Transcriptional Profiles in Liver from Mice Treated with Hepatotumorigenic and Nonhepatotumorigenic Triazole Conazole Fungicides: Propiconazole,Triadimefon,and Myclobutanil

Abstract

Conazoles are environmental and pharmaceutical fungicides. The present study relates the toxicological effects of conazoles to alterations of gene and pathway transcription and identifies potential modes of tumorigenic action. In a companion study employing conventional toxicological bioassays (Allen et al., 2006), male CD-1 mice were fed triadimefon, propiconazole, or myclobutanil in a continuous oral-dose regimen for 4, 30, or 90 days. These conazoles were found to induce hepatomegaly, to induce high levels of hepatic pentoxyresorufin-O-dealkylase activity, to increase hepatic cell proliferation, to decrease serum cholesterol, and to increase serum triglycerides. Differentially expressed genes and pathways were identified using Affymetrix GeneChips. Gene-pathway associations were obtained from the Kyoto Encyclopedia of Genes and Genomes, Biocarta, and MetaCore compendia. The pathway profiles of each conazole were different at each time point. In general, the number of altered metabolism, signaling, and growth pathways increased with time and dose and were greatest with propiconazole. All conazoles had effects on nuclear receptors as evidenced by increased expression and enzymatic activities of a series of related cytochrome P450s (CYP). A subset of altered genes and pathways distinguished the three conazoles from each other. Triadimefon and propiconazole both altered apoptosis, cell cycle, adherens junction, calcium signaling, and EGFR signaling pathways. Triadimefon produced greater changes in cholesterol biosynthesis and retinoic acid metabolism genes and in selected signaling pathways. Propiconazole had greater effects on genes responding to oxidative stress and on the IGF/P13K/AKt/PTEN/mTor and Wnt-β-catenin pathways. In conclusion, while triadimefon, propiconazole, and myclobutanil had similar effects in mouse liver on hepatomegaly, histology, CYP activities, cell proliferation, and serum cholesterol, genomic analyses revealed major differences in their gene expression profiles.

Introduction

Conazoles are antifungal agents most commonly used as pesticides in the protection of fruit, vegetable, and cereal crops (Zarn et al., 2003). In addition to their wide agricultural use, many conazoles are also used as pharmaceuticals in the treatment of local and systemic fungal infections in the human population (Georgopapadakou and Walsh, 1996). The mechanism of antifungal action of these agents is based in their inhibition of ergosterol biosynthesis by inhibiting lanosterol 14α-demethylase (CYP51) activity (Vanden Bossche et al., 1989; Ronis et al., 1994; Debeljak et al., 2003). This is accomplished through the triazole or imidazole moiety coordinating with the heme iron of CYP51 (Podust et al., 2001).

Conazole fungicides demonstrate a wide range of toxicological properties in mammalian systems (Zarn et al., 2003). They can induce both hepatomegaly and cytochrome P450 (CYP) isozymes (Sun et al., 2005; Juberg et al., 2006; Sun et al., 2006; Tully et al., 2006). Conazoles can act as both inducers and inhibitors of hepatic CYP activities depending on the specific conazole. For example, propiconazole and fluconazole induce CYP2B and CYP3A activities in mice and rats (Ronis et al., 1994; Sun et al., 2005, 2006), while tioconazole can inhibit CYP activites in the CYP1A, 2A, 2C, 2D and 3A families (Zhang et al., 2002). Conazoles can induce hepatic cell proliferation and alter cholesterol levels (Strandberg et al., 1987; Juberg et al., 2006; Peffer et al., 2006). In rodents, chronic exposure to several conazoles can also induce hepatotoxicity, neurotoxicity, and tumorigenesis (Hurley, 1998; Moser et al., 2001; Reeves et al., 2003; Biagini et al., 2006).

Myclobutanil (Systhane or Eagle), propiconazole (Banner), and triadimefon (Bayleton) are triazole-based conazoles. In 2-year feeding studies in mice and rats, high-dose (1800 ppm) triadimefon was associated with an increased incidence of hepatocellular adenomas in male and female NMRI and CF₁/W74 mice, and a slightly higher incidence of thyroid follicular cell adenomas in male Wistar/Han rats compared to control (INCHEM, 1981; EPA, 1996). In 2-year feeding studies in mice and rats with propiconazole, significant increases were noted in the incidence of benign and malignant liver tumors in male CD-1 mice at the highest feeding level (2500 ppm); however, there were no tumors in the rat at any feed level (INCHEM, 1987). Myclobutanil did not show any treatment-related tumorigenic effects in 2-year studies in mice or rats at 2000 ppm (INCHEM, 1992).

