain amino acids (BCAA) (box A), urea cycle and amino acid metabolism (box B), carbohydrate metabolism (box C), lipid metabolism and small molecule transport network function (box D and E), and citrate cycle, taurine metabolism, and BCAA biosynthesis (box F and G). Panels (B) and (C) show signaling and toxicology pathways significantly affected by fenofibrate (FF) and rosiglitazone (RSG) treatment compared to the control group and identified with the canonical pathway (B) and toxicity function (C) features of IPA.

metabolism. Additionally, RSG yielded the greatest effect on long-chain fatty acid metabolism with respect to that observed with FF (Figure 4C). Such a differential effect could imply a different capacity to form triglycerides or altered capacity for oxidation as a fuel source, although no treatment showed a significant increase in 3-hydroxybutyrate levels, a marker of elevated β-oxidation with respect to controls.

conversion of these amino acids to either glucose or fatty acid precursors. There was a significant increase in a number of lysolipids (Figure 4C), metabolites with known bioactive properties as signaling molecules, in response to both treatments, a finding that may also corroborate similar effects in the regulation of insulin sensitivity and/or resistance. Finally, RSG treatment decreased various essential fatty acids, an effect which is limited in FF to a significant reduction in linolenate and docosahexanoate acids. This may be important, because these metabolites may modulate membrane fluidity and may serve as precursors for the synthesis of molecules with intense effects on cell

Functional Interpretation of the Changes in the Transcriptomic and Metabolomic Profiles

Clustering of the most representative changes in transcriptomic and metabolomic profiles (rows) is shown in Figure 5A. The 1737

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Figure 6. Scheme shows the top differentially regulated pathways in fenofibrate (FF)-treated mice compared to the control group, identified by IPA. Genes and metabolites significantly different between both groups are shaded. The intensity of the shading shows to what degree each gene or metabolite was up-regulated (red) or down-regulated (green). Solid line represents a direct interaction, and a dotted line means indicates an indirect interaction. Icon legend is at the bottom of the figure.

patterns between groups. To interpret the network functions and the canonical signaling pathways associated to each selected cluster (box), genes and metabolites with known gene symbol were submitted to IPA (Supplementary Table 4, Supporting Information). The metabolites depicted in box A, which include the branched-chain amino acids (BCAA), were highly associated to aminoacyl-t-RNA biosynthesis and BCAA

distinctness of the clusters, which represents how different one cluster is from its closest neighbor, was represented by the distance along the horizontal axis. The dendrogram of the hierarchical clustering of all animals studied (columns) showed a good degree of similarity between the experimental groups. Hierarchical clustering of genes identified various clusters with the most distinct transcriptomic and metabolomic expression 1738

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Figure 7. Scheme shows the top differentially regulated pathways in rosiglitazone (RSG)-treated mice compared to the control group, identified by IPA. Genes and metabolites significantly different between both groups are shaded. The intensity of the shading shows to what degree each gene or metabolite was up-regulated (red) or down-regulated (green). A solid line represents a direct interaction, and a dotted line means there is an indirect interaction. Icon legend is at the bottom of the figure.

to Type II diabetes mellitus signaling. Concretely, box E showed a cluster of genes related to FXR/RXR and PXR/RXR activation, both pathways related to toxicity effects. Clusters G and F, showing similarities between FF-treated mice and the untreated group were basically related to the citrate cycle, taurine metabolism, and BCAA biosynthesis. In addition, to interpret the specific canonical signaling pathways affected and the toxicity effects produced by each

biosynthesis. Changes in box B, which revealed similar expression patterns in the FF-treated and the untreated group, were significantly associated to urea cycle and metabolism of amino groups. Box C shows a cluster of metabolites involved in carbohydrate metabolism, specifially associated to fructose/mannose metabolism. Changes in box D and E, which were strongly associated to lipid metabolism and small molecule transport network functions, were highly related 1739

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research



treatment compared to the control group, differentially expressed genes and metabolites with known gene symbol were also submitted to IPA (Supplementary Tables 5 and 6, Supporting Information). The canonical signaling pathway showed that Type II diabetes mellitus signaling, PPARα/RXRα activation, and AMPk signaling were the most affected pathways by both FF and RSG treatments (Figure 5B). As expected from metabolomic analyses, aminoacyl-t-RNA biosynthesis and BCAA biosynthesis were highly associated to FF treatment. With respect to the possible toxicological effects (Figure 5C), both treatments are associated to PPARα/RXRα activation. However, the most important effect may be related to RSG treatment by activation of PXR/RXR and FXR/RXR functions.

