Abstract
Lipid homeostasis is controlled in part by the nuclear receptors peroxisome proliferator (PP)-activated receptor α (PPARα) and liver X receptor (LXR) through regulation of genes involved in fatty acid and cholesterol metabolism. Exposure to agonists of retinoid X receptor (RXR), the obligate heterodimer partner of PPARα, and LXR results in responses that partially overlap with those of PP. To better understand the gene networks regulated by these nuclear receptors, transcript profiles were generated from the livers of wild-type and PPARα-null mice exposed to the RXR pan-agonist 3,7-dimethyl-6S,7S-methano, 7-[1,1,4,4-tetramethyl-1,2,3,4-tetrahydronaphth-7-yl]-2E,4E-heptadienoic acid (AGN194,204) or the PPAR pan-agonist WY-14,643 (WY; pirinixic acid) and compared with the profiles from the livers of wild-type and LXRα/LXRβ-null mice after exposure to the LXR agonist N-(2,2,2-trifluoroethyl)-N-[4-(2,2,2-trifluoro-1-hydroxy-1-trifluoromethylethyl)phenyl] sulfonamide (T0901317). All 218 WY-regulated genes altered in wild-type mice required PPARα. Remarkably, ∼80% of genes regulated by AGN194,204 required PPARα including cell-cycle genes, consistent with AGN-induced hepatocyte proliferation having both PPARα-dependent and -independent components. Overlaps of ∼31 to 62% in the transcript profiles of WY, AGN194,204, and T0901317 required PPARα and LXRα/LXRβ for statistical significance. Ofthe 50 overlapping genes regulated by T0901317 and WY, all but one were regulated in a similar direction. These results 1) identify new transcriptional targets of PPARα and RXR important in regulating lipid metabolism and liver homeostasis, 2) illustrate the importance of PPARα in regulation of gene expression by a prototypical PP and by an RXR agonist, and 3) provide support for an axis of PPARα-RXR-LXR in which agonists for each nuclear receptor regulate an overlapping set of genes in the mouse liver.
Nuclear receptors (NRs) are critical modulators of developmental and physiological processes and are both targets of drugs as well as chemicals of environmental significance. The NR peroxisome proliferator-activated receptor (PPAR) subtypes α, β, and γ are activated by a structurally diverse group of chemicals, including peroxisome proliferators (PPs) that increase the number and size of peroxisome organelles in mouse and rat liver (Klaunig et al., 2003). The PPAR subtypes have unique ligand-specificities (Corton et al., 2000) as well as expression patterns in the liver with PPARα expressed primarily in hepatocytes, PPARβ expressed in multiple cell types and PPARγ expressed in Kupffer cells (Braissant et al., 1996). In mice and rats, PPARα agonists elicit a predictable course of adaptive responses in the liver, including peroxisome proliferation, induction of lipid-metabolizing genes, and hepatomegaly (Klaunig et al., 2003). PPARα agonists have been used clinically for many years because of their ability to lower cholesterol and triglyceride levels in patients at risk for coronary heart disease (Hihi et al., 2002). Studies in wild-type and PPARα-null mice demonstrated that the phenotypic effects of PP exposure in the liver, including peroxisome proliferation, cell proliferation, and alteration of fatty acid metabolism genes, depend on a functional PPARα (Lee et al., 1995; Klaunig et al., 2003). Although there is a general assumption that genes altered by PPs in the liver are regulated by PPARα, a number of PPs, including WY-14,643 (WY) can act as PPAR pan-agonists activating PPARβ and γ in trans-activation assays (Kliewer et al., 1996). Notably, the PP bezafibrate can alter gene regulation in the livers of PPARα-null mice through PPARβ (Peters et al., 2003). Comprehensive transcript profiling in PP-treated wild-type and PPARα-null mice will be useful to determine the extent of PPARα-independent gene changes upon PP exposure.
The three retinoid X receptors (RXRα, β, and γ) bind to the naturally occurring vitamin A derivative 9-cis-retinoic acid and are the targets of experimental drugs known as rexinoids. Rexinoids act as insulin sensitizers and were beneficial in treating non-insulin-dependent (type 2, or adult-onset) diabetes and obesity in experimental models (Faul and Grese, 2002). The exact mechanisms by which RXR agonists modulate these responses have not been identified, partly because of the complexity of the interactions between RXRs and NRs. In addition to binding DNA response elements as a homodimer, RXR is required as a heterodimeric partner for a large number of NRs (Rastinejad, 2001). These NRs collectively called the class II receptors include all PPAR subtypes as well as other receptors expressed in the liver, including the liver X receptor. This promiscuity of RXR makes it difficult to separate the effects of RXR activation alone from activation through its heterodimeric partners. If RXR is to be exploited as a drug target, there is a need to understand the involvement of the class II receptors in the regulation of rexinoid gene targets.
Liver X receptors are key regulators of cholesterol metabolism. The two subtypes, LXRα and LXRβ, are activated by oxysterols and synthetic compounds, including T0901317 (T1317) to regulate genes involved in cholesterol transport and metabolism to bile acids (Tontonoz and Mangelsdorf, 2003; Steffensen and Gustafsson, 2004). LXR also controls the expression of sterol regulator element-binding protein-1c, which regulates several lipogenic enzymes. Additional gene targets regulated by LXR include those involved in steroid hormone synthesis, growth hormone signaling, and inflammation (Stulnig et al., 2002; Joseph et al., 2003).
