Abstract
Peroxisome proliferator-activated receptor gamma (PPARγ) agonists of the thiazolidinedione family are used for the treatment of type 2 diabetes mellitus due to their ability to reduce glucose and lipid levels in patients with this disease. Three thiazolidinediones that were approved for treatment are Rezulin (troglitazone), Avandia (rosiglitazone), and Actos (pioglitazone). Troglitazone was withdrawn from the market due to idiosyncratic drug toxicity. Rosiglitazone and pioglitazone are still on the market for the treatment of type 2 diabetes. The authors present data from a gene expression screen that compares the impact these three compounds have in rats, in rat hepatocytes, and in the clone 9 rat liver cell line. The authors monitored the changes in expression in multiple genes, including those related to xenobiotic metabolism, proliferation, DNA damage, oxidative stress, apoptosis, and inflammation. Compared to the other two compounds, troglitazone had a significant impact on many of the pathways monitored in vitro although no major perturbation was detected in vivo. The changes detected predict not only general toxicity but potential mechanisms of toxicity. Based on gene expression analysis, the authors propose there is not just one but multiple ways troglitazone could be toxic, depending on a patient’s environment and genetic makeup, including immune response-related toxicity.
Peroxisome proliferator-activated receptor gamma (PPARγ) is a member of the steroid/thyroid nuclear receptor superfamily (Evans 1988) and is important in coordinating energy systems, including glucose and lipid metabolism. Binding of PPARγ agonists such as the thiazolidinediones results in increased insulin sensitivity and lower glucose levels in patients with type 2 diabetes by enhancing glucose metabolism in muscle and decreasing glucose biosynthesis in liver (Spiegelman 1998; Saltiel and Olefsky 1996). Three thiazolidinedione-based glitazones, Rezulin (troglitazone), Avandia (rosiglitazone), and Actos (pioglitazone), are all PPARγ agonists that have been used successfully for the treatment of type 2 diabetes in humans (Spiegelman 1998; Willson et al. 2000; Saltiel and Olefsky 1996). Of the three compounds, troglitazone alone has exhibited significant adverse effects.
A small percentage of patients receiving troglitazone developed hepatotoxic reactions including fulminant (Murphy et al. 2000), subacute (Fukano et al. 2000), and late-onset (Bell and Ovalle 2000), some of which resulted in hepatic failure (Gitlin et al. 1998; Neuschwander-Tetri et al. 1998; Vella, de Groen, and Dinneen 1998; Misbin 1999; Herrine and Choudary 1999). Due to the severe hepatotoxicity in some cases that resulted in death, and the availability of pioglitazone and rosiglitazone for the treatment of type 2 diabetes, troglitazone was removed from the market on March 21, 2000. During clinical testing, 1.9% of the patients receiving troglitazone developed reversible elevations in serum alanine aminotransferase (ALT), yet no hepatic toxicity was detected in any of the animals tested, including monkeys, which show similar metabolic profiles as humans (Watkins and Whitcomb 1998; SBA 1997). There was no association between elevated ALT levels, sex, age, or amount of troglitazone taken. These patients did not have any symptoms such as fever or rash that would suggest an immunological reaction. As a consequence, the hepatotoxicity associated with troglitazone is unexplained and referred to as idiosyncratic.
However, different studies monitoring the impact of troglitazone on various genes in different models have shown how it may be toxic through multiple physiological pathways. MCF-7 cells driven to apoptosis by troglitazone were found to have elevated levels of GADD45, and when this gene was silenced by RNA intereference (RNAi), MCF-7 cells were refractory to cell death, supporting a role for this gene in troglitazone mediated apoptotic events (Yin et al. 2004). The elevated expression of a second DNA damage response gene, GADD153, was correlated to troglitazone-induced growth inhibition of non-small cell lung carcinoma cells, and suppression of GADD153 gene expression with antisense oligonucleotides attenuated the troglitazone driven growth inhibition (Satoh et al. 2002). Another study showed increased expression of cell cycle arrest-associated genes p53, p27, and p21, coinciding with reduced levels of cyclin D1 expression in hepatoma cells driven to apoptosis by troglitazone, when compared to untreated or cells treated with rosiglitazone (Bae, Rhee, and Song 2003). Cytochrome induction has also been implicated as a possible mechanism for troglitazone’s toxicity. Troglitazone was found to increase the levels of CYP3A in rats, resulting in enhanced toxicity of acetaminophen when they were administered together as evidenced by increased levels of ALT, aspartate aminotransferase, and alpha-glutamine S-transferase in the plasma of the rats and low glutathione concentration in the liver (Li et al. 2002).
