Abstract
Hepatic enzyme inducers such as phenobarbital are often nongenotoxic rodent hepatocarcinogens. Currently, nongenotoxic hepatocarcinogens can only be definitively identified through costly and extensive long-term, repeat-dose studies (e.g., 2-year rodent carcinogenicity assays). Although liver tumors caused by these compounds are often not found to be relevant to human health, the mechanism(s) by which they cause carcinogenesis are not well understood. Toxicogenomic technologies represent a new approach to understanding the molecular bases of toxicological liabilities such as nongenotoxic carcinogenicity early in the drug discovery/development process. Microarrays have been used to identify mechanistic molecular markers of nongenotoxic rodent hepatocarcinogenesis in short-term, repeat-dose preclinical safety studies. However, the initial “noise” of early adaptive changes may confound mechanistic interpretation of transcription profiling data from short-term studies, and the molecular processes triggered by treatment with a xenobiotic agent are likely to change over the course of long-term treatment. Here, we describe the use of a differential display technology to understand the molecular mechanisms related to 13 weeks of dosing with the prototype rodent nongenotoxic hepatocarcinogen, phenobarbital. These findings implicate a continuing role for oxidative stress in nongenotoxic carcinogenicity.
An Excel data file containing raw data is available in full at http://taylorandfrancis.metapress.com/openurl.asp?genre=journal&issn=0192-6233. Click on the issue link for 33(1), then select this article. A download option appears at the bottom of this abstract. The file contains raw data for all gene changes detected by AFLP, including novel genes and genes of unknown function; sequences of detected genes; and animal body and liver weight ratios. In order to access the full article online, you must either have an individual subscription or a member subscription accessed through www.toxpath.org.
Keywords
Introduction
Phenobarbital is a rodent nongenotoxic hepatic carcinogen that has been safely and effectively used in humans as a sedative and antiepileptic drug for many years. Phenobarbital (PB) is a representative of a class of xenobiotics that typically induce hepatocellular and smooth endoplasmic reticulum proliferation as well as changes in the expression of several detoxification gene families (Waxman, 1999). Across species a similar enzyme induction and detoxification response to phenobarbital and phenobarbital-like compounds is seen (Gonzalez, 1988; Okey, 1990; Corocos and Lagadic-Gossmann, 2001). In present day drug development, a compound like PB would face many challenges prior to FDA approval due to the tumor-promoting effects in the rodent. Further elucidation of the mechanism of phenobarbital-mediated hepatocarcinogenesis could greatly benefit future studies involving nongenotoxic hepatic carcinogens.
Presently the only method for the identification of carcinogens is through long-term repeat-dose studies such as the 2-year rodent carcinogenicity assay. Such studies are required for the registration of new pharmaceutical agents intended for chronic or intermittent use over 6 months duration (IHC M5, 1998). Two-year rodent carcinogenicity studies require a significant expenditure of both capital and active pharmaceutical ingredient and use large numbers of animals. Efforts have been initiated toward the use of transgenic or knockout mouse lines to replace the standard 2-year mouse carcinogenicity studies with shorter duration studies (Storer et al., 2001; Van Kreijl et al., 2001; Usui et al., 2001), however these studies are at least 6 months in duration, and are only alternatives to the mouse 2-year bioassay. These alternative models do not afford relief from conducting a chronic bioassay in the rat. For these reasons, rodent carcinogenicity studies are often conducted late in the development process. A treatment-related tumor response in the 2-year rodent carcinogenicity bioassay, left unaddressed, could be a major setback to a drug development program, and may significantly increase registration time or even cause the discontinuation of a compound in some therapeutic areas.
A more thorough understanding of the mechanism(s) of nongenotoxic hepatic carcinogenesis would help to identify the molecular mechanism of nongenotoxic carcinogenicity. Although there are likely to be numerous mechanisms involved in nongenotoxic rodent hepatic carcinogenesis, 2 broad mechanistic hypotheses have been proposed. Specifically, a role for chronic oxidative stress has been postulated to result in the eventual overwhelming of the tissue’s ability to respond, ultimately resulting in tumorigenesis (Klaunig et al., 1995). Other researchers have suggested that nongenotoxic carcinogens act via a disruption in the balance between proliferation and apoptosis (Schulte-Hermann, 1979; Cohen and Ellwein, 1990). In the present study we used amplified fragment length polymorphism (AFLP) to evaluate changes in gene expression after 13 weeks of dosing with phenobarbital. The results confirm a continuing role for oxidative stress in nongenotoxic carcinogenicity.