Article

DISCUSSION

Glitazones and fibrates are widely employed for the treatment of glucose and lipid alterations. However, their use is controversial, because they have been proven to be associated with adverse secondary effects.32,33 Indeed, the use of RSG was banned in Europe in 2010, and it is increasingly restricted by the USA Food and Drug Administration.34 In particular, there is a significant debate in the recent literature on whether these compounds increase or decrease the accumulation of lipids in the liver in various experimental models or in patients with NAFLD. The present study shows that RSG and FF decrease body weight and improve insulin resistance but exacerbate liver steatosis in ob/ob and LDLR-double deficient mice given a standard chow diet. Both drugs simultaneously activate prosteatotic (i.e., Adfp, Scl27a1, Scd1)35−37 and antisteatotic genes (i.e, Acox, Cpt1, Foxo1)38−40 with a well-ordered result of aggravation of the hepatic lipid accumulation; similar effects have been reported in obese mice fed with a linoleic-acidsupplemented diet.41 An interesting observation from our study is the upregulation of the mammalian target of rapamycin complex (mTOR) gene expression. mTOR is a key activator of protein kinases that act downstream of insulin and growth factor signaling, and it is a key regulator of the autophagy process. As a consequence of mTOR action, autophagy is inhibited;42 conversely, mTOR inhibition induces autophagy.43 Hepatic autophagy is already decreased in obese mice, and restoration of autophagy amelioraties liver steatosis in mice.44,45 It is therefore likely that the characteristic deficiency in autophagy observed in obesity may be aggravated by the concomitant use of the studied drugs. Consequently, an alteration in vesicle transport in hepatocytes, a substantial contribution to liver steatosis, endoplasmic reticulum stress, and progression of liver disturbances could be expected. In the present study, we show that FF and RSG may play a role in the modulation of autophagy. Whether this is important to explain our findings and the possibility of clinical translation requires further research. The net result in this experimental model is the aggravation of the degree of “spontaneous” steatosis. However, the effects of thiazolidinediones and fibrates on liver structure and function have generated considerable debate with contradictory results among species.46−49 For instance, a clinical trial in patients with nonalcoholic steatohepatitis found an improvement in the liver histology of some patients after RSG administration for 48 weeks, but results were far from being homogenenous.50 Similarly, a recent meta-analysis concluded that whether fibrates have significant effects on hepatic steatosis in humans or not still remains to be established.51 These results indicate that the effects of RSG or FF in the liver are complex and involve multiple mechanisms. Our study is novel in that using an integrated transcriptomic and metabolomic approach may give us a more accurate picture of the multitude of biochemical pathways altered by PPARα or PPARγ activation. Additionally, our data provides us with increased knowledge as to how the metabolic network is perturbed in obesity combined with alterations in the management of lipoproteins, a common accompanying condition. This is expected to facilitate the development of metabolic models to understand the alterations of metabolism during related diseases such as NAFLD. We showed that both compounds simultaneously induce pro-steatotic and antistea-

