ToxSci Advance Access originally published online on August 4, 2005
Toxicological Sciences 2005 88(1):250-264; doi:10.1093/toxsci/kfi273
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gene Expression Profiling of the PPAR-alpha Agonist Ciprofibrate in the Cynomolgus Monkey Liver
* GlaxoSmithKline Inc., Safety Assessment, Research Triangle Park, North Carolina 27709, and GlaxoSmithKline Inc., Human Biomarker Center, Research Triangle Park, North Carolina 27709
1 To whom correspondence should be addressed at GlaxoSmithKline, Building 9, Room 2011, Research Triangle Park, NC 27709. Fax: (919) 483-6858. E-mail: Neal.F.Cariello{at}gsk.com.
Received June 23, 2005; accepted July 27, 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Fibrates, such as ciprofibrate, fenofibrate, and clofibrate, are peroxisome proliferator-activated receptor-
(PPAR) agonists that have been in clinical use for many decades for treatment of dyslipidemia. When mice and rats are given PPAR agonists, these drugs cause hepatic peroxisome proliferation, hypertrophy, hyperplasia, and eventually hepatocarcinogenesis. Importantly, primates are relatively refractory to these effects; however, the mechanisms for the species differences are not clearly understood. Cynomolgus monkeys were exposed to ciprofibrate at various dose levels for either 4 or 15 days, and the liver transcriptional profiles were examined using Affymetrix human GeneChips. Strong upregulation of many genes relating to fatty acid metabolism and mitochondrial oxidative phosphorylation was observed; this reflects the known pharmacology and activity of the fibrates. In addition, (1) many genes related to ribosome and proteasome biosynthesis were upregulated, (2) a large number of genes downregulated were in the complement and coagulation cascades, (3) a number of key regulatory genes, including members of the JUN, MYC, and NF
B families were downregulated, which appears to be in contrast to the rodent, where JUN and MYC are reported to upregulated after PPAR agonist treatment, (4) no transcriptional signal for DNA damage or oxidative stress was observed, and (5) transcriptional signals consistent with an anti-proliferative and a pro-apoptotic effect were seen. We also compared the primate data to literature reports of hepatic transcriptional profiling in PPAR-treated rodents, which showed that the magnitude of induction in ß-oxidation pathways was substantially greater in the rodent than the primate.
Key Words: ciprofibrate; PPAR.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Peroxisome proliferator-activated receptor-
(PPAR) agonists comprise a wide variety of compounds, including pharmaceuticals, industrial chemicals, endogenous fatty acids, and eicosanoids.
PPAR plays a central role in the uptake and ß-oxidation of fatty acids, especially in the liver (Corton et al., 2000
; Reddy and Hashimoto, 2001). Stimulation of the PPAR receptor causes an increase in the transcription of genes related to fatty acid transport across the cell membrane, intracellular lipid trafficking, mitochondrial and peroxisomal fatty acid uptake, and both mitochondrial and peroxisomal fatty acid ß-oxidation (Mandard et al., 2004).
Administration of these agents to rats and mice typically causes hepatic peroxisome proliferation, hypertrophy, hyperplasia, and eventually hepatocarcinogenesis; importantly, primates are relatively refractory to these effects (Klaunig et al., 2003). The mechanism of PPAR-induced rat and mouse hepatocarcinogenesis is not completely understood; however, several hypothesis have been put forth and mainly fall into two camps, one relating to increased oxidative stress caused by peroxisome proliferation, and the other centering on alteration in apoptosis and/or mitogenesis (Corton et al., 2000; Gonzalez, 2002; Klaunig et al., 2003; Oliver and Roberts, 2002).
Fibrates, such as ciprofibrate and fenofibrate, are PPAR agonists that have been in clinical use for many decades and are generally considered safe; epidemiological data has not shown an increased risk for human liver tumors (Bentley et al., 1993; Cattley et al., 1998; IARC, 1995). It has been noted, however, that the epidemiological studies were fairly small and therefore may not be sufficiently robust to detect small increases in human liver carcinogenesis.
The data regarding the ability of fibrates to cause human and nonhuman primate liver peroxisome proliferation in vivo is mixed. Examination of liver biopsy samples from patients receiving therapeutic doses of PPAR agonists showed a slight increase in peroxisome number and peroxisome volume density, while other studies showed no increase in peroxisome number (Bentley et al., 1993; Hanefeld et al., 1983; Hinton et al., 1986). Likewise, in the nonhuman primate, there are conflicting reports on the ability of PPAR agonists to cause peroxisome proliferation in vivo (Bentley et al., 1993; Lake, 1995; Lalwani et al., 1985; Lock et al., 1989; Reddy et al., 1984). Peroxisome proliferation, when reported to occur in primates, is greatly reduced compared to what can occur in rats and mice.
However, a recent report by Hoivik et al. has clearly shown that treating cynomolgus monkeys for 2 weeks with ciprofibrate or fenofibrate causes hepatic peroxisome proliferation (Hoivik et al., 2004). Briefly, that report showed a dose-related increase in relative liver weight, peroxisome number, and mitochondrial number; the maximum observed increase in liver weight and peroxisome number was about two- and three-fold over controls, respectively.
This paper reports our interpretation of hepatic transcriptional profiling using Affymetrix human GeneChips® on the same primates treated with ciprofibrate as in Hoivik et al. (Hoivik et al., 2004). We apply a variety of unsupervised and supervised approaches, with emphasis on transcriptional responses for cell proliferation, apoptosis, oxidative stress, and DNA repair. Hepatic transcriptional profiling has been reported for a variety of PPAR agonists in the rodent, but to our knowledge, this report is the first in the primate. The entire dataset has been placed in the public domain.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Animal treatment.
The animals in the present study are from Hoivik et al. (2004)
, and the reader is referred to that publication for a detailed description of the experimental design, as well as clinical and pathology results. Liver gene expression was examined from animals treated for 4 days with 400 mg/kg/day ciprofibrate or 15 days with 3, 30, 150 or 400 mg/kg/day ciprofibrate. Animals receiving only vehicle constituted the control groups. The Institutional Animal Care and Use Committee (IACUC) reviewed the protocol to ensure that the animals were treated humanely and that distress and pain were minimized.