The modes of hepatotumorigenic action of triadimefon and propiconazole have not been characterized. Genotoxicity studies with several of these tumorigenic conazoles have provided negative results (INCHEM, 1981, 1987, 1992; EPA, 1996). Based on the lack of genotoxicity, and increases in liver hypertrophy, Cyp2b20(10) induction, and cell proliferation, it has been proposed that the tumorigenic mode of action of fenbuconazole in mice is similar to that of phenobarbital (Juberg et al., 2006).

An accompanying paper (Allen et al., 2006) describes in-life studies in mice with triadimefon, propiconazole, and myclobutanil in which traditional toxicity endpoints were evaluated. The present analyses of conazole-altered gene expression represent an extension of these studies aimed at the identification of potential modes(s) of tumorigenic action.

Toxicogenomic analyses using DNA microarrays provide a powerful tool to examine transcriptional events and relate them to biochemical and toxicological alterations (Hayes and Bradfield, 2005; Lettieri, 2006). Gene expression profiles induced by pharmaceuticals and chemicals in the livers of rodents have been used to identify potential mechanisms of toxicity (Waring et al., 2001; Meneses-Lorente et al., 2003; Huang et al., 2004; McMillian et al., 2004). Gene expression profiles in mice exposed to chemical agents have also been associated with modes of toxic action (Iida et al., 2003, 2005; Malarkey et al., 2005; Okawa et al., 2006). In order to obtain information on the mode of tumorigenic action of these conazoles, we have characterized hepatic gene expression profiles in mice following acute and subchronic exposures to tumorigenic conazoles, propiconazole, and triadimefon, and the nontumorigenic conazole, myclobutanil. The design of the present study mimics the experimental conditions used in the chronic bioassays of these conazoles. The present study utilized transcriptional profiling as a means to differentiate between 3 conazole fungicides and to identify potential molecular pathways that may be involved in the induction of mouse liver tumors. These differences are discussed in relationship to their hepatotumorigenic activities.

Materials and Methods

Experimental Design

The complete details and results of the in-life experiments are described in Allen et al. (2006) and this study provided the liver tissues used in the current study. Briefly in the Allen et al. (2006) study, male CD-1 mice received triadimefon (100, 500, 1800 ppm), propiconazole (100, 500, 2500 ppm), myclobutanil (100, 500, 2000 ppm) or vehicle-treated feed for 4, 30, and 90 days. Feed exposure levels were determined from previously conducted long-term chronic bioassays in the mouse at the maximum tolerated dose (INCHEM, 1981, 1987, 1992). For triadimefon the high-dose level of 1800 ppm gave hepatic tumors in mice (INCHEM, 1981). For propiconazole the high-dose level of 2500 ppm gave hepatic tumors in mice (INCHEM, 1987). For myclobutanil the high-dose level 2000 ppm did not give tumors in mice (INCHEM, 1992). The lower doses: mid-dose (500 ppm) and low-dose (100 ppm) were nontumorigenic for all 3 agents. At each time point, mice were euthanized by CO₂ asphyxiation and then necropsied. Livers were removed, frozen in liquid N₂ within 2 minutes of death, and stored at −80°C until further processing for RNA isolation.

RNA Preparation

Total RNA was isolated from mouse livers (n = 3) for 4 days, mid- and high-dose, for 30 days, low-, mid-, and high-dose, and for 90 days, mid- and high-dose using TRI reagent (Molecular Research Center, Inc., Cincinnati, OH) and following the manufacturer’s directions. Samples were not pooled. Briefly the liver tissues were ground to a fine powder in liquid N₂, and added to 5 ml of TRI reagent. One ml of BCP (1-bromo-3-chloropropane) was added and tubes were vortexed and allowed to sit at room temperature for 10 minutes, and then centrifuged at 10,000 × g for 15 minutes at 4°C. The upper aqueous phase containing the RNA was removed to a clean tube and 2.5 ml of ice-cold isopropanol was added. The samples were left at −70°C overnight followed by a centrifugation at 10,000 × g for 15 minutes at 4°C to precipitate the RNA. The RNA pellet, visible after centrifugation, was washed with 70% ethanol and air dried. The RNA was resuspended in RNAse free water and its RNA purity and concentration was measured on an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA).

Microarray Analysis

Sample preparation, processing and hybridization to the Affymetrix Mouse 430 2.0 array was performed at Expression Analysis, Durham, NC as described in the GeneChip Expression Analysis Manual (Affymetrix; Santa Clara, CA). In the study, 1 chip was used per animal and liver sample. Information on the Mouse 430 2.0 array, which analyzes approximately 34,000 mouse genes is available on the internet <http://www.affymetrix.com/support/technical/datasheets/mogarrays-datasheet.pdf>.