The purpose of this study was to examine in further detail specific gene expression responses that would explain the potential different mechanisms responsible for the idiosyncratic toxicity of troglitazone versus pioglitazone and rosiglitazone. In vitro analysis was performed using the rat clone 9 (C9) hepatocyte-derived cell line and primary rat hepatocytes treated with different glitazone compounds to monitor response differences that could support the effects reported above and suggest one or more mechanisms of action that may result in troglitazone’s toxicity.
MATERIALS AND METHODS
Chemicals
The compounds used to treat rat C9 cells and hepatocytes were prepared as follows: troglitazone, rosiglitazone, and pioglitazone were purchased as a set from Cayman Chemical Corporation (Ann Harbor, MI; catalog no. 71000) and resuspended as 10 mM solutions in DMSO. The compounds and DMSO vehicle controls were then prepared at the final concentrations in the appropriate cell culture medium for each cell type as described below.
Treatment of Clone 9 Rat Liver Cell Line and Rat Hepatocytes
Normal C9 rat liver cells (ATCC, Manassas, VA; catalog no. CRL-1439) were maintained in log growth phase in RPMI 1640 medium supplemented with 10% fetal bovine serum and antibiotics (all tissue culture reagents were obtained from HyClone, Logan, UT). One hundred thousand cells per well were plated onto a 96-well tissue culture plate and incubated at 37°C in a humidified incubator at 5% CO2. The cells were allowed to stabilize and attach for 18 h before treatment.
Freshly isolated rat hepatocytes were obtained from CellzDirect (Pittsboro, NC; catalog no. RTP050-96). The cells were isolated from a 10-week-old male Sprague-Dawley rat. The cell viability after isolation was 86%. The hepatocytes were plated to confluence onto a collagen type 1 substratum with a Matrigel overlay in a 96-well plate. The plate of cells was shipped to our laboratory on ice. Upon receiving the cells, the preservation medium was replaced with cold serum-free hepatocyte maintenance medium which consisted of modified Chee’s medium supplemented with 1 mg/ml ITS+ Premix containing insulin, human transferrin, and selenous acid (BD Biosciences, Franklin Lakes, NJ; catalog no. 354352), dexamethasone at a final concentration of 0.1 μM, 50 units/ml penicillin, and 50 μg/ml streptomycin. The cells were allowed to acclimate for 24 h in a humidified incubator at 37°C with 5% CO2.
C9 and rat hepatocyte cells were treated in triplicate with 10, 50, or 100 μM final concentrations of troglitazone, rosiglitazone, and pioglitazone and their corresponding DMSO-only vehicle controls, which were 0.1%, 0.5%, and 1.0%, respectively. The compounds were prepared by diluting the 10 mM compounds in the appropriate cell culture medium. The overnight acclimation medium was gently aspirated from the wells and replaced with the above compounds at their final concentrations by adding 200 μl to the wells very slowly. A set of wells were given medium only as 0-h and 24-h medium controls. The 0-h control wells were aspirated, lysed by the addition of 100 μl SV96 Lysis Buffer (Promega, Madison, WI; SV96 Total RNA Isolation Kit, catalog no. Z3505) and frozen at −20°C until RNA isolation. The treatment plates were placed back in the incubator for 24 h.
Following the 24-h incubation, the medium was aspirated and the cells were lysed by the addition of 100 μl SV96 Lysis Buffer (Promega). Lysates were transferred to another plate and stored at −20°C until RNA isolation.
Isolation of RNA
Total RNA from the treated cell culture lysates was isolated and DNased using the SV96 Total RNA Isolation Kit (Promega catalog no. Z3505). RNA concentrations were determined by a RiboGreen assay (Molecular Probes, Carlsbad, CA; catalog no. R11490).