RNA isolation, microarray hybridization, and quality control.
Animals were fasted overnight prior to sacrifice. Approximately 3 g from the right lateral lobe of the liver was removed and was flash frozen in liquid nitrogen. Total RNA was isolated using TRIzol (Invitrogen Life Technologies, Inc., Invitrogen Corp., Carlsbad, CA) and was processed and hybridized to Affymetrix GeneChip® HGU95Av2 following the manufacturer's protocol (Affymetrix GeneChip® Expression Analysis, Technical Manual, rev 1). Poor-quality RNA from one animal was found, and this animal was excluded from the analysis, thus the vehicle control group for the 15-day treatment consisted of three animals, while all other 15-day treatment groups had four animals.
Identification of differentially expressed probesets and data availability.
Rosetta Resolver® Version 3.2, Build 3.2.1.1
[EC]
.13 was used to determine which probesets in a treatment group differed from the control group. All animals in each treatment group were compared to all animals in the appropriate vehicle control group. Rosetta Resolver® examines the variance in each group and determines a p-value for each probeset which indicates the probability that the expression level in the treated group is different than the control group; probesets with p 
0.01 were selected as differentially expressed. Data is expressed as a fold-change.
The dataset is available at NCBI GEO (http://www.ncbi.nlm.nih.gov/geo/) with accession number GSE2853. The following have been deposited: (1) Affymetrix CEL files, (2) Affymetrix MAS5-processed data, and (3) fold-changes and p-values as determined by Rosetta Resolver. The analysis in this manuscript is based on the Rosetta Resolver data, and the MAS5-processed data is provided as a convenience.
Principle components analysis (PCA), pathway analysis, and protein–protein interaction analysis.
The signal intensity for all probesets considered present by the Affymetrix MAS5 software was used for PCA, and all data was univariate scaled. PCA was performed using SIMCA-P version 10.0.2.0
[EC]
.
Three software programs were used to map differentially expressed probesets to metabolic and biochemical pathways, (1) a GlaxoSmithKline software tool called Biological Networks Information Integration (BNII), which maps to KEGG and BioCarta pathways, (2) GenMAPP (http://www.genmapp.org/default.html), and (3) DAVID / EASE (http://david.niaid.nih.gov/david/ease.htm).
Interaction networks and maps for differentially expressed probesets were obtained using a commercial product called the Ingenuity Pathway Analysis (IngenuityTM of Mountain View, CA; http://www.ingenuity.com/index.html).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Summary of Pathology, Toxicokinetics, Relative Liver Weights, Liver Peroxisome Counts, and Significantly Changed Probesets
The primate exposure in this study can be compared to a typical human clinical AUC. A single 100 mg dose of ciprofibrate produced a human AUC of approximately 2000 h·µg/ml (Ferry et al., 1989
), and the primate exposure ranged from 10 to 940% of the human exposure (See Table 1).
|
Hepatocellular hypertrophy was observed for animals treated for 15 days with ciprofibrate at 150 and 400 mg/kg/day; these doses produced an AUC between 5- and 10-fold over the human therapeutic AUC. In addition, there was increased subcapsular single-cell hepatocellular necrosis that was consistent with apoptotic cell death after treatment with ciprofibrate at all dose levels.
The relative liver weights and the number of hepatic peroxisomes increased with dose, reaching a maximum at 400 mg/kg/day; the fold-increase for relative liver weight and peroxisome proliferation compared to control was about 1.9-fold and 2.8-fold, respectively. Table 1 summarizes some of the pertinent information from Hoivik et al. (2004), as well the number of Affymetrix probesets altered in the liver of each treatment group.
Assumptions for hybridization of monkey RNA on human Affymetrix GeneChip®
No cynomolgus monkey Affymetrix GeneChip® exists and we used GeneChip® HGU95Av2, which is based on the human sequence. One publication examined the performance of rhesus monkey RNA on a human-based GeneChip®, and the authors concluded that a large fraction of probesets are effective at transcript detection in the cross-species hybridization (Chismar et al., 2002).
Our assumptions for a cross-species hybridization are: (1) human-based probesets which report as Present are hybridizing to the correct homologous monkey sequence, and the information is valid and interpretable; (2) probesets which report as Absent are uninformative and not interpretable, since a probeset can report as Absent for two reasons: the transcript is not present or the transcript is present but differences in the monkey sequence prevent hybridization.
Principle Components Analysis
Principle Components Analysis (PCA) provides a means to visualize high-dimensional gene expression data. Each animal is represented as a single point, and animals that are found in close proximity in PCA space typically have similar gene expression profiles. The PCA analysis shows, at a very high level, the variability between individual animals and the effect of treatment. In general, between 20 and 30% of the variance (R2) was captured in the first two PCA dimensions.
Figure 1 shows the PCA plot of the control and ciprofibrate-treated animals. Generally, all dose groups form reasonably tight clusters. A dose response exists in PCA space, and this indicates that ciprofibrate is having a pronounced and dose-related effect on gene expression.
|
Somewhat surprisingly, the 4- and 15-day controls are not in the same region of the PCA plot. The 4- and 15-day Affymetrix GeneChips® were processed on different days by different individuals, and that may have contributed to the observed divergence in the controls. The gene expression results for the 4- and 15-day treatment groups are considered valid, even given the difference in the controls, since each treatment group is compared to the appropriate control.
Mapping to Metabolic and Biochemical Pathways
Several software tools exist which map Affymetrix probesets onto classical metabolic and biochemical maps, and we used the following: (1) GENMAPP, (2) DAVID / EASE, and (3) a proprietary software tool developed at GlaxoSmithKline (GSK) called BNII. The GSK BNII software program maps to both KEGG (http://www.genome.jp/kegg) and BioCarta (http://www.biocarta.com/index.asp) pathways and has excellent visualization properties.
