Transcriptional Profiles in Liver from Rats Treated with Tumorigenic and Non-tumorigenic Triazole Conazole Fungicides: Propiconazole,Triadimefon,and Myclobutanil

Abstract

Conazoles are a class of fungicides used as pharmaceutical and agricultural agents. In chronic bioassays in rats, triadimefon was hepatotoxic and induced follicular cell adenomas in the thyroid gland, whereas, propiconazole and myclobutanil were hepatotoxic but had no effect on the thyroid gland. These conazoles administered in the feed to male Wistar/Han rats were found to induce hepatomegaly, induce high levels of pentoxyresorufin-O-dealkylase, increase cell proliferation in the liver, increase serum cholesterol, decrease serum T3 and T4, and increase hepatic uridine diphospho-glucuronosyl transferase activity. The goal of the present study was to define pathways that explain the biologic outcomes. Male Wistar/Han rats (3 per group), were exposed to the 3 conazoles in the feed for 4, 30, or 90 days of treatment at tumorigenic and nontumorigenic doses. Hepatic gene expression was determined using high-density Affymetrix GeneChips (Rat 230_2). Differential gene expression was assessed at the probe level using Robust Multichip Average analysis. Principal component analysis by treatment and time showed within group sample similarity and that the treatment groups were distinct from each other. The number of altered genes varied by treatment, dose, and time. The greatest number of altered genes was induced by triadimefon and propiconazole after 90 days of treatment, while myclobutanil had minimal effects at that time point. Pathway level analyses revealed that after 90 days of treatment the most significant numbers of altered pathways were related to cell signaling, growth, and metabolism. Pathway level analysis for triadimefon and propiconazole resulted in 71 altered pathways common to both chemicals. These pathways controlled cholesterol metabolism, activation of nuclear receptors, and N-ras and K-ras signaling. There were 37 pathways uniquely changed by propiconazole, and triadimefon uniquely altered 34 pathways. Pathway level analysis of altered gene expression resulted in a more complete description of the associated toxicological effects that can distinguish triadimefon from propiconazole and myclobutanil.

Introduction

Conazoles are a class of azole fungicides used both in agriculture and as pharmaceuticals. In agriculture, conazoles are used to prevent fungal growth on fruit, vegetables, and seeds. As pharmaceuticals, they are used for the treatment of local and systemic fungal infections. Conazoles can be divided into two major categories according to their chemical structure. One group, such as ketoconazole and miconazole, contain an imidazole ring. The other group, such as propiconazole, triadimefon, and myclobutanil, contain a 1,2, 4-triazole ring. Conazoles were designed to specifically inhibit CYP51, a sterol 14-α-demethylase which catalyzes the 14-α-demethylation of lanosterol in the ergosterol biosynthesis pathway in fungi. The biosynthesis of ergosterol is critical to the formation of fungal cell walls. The lack of normal sterol production slows or stops the growth of the fungus, effectively preventing further infection and invasion of host tissues (Ghannoum and Rice, 1999). Mammalian cholesterol synthesis is also blocked by conazoles at the step of the sterol 14-α-demethylase, but a much higher concentration is required to cause the same degree of inhibition as in fungi (de Brabander et al., 1980; Van de Bossche, 1980).

Myclobutanil (Systhane or Eagle), propiconazole (Banner), and triadimefon (Bayleton) are triazole based conazoles used in agricultural settings. In 2-year feeding studies in rats and mice with propiconazole, there were no tumors in the rat at any level of the chemical in the feed; however, significant increases were noted in the incidence of benign and malignant liver tumors in male mice at the highest feeding level (2500 ppm) (Extoxnet, 1997; INCHEM, 1987). Propiconazole was shown to be a rat liver tumor promoter as rats treated with diethylnitrosamine then subjected to a partial hepatectomy and fed propiconazole produced increases in foci positive for the placental form of glutathione S-transferase compared to controls (Hasegawa and Ito, 1992) In 2-year feeding studies, high-dose (1800 ppm) triadimefon was associated with a slightly higher incidence of thyroid follicular cell adenomas in male Wistar/Han rats and an increased incidence of hepatocellular adenomas in male and female NMRI and CF₁/W74 mice (EPA, 1996; INCHEM, 1981). Myclobutanil did not show any treatment-related tumorigenic effects in 2-year studies in mice or rats (2000 ppm) (INCHEM, 1992).

Although myclobutanil, propiconazole, and triadimefon are all triazole-based conazoles and are all CYP51 inhibitors, as described here, their profiles of toxic responses in rats are different. The mode of action ascribed to triadimefon in the induction of thyroid tumors in rats was suggested to be through endocrine effects, disruption of the hypothalamic-pituitary-thyroid hormonal balance. Since liver is the major site for the metabolism of these compounds as well as the major site of thyroid hormone metabolism, we investigated the gene expression profiles in rat liver after conazole treatment.

In an accompanying paper (this issue) Wolf et al. describe in-life studies in rats with triadimefon, propiconazole, and myclobutanil in which traditional toxicity endpoints were evaluated. No unique set of toxicological responses was found that could differentiate the conazoles from each other. The present analyses of conazole-altered gene expression represent an extension of these studies aimed at the identification of characteristic changes in transcriptional profiles.