Data Analysis

Gene expression measures for each gene in each sample were obtained using Robust Multichip Average analysis (Irizarry et al., 2003). The subsequent gene lists and associated expression values were uploaded into GeneSpring (Version 7.2) and log2 transformed. For each sample at each time point, unchanging genes whose normalized value (ratio to median) fell between 0.9 and 1.2 were removed from further analysis. Of the remaining genes, 1-way ANOVA was used to identify differentially expressed genes. The variance for each gene was estimated using the Cross-Gene Error Model (Rocke-Lorenzo; GeneSpring version 7.2; Redwood City, CA). A False Discovery Rate (FDR) of 0.05 was applied to control for family-wise errors. These lists of differentially expressed genes constituted the basis for all subsequent analyses. The numbers of differentially expressed genes differed by exposure scenario (Table 1).

When appropriate, fold change was calculated as the ratio of the mean of 3 experimental expression measures to the mean of 3 control expression measures. At each time point, principal component analysis (Cluster 3.0, University of Tokyo, Human Genome Center, June, 2002) was conducted on all the genes that were differentially expressed at that time point. Pathway assignments of genes were made by annotations from the Kyoto Encyclopedia of Genes and Genomes (KEGG), the Bio Resource for Array Genes (Bioinformatics Group at Arizona Cancer Center) and MetaCore from GeneGo Inc. Overrepresentation of a pathway was determined using the hypergeometric function in Microsoft Excel. Pathways that contained 4 or more genes and that were over-represented based on a hypergeometric test with p-value < 0.05 were considered altered.

Results

Differentially Expressed Genes

The numbers of differentially expressed genes (DEG) in each exposure group tested after a 1-way ANOVA analysis are presented in Table 1. At 4 and 30 days of treatment, propiconazole induced the largest number of DEG at the high dose of the 3 conazoles. However, at 90 days of treatment triadimefon produced the most DEG with propiconazole producing slightly less. At 30 days of treatment the number of DEG monotonically increased with dose for myclobutanil and propiconazole, with the DEG induced by triadimefon reaching a plateau at the mid-dose. Except for 4-day myclobutanil and 30-day triadimefon, high doses of conazoles induced the largest numbers of DEG (Table 1).

Principal Component Analyses

Principal component analysis (PCA) was applied to all of the DEG at each time point after high dose treatment. Replicate samples tightly clustered for control and myclobutanil at 4 days of treatment, for control and propiconazole at 30 days of treatment, and for control, myclobutanil, and propiconazole at 90 days of treatment. PCA differentiated conazole treatment from control after 4 days and each conazole and control from each other after 90 days (Figure 1). The PCA showed that as treatment time increased the conazole-treated samples were more differentiated from each other and from control. After 90 days of treatment, each treatment group was clearly distinct. The PCA analysis demonstrated that the treatments could be differentiated from each other.

Venn Analyses of DEG

Venn analyses of DEG were performed on the DEG at each time point to partition those DEG that were common to all conazoles, common to groups of 2 conazoles, and unique to 1 conazole (Figure 2). The numbers of DEG common to all conazoles increased with time (202 to 462), as did the numbers of DEG that were unique to triadimefon treatment from 96 to 997. The numbers of DEG common to triadimefon and propiconazole rose from 4 days of treatment to 30 days of treatment and then remained about the same at 90 days. The numbers of DEG unique to propiconazole presented a complex time course, increasing from 4 days to 30 days of treatment from 1097 to 1673 and then lowering to 454 at 90 days. These results suggested that at 4 days of treatment, propiconazole altered many DEG unique to that treatment, while triadimefon altered approx. one-tenth that number (96). At 90 days of treatment, the inverse was observed where triadimefon uniquely altered 997 DEG compared to propiconazole’s 454 DEG.

