Gene Expression Profiles in Rat Liver Slices Exposed to Hepatocarcinogenic Enzyme Inducers,Peroxisome Proliferators,and 17 α -Ethinylestradiol

Abstract

Transcription profiling is used as an in vivo method for predicting the mode-of-action class of nongenotoxic carcinogens. To set up a reliable in vitro short-term test system DNA microarray technology was combined with rat liver slices. Seven compounds known to act as tumor promoters were selected, which included the enzyme inducers phenobarbital, α-hexachlorocyclohexane, and cyproterone acetate; the peroxisome proliferators WY-14,643, dehydroepiandrosterone, and ciprofibrate; and the hormone 17α-ethinylestradiol. Rat liver slices were exposed to various concentrations of the compounds for 24 h. Toxicology-focused TOXaminer^™ DNA microarrays containing approximately 1500 genes were used for generating gene expression profiles for each of the test compound. Hierarchical cluster analysis revealed that (i) gene expression profiles generated in rat liver slices in vitro were specific allowing classification of compounds with similar mode of action and (ii) expression profiles of rat liver slices exposed in vitro correlate with those induced after in vivo treatment (reported previously). Enzyme inducers and peroxisome proliferators formed two separate clusters, confirming that they act through different mechanisms. Expression profiles of the hormone 17α-ethinylestradiol were not similar to any of the other compounds. In conclusion, gene expression profiles induced by compounds that act via similar mechanisms showed common effects on transcription upon treatment in vivo and in rat liver slices in vitro.

Gene expression profiling allows to simultaneously analyze the expression levels of thousands of different genes (Waring et al. 2001a). Its application in toxicology, referred to as toxicogenomics, is based on the assumption that most, if not all, toxic compounds alter gene expression to some extent and, therefore, the resulting expression pattern may be characteristic for specific classes of toxic compounds (Hong et al. 2003; Nuwaysir et al. 1999; Storck et al. 2002; Thomas et al. 2001). This approach not only allows us to investigate mechanistic aspects of toxicologically relevant compounds but also can be used in predictive toxicology (Hong et al. 2003; Nuwaysir et al. 1999; Storck et al. 2002; Thomas et al. 2001). Hereby the analysis of a large number of compounds with known mechanisms of action generates data banks that showed to be able to predict the toxic potential of newly developed drugs (Scheel, von Brevern, and Storck 2003). For that purpose we have selected three classes of compounds with known mode of action, i.e., enzyme inducers, peroxisome proliferators, and one estrogenic hormone.

Many xenobiotics and hormones are primarily metabolized in the liver, thus alterations in hepatic gene expression in response to treatment provides insights into the effects that may occur in other organs (Gerhold et al. 2003). Most toxicogenomic analyses have been conducted in the liver after in vivo drug treatment. However, in vitro test systems are highly desirable, offering the advantage of higher throughput, fewer test compounds and less animal experimentation. Cell lines such as the human hepatoma HepG2 cells have been used for toxicogenomic studies (Burczynski et al. 2000; Frueh et al. 2001; Harries et al. 2001; Hong et al. 2003; van Delft et al. 2004), but because of their metabolic deficiencies, the utility is limited. Cultured primary hepatocytes have been intensively investigated by gene expression analysis (Baker et al. 2001; Boess et al. 2003) and applied in predictive toxicogenomics (Jessen et al. 2003; Waring et al. 2001a). Recently gene expression of precision-cut rat liver slices have been explored as an in vitro system, with increasing application in toxicological studies (Cross and Bayliss 2000). For all these reasons, we used as in vitro test system precision-cut rat liver slices, which maintain the functional acinar architecture and contain the variety of different cell types of the liver (Lupp, Danz, and Muller 2001; Muller et al. 2000). Rat liver slices have been used extensively to evaluate liver function, toxicity, metabolism (reviewed in Lerche-Langrand and Toutain 2000), and induction of cytochrome P450 (Glockner et al. 1999; 2003; Muller et al. 1998).