Materials and Methods
In vivo Experiments
All experimental procedures were approved by the Institutional Animal Care and Use Committee and were performed in compliance with laws regarding humane treatment of laboratory animals. Two studies were performed, a 13-week “AFLP study,” and a 5-day “transcript profiling study.” For the 13-week study, phenobarbital was suspended in distilled water and dosed at concentrations of 150 mg/kg/day for the first week and 200 mg/kg/day for the remainder of the study. Prior to treatment, male CD:IGS rats (Charles River Laboratories, Wilmington, MA; n = 14), aged ~7 weeks, were acclimated for 1 week then randomly assigned to 2 dosage groups. Animals were housed individually in rooms set to maintain 72° ± 5°F (22° ± 3°C) and 40% humidity with a 12-hour light, 12-hour dark cycle. Animals were dosed once daily by oral gavage, were fed ad libitum during the course of the study, and were not fasted prior to sacrifice. Animals were weighed at the outset of the study, weekly thereafter, and immediately prior to sacrifice. Animals were anaesthetized using CO2-O2 and sacrificed by exsanguination. Livers were collected from each animal, weighed, and sections of each were frozen in liquid nitrogen for analysis by AFLP. Portions of the livers were also sampled for histology (1 slice of the right medial lobe and 1 slice of the left lateral lobe), fixed in 10% neutral buffered formalin for 24 hours, and subsequently processed and embedded in paraffin.
The conduct of the 5-day study has been described previously (Kramer et al., 2004), and utilized phenobarbital dosed at 200 mg/kg/day. Appropriate amounts of test article were suspended in 0.5% methylcellulose (w/v) plus 0.1% polysorbate 80 (v/v) in distilled water. Prior to treatment, male Charles River CD:IGS rats, aged 6–7 weeks, were acclimated for 1 week then randomly assigned to dosage groups. Animals were housed individually and fed ad libitum in rooms set to maintain 72° ± 5°F (22° ± 3°C) and 40% humidity with a 12-hour light, 12-hour dark cycle. Compound was administered once daily for 5 days by oral gavage at 10 mL/kg. Animals were weighed prior to day 1, on day 3, and on the day of necropsy (day 5). Livers were collected from each animal and weighed, and sections were frozen in liquid nitrogen for transcription profiling.
Amplified Fragment Length Polymorphism
The AFLP technique involves 5 basic steps: creation of cDNA from mRNA; restriction endonuclease digestion of the DNA and ligation of adapters; selective amplification of sets of restriction fragments; analysis of the amplified fragments via gel electrophoresis; and sequencing. RNA stat reagents and protocol were used to isolate total RNA from frozen liver sections as specified by Leedo Medical Laboratories (Houston, TX). RNA was shipped on dry ice to Keygene/PE Genscope where reverse transcription, gel electrophoresis, and sequence analysis were conducted (Rockville, MD). All AFLP typing was conducted at Keygene/PE Genscope, following the protocol previously described (Zabeau and Vos, 1993). Briefly, complementary DNA was digested using 4 restriction endonucleases, BstYI, ApoI, MseI, and BfaI. The first 2 enzymes recognize pentanucleotide consensus sequences, the latter 2 recognize a 4-nucleotide motif. Each combination of a penta and tetra cutter provides reasonable coverage (50% or more) of the transcripts and yields fragments in the size range of 10 to 500 base pairs.
Analysis of short phosphoimage exposure was used to measure more abundant fragments, while the scarce transcripts were identified from exposures 4 to 5 times longer. Proprietary software developed by Keygene/PE Genscope was used to find and trace bands and generate band tables containing signal intensity for each AFLP profile. Further analysis was performed at Pfizer using Excel (Microsoft) and GeneSpring Software (Silicon Genetics, Palo Alto, CA). Band intensities were normalized against the total lane intensity, correcting for sample variation. Genes induced or repressed on average by 2-fold or more were placed into general functional categories.