The fold-change for probesets which had a p-value of 0.01 for one or more treatment conditions was submitted to the mapping programs; no value for fold-change was submitted for treatment groups that had a p value of >0.01. We divided the ciprofibrate dataset into two sections, upregulated and downregulated, and submitted these datasets separately to the mapping tools.
Metabolic and biochemical pathways upregulated by ciprofibrate treatment.
Generally, the results from all mapping tools were very consistent, with the following pathways most strongly upregulated: oxidative phosphorylation, fatty acid metabolism, tryptophan metabolism, ribosome, and proteasome. Figure 2 shows that many genes in the fatty acid metabolism pathway were upregulated; some genes showed upregulation at all dose levels; other genes showed upregulation only at higher dose levels. The efficacy of PPAR agonists to positively influence dyslipidemia is based on increasing fatty acid ß-oxidation and oxidative phosphorylation (Staels et al., 1998), so upregulation of these pathways was expected; the fact that these pathways are among the most strongly upregulated provides reassurance that the cross-species GeneChip® hybridization is yielding valid results. Greater detail about the mapping software and additional upregulated pathways are shown in Supplementary Figure 1 and Supplementary Table 1.
|
A large number of genes in the tryptophan pathway were upregulated, in part because many of the genes are common between the fatty acid metabolism and tryptophan metabolism pathways, including many aldehyde dehydrogenases and cytochrome P450s. The pathway is shown in Supplementary Figure 1.
Tryptophan is a key amino acid in the biosynthesis of nicotinamide adenine dinucleotide (NAD+) (Murray et al., 2000). A variety of PPAR ligands, including fibrates, increase NAD+ in vivo in the rat liver and in vitro in rat hepatocytes by alteration of the tryptophan–NAD+ pathway (Shibata et al., 1996; Shin et al., 1998, 1999a,b). Interestingly, a group has proposed using urinary metabolites from the tryptophan–NAD+ pathway as a marker of peroxisome proliferation (Ringeissen et al., 2003). A series of genes involved in the metabolic production of acetyl CoA in the tryptophan pathway were upregulated, including ACAT1, ECHS1, EHHADH, HADHA, and HADH2. Acetyl CoA is used by fatty acid ß-oxidation, and the upregulation of these genes can be understood in this light.
A large number of genes related to ribosomal proteins were upregulated (Supplementary Fig. 1). Ribosomes are central to protein synthesis, and the upregulation of this pathway is likely contributing to the observed liver weight increases. The upregulation is an early event, in that a fair number of ribosome-related genes were upregulated at the 4-day timepoint, but not at the 15-day timepoint.
The proteasome also had a large number of upregulated transcripts (Supplementary Fig. 1). Generally, the response appears to be dose related, and no great difference exists at the early and late 400-mg/kg/day timepoints. The 26S proteasome is part of the ubiquitin–proteasome pathway and removes aged, damaged, and misfolded proteins in mammalian cells (Goldberg, 2003). The proteasome was initially thought to simply function as a recycler for damaged proteins, but recent evidence shows the ubiquitin–proteasome pathway to be of central importance in cell cycle, cell survival, and apoptosis (Adams, 2003; Bach and Ostendorff, 2003).
A recent publication showed increased hepatic expression of a large set of proteome maintenance genes in wild-type mice after treatment with a PPAR agonist (Anderson et al., 2004). PPAR-null mice are more sensitive to some hepatotoxic compounds, and the authors suggest that PPAR may regulate the timing and extent of hepatocyte repair.
While the ubiquitin-proteasome pathway is involved in fundamental cell processes, many proteasome genes and genes related to lipid metabolism can be upregulated in the rat by simply fasting the animals, including the very long, long, and medium chain acyl CoA dehydrogenases, lipoprotein lipase, carnitine palmitoyl transferase 1 and 2 and acyl CoA synthetase (de Lange et al., 2004). We observed upregulation of both proteasome and lipid metabolism genes in the primate liver, and it is possible that the transcriptional increase in genes relating to the proteasome reflects a basic aspect of both lipid metabolism and PPAR pharmacology that is not well appreciated.
In summary, the pathway analysis shows that the genes relating to fatty acid metabolism, especially ß-oxidation and mitochondrial oxidative phosphorylation, were strongly upregulated. Thus, the broad outline of known PPAR pharmacology is recapitulated in the upregulated genes.
Comparison of primate upregulated genes to rodent transcript profiling.
Hepatic transcriptional profiling for a variety of PPAR agonists has been reported in the rat and mouse; however, differences in PPAR agonists used, biological or hepatic effects induced, microarray platforms, statistical methodology, and dosing regimes, to name just a few parameters, make direct comparison with the ciprofibrate primate data problematical (Cherkaoui-Malki et al., 2001; Cornwell et al., 2004; Frederiksen et al., 2003, 2004; Hamadeh et al., 2002; Kramer et al., 2003; Orr MS, personal communication, Gene Logic Inc., Gaithersburg, MD; Yadetie et al., 2003; Yamazaki et al., 2002). Nonetheless, it is possible to draw some broad conclusions, and one of the striking features is that the magnitude of the rodent response for both mitochondrial and peroxisomal ß-oxidation genes is often an order of magnitude greater than that for the primate. Supplementary Table 7 shows the primate and rodent response for five genes related to ß-oxidation; while the primate rarely achieves greater than a 2-fold upregulation, the rodent upregulation is often 10-fold or greater. Peroxisomal ß-oxidation produces H2O2, and the potential for oxidative damage in the rodent appears to be greater than in the primate, based on the strong upregulation of many ß-oxidation genes in the rodent liver.
Metabolic and Biochemical Pathways Downregulated by Ciprofibrate Treatment
Complement and coagulation pathway.