Selected Networks Affected by FF and RSG Treatments

The IPA analysis identified three networks with more than seven focus genes each among the differentially expressed molecules (DEM) in FF-treated mice compared to untreated animals (Figure 6). Analyses of these networks revealed links between fatty acid synthesis, immune functions, lipid transport, fructose/mannose metabolism, autophagy, and energy metabolism. IPA also revealed that the network with a greater number of DEM (Figure 6A) contained the nuclear receptors PPARα and PPARδ as central regulators of the expression of genes involved in fatty acid synthesis and oxidation (Cpt1, Scd1), lipid transport (Scl27a1), peroxysomal β-oxidation (Acox), and immune response (Cat, NFKβ). All these genes were markedly up-regulated by FF treatment. IPA also highlighted links between palmitic acid, palmitoleate acid, carnitine, and the nuclear receptors PPARα and PPARδ, as well as between insulin signaling and amino acid metabolism. PPARα was linked to the transcription factor Acox, to proinflammatory cytokines, and to the expression of linoleic acid, related to bile acid functions and diseases. This finding is consistent with the hypothesis that altered bile acid metabolism in diabetes that may be related to glucose and lipid control. When we compared the RSG group with untreated animals, IPA analyses revealed five networks with more than eight focus genes each among the DEM (Figure 7), although the most marked changes linked two major networks. The principal network (Figure 7A) contained the nuclear receptors PPARα and PPARδ which, together with PPARγ, act as central regulators of the expression of genes involved in fatty acid oxidation (Foxo1, Cpt1), glycolysis and gluconeogenesis (Slc2a2, PCK1), and triglyceride synthesis (Dgat2). This network also revealed regulatory loops between PPARγ and genes associated with lipid transport, such as Fabp4 and Slc27a1, by the up-regulation of adiponectin gene expression, which was also related to taurine expression. Previously published results from our laboratory revealed a link between decreased taurine levels and body weight increase.30,31 IPA also highlighted links between the nuclear receptor PPARγ and adiponectin expression to immune response (Emr1), which was markedly down-regulated by RSG treatment. Similar to the FF group, PPARα linked to the transcription factor Acox and the expression of linoleic acid were related to bile acid related functions and diseases, although in that case, up-regulation of taurine levels was also included. The other key network revealed an increased mTOR gene expression by RSG treatment. 1740

dx.doi.org/10.1021/pr401230s | J. Proteome Res. 2014, 13, 1731−1743

Journal of Proteome Research

Article

Madrid, Spain, and by the Bijzonder Onderzoeksfonds of the KU Leuven (PF/10/014), by the Interdisciplinair Ontwikkelingsfonds−Kennisplatform (KP/12/009), and by the Fonds voor Wetenschappelijk Onderzoek-Vlaanderen (G0846.11, and Vascular Biology Network).

totic pathways and that both elevated the potential oxidative stress. Curiously, these potential mechanisms are more important than those derived from an improved utilization of glucose. The result on our experimental model is an imbalance favoring the increase of the degree of steatosis, but it is likely that balance can change from one side to the other depending on the species investigated, the dose and time schedule of administration or, perhaps, the associated comorbidities. This is a cautionary note, which is addressed not only to researchers studying basic mechanism of of PPAR activators but also to clinicians prescribing these compounds in the clinical practice. This is particularly important, because it is often difficult to establish the presence of NAFLD, and because a significant number of individuals are asymptomatic until their liver begins to fail, at which point potential treatments are useless.