Real-Time Quantitative RT-PCR
Eight genes detected via the AFLP technique were further analyzed using quantitative real-time polymerase chain reaction (RT-PCR) on an ABI Prism 7900 Sequence Detector (Applied Biosystems, Foster City, CA). Reverse transcription was primed with oligo(dT)12–18, and used 100 ng total RNA and RT-PCR reagents (Applied Biosystems), following the manufacturer’s instructions. After reverse transcription, 2.5 uL of cDNA was employed in each reaction. The concentration of amplification primers and reporter were 600 nM and 200 nM, respectively. Probes were labeled with 6-carboxyfluorescein (FAM) on the 5′ nucleotide and a nonfluorescent quencher on the 3′ nucleotide. A passive reference dye, 6-carboxy-X-rhodamine, provided an internal standard for normalization of FAM fluorescence, correcting for fluctuations due to volume changes. Primer and probe sequences are listed in Table 1. Abundance of each gene was determined relative to a standard transcript, cyclophillin. Each cDNA was assayed in duplicate PCR reactions. Average Ct values from duplicate PCR reactions were normalized to average Ct values for cyclophilin from the same cDNA preparations. The ratio of expression of each gene in drug-treated vs. vehicle-treated samples was calculated as 2−(mean ΔΔ Ct ) of that treatment as recommended by ABI.
Transcript Profiling
Transcript profiling on total liver RNA from the five day study has been previously described (Kramer et al., 2004). In brief, total RNA was prepared from frozen livers using RNA-Stat reagent and protocol (Leedo Medical Laboratories, Houston, TX). Poly(A)+ RNA was isolated using micro poly(A)plus kits (Ambion, Austin, TX). Aliquots of 300 ng of mRNA were used to generate end-labeled complementary DNA using Incyte GEM Brite probe labeling kits (Incyte Pharmaceuticals, Palo Alto, CA). Pooled mRNAs from vehicle-treated rats were used to prepare Cy3 probes, whereas individual and pooled mRNA from animals dosed with 200 mg/kg/day phenobarbital were used to prepare Cy5 probes. Probes were separated from unincorporated primer using Chromaspin TE-30 columns (Clontech, Palo Alto, CA). Microarray hybridization was performed in duplicate at Incyte Pharmaceuticals using RatGEM 1.4, containing ~7,800 cloned rat cDNAs, as previously described (Schena et al., 1996). Hybridization data were balanced using total average signal intensity to generate a balance coefficient. Balanced differential expression (BDE) values were calculated using the equation:
where P1 represents the signal intensity of the Cy3 labeled control sample and bP2 represents the balanced signal intensity of the Cy5 labeled treated sample. A BDE value was generated for all elements with signal intensity P1 + P2 ≥ 500 (corrected for global background), signal to background values P1STB + P2STB ≥ 10, and area of coverage greater than 40%. All other elements were recorded as absent values and were omitted from further consideration. Statistical analysis was performed using GeneSpring and Excel.
Results
In vivo Study Results
After 1 week of once daily oral gavage dosing at 150 mg/kg/day, attenuation of the sedative effects of phenobarbital was observed, likely due to induction of hepatic metabolism enzymes. Daily dosing was increased to 200 mg/kg/day, and maintained at this level throughout the remainder of the study. Three animals had to be sacrificed early in the 13-week study due to moribundity, and 1 high dose (200 mg/kg/day) phenobarbital treated animal died on day 2 in the 5-day study. Significant increases in absolute liver weight and liver/body weight ratios were observed in both studies. At 13 weeks, liver weights were increased by 65% (p < 0.0001), and liver to total body weight ratios were increased by 88% (p < 0.0001). In the 5-day study, liver weight and liver to total body weight were increased by 34% (p = 0.0017), and 42% (p = 0.0013), respectively. Histological examination revealed diffuse centrilobular hepatocellular hypertrophy and hepatomegaly.
Expression Changes Identified by Amplified Fragment Length Polymorphism
Amplified Fragment Length Polymorphism analysis indicates that treatment with phenobarbital dramatically altered genetic expression out to 13 weeks of treatment (Tables 2 and 3). Genes with altered transcriptional activity of greater than 2-fold as compared to control following phenobarbital treatment were categorized into 8 classes according the major functions as follows: biotransformation; cell cycle and growth regulation and apoptosis; signal transduction; cellular metabolism; membrane transport and clearance; stress response; cytoskeletal and structural; and novel sequences and genes of unknown function (Figure 1, Table 2). A total of 319 restriction fragments comprising 168 contigs demonstrating altered expression patterns were identified. Of the 168 sequence contigs, 53 encoded genes (31.5% of total) of unknown function, and 32 (19% of total) represented novel sequences. A novel sequence was defined as a sequence for which no resulting match could be identified when the sequence was BLASTed against public and private sequence databases. For a complete list of AFLP gene changes, see http://taylorandfrancis.metapress.com/openurl.asp?genre=journal&issn=0192-6233. Of the remaining 83 genes with known functions, 69 were up-regulated and 14 genes were down-regulated.