The profiling of gene expression alterations in response to pharmaceuticals and chemicals by DNA microarray approaches in rat liver has been used to identify potential mechanisms of toxicity (de Longueville et al., 2002; Huang et al., 2004; Kier et al., 2004; McMillian et al., 2004; Meneses-Lorente et al., 2003; Tully et al., 2006; Waring et al., 2001). Using low-density arrays (973 genes), gene expression studies of cultured rat hepatocytes exposed to 15 structurally different chemical hepatotoxins were able to identify clusters of chemicals with similar toxic mechanisms (Waring et al., 2001). In similar studies with a series of 11 pharmaceuticals and related chemicals, including phenobarbital, a low-density array containing 59 genes was able to classify those chemicals which caused the hepatic alterations of cholestatis, steatosis, necrosis, or cytochrome P450 (CYP) induction (Meneses-Lorente et al., 2003). RNAs from rats treated with CYP inducers including phenobarbital, pregneneolone-16α-carbonitrile, miconazole, or clotrimazole and analyzed by genomic techniques using low density arrays (de Longueville et al., 2002 ; Meneses-Lorente et al., 2003) identified genes that related histopathological changes to expression profiles.

Similar studies in rats have been conducted focusing on a set of hepatic oxidant stressors/reactive metabolites (McMillian et al., 2004). Recently, microarray gene expression profiling in liver and testis has been used to characterize 4 conazoles, triadimefon, propiconazole, myclobutanil, and fluconazole. In this study, adult Sprague–Dawley rats were exposed through gavage for 14 days, and gene expression profiles were analyzed on medium density (4273 genes) arrays (Tully et al., 2006). The major findings were increases in CYP and other xenobiotic metabolizing genes, and increased transporter gene expression.

The objective of the present study was to provide a method to differentiate the various treatments through the systematic analysis of hepatic gene expression using high-density arrays. This report characterizes the transcriptional responses in rat liver after in vivo exposure to 3 conazoles, compares these findings to the observed phenotype and conventional toxicologic measurements, and describes potential cellular and molecular pathways that might be involved in the toxic effects of these conazoles and that could be used to uniquely identify each treatment.

Materials and Methods

Experimental Design

The complete details of the in life experiments are described in Wolf et al. (2006). All studies mentioned in that report were approved by an Institutional Animal Care and Use Committee and conform to the Guide for the Care and Use of Laboratory Animals. Liver samples from that study were used herein for genomic analysis. Briefly, male Wistar/Han rats (3/dose) received triadimefon (100, 500, 1800 ppm), propiconazole (100, 500, 2500 ppm), myclobutanil (100, 500, 2000 ppm) in the feed, or untreated feed for periods of 4, 30, or 90 days. At each time point, rats were euthanized by CO₂ asphyxiation and then necropsied. Livers were removed, a slice from each lobe combined into a tube, frozen in liquid N₂ within 2 minutes of death, and stored at −80°C for subsequent RNA isolation.

RNA Preparation

Total RNA was isolated from rat livers using TRI Reagent (Molecular Research Center Inc., Cincinnati, OH) and following the manufacturer’s directions. Briefly, the liver tissue was ground up to a fine powder in liquid N₂, and added to 5 ml of TRI Reagent. Then, 1-bromo-3-chloropropane (1 ml) was added and tubes were vortexed and remained 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 ice cold isopropanol (2.5 ml) was added. The samples were left at −70°C overnight followed by centrifugation at 10, 000 × g for 15 minutes at 4°C to precipitate the RNA. The RNA pellet, visible after centrifugation, was washed with cold 70% ethanol and air-dried. The RNA was resuspended in Rnase-free water and its purity and concentration was measured on an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA).

Microarray Experiments

Sample preparation, processing, and hybridization to the Affymetrix Rat 230_2 chips were performed at Expression Analysis, Durham, NC as described in the GeneChip Expression Analysis Manual (Affymetrix; Santa Clara, CA). Information on the Rat 230_2 Genome chip, which allows for the analysis of approximately 22K annotated genes and approximately 9K untranscribed sequence (ESTs) clusters, is available at <http://www.affymetrix.com/products/ arrays/specific/rgu230_2.affx>. In the present study, 1 chip was used per animal and sample (3 chips/group).

Overall Data Analysis Strategy

Our approach to the analysis of this data set, consisting of 3 time points, 3 conazoles, and 3 doses, was to

1. evaluate the data quality;

2. perform unsupervised principal components analysis (PCA) as a global inspection of within-group sample correspondence;

3. perform a supervised statistical analysis to find differentially expressed genes between each treatment and concurrent control and between each conazole;

4. examine these statistical lists to find pathways overrepresented;

5. mine the data to find pathways that were unique or in common among the treatments by mapping filtered genes to functional pathways;

6. analyze for a possible dose-response effect; and 7. inspect the gene expression responses in context of the phenotypic response measures of histopathology, serum lipid profiles, thyroid hormone levels, and liver enzyme activity.

