Release of (and lessons learned from mining) a pioneering large toxicogenomics database.

AIM: We release the Janssen Toxicogenomics database. This rat liver gene-expression database was generated using Codelink microarrays, and has been used over the past years within Janssen to derive signatures for multiple end points and to classify proprietary compounds. MATERIALS & METHODS: The release consists of gene-expression responses to 124 compounds, selected to give a broad coverage of liver-active compounds. A selection of the compounds were also analyzed on Affymetrix microarrays. RESULTS: The release includes results of an in-house reannotation pipeline to Entrez gene annotations, to classify probes into different confidence classes. High confidence unambiguously annotated probes were used to create gene-level data which served as starting point for cross-platform comparisons. Connectivity map-based similarity methods show excellent agreement between Codelink and Affymetrix runs of the same samples. We also compared our dataset with the Japanese Toxicogenomics Project and observed reasonable agreement, especially for compounds with stronger gene signatures. We describe an R-package containing the gene-level data and show how it can be used for expression-based similarity searches. CONCLUSION: Comparing the same biological samples run on the Affymetrix and the Codelink platform, good correspondence is observed using connectivity mapping approaches. As expected, this correspondence is smaller when the data are compared with an independent dataset such as TG-GATE. We hope that this collection of gene-expression profiles will be incorporated in toxicogenomics pipelines of users.

Oxidative stress/reactive metabolite gene expression signature in rat liver detects idiosyncratic hepatotoxicants.

Previously we reported a gene expression signature in rat liver for detecting a specific type of oxidative stress (OS) related to reactive metabolites (RM). High doses of the drugs disulfiram, ethinyl estradiol and nimesulide were used with another dozen paradigm OS/RM compounds, and three other drugs flutamide, phenacetin and sulindac were identified by this signature. In a second study, antiepileptic drugs were compared for covalent binding and their effects on OS/RM; felbamate, carbamazepine, and phenobarbital produced robust OS/RM gene expression. In the present study, liver RNA samples from drug-treated rats from more recent experiments were examined for statistical fit to the OS/RM signature. Of all 97 drugs examined, in addition to the nine drugs noted above, 19 more were identified as OS/RM-producing compounds-chlorpromazine, clozapine, cyproterone acetate, dantrolene, dipyridamole, glibenclamide, isoniazid, ketoconazole, methapyrilene, naltrexone, nifedipine, sulfamethoxazole, tamoxifen, coumarin, ritonavir, amitriptyline, valproic acid, enalapril, and chloramphenicol. Importantly, all of the OS/RM drugs listed above have been linked to idiosyncratic hepatotoxicity, excepting chloramphenicol, which does not have a package label for hepatotoxicity, but does have a black box warning for idiosyncratic bone marrow suppression. Most of these drugs are not acutely toxic in the rat. The OS/RM signature should be useful to avoid idiosyncratic hepatotoxicity of drug candidates.

Erratum to "Prediction of non-genotoxic carcinogenesis in rats using changes in gene expression following acute dosing".

Carcinogenicity of chemicals can currently only be evaluated in 2-year rodent bioassays. Therefore, the development of early biomarkers for carcinogenesis would result in substantial savings in time and expense. The current study investigates whether early changes in gene expression may be developed as markers for cancer. Animals were treated for 1 or 5 days with either non-genotoxic carcinogens or non-carcinogens and gene expression was analyzed by quantitative PCR (qPCR).We tested two gene signatures previously reported to detect non-genotoxic carcinogens. Using one gene signature it was confirmed that 3/3 nongenotoxic carcinogens and 2/2 non-carcinogens are correctly identified with data from 1 or 5 days of dosing. In contrast an alternative signature correctly identified 0/3 and 2/3 nongenotoxic carcinogens at 1 and 5 days of treatment, respectively and 2/2 non-carcinogens at both time-points. Additionally, we evaluated a novel panel of putative biomarker genes, from the literature, many of which have roles in cell growth and division, including myc, cdc2 and mcm6. These genes were significantly induced by non-genotoxic carcinogens and not by non-carcinogens. Using the average fold-induction across this panel, 2/3 non-genotoxic carcinogens were detected on both day 1 and day 5. These data support the idea that acute changes in gene expression may provide biomarkers for non-genotoxic carcinogenesis but also highlight interesting differences in the sensitivities of distinct gene signatures.