A pathway very strongly downregulated is the complement and coagulation pathway, and this pathway was identified by all mapping tools. Figure 3 shows the KEGG complement and coagulation cascade. This figure shows that the downregulation is both time and dose dependent, in that often only the early timepoint and the highest dose level produced a statistically significant downregulation. Greater detail about the mapping software and additional downregulated pathways are shown in Supplementary Figure 2 and Supplementary Table 2.
|
|
Downregulated genes in the primate liver in the coagulation pathway include fibrinogen, plasma kallikrein B (Fletcher factor), and coagulation factors VII, XI, XII, and XIII; the increase in activated partial thromboplastin time (Supplementary Table 5) in the primates may reflect the downregulation of these genes. In human clinical studies, fibrates have been shown to have an effect on coagulation and fibrinolysis, and human plasma fibrinogen levels can be reduced 12–15% by fibrates (Watts and Dimmitt, 1999
). Thus, the downregulation of many genes in the coagulation cascade in the primate liver is consistent with known pharmacological fibrate effects.
There is also a potent downregulation of many primate genes in the complementation cascade, including complement components (1, 2, 3, 3A, 4A, 4B, 5, 6, 7, 8, 9), complement factor H, complement I factor, and others. There are relatively few literature reports linking complement with PPAR agonists, but the reports that exist focus on the role of complement 3 (C3), hormone-sensitive lipase (HSL), lipoprotein lipase (LPL), and PPAR
(Imbeault et al., 2001; Ylitalo et al., 2002).
Two papers provide sufficient detail to compare primate downregulation in the coagulation and complement pathways to the rat liver (Kramer et al., 2003; Yadetie et al., 2003). Unlike the ß-oxidation genes, the magnitude of the response in the primate and the rat liver appear similar, in that the downregulation in both species is modest, often not exceeding two-fold.
There are several complement cascades, and all can trigger potent inflammatory responses. PPAR agonists have anti-inflammatory characteristics, and the first evidence showing a role in the inflammatory response was the demonstration that PPAR agonists increase the degradation of the pro-inflammatory leukotriene B4 (LTB4) (Devchand et al., 1996). A general mechanism by which PPAR agonists may exhibit anti-inflammatory action appears to be by antagonizing nuclear factor-B (NFB) and activating protein-1 (AP-1) (Delerive et al., 1999, 2001). Many pro-inflammatory genes are controlled by NFB and AP-1, and PPAR agonists may exhibit a spectrum of anti-inflammatory activities by modulating these two key gene products. It is probable that the general anti-inflammatory effect of PPAR agonists is causing a downregulation of the classical complement cascade.
Key regulatory genes.
Many pathways contained a number of downregulated key regulatory genes and transcription factors, such as RELA (also known as NFB), NFKBIA (also called IBa), JUN, and MYC. The pathway mapping of these genes, and others, is given in Supplementary Figure 2.
We observed downregulation of RELA (labeled as NFB in Supplementary Fig. 2 parts 2 and 3), a member of the NFB family, in 4 of 5 ciprofibrate treatment conditions and a downregulation of NFKBIA (given as IBa in Supplementary Fig. 2 part 2) in the high-dose ciprofibrate group. The NFB proteins are a family of transcription factors that, in mammals, consist of five members including RELA. All five members of the NFB family can form homodimeric and heterodimeric combinations, which when activated, bind to the B consensus sequences in the NFB-regulated genes (Yamamoto and Gaynor, 2004). A key regulator of NFB is IB, which is bound to the inactive NFB dimer; once IB is phosphorylated and degraded, the NFB dimer is activated and translocates to the nucleus.
PPAR agonists, including ciprofibrate, will activate NFB in the rat and mouse liver, but not the hamster (Rusyn et al., 1998; Tharappel et al., 2001, 2003). Rats and mice are sensitive to the hepatocarcinogenic effects of PPAR ligands, while hamsters are not. NFB also has shown anti-apoptotic activity in several cell types, including hepatic cell lines. Our gene expression results in the primate (if downregulation reduces the activity of NFB) are consistent with the idea that NFB-modulated effects are promoting cell proliferation and anti-apoptosis in the species sensitive to hepatocarcinogenesis, but not in species insensitive to PPAR-induced liver growth regulating effects.
Ruth Roberts and others have proposed a hypothesis for PPAR agonist-mediated hepatic effects which involves cell division and apoptosis (Boitier et al., 2003; Klaunig et al., 2003; Roberts et al., 2002, 2004). Roberts has proposed a role for both tumor necrosis factor (TNF) and transforming growth factor ß1 (TGFß1). The basic hypothesis is that apoptosis is suppressed and cell proliferation is increased in the rodent. As discussed below in the Supervised Analysis section, in the primate liver we see some indication of increased expression of a number of pro-apoptotic genes and no indication of cell proliferation, which again highlights species differences.
TGFß1 is a key cytokine in the regulation of hepatic apoptosis which acts via the SMAD pathway to alter the transcription of caspases and other genes. Treating rat cells with TGFß1 upregulated a series of genes related to apoptosis and several of the so-called "early response" genes, including JUN (Coyle et al., 2003). Several PPAR agonists, at least in vitro, suppress both spontaneous rodent hepatocyte apoptosis and apoptosis induced by TGFß1 (Bayly et al., 1994; Oberhammer and Qin, 1995). Suppression of apoptosis by PPAR agonists would allow cells that would be normally removed to persist for further mitogenic stimulation.
We saw no alteration of TGFß1 gene expression in any ciprofibrate treatment condition in the primate. Several probesets hybridize to TGFß1, and importantly, one or more probesets was reported as present in all ciprofibrate treatment conditions. Thus, the lack of response of the TGFß1 gene is likely not due to the fact that the primate sequence and human sequence are divergent and thus prevent hybridization on the GeneChip®.
The data regarding modulation of TGFß1 gene expression by PPAR agonists in the rat is mixed. Treatment of rats for up to 2 days with 100 mg/kg/day of nafenopin, a PPAR agonist, caused no change in hepatic TGFß1 expression (Grasl-Kraupp et al., 1998), while the same compound administered for 7 days at 80 mg/kg/day caused more than a 150% increase in TGFß1 mRNA levels (Rumsby et al., 1994). Several reports indicate that a variety of PPAR agonists may increase TGFß1 expression and that greater expression occurs in sensitive species (Lake et al., 2000; Rumsby et al., 1994).