(1) Kalaitzidis, R. G.; Sarafidis, P. A.; Bakris, G. L. Effects of thiazolidinediones beyond glycaemic control. Curr. Pharm. Des. 2009, 15 (5), 529−536. (2) Bojic, L. A.; Huff, M. W. Peroxisome proliferator-activated receptor δ: a multifaceted metabolic player. Curr. Opin Lipidol. 2013, 24 (2), 171−177. (3) Wang, Y. X. PPARs: diverse regulators in energy metabolism and metabolic diseases. Cell. Res. 2010, 20 (2), 124−137. (4) Geeraert, B.; Crombé, F.; Hulsmans, M.; Benhabilès, N.; Geuns, J. M.; Holvoet, P. Stevioside inhibits atherosclerosis by improving insulin signaling and antioxidant defense in obese insulin-resistant mice. Int. J. Obes. 2010, 34 (3), 569−577. (5) Mertens, A.; Verhamme, P.; Bielicki, J. K.; Phillips, M. C.; Quarck, R.; Verreth, W.; Stengel, D.; Ninio, E.; Navab, M.; Mackness, B.; Mackness, M.; Holvoet, P. Increased low-density lipoprotein oxidation and impaired high-density lipoprotein antioxidant defense are associated with increased macrophage homing and atherosclerosis in dyslipidemic obese mice: LCAT gene transfer decreases atherosclerosis. Circulation. 2003, 107 (12), 1640−1646. (6) Angulo, P. Obesity and nonalcoholic fatty liver disease. Nutr. Rev. 2007, 65 (6 Pt. 2), S57−S63. (7) Rull, A.; Camps, J.; Alonso-Villaverde, C.; Joven, J. Insulin resistance, inflammation, and obesity: role of monocyte chemoattractant protein-1 (or CCL2) in the regulation of metabolism. Mediators Inflamm 2010, 1−11, DOI: 10.1155/2010/326580. (8) Lee, A. H.; Scapa, E. F.; Cohen, D. E.; Glimcher, L. H. Regulation of hepatic lipogenesis by the transcription factor XBP1. Science 2008, 320 (5882), 1492−1496. (9) Nishimura, S.; Manabe, I.; Nagasaki, M.; Eto, K.; Yamashita, H.; Ohsugi, M.; Otsu, M.; Hara, K.; Ueki, K.; Sugiura, S.; Yoshimura, K.; Kadowaki, T.; Nagai, R. CD8+ effector T cells contribute to macrophage recruitment and adipose tissue inflammation in obesity. Nat. Med 2009, 15 (8), 914−920. (10) Winer, S.; Chan, Y.; Paltser, G.; Truong, D.; Tsui, H.; Bahrami, J.; Dorfman, R.; Wang, Y.; Zielenski, J.; Mastronardi, F.; Maezawa, Y.; Drucker, D. J.; Engleman, E.; Winer, D.; Dosch, H. M. Normalization of obesity-associated insulin resistance through immunotherapy. Nat. Med. 2009, 15 (8), 921−929. (11) Feuerer, M.; Herrero, L.; Cipolletta, D.; Naaz, A.; Wong, J.; Nayer, A.; Lee, J.; Goldfine, A. B.; Benoist, C.; Shoelson, S.; Mathis, D. Lean, but not obese, fat is enriched for a unique population of regulatory T cells that affect metabolic parameters. Nat. Med. 2009, 15 (8), 930−939. (12) Johnson, N.; Walton, D.; Sachinwalla, T.; Thompson, C.; Smith, K.; Ruell, P.; Stannard, S.; George, J. Noninvasive assessment of hepatic lipid composition: Advancing understanding and management of fatty liver disorders. Hepatology 2008, 47 (5), 1513−1523. (13) Ratziu, V.; Giral, P.; Jacqueminet, S.; Charlotte, F.; HartemannHeurtier, A.; Serfaty, L.; Podevin, P.; Lacorte, J. M.; Bernhardt, C.; Bruckert, E.; Grimaldi, A.; Poynard, T. LIDO Study Group. Rosiglitazone for nonalcoholic steatohepatitis: one-year results of the randomized placebo-controlled fatty liver improvement with rosiglitazone therapy (FLIRT) trial. Gastroenterology 2008, 135 (1), 100−110. (14) Sanyal, A. J.; Chalasani, N.; Kowdley, K. V.; McCullough, A.; Diehl, A. M.; Bass, N. M.; Neuschwander-Tetri, B. A.; Lavine, J. E.; Tonascia, J.; Unalp, A.; van Natta, M.; Clark, J.; Brunt, E. M.; Kleiner, D. E.; Hoofnagle, J. H.; Robuck, P. R. NASH CRN. Pioglitazone, vitamin E, or placebo for nonalcoholic steatohepatitis. N. Engl. J. Med. 2010, 362 (18), 1675−1678.



LIMITATIONS OF THE STUDY We have studied atherosclerosis in mice, and mechanisms of murine atherosclerosis can be different from these of human atherosclerosis. Furthermore, double mutant mice had very high cholesterol levels, whereas most obese persons do not. However, previously we found that the increased oxidative stress in obese dyslipidemic mice in association with the metabolic syndrome and the accelerated atherogenesis in relation with metabolic syndrome components in these mice is in agreement with the increased oxidation of LDL in obese people and its relation with metabolic syndrome, insulin resistance, and atherogenesis.19,52−54



ASSOCIATED CONTENT

S Supporting Information *

Primers used for quantitative real-time reverse transcriptionPCR analysis. Differential expression of the 46 selected genes in FF- and RSG-treated animals, calculated by relative quantification normalized respect to control group. Increased and decreased gene expressions that achieve statistical significance are shown in red and green, respectively. Differential expression of metabolites in FF- and RSG-treated animals. Increased and decreased gene expressions that achieve statistical significance are shown in red and green, respectively. Network functions and top canonical signaling pathways associated to the differed clusters (box) of genes and metabolites. Canonical signaling pathways associated to each treatment with p-value