Microarray Data Analysis

Data were obtained using the Affymetrix GCOS 1.3 software version (Affymetrix; Santa Clara, CA). The Rat 230_2 GeneChip array consists of 31,099 probe sets (including 58 control genes). A total of 79 genechips (3 treatment groups, 3 time points, 2 or 3 doses, and concurrent day controls) were hybridized. To survey the data for within-group outliers and trends in data quality, principal component analysis (PCA; Array Tools; Biometric Research Branch, NCI), pairwise correlation analysis (R- Bioconductor; AffylmGUI), and signal intensity histograms for each chip (Hierarchical Cluster Explorer, Univ. of Maryland) were conducted. Data used in PCA analysis was log2 transformed prior to calculations. Data were quantile normalized using Robust Multichip Average (RMA) in R-Bioconductor) to generate estimated expression summaries. A gene expression filter, using the normalized value for each gene was applied to exclude unchanging genes (0.9 to 1.2 of mean expression in log2 space) leaving the variable genes for subsequent analyses.

For each time point, high- and mid-doses were grouped for combined analysis, yielding a total of 21 analytic samples, to increase the total number of samples for subsequent statistical analysis. To investigate differentially expressed genes between treatment and control groups, 1-way ANOVA models (across all groups) were fitted to each probe set individually, using chemical (myclobutanil, propiconazole, tri-adiemefon) as the variable. This was followed by pairwise group Student’s t-test (using 3 untreated and 3 treated). The variance statistic was generated using the Cross-Gene Error Model (Rocke-Lorenzo; GeneSpring version 7.2; Redwood City, CA). A False Discovery rate (FDR) of 5% was applied to control for family-wise errors. Data were log2 transformed prior to the calculations. The low-dose samples were used to construct a separate experimental grouping; the same procedures were carried out with these samples as described for the high-mid experiment. For inspection of individual genes, a 2-sample (control vs. treated) Student’s t-test was performed using a p value < 0.05. We refer to the transcript detected by 1 or more probesets by its gene name in the following text.

Pathway Level Analysis

Four software programs were used to map differentially expressed genes to signaling and metabolic/biochemical pathway maps: DAVID/EASE <http://david.niaid.nih.gov/ david/ease.htm>, KEGG and Biocarta pathways Array-Track <http://edkb.fda.gov/webstart/arraytrack/>, Meta-Core GENEGO, <http://trials.genego.com/cgi/index.cgi#Information>, and Ingenuity Pathway Analysis (Mountain View, CA; <http://www.ingenuity.com/index.html>.

The p values throughout MetaCore, for maps, networks, and processes were all calculated using the same basic formula: a hypergeometric distribution in which the p value essentially represents the probability of particular mapping arising by chance, given the numbers of genes in the set of all genes on maps/networks/processes, genes on a particular map/network/process, and genes in the experiment. We considered pathways significant at p ≤ 0.05 and defined a pathway as having a minimum number of 3 genes. Selected groups of genes are discussed after first verifying those individual genes as genes participating in significant pathways reflected in Tables 2–9.

Results

Gene Expression Analyses: Microarray Filtering

A gene expression filter was applied to remove invariant genes (genes with normalized values centered around 1 across all groups) before statistical analysis. The total number of changed genes (total number of genes assayed less unchanged genes) at 90 days for high-dose and mid-dose expression was 14,840. Expression of a gene was considered altered by conazole treatment if the change was statistically significant at FDR/0.05. The total number of genes passing a 1-way ANOVA (Cross-Gene Error estimate of variance with a FDR/0.05) was 11,184 genes for the 90-day data set. In contrast, the total number of genes passing a 1-way ANOVA (Cross-Gene Error estimate of variance with a FDR/0.05) was 2,395 at 30 days and 113 at the 4-day time point.

The number of genes with altered expression resulting from statistical pairwise (treated to control) comparisons, over the 3 time points are presented in Table 1. At the 4-day time point, only the highest dose produced over 200 altered genes for each conazole. At the 30-day time point, the numbers of significantly different genes increased between the low- and mid-dose levels for myclobutanil and triadimefon compared to 4 days. The numbers of genes was lower at the high-dose compared to the mid-dose for these 2 conazoles. An inverse dose for gene response was observed with propiconazole at 30 days. At the 90-day time point, myclobutanil treatment did not produce substantial numbers of significantly changed genes. At 90 days greater than 10,000 genes changed after the mid- and high-dose triadimefon treatment, while propiconazole treatment had 8,452 altered genes and a dose-response for the numbers of altered genes.

PCA at the Chip Level

PCA provides a means to view multidimensional gene expression data in 3-dimensional space to reveal clusters in the experimental data. PCA was applied to the statistically significant genes from high dose treatment of each conazole after 4, 30, and 90 days of exposure. This analysis reduces the dimensionality of the data and captures the variability of the datasets (Figure 1). Good separation of groups was observed between the high doses of the 3 chemicals and control, while the 3 samples within each treatment group clustered together. Each chip is represented as a single point. The points that cluster together in space reflect similar expression profiles. The PCA analysis illustrates the amount of variability between individual animals associated with a chemical treatment. For each time point, 40–50% of the variability was captured in the first PCA dimension. The other PCA dimensions represent the lower percentages of the sample variability.