TNF is able to suppress apoptosis and increase proliferation in hepatocytes, and the Kupffer cells, in particular, have been implicated in the production of interleukins and TNF. Hepatocyte growth in response to PPAR agonists can be prevented by antibodies to TNF or the TNF receptor. Some groups reported an increase in rodent TNF gene expression after PPAR treatment; however, this was not seen by others (Klaunig et al., 2003).
In the present study, we saw no change in gene expression for TNF, but the receptor for TNF (member 1A, TNFRSF1A) was downregulated at the early ciprofibrate timepoint. Additional TNF receptors were also downregulated, as is shown in Table 2. The TNF receptor gene expression is dose related, in that only the higher dose levels show a statistically significant downregulation. Thus, in the primate liver we saw no change in expression of TGFß1 or TNF and a downregulation of several TNF receptors, which appears to be in contrast to the rat and mouse.
The c-JUN N-terminal kinases (JNK) form a subgroup of the MAPK superfamily, and JNK can affect expression of c-JUN, JUNB, and JUND by phosphorylation, as well as activation of NFB (Varfolomeev and Ashkenazi, 2004). Several probesets reporting on JUN members were downregulated in the primate liver (Table 2). In rodent hepatocytes, a variety of PPAR agonists cause upregulation of the so-called early response genes, including C-FOS, C-JUN, JUNB, and JUND (Klaunig et al., 2003; Ledwith et al., 1993, 1996). The upregulation of JUN occurs in the rodent hepatocyte cultures within hours; nonetheless, the downregulation of the JUN members in the primate at 4 and 15 days is in contrast to the reported rodent JUN gene upregulation.
c-MYC is another key growth regulatory gene that appears to be moving in opposite directions in the rodent and primate. Several studies have shown that c-MYC gene expression is increased in the rodent liver after PPAR agonist treatment (Miller et al., 1996; Peters et al., 1998). Table 2 shows that a statistically significant downregulation of c-MYC occurred in two treatment conditions in the primate. While statistical significance of c-MYC downregulation only occurred in two of five ciprofibrate treatment condition, all Affymetrix probesets reporting on c-MYC showed downregulation, and Taqman analysis likewise showed downregulation in all treatment groups (data not shown).
Members of several key regulatory networks in the primate liver relating to apoptosis and cell division may be responding differently than in the rodent liver. c-MYC and JUN are downregulated in the primate liver and are reported to be upregulated in the rodent liver. TNF receptors are downregulated in the primate liver, and TNF is able to suppress apoptosis and increase proliferation in cultured hepatocytes. Also, some members of the MAPKinase pathways are downregulated (Supplementary Fig. 2) and may reflect inhibition of mitosis in the primate liver. We believe that the primate data are consistent with the hypothesis that apoptosis is not inhibited in the primate as it is in the rodent and that cell proliferation is not stimulated in the primate liver as it is in the rodent liver. In addition, no indication of increased proliferation was observed in the primate livers, as determined by qualitative assessment of mitotic activity and immunohistochemical staining for Ki-67 (Hoivik et al., 2004).
Protein–protein interaction pathway analysis.
The gene expression data can also be mapped using a different approach, namely protein–protein interactions. The software tool we used for this is a commercial product called the Ingenuity Pathway Analysis (IngenuityTM of Mountain View, CA; http://www.ingenuity.com/index.html). The method used by Ingenuity makes extensive use of the literature and some protein–protein interaction experiments. The output is a map; however, it is unlike classical biochemical maps in that the genes will probably not form a well-recognized biochemical pathway or process. The maps are ranked in terms of probability, and the map least likely to have occurred by chance will contain the largest number of disregulated genes and presumably be of the greatest interest and indicative of biologically relevant effects.
The input for the program was the same set of disregulated genes determined by Rosetta Resolver that we have been using for the other pathway methods for ciprofibrate; all altered probesets for a given treatment condition were submitted to the Ingenuity Pathway AnalysisTM program.
Figure 4 shows the network least likely to have occurred by chance for the 4-day treatment condition is a network centered on MYC. This reinforces our observation above that the regulation of MYC may play an important role in species differences to PPAR-induced hepatic effects. Many of the genes that were noted tangentially in the mapping to conventional pathways are highlighted in the Ingenuity analysis, including genes in the NFB, ERK/MAPK, p38/MAPK, and G1/S Checkpoint pathways. For example, in a previous section we noted that RELA (a member of the NFB family) was downregulated; however, the network in this section highlights the fact that RELA can directly affect c-MYC transcription (Kim et al., 2000; La Rosa et al., 1994). Supplementary Figure 3 shows canonical pathways and provides full gene names.
|
IGF1 and IGF2 were strongly downregulated and appear in the IngenuityTM pathway; however, these genes were not noted in the mapping to classical biochemical pathways. Hyperinsulinemia resulting from insulin resistance leads to increased levels of IGF1, which can promote cell proliferation (Wu et al., 2002
). The mitogenic and anti-apoptotic effects of insulin appear to be mediated through both IGF1 and NFB signaling. IB, a key regulator of NFB, may be a key mediator in insulin resistance as was suggested by recent work demonstrating that high doses of salicylates, which inhibit IB activation, reversed hyperglycemia, hyperinsulinemia, and dyslipidemia in obese rodents by sensitizing insulin signaling (Yuan et al., 2001). PPAR ligands, especially PPAR agonists, exert anti-proliferative, anti-inflammatory effects that may be mediated by antagonizing the activities of NFB (Pineda Torra et al., 1999). Downregulation of IGF1 and IGF2 may be contributing to the anti-inflammatory response seen with PPAR compounds and may serve as a brake on cell proliferation. IngenuityTM reveals a different set of genes than the conventional biochemical mapping. For example, fatty acid oxidation was predominate in the conventional biochemical mapping, but these genes were not highlighted in the IngenuityTM analysis and were often scattered through multiple networks. At present, no single software tool can give a complete view of complex gene expression patterns where thousands of genes can be affected, and only by applying different methods can multiple valid views of the data be realized.