Interlaboratory evaluation of genomic signatures for predicting carcinogenicity in the rat.

The Critical Path Institute recently established the Predictive Safety Testing Consortium, a collaboration between several companies and the U.S. Food and Drug Administration, aimed at evaluating and qualifying biomarkers for a variety of toxicological endpoints. The Carcinogenicity Working Group of the Predictive Safety Testing Consortium has concentrated on sharing data to test the predictivity of two published hepatic gene expression signatures, including the signature by Fielden et al. (2007, Toxicol. Sci. 99, 90-100) for predicting nongenotoxic hepatocarcinogens, and the signature by Nie et al. (2006, Mol. Carcinog. 45, 914-933) for predicting nongenotoxic carcinogens. Although not a rigorous prospective validation exercise, the consortium approach created an opportunity to perform a meta-analysis to evaluate microarray data from short-term rat studies on over 150 compounds. Despite significant differences in study designs and microarray platforms between laboratories, the signatures proved to be relatively robust and more accurate than expected by chance. The accuracy of the Fielden et al. signature was between 63 and 69%, whereas the accuracy of the Nie et al. signature was between 55 and 64%. As expected, the predictivity was reduced relative to internal validation estimates reported under identical test conditions. Although the signatures were not deemed suitable for use in regulatory decision making, they were deemed worthwhile in the early assessment of drugs to aid decision making in drug development. These results have prompted additional efforts to rederive and evaluate a QPCR-based signature using these samples. When combined with a standardized test procedure and prospective interlaboratory validation, the accuracy and potential utility in preclinical applications can be ascertained.

Characterization of genetic alteration patterns in human esophageal squamous cell carcinoma using selected microsatellite markers spanning multiple loci.

In order to identify representative genetic alterations in esophageal squamous cell carcinomas (ESCC) and useful markers for future early detection, 34 ESCC samples with neighboring normal epithelia and 30 esophageal biopsy samples from Linzhou, P.R. China, were studied. Of the 38 microsatellite markers selected, half were linked with tumor suppressors. More than 40% of the tumor samples showed loss of heterozygosity (LOH) in at least one of the eight markers, D3S1067 and D3S1561 (both linked to hMLH1 locus), FABP2, D4S1613, D9S171 (p14ARF, p15INK4b, p16INK4a loci), Rb1 (intron), p53-2 (intron), and NM23-H1. Most of the 38 microsatellite markers did not display microsatellite instability (MSI) in more than 30% of the tumor samples, except D9S942 (p14ARF, p15INK4b, p16INK4a loci) and Bat26, which showed frequency at 32 and 41%, respectively. Of all the ESCC samples examined, 20 samples exhibited LOH in 25% or more of the informative markers. Three samples displayed MSI in more than 30% of the markers, indicating that MSI might be an important event in these subset ESCC cases. Statistically significant correlations were found between LOH of the hMLH1 locus and the general LOH status of the sample, and between the LOH of the hMLH1 locus and p53 mutations. In addition, correlation was found between MSI in D3S1067/D3S1561 and the general MSI status in the samples. However, MSI in the introns of hMLH1 and hMSH2 were not correlated with the general MSI status of the tumors. LOH analysis was also performed in 30 esophageal biopsy samples containing precancerous lesions with matching blood samples using nine microsatellite markers selected from the above studies. LOH frequence ranged from 0 to 33% in informative cases, mostly in the 9p21 and p53 gene regions, suggesting these regions are possible targets of genomic instability in early stage ESCC carcinogenesis. The results demonstrate the degree of genetic alterations at different loci of the chromosomes. Some of the microsatellite markers may be useful for the early detection of ESCC.

Evaluation of felbamate and other antiepileptic drug toxicity potential based on hepatic protein covalent binding and gene expression.