Venn Analyses of Genes

The lists of statistically significant genes (FDR/0.05) represented in Table 1 were applied to a Venn visualization to identify the number of genes that were unique and common to the chemical treatments (Figure 2). This analysis identified a number of genes that were common to the 3 conazoles at each time point: 190, 112, 23 genes for 4 days, 30 days, and 90 days, respectively. Higher numbers of genes (up to 7,438) that were common to 2 conazole pairs were observed depending on the time point and the conazole pair. At each time point, the numbers of genes unique to triadimefon was greater than the other two conazoles (Figure 2). The numbers of genes unique to propiconazole or triadimefon dramatically increased from the 30-day time point to the 90-day time point. At 90 days, the expression of 979 and 2876 genes was uniquely associated with propiconazole and triadimefon, respectively.

Mapping to Biochemical Pathways

Genes that had a FDR/0.05 in pairwise t-tests were submitted for pathway level analysis. Several pathway programs (described in Materials and Methods) were employed to identify which functional pathways were reflected in the differentially expressed pairwise comparisons. The lists represented filtered genes statistically significant with a t-test, FDR of p < 0.05. A pathway was identified based on the presence of 3 or more genes that can be put together to define a coordinated biological function or process. The results from the various pathway analysis approaches were consistent with the numbers of significant genes. Triadimefon and propiconazole altered more pathways compared to myclobutanil across the 4 and 30 day time points (Figure 3). After 4 days of treatment triadimefon altered 48 pathways, while propiconazole and myclobutanil altered, 11 and 7 pathways, respectively (Tables 2–4).

The majority of the pathways could be functionally characterized as cell signaling, and growth and differentiation pathways. After 30 days of treatment propiconazole and myclobutanil altered 11 and 7 pathways, respectively whereas triadimefon altered 18 pathways (Tables 2–4). The greatest number of pathways was altered after 90 day of treatment for both propiconazole and triadimefon and was associated with cell signaling, metabolism, and growth (Figure 4). There was no significant alteration of pathways after 90 days of treatment with myclobutanil. Triadimefon and propiconazole altered 105 and 108 pathways, respectively (Tables 8 and 9). Pathways were grouped into signaling, metabolism, and growth categories (Figure 4). Triadimefon and propiconazole altered categories of pathways in the order: growth> metabolism> signaling. Within each category both conazoles altered similar numbers of categories (Figure 4).

Venn Analyses of Pathways

A Venn analysis revealed pathways common to all conazoles and those associated with only one or two conazoles (Figure 5, see also Tables 2–9). After 4 or 30 days of treatment, there were 4 and 3 pathways, respectively, common to the 3 conazoles. There were no common pathways between the 4-day and 30-day time points. After 4 days of treatment growth and differentiation pathways were most affected, whereas after 30 day of treatment metabolism regulation pathways were most affected. Triadimefon had the most unique pathways at 4-day and 30-day time points, with 42 after 4 days, and 9 after 30 days. At the 90-day time point, propiconazole had 37 unique pathways compared to triadimefon’s 34 unique pathways.

Correspondence of Hepatic Xenobiotic Enzyme Activities and Serum Lipid Levels With Gene Overexpression

All 3 conazoles evaluated induced CYP metabolizing enzyme activities related to CYP1A1 (ethoxyresorufin-O-dealkylase, EROD) and CYP2B1/2 (pentoxyresorufin-O-dealkylase, PROD), and propiconazole induced CYP1A2 (methoxyresorufin-O-dealkylase, MROD) activities (Wolf et al., 2006). All three conazoles also induced UDP-glucuronosyltransferase activity (UGT), and caused altered thyroid hormone levels (Wolf et al., 2006). A dose related increase in expression of several of these enzymes after 30 days of treatment is presented in Figure 6, where CYP2B1/2, CYP1A1, Udpgtr2 and the related Aldh1a1, and Gsta2 genes increase with dose.

Of the 3 conazoles evaluated, myclobutanil and propiconazole increased serum cholesterol levels after 4 days of treatment while only triadimefon increased serum cholesterol levels after 90 days of exposure (Wolf et al., 2006). After 90 days of triadimefon treatment there was a significant overexpression of transcripts involved in cholesterol biosynthesis. These genes were: mevalonate pyrophosphate, decarboxylase (Mvd) isopentenyl-diphosphate delta isomerase (Idi1), isopentenyl-3-hydroxy-3-methylglutaryl-Coenzyme A reductase (Hmgcr), 3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1(Hmgcs1), and cytochrome P450, subfamily 51 (CYP51).

Comparing expression levels of genes involved in thyroid hormone metabolism, the UGTs, to their corresponding function, 3 UGT genes including UGT1A6, UGT2B, and UGT8 were significantly induced at 90 days in concordance with the functional assays at all 3 time points. This analysis was able to link the transcriptional response with the alterations in functional proteins previously analyzed (Wolf et al., 2006).