Analyses of DEG

An algorithm was applied to the 4- and 90-day DEG to identify potentially important genes altered by propiconazole or triadimefon that might be associated with potential mechanisms of carcinogenesis. The algorithm took advantage of 2 observations: (1) Propiconazole and triadimefon only induced tumors at the high dose and (2) myclobutanil did not. This algorithm consisted of 2 steps: (1) For triadimefon the expression data at the tumorigenic high-dose and nontumorigenic mid-dose were compared to identify those genes which were significantly expressed only at the high dose (i.e., those genes that were significant at both dose levels were disregarded). This was done for the 4- and 90-day treatments and (2) the expression levels for each gene in resultant lists generated in 1 (from triadimefon) were compared with the expression values of the same genes found with myclobutanil at the high-dose treatment. A triadimefon/myclobutanil fold ratio was calculated (e.g., expression fold (gene x) in triadimefon/expression fold gene-x in myclobutanil). Genes were retained if the fold ratio was less than 0.5 or greater than 2. This yielded genes only associated with both high-dose triadimefon treatment and with overexpression levels greater or underexpression levels less than those found with myclobutanil treatment. These 2 steps were repeated for propiconazole. This process created the following groups of DEG. Propiconazole: 4-day, 51 genes; 90-day, 67 genes: triadimefon: 4-day, 29 genes; 90-day, 95. These analyses revealed the following genes and their fold expression levels for triadimefon at 4 days; Akr1b7, 4.1-fold; Cyp26a1, 4.9-fold; Gstα 2, 3.3-fold; Hspa1b, 2.9-fold; and PPARα, 0.35-fold. Akr1b7 is an aldo-keto reductase protein that is responsible for detoxifying isocaproaldehyde generated by the conversion of cholesterol to pregnenolone and detoxifying 4-hydroxynonenal, a lipid peroxidation byproduct (Lefrancois-Martinez et al., 1999; Kotokorpi et al., 2004). Cyp26a1 is a retinoic acid hydroxylase which converts all trans retinoic acid to 4-hydroxy retinoic acid (Marikar et al., 1998; Wolf, 2001). Gstα 2, glutathione transferase α2, is an antioxidant gene and is transcriptionally activated by nuclear-factor-E2-related factor (Nrf-2) (Kang et al., 2005). Hspa1b is a member of the HSP 70 family of stress response proteins which is induced in response to heat and other stresses (Didelot et al., 2006). PPARα, peroxisome proliferator-activated receptor α is a transcription factor that controls proteins involved in lipid transport and metabolism (Kiec-Wilk et al., 2005). After 90 days of triadimefon treatment, the following genes were identified: Cyp26a1, 2.5-fold; Gadd45b, 5.7-fold; and Ptges, 2.4-fold. Gadd45b is a nuclear protein associated with cell cycle checkpoints, genome stability, and DNA repair (Gupta et al., 2006). Ptges, prostaglandin e synthetase, is at the terminal step in the biosynthesis of prostaglandin e2 (Degousee et al., 2006). For propiconazole at 4 days the following genes were identified; Akr1b7, 8-fold; Cyp4a10, 0.38-fold; Pparα, 0.18-fold. Cyp4A10 is a fatty acid hydroxylase that metabolizes arachidonic acid into hydroxyeicosatetraenoic acids (HETEs) and is a marker gene for PPARα-activation (Patsouris et al., 2006). For propiconazole at 90 days the following genes were identified: Gadd45b, 4.5-fold; Hspa1a, 3.1-fold; Hspa1b, 5.3-fold; Pparα, 0.22-fold; Ptges, 2.5-fold; Tgfbr2, 2.9-fold; and Tnfrsf19, 3-fold. Tnfrsf19 is a tumor necrosis factor receptor found in most mouse tissues (Hu et al., 1999). Tgfbr2 is transforming growth factor, beta receptor II. It is a required component of the transforming growth factor beta (TGF-β) receptor and one gene in the microRNA expression signature of human solid tumors (Volinia et al., 2006).

Gene Expression Dose Response

Within this data set there are many instances of genes with mRNA expression increasing with dose. One example of this is shown in Figure 3. A clear dose response in gene expression is evident with propiconazole at 30 days of treatment for selected genes (Figure 3). The greatest change of increased gene expression after conazole treatment was the xenobiotic metabolizing P450 genes coding for Cyp2b20, Cyp2c55, and Cyp2c65. Cyp2b20 is a phenobarbital-inducible mono-oxygenase; Cyp2c55 is a recently discovered mono-oxygenase that metabolizes arachidonic and linoleic acid (Wang et al., 2004); Cyp2c65 function has not yet been determined. Other xenobiotic metabolizing genes that had a related increased expression were Gstt3, Gstm3, andAkr1b7. Gst theta 3 and Gst mu3 are glutathione transferases and are Phase II conjugating enzymes.

Cholesterol Biosynthetic Genes

Table 2 summarizes the transcriptional changes of cholesterol biosynthesis genes at four days induced in the mouse liver in high dose animals compared to control animals. Tri-adimefon significantly up-regulated 8 of the 10 enzymes in this pathway, while propiconazole significantly up-regulated 3. Myclobutanil did not significantly up-regulate any of these genes. In accord with negative feedback for these transcriptional changes, only triadimefon significantly lowered blood serum cholesterol levels at 4 days (Allen et al., 2006). These lower serum cholesterol levels could potentially be sensed via 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase, a key regulator of the cholesterol biosynthetic pathway (Ness and Chambers, 2000). This could lead to allosteric alterations of enzymatic activity as well as transcriptional feedback. Increased transcription of cholesterol biosynthetic genes accompanied by blockage of lanosterol α-demethylase near the end of the cholesterol biosynthetic pathway could alter the distribution of molecular intermediates along this pathway. After 30 or 90 days of treatment at high dose there were no significant transcriptional alterations of cholesterol biosynthetic genes by any of the 3 conazoles. However, after 30 days all the chemicals at high doses reduced serum cholesterol, and after 90 days myclobutanil and propiconazole reduced serum cholesterol (Allen et al., 2006).