Confirmation of AFLP Data by RT-PCR
To confirm the AFLP data, relative transcript levels of selected genes were measured by RT-PCR using RNA isolated from the same liver samples used in the AFLP study. Eight genes demonstrating varying levels of regulation by AFLP were chosen for RT-PCR analysis. Seven of the eight genes assessed were confirmed to have altered transcript levels by RT-PCR (Figure 2). This indicated strong positive correlations (r = 0.78) between results obtained with the 2 techniques. Differences in sensitivity and/or specificity could be attributed to the minor discrepancies between the 2 techniques.
Expression Changes Identified by Transcription Profiling
Selected results from transcription profiling on the 5-day study have been reported previously (Kramer et al., 2004), and data has been deposited in the European Bioinformatics Institute MIAME-Express database (accession # E-MEXP-89). Among the 68 genes that were regulated 2-fold or greater relative to control by treatment with 200 mg/kg/day phenobarbital for 5 days, 34 were induced and 34 repressed (Table 4). Of the 29 biotransformation genes that were regulated at 5 days, 26 were induced. Similarly, 3 of the 4 stress response pathway genes identified were induced. In contrast, nearly all of the cellular metabolism, cell cycle and growth regulation and apoptosis, and membrane transport and clearance genes identified were repressed.
Discussion
In the present study AFLP was used to evaluate changes in gene expression after 13 weeks of dosing with phenobarbital, a prototypical rodent nongenotoxic hepatic carcinogen. A 13-week study is typically run as a range finder for the longer-term rodent carcinogenicity assay. Although there is some debate in the literature regarding the utility of the lifetime bioassay (Boorman et al., 1994; Cohen, 1995; Alden et al., 2002), rodent carcinogenicity assays must be performed prior to the registration of many new pharmaceutical agents intended for chronic or intermittent use over 6 months’ duration. The present study was undertaken to provide insight into the mechanism(s) of nongenotoxic carcinogenicity resulting from chronic PB treatment. A number of general cellular targets, responses and outcomes can be proposed for nongenotoxic hepatocellular carcinogenesis (Figure 3). Among these, increased oxidative stress due to chronic induction of metabolism and biotransformation genes appears to be pertinent.
Comparison of AFLP and RT-PCR Results
The effects of phenobarbital on the expression of several sequences were verified via RT-PCR. Of the 8 genes analyzed, all but 1 gene demonstrated the same trend and many showed the same degree of regulation (Figure 2). Disagreement between AFLP and RT-PCR may be explained by one of a number of possibilities. The 2 techniques may be evaluating different regions of the gene and/or alternative splice variants. Alternatively, the AFLP sequence could be a close match to that of the known sequence identified, but is actually a different sequence, resulting in improper primer and probe design. Finally, minor changes in the expression of some genes may have been exaggerated during the amplification process by either methodology. However, good agreement was demonstrated overall between AFLP and RT-PCR analysis.
Comparison of 13-week AFLP with 5-Day Microarray Results
Agreement in the identity and regulation of genes affected at 5 days and 13 weeks varied based upon their classification. Results from these 2 methodologies might not be expected to compare very well. It is possible that the use of an “open” technology (AFLP) in the 13-week study compared to a “closed” technology (cDNA microarrays) at 5 days could have resulted in the differences in the genes identified in these 2 studies. For financial reasons, not all differentially expressed bands were sequenced in the AFLP experiment. In general, only the most highly expressed and obviously changed bands were chosen for sequence analysis. Similarly, only those genes present on the RatGEM cDNA array could be measured in the 5-day experiment. Despite these obstacles, there was generally good agreement in the nature of the genes identified between AFLP and transcription profiling. In particular, “biotransformation” and “stress response” pathway genes were mostly all induced at both time points. In contrast, whereas most of the “cellular metabolism”; “cell cycle and growth regulation and apoptosis”; and “membrane transport and clearance” pathway genes were repressed at 5 days, members of these categories were mostly induced at 13 weeks. Within the biotransformation category, a large number of the genes identified as changed were identical between the two methodologies at the time points assessed. Although this may reflect better coverage on the microarray of genes in this category, it demonstrates a persistent effect on a specific subset of biotransformation genes. The persistant induction of many of these genes supports a role for oxidative stress in the mechanism(s) of rodent nongenotoxic hepatocarcinogenesis.