Supervised Analysis: Examination of Genes Related to Apoptosis, DNA Repair, Oxidative Stress, and Cell Proliferation
In an effort to understand species differences in the response to PPAR agonists, we have examined the gene expression in the primate liver by examining genes related to apoptosis, DNA repair, oxidative stress, and cell proliferation. The selection of genes was based on a variety of methods, including Gene Ontology categorization, literature review, and expert knowledge. The probesets used for this analysis are given in Supplementary Table 3, and all probesets were examined for altered regulation. Any probeset with a statistically significant change in gene expression is given in Supplementary Table 4.
Apoptosis.
90 probesets were examined for an apoptosis-related signal, and only 9 probesets showed any disregulation for any treatment condition. The 400-mg/kg/day ciprofibrate treatment at both the 4- and 15-day timepoints produced the largest number of disregulated genes.
Several pro-apoptosis genes were upregulated in the high-dose ciprofibrate treatment groups, including CASP9, CRADD, BNIP3, EI24, NME3, and HTATIP2. CASP9 protein is processed by caspase APAF1 (apoptotic protease activating factor), and this step is thought to be one of the earliest events in the caspase activation cascade (Chen and Wang, 2002). CRADD, also known as RAIDD, binds to the prodomain of caspase-2 and recruits it to the signaling complex (Cohen, 1997). BNIP3L codes for a gene which is a functional homolog of BNIP3 (BCL2/adenovirus E1B 19kDa interacting protein 3), a pro-apoptotic protein. EI24 is a p53 response gene, and overexpression of EI24 suppresses cell growth by inducing apoptotic cell death (Gu et al., 2000). Overexpression of NME3 induced apoptosis in myeloid cells (Venturelli et al., 1995). HTATIP2, also known as CC3 and TIP30, predisposes some cell types to apoptosis (Xiao et al., 2000). The upregulation of these genes may constitute a pro-apoptotic signal.
However, the observed downregulation of TIAL1 and TNFSF10 are not consistent with an increase in apoptosis. The TIAL1 gene product is a cytotoxic granule-associated protein and has been shown to bind specifically to poly(A) homopolymers and to fragment DNA in permeabilized target cells. It has been suggested that members of the TIAL protein family may be involved in the induction of apoptosis (Kawakami et al., 1992). Likewise TNFSF10, also known as TRAIL, can induce apoptosis in a wide variety of transformed cell lines of diverse lineage; however it does not appear to be capable of inducing cell death in normal cells (Degli-Esposti et al., 1997; Kawakami et al., 1992).
Taken together, we believe that the gene expression indicates a pro-apoptotic signal, especially for the animals treated for 15 days with 400 mg/kg/day of ciprofibrate; this is consistent with the subcapsular single-cell necrosis observed microscopically. Animals treated with this dose level for 4 days show a weaker signal.
DNA repair.
148 probesets were selected as indicative of DNA repair, and 10 probesets were disregulated in at least one treatment condition. The signal for DNA repair is relatively unconvincing, in that (1) 6 of the 10 probesets only showed a disregulation at a single treatment condition, and (2) many disregulated genes are not altered in a dose-responsive manner.
If a DNA repair signal exists, it may be reflected in the response of APEX1 and DNAJA1. APEX1 expression was increased in three of the four 15-day ciprofibrate dose groups. APEX1 codes for the major apurinic/apyrimidinic (AP) endonuclease in human cells. AP sites can occur in DNA by spontaneous hydrolysis, by DNA damaging agents, or by DNA glycosylases that remove specific abnormal bases (Dianov et al., 2003).
DNAJA1 is a homolog of the bacterial DNAJ gene (also known as HSP40, Heat Shock Protein 40). The function of DNAJA1 in mammalian cells in poorly understood, but in prokaryotes, DNAJ functions as a molecular chaperone, and DNAJ is induced in response to cell stress (Macario et al., 1999). DNAJA1 was upregulated in three of the four 15-day ciprofibrate dose groups. However, taken together, our data does not show a clear indication of DNA repair.
Oxidative stress.
51 probesets were selected as being indicative of oxidative stress, and 13 probesets were altered at one or more treatment condition. Catalase (CAT) expression was reduced for one treatment condition, and glutathione peroxidase (GPX1) was downregulated for two treatment conditions; no treatment condition increased the expression of CAT or GPX1. Several publications have shown that transcription of both CAT and GPX1 can be increased after oxidative stress in a variety of tissues (Limaye et al., 2003; Nemali et al., 1988; Rohrdanz et al., 2001a,b).
The signal for oxidative stress is not convincing in that (1) the expression of several important genes (CAT and GPX1) has been repressed, (2) upregulation and downregulation of the same gene occurred for different treatment conditions (MT1F and MT2A), and (3) there was a lack of dose-response.
Furthermore, several of the genes that did show a statistically significant change in gene expression may also respond to cellular phenomenon other than oxidative stress. For example, HSPA9B was upregulated in four treatment groups. However, HSPA9B (also called mortalin-2) is a member of the heat shock 70 family, and mortalin has been shown to be involved in stress response, muscle activity, mitochondrial biogenesis, intracellular trafficking, antigen processing, control of cell proliferation, differentiation, cell fate determination, and tumorigenesis (Kaul et al., 2002) and, consequently, may be upregulated by the pharmacological activity of ciprofibrate rather than oxidative stress.
Likewise, while metallothionein certainly can be induced by reactive oxygen species, it can also be induced by glucocorticoids and induced in the liver by inflammation and infection (Ghoshal and Jacob, 2001). There are two probesets reporting on different metallothionein genes, and each probeset shows some treatment conditions with an upregulation of the gene and other treatment conditions with a downregulation. Taken together, we believe the data do not suggest a significant response to oxidative stress.
Regulators of cell cycle and proliferation.
99 probesets were annotated as being indicative of positive regulation of the cell cycle and cellular proliferation. 11 probesets had a statistically significant alteration in gene expression at one or more treatment conditions.