Pathway Analyses

The DEG found in Table 1 were mapped to known compendia of pathways (GeneGo, Biocarta, and KEGG) using the criteria that pathways must have 4 or more DEG and the DEG in that pathway are overrepresented based on a hypergeometric test with p < 0.05. The pathways were functionally categorized as metabolism, cell signaling, or cell growth pathways and their distribution over time by conazole is found in Figure 4. The greatest effect observed was in metabolism pathways for all exposure groups except 90-day propiconazole. Propiconazole altered the most pathways at all time points and both propiconazole and triadimefon had a much greater effect than myclobutanil at all time points. After 4 days propiconazole altered the greatest number of pathways (61), divided between metabolism, signaling, and growth categories. The 10 pathways altered by triadimefon were metabolic pathways, while myclobutanil had similar numbers of pathways (9), but with a different distribution (Table 3).

After 30 days, the number of altered pathways increased for each conazole, with propiconazole again altering the greatest number (136) compared to triadimefon (66) and myclobutanil (11). The distribution of categories of pathways at day 30 changed favoring increasing numbers of metabolic and signaling pathways for both propiconazole and triadimefon compared to day 4. At days 30 and 90 the numbers of altered pathways increased with myclobutanil to a total of 28, while the numbers of pathways decreased for both propiconazole (77) and triadimefon (59) compared to day 30. At 90 days of treatment there was a strong representation of both metabolic and signaling pathways for all 3 conazoles. The complete lists of conazole-altered pathways at high-dose treatment after 4, 30, and 90 days of exposure are found in Tables 3, 4, and 5, respectively.

Venn Analyses of Pathways

A Venn analysis was applied to the significantly altered pathways. At each time point propiconazole had the greatest number of altered pathways (Figure 5). As exposure time increased the percentage of pathways that propiconazole and triadimefon had in common also increased. There were few pathways that propiconazole and triadimefon individually had in common with myclobutanil at any time point. At 4 days of treatment, there were 2 pathways common to all 3 conazoles, 4 pathways common to triadimefon and propiconazole, and 4 and 51 pathways unique to triadimefon and propiconazole, respectively. After 90 days of treatment there were 20 altered pathways that were common to all conazole treatments, triadimefon and propiconazole had 24 altered pathways in common and there were 14 and 32 pathways unique to triadimefon and propiconazole, respectively.

Discussion

The goal of the present study was to differentiate 3 conazoles with transcriptional profiles, to characterize the toxicity pathways, and to identify potential modes of hepatotumorigenic action for triadimefon and propiconazole. In chronic studies in high dose in mice triadimefon induced mouse liver adenomas and was hepatotoxic; propiconazole induced hepatic adenomas and carcinomas and was hepatotoxic; and myclobutanil was devoid of tumorigenic properties, but was hepatotoxic (INCHEM, 1987, 1992; EPA, 1996). In a companion study under conditions that mimicked the chronic bioassay, these 3 conazoles were administered in the feed to male CD-1 mice for 4, 30, and 90 days. All 3 conazoles were found to induce hepatomegaly, induce high levels of pentoxyresorufin-O-dealkylase (PROD) activity, increase cell proliferation in the liver, decrease serum cholesterol levels, and at 30 days of treatment to increase triglyceride levels (Allen et al., 2006). In a recently published study using medium density arrays, hepatic gene expression profiles were obtained from male CD-1 mice after gavage treatment for 14 days with triadimefon, propiconazole, and myclobutanil. CYP genes, xenobiotic response genes, and transporter genes were altered by the 3 conazoles (Goetz et al., 2006). The present study examined gene expression response over extended time and dose ranges in the livers of male CD-1 mice fed these 3 conazoles using high-density microarrays.

Nuclear Receptors, CYPs, and Xenobiotic Metabolizing Enzymes

The observed genomic and cellular responses induced by the 3 conazoles could be partially attributed to the activation of a series of nuclear receptors. The overexpression of Cyp1a2 and the induction of hepatic MROD activities are related to activation of the aryl hydrocarbon receptor (Ahr). The overexpression of Cyp2b20 and the associated induction of PROD activities are related to the activation of constitutive androstane receptor (CAR) (Maglich et al., 2002). The over-expression of Cyp3a11 andCyp3a13 are related to pregnane X receptor (PXR) (Maglich et al., 2002), the altered expression of CYP4A is related to the persoxisome proliferator activated receptor (PPAR) (Jeffery et al., 2004), and the altered expression of CYP7A is related to the farnesoid X receptor (FXR) and liver X receptor (LXR) (Peet et al., 1998; Wang et al., 2006). While the transcriptional regulation of Cyp2c55 is unknown, other human CYP2C family members, CYP2C8, and CYP2C9 are regulated by CAR (Ferguson et al., 2002, 2005). CAR activation and subsequent Cyp2b10(20) over expression have been linked to hepatic hypertrophy using phenobarbital and 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene in CAR-deficient mice (Wei et al., 2000). The increased mouse liver weights and PROD activities induced by triadimefon, propiconazole, and myclobutanil at 4, 30, and 90 days of treatment (Allen et al., 2006) coupled with the increased Cyp2b20 gene expression at these time points by all 3 conazoles are consistent with a CAR-mediated hepatic hypertrophy.