Nuclear receptors can “cross-talk” to ensure that antagonistic pathways are not simultaneously activated. For the class II receptors, cross-talk can occur through competition for a limiting amount of RXR so that activation of one NR-RXR heterodimer may have negative effects on the ability of other NRs to heterodimerize with RXR and regulate gene expression. PPARα can antagonize the ability of LXR to activate the sterol regulator element-binding protein-1c promoter by titration of a limiting amount of shared RXR (Yoshikawa et al., 2003). Likewise LXR can antagonize the ability of PPARα to activate gene expression, and this may occur through titration of RXR or through direct interactions between PPAR and LXR (Miyata et al., 1996; Ide et al., 2003). Comparison of comprehensive transcript profiles elicited by T0901317 (Stulnig et al., 2002), PPs, and rexinoids would help to characterize any cross-talk between these receptors.
The objectives of the present study were to 1) determine the role of PPARα in mediating the transcriptional response to a PP; 2) classify PPARα-dependent and -independent targets of an RXR agonist; and 3) determine the extent of the overlap in the transcriptional programs regulated by PPARα, RXR, and LXR in the mouse liver.
Materials and Methods
Animal Treatments. All animal studies were conducted under federal guidelines for the use and care of laboratory animals and were approved by the Institutional Animal Care and Use Committee of the CIIT Centers for Health Research. Male wild-type and PPARα-null mice on a mixed SV129/C57BL/6 background were used in these studies and have been described previously (Lee et al., 1995). Mice were provided with NIH-07 rodent chow (Zeigler Brothers, Gardeners, PA) and deionized, filtered water ad libitum. Lighting was on a 12-h light/dark cycle. Mice received a single daily gavage dose between 9:00 and 10:00 AM of either the RXR pan-agonist AGN-194,204 (AGN; 3 mg/kg/day; Allergan, Irvine, CA), the PPAR panagonist WY-14,643 (50 mg/kg/day; ChemSyn Science, Lenexa, KS) or the carrier methylcellulose (0.1%) each day for 3 days. These doses were selected based on previous studies showing that maximal transcriptional responses are induced in the absence of overt toxicity (R. Chandraratna, unpublished observations; S. Anderson and J. C. Corton, unpublished observations). Twenty-four hours after the last dose, animals were deeply anesthetized with pentobarbital injection and killed by exsanguination. Portions of the liver were snap frozen in liquid nitrogen and stored at -70°C until processed for analysis of mRNA. Sections of liver were fixed in 10% neutral buffered formalin for 48 h, transferred to 70% ethanol, and embedded in paraffin. Subsequently, 5-μm sections were mounted on slides, stained with hematoxylin and eosin, and examined by light microscopy.
RNA Isolation and Analysis of Gene Expression Using Oligonucleotide Arrays. Three mice were analyzed from each of six treatment groups, for a total of 18 analyses. Hepatic RNA was isolated using a modified guanidium isothiocyanate method (TRIzol; Invitrogen, Carlsbad, CA) and was further purified using silica membrane spin columns (RNEasy total RNA kit; QIAGEN, Valencia, CA). RNA integrity was assessed by ethidium bromide staining followed by resolution on denaturing agarose gels and also by the RNA 6000 LabChip kit using a 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). For each sample from 18 individual mice, 15 μg of biotin-labeled cRNA was generated from 10 μg of total RNA and hybridized to GeneChip Test3 arrays (Affymetrix, Inc., Santa Clara, CA) to determine quality. Subsequently, the same samples were hybridized to Murine GeneChip U74Av2 oligonucleotide arrays (Affymetrix, Inc.). All procedures were carried out according to the manufacturer's recommendations, using the antibody amplification technique. Images were initially processed using the MAS 5.0 software (Affymetrix, Inc.). Hybridization quality was assessed by visual inspection of the image and from a report generated by MAS 5.0. Criteria for an acceptable hybridization were as follows: background <100, noise (RawQ) < 5, 3′/5′ ratio for select housekeeping genes <4. Hybridizations not meeting these criteria were repeated, beginning at the target preparation step. The data were analyzed and statistically filtered using Rosetta Resolver version 3.0 software (Rosetta Inpharmatics, Kirkland, WA). The threshold for significance was set at p ≤ 0.001, and genes that exhibited a ≥1.5- or ≤-1.5-fold change were reported as a fold-change relative to the corresponding control. Genes altered by WY or AGN in wild-type and PPARα-null mice are reported. Sixty-four of these genes were also altered in control PPARα-null mice compared with control wild-type mice. An additional 59 genes were altered only in control PPARα-null mice compared with wild-type mice and will be characterized elsewhere. Similarly regulated genes were visualized using CLUSTER and TreeView (Eisen et al., 1998).