Felbamate is an antiepileptic drug that is associated with minimal toxicity in preclinical species such as rat and dog but has an unacceptable incidence of serious idiosyncratic reactions in man. Idiosyncratic reactions account for over half of toxicity-related drug failures in the marketplace, and improving the preclinical detection of idiosyncratic toxicities is thus of paramount importance to the pharmaceutical industry. The formation of reactive metabolites is common among most drugs associated with idiosyncratic drug reactions and may cause deleterious effects through covalent binding and/or oxidative stress. In the present study, felbamate was compared to several other antiepileptic drugs (valproic acid, carbamazepine, phenobarbital, and phenytoin), using covalent binding of radiolabeled drugs and hepatic gene expression responses to evaluate oxidative stress/reactive metabolite potential. Despite causing only very mild effects on covalent binding parameters, felbamate produced robust effects on a previously established oxidative stress/reactive metabolite gene expression signature. The other antiepileptic drugs and acetaminophen are known hepatotoxicants at high doses in the rat, and all increased covalent binding to liver proteins in vivo and/or to liver microsomes from human and rat. With the exception of acetaminophen, valproic acid exhibited the highest covalent binding in vivo, whereas carbamazepine exhibited the highest levels in vitro. Pronounced effects on oxidative stress/reactive metabolite-responsive gene expression were observed after carbamazepine, phenobarbital, and phenytoin administration. Valproic acid had only minor effects on the oxidative stress/reactive metabolite indicator genes. The relative ease of detection of felbamate based on gene expression results in rat liver as having potential oxidative stressor/reactive metabolites indicates that this approach may be useful in screening for potential idiosyncratic toxicity. Together, measurements of gene expression along with covalent binding should improve the safety assessment of candidate drugs.

Hepatic expression of heme oxygenase-1 and antioxidant response element-mediated genes following administration of ethinyl estradiol to rats.

Heme oxygenase-1 (HO-1) is one of several enzymes induced by hepatotoxicants, and is thought to have an important protective role against cellular stress during liver inflammation and injury. The objective of the present study was to evaluate the role of HO-1 in estradiol-induced liver injury. A single dose of ethinyl estradiol (500 mg/kg, po) resulted in mild liver injury. Repeated administration of ethinyl estradiol (500 mg/kg/day for 4 days, po) resulted in no detectable liver injury or dysfunction. Using RT-PCR analysis, we demonstrate that HO-1 gene expression in whole liver tissue is elevated (>20-fold) after the single dose of ethinyl estradiol. The number and intensity of HO-1 immunoreactive macrophages were increased after the single dose of ethinyl estradiol. HO-1 expression was undetectable in hepatic parenchymal cells from rats receiving Methocel control or a single dose of ethinyl estradiol, however cytosolic HO-1 immunoreactivity in these cells after repeated dosing of ethinyl estradiol was pronounced. The increases in HO-1 mRNA and HO-1 immunoreactivity following administration of a single dose of ethinyl estradiol suggested that this enzyme might be responsible for the observed protection of the liver during repeated dosing. To investigate the effect of HO-1 expression on ethinyl estradiol-induced hepatotoxicity, rats were pretreated with hemin (50 micromol/kg, ip, a substrate and inducer of HO-1), with tin protoporphyrin IX (60 micromol/kg, ip, an HO-1 inhibitor), or with gadolinium chloride (10 mg/kg, iv, an inhibitor/toxin of Kupffer cells) 24 h before ethinyl estradiol treatment. Pretreatment with modulators of HO-1 expression and activity had generally minimal effects on ethinyl estradiol-induced liver injury. These data suggest that HO-1 plays a limited role in antioxidant defense against ethinyl estradiol-induced oxidative stress and hepatotoxicity, and suggests that other coordinately induced enzymes are responsible for protection observed with repeated administration of high doses of this compound.

Predictive toxicogenomics approaches reveal underlying molecular mechanisms of nongenotoxic carcinogenicity.