Gene Expression Analysis
The gene expression patterns of 23 genes were examined by subjecting 25 ng of total RNA from each of the above samples to the eXpress Profiling Multiplex RT-PCR Assay. Briefly, 25 ng of RNA from each sample was reverse transcribed using 23 reverse primers. Each reverse primer is chimeric with the 19 nucleotides of the 5′ primer end containing a sequence from a T7 bacteriophage genome that provides a template for the universal reverse primers in the polymerase chain reaction (PCR) reaction. The 3′ primer end contains the gene specific sequence. Each of the primer pairs were designed to yield PCR products 7 base pairs apart, ranging from 149 base pairs up to 348 base pairs. A similarly designed 24th primer is also included that binds to a 1200–base pair kanamycin RNA transcript (Promega) that is spiked into each reverse transcription reaction and serves as an internal reaction integrity control. The reverse transcription reactions were carried out with 20 units of Moloney Murine Leukemia Virus (MMuLV) reverse transcriptase (ABgene, Rochester, NY; catalog no. AB-0322/B); PCR Buffer II (ABI, Foster City, CA; catalog no. N8080006) containing 10 mM Tris-HCl and 50 mM KCl; 2.5 mM MgCl2 (USB, Cleveland, OH; catalog no. 78641); 10 mM dithiothreitol (EMD Biosciences, San Diego, CA; catalog no. 69714); and 1 mM of each dNTP (USB, catalog no. 77119). The concentration of each primer varied from 0.0004 to 0.05 μM in order to adjust the final signals of each amplified gene. The reverse transcription reactions were incubated for 1 min at 48°C, 5 min at 37°C, 60 min at 42°C, and then 5 min at 95°C. The 20-μl reactions were carried out in a Thermo-Fast 96-well PCR Detection Plate (ABgene, catalog no. AB-1100).
Multiplex PCR was then performed on each sample as follows: 10 μl of cDNA from each above reverse transcription reaction was added to the wells of a new 96-well PCR plate and 10 μl of a PCR reaction mix containing the 24 chimeric forward primers at 2 μM each. Each forward primer is chimeric with the 18 nucleotides of the 5′ primer end containing a sequence of the SP6 bacteriophage genome that provides a template for the universal forward primers in the PCR reaction. The 3′ primer end contains the gene-specific sequence. Other PCR reaction constituents included the kanamycin forward chimeric primer; 1 μM T7 Universal reverse primer (19 nucleotides); 1 μM D4 dye–labeled SP6 Universal forward primer (18 nucleotides); PCR Buffer II (ABI, catalog no. N8080006) containing 10 mM Tris-HCl and 50 mM KCl; 7 mM MgCl2 (USB, catalog no. 78641); 0.3 mM each of dNTPs (USB, catalog no. 77119); and 3.5 units of Taq Polymerase (ABgene, catalog no. AB-0908/b). The reactions were first subjected to 95°C for 10 min followed by 35 PCR cycles. Each PCR cycle consisted of the following conditions: 94°C for 30 s, 55°C for 30 s, and 68°C for 1 min.
The PCR products were then prepared for capillary electrophoresis by adding 1 μl of each reaction to its corresponding well in a Beckman 96-well CEQ electrophoresis plate (Beckman, Fullerton, CA; catalog no. 609801) containing 39μl of CEQ Sample Loading Solution (Beckman catalog no. 608082) and 1 μl of CEQ DNA Size Standard 400 (Beckman, catalog no. 608098). The samples were mixed and placed in a Beckman CEQ 8800 for capillary electrophoresis and fragment size analysis. The fragment results from the CEQ 8800 were analyzed on Althea Technologies, eXpression Profiling Analysis Software. This software associates each PCR product with its corresponding gene and reports its peak area. The gene expression data were normalized by dividing the peak area result for each gene by the peak area result of either glyceraldehyde phosphate dehydrogenase (GAPDH), or cyclophilin A. These expression ratios were then compared to the corresponding vehicle control to ascertain whether any gene expression changes were due solely to the compound or to the DMSO vehicle. Comparisons were also made between the 0-h and 24-h medium-only controls to ascertain gene expression changes due to cell growth or any changes in the culture conditions.
