|Home | About | Journals | Submit | Contact Us | Français|
The peroxisome proliferator–activated receptor (PPAR) family of nuclear hormone transcription factors (PPARα, PPARβ/δ, and PPARγ) is regulated by a wide array of ligands including natural and synthetic chemicals. PPARs have important roles in control of energy metabolism and are known to influence inflammation, differentiation, carcinogenesis, and chemical toxicity. As such, PPARs have been targeted as therapy for common disorders such as cancer, metabolic syndrome, obesity, and diabetes. The recent application of metabolomics, or the global, unbiased measurement of small molecules found in biofluids, or extracts from cells, tissues, or organisms, has advanced our understanding of the varied and important roles that the PPARs have in normal physiology as well as in pathophysiological processes. Continued development and refinement of analytical platforms, and the application of new bioinformatics strategies, have accelerated the widespread use of metabolomics and have allowed further integration of small molecules into systems biology. Recent studies using metabolomics to understand PPARα function, as well as to identify PPARα biomarkers associated with drug efficacy/toxicity and drug-induced liver injury, will be discussed.
The peroxisome proliferator–activated receptors (PPARs) are ligand-activated transcription factors belonging to the nuclear receptor (NR) superfamily. Structurally, PPARs (α, (β/δ, and γ) are highly conserved sharing similar domains with other NRs; with a ligand-independent and a ligand-dependent transcriptional activation function (AF-1 and AF-2, respectively) that sandwiches the zinc-finger DNA-binding domain, and a ligand-binding domain (Peters, Shah, and Gonzalez 2012; Francis et al. 2003). Based on amino acid sequence similarity, the PPARs belong to group one of the six-member NR superfamily. More generically, the PPARs are considered part of the metabolic sensors of the NR superfamily that are distinct from the steroid and orphan receptors (Desvergne, Michalik, and Wahli 2006). Given its important and documented role in metabolism and disease, this review focuses predominantly on PPARα. However, there have been fundamental metabolomics studies (albeit few compared to PPARα) with respect to PPARβ/δ (Roberts et al. 2011; Patterson and Peters 2011) and PPARγ (Watkins et al. 2002; van Doorn et al. 2007), and it is anticipated that metabolomics will have a similarly significant impact on understanding the function of these receptors.
PPARα was the first of the PPARs to be discovered and is so named because it was shown to mediate the proliferation of peroxisomes in hepatic tissue of rodents following administration of the hypolipidemic drug clofibrate (Issemann and Green 1990). PPARα is expressed mainly in the liver, but is also expressed in the kidney, heart, skeletal muscle, and brown adipose tissue (Braissant et al. 1996). Endogenous ligands that bind with the highest affinity to PPARα are saturated/unsaturated fatty acids, leukotriene derivatives, and VLDL hydrolysis products. Examples of synthetic ligands that bind PPARα are the fibrate class of hypolipidemic drugs, the experimental ligand Wy-14,643 ([4-chloro-6-(2,3-xylidino)-2-pyrimidinylthio] acetic acid) as well as some phthalate monoesters (monoethylhexyl phthalate), and herbicides (lactofen; Bility et al. 2004; Gonzalez, Peters, and Cattley 1998). PPARα is a major regulator of the mitochondrial and peroxisomal β-oxidation pathway, and as discussed below, these pathways have been implicated in the pathogenesis of various liver complications.
Downstream of the genome, transcriptome, and proteome, the metabolome (Table 1) may in fact be the most accurate indicator of cellular physiology (Idle and Gonzalez 2007). The metabolome represents the complete set of small molecules found in biofluids (blood, plasma, serum, urine, sweat, saliva) and, unlike the genome, only a small fraction of the metabolome has been annotated. However, efforts as part of the Human Metabolome Database and the Human Serum Metabolome in Health and Disease initiatives have begun to systematically identify and describe metabolites found in various biofluids (Cottingham 2008). Once thought to be a simple source (Pearson 2007) of biological information (compared with the genome, transcriptome, and proteome), appreciation for the complexity and richness of the metabolome has grown. In 2007, the initial “draft” of the human metabolome containing 2,500 metabolites was reported (Wishart et al. 2007); however, in just over 5 years, the Human Metabolome Database (HMDB) now contains nearly 8,000 metabolites. Further, given that discrete chemical exposures in humans are thought to be on the order of 2 to 3 million in a lifetime (Idle and Gonzalez 2007), metabolomics will also be important for capturing information regarding exposure to xenobiotics. This is particularly relevant for human studies where metabolomics approaches are likely to capture not only endogenous but also xenobiotics and their metabolites (Johnson et al. 2012; Patterson, Gonzalez, and Idle 2010).
