Marys Medicine

Blumberg-serv.bio.uci.edu

HAMADEH ET AL.
creased by phenobarbital treatment, while fatty acid synthesiswas stimulated (Argaud et al., 1991; Thurman and Marazzo,1975). Although some of these gene inductions were previ-ously reported, there has been little effort to integrate theobservations in the context of phenobarbital's mechanism ofaction. Our results offer a more comprehensive overview ofmolecular responses to toxicant exposure by revealing thecoordinate expression of multiple genes in homeostasis andmetabolic pathways. The agreement of the expression profilesfor phenobarbital and peroxisome proliferators with past, tra-ditional studies lends confidence in the use of these geneexpression profiles in further pattern recognition applications.
Gene expression validation. Our gene expression data is validated in 3 ways. The first is by replicate analysis. We used3 animals for each compound and measured the gene expres-sion for each animal on 3 chips. This approach generated 9measurements for each gene. The combined use of a confi-dence interval and the binomial probability aided in eliminat-ing genes with high biological or technical variability fromfurther analyses (i.e., clustering). Secondly, we routinely con- Assessment of amplification of the RT-PCR specificity validation experiment using agarose gel electrophoresis: 60 ␮l of PCR reaction product duct resequencing of the clones we find significantly changed.
corresponding to each gene were concentrated, then separated on a 2% agarose Currently, our clone set shows an accuracy of approximately gel. Marker ladders (1 kb and 100 bp; Invitrogen, Carlsbad, CA) were run on 90%. Identification of clones is updated on our Web site, either side of the samples. Each lane shows a representative product for the http://dir.niehs.nih.gov/microarray/chips.htm. Finally, we vali- gene indicated.
dated the expression profile of 10 genes across samples derivedfrom individual animals exposed to peroxisome proliferators or fibrozil, or clofibrate. Furthermore, in the latter node, individ- phenobarbital at the 24-h and 2-week time points. Sample PCR ual animals were clustered in separate subnodes denoting the products were run on a 2% agarose gel and visualized by specific peroxisome proliferator to which they were exposed.
ethidium bromide staining. A representative gel indicating the We used principal components analysis (PCA), which is a quality of the reactions is shown in Figure 1. Comparison of pattern recognition technique that represents a multivariate data from cDNA microarray and real-time polymerase chain statistical method that is useful for reducing multidimensional reaction (RT-PCR) evaluations demonstrated a high level of data (such as high-density gene expression data) down to 2 or correlation between the 2 approaches (Table 3), where the 3 dimensions that can be readily comprehended. The principal induction or repression of each gene was confirmed across components were new variables created from linear combina- multiple samples. Microarray measurements are typically only tions of the starting variables (genes), where each principal semiquantitative, with compression of values occurring at component is orthogonal or not correlated with all others. The high-fold changes. The RT-PCR measurements are likely to first principal component contained the largest part of the provide better quantitation for genes such as p4502B2 (pheno- variance of the data set (37.5%) with the subsequent principal barbital 12.42 fold on microarray, 32-fold by RT-PCR), but components containing correspondingly smaller amounts of generally the quantitative measurements with the 2 approaches variance (18.7 and 12.1% for 2nd and 3rd, respectively). This are well correlated (Table 3).
analysis allowed the visualization, in 3-dimensional space for A critical question in toxicogenomics simplicity, of the discrimination between the gene expression is whether gene expression information may be used to reveal responses elicited by these 2 classes of compounds (Fig. 3).
chemical-specific signature patterns. We used several compu- PCA demonstrated close proximity in the gene expression tational analyses to determine whether different toxicants result pattern between clofibrate-, Wyeth 14,643-, and gemfibrozil- in distinguishable gene expression patterns. Application of exposed animals, but indicated a distinct partition (separation) hierarchical cluster analysis (Eisen et al., 1998) confirmed that between these compounds and phenobarbital-exposed animals.
individual animals could be distinguished by the class of tox- Finally, pairwise comparisons (Fig. 4) of gene expression icants to which they were exposed (Fig. 2) and revealed 2 profiles of individual animals exposed to chemicals showed distinct nodes containing animals treated with either of the 2 confirmation of the discerning potential of microarray data.
classes of chemicals. Across experiments, one node represents Comparison of gene expression profiles of two different ani- the 3 animals that were exposed to phenobarbital, while the mals exposed to the same compound resulted in a relatively other node includes animals treated with Wyeth 14,643, gem- high correlation (e.g., r ⬎ ⫹0.8 for Wyeth 14,643 vs. Wyeth GENE EXPRESSION ANALYSIS AND TOXIC MECHANISMS Validation of cDNA Microarray Data by Real-Time PCR