Discussion

The goal of the present study was to comprehensively investigate relationships of the toxicity of three different conazoles, triadimefon, propiconazole, and myclobutanil, with their transcriptional profiles. Each of the conazoles had been previously evaluated in chronic studies in rats. Triadimefon induced rat thyroid tumors, propiconazole was a hepatic tumor promoter, and myclobutanil was devoid of tumorigenic properties in rats (EPA, 1996; INCHEM, 1987, 1992). In a companion study under conditions that mimicked the chronic bioassay, these 3 conazoles were administered in the feed to male Wistar/Han rats for 4, 30, and 90 days. All 3 conazoles were found to induce hepatomegaly, to induce high levels pentoxyresorufin-O-dealkylase (PROD) activity, to increase cell proliferation in the liver, to increase serum cholesterol levels, to decrease T3, T4 levels, and increase UGT enzyme activities (Wolf et al., 2006).

Propiconazole and triadimefon decreased TSH levels and only triadimefon increased cell proliferation in the thyroid gland (Wolf et al., 2006). The triadimefon-induced rat thyroid tumors were proposed to arise secondary to hepatic metabolism of T4 with an associated persistent increase in TSH. However, the data from Wolf et al. (2006) suggested that although there was induction of UGT enzyme activity and a mild decrease in serum T3 and T4 levels, there were no associated increases in TSH levels. Thus triadimefon-induced thyroid tumors are likely due to another mode of action. The purpose of the present study was to identify potential modes of toxic responses of conazoles through the use of a systematic analysis of hepatic gene expression. A previous study comparing gene expression profiles of triadimefon, propiconazole and myclobutanil in livers of Sprague–Dawley rats after gavage exposure for 14 days showed altered expression of CYP genes, xenobiotic response genes, and transporter genes (Tully et al., 2006).

Similar to the present study, Tully et al. (2006) found that triadimefon and propiconazole altered similar numbers of genes, with myclobutanil altering far fewer genes. It was of interest to relate toxicological findings to concurrent changes in gene expression for a series of endpoints based on high dose exposure.

Serum Cholesterol

The triazole-containing conazoles are designed to inhibit CYP51, lanosterol-14-α-demethylase, which blocks the progression of lanosterol to ergosterol in fungi. Biosynthesis of ergosterol is critical to the formation of fungal cell walls. It has been reported that conazole exposure in mammals results in a similar effect as cholesterol biosynthesis is also blocked at the step of lanosterol 14-α-demethylase (Strandberg et al., 1987).

The inhibition of CYP51 by conazoles would be expected to decrease serum cholesterol levels as reported for another CYP51 inhibitor (Harwood et al., 2005). We also observed, in a parallel study in mice, that all 3 conazoles decreased serum cholesterol levels (Allen et al., 2006). However, in the rat, serum cholesterol was increased by propiconazole and myclobutanil after 4 days of treatment and by triadimefon after 90 days of treatment. This result could be a result of two related events; overexpression of 3-hydroxy-3-methylglutaryl-CoA, the rate-limiting step in cholesterol biosynthesis, and a simultaneous lack of CYP51 inhibition due to insufficient conazole levels in the rat liver.

Rats fed dietary ketoconazole showed lower oxysterol levels, which may release the inhibitory effect on 3-hydroxy-3-methylglutaryl-CoA reductase enzyme activity leading to increased cholesterol levels (Tamasawa et al., 1997). Higher concentrations of conazoles are required to inhibit mammalian CYP51 compared to fungal CYP51 (de Brabander et al., 1980). A second possibility is that increased cholesterol levels might be associated with the increased endoplasmic reticulum proliferation that occurs in conjunction with induction of xenobiotic metabolizing enzymes (Glaumann and Dallner, 1968; Ridsdale et al., 2006).

A third possibility to explain increased serum cholesterol levels involves reverse cholesterol transport or the transport of cholesterol from peripheral tissues to the liver and then into the serum (Olkkonen and Levine, 2004). This process is mediated by oxysterols which are signaling lipids that activate liver X receptor (LXR) and oxysterol binding proteins (OSBP) (Olkkonen and Levine, 2004). LXRs heterodimerize with the retinoic X receptor. Oxysterols also activate pregnane X receptor (PXR) which is involved in CYP3A induction (Shenoy et al., 2004) and CYP3A gene expression was increased by all 3 conazoles in this study. Taken together, one or more of the events described here could account for increased serum cholesterol levels in the presence of a CYP51 inhibitor.

Cholesterol Biosynthesis

Strong up-regulation (increased 2-fold compared to control) of 2 key genes involved in cholesterol biosynthesis, 3-hydroxy-3-methylglutaryl-CoA and squalene epoxidase was observed, with the greatest effect seen is after high-dose triadimefon treatment for 90 days. The product of the gene encoding cholesterol 7 alpha-hydroxylase (CYP7A1), is known to catalyze the rate-limiting step in the bile acid pathway (Goodwin et al., 2003). CYP7A1 is also stimulated by LXR-α, which is a receptor for oxysterol metabolites of cholesterol. This then provides a mechanism for clearance of cholesterol from the body. In rodents, the forward regulation of CYP7A1 by oxysterolsacting through LXRαresults in increased catabolism of cholesterolto bile acids (Goodwin et al., 2003). In the present study, CYP7A1 was strongly induced by propiconazole and less so by triadimefon and myclobutanil at 90 days. The nuclear PXR receptor regulates the expression of genes involved in the biosynthesis, transport, and metabolism of bile acids including cholesterol 7alpha-hydroxylase (CYP7A1) and the Na(+)-independent organic anion transporter 2 (Oatp2) (Staudinger et al., 2001a, 2001b).