While the precise definition of metabolomics varies throughout the literature, it can be simply defined as the global, unbiased measurement of small molecules found in biofluids, or extracts from cells, tissues, or organisms (Griffin and Nicholls 2006). In more general terms, metabolomics can be described as the chemical fingerprints left behind by endogenous biological processes or the biological activity on chemicals derived from the diet and/or environment (Daviss 2005). However defined, metabolomics has already provided unprecedented views of PPAR biology and function yielding important insights and validation of their roles in health and disease.
Metabolomics approaches have been described in great detail elsewhere (Dettmer, Aronov, and Hammock 2007; Patterson et al. 2010). Briefly, the process (Figure 1) involves extraction of metabolites from a biofluid (urine, serum, plasma), data acquisition using a variety of platforms including 1H-NMR, UPLC-ESI-QTOFMS, or GC-MS, followed by peak alignment and normalization using a variety of commercial (Waters MarkerLynx, AB SCIEX MarkerView) and public (XCMS, MZMine) software tools (Smith et al. 2006; Tautenhahn et al. 2012; Eliasson et al. 2012; Katajamaa, Miettinen, and Oresic 2006; Pluskal et al. 2010), and ultimately multivariate data analysis (MDA) to identify a metabolite or metabolites that can distinguish one sample group from the other. It is important to note that despite many technological advances in chromatographic separations (GC, LC) and detection platforms (1H-NMR, MS), there does not yet exist a single platform that can capture all metabolites as say a microarray chip would for gene expression. Therefore, a combination of approaches (1H-NMR, LC-MS, GC-MS) is necessary to increase coverage of the metabolome. Metabolite identification typically involves structural elucidation using tandem mass spectrometry of the metabolite compared with that from an authentic compound. The last highly recommended steps involve quantitation of the metabolite on platforms such as triplequadrupoles that afford high degrees of linear dynamic range.
Metabolomics studies of PPARα function have benefited tremendously from the generation of the Ppara-null mouse (a summary of some significant metabolomic studies using Ppara-null mice is described in Table 2; Lee et al. 1995; Akiyama et al. 2001). Many such studies have contributed significantly to better defining the role of PPARα in general metabolism as well as in response to various synthetic and diet-derived ligands. However, the Ppara-null mouse is also essential to identify and validate biomarkers that are specifically associated with activation of the PPARα pathway. In addition, diet, age, gender, environment, and gut microbiota can be well controlled in mouse models helping to reduce any variability or “background noise” that is common to many human metabolomics studies (Patterson and Idle 2009).
An elegant metabolite flux study aimed at understanding the role of PPARα in hepatic glucose production was first used to better understand the hypoglycemic state seen in fasted Ppara-null mice (Xu et al. 2002).This study would not fall under the true definition of metabolomics as it is not unbiased and global, but is instead targeted. However, it is discussed here to provide compelling evidence that understanding the routes by which metabolites are synthesized or catabolized is as equally important as understanding the absolute concentrations of metabolites. To accomplish the flux studies in mice, mini-osmotic pumps were implanted to supply mice (wild-type and Ppara-null) with stable isotope tracers such as [U-13C6]glucose, [U-13C3]lactate, or [2-13C]glycerol. Through this approach, the authors report that the Ppara-null mouse has defective glucose carbon recycling (lactate/pyruvate and glucose) and that this change, in addition to reduced liver glycogen levels, contributes to the sensitivity of the Ppara-null mouse to fasting. Others using more classical metabolomics approaches (nuclear magnetic resonance spectroscopy [NMR]-, gas chromatography coupled with mass spectrometry [GC-MS]-based analysis of tissue extracts) have arrived at similar conclusions (Atherton et al. 2006).