Impacts of Altered Cholesterol Levels

Conazoles inhibit lanosterol 14α-demethylase (CYP51), which can reduce serum cholesterol levels. Mice fed cyproconazole in the diet had decreased serum cholesterol levels (Peffer et al., 2006). Triadimefon reduced serum cholesterol levels after 4 days of treatment. Reduced cholesterol levels in the presence of normal or increased expression of cholesterol biosynthetic genes can lead to altered arachidonic acid (AA) metabolism, altered cell growth and division, and potential adverse effects from overproduction of cholesterol intermediates.

Cholesterol constitutes a major component of animal plasma cell membranes and alterations in intracellular levels of cholesterol could impact membrane integrity. Phospholipase A2 group 6, an enzyme that participates in maintaining cell membrane integrity and liberates AA from the membrane, is significantly up-regulated after 90 days in all 3 conazoles but more so with propiconazole and triadimefon. If phospholipase A2 group 6 activity was increased, intracellular levels of AA should increase. Although levels of AA were not measured directly, there were significant increases in the transcription of Cyp2c55 andCyp2c37, both of which metabolize AA (Luo et al., 1998; Wang et al., 2004). In fact, Cyp2c55 was one of the most up-regulated genes at all time points for all conazoles. The liberation of AA by the hydrolysis of cell membrane glycerophospholipids is an important event that can lead to the generation of biologically active lipid mediators, such as prostaglandins and leukotrienes (Jampilek et al., 2006; Luo et al., 2006). In addition prostaglandin E synthase, which converts AA to prostaglandins was induced by propiconazole and triadimefon. These COX-2-derived bioactive lipids are known to stimulate cell migration, proliferation and tumor-associated neovascularization while inhibiting cell death (Backlund et al., 2005).

Cell growth and division are more generally dependent on cholesterol (Brown and Goldstein, 1974; Chen et al., 1974, 1975). Prolonged cholesterol depletion inhibited cytokinesis, induced the formation of polyploid cells (Fernandez et al., 2004), and prevented interaction with oxysterol binding proteins and their control of key signaling pathways including mitogenic pathways through pERK (Wang et al., 2005).

The overproduction of cholesterol intermediates could exert the following 3 potential adverse effects. 1. The normal or increased expression of cholesterol biosynthesis genes suggests that the gene products are active in the biosyntheses of cholesterol intermediates up to the step of lanosterol demethylation (CYP51). This suggests increased levels of lanosterol. The further conversion of lanosterol to Vitamin D3 involves a series of sterol intermediates as well as the conversion of the lanosterol precursor, squalene-2,3-epoxide to Vitamin D2. Some of these intermediates are subject to metabolism by enzymatic or free radical processes to oxysterols, which can be cytotoxic, mutagenic, and tumorigenic (Guardiola et al., 1996). 2. The inhibition of CYP51 could produce increasing levels of squalene-2,3-oxide, a substrate and an inhibitor for epoxide hydrolase (Oesch et al., 1971), an enzyme involved in xenobiotic detoxification and arachidonic acid catabolism (Newman et al., 2005). 3. Mevalonic acid is the product of the rate-limiting step in cholesterol biosynthesis and serves as an initiator for DNA replication. The overproduction of mevalonate is linked to malignant transformation (Siperstein, 1984). Moreover, mevalonate has been found to promote the growth of tumors in vivo and stimulate cancer cell proliferation in vitro (Duncan et al., 2004). The combination of these altered pathways and associated functional outcomes may contribute to the hepatic toxicity and associated liver tumor response in triadimefon- and propiconazole-treated mice.

After 30 days of treatment, all 3 conazoles decreased serum cholesterol levels and increased serum triglyceride levels (Allen et al., 2006). The biosynthesis of triglycerides requires factors such as acetyl CoA and pyruvate, intermediates also required for cholesterol biosynthesis. The overproduction of triglycerides may be a result of the shunting of acetyl-CoA and pyruvate from the cholesterol pathway to triglyceride biosynthesis. These possible metabolic shifts are consistent with transcriptional alterations in the pyruvate and fatty acid metabolic pathways induced by all 3 conazoles. Alternatively, the observed changes in serum cholesterol and triglycerides are consistent with the activation of LXR/RXR (Peet et al., 1998; Edwards et al., 2002) by oxysterols, other cholesterol pathway intermediates.