Applying toxicology-focused TOXaminer^™ DNA microarrays, containing approximately 1500 genes (Scheel, von Brevern, and Storck 2003) on rat liver slices, mRNA expression profiles of several tumor promoters known to act by different mechanisms, were generated. These included the enzyme inducers phenobarbital (PB; a sedative and anticonvulsive drug) and α-hexachlorocyclohexane (α-HCH; an isomer of the insecticide Lindane), the anti-androgenic drug cyproterone acetate (CPA; which does not promote liver tumors but induces liver hyperplasia), and the peroxisome proliferators (PPs) WY-14,643 (WY; a hypolipidemic compound), dehydroepiandrosterone (DHEA; an anti-aging drug), and ciprofibrate (CF; a lipid-lowering agent). As a hormone, 17α-ethinyl estradiol (EE), a contraceptive drug, was tested.

The aim of this study was (i) to determine changes in gene expression in precision-cut rat liver slices after exposure to the test compounds at various concentrations; (ii) to examine whether the gene expression profiles generated in vitro permit a classification of compounds according to their mechanism of action; (iii) to compare expression profiles in vitro with those observed after in vivo treatment by the same compounds, which have been reported recently (Scheel, von Brevern, and Storck 2003).

MATERIALS AND METHODS

Chemicals and Reagents

Unless stated otherwise, all chemicals were obtained from Sigma Aldrich (Seelze). DHEA was a gift of D. Mayer (Heidelberg). All chemicals used were of the purest grade available.

Preparation, Culture, and Treatment of Rat Liver Slices

Precision-cut liver slices were prepared from male Han:Wistar rats (10 weeks of age) using a Krumdieck slicer, filled with Krebs-Henseleit HEPES buffer, pH 7.4 (slice thickness 0.25 mm) as described by Muller et al. (1998). For the treatment groups, eight liver slices of four rats were immediately incubated in 50-ml Erlenmeyer flasks containing 10 ml William’s medium E supplemented with insulin (1 μM), gentamicin (50 mg/L) and tylosin (100 mg/L), gassed with carbogen (95%O₂/5% CO₂) and shaken in a water bath at 37°C. After preincubation for 2 h the medium was changed and the test compounds PB, α-HCH, CPA, EE, DHEA, and WY were added at different concentrations (0.1, 1, 10, and 100μM), CF was tested at concentrations of 0.02, 0.2, and 2 mM. After treatment for 24 h the rat liver slices were frozen in liquid nitrogen and stored at −80°C. For comparison of the gene expression in non-cultured and for 24 h cultured untreated liver slices, these were frozen directly after isolation.

RNA Isolation

Liver slices were homogenized in TRIzol (Life Technologies, Gaithersburg MD) using a polytron (Art Miccra D-8; Labortechnik, Mülheim, Germany) and total RNA was isolated from each sample according to the manufacturer’s instructions. Total RNA was precipitated with isopropanol and washed with 75% ethanol. Afterwards the RNA was quantified spectrophotometrically and total RNA was routinely quality-controlled on agarose gels. Aliquoted samples were stored at −80°C.

Synthesis of cDNA

Five micrograms total RNA were used for cDNA synthesis. Reactions were carried out with the Superscript Choice System (Gibco-BRL Life Technologies, Karlsruhe, Germany), using an oligo(dT)18VN-primer as a reagent during the first-strand synthesis at 42°C for 2 h. Reaction was stopped (0.1 M EDTA; 1 mg/ml glycogen) and the unincorporated nucleotides were removed with Bio-Spin 6 Chromatography Columns (BioRad, München, Germany). The purified samples were ethanol-precipitated in the presence of 3 M sodium acetate and glycogen for 2 h or overnight. After centrifugation (12,000 × g at 4°C for 30 min), pellets were washed twice with 80 % ethanol, dried, and resuspended in 20 μl DEPC-treated water.

Labeling

Nine microliters of the first-strand synthesis were used for random priming. Enzyme reaction was done according to the manufacturer’s instructions (Roche Diagnostics, Mannheim, Germany) and cDNA was labeled with 50 μCi [α ³³P]dCTP (Amersham, Freiburg, Germany). Subsequently, reaction was stopped with 30 μl 50 mM EDTA and unincorporated nucleotides were removed with Bio-Spin 6 Chromatography Columns. The amount of radioactivity before and after purification was determined with a scintillation counter (Packard) to calculate the specific activity of the cDNAs (cpm/μg) and the yield of cDNA synthesis (%).