Hepatic Hypertrophy

Most studies on pesticides suggest that the liver one of the primary target tissues (Hurley, 1998). With the exception of amitrole and terbutryn, there is histological evidence that pesticides generally induce hepatocellular hypertrophy, increase liver weight, and/or increase smooth endoplasmic reticulum in the liver of at least one species. Other indicators of hepatic activity include increased mixed-function oxidase activity and biliary excretion (Hurley, 1998). The transcriptional response in the liver after conazole exposure in the present study identified cellular pathways and processes that differentiated each of the three conazoles and improved our understanding of the toxicity pathways associated with exposure.

The hepatocyte hypertrophy observed for all 3 conazoles, and most markedly for triadimefon, is thought to be associated with increased smooth endoplasmic reticulum proliferation (SER) as seen in Hurley (1998) and as reported for phenobarbital (Higgins, 1976). A function of SER is to hold the enzymes in the hepatocytes that synthesize steroid hormones, metabolize endogenous and exogenous substrates, synthesize glucose as well as perform other related functions. The overexpression of microsomal P450s (fatty acid ω-hydroxylases) typically results in extensive proliferation of SER (Szczesna-Skorupa et al., 2004) In our studies we observed significant increases in the expression of CYP4A12. Rat CYP4A proteins possess fatty acid ω-hydroxylase activities (Su et al., 2005), thus linking induction of CYP4A proteins and hepatic hypertrophy by triadimefon.

Nuclear Receptors and Xenobiotic Metabolizing Enzyme Induction

Some of the observed genomic and cellular responses induced by propiconazole and triadimefon are attributable to the activation of nuclear receptors. A significant alteration of genes reflecting phase I metabolism including CYP2B15, CYP4A, Aldh1a1, Aldh1a2, Aldh1, Adh4, andCes2 was identified for all conazoles at 90 days. Similarly, Phase II metabolism expression showed a significant consistent induction as shown by upregulation of Ugt2a1 and Gstm5 in propiconazole samples. These observations are consistent with nuclear receptors constitutive androstane receptors (CAR) and PXR activation (Maglich et al., 2002; Xu et al., 2005).

The overexpression of CYP1A1 and the induction of hepatic EROD activities are related to activation of the aryl hydrocarbon receptor (Ahr) and the overexpression of CYP2B and the induction of PROD activities are related to the activation of the CAR. The overexpression of CYP3A is related to PXR, the overexpression of CYP4A is related to the persoxisome proliferator activated receptor-alpha (PPARα), and the overexpression of CYP7A is related to the farnesoid X receptor (FXR) and LXR.

Many of the phase II enzymes, the UGTs and GSTs, which are overexpressed by propiconazole and triadimefon are regulated by CAR, PXR, and Ahr (Gong et al., 2005). In related studies that characterized CYPs induced by propiconazole and fluconazole in rats and mice, CYP1A2, CYP2B1/2, CYP3A1/23, and CYP3A2 were induced in rats (Sun et al., 2005, 2006) and are related to the activation of the CAR and PXR receptors.

Thyroid Hormone-Regulating Genes

A number of genes related to thyroid gland function were examined from the livers of the conazole-treated rats. This was done to identify an association between molecular functional changes and histological alterations present in the thyroid gland or serum thyroid hormone levels after conazole treatment. All 3 conazoles decreased serum T4 levels after 4 days, but only the high dose of propiconazole and triadimefon had decreased serum T4 level after 30 days (Wolf et al., 2006). Serum T3 was decreased by high-dose triadimefon after 4 days, but all conazoles decreased serum T3 levels after 30 days. Only 30 days of treatment with the high dose of triadimefon resulted in histologic alterations in the thyroid that included follicular cell hypertrophy, colloid depletion, and increased follicular cell proliferation (Wolf et al., 2006).

While a number of thyroid function associated genes present in the liver appeared to have a trend toward decreased expression compared to control for their pattern of changed expression after triadimefon treatment, only thyroid releasing hormone (TRH), thyroid releasing hormone receptor (TRHR), and thyroid stimulating hormone beta (TSHβ) were statistically significantly repressed by high-dose triadimefon after 90 days of treatment. Further, only triadimefon significantly repressed thyroperoxidase (TPO). In contrast, the expression level of thyroid hormone receptor-α (Thra) was significantly increased by triadimefon and propiconazole at 90 days. The thyroid transcription factors TTF-1 and Pax8 are known to work together inthe transcriptional activation of thyroid-specific genes suchas thyroglobulin (Tg), TPO, and the sodium/iodidesymporter (NIS) (Altmann et al., 2005). The observed significant repression of TPO by triadimefon may suggest an upstream transcriptional block accounting for the changes in circulating thyroid hormone levels and tissue response present in the triadimefon-treated rats.