Possible Role of Reactive Aldehydes and Reactive Oxygen Species

Propiconazole induced Akr1b7 overexpression at all days of treatment with the highest recorded level of 8.7-fold after 30 days of treatment. This aldo-keto reductase is associated with the detoxification of lipid peroxidation products, the cleavage of isocaproaldehyde from the cholesterol side chain, and is regulated by LXRα (Volle et al., 2004). Lipid peroxidation products can arise from reactive oxygen species generated from high activities of CYP. High levels of CYP activities induced by phenobarbital have been associated with oxidized DNA and hydroxyl radical formation in rat liver (Imaoka et al., 2004), and high levels of CYP activities induced by procimidone are associated with reactive oxygen species in the livers of mice (Sapone et al., 2003). Concomitant with Akr1b7 overexpression induced by propiconazole was Gadd45a (2.3-fold) and Gadd45b (4.7-fold) overexpression after 30 days of treatment. Gadd45a and Gadd45b over-expression have been associated with DNA damage, apoptosis, and general genotoxic stress (Hollander et al., 2001; Gupta et al., 2005, 2006).

Retinoic Acid Biosynthesis and Metabolism

Triadimefon, after 4 days of treatment, resulted in only 4 uniquely altered pathways, 1 of which controlled retinol metabolism. Cyp26a1 (retinoic acid 4-hydroxylase) was overexpressed at every time point with triadimefon high-dose treatment. These results suggested that retinoic acid levels might be depressed in triadimefon-treated mice. The impact of depressed retinoic acid levels in the mouse liver tissues is unknown, however, Cyp26a inhibitors have been developed as potential anti-cancer agents (Njar, 2002). Furthermore, retinoic acid has anti-proliferative and cancer preventative properties and has been used to treat human cancers. Finally, Vitamin A deficiency has been linked to increased susceptibility to carcinogenesis in animal models and is associated with tumor progression (Lotan, 1996a, 1996b; Osanai and Petkovich, 2005).

Cell Signaling

Propiconazole significantly altered the Wnt signaling pathway at all time points. Cadherin-catenin complexes are important in mediating Wnt signaling, (Conacci-Sorrell et al., 2002). Increased β-catenin protein is a common marker for Wnt signaling in multiple types of cancer including liver cancer (Behrens, 2000; Sangkhathat et al., 2006). β-catenin complexes with the adenomatous polyposis coli (Apc) gene and axin to regulate its phosphorylation by glycogen synthase kinase (Gsk3b) and subsequent proteosomal degradation (Henderson and Fagotto, 2002).

Propiconazole also altered the IGF/PI3K/AKT/PTEN/mTOR pathways after 30 days of treatment. The IGF/PI3K/AKT/PTEN/mTOR pathways are critical mediators of oncogenic signaling. Exposure to propiconazole significantly decreased expression of insulin like growth factor1 (IGF-1). IGF-1 and its receptor (IGF-1R) provide a potent proliferative signaling system that stimulates growth and blocks apoptosis in many different cell types. PTEN, a tumor suppressor gene, is a downstream regulator of IGF signaling and its expression was down-regulated more than 2-fold by high-dose propiconazole after 90 days of treatment. Inactivation of PTEN activates the AKT pathway (Lu et al., 2003; Radu et al., 2003; Puc et al., 2005). The serine-threonine protein kinase Akt mediates many of the downstream effects of PI3K and plays a central role in signaling by PI3K (Aoki et al., 1998). Propiconazole after 90 days of treatment, primarily altered a set of pathways which are activated by a series of receptor ligands: IGF, insulin, calcium, cytokines, EGF, PDGF, FGF, estrogen, vascular endothelial growth factor (VEGF), and TGF-β. This transcriptional profile characterizes the cellular changes associated with inflammation, cell transformation, cell cycle arrest, and subsequently apoptotic signals.

By contrast triadimefon exposure at 4 days altered only 2 signaling pathways, cytokine receptor interaction, and a nuclear receptor pathway. After 30 days of treatment, triadimefon altered no cell growth pathways and 21 signaling pathways, 12 of which were common to propiconazole. After 90 days of treatment, triadimefon altered a smaller set of signaling pathways.

After 90 days of treatment, both triadimefon and propiconazole altered the expression of genes and molecular pathways that control the adherens junctions and the actin cytoskeleton. The adherens junction and the actin cytoskeleton are physically juxtaposed and coordinately regulated. Adherens junctions are composed of cadherin-catenin complexes linked to the actin cytoskeleton (Tian and Phillips, 2002). Chemically mediated disruption of these complexes can result in de-differentiation of cells and proliferation. Among the adherens junction genes most influenced by triadimefon and propiconazole exposure was transforming growth factor β receptor II (Tgfbr2). Altered transcript levels of Tgfbr2 have been associated with liver cancer (Im et al., 2001).