Probesets reporting on insulin-like growth factor 1 (IGF1) showed a very marked downregulation at high doses of ciprofibrate. IGF1 plays a critical role in cell survival, proliferation, and terminal differentiation for a variety of different cell types, and IGF1 is produced by the liver in response to growth hormone stimulation (Benito et al., 1996; Grimberg and Cohen, 2000). IGF1 plays an important role in prevention of apoptosis and cell survival signaling pathways (Vincent and Feldman, 2002). IGF1 has been shown to protect cells from apoptosis induced by a variety of stimuli (Murray et al., 2003). Signaling through the IGF system plays an important role in prevention of apoptosis in a variety of cell types (Vincent and Feldman, 2002). Growth of many cell types in culture is stimulated by IFG1 (Benito et al., 1996), and there are several reports that hepatocytes in culture will proliferate when treated with IGF1 (Kimura and Ogihara, 1998; Kundu et al., 2003). Therefore, the gene expression results with ciprofibrate for IFG1 are consistent with an anti-proliferative and perhaps a pro-apoptotic effect.
VEGF was downregulated at the highest dose level of ciprofibrate at 4 and 15 days. Increased expression of VEGF is well known to be associated with a variety of tumor types, including human hepatocellular carcinomas, where angiogenic factors appear to play an important role in the development of hepatocellular carcinomas and liver metastasis by stimulating vascularization (Kim et al., 2002; Takeda et al., 2002). Downregulation of VEGF is consistent with an anti-proliferative effect.
CAPNS1 was upregulated at the highest dose of ciprofibrate (400 mg/kg/day) at the 15-day timepoint. Calpains form part of the proteolytic pathways that play a role in apoptosis (Solary et al., 1998). Treating rat hepatocytes with an agent that induces oxidative stress led to apoptosis by activation of the calpain pathway (Ding and Nam Ong, 2003). This is consistent with a pro-apoptotic anti-proliferative effect. Taken together, we believe that the gene expression results are not consistent with a proliferative effect, but rather an anti-proliferative and pro-apoptotic effect.
Correlation of Clinical Chemistry Data to Gene Expression on an Individual Animal Basis
The analysis in this manuscript to this point has been done by comparing gene expression in a treated group to gene expression in the untreated group; using this method, a single fold-change value for each treatment group was produced. However, a finer level of granularity can be obtained by an examination of gene expression on an individual animal basis. Supplementary Table 5 gives toxicokinetic, clinical chemistry, and liver biochemistry data on a per animal basis, and we use the activated partial thromboplastin time (PTT) to illustrate how gene expression data may be correlated to other endpoints.
PTT measures the clotting time from the activation of factor XII through the formation of fibrin clot, and this measures the integrity of the intrinsic and common pathways of coagulation. In the current study, plasma PTT increased with ciprofibrate dose, indicating that the drug is probably reducing one or more of the clotting factors, such as VIII, IX, XI, and/or XII. Figure 5A shows the gene expression intensity for several thousand genes on an individual animal basis, and we have added a user-defined trend that matches the PTT for each animal.
|
PTT generally increased with dose, and since many of the clotting factors are made in the liver, we chose to examine probesets that have gene expression patterns that are the opposite of the PTT pattern (i.e., anti-correlated gene expression patterns; these gene expression profiles will generally decrease with dose). The gene expression pattern for 77 probesets were anti-correlated with PTT, and the shape of the gene expression for these probesets is given in Figure 5B, and the probesets themselves are given in Supplementary Table 6.
Many genes related to complement and coagulation were identified by this approach. Genes previously mapped to the complement and coagulation cascades (Fig. 3) and also identified by the individual animal method include: BF (B-factor, properdin), C2 (complement component 2), C4A/C4B (complement component 4A and 4B), C8A (complement component 8, alpha polypeptide), IF (I factor–complement), KLKB1 (kallikrein B, plasma (Fletcher factor) 1), and PROC (protein C (inactivator of coagulation factors Va and VIIIa)). Genes identified only by the individual animal approach include C1S (complement component 1, s subcomponent), CFHL4 (complement factor H-related 4), FGL1 (fibrinogen-like 1), and HMOX1 (heme oxygenase (decycling) 1); the genes uniquely identified may be candidates for further investigation into PPAR-mediated effects on coagulation and clotting. Again, this illustrates the need to apply different methods to analyze complex patterns of gene expression.
Conclusion
In conclusion, this is the first report of large-scale hepatic transcriptional profiling in PPAR-ligand-treated nonhuman primates. We applied a variety of analysis methods, both unsupervised and supervised, with emphasis on network and pathway analysis. We also compared the primate data to literature reports of hepatic transcriptional profiling in PPAR-ligand-treated rodents, which showed that the magnitude of induction in ß-oxidation pathways was substantially greater in the rodent than the primate. Upregulation of genes relating to fatty acid metabolism, mitochondrial oxidative phosphorylation, ribosomal machinery, and proteasome biosynthesis was observed in the primate. A large number of genes downregulated were in the complement and coagulation cascades, as well a number of key regulatory genes, including members of the JUN, MYC, and NFB families. The downregulation of some key regulatory genes appears to be in contrast to the rodent. In addition, we saw no signal for DNA damage or oxidative stress, while we did observe gene expression consistent with an anti-proliferative and a pro-apoptotic effect.
Primates were treated with both ciprofibrate and fenofibrate in Hoivik et al. (2004), and here we report transcriptional profiling only for the ciprofibrate-treated animals. We omitted the fenofibrate-treated animals from this publication for several reasons, namely, (1) the transcriptional response for ciprofibrate was more robust than fenofibrate, (2) the fenofibrate dataset appears to be largely a subset of the ciprofibrate dataset, and (3) consideration of manuscript size and complexity.
|
SUPPLEMENTARY DATA |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Supplementary data are available online at www.toxsci.oupjournals.org.
|
NOTES |
|---|
2 Current Address: Bayer HealthCare LLC, 85 T.W. Alexander Drive, Research Triangle Park, North Carolina 27709.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUPPLEMENTARY DATAREFERENCES |
|---|
Adams, J. (2003). The proteasome: Structure, function, and role in the cell. Cancer Treat. Rev. 29(Suppl. 1), 3–9.[ISI][Medline]
Anderson, S. P., Howroyd, P., Liu, J., Qian, X., Bahnemann, R., Swanson, C., Kwak, M. K., Kensler, T. W., and Corton, J. C. (2004). The transcriptional response to a peroxisome proliferator-activated receptor alpha agonist includes increased expression of proteome maintenance genes. J. Biol. Chem. 279, 52390–52398.