Stress Response

A number of genes participating in a general cellular stress response were consistently overexpressed, including GSTs and glutathione peroxidase 2 and 4, Hsp70 and HSP70 interacting proteins, alpha crystalline (Cryab-Hsp20 family), Hspb4, and Hspa5. NAD(P)H: quinone oxidoreductase was significantly altered as was heme oxygenase 2 at 90 days. Both are known to be associated with the presence of increased cellular oxidative stress and their induction is a response to scavenge reactive oxygen species produced secondary to increasing oxidative stress (Radjendirane and Jaiswal, 1999; Jaiswal, 2000).

Cell Cycle Genes

All 3 conazoles induced increased hepatocyte proliferation after 4 days of treatment, which returned to control levels after 30 days for triadimefon and propiconazole, and by 90 days for myclobutanil (Wolf et al., 2006), suggesting a general proliferative conazole effect. A number of cell cycle related genes had uniformly changed expression relative to control after treatment with the various conazoles but only a few genes achieved significance. Cyclin B1 and Cdc20 were both significantly increased in expression after 4 and 30 days of myclobutanil treatment but not different after 90 days.

After 90 days of treatment Cdc20 was increased with propiconazole treatment however cyclin B1 was decreased after triadimefon treatment. Cdc 20 is associated with mitosis and cyclin B1 is associated with G2/M progression (Golias et al., 2004; Takeda and Dutta, 2005). In addition, triadimefon increased the expression of insulin-like growth factor 2 (IGF-2), and CSF2 (colony-stimulating factor 2) in the liver which have been associated with liver cell growth (Norstedt et al., 1988; Pinzani et al., 1989; Theocharis et al., 1999; Reeves et al., 2000). While these results correlate with the findings of increased cell proliferation in rat liver, it also suggests different control mechanisms that could account for the variable time course to the changes in proliferative rates.

DNA Damage

Gadd45α, a transcript involved in DNA damage signaling, was significantly increased in expression by high dose triadimefon as was Gadd45γgt 90 days. The p53-regulated Gadd45α, gene is one of the important players in cellular response to DNA damage, and thought to be involved in the control of cell cycle checkpoint, apoptosis and DNA repair (Jin et al., 2001). Gadd45α and Gadd45γ are cdc2/cyclinB1 kinase inhibitors and are thought to play a role in S and G2/M cell cycle checkpoints induced by genotoxic stress (Vairapandi et al., 2002). In this present study, the transcriptional profile of increased Gadd45α and Gadd45γ at 90 days by triadimefon, suggests DNA damage. Combining this profile with the observed decreased cyclin B1 known to function in G2/M cell cycle arrest (Vairapandi et al., 2002) by triadimefon at 90 days but not propiconazole, may provide a molecular environment consistent with G2/M cell cycle arrest, thereby allowing damaged cells to accumulate.

Complex Pathway Responses to Different Conazoles

In the present study, the largest number of altered pathways was associated with triadimefon and propiconazole exposure at 90 days. The cellular processes associated with each conazole were different and resulted in patterns that distinguished triadimefon from propiconazole. Overall, triadimefon exposure altered the molecular pathways associated with M-Ras regulation, cytoskeleton remodeling, and apoptosis by mitochondrial proteins, whereas, propiconazole uniquely altered transcription regulation of amino-acid metabolism, ubiquitin-metabolism, inflammation signaling via the adenosine receptor, and glutathione metabolism. These responses suggest that triadimefon exposure was associated with a cellular toxicity response characterized by cellular remodeling whereas propiconazole may impact protein homeostasis as evidenced by acute-phase responses of inflammation and regenerative amino-acid regulation.

Common Pathways

There were 71 pathways common to triadimefon and propiconazole after 90 days of exposure. These pathways could be grouped into 2 major areas of cell function: (1) cell cycle regulation, growth factors, and signaling oncogenes; and (2) inflammation and metabolism. Central to the first major area is an important family of genes coding for a group of Ras proteins. N-Ras andK-Ras pathways were among the commonly altered pathways after 90 days of treatment. These basic pathways are integrally involved in many complex signaling pathways involving growth factors, hormones, and cytokines acting both upstream and downstream from Ras. Ras proteins are required for signal transduction (Pellicer, 1998).

Extracellular signals that activate Ras proteins can act through different receptors such as, tyrosine kinase receptors (platelet derived growth factor receptor and epidermal growth factor receptor), cytokine receptors (IL-2, 4, 6, and 9 receptors), T-cell receptors, and subunits of heterotrimeric G proteins (Rodriguez-Viciana et al., 2004). ERK, IGF-RI, TGF-beta receptor, and PDGF signaling pathways were altered by both triadimefon and propiconazole. Both positive and negative regulators of the cyclin-dependent kinases (CDK), cyclins and CDK inhibitors, are targets of Ras, giving rise to different effects in the cell cycle progression and represent the complexity of the signal transduction networks and the diversity of pathways used by Ras to propagate molecular signals (Pawson, 1995).

The second major area of cellular function that was altered by both triadimefon and propiconazole was inflammation and metabolism. Specifically many of the interleukins (2, 4, 6, 9) and a chemokine receptor CXCR4 were among the common pathways altered by triadimefon and propiconazole. The shared metabolic pathways were those for tryptophan, urea, nucleotide and fatty-acid biosynthesis. A number of genes in the tryptophan biosynthesis pathway were over expressed. Many genes are common between fatty-acid metabolism and tryptophan metabolic pathways, for example acetyl CoA is in the tryptophan pathway and in addition, acetyl CoA is used for fatty acid β 0 xidation. Therefore, induction of genes involved in these processes was interpreted as shared functions in both amino acid and fatty acid metabolism.