The first urinary metabolomic study to assess PPARα activation and function was conducted using wild-type and Ppara-null mice in order to identify urinary biomarkers indicative of its activation with the experimental ligand Wy-14,643 (Zhen et al. 2007). Here extensive use of the Ppara-null mouse along with ultra pressure liquid chromatography coupled with electrospray ionization quadrupole time-of-flight mass spectrometry (UPLC-ESI-QTOFMS) and MDA was used to unequivocally identify metabolites specific to PPARα activation. Excretion of a series of urinary metabolites in the tryptophan–nicotinamide pathway (e.g., nicotinamide-1-oxide, nicotinamide, and 1-methylnicotinamide) was increased in wild-type compared to Ppara-null mice. This was consistent with the known effect of PPARα on the tryptophan-nicotinamide pathway, specifically its indirect involvement in downregulating the α-amino-β-carboxymuconate-ε-semialdehyde decarboxylase (ABCSMD) gene. Furthermore, NAD is an important cofactor for many oxidation-reduction enzymes and is particularly important during fatty acid β-oxidation. While this observation was anticipated and could be explained based on previous studies (Shin et al. 2006), the effect of PPARα on steroid 21-oic acid synthesis (an increase in the urinary excretion of the 11β, 20-dihydroxy-3-oxopregn-4-en-21-oic and 11β-hydroxy-3,20-dioxopregn-4-en-21-oic acids) was unexpected. Those observations were later explained mechanistically in a separate follow up metabolomics study that identified a new link between PPARα and the hypothalamic-pituitary-adrenal axis (Wang et al. 2010).
As mentioned above, aging produces dramatic changes in the metabolome and if not properly controlled for can produce artifactual metabolomics results. In order to better understand the impact of the aging process on the metabolome, changes in various mouse tissue metabolomes with respect to age and PPARα status under normal conditions were examined (Atherton et al. 2009). Age-related metabolic changes were similarly identified in both wild-type and Ppara-null mice (decreased glycogen and glucose in the liver, decreased lactate in muscle) through a combination of NMR, GC-MS-based metabolomics, and MDA including regression analysis. However, compared to the wild-type mouse, these changes (perturbed glycolysis/gluconeogenesis, reduced fatty acid catabolism) were much more pronounced in the Ppara-null mouse, and with respect to the liver, were associated with more severe steatosis. These results are clearly beneficial for generating a more complete understanding of PPARα function. However, given that formal study of the metabolome is still quite new, the importance of investigating how basic metabolic processes such as aging influences the metabolome cannot be overstated.
The fibrate class of PPARα ligands is used to treat hyperlipidemia by reducing serum triglyceride levels (Guerre-Millo et al. 2000; Fruchart, Duriez, and Staels 1999). In rodents, prolonged treatment with PPARα ligands including fibrates results in hepatocarcinogenesis, yet humans and mice expressing the human PPARα receptor (on the Ppara-null background) appear resistant to these carcinogenic effects (Yang et al. 2008). Since adverse effects of fibrates have been reported in humans, investigation of fibrate-induced toxicity and the identification of biomarkers of PPARα activation remains an area of active research.
In 2009, Ohta and colleagues used LC-MS- and GC-MS-based metabolomics analysis of urine and serumto determine the global effects of fenofibrate treatment in rats (Ohta et al. 2009). Consistent with that reported in the above studies using wild-type and Ppara-null mice treated with the experimental PPARα ligand Wy-14,643, dramatic increases in nicotinamide (>13-fold) and 1-methylnicotinamide (>4-fold) were found in the urine of Fischer 344 male rats after 2 weeks treatment with fenofibrate. Metabolites associated with increased fatty acid catabolism (carnitine, pantothenate) were also decreased in the urine of fenofibrate-treated rats consistent with the role of PPARα in activating the fatty acid β-oxidation pathway. Some signs of both liver and kidney toxicity were also evident including increased quinolinate and 1-methylguanidine, respectively, although these metabolites were not observed with Wy-14,643 treatment in mice.