Description, Hybridization, and Washing of Arrays

Based on a collaboration in the Bio-Regio-Project 0311942, the TOXaminer^™ DNA microarrays were provided by Axaron Bioscience AG (Heidelberg, Germany) (Scheel, von Brevern, and Storck 2003). The membrane-based arrays were generated carrying approximately 1500 cDNAs, including known toxicological relevant genes and novel potentially relevant marker genes, identified by open systems, such as MPSS (massively parallel signature sequencing) (Brenner et al. 2000) and RMDD (restriction-mediated differential display) (Storck et al. 2002).

The arrays were prehybridized at 68°C for 1 h in 10 ml Church buffer (7% sodium dodecyl sulfate [SDS]; 0.5 M sodium phosphate pH 7.2 and 1 mM EDTA). After changing the Church buffer, hybridization was performed at 68°C for 3 days with 30 × 10⁶ cpm of probe. The filters were washed twice for 30 min with 2× SSC/0.1% SDS and once with 0.2×SSC/0.1% SDS at 68°C. Filters were sealed in plastic foil and exposed to phosphorus screens (Fuji, Düsseldorf, Germany) for 16 h. For each filter the hybridization signals were scanned using the Phosphoimager (Fujifilm FLA-3000) at a dissolution of 100 μ and 16 bit.

Stripping the filters with 0.1% SSC/0.1% SDS, they were reused 6 to 7 times in further hybridization experiments.

Data Analysis and Clustering Algorithm

One hybridization round included two filters that were chosen randomly from a filter batch number. Spot intensities on the arrays were analyzed using ArrayVision software (Imaging Research). Normalization was performed by comparing the density of each spot with the average intensity of all genes on that array. To ensure reproducibility, we repeated the whole experiment, starting from cDNA synthesis. For analysis we used data obtained from at least two experiments. Statistical analyses for gene changes were performed separately for each experiment, individual compound, and treatment dose. Gene expression data from two arrays, each with two spots of each gene, were calculated for each of both experiments as the average change versus gene expression of the vehicle-treated control group. Differently regulated genes in both experiments were identified using a modified t-test (Baldi and Long 2001). A p-value ≤.05 was accepted as significant. The significant changes were further filtered using a cutoff value of 2.0-fold over the control groups. Average fold-changes of both experiments were calculated only for those genes with a fold-change of 2.0 or more in both experiments.

Cluster analysis was performed using the software Gene-Spring (Silicon Genetics). The clustering is based on genes significantly regulated in either one of the treatments. The transcription profiles of these genes were classified according to similarities into hierarchical clusters using the Pearson correlation.

RESULTS

Gene Expression after Cultivation of Rat Liver Slices for 24 Hours

Changes induced by the cultivation were determined by gene expression analyses of freshly cut rat liver slices in comparison with slices that had been incubated for 24 h. Different expression of numerous genes was found. In phase I enzymes, especially genes of the CYP2 family, such as CYP 2E1, 2C11 and 2C23 were down-regulated (Table 1). Phase II biotransformation enzymes showed variable expression patterns. Glutathione S- transferases (GSTs) were regulated differentially: Transcripts of GST Yb1 (GST M1), GST Yb4 (GST M4), and GST P were up-regulated, whereas GST Ya (GST A) transcripts remained unchanged. GSH peroxidase, involved in the cellular redox cycling of GSH, was down-regulated.

Expression of sulfotransferases (STs) was changed differentially: A decrease in expression of arylsulfotransferase (ST1A1) was observed, whereas ST 1B1 was up-regulated. Glucuronosyltransferases (UDP-GT) did not show changes in expression levels. Thus the main changes after 24 h of cultivation of rat liver slices was repression of the cytochrome P450 mRNAs and differential modulation of phase II enzymes. Genes involved in liver cell proliferation and differentiation also varied in their expression after 24 h of cultivation: insulin-like growth factor (IGF)-II (unchanged), IGF-binding protein-1 (up-regulated), or IGF I (down-regulated). The G0/G1 switch gene (Gos-2), involved in cell cycle regulatory pathways, and the transformation-transactivation domain-associated protein (TRRAP) were found to be induced. Genes involved in protein synthesis, such as elongation factor-1 and ribosomal proteins showed elevated expression.