Toxicogenomics technology defines toxicity gene expression signatures for early predictions and hypotheses generation for mechanistic studies, which are important approaches for evaluating toxicity of drug candidate compounds. A large gene expression database built using cDNA microarrays and liver samples treated with over one hundred paradigm compounds was mined to determine gene expression signatures for nongenotoxic carcinogens (NGTCs). Data were obtained from male rats treated for 24 h. Training/testing sets of 24 NGTCs and 28 noncarcinogens were used to select genes. A semiexhaustive, nonredundant gene selection algorithm yielded six genes (nuclear transport factor 2, NUTF2; progesterone receptor membrane component 1, Pgrmc1; liver uridine diphosphate glucuronyltransferase, phenobarbital-inducible form, UDPGTr2; metallothionein 1A, MT1A; suppressor of lin-12 homolog, Sel1h; and methionine adenosyltransferase 1, alpha, Mat1a), which identified NGTCs with 88.5% prediction accuracy estimated by cross-validation. This six genes signature set also predicted NGTCs with 84% accuracy when samples were hybridized to commercially available CodeLink oligo-based microarrays. To unveil molecular mechanisms of nongenotoxic carcinogenesis, 125 differentially expressed genes (P<0.01) were selected by Student's t-test. These genes appear biologically relevant, of 71 well-annotated genes from these 125 genes, 62 were overrepresented in five biochemical pathway networks (most linked to cancer), and all of these networks were linked by one gene, c-myc. Gene expression profiling at early time points accurately predicts NGTC potential of compounds, and the same data can be mined effectively for other toxicity signatures. Predictive genes confirm prior work and suggest pathways critical for early stages of carcinogenesis.

Application of genomics in preclinical drug safety evaluation.

Understanding the response of biological systems to xenobiotics is fundamental to the evaluation of drug safety. Toxicologists have traditionally gathered pathological, morphological, chemical and biochemical information from in vivo studies of preclinical species in order to assess drug safety and to determine how new drugs can be safely administered to the human patient population. In recent years the emerging "-omics" technologies have been developed and integrated into preclinical studies in order to better assess drug safety by gaining information on the cellular and molecular events underlying adverse drug reactions. Genomics approaches in particular have become readily available and are being applied in several stages of drug development. The burgeoning literature on what has become known as "toxicogenomics" has for the most part highlighted successful applications of gene expression profiling in predictive toxicology, enabling decisions to be made on the developability of a compound early in the drug development process. It is also becoming apparent that toxicogenomic approaches are good starting points to develop experiments designed to gain a mechanistic insight into drug toxicities within and across species. Gene expression arrays permit the measurement of responses of essentially all the genes in the entire genome to be monitored, and knowledge of the function of the genes affected can identify the potential mechanisms to then be confirmed using conventional biochemical, toxicological and pathological approaches. As toxicologists put these technologies into practice they build up a knowledge base to better characterize toxicities at the molecular level and to make the search for much needed, novel biomarkers of toxicity more achievable.

The evolution of gene expression studies in drug safety assessment.

Since the identification in the 1950s of deoxyribonucleic acid as the building block of life, the impact of molecular biology has been far-reaching. Understanding the processes of how DNA is replicated, transcribed into RNA and then translated into protein products has not only provided a fundamental knowledge of life but has also spawned a plethora of applications. Molecular biology has been high profile and widespread in research into the biology of disease and in drug discovery. It has additionally found application in understanding the adverse effects, or toxicity, of candidate drugs and how they interfere with biochemical and biological processes. In recent times the biggest impact of molecular biology in toxicology has been through the study of differential gene expression, largely as a result of the advent of genomics. This review seeks to describe how toxicogenomics strategies have been implemented and integrated into nonclinical studies of drug safety.

Drug-induced oxidative stress in rat liver from a toxicogenomics perspective.

Macrophage activators (MA), peroxisome proliferators (PP), and oxidative stressors/reactive metabolites (OS/RM) all produce oxidative stress and hepatotoxicity in rats. However, these three classes of hepatotoxicants give three distinct gene transcriptional profiles on cDNA microarrays, an indication that rat hepatocytes respond/adapt quite differently to these three classes of oxidative stressors. The differential gene responses largely reflect differential activation of transcription factors: MA activate Stat-3 and NFkB, PP activate PPARa, and OS/RM activate Nrf2. We have used gene signature profiles for each of these three classes of hepatotoxicants to categorize over 100 paradigm (and 50+ in-house proprietary) compounds as to their oxidative stress potential in rat liver. In addition to a role for microarrays in predictive toxicology, analyses of small subsets of these signature profiles, genes within a specific pathway, or even single genes often provide important insights into possible mechanisms involved in the toxicities of these compounds.