We compared the genes regulated by WY and AGN with those regulated by other class II nuclear receptors in the liver. An initial analysis of available published gene lists for constitutive androstane receptor, LXR, PXR, farnesoid X receptor, and thyroid hormone receptor indicated that only the genes regulated by LXR exhibited significant overlap. We next determined the extent of the overlap directly by comparing the WY and AGN data set to that generated in wild-type and LXRα/LXRβ-null mice treated with the LXR agonist T0901317 using the Affymetrix U74Av2 mouse chips (Stulnig et al., 2002). Genes that were reported as significant (p ≤ 0.001) and exhibited a ≥1.5- or ≤-1.5-fold change were compared using CLUSTER and TreeView. Genes were grouped into functional classes with the help of KEGG (http://www.kegg.org) and using Gene Ontology (http://www.geneontology.org) identifiers in the U74Av2 template (http://www.affymetrix.com). Identification of expressed sequence tags was facilitated by euGenes (http://iubio.bio.indiana.edu:89/mouse/).
Real-Time PCR Analysis of Gene Expression. Real-time PCR was performed as follows. After DNase treatment, total RNA was quantified (Ribogreen; Molecular Probes, Eugene, OR) and diluted with water. Fifty nanograms of RNA and PCR reagents were separated into aliquots on 96-well plates using a Prism 6700 Automated Nucleic Acid Workstation (Applied Biosystems, Foster City, CA) and subjected to real-time PCR (TaqMan; Applied Biosystems) using gene-specific primers and fluorescently labeled probes (Molecular Probes) (Supplemental Table 1) designed by the Primer Express software (Applied Biosystems). Amplification curves were generated using the ABI Prism 7900HT Sequence Detection System (Applied Biosystems). Expression relative to vehicle control animals was determined after normalizing to the ribosomal 18S gene (Applied Biosystems). There were three animals per treatment group, and each sample was analyzed in duplicate. Variability is expressed as standard error of the mean.
Western Blot. Liver lysates were prepared in 250 mM sucrose, 10 mM Tris-HCl, pH 7.4, 1 mM EDTA with protease inhibitors (0.2 mM phenylmethylsulfonyl fluoride, 0.1% aprotinin, 1 μg/ml pepstatin, and 1 μg/ml leupeptin) as described previously (Fan et al., 2003). Fifty micrograms of whole cell lysate was subjected to 12% SDS-polyacrylamide gel electrophoresis followed by transfer to nitrocellulose membranes. Immunoblots were developed using primary antibodies against acyl-CoA oxidase (ACO) (a gift from Dr. S. Alexson, Huddinge University Hospital, Huddinge, Sweden), MFP-II and thiolase (gifts from Dr. T. Hashimoto, Shinshu University, Nagano, Japan), or Cyp4A (GenTest, Waltham, MA) and appropriate secondary antibodies conjugated with horseradish peroxidase (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in the presence of chemiluminescent substrate ECL (Amersham Biosciences, Piscataway, NJ). Blots were quantitated using Gel-Pro (Media Cybernetics, Silver Spring, MD).
Determination of Hepatocellular Proliferation. Osmotic minipumps (Alzet model 2001, 7 day pumps, 1 μl/h; Alza Corporation, Palo Alto, CA) filled with 16 mg/ml 5-bromo-2′-deoxyuridine (BrdU) in phosphate-buffered saline were implanted into the mice the day before the start of treatment. Nuclei that incorporated BrdU were identified by immunohistochemistry (Miller et al., 2000). Light microscopy was performed using a Microphot microscope (Nikon, Melville, NY) with a Dage charge-coupled device color videocamera (Dage-MTI, Michigan City, IN). The hepatocytes were analyzed using the Cytology Histology Recognition Identification System (Sverdrup Medical/Life Sciences Imaging Systems, Ft. Walton Beach, FL). At least 1000 cells were counted for each animal. Cells that incorporated BrdU were identified by red pigmented nuclei. Ten to 15 fields were counted. The labeling indices of hepatocytes were determined in different zones as described previously (Bahnemann and Mellert, 1997). To guarantee lobule comparison, the distance from the portal tract to the central vein was determined. Each lobule was subdivided into three parts, representing the three zones: zone 1, periportal; zone 2, intermediary; and zone 3, perivenous. The labeling index was calculated by dividing the number of labeled hepatocyte nuclei by the total number of hepatocyte nuclei counted, and the results expressed as a percentage.
Statistical Analysis of Data. Statistical test of significance was done by analysis of variance post hoc testing performed using the Tukey-Kramer test with a p value of ≤0.05 (JMP; SAS Institute, Research Triangle Park, NC). Spearman rank correlation test was performed using SAS (SAS version 6.12; SAS Institute).
Results
Transcriptional Programs Regulated by Agonists of PPARα and RXR. We identified 383 genes that were significantly (p ≤ 0.001; ≥1.5- or ≤-1.5-fold change) different between two or more groups as outlined under Materials and Methods. The identified genes were classified into major groups based on their expression behavior (Fig. 1). The first three classes were dependent on PPARα for altered expression after exposure. The largest class of genes (class I) was regulated solely by WY. Class II genes were regulated by both WY and AGN. The regulation of these genes by AGN was not caused by direct activation of PPARα, because AGN does not activate any of the PPAR subtypes (R. Chandraratna, unpublished observations). Class III genes were solely regulated by AGN. The class IV genes, which were predominantly up-regulated, were altered by AGN in a PPARα-independent manner because regulation was approximately the same in both wild-type and PPARα-null mice. Class V genes were regulated by AGN only in PPARα-null mice. Class VI genes were defined as regulated by WY in a PPARα-dependent manner and AGN in a PPARα-independent manner. Class VII genes consisted of genes altered by WY only in PPARα-null mice. Class VIII genes were regulated by either WY or AGN in PPARα-null mice only. Class IX genes had mixed regulation. The classes I-IX contained 174, 31, 79, 18, 34, 9, 28, 5, and 5 genes, respectively. In addition, we identified genes that were dependent on PPARα for basal expression (discussed below), reinforcing the concept that PPARα controls the expression of many genes through activation by endogenous lipids including fatty acids (Aoyama et al., 1998).