Bach, I., and Ostendorff, H. P. (2003). Orchestrating nuclear functions: Ubiquitin sets the rhythm. Trends Biochem. Sci. 28, 189–195.[CrossRef][ISI][Medline]
Bayly, A. C., Roberts, R. A., and Dive, C. (1994). Suppression of liver cell apoptosis in vitro by the non-genotoxic hepatocarcinogen and peroxisome proliferator nafenopin. J. Cell Biol. 125, 197–203.
Benito, M., Valverde, A. M., and Lorenzo, M. (1996). IGF-I: A mitogen also involved in differentiation processes in mammalian cells. Int. J. Biochem. Cell Biol. 28, 499–510.[CrossRef][ISI][Medline]
Bentley, P., Calder, I., Elcombe, C., Grasso, P., Stringer, D., and Wiegand, H. J. (1993). Hepatic peroxisome proliferation in rodents and its significance for humans. Food Chem. Toxicol. 31, 857–907.[CrossRef][ISI][Medline]
Boitier, E., Gautier, J. C., and Roberts, R. (2003). Advances in understanding the regulation of apoptosis and mitosis by peroxisome-proliferator activated receptors in pre-clinical models: Relevance for human health and disease. Comp. Hepatol. 2, 3.[CrossRef][Medline]
Cattley, R. C., DeLuca, J., Elcombe, C., Fenner-Crisp, P., Lake, B. G., Marsman, D. S., Pastoor, T. A., Popp, J. A., Robinson, D. E., Schwetz, B., Tugwood, J., and Wahli, W. (1998). Do peroxisome proliferating compounds pose a hepatocarcinogenic hazard to humans? Regul. Toxicol. Pharmacol. 27, 47–60.[CrossRef][ISI][Medline]
Chen, M., and Wang, J. (2002). Initiator caspases in apoptosis signaling pathways. Apoptosis 7, 313–319.[CrossRef][ISI][Medline]
Cherkaoui-Malki, M., Meyer, K., Cao, W. Q., Latruffe, N., Yeldandi, A. V., Rao, M. S., Bradfield, C. A., and Reddy, J. K. (2001). Identification of novel peroxisome proliferator-activated receptor alpha (PPARalpha) target genes in mouse liver using cDNA microarray analysis. Gene Expr. 9, 291–304.[ISI][Medline]
Chismar, J. D., Mondala, T., Fox, H. S., Roberts, E., Langford, D., Masliah, E., Salomon, D. R., and Head, S. R. (2002). Analysis of result variability from high-density oligonucleotide arrays comparing same-species and cross-species hybridizations. Biotechniques 33, 516–524.[ISI][Medline]
Cohen, G. M. (1997). Caspases: The executioners of apoptosis. Biochem. J. 326, 1–16.[ISI][Medline]
Cornwell, P. D., De Souza, A. T., and Ulrich, R. G. (2004). Profiling of hepatic gene expression in rats treated with fibric acid analogs. Mutat. Res. 549, 131–145.[ISI][Medline]
Corton, J. C., Anderson, S. P., and Stauber, A. (2000). Central role of peroxisome proliferator-activated receptors in the actions of peroxisome proliferators. Annu. Rev. Pharmacol. Toxicol. 40, 491–518.[CrossRef][ISI][Medline]
Coyle, B., Freathy, C., Gant, T. W., Roberts, R. A., and Cain, K. (2003). Characterization of the transforming growth factor-beta 1-induced apoptotic transcriptome in FaO hepatoma cells. J. Biol. Chem. 278, 5920–5928.
de Lange, P., Ragni, M., Silvestri, E., Moreno, M., Schiavo, L., Lombardi, A., Farina, P., Feola, A., Goglia, F., and Lanni, A. (2004). Combined cDNA array/RT-PCR analysis of gene expression profile in rat gastrocnemius muscle: Relation to its adaptive function in energy metabolism during fasting. FASEB J. 18, 350–352.
Degli-Esposti, M. A., Dougall, W. C., Smolak, P. J., Waugh, J. Y., Smith, C. A., and Goodwin, R. G. (1997). The novel receptor TRAIL-R4 induces NF-kappaB and protects against TRAIL-mediated apoptosis, yet retains an incomplete death domain. Immunity 7, 813–820.[CrossRef][ISI][Medline]
Delerive, P., De Bosscher, K., Besnard, S., Vanden Berghe, W., Peters, J. M., Gonzalez, F. J., Fruchart, J. C., Tedgui, A., Haegeman, G., and Staels, B. (1999). Peroxisome proliferator-activated receptor alpha negatively regulates the vascular inflammatory gene response by negative cross-talk with transcription factors NF-kappaB and AP-1. J. Biol. Chem. 274, 32048–32054.
Delerive, P., Fruchart, J. C., and Staels, B. (2001). Peroxisome proliferator-activated receptors in inflammation control. J. Endocrinol. 169, 453–459.[Abstract]
Devchand, P. R., Keller, H., Peters, J. M., Vazquez, M., Gonzalez, F. J., and Wahli, W. (1996). The PPARalpha-leukotriene B4 pathway to inflammation control. Nature 384, 39–43.[CrossRef][Medline]
Dianov, G. L., Sleeth, K. M., Dianova, II, and Allinson, S. L. (2003). Repair of abasic sites in DNA. Mutat. Res. 531, 157–163.[ISI][Medline]
Ding, W. X., and Nam Ong, C. (2003). Role of oxidative stress and mitochondrial changes in cyanobacteria-induced apoptosis and hepatotoxicity. FEMS Microbiol. Lett. 220, 1–7.