Class prediction in toxicogenomics.

The intent of this article is to discuss some of the complexities of toxicogenomics data and the statistical design and analysis issues that arise in the course of conducting a toxicogenomics study. We also describe a procedure for classifying compounds into various hepatotoxicity classes based on gene expression data. The methodology involves first classifying a compound as toxic or nontoxic and subsequently classifying the toxic compounds into the hepatotoxicity classes, based on votes by binary classifiers. The binary classifiers are constructed by using genes selected to best elicit differences between the two classes. We show that the gene selection strategy improves the misclassification error rates and also delivers gene pathways that exhibit biological relevance.

A gene expression signature for oxidant stress/reactive metabolites in rat liver.

Formation of free radicals and other reactive molecules is responsible for the adverse effects produced by a number of hepatotoxic compounds. cDNA microarray technology was used to compare transcriptional profiles elicited by training and testing sets of 15 oxidant stressors/reactive metabolite treatments to those produced by approximately 85 other paradigm compounds (mostly hepatotoxicants) to determine a shared signature profile for oxidant stress-associated hepatotoxicity. Initially, 100 genes were chosen that responded significantly different to oxidant stressors/reactive metabolites (OS/RM) compared to other samples in the database, then a 25-gene subset was selected by multivariate analysis. Many of the selected genes (e.g., aflatoxin aldehyde reductase, diaphorase, epoxide hydrolase, heme oxgenase and several glutathione transferases) are well-characterized oxidant stress/Nrf-2-responsive genes. Less than 10 other compounds co-cluster with our training and testing set compounds and these are known to generate OS/RMs as part of their mechanisms of toxicity. Using OS/RM signature gene sets, compounds previously associated with macrophage activation formed a distinct cluster separate from OS/RM and other compounds. A 69-gene set was chosen to maximally separate compounds in control, macrophage activator, peroxisome proliferator and OS/RM classes. The ease with which these 'oxidative stressor' classes can be separated indicates a role for microarray technology in early prediction and classification of hepatotoxicants. The ability to rapidly screen the oxidant stress potential of compounds may aid in avoidance of some idiosyncratic drug reactions as well as overtly toxic compounds.

Inverse gene expression patterns for macrophage activating hepatotoxicants and peroxisome proliferators in rat liver.

Macrophage activation contributes to adverse effects produced by a number of hepatotoxic compounds. Transcriptional profiles elicited by two macrophage activators, LPS and zymosan A, were compared to those produced by 100 paradigm compounds (mostly hepatotoxicants) using cDNA microarrays. Several hepatotoxicants previously reported to activate liver macrophages produced transcriptional responses similar to LPS and zymosan, and these were used to construct a gene signature profile for macrophage activators in the liver. Measurement of cytokine mRNAs in the same liver samples by RT-PCR independently confirmed that these compounds are associated with macrophage activation. In addition to expected effects on acute phase proteins and metabolic pathways that are regulated by LPS and inflammation, a strong induction was observed for many endoplasmic reticulum-associated stress/chaperone proteins. Additionally, many genes in our macrophage activator signature profile were well-characterized PPARalpha-induced genes which were repressed by macrophage activators. A shared gene signature profile for peroxisome proliferators was determined using a training set of clofibrate, WY 14643, diethylhexylphthalate, diisononylphthalate, perfluorodecanoic acid, perfluoroheptanoic acid, and perfluorooctanoic acid. The signature profile included macrophage activator-induced genes that were repressed by peroxisome proliferators. NSAIDs comprised an interesting pharmacological class in that some compounds, notably diflunisal, co-clustered with peroxisome proliferators whereas several others co-clustered with macrophage activators, possibly due to endotoxin exposure secondary to their adverse effects on the gastrointestinal system. While much of these data confirmed findings from the literature, the transcriptional patterns detected using this toxicogenomics approach showed relationships between genes and biological pathways requiring complex analysis to be discerned.