Two major conclusions can be drawn from the profiles. First, PPARα controls the majority of changes altered by WY exposure. Using our standard filtering criteria (p ≤ 0.001; ≥1.5- or ≤-1.5-fold change), all 219 genes regulated by WY in wild-type mice were PPARα-dependent; i.e., similar expression was not observed in PPARα-null mice. Under these selection conditions, however, there were five genes that exhibited opposite regulation by WY in PPARα-null mice versus wild-type mice. If WY was activating PPARβ and PPARγ in PPARα-null mice, we would expect any changes to be relatively subtle because of the dilution of these transcripts from PPARβ- or PPARγ-containing cell types by hepatocyte transcripts. We therefore repeated the selection of genes in the absence of a fold-change cutoff. Eleven genes were identified that exhibited similar regulation in the two strains (Supplemental Table 2) and 25 genes exhibited opposite regulation in the two strains (Supplemental Table 3). In this second group of genes, there was a significant negative correlation between gene expression in each strain (R = -0.95; p < 0.001 by Spearman analysis) as the fold-change rank order was roughly opposite in each strain.
The profiles also revealed that AGN required PPARα for the majority of gene expression changes. Using the standard filtering criteria, 111 of a total 138 AGN-regulated genes (80.4%) were dependent on PPARα because expression changes after AGN treatment were observed in wild-type but not PPARα-null mice. To ensure that our selection criteria were not biasing our results, we decreased the stringency of the cutoffs. When we removed the fold-change cutoff but retained the significance cutoff, more AGN-regulated genes were observed (372), but the percentage of those genes that required PPARα (303) remained approximately the same (81.5%). When we relaxed our filtering criteria to include only a p < 0.05 as a cutoff, 76.8% of the genes (602 of 784) were PPARα-dependent. These findings indicate that PPARα plays a crucial role for an RXR agonist to elicit a transcriptional response in the mouse liver.
Overlap in the Genes Regulated by WY, AGN, and an LXR Agonist. Our gene expression profiles identified genes regulated by AGN that were PPARα-independent (classes IV, V, VI, VIII, and IX), indicating that AGN alters gene expression through other nuclear receptor-RXR heterodimers. To determine whether there was an overlap in the transcript profiles determined by PPARα, RXR, and LXR in the mouse liver, the profiles of WY and AGN were compared with those of the LXR agonist T1317 in wild-type and LXRα/LXRβ-null mice as described under Materials and Methods. All genes were filtered using the same criteria (p ≤ 0.001; ≥1.5- or ≤-1.5-fold change). To determine the statistical significance of any overlapping genes, we performed a Spearman rank correlation test on the 499 genes regulated by at least one of the three compounds in both wild-type and nullizygous mice (Table 1). There was a significant negative correlation in the five genes regulated by WY in wild-type versus PPARα-null mice, as discussed above. Significant overlaps existed between WY in wild-type mice and AGN in wild-type but not PPARα-null mice. Significant overlaps were also observed between WY and T1317 in wild-type but not in nullizygous mice, indicating that the overlap between WY and T1317 regulated genes requires PPARα and LXRα/LXRβ. The transcript profile of AGN in wild-type or PPARα-null mice had a significant overlap with T1317 in wild-type but not LXRα/LXRβ-null mice pointing to the possibility that some of the PPARα-independent genes altered by AGN are regulated through LXR-RXR heterodimers.
The expression of T1317-responsive genes was compared with those regulated by WY and AGN. The overlapping genes altered by exposure to the three compounds in wild-type mice are shown in Fig. 2A, and the number of genes altered in wild-type mice are shown in Fig. 2B. Remarkably, only one of 87 overlapping genes was regulated in an opposite manner by WY versus AGN or WY versus T1317 (Armet, mutated in early stage tumors). The extent of the overlap ranged from ∼31 to 62%. For up-regulated genes, T1317 had the greatest overlap with the other two compounds (42%) followed by AGN (34%) and WY (32%). For down-regulated genes AGN had the greatest overlap (62%) followed by T1317 (33%) and WY (31%). Overall, these results indicate that WY, AGN, and T1317 regulate an overlapping set of genes in wild-type mice primarily dependent on PPARα or LXRα/LXRβ.