A similar study conducted in healthy human volunteers revealed biomarkers reflecting increased fatty acid β-oxidation (Patterson et al. 2009). In this study, fenofibrate was given to healthy individuals for 14 days, the urine profiled by UPLC-ESI-QTOFMS, and data analyzed by random forests. In addition to identifying acetylcarnitine (>14-fold reduction) and the CoA precursor pantothenate (>5-fold reduction) as being dramatically decreased in urine following fenofibrate treatment, these putative biomarkers of PPARα activation were subsequently validated using the Ppara-null mouse. While wild-type mice exhibited a dramatic decrease in urinary acetylcarnitine and pantothenate following treatment with Wy-14,643, the Ppara-null mouse exhibited no significant change. This study demonstrated for the first time a specific urinary biomarker of PPARα activation that may be useful in drug development and design, as well as for identifying interesting outliers as potential nonresponders (Figure 2) especially as new therapeutics targeting these pathways are developed. This study further exemplifies the utility of conducting these studies in animal models (mice, rats, monkeys) as similar changes in metabolites were observed across many different species.
As indicated above, metabolomics is an invaluable tool for understanding not only PPARα activation but also, as discussed below, potential toxicity associated with its downregulation. Acetaminophen (APAP) is a common nonprescription drug used as an analgesic (pain reliever) and antipyretic (fever reducer). It is known to cause liver toxicity when taken in excess of the recommended dose and is a concern due to frequent overdose in children and adults (D’Arcy 1997; Mazer and Perrone 2008). It also serves as an excellent model to study the general effects of drug-induced liver injury. APAP is metabolized to an active metabolite N-acetyl-p-benzoquinone imine (NAPQI) by CYP2E1 (Gonzalez 2007). NAPQI is a highly reactive quinone that can bind to cellular nucleophiles and cause cell toxicity. Mice lacking expression of CYP2E1 are resistant to APAP toxicity (Chen et al. 2009). To determine the role of metabolism in APAP toxicity and to find biomarkers for liver toxicity, metabolomics was used to monitor the serum metabolome of wild-type and Cyp2e1-null mice treated intraperitoneally with a toxic dose (400 mg/kg) of APAP. Using UPLC-ESI-QTOFMS and MDA, potential biomarkers of hepatotoxicity were identified including the long chain acylcarnitine, palmitoylcarnitine. Furthermore, increased palmitoylcarnitine levels in the serum correlated well with the accumulation of triglycerides and free fatty acids in mice treated with a hepatotoxic dose of APAP. These observations suggested that the fatty acid β-oxidation pathway was impaired and that these serum biomarkers may be reflective of mitochondrial damage associated with APAP treatment. A subsequent examination of fasted Ppara-null mice revealed that acylcarnitine levels were elevated and suggest that the serum acylcarnitines may be useful indicators of hepatotoxicity.
In a subsequent study, the experimental PPARα agonist Wy-14,643 was found to protect against APAP-induced hepatotoxicity (Patterson et al. 2012). Mice fed a 0.1% Wy-14,643 diet for 24 hr before an injection of 400mg/kg APAP were completely protected from APAP-induced liver damage, and levels of serum acylcarnitines were well correlated with sensitivity (increased serum levels) and protection (normal levels). While not implicated directly by metabolomics, although a fatty acid β-oxidation defect was, the mitochondrial uncoupling protein 2 (UCP2) was identified as one of the key mitochondrial proteins responsible for mediating the protective effects. Future studies examining the impact of UCP2 expression on the cellular metabolome are currently underway. Interestingly, several UCP2 variants are known to be associated with obesity and it is anticipated that metabolomics will help uncover the mechanisms by which UCP2 contributes to the fatty acid β-oxidation pathway.