Gene Expression after Treatment of Rat Liver Slices with Different Test Compounds

Transcription profiles were generated in rat liver slices which had been treated with known tumor promoters. Differentially expressed genes after treatment of rat liver slices by two enzyme inducers (PB and HCH), two peroxisome proliferators (WY and DHEA), and the hormone (EE) were grouped according to their biological functions (Table 2).

The enzyme inducers and tumor promoters PB and HCH, as well as the peroxisome proliferators WY and DHEA are known to increase the metabolic activity in the rat liver. Accordingly, induction of the phase I enzymes (cytochrome P450 subfamily members) and phase II enzymes (GSTs and UDP-GTs) was observed in treated rat liver slices. In addition, peroxisome proliferators known to affect lipid metabolism, changed expression of genes involved in peroxisomal ß-oxidation (peroxisomal 3-ketoacyl-CoA thiolase) and lipid metabolism (acetoacetyl-CoA-acetyltransferase).

Expression of genes involved in signal transduction and cell proliferation was differentially regulated: Serine/threonine protein kinase (PCTAIRE-2) was up-regulated by WY, whereas HCH caused down-regulation. In addition, the expressions of both extracellular signal-regulated kinases 2 and 3 (ERK2 and ERK3) was changed; ERK2 was repressed and ERK3 was increased by treatment with HCH, whereas DHEA down-regulated ERK3. A strong reduction of the GTP-binding protein RAB and DNA topoisomerase IIB was induced by DHEA, whereas BAD (Bcl-2–associated death promoter) which was down-regulated by PB was up-regulated by DHEA.

Genes involved in proteolytic processes showed variable expression. For example, the ubiquitin-like carrier protein (SUMO-1) was up-regulated by CPA and CF (data not shown), as was the spermatogenic cell/sperm-associated Tat-binding protein homolog (Sata), the expression of which was enhanced by HCH treatment. In addition, HCH also induced the expression of polyubiquitin and ubiquitin whereas the expression of cathepsin was reduced. DHEA uniformly reduced the expression of proteolytic enzymes.

Among the transport and membrane proteins, a strong reduction of megalin (low-density lipoprotein receptor-related protein 2 precursor) was observed after treatment with HCH or DHEA, whereas treatment with EE caused an induction—additionally to that of the LDL receptor. Furthermore, one of the transporter systems for endogenous compounds and xenobiotics, the P-glycoprotein, a member of the ATP binding cassette transporter family, was reduced after EE exposure of rat liver slices. PB reduced the expression of clathrin (clathrin coated vesicles mediate intracellular membrane traffic) in a dose dependent manner.

In the class of stress- and cell damage–related genes, HCH (10 μM) caused a negative regulation of carbonic anhydrase III mRNA expression. The antioxidant enzymes, manganese-containing superoxide dismutase (SOD-2) and SOD-1 [Cu-Zn], were down-regulated by DHEA. Expression of other typical oxidative stress response genes such as metallothionein-I and the heat shock proteins HSP60 and HSP90 showed little change.

Genes involved in steroid metabolism were mainly affected by the hormones EE and DHEA. Tumor suppressors like RB1 and p53 were strongly down-regulated by HCH and up-regulated by EE. Furthermore, expression of RB1 was reduced through PB, whereas DHEA as well as CF enhanced the expression, the latter also increased the transcription level of the RB like protein 2 (data not shown). Induction of cytochrome c oxidase subunits I, II, and III by EE, DHEA, and CPA as well as of the mitochondrial genome (MTCO1) by EE and DHEA revealed increased respiratory activity in cells. In contrast, WY reduced the expression of MTCO1.

Cluster Analyses

Using the software GeneSpring (Silicon Genetics), the transcription profiles of genes that had been changed in the expression analyses were classified according to similarities into hierarchical clusters using the Pearson correlation. The three enzyme inducers PB, HCH, and to a lesser degree CPA, also acting as a hormone, form one cluster; the three peroxisome proliferators WY, CF, and DHEA form another (Figure 1). The pattern of the hormone EE is dissimilar to all other of the above mentioned classes of compounds.