Pathways Unique to Propiconazole

There were 37 pathways unique to propiconazole after 90 days of exposure and some of these pathways were relevant to the tumor promoting activity propiconazole in rat liver. These pathways were those involving inflammation mediation through A2BR adenosine signaling. Specifically, they include A2BR signaling though G-alpha-q. Adenosine is an endogenous signaling molecule that is frequently induced during inflammatory processes (Kolachala et al., 2005). Moreover, adenosine signaling via the A2B receptor that acts through the G-alpha-q pathway, which activates phospholipase C to produce diacylglycerol and IP3. Diacylglycerol in turn activates protein kinase C (PKC). PKC also serves as the receptor for phorbol esters, a class of tumor promoters (Nishizuka, 1995; Cenni et al., 2002). Phorbol ester tumor promoters activate PKC by a mechanism similar to that of diacylglycerol, providing evidence that PKC activation is a critical event in tumor promotion (O’Brian and Ward, 1989). In turn, PKC signals downstream to activate phosphatidylinositol-kinase 3 (PI-3K) which also transmits signals from receptor tyrosine kinases (Vivanco and Sawyers, 2002).

The insulin-like growth factor receptor (IGFR1) is a trans-membrane protein tyrosine kinase which mediates a cascade of biological effects upon binding its ligand IGF-1 (Blakesley et al., 1997). The IGF-1 receptor has been implicated in tumor growth and is known to increase PI-3K activity (Altschuler et al., 1994). Further, the insulin receptor-signaling pathway in our study was significantly altered by propiconazole but not by triadimefon.

A third group of propiconazole-altered pathways that could contribute to its tumor promotion potential are the MAP kinase and the G-protein regulation of MARK-ERK signaling pathway. Ornithine decarboxylase (ODC) was shown to be a necessary step for MAPK-induced mouse skin tumorigenesis (Feith et al., 2005). That result established that ODC was an important component of the Raf/MEK/ERK pathways. In the present study, the gene coding for ODC was only significantly ( p < 0.01) induced only by high-dose propiconazole at 90 days.

Pathways Unique to Triadimefon

In contrast to propiconazole at 90 days, triadimefon uniquely altered 34 pathways. Of these pathways one-third were involved in metabolic processes including amino acids, prostaglandin, fatty acids, and leukotrienes. The remaining two-thirds of the 34 pathways could be grouped into apoptotic, growth factors, cytoskeletal remodeling, or G-protein signaling. These observed pathways, when linked to the histological findings of marked hepatic hypertrophy associated with triadimefon exposure, are consistent with increased protein synthesis in response to an increase in growth-factor signaling and a concomitant increase in apoptosis.

In conclusion, using a transcriptional profiling pathway approach we were able to distinguish the hepatic effects of triadimefon from propiconazole and myclobutanil, a result not achieved by traditional toxicologic methods. This comprehensive approach, at the cellular and molecular level, combined with traditional toxicology studies has provided a clearer picture of the events occurring in the liver after conazole treatment. The approach described herein has enabled the differentiation of the actions of related chemicals based on uniquely altered molecular pathways in contrast to the similarity of responses observed using traditional toxicity methods. The companion study in this issue (Wolf et al., 2006) showed that altered metabolism in the liver is a common response to all the conazoles studied and was related to the development of thyroid hormone disruption. Those results suggested that thyroid tumors induced by triadimefon would likely develop by a mode of action that is not consistent with an excess of circulating TSH as a mitogenic stimulus, since TSH levels were not increased.

From our transcriptional analysis, we hypothesize that the repressed expression of the thyroid specific genes (TPO, TRH, TRHR, and TSHβ) observed only with triadimefon treatment may contribute to its thyroid tumorigenic mode of action. Hepatic transcriptional profiles identified several growth pathways associated with triadimefon exposure that could initiate hepatic growth signaling which could account for the observed hypertrophy and increased liver weight, significant for triadimefon after 90 days but not significant for propiconazole or myclobutanil. Thus, we hypothesize that dysregulation of cell cycle and metabolic growth processes may represent key events for hepatotoxicity in rats treated with triadimefon but not in propiconazole or myclobutanil groups.

In contrast, the transcriptional profiles associated with propiconazole are consistent with an acute stress response characterized by induction of many immediate inflammatory pathways and pathways associated with amino acid metabolism and protein turnover. Inflammation responses and damaged proteins could be a source of toxicity in and of itself or could certainly amplify any initial hepatic toxicity. Combining the evidence presented in Wolf et al. (2006) with the present systematic analysis of transcription after conazole exposure provides a differential response by triadimefon, associated with dysregulation of cell-regulating processes whereas propiconazole appeared to have a unique toxicity profile of acute inflammation and protein metabolism. Both of these profiles certainly could underlie the observed hepatic toxicity of hypertrophy and increased liver weight, and could serve as key events to characterize and distinguish the rat tumorigen, triadimefon, from nonthyroid tumorigen propiconazole.