Microarray analysis of hepatotoxins in vitro reveals a correlation between gene expression profiles and mechanisms of toxicity

Abstract

A rate-limiting step that occurs in the drug discovery process is toxicological evaluation of new compounds. New techniques that use small amounts of the experimental compound and provide a high degree of predictivity would greatly improve this process. The field of microarray technology, which allows one to monitor thousands of gene expression changes simultaneously, is rapidly advancing and is already being applied to numerous areas in toxicology. However, it remains to be determined if compounds with similar toxic mechanisms produce similar changes in transcriptional expression. In addition, it must be determined if gene expression changes caused by an agent in vitro would reflect those produced in vivo. In order to address these questions, we treated rat hepatocytes with 15 known hepatoxins (carbon tetrachloride, allyl alcohol, aroclor 1254, methotrexate, diquat, carbamazepine, methapyrilene, arsenic, diethylnitrosamine, monocrotaline, dimethyl-formamide, amiodarone, indomethacin, etoposide, and 3-methylcholanthrene) and used microarray technology to characterize the compounds based on gene expression changes. Our results showed that gene expressional profiles for compounds with similar toxic mechanisms indeed formed clusters, suggesting a similar effect on transcription. There was not complete identity, however, indicating that each compound produced a unique signature. These results show that large-scale analysis of gene expression using microarray technology has promise as a diagnostic tool for toxicology.

Introduction

Toxicity evaluation of chemicals is often the rate-limiting step in the discovery and development of new pharmaceuticals. The rate is established by both the necessary level of detail needed to assess compound liabilities fully and by the time needed to produce the large amount of an experimental compound required to conduct such a detailed evaluation in animals. With the increasing cost of new drug development there is the need to conduct toxicity evaluation as early as possible and on as many potential chemical leads as possible. Consequently there is increasing need to develop assays for toxicity evaluation that use small amounts (preferably milligrams) of experimental compounds while providing an accurate prediction of toxicity. Though cell culture systems can reduce the compound requirement to the necessary level, traditional in vitro cytotoxicity assays that measure endpoints such as cell lysis are of little predictive value. Mechanism-based assays that measure cell functional impairment or endpoints such as enzyme induction are of more predictive value, but determining toxic mechanism generally requires a considerable commitment of time and resources. The application of genomic technologies to toxicology research may greatly reduce this commitment.

Over the past few years numerous technologies have been developed that are revolutionizing pharmaceutical research. One such technology, DNA microarrays, allows one to monitor the expression of hundreds or thousands of genes at the same time. While this technology is immediately useful to quantitate the expression of select genes it also offers the ability to obtain information about the total transcriptional profile of a cell. The utility of transcriptional profiling as a sensitive method to classify cells, identify mechanisms of action for compounds and determine the function of previously unidentified genes has been demonstrated. For example, gene expression signatures were recently used to cluster acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) cells. Using microarray analysis, 29 out of 34 bone marrow or peripheral blood samples were classified as containing either AML or ALL cells with 100% accuracy (Golub et al., 1999). A similar analysis was performed using gene expression analysis from microarrays to classify 60 cell lines from diverse tumor tissues. From the gene expression signatures of the tumor cells alone it was possible to determine the tissue of origin from which the cell lines originated (Ross et al., 2000). In another example, it was shown that the functions of eight previously uncharacterized genes in the yeast system could be characterized using gene expression profiles. Also, a novel target of the drug dyclonine was identified (Hughes et al., 2000). Finally, it has also been shown that cell responses to xenobiotic treatment can be an indicator of the mechanism of action of the xenobiotic agent (Scherf et al., 2000). Recently, microarray and cluster analysis was used to study cell responses to 118 cancer drugs with known mechanisms of action. These studies indicated that cellular responses to cancer drugs clustered according to the mechanism of action of the drug regardless of the type of cell used.

The potential utility of using microarrays for toxicity evaluation is enormous. The immediate applications might include using microarrays to generate unique gene expression profiles (also called fingerprints or signatures) for compounds of known toxic mechanisms. Used as a database, these compound profiles could then be used as a reference to classify responses to new compounds according to pharmacologic or toxic phenotype (guilt by association). Identification of specific gene responses within a profile can also provide clues that may lead to the identification of specific mechanisms. Combined with an appropriate cell culture model much of this work could be done in vitro, which would reduce the need for large-scale synthesis of new compounds and allow for the comparison of multiple analogues. Identification of transcriptional changes that are indicative of specific compound liabilities would also provide markers for developing higher throughput assays that are more sensitive and predictive than cell death assays.

Before the potential utility of microarray technology can be realized, however, many questions must be addressed. For example, it remains to be determined if compounds with similar toxic mechanisms produce similar changes in transcriptional expression or, conversely, if compounds with very different mechanisms of toxicity can be differentiated based on gene expression profiles. It must also be determined if responses in cultured cells derived from a specific target tissue will recapitulate in vivo responses in that tissue; that is, if there is similarity or identity between in vitro and in vivo responses. In order to address some of these questions, we have treated primary rat hepatocytes with 15 different known hepatotoxic agents including carbon tetrachloride, allyl alcohol, aroclor 1254, methotrexate, diquat, carbamazepine, methapyrilene, arsenic, diethylnitrosamine, monocrotaline, dimethylformamide, amiodarone, indomethacin, etoposide, and 3-methylcholanthrene. These agents were selected based on the diversity of hepatic injuries they produce including necrosis, cirrhosis, hepatomegaly and carcinogenicity (Badr et al., 1986, Barak et al., 1984, Butterworth et al., 1978, Custer et al., 1977, Lijinsky et al., 1992, Lijinsky et al., 1980, Mayes et al., 1998, Ohshima et al., 1984, Recknagel, 1967). They also represent a variety of toxic mechanisms including DNA damage, generation of reactive oxygen species, P450 enzyme induction and others. Gene expression analysis was performed on RNA from treated hepatocytes and compared with that using vehicle-treated controls and the resultant gene expression profiles were clustered. Results show that the 15 hepatotoxic agents did separate into clusters and many of the compounds with similar known hepatic effects clustered together. The results show that microarray analysis may prove to be a highly sensitive technique for chemical toxicity evaluation.