Expression of Fatty Acid Metabolism Genes. The role for PPARα in regulating genes involved in fatty acid metabolism is well known (Hihi et al., 2002). However, there is little information about regulation of these genes by RXR agonists. Genes involved in fatty acid metabolism altered by WY or AGN were grouped into functional categories and compared with those regulated by T1317 (Table 2). Additional genes regulated by WY, AGN, and T1317 are found in Supplemental Table 4. The expression of the PPARα gene Ppara was repressed in control PPARα-null mice compared with control wild-type mice (-1.72 fold-change; p ≤ 9.3 × 10-5), as expected, and expression did not change after compound treatment (data not shown). All of the genes involved in peroxisomal and mitochondrial fatty acid β- and ω-oxidation identified here were induced by WY. The genes included three forms of peroxisomal 3-oxoacyl-coenzyme A thiolase (Acaa1) and AW122615, similar to trifunctional protein, β subunit. The genes also included 1300002P22Rik, 83% similar to human enoyl-coenzyme A, hydratase/3-hydroxyacyl coenzyme A dehydrogenase (Ehhadh), and 4930479F15Rik, known as hydroxyacyl-coenzyme A dehydrogenase/3-ketoacyl-coenzyme A thiolase/enoyl-coenzyme A hydratase (trifunctional protein), β subunit (Hadhb). Induction by WY was abolished in PPARα-null mice (WY, WT versus WY, Null; Table 2). Eight of the genes, including one of the Acaa1 forms, were down-regulated in control PPARα-null mice compared with control wild-type mice (CON, Null; Table 2), consistent with earlier studies describing constitutive activation of β-oxidation genes by PPARα, probably by endogenous activators such as fatty acids (Aoyama et al., 1998).
AGN treatment induced four of the β-oxidation genes and both Cyp4a genes that were also induced by WY. The AGN-responsive genes included 1300002P22Rik, two forms of Acaa1, Acat2, MGC29978, Cyp4a10, and Cyp4a14. Regulation of all of these genes except Cyp4a10 was partially or completely dependent on PPARα because increased expression was lost in PPARα-null mice (AGN, WT versus AGN, Null; Table 2).
Extra- and intracellular fatty acid transport is facilitated by proteins located on cell membranes (e.g., fatty acid transporters) and within the cytoplasm (e.g., fatty acid binding proteins). Three fatty acid binding proteins (Fabp2, Fabp4, and Fabp5) and a fatty acyl-CoA binding protein (Dbi) were induced by WY. Other fatty acid transporters, located on the cell surface or that act as intercellular transporters, were either induced (Abcd3 and Slc25a20) or decreased (Apom) by WY. Of all of these transporters, only Fabp2 was induced by AGN and in a PPARα-dependent manner. Slc25a20 was the only transporter dependent on PPARα for basal level expression. These results suggest that WY coordinately induces genes involved in both fatty acid β-oxidation and transport, but that under these conditions, AGN induces only a subset of these genes.
Genes involved in fatty acid and triglyceride synthesis were regulated by WY or AGN. Fatty acid synthesis genes induced by WY included Fasn, Helo1-pending, and two forms of Scd1. AGN induced Scd1 and down-regulated Elov3. All genes required PPARα for alteration by WY or AGN. The Fasn gene was up-regulated by AGN only in PPARα-null mice. Akr1b7 involved in glycerol lipid synthesis was induced by AGN only. WY and AGN induced the triglyceride synthesis genes Gpam and Chk, respectively. Induction of Gpam may serve as a way to funnel free, potentially toxic fatty acids into storage.
Lipases that govern the release of fatty acids from intra- or extracellular stores were up-regulated by WY (Lpl and Mgl1). The preferential up-regulation of these secreted lipases by WY may indicate an increased use of extracellular fat stores for energy (Auwerx et al., 1996). Consistent with this, Cte1 and Pte1, which release fatty acids from acyl-CoA stores, were up-regulated by WY or AGN and may be involved in the use of fatty acids for β-oxidation (Hunt et al., 1999). Miscellaneous genes involved in fatty acid metabolism and peroxisome proliferation regulated only by WY included Aldh3a2, Cyp2c37, Cyp2c40, Pex11a, and Pltp.
LXR agonists have well known effects on fatty acid metabolism, including increases in triglyceride synthesis that lead to increases in circulating triglyceride levels. T1317 exposure led to increases in five fatty acid β-oxidation genes, two ω-oxidation genes, two Fabp family members, and four genes involved in fatty acid and triglyceride synthesis. The number of fatty acid metabolism genes that overlapped was greater for WY and T1317 than for WY and AGN (16 versus 10 genes).
We confirmed the expression of some genes involved in fatty acid metabolism by real-time PCR (TaqMan) and by Westerns. Cyp4a14, Acox1, and Lpl gene expression exhibited the same trends that were observed by transcript profiling (i.e., preferential PPARα-dependent up-regulation by WY over AGN) (Fig. 3A). Expression of the ACO, MFP-II, thiolase, and CYP4A proteins was increased in wild-type mice by WY exposure (Fig. 3B). AGN exposure led to weaker induction of ACO and CYP4A compared with WY. AGN did not appreciably induce MFP-II or thiolase. Induction of all proteins was dependent on PPARα. These studies illustrate the coordinate regulation of a large number of fatty acid metabolism and transport genes by WY, and to a lesser extent, AGN.