Verification of Selected Multiplex PCR Results
eXpress Profiling-based changes in expression of several genes were verified using real time PCR. A Taqman probe was designed for each of the four assays (GADD45, GADD153, NQO-1, and p21). The probes are located in a region directly between the forward and reverse primers. The sequences of the primers used were the same sequences of the primers used in the eXpress Profiling assay minus the universal tail sequences. cDNA was generated using a random hexamer–primed synthesis (5 μM random hexamers, 1 × PCR buffer, 5.5 mM MgCl2, 500 μM dNTPs, 0.5 units RNase inhibitor, and 1.5 units MMuLV reverse transcriptase). A total of 10 μl of cDNA was used in the Taqman PCR reactions (1 × Universal Master Mix, 300 nM gene specific forward and reverse primers, and 100 nM gene specific probe).
Analysis of the Taqman assay data was based on comparison to gene-specific (GADD45, GADD153, NQO-1, and p21) threshold cycle (Ct) values relative to cyclophilin A Ct values. A delta-delta Ct comparison was performed for each gene relative to cyclophilin A using a medium control as a baseline Ct value.
In order to compare eXpress Profiling values to Taqman Real Time PCR values, normalization of target genes was utilized. A normalization ratio of the target gene signal to a media control was first calculated, including the target gene values and adjusted according to the obtained ratios. This initial normalization was performed for each target gene. A subsequent normalization of the adjusted target gene values was performed relative to cyclophilin A.
RESULTS
The gene expression levels of 23 genes associated with specific mechanisms were monitored using multiplexed reverse transcriptase (RT)-PCR. The specific genes are listed in Table 1. The genes selected fall into several categories including the drug metabolizing enzymes and related genes CYP2B1, CYP3A1, CYP1A1, CYP4A1, CYP2E1, CYP2D4, CYP1A2, NADPH CYP, and UGTB1, stress response–related genes heme oxygenase (HMOX-1) and NADPH:quinone reductase (NQO); cell cycle and cell proliferation–related genes PCNA, p53, p21, and cyclin D1; DNA damage indicators GADD45 and GADD153; inflammatory marker COX-2; and apoptotic gene caspase-3. In an earlier study by our group, in vivo gene expression analysis of these 19 genes in rat liver treated with troglitazone, rosiglitazone, and pioglitazone showed no significant changes in gene expression (data not shown), and supported the null findings in preclinical studies for troglitazone.
The analysis was then extended to treatments in two in vitro models, including primary rat hepatocytes and the rat hepatocyte-derived C9 cell line. In preliminary studies, cells were treated with a range of concentrations of compounds from levels that had no impact on gene expression, to those that resulted in cell death, which was monitored by morphology and overall gene expression. A concentration of 100 μM for the C9 cell line and 50 μM for the primary hepatocytes was found to have the most profound impact on gene expression for the majority of genes of interest and this was supported by previous studies of troglitazone in a liver-derived cell line (Yamamoto et al. 2002) and rat hepatocytes (Davies et al. 2002). Sprague-Dawley rat hepatocytes were obtained from CellzDirect in a 96-well plate format with a collagen and matrigel overlay at approximately 75,000 cells per well. The rat hepatocyte were incubated for 24 h in CellzDirect-recommended medium and then treated for 24 h with 10, 50, and 100 μM of each glitazone, and then harvested for RNA isolation. The gene expression results for the primary hepatocyte 50 μM treatments are shown in Figure 1. The most significant gene expression responses observed from all three glitazone treatments were the inductions of the cytochromes 2B1 and 3A1 as well as the cytochrome-related NADPH cytochrome reductase (Figure 1). The thiazolidinediones had little impact on CYP1A1 and CYP1A2. Similar results had been seen for troglitazone in a previous study (Sahi et al. 2000). Induction of these cytochromes or a different subtype not monitored here may lead to activation of troglitazone or one of its metabolites into a more toxic form (Prabhu et al. 2002) or may result in increased generation of damaging radical oxygen species (Paolini et al. 2001) compared to the other glitazones. Two gene expression changes in hepatocyte cells treated with troglitazone that were different from the pioglitazone, and rosiglitazone-treated cells support this possibility (Figure 1C). Heme oxygenase, which is a key indicator of increased oxidative stress (Tyrrell and Basu-Modak 1994; Horvath et al. 1998; Cossi et al. 2001; Kamalvand, Pinard, and Ali-Khan 2003), was induced more by troglitazone, and NQO reductase was also more elevated, which is a signal of increased quinone epoxide radicals. These differences were more pronounced at higher concentrations (Figure 2). Also, the GADD genes, which can be induced by radical oxygen species, were induced by troglitazone. Conversely, CYP2B1 showed little induction by troglitazone as compared to the other two compounds (Figure 1), and so any direct toxic effect of troglitazone may be exacerbated due to reduced cytochrome driven troglitazone metabolism.