Detoxification Response
In the oxidative stress hypothesis, the generation of reactive oxygen species (ROS) occurs endogenously as a result of cellular metabolism and oxidative phosphorylation (Sies, 1991). Exogenous xenobiotic chemicals may contribute to overall ROS levels by inducing cytochromes P450 (Halliwell, 1996). The P450 enzymes generate oxygen free radicals in the process of metabolizing xenobiotic chemicals (Parke and Ioannides, 1990), including phenobarbital. Kraupp-Grasl et al. (1991) provided evidence for the oxidative stress hypothesis. After treating young rats, aged 13 weeks, and old rats, aged 57 weeks, with phenobarbital or the peroxisome proliferator nafenopin for 13 months, old animals produced numerous hepatocellular adenomas and carcinomas relative to age matched control animals, while young animals produced very few lesions relative to control. As animals age, their ability to repair ROS-induced cellular damage is compromised, supporting the hypothesis that oxidative stress is a key mediator of phenobarbital-mediated promotion and tumorgenesis. Imaoka et al. (2004) also suggest oxidative processes mediated by PB, as seen by induction of hydroxyl radicals produced by P450s, both in vitro and in vivo. The ROS produced by P450 oxidized genomic DNA and increased oxidative stress, possibly contributing to tumor initiation and promotion. In the present study, the expression of numerous cytochrome P450 monooxygenases remains elevated out to 13 weeks. Additionally, induction of flavin containing monooxygenase-5 and epoxide-containing metabolic intermediates, suggested by the induction of epoxide hydrolase, may also produce reactive oxygen intermediates. Aldehyde dehydrogenase 1A1, which has also been shown to be induced in short-term studies by both genotoxic and nongenotoxic carcinogens (Pappas et al., 1998; Ellinger-Ziegelbauer et al., 2004; Kramer et al., 2004), also remained highly induced at 13 weeks.
Drug Conjugation and Clearance Response
Among the more pronounced effects observed at 13 weeks was the continued induction of genes encoding phase-II metabolism and clearance enzymes. This included 5 UDP-glucuronosyl transferase (UGT) isoforms from the 1A, 1B, 2A, and 2B families (Table 3). Four members of the ATP-binding cassette family of transporters were elevated at 13 weeks, including multiple drug resistance protein (also called P-glycoprotein), multidrug resistance protein 3, and the canalicular multispecific organic anion transporter. Genes encoding five different glutathione-S-transferase (GST) subunits remained induced relative to control at 13 weeks. Many of these genes were also elevated after 5 days of dosing in the microarray data (Table 4). The induction of GSTs may reflect general cellular stress due to an increase in electrophilic and reactive intermediates, lending support to the oxidative stress hypothesis. Among the xenobiotic conjugation genes, several members of the sulfotransferase family were repressed both at 5 days and 13 weeks.
Stress Response
As well as the GSTs, several other general and oxidative stress response-related genes were induced relative to control (Table 3). Among these, the gene encoding α-crystallin, a protein whose expression correlates with high oxidative activity (Iwaki et al., 1990), was induced greater than 10-fold. Genes encoding 4 members of the heat shock protein family were induced between 3- and 4-fold. Finally, a rat gene similar to the mouse growth arrest and DNA-damage-inducible (GADD)-45β was elevated approximately 3.5-fold relative to vehicle treated rats at 13 weeks. The persistant induction of genes encoding enzymes involved in stress responses supports a role for chronic low levels of oxidative stress in the mechanism of phenobarbital-mediated rodent hepatocellular carcinogenesis.