Chronic alcohol consumption is a major cause of nonaccident-related death particularly due to the development of alcohol-induced liver disease (ALD; Manna et al. 2011, 2010). UPLC-ESI-QTOFMS and MDA were used to view the global effects of chronic alcohol consumption in wild-type and Ppara-null mice to identify potential biomarkers of ALD. The metabolomics approach taken was able to show a distinct separation between the alcohol-treated versus control mice as well as between the wild-type and the Ppara-null mice. In particular, Ppara-null mice fed an alcohol diet were shown to separate not only from control but also from wild-type mice fed an alcohol diet. The urinary metabolomics data supported observations that disruption of the tryptophan/quinolinic acid/NAD pathway normally regulated by PPARα inhibits fatty acid β-oxidation, and leads to enhanced and accelerated fat deposition in the liver. Additional studies in the 129SvJ background found indole-3-lactic acid and phenyllactic acid increased in Ppara-null mice that correlated well with increased levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) as well as changes in NAD+/NADH. Given that subtle differences in the background strain could contribute to any of the observed biomarkers, it was important to establish a strain-independent biomarker of ALD. This study, in addition to its contribution to the field of ALD research, provides compelling evidence to examine the variation in metabolomes of common strains of laboratory mice.
It is clear from the studies described here that metabolomics has provided a new vantage from which to study PPARα function (Figure 2) that is made only more relevant with the strategic use of the Ppara-null mouse. In addition to uncovering novel pathways involving PPARα activation (steroid 21-oic acid synthesis) and its repression (APAP-induced liver injury), metabolomics has provided new insights into the influence of basic physiological factors such as age and the rather profound impact of mouse strain on the metabolome. As the field continues to grow, it is hoped that studies addressing the basic fundamentals of a metabolomics study (e.g., how does the procedure used to extract metabolites from a biofluid or tissue influence the observed metabolomic profile?) will be completed. The analytical technology (advanced liquid chromatography systems, new high resolution mass spectrometers) has dramatically outpaced the development of rigorous informatics tools, comprehensive databases, and the refinement of standardized reporting of not only the raw data but also the conditions (extraction procedure, chromatography gradient, mass spectrometer settings) by which the data were collected.
In the last few years, there have been several key advances in mass spectrometry-based metabolomics that will likely shape metabolomics studies designed not only for PPARα but also for those wishing to obtain metabolomic profiles in situ.
With regard to the former, a technique was developed that may be useful for identifying natural metabolites that are ligands for nuclear receptors (Kim, Lou, and Saghatelian 2011); this may be especially relevant for orphan nuclear receptors for which there is no known ligand. The technique is LC-MS-based for an untargeted, global analysis compared to the techniques that are used now such as the luciferase reporter assay for which candidate ligands must be known while also having a robust reaction in order to be detected. Recombinant nuclear receptors are attached to a solid support, mixed, and incubated with an extract from tissues where the receptor is expressed. To account for any background, synthetic ligands specific for a nuclear receptor (e.g., Wy-14,643 for PPARα) can be added in conjunction with small beads that lack protein. Once the lysates are filtered through, beads and proteins are washed followed by the elution of the nuclear receptors bound to their ligands. Enriched metabolites are then identified via LC-MS and MDA by comparison with control samples. In this report, the authors identified known ligands for both PPARα and PPARγ.
The very latest and perhaps most exciting addition to the field of mass spectrometry-based metabolomics is the recent development of tissue imaging based on nanostrucrure initiator mass spectrometry (NIMS; Patti, Shriver, et al. 2010; Patti, Woo, et al. 2010; Woo et al. 2008). Unlike traditional matrix-assisted laser desorption ionization (MALDI) imaging which relies on the deposition of a MALDI matrix, the NIMS approach appears to be more reproducible, has greater mass range, and requires little sample preparation for tissue. The NIMS approach compared with traditional metabolomics analysis of biofluids or tissue extracts has the added advantage of providing spatial information for metabolites (i.e., where in the tissue is the metabolite changing). Not only does this have important applications in general metabolomics studies, it will also be useful in drug metabolism and uncovering where a particular drug is metabolized or where a toxic metabolite may be formed.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: ES022186 (A.D.P), CA124533 (J.M.P), CA141029 (J.M.P), CA140369 (J.M.P), and the NIH Intramural Research Program (F.J.G.).
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.