Changes in the expression of actin cytoskeleton genes are consistent with disturbance of adherens junction complexes and their signaling. Adherens complex signaling and linkage between membrane-associated proteins and the actin cytoskeleton are controlled by radixin (Rdx) and phosphatidylinositol 3-kinase C2 domain containing, gamma peptide (Pik3c2g) (Yonemura et al., 2002; Larue and Bellacosa, 2005). Rdx was down-regulated and Pik3c2g was up-regulated by high-dose propiconazole and high-dose triadimefon at 90 days. These expression changes could be associated with destabilized cell attachment and could contribute to conazole-induced hepatotumorigenesis.

Conclusions

There are a number of well-established mechanisms of action for rodent hepatic tumorigens. These include both genotoxic and nongenotoxic mechanisms. Based on short-term studies many conazoles are considered to be nongenotoxic. One of the most common nongenotoxic mechanisms is a phenobarbital-like mechanism that features CAR activation, hepatocyte hypertrophy, induction of the CYP2B enzyme family, induction of cell proliferation, and inhibition of apoptosis (Holsapple and Lehman-McKeeman, 2005). All 3 conazoles, regardless of their tumorigenic potential, produced the same responses. However, since myclobutanil was not tumorigenic, we used comparative genomic analyses to identify clues that might explain triadimefon’s and propiconazole’s tumorigenic activities. Current evidence suggests that triadimefon and propiconazole have different complex patterns of genomic profiles with some common features. This suggests that the aggregate events that might contribute to each of their modes of tumorigenic action would consist of a number of factors, some common to both conazoles and some unique to each. It seems reasonable that the initial cellular events occurring after high-dose conazole exposure are those involving the activation of a series of nuclear receptors (Ahr, CAR, PXR, LXR, PPAR, FXR) associated with the inhibition of CYP51. For triadimefon this results in high levels of CYPs (Cyp1a2, Cyp2b20, Cyp3a11, Cyp26a1, Cyp2c37, Cyp2c55), decreased serum cholesterol, and increased expression of sterol and cholesterol biosynthesis. These changes might induce oxidative stress as seen by the overexpression of Akr1b7, Hspa1, and Gstα, and result in decreased retinoic acid levels. After 90 days of treatment, triadimefon alters some metabolic pathways (amino acid, carbohydrate, lipid) and some cell growth and apoptosis pathways as well as cell cycle-G2/M checkpoint, calcium signaling, and EGFR signaling pathways. Since Akr1b7 and Cyp26a1 are still over-expressed, oxidative stress and decreased retinoic acid levels are possible. Oxidative stress has been linked to chemical carcinogenesis and the production of free radicals (Klaunig and Kamendulis, 2004). These changes in toto might have implications in the mode of action of triadimefon as discussed before.

Propiconazole exposure after 4 days induced high levels of CYPs (Cyp1a2, Cyp2b20, Cyp3a11, Cyp26a1, Cyp2c37, Cyp2c55). These changes might also result in the induction of oxidative stress as seen in the marked overexpression of Akr1b7. Propiconazole, unlike triadimefon, alters a large number of metabolic pathways (amino acid, carbohydrate, lipid) and cell growth, apoptosis, and signaling pathways. After 90 days of treatment of propiconazole, oxidative stress is indicated by Hspa1a, Hspa1b, and Gadd45b overexpression. Among the many pathways altered are apoptosis, cell cycle-G2/M checkpoint, calcium signaling, EGFR and WNT-β-catenin signaling pathways. Propiconazole is the only conazole to alter the WNT-β-catenin pathway and depress the negative regulator of the cell-survival signaling pathway, PTEN, suggesting an increase in proliferative signals.

While the tumorigenic triadimefon and propiconazole, and the non tumorigenic myclobutanil exerted similar effects on liver hypertrophy, histological effects, AROD activities, cell proliferation and serum cholesterol in the mouse, genomic analyses described different gene expression profiles and potential tumorigenic modes of action. Toxicogenomic analyses of these conazoles enabled us to develop general conclusions and formulate novel hypotheses regarding their mode of action as hepatotumorigens. Both propiconazole and triadimefon likely activate nuclear receptors, particularly CAR, PXR, LXR, PPAR, and FXR leading to overexpression of CYPs and associated induction of oxidative stress. These effects can result in genomic damage subsequent to elaboration of reactive oxygen species and reactive aldehydes. Both propiconazole and triadimefon decreased cholesterol levels that have been associated with polyploidy and disruption of mitotic cell cycling. The hypothesis for triadimefon also includes the up-regulation of the cholesterol biosynthetic pathway and an influence on the retinoic acid catabolism pathway. Altered retinoic acid levels have been shown to stimulate cell proliferation through key signaling pathways. The hypothesis for propiconazole includes down-regulation of the PTEN pathway and up-regulation of the WNT-β-catenin signaling pathway, which would stimulate cell proliferation. These proposed mechanistic considerations are key to the identification of important events that lead to their respective tumorigenic outcomes.