Altered Expression of Cholesterol Synthesis Genes. Nine genes involved in cholesterol synthesis were up-regulated by AGN exposure. Four of these genes were PPARα-dependent. We examined the expression of four cholesterol synthesis genes by TaqMan (Fig. 3C). Cyp51, Fdft1, Fdps, and Idi1 exhibited increased expression by AGN in wild-type mice. Although expression of the four genes seemed to be increased by AGN in PPARα-null mice, the increases did not reach significance. In addition, AGN altered expression of bile acid synthesis genes including decreases in Cyp7b1 and increases in Ltb4dh. The AGN-regulated genes involved in cholesterol and bile acid synthesis overlapped with those regulated by T1317. Although the results are consistent with AGN coordinately regulating cholesterol synthesis genes, a role for PPARα needs to be examined further. This is the first study that we are aware of that demonstrates the coordinated regulation of cholesterol synthesis genes by a rexinoid. Further work is required to determine whether RXR activation by AGN directly or indirectly regulates these genes.
Coordinate Regulation of Inflammation Genes by WY. WY exposure resulted in coordinate decreases in the expression of genes elevated during times of acute or chronic inflammation, including the acute phase response. Acute phase proteins (APPs) included Apcs, Hpsn, Orm1, Orm2, and Saa2. Expression of serum amyloid components is also positively associated with increases in very low-density lipoprotein and triglyceride levels (Kindy et al., 2000). Seven components of the complement cascade that are also APPs were down-regulated by WY. Decreases in APP expression may be due in part to the decrease in expression of leukemia inhibitory factor receptor (Lifr), which dimerizes with gp130 and controls interleukin-6-dependent expression of the APPs. Although AGN and T1317 down-regulated Lifr, there was no coordinate down-regulation of the same inflammation genes as with WY, indicating other genes may be targets of the anti-inflammatory properties of T1317 (Tontonoz and Mangelsdorf, 2003).
PPARα-Dependent and -Independent Induction of Cell Proliferation in Mouse Liver by AGN. Both PPARα and RXR agonists increase cell proliferation in rat liver (Standeven et al., 1997; Corton et al., 2000). The ability of RXR agonists to increase cell proliferation in mouse liver and dependence on PPARα is not known. Increases in liver to body weights were observed after WY or AGN treatment in wild-type but not PPARα-null mice. (Fig. 4A). We measured hepatocellular proliferation after AGN exposure and compared that to induction by WY. Because chemicals can increase cell proliferation differentially across the liver lobule (Bahnemann and Mellert, 1997), we measured hepatocellular proliferation in different lobular zones. In wild-type mice, WY increased hepatocyte proliferation in the centrilobular region only (zone 3) (Fig. 4B). AGN induced hepatocyte proliferation in wild-type mice in zones 2 and 3, being strongest in zone 3. Increases in hepatocyte proliferation by WY were abolished in PPARα-null mice. PPARα-null mice exposed to AGN retained levels of hepatocyte proliferation in zone 2 but exhibited decreases in zone 3 proliferation compared with wild-type mice. Thus, the increases in hepatocyte proliferation by AGN had both PPARα-dependent and -independent components. Cell proliferation occurred in the absence of cell death either through cytotoxicity or overt increases in apoptosis (data not shown).
The cell cycle is controlled by the regulated expression and degradation of cyclins, cyclin-dependent kinases (Cdk), and Cdk inhibitors. The maturation promoting factor (MPF) is a universal regulator of the G2/M transition (Doree and Hunt, 2002). AGN exposure led to increased expression of MPF components, including Ccnb1, Ccnb2, and Cdc2a as well as Ccna2 that may interact with MPF (Arooz et al., 2000) (Table 2).
Components of the anaphase promoting complex are required for the degradation of CDK1 and entrance into telophase and cytokinesis. There were AGN-induced increases in the components of APC (Anapc5) or proteins that negatively regulate the APC (Mad2l1; Cdc20 and Kif4). Negative regulators of the cell cycle also exhibited increased expression, including p21 (Cdkn1a) and p18 (Cdkn2c). Most of the cell cycle, kinetochore, anaphase, and DNA structural genes were regulated by AGN through a PPARα-dependent mechanism. Compared with AGN very few cell cycle genes were altered by WY. This may be attributed to the kinetics of WY-induced cell proliferation, which reaches a peak approximately 48 h after initial exposure (J. C. Corton, unpublished observations). The kinetics of AGN induction of cell proliferation in the mouse liver is not known, but the preponderance of AGN-regulated cell cycle genes may reflect delayed cell proliferation compared with WY.
We confirmed the expression of cell cycle genes identified by transcript profiling (Fig. 4C). The genes Ccnb1, Cdc20, Cdc2a, and Mad2l1 exhibited similar expression profiles. AGN strongly induced expression in wild-type mice, whereas induction in PPARα-null mice was greatly attenuated but in most cases still significantly altered. Ccna2 was induced by AGN in wild-type mice but not in PPARα-null mice. Changes in Cdkn1a did not reach statistical significance. Discrepancies between the transcript profile and TaqMan studies exist (i.e., Ccnb1, Cdc2a, and Cdkn1a) highlighting the need to determine more subtle expression results using multiple methods. Collectively, these results identify genes that may be involved in both PPARα-dependent and -independent hepatocyte proliferation induced by AGN.
Miscellaneous Genes Regulated by WY and AGN. Exposure to WY or AGN resulted in altered expression of additional genes which fell into multiple categories (Supplemental Table 4). The type I deiodinase gene (Dio1), down-regulated by AGN in mice (Macchia et al., 2002), was down-regulated by AGN but not WY in both wild-type and PPARα-null mice. Many genes involved in proteolysis were altered by WY, including up-regulation of proteasomal components and down-regulation of genes involved in proteolytic cascades. A large number of genes have little or no information regarding their function or their possible role in the phenotypic effects of these compounds.