The treatment of primary hepatocytes was followed with the glitazone treatment of the hepatocyte-derived C9 cell line to determine if any potential toxicity could be predicted by looking at the impact of these compounds on gene expression in a proliferating cell line. Pioglitazone and rosiglitazone had a small impact on the expression of a few of the genes in the assay, including a 1.5-fold increase in heme oxygenase and a slight increase in NQO (Figure 3A and 3B). Troglitazone, however, had a strong impact on many genes, including ones that had been impacted by this compound in other cell lines (Figure 3C). GADD45 and GADD153 were both strongly induced, indicating potential DNA damage. Troglitazone also had a stronger impact on heme oxygenase, resulting in double the induction caused by rosiglitazone and pioglitazone. There was also a very strong induction of the quinone reducer NQO and the proinflammatory gene COX-2 (Figure 2C). Troglitazone also substantially upregulated the proliferation inhibitory associated gene p21. Verification of the GADD, NQO, and p21 C9 results was carried out using real time PCR (Figure 4).
DISCUSSION
In this study we sought to determine if data obtained from monitoring the impact of troglitazone treatment of cells in culture with gene expression analysis would suggest in vivo drug toxicity. This study included performing the same analysis with the currently available glitazones, pioglitazone and rosiglitazone, to compare potential toxicity profiles of the three drugs. Ours results suggest that the impact of troglitazone in vitro on the expression of the genes monitored reveal a profile of potential toxicity that is not apparent when analyzing the gene profiles produced by rosiglitazone and pioglitazone in vitro. Additionally, the potential toxicity of troglitazone was not apparent from the in vivo studies (data not shown).
Specifically we have identified expression patterns of genes in vitro representing multiple pathways that are predictive of potential toxicity in vivo. Our in vitro experiments utilized primary rat hepatocytes and rat hepatocyte-derived C9 cells. Rat hepatocytes were used to observe the glitazone’s impact in metabolically active liver cells and the C9 cell line was utilized to see the effect of glitazones on a proliferating cell type. C9 cells are an epithelial cell line isolated from a normal liver of a young, male rat. All cells were plated in a 96-well plate format and all treatments were treated at concentrations of 10, 50, and 100 μM for 24 h. After 24 h of treatment, cells were harvested, the total RNA from each well extracted, and expression levels of the selected genes determined by multiplex PCR reactions that were also carried out with RNA from untreated control C9 and rat hepatocyte cells for comparison.
The gene expression screen that was performed in vitro with troglitazone predicts that the drug could be toxic by a variety of mechanisms. One of the current explanations of idiosyncratic toxicity is the “multiple determinant hypothesis.” The hypothesis states that idiosyncratic toxicity requires the occurrence of multiple critical and discrete events with the probability of an idiosyncratic drug toxicity event is the product of the probabilities of each event (Li 2002). This hypothesis as described by P (probability of troglitazone toxicity) = P 1 × P 2 × P 3 × P 4 × · × P n wherein P n values represent the probability that multiple independent biochemical events could synergistically contribute to the mechanisms by which troglitazone becomes toxic in a small percentage of the human population.