Figure 2 illustrates the comparison between the cluster analyses of the liver slices treated in vitro, and the cluster analyses of the liver of rats treated in vivo (Scheel, von Brevern, and Storck 2003). As apparent, for some drugs, there is concordence between the transcription profiles in vitro (liver slices) and those obtained in vivo. With the exception of CPA in vitro, the transcription profiles of HCH and PB, both in vivo, and in vitro, clustered together. Similar coherent transcriptional changes were apparent for the peroxisome proliferators, whereby DHEA, WY, and CF formed a second cluster. HCH and DHEA in vivo and in vitro are located both in different subbranches within their main cluster. EE and CPA in vitro were dissimilar from all other treatments. Thus, the most of the substances we have tested can be classified in groups according to their different mechanism of action.

DISCUSSION

Cultivation of Rat Liver Slices

Conditions that allow the maintenance of differentiated hepatocytes should provide for an optimal in vitro test system. Cultivation of primary cells or liver slices is known to lead to cellular dedifferentiation, with partial loss of specialized cell functions. As revealed by our gene expression analyses of rat liver slices cultured for 24 h, numerous genes were found to be altered due to dedifferentiation and other adaptation processes. Within many functional gene classes, trend changes among the expression patterns were observed. As already reported in cultured hepatocytes (Price et al. 1997) phase I enzymes, notably members of the CYP family, were down-regulated during the course of cultivation. In our study especially genes of the CYP2 family were down-regulated, whereas members of the CYP3 and CYP4 families maintained their expression, confirming other studies on CYP 3A1 (Glockner et al. 2003). Furthermore, transcriptional changes of the phase II biotransformation enzymes were observed. In agreement with findings in primary hepatocytes (Baker et al. 2001), expression of arylsulfotransferase (ST 1A1) and thyroid hormone sulfotransferase (ST 1B1) were decreased and up-regulated, respectively. In contrast to primary hepatocytes, in slices UDP-glucuronosyltransferases (UGTs) did not show any alterations of transcription levels, possible due to the absence of dexamethasone in the medium (Jemnitz et al. 2000). GSTs catalyzing the initial step of GSH conjugation were up-regulated. Unlike primary hepatocytes (Baker et al. 2001) showing no changed expression, GST Yb1 (GSTM1) and GST Yb4 (GSTM4) were up-regulated in our study. GSH plays an important role in cell survival and defense mechanisms by detoxifying electrophiles and scavenging free radicals. An elevated GSH synthetase mRNA level was observed in primary hepatocytes (Baker et al. 2001) but not in slices. GSH peroxidase, an enzyme implicated in oxidative stress, was down-regulated along with other antioxidant enzymes such as SOD-2. In comparison to primary hepatocytes (Baker et al. 2001), GSSG reductase mRNA level was not increased. However, expression of other genes typically responding to oxidative stress, such as metallothionein-I and the heat shock proteins HSP60 and HSP90, were induced. Normally, parenchymal hepatocytes in adult liver are quiescent and replicate infrequently. Genes involved in hepatocellular proliferation were found to be modulated to a various degree. Changes in transcripts of IGF-binding protein-I or IGF I revealed a pattern indicative of nonproliferation. However, the G0/G1 switch gene (GOS2), thought to be involved in the cell cycle regulation (Cristillo et al. 1997), was up-regulated, whereas it was down-regulated in hepatocyte culture. Up-regulation of cell cycle inhibitory genes p53 and p21, as shown in hepatocytes, was not observed in rat liver slices. Induction of the ATM-related transformation-transactivation domain-associated protein (TRRAP) indicated transcriptional activation (Liu et al. 2003).

Treatment of Rat Liver Slices with Different Test Compounds

To evaluate rat liver slices as a suitable in vitro system in predicting the toxic potential of chemicals, gene expression profiles induced by seven known liver tumor promoters were analyzed to permit classification according to mechanism of action as previously shown for gene expression profiles generated in vivo (Scheel, von Brevern, and Storck 2003).

The gene expression profiles in rat liver slices after treatment with tumor-promoting agents affected a number of cellular pathways, including signal transduction, lipid metabolism, and biotransformation. The barbiturate PB is known to possess tumor-promoting effects on the rat liver (Hagiwara et al. 1999) and to enhance the expression of several phase I and II enzymes such as members of the CYP 2B and 3A families, GSTs, and UGTs. In response to PB, the nuclear constitutive androstane receptor (CAR) forms heterodimers with the retinoid X receptor (RXR), leading to transcriptional activation of CYP 2B and other genes. In accordance, expression of CYP 2B and CYP 3A was induced in our study whereas expression of CYP 4A1 was not repressed as reported for primary cultures of rat hepatocytes (de Longueville et al. 2003). Altered expression of several UGTs and GSTs induced by PB was also observed. An as yet undescribed finding is that PB repressed the Sentrin/Sumo-specific protease (SENP), a desumoylation enzyme involved in the release of proteins from nuclear domains (Gong et al. 2000).