Discussion
The obligate role for RXR as a heterodimer partner for class II NRs is well known, but the genes regulated by rexinoid signaling have not been systematically identified. Here, transcript profiling has been used to generate a comprehensive view of transcriptional alterations in liver genes after exposure to an RXR agonist or a PP in wild-type and PPARα-null mice. The regulated genes fell into a number of categories and their regulation can be explained based on prior knowledge of NR-RXR interactions in the liver (Fig. 5). The class I and II genes are those activated by WY or WY and AGN and are completely dependent upon PPARα for chemical-induced responses. A large body of evidence supports a dominant role for the PPARα-RXR heterodimer in PP-regulated gene expression (Corton et al., 2000; Hihi et al., 2002). WY-treated mice harboring a liver-specific deletion of the RXRα gene showed greatly reduced hepatic induction of Acox1, Cyp4a1, and Fabp1 (Wan et al., 2000). RXR agonists increased the expression of genes under control of PPARα, including Fabp1 (Poirier et al., 1997), bifunctional enzyme (also known as MFP-I), and Acox1 (Standeven et al., 1997; Mukherjee et al., 1998), and fatty acid transfer protein and acyl-CoA synthetase (Martin et al., 2000). The RXR agonist LGD1069 increased the expression of Cyp4a, thiolase, Acox1 as well as cholesterol levels in wild-type but not PPARα-null mice (Ouamrane et al., 2003). Many of the class I and II genes are probably activated by endogenous activators of PPARα because we identified many of these genes down-regulated in control PPARα-null mice versus control wild-type mice, especially genes involved in fatty acid β-oxidation. PPARα interacts with other NR and coactivators (summarized in Corton et al., 2000) that may alter DNA binding specificity or transcriptional activity. Thus, some of the class I and II genes could be regulated by mechanisms other than through a PPARα-RXR heterodimer. In particular, there is no evidence that a PPARα-RXR heterodimer is directly involved in down-regulating expression of PP-regulated genes. PPARα negatively regulates genes under control of c-Jun or nuclear factor-κB through interactions with and sequestration of their subunits (Delerive et al., 1999). This mechanism is consistent with a large number of class I genes involved in inflammatory responses down-regulated by WY that are known targets of nuclear factor-κB.
The majority of genes regulated by WY were PPARα-dependent, an expected finding given the important role PPARα plays in mediating the phenotypic responses of PPs (Klaunig et al., 2003). However, a modest number of genes altered by WY in PPARα-null mice were identified, and they fell into three groups. The first group, consisted of genes regulated only in PPARα-null mice (class VII) and were dominated by down-regulated genes. Genes in the second group were regulated similarly by WY in both strains. The overall expression changes in this group were relatively weak, possibly because of dilution by hepatocyte transcripts of differentially regulated transcripts originating in minor cell types of the liver. Genes in the third group were altered in an opposite manner in the two strains, reminiscent of estrogen receptor ligands that demonstrate agonistic or antagonistic properties depending on the cell type in which they are measured (McDonnell et al., 2001). Additional experiments are required to determine whether these genes are regulated through complexes of PPARβ or PPARγ with cell type-specific coactivators/corepressors or through alternate mechanisms.
We unexpectedly identified a large percentage of genes (∼80%) that were regulated predominantly by AGN and require PPARα for most, if not all, of their changes (class III genes). The identification of cell cycle genes in this class is consistent with the PPARα-dependent increases in centrilobular hepatocyte proliferation observed in AGN-treated wild-type but not PPARα-null mice. Given the dependence of regulation on PPARα and the specificity of AGN for RXR family members, these genes are most likely regulated through PPARα-RXR heterodimers. Some nuclear receptors within a heterodimer are transcriptionally silent under different promoter-specific contexts (Westin et al., 1998), but this is the first example, to our knowledge, of the identification of genes that require PPARα for expression but that are not activated by a PPARα ligand to a significant extent. It will be of interest to determine whether other rexinoids primarily rely on PPARα for gene expression changes in the liver and whether these genes are nonresponsive to other PP.
RXR is a promiscuous heterodimer partner of many nuclear receptors in the liver, and we identified genes altered by AGN in a PPARα-independent manner (class IV and VI genes). AGN might regulate these genes either through RXR-RXR homodimers or through heterodimers of RXR and class II NR other than PPARα. Given the significant overlap in the transcript profiles of AGN and T1317, it is possible that many of the class IV genes are regulated through LXR-RXR heterodimers. Additional evidence for class VI genes comes from studies in which 9-cis-retinoic acid or a rexinoid could activate PPARα target genes or responses in PPARα-null mice (Ouamrane et al., 2003; IJpenberg et al., 2004) and 9-cis-retinoic acid could rescue the hypothermic phenotype observed in fasted PPARα-null mice (IJpenberg et al., 2004).