The gene expression data presented here provide support for the multiple determinant hypothesis with different gene responses, indicating a plurality of ways troglitazone could be toxic in small population subsets. Of these events, one of the most straightforward mechanisms of idiosyncratic troglitazone toxicity could result from direct action of the glitazone itself. A combination of troglitazone toxicity and low activity of the major enzymes that metabolize it could be a triggering event. Major metabolites that are formed from troglitazone are sulfate and quinone, with troglitazone sulfate accounting for approximately 70% of the metabolites found in human plasma (Shibata et al. 1993; Loi et al. 1997). Honma and colleagues found troglitazone caused concentration-dependent lactate dehydrogenase leakage in hepatoma cells, whereas the troglitazone metabolites (sulfate, glucuronide, and quinone forms) caused insignificant leakage (Honma et al. 2002). For individuals with reduced or less active enzymes in the sulfation pathway, standard dose therapy could lead to prolonged exposure to troglitazone at higher concentrations and/or increased levels of quinone metabolite, creating a toxic event within this population. Thus theoretically, sulfation inactivity could be the principal event that predisposes some individuals to troglitazone idiosyncratic toxicity. Similarly, certain cytochromes could be important for thiazolidine metabolism and lack of activation could result in increased toxicity. In rat hepatocyte cells, CYP2B1 was induced 2.5 to 4 times more than troglitazone by pioglitazone and rosiglitazone, respectively. Little or no induction was seen from troglitazone. Higher levels of unmetabolized troglitazone could contribute to the toxicity scenario described above.
Cytochrome activity, including induction, can also be the beginning of a series of events leading to toxicity. The impact of the glitazones on the metabolizing enzymes cytochrome 2B1 and cytochrome 3A1 (CYP3A1), especially CYP3A1, was apparent in the treated primary hepatocytes. Cytochrome 1A1’s expression level was not effected. Cytochrome induction detoxifies, but could also lead to toxicity in some situations. Induction of cytochromes by xenobiotics allows for an adaptive response to exposure to noxious substances. Cytochromes are involved in the detoxification of xenobiotic compounds but this action can result in toxic by-products. The P-450 catalytic reaction cycle, including CYP2B1’s (Imaoka et al. 2004), generates radical oxygen species (ROS) (Heinemeyer, Nigan, and Hildebrandt 1980) and is probably a major source of intracellular ROS in the liver (Bondy and Naderi 1994; Puntarulo and Cederbaum 1998). Enhanced exposure to ROS during prolonged induction of cytochromes could result in oxidant damage to the cells. In addition, P-450 activity is known to produce other potentially damaging products during xenobiotic metabolism. Troglitazone may be metabolized by the induced cytochrome 3A1 to yield benzoquinone and reactive intermediates in a manner similar to quinone production by cytochrome 3A4 in human liver microsomes (He et al. 2001). Formation of a cytotoxic epoxide metabolite from troglitazone has been reported in studies with human HepG2 cells (Yamamoto et al. 2002). Our data support this possibility with the induction of NQO at increased concentrations of troglitazone, a reducer of quinones that is induced by these substrates (Figure 2A). Some cytochromes, like cytochrome 1A1, in addition to producing ROS, have also been shown to convert procarcinogenic metabolites into carcinogenic metabolites (Barouki and Morel 2001), so a CYP induction could also result in troglitazone’s toxicity by the production of deleterious metabolites through bioactivation by CYP3A1.
A second form of toxicity event potentially present with troglitazone is immunotoxicity. A delay between adverse events and starting a drug that results in idiosyncratic toxicity strongly suggests idiosyncratic toxicity is immune mediated and this observation has led to the danger hypothesis (Park, Primohamed, and Kitteringham 1998). In the danger hypothesis helper T cells are activated by endogenous signals from cells that are stressed or undergoing unprogrammed cell death (Seguin and Uetrecht 2003). Radical oxygen species (ROS) produced by metabolism of troglitazone could damage cells, resulting in the release of danger signals and the induction of a damaging immune response. That stress could result in the release of these danger signals is apparent in the induction of heme oxygenase and NQO by troglitazone in the C9 cell line (Figure 3). Heme oxygenase (HMOX-1) is induced by troglitazone in both cell types and its overexpression could also contribute to cell damage and the release of danger signals (Figure 2B and 3). HMOX-1 is classic indicator of oxidative stress whose moderate induction results in the production of two potent antioxidants biliverdin and bilirubin. HMOX-1 cleaves heme to form biliverdin iron and carbon monoxide. Overproduction of biliverdin could result in large amounts of free iron being released that could then be involved in deleterious reactions that compete with iron reutilization and protective sequestration pathways (Balla et al. 2003; Ryter and Tyrrell 2000).