Industrial HCH production results in a mixture of different isomers, mainly α- HCH. α-HCH is known to promote liver tumor development (Schroter et al. 1987) and to induce enzyme expression and cytochrome P450 activity (Barros, Simizu, and Jungueira 1991; Beurskens et al. 1991), the latter being confirmed by our study which yielded increased expression of CYP 3A1 and CYP 4A1 genes. Whereas in previous studies HCH was shown to change GSH-redox cycle enzymes (Garcia-Fernandez et al. 2002), the present investigation revealed elevated expression of several GSTs as well as of different UGTs. Furthermore, α-HCH is known to affect reproductive function, cause testicular toxicity (Pius et al. 1990) and inhibit steroidogenesis (Ronco et al. 2001; Walsh and Stocco 2000). We revealed a twofold induction of the spermatogenic cell/sperm-associated Tat-binding protein (TBP-1) homolog Sata, a member of the proteasome 26, known to be involved in the control of cell proliferation. (Nakamura et al. 1998). Overexpression of TBP-1 is shown to result in an increase of p53 protein levels and activity (Pollice et al. 2004). In this study an enhanced expression of p53 mRNA was observed with low HCH concentration, whereas high concentrations repressed the p53 expression. Induction of oxidative stress (Samanta, Sahoo, and Chainy 1999), accompanied by a reduced activity of antioxidant enzymes such as catalase and superoxide dismutase, in the testis of HCH-treated rats (Samanta and Chainy 2002) is involved in HCH-induced testicular toxicity. In accordance, we observed a negative regulation of carbonic anhydrase III (CA III) mRNA expression, which functions as an oxyradical scavenger and protects cells from oxidative damage (Raisanen et al. 1999). Furthermore, within the family of membrane proteins, the expression of low-density lipoprotein receptor (LDL receptor) was reduced in α- HCH–treated rat liver slices. This is in keeping with findings that female Wistar rats that had been given α-HCH develop a hyperlipemia, due to an increase in serum very-low-density lipoprotein (Grajewski and Oberdisse 1977). Regarding signal transduction, the expression of both of the mitogen-activated protein (MAP) kinase family members ERK2 and ERK3 was changed, ERK3 being increased and ERK2 repressed. Recent findings show high similarity in their kinase catalytic domain. Yet, ERK2 translocates into the nucleus following activation whereas ERK3 is constitutively localized to the nucleus (Robinson et al. 2002). α-HCH also strongly induced the expression of D123, most highly expressed in testis, which appears necessary for S-phase entry of the cell cycle.

DHEA occurs naturally as adrenal steroid in the human body, and circulates largely as the sulfate conjugate and serves as precursor for estrogens and androgens (Ripp et al. 2003). In rodents DHEA is a potent peroxisome proliferator, inducing peroxisomal and microsomal enzymes through activation of the nuclear receptor, PPARα(Peters et al. 1996; Prough et al. 1994). In contrast to other peroxisome proliferators like nafenopin, DHEA induces also genes unrelated to peroxisome proliferation (Gu et al. 2003; Singleton et al. 1999). Furthermore, DHEA exerted protective effects against chemically induced tumor formation and has lipid-lowering effects combined with down-regulated PPAR γ expression (Inano et al. 1995; Kajita et al. 2003; Nyce et al. 1984; Pashko et al. 1984; Schwartz et al. 1981). Several regulated genes primarily involved in peroxisome proliferation and previously shown to be affected by DHEA were shown to be changed in expression: members of xenobioticmetabolizing CYP 4A family (Webb et al. 1996) and of peroxisomal β-oxidation such as acyl-CoA oxidase and synthase, 3-ketoacyl-CoA thiolase, phospholipase C, and the bifunctional enzyme enoylCoA:hydratase-3-hydroxyacyl-CoA dehydrogenase (Osumi 1993; Rao et al. 1994; Yamada et al. 1991). In addition, genes not known to be peroxisome proliferator reponsive were identified. Genes with elevated expression are genes of the mitochondrial genome, anti-apoptotic Bcl-2– associated death promoter, and the tumor suppressor RB1. Genes with decreased expression include DNA topoisomerase (TOP) IIB, extracellular signal-related kinase (ERK3), Rab-related GTP-binding protein (involved in intracellular vesicle transport; Osanai et al. 2001), SH3P7r3 (an actin-binding protein implicated to cytoskeleton organization, Yamazaki et al. 2001), and proteolytic enzymes. Little is known about the effect of DHEA on extracellular signal-regulated kinases (ERK), but WY was shown to induce phosphorylation of ERK and activate the ERK1/2 signaling pathway independent on PPAR α activation (Banfi et al. 2003; Gardner et al. 2003). The strong decrease of TOP IIB caused by DHEA may be important for transcription because the TOP IIß–DNA covalent complex has been proposed to arrest transcription (Xiao et al. 2003). In our study, no down-regulation of p21 as previously reported (Depreter et al. 2002) was observed.