LXRα/LXRβ and PPARα regulate alternate pathways of fatty acid synthesis and catabolism. Functional antagonism between PPARα and LXR would help ensure that these opposing pathways are not simultaneously activated. Evidence for antagonism between PPARα and LXR comes from in vitro trans-activation studies as well as analysis of a small number of genes in the mouse liver after short-term exposure to WY or T1317 (Ide et al., 2003; Yoshikawa et al., 2003). PPARα activation was shown to inhibit activation of LXR-regulated genes and reciprocally, LXR activation inhibited PPARα-regulated gene expression. Antagonism under these conditions probably comes from titration of shared RXR as well as the formation of inactive PPARα-LXR heterodimers (Miyata et al., 1996; Ide et al., 2003; Yoshikawa et al., 2003). If this antagonism existed in the mouse liver, we would predict that activation of PPARα would lead to opposite regulation of LXR-activated genes and vice versa. Under the conditions of our exposures, we found no evidence for antagonism. First, all but one of the 49 genes regulated by both WY and T1317 were regulated in the same direction. Although T1317 suppressed expression of PPARα targets Hmgcs, Acox1, Cpt1, and CYP4A2 in the livers of fasted mice after 18 h (Ide et al., 2003), a 7-day exposure to T1317 had either no effect on these genes or increased expression of Cyp4a family members (Stulnig et al., 2002). Second, of the 64 genes constitutively regulated by PPARα in control mice, we found only one that was regulated in a manner that indicated antagonism by T1317; i.e., constitutively up-regulated in control wild-type mice versus control PPARα-null mice and down-regulation by T1317 in wild-type mice versus control wild-type mice. Discrepancies between our work and those reported earlier could be caused in part by differences in exposure times for T1317 and WY or to the use of in vitro systems to determine antagonism in which PPARα and LXR were highly expressed to nonphysiological levels. Our work indicates that PPARα and LXR activate an overlapping set of genes involved in both fatty acid catabolism as well as synthesis. The physiological advantage of activating these opposing pathways remains to be determined.
Mechanisms of cross-talk between PPARα and LXR probably exist that do not involve competition for RXR. Increased expression of the LXR gene itself and LXR-regulated genes by PP and PPARα could occur through the peroxisome proliferator response elements found in the promoters of the mouse (Tobin et al., 2000) and human (Laffitte et al., 2001) LXR genes. LXR gene transcripts are increased in in vitro models and in mouse liver after exposure to PPs (Tobin et al., 2000). The PPARα agonist clofibrate increases the levels of the endogenous activators of LXR, 25- and 27-hydroxycholesterol, in the livers of rats (Guan et al., 2003). Although the basis for LXR regulation of PPARα genes remains elusive, WY may regulate LXR genes through increasing LXR-RXR heterodimer formation and activation. Transcript profiling in PP-treated wild-type and LXR-null mice will help to identify LXR-dependent and -independent genes regulated by PPARα.
In summary, these studies reinforce the dominant role played by PPARα in the regulation of multiple classes of genes by PPAR and RXR agonists in the mouse liver. The significant overlap in the transcript profiles regulated by PPARα, RXR, and LXR support the concept of a PPARα-RXR-LXR axis in the liver that acts to control lipid homeostasis.
Acknowledgments
We thank Drs. Kevin Morgan and Ron Tyler for support, Dr. Steve Clark for assistance in performing some of the bioinformatics aspects of the study, Drs. S. Alexson and T. Hashimoto for antibodies, Dennis House for assistance in performing some of the statistics, Dr. Paul Howroyd for support of pathology, the CIIT Animal Care and Necropsy and Histology Units for assistance in performing these studies, and Drs. Kevin Gaido and Rusty Thomas for critical review of the manuscript.
Footnotes
-
This work was supported in part by National Institute of Environmental Health Sciences grant ES09775-01 (to J.C.C.), a Marie Curie Fellowship of the European Community program Human Potential (contract number HPMF-CT-2000-00898) (to T.M.S.), and the Swedish Science Council and KaroBio AB (to J.-Å.G.).
-
ABBREVIATIONS: NR, nuclear receptor; PPAR, peroxisome proliferator-activated receptor; PP, peroxisome proliferator; RXR, retinoid X receptor; LXR, liver X receptor; PCR, polymerase chain reaction; PXR, pregnane X receptor; ACO, acyl-CoA oxidase; BrdU, 5-bromo-2′-deoxyuridine; APP, acute phase protein; Cdk, cyclin-dependent kinase; MPF, maturation promoting factor; WY, WY-14,643 [pirixinic acid (4-chloro-6-(2,3-xylidino)-2-pyrimidinyl)thioacetic acid)]; T1317, T0901317 [N-(2,2,2-trifluoroethyl)-N-[4-(2,2,2-trifluoro-1-hydroxy-1-trifluoromethylethyl)phenyl]sulfonamide]; LGD1069, targretin; AGN, AGN194,204 (3,7-dimethyl-6S,7S-methano, 7-[1,1,4,4-tetramethyl-1,2,3,4-tetrahydronaphth-7-yl]-2E,4E-heptadienoic acid).
-
↵s⃞ The online version of this article (available at http://molpharm.aspetjournals.org) contains four supplemental tables.
-
↵1 Current address: Bayer Corporation, Research Triangle Park, NC 27709.
- Received July 27, 2004.
- Accepted September 14, 2004.
- The American Society for Pharmacology and Experimental Therapeutics