Another gene expression response that implicates both chemotoxic and immunotoxic pathways in idiosyncratic toxicity is the induction of cyclooxygenase-2 (COX-2). COX-2, a gene involved in the inflammatory process, is also induced by troglitazone in the C9 cells (Figure 3C). The induction of COX-2 results in increased eicosanoid production, which increases the sensitivity of hepatocytes to allyl alcohol toxicity (Maddox et al. 2004). Troglitazone may have a similar effect. Induction of COX-2 may contribute to an inflammatory state that increases the body’s sensitivity to potential toxic compounds (Ganey et al. 2004). Thus, a person taking troglitazone may have a lower threshold for certain toxic compounds and exposure to these compounds could also yield an idiosyncratic toxic event linked to troglitazone. Additionally, COX-2 action could participate in the response described by the danger hypothesis, eliciting a stronger deleterious immune response because its action results in the release of danger signals.
Two other genes that were monitored were GADD45 and GADD153. GADD45 is a cell cycle regulator, which encodes a protein that is induced by genotoxic as well as various other stressors (Yin et al. 2004). It is shown to induce apoptosis in vitro (Takekawa and Saito 1998), and studies in the cancer cell line MCF-7 strongly suggest troglitazone initiates apoptosis in these cells through the induction of GADD45 (Yin et al. 2004). GADD153 also is induced by stressors and modulates apoptosis through members of the BCL2 family (Corazzari et al. 2003). GADD45 and GADD153 were induced exclusively by troglitazone in the primary hepatocytes as well as the C9 cells (Figures 1C and 3C). The other two PPARγ agonists did not induce the GADD genes but this is not completely unexpected as previous studies have shown GADD45 to be induced by troglitazone through a PPARγ independent mechanism (Yin et al. 2004). Additionally, it has been shown that ROS and quinones can induce GADD45 and GADD153 in culture, both of which are most likely generated in the presence of troglitazone (Stokes et al. 2002). Induction of the GADD genes may be indicative of direct cell damage produced by troglitazone and/or its metabolites along with predisposing cells to apoptosis. Cell damage leading to necrosis or apoptosis in the presence of ROS both could be contributing danger signals that could result in an immune-based idiosyncratic toxicity. Three genes related to cell growth p53, p21, and cyclin D1 were also monitored by this assay (Figure 3). Both p53 and p21 were induced by troglitazone and cyclin D1, a cell cycle regulatory protein, which acts as a growth factor sensor to integrate extracellular signals with the cell cycle, displayed lowered expression levels in the majority of experiments carried out with the described conditions. In the experiment shown here, cyclin D1 was only slightly down-regulated but this may be due to minor variation of growth phase of the cells from experiment to experiment and the impact of inductions of p53 and p21, which are inhibitory to cyclin D1, were not yet reflected in its expression level (Bae, Rhee, and Song 2003). p53 and p21 were shown to be induced in every experiment carried out. Troglitazones impact on the expression of these genes could result in G1 arrest and apoptosis (Bae, Rhee, and Song 2003). This activity could also result in the release of signals that could generate a deleterious immune response.
Based on the diversity of gene expression response observed, the idiosyncratic toxicity of troglitazone is most likely a genotype and or environment-specific multiple determinant process, with immune-related responses potentially acting as a major contributor to inducing adverse events. In this study an in vitro screen of the glitazones predicted the probability of troglitazone toxicity in a more robust manner than a 10-day in vivo study. This is not to say that in vivo studies can not predict toxicity. Even with troglitazone there have been some in vivo studies that upon closer scrutiny suggested troglitazone might be toxic, including a chronic monkey study and a dog metabolism study (SBA 1997). However, it would be useful to have a screening method that is more sensitive, cheaper, and more amendable to the early screening process than in vivo models. Early screening of compounds in a genetically homogenous population in vivo may mask the impact a compound may have in the human population that has a highly diverse genetic makeup. In the case of troglitazone in vitro monitoring revealed its toxicological potential. The in vitro system with its apparent increased sensitivity as well as higher throughput adaptability make it a strong candidate as an early toxicological screening platform, especially when combined with gene expression analysis.