The rodent liver carcinogen and hypolipidemic drug WY activates transcription of the CYP 4A family through the PPARα. (Simpson 1997) and is known to increase lipoprotein uptake and fatty acid ß-oxidation (Staels et al. 1998). This is reflected by the induction of peroxisomal genes such as acyl-CoA oxidase and 3-ketoacyl-CoA thiolase, and mitochondrial genes such as carnitine palmitoyl transferase II and acyl-CoA dehydrogenase. Several studies describing WY-induced gene expression in mouse liver (Iida et al. 2003; Thomas et al. 2001; Yamazaki, Kuromitsu, and Tanaka 2002), rat liver (Crunkhorn et al. 2004; Hamadeh et al. 2002; Ellinger-Ziegelbauer et al. 2005), HepG2 cells (Barbier et al. 2003) and rat liver slices as well as in primary hepatocytes (Jessen et al. 2003) have been published. With the exception of acyl-CoA dehydrogenase, these studies support our findings.

EE is a strong tumor promoter and weak hepatocarcinogen in the rat. It was reported to enhance the transcription of nuclear and mitochondrial genes and respiratory chain activity in liver and cultured hepatocytes of female rats (Chen et al. 1996, 2003). These data are in accordance with increased mRNA levels of genes involved in the respiratory pathway such as cytochrome c oxidase subunits I, II, III and MTCO1 in male rats in our study.

Classification of Test Compounds

In order to determine if gene expression changes induced by tumor-promoting agents permit classification of the compounds according to their mechanism of action, hierarchical cluster analyses have been reported as a useful method in several studies (Bulera et al. 2001; Burczynski et al. 2000; de Longueville et al. 2003; Jessen et al. 2003; Waring et al. 2001a, 2001b).

In the present study hierarchical cluster analysis showed that in most cases, gene expression profiles were specific and compounds with similar mode of action clustered together (Figure 1). The enzyme inducers PB, α-HCH and CPA, the latter also acting as hormone, and the peroxisome proliferators WY, CF, and DHEA formed two separate clusters, which confirmed that enzyme inducers and peroxisome proliferators functionally regulate different genes and act through different mechanisms. Expression profiles of the hormone 17α-ethinylestradiol did not show similarity to any of the other compounds.

Additionally, hierarchical cluster analysis of gene expression profiles obtained in rat liver slices exposed in vitro (this study) and those induced after in vivo treatment (Scheel, von Brevern, and Storck 2003) of most of the investigated compounds revealed striking similarities of gene expression responses between agents belonging to the same mechanism of action class. This observation is in contrast to a study (Jessen et al. 2003) that reported that gene expression profiles generated in vitro in rat liver slices and primary hepatocytes as well as in vivo in rat liver clustered mainly according to the test system used. In support of our results, Boess et al. (2003) observed that gene expression in rat liver slices was more similar to that in intact liver than the gene expression from different culture systems of primary hepatocytes. Dissimilarities in gene expression observed with the hormones EE and CPA may depend on differences in age or hormonal situation known to influence gene expression (Pankiewicz et al. 2003).

Taken together, cluster-based analysis of our array data reinforces the notion that even with a limited set of compounds, transcription profiling of rat liver slices exposed in vitro is a useful tool for predictive toxicology screening.