Proliferative Response
The proliferation hypothesis implicates a compound-mediated imbalance between hepatocellular proliferation and apoptosis. Many rodent hepatic carcinogens that are negative in genotoxicity assays have been shown to stimulate adaptive liver growth and hepatocellular proliferation (Schulte-Hermann, 1979), and several investigators have suggested that induction of cell proliferation may be a primary cause of nongenotoxic carcinogenicity (Ames and Gold, 1990; Cohen and Ellwein, 1991). A role for organ specific DNA synthesis and cell proliferation has been suggested to explain organ-specific and species-specific carcinogenesis (Kolaja et al., 1996). This hypothesis suggests that nongenotoxic hepatic carcinogens may affect this balance by increasing hepatocellular proliferation via direct mitogenic activity, or through a regenerative response to cytotoxic or apoptotic effects (Dragan et al., 2001). It has even been suggested that dose-response relationships observed with some genotoxic carcinogens may be due in part to effects on cell proliferation (Cohen and Ellwein, 1990). Similarly, a number of nongenotoxic carcinogens have been demonstrated to stimulate DNA synthesis (Roberts et al., 1995). In contrast, other investigators have challenged the relationship between nongenotoxic carcinogenicity and proliferation (Melnick, 1992; Melnick and Huff, 1993). In fact, proliferation and DNA synthesis stimulated by many nongenotoxic carcinogens peaks by 1–4 weeks and returns to baseline by about 8 weeks (Marsman et al., 1992; Kolaja et al., 1996). Schulte-Hermann et al. (1999) have suggested that chemically mediated increases in cell proliferation and/or cell survival may not necessarily lead to cancer, though these effects may indicate carcinogenic potential.
Among the genes that were regulated after 13 weeks of treatment with phenobarbital was transforming growth factor-β stimulated clone (TSC)-22. We have previously reported on regulation of TSC-22 in short-term studies as a possible early in vivo marker of hepatic carcinogenesis (Kramer et al., 2004). The gene-product is a leucine zipper transcription factor (Kester et al., 1999), induced by TGF-β, phorbol-12-myristate 13-acetate, and glucocorticoid receptor agonists (Shibanuma et al., 1992). TSC-22 is highly similar to the Drosophila gene bunched (Dobens et al., 1997), which acts downstream of the fruit fly homologue of transforming growth factor-β, where it exerts an inhibitory role in transcription mediated by this signaling cascade (Dobens et al., 2000). Ornithine aminotransferase, a marker of mitochondrial damage and de-differentiation was also among the genes repressed by PB at 13 weeks. Ellinger-Ziegelbauer et al. (2004) have shown this gene to be repressed by short-term treatment with a number of genotoxic carcinogens. Finally, senescence marker protein 2B, the expression of which is induced during senescence, demonstrated remarkable induction relative to the age-matched vehicle treated control animals.
Conclusions
In the present study, biotransformation and oxidative stress response genes remain induced at 13 weeks, whereas a number of cell cycle and growth regulation, and apoptosis genes are also induced. The fact that many novel genes and genes of unknown function are affected may confound mechanistic interpretation. The involvement of gene expression changes not related to toxicity or carcinogensis (i.e., adaptive responses, drug clearance, etc.) must also be taken into consideration, and lends further complexity to transcription profiling data interpretation. Neither the oxidative stress nor the proliferation hypotheses can explain all the mechanisms of nongenotoxic hepatic carcinogenesis. Reports from Klaunig et al. (1998) and Rusyn et al. (2000) have suggested that these pathways need not be considered mutually exclusive. Specifically, during the promotion stage, ROS and oxidative stress can contribute to abnormal gene expression, blockage of cell-to-cell communication, and modification of second messenger systems resulting in increased cell proliferation and decreased apoptosis (Klaunig et al., 1998). Similarly, Rusyn et al. (2000) propose that both mechanistic hypotheses may work conjointly, since NF-κB-mediated hepatocellular proliferation and DNA synthesis may be activated in response to reactive oxygen species (Nilakantan et al., 1998; Rusyn et al., 1998). Our results support a continuing role for metabolic oxidative stress in the mechanism of rodent nongenotoxic hepatocarcinogenesis.
Butterworth et al. (1995) have suggested a strategy for establishing a mode of action to enable risk assessment for chemical carcinogens. Every pharmaceutical company employs a strategy to identify and screen out genotoxic compounds very early in the discovery/development paradigm, however there are as yet no early assays to assess nongenotoxic carcinogenicity. We have suggested that measuring molecular markers in early preclinical safety studies may detect nongenotoxic carcinogenicity, and could be used to trigger more predictive assays both to identify and understand nongenotoxic carcinogens, as well as to assess the risks of continuing their development. Although the exact mechanistic role of some of the genes identified in this study is uncertain, many are involved in common responses to interactions with potential cellular targets for xenobiotics and ROS (Figure 3). These genes and the mechanisms in which they are involved may serve as a foundation for risk assessment prior to running a carcinogenicity assay and for triggering more thorough mechanistic analyses of potential carcinogenicity.