Accession no.
RT-PCR cDNA RT-PCR cDNA RT-PCR cDNA RT-PCR cDNA RT-PCR cDNA RT-PCR cDNA RT-PCR cDNA RT-PCR cDNA Carnitine Palmitoyl Transferase 1 Aflatoxin aldehyde reductase Cytosolic phosphoprotein (p19) Enoyl CoA isomerase Ah receptor Note. Values are fold induction/repression over control. C, clofibrate; W, Wyeth 14,643; G, gemfibrozil; P, phenobarbital; RT-PCR, real-time polymerase chain reaction-derived data (average of 3 measurements); cDNA, cDNA microarray-derived data (average of 3 measurements).
14,643) as compared to animals exposed to different com- exposure to the chemicals used in this study. Figure 6 illus- pounds belonging to the same class (e.g., r ⬃ ⫹0.6 for Wyeth trates the value of studying multiple time points, and the cluster 14,643 vs. gemfibrozil or clofibrate). However, comparisons of indicates genes that are altered at the delayed time point. These animals treated with toxicants belonging to different classes of genes may be further investigated or corroborated in future compounds resulted in a relatively low correlation (e.g., r ⬍ ⫹ studies for their association with chronic toxicity or adaptation 0.4 for Wyeth 14,643 vs. phenobarbital). In summary, the data in Figures 2, 3, and 4 indicate that through multiple approachesand bioinformatics tools, it is possible to discriminate gene expression profiles generated as a response to distinct livertoxicants. Proof of this concept aids in validation of the future The goal of this study was to determine whether generation potential of a predictive toxicogenomic strategy.
of chemically associated gene expression profiles, using mi- Time dependency of gene expression. One potential appli- croarray technology, would permit classification of compound cation of microarrays in toxicology is their use in predicting associated signatures. We generated gene expression profiles toxicity of undefined chemicals prior to the appearance of corresponding to 3 peroxisome proliferators, clofibrate, Wyeth pathological or disease outcomes. Gene expression profiles 14,643, and gemfibrozil, and an enzyme inducer, phenobarbi- from animals treated with a compound of unknown toxicity tal, and demonstrated that gene expression profiling is indeed a may be compared with a database of DNA microarray-gener- powerful tool for distinguishing gene expression generated by ated gene expression data corresponding to known compounds.
structurally unrelated toxicants in an in vivo model. Those However, toxicological effects are confounded by time and distinctions were made even when compounds shared some dose dependency of lesions that lead to differences in gene endpoints such as peroxisome proliferation. A greater similar- expression signatures. Therefore, we evaluated whether time- ity was found among peroxisome proliferators than among the independent gene expression responses that represent signa- peroxisome proliferators and phenobarbital. Clustering of tures of chemical-specific alterations occur. Application of genes that were significantly affected by the 24-h exposures, clustering methods to the data from phenobarbital and the where no histopathological abnormalities were detected, dem- peroxisome proliferators at 2 time points allowed the identifi- onstrated that gene expression profiles might be successfully cation of time-independent regulation (up or down) of genes in used for compound classification.
response to compound treatment (Fig. 5). Those genes that are Whether animals should be grouped together as a pool or regulated in the same fashion upon compound exposure at examined individually represents one issue in the design of multiple time points may potentially serve as reliable biomar- toxicogenomics studies. Some investigators advocate pooling kers of effect when establishing time-independent gene expres- for the costly microarray analysis and using individual animals sion profiles.
for the follow-up verification steps. However, pooling may RNA based gene expression analysis may be regarded as a cause misinterpretation of data if one animal shows a remark- snapshot of molecular occurrences in time/space. Analysis of ably distinct response, or lack of response. In this study we expression at one time point can be misleading due to transient analyzed individual chemically exposed animals against a pool and delayed gene expression events. We studied transient and of control animals. The generation of gene expression profiles delayed alterations in gene expression in response to the daily corresponding to 3 animals exposed to the same compound, as

HAMADEH ET AL.
Different class compound toxicants generate discrete gene expression profiles. Genes whose expression profile was significantly altered in any toxicant-exposed animal at the 24-h time point were clustered according to their expression levels across animals. Samples were also clustered based on thepattern of expression of the studied genes. A node is highlighted showing the expression pattern of a subset of genes representative of the overall grouping ofsamples derived from the livers of chemically exposed rats.
opposed to pooling, allowed for the detection of interanimal There is high concordance of the expression changes found differences. This facilitated the testing of the robustness of in the microarray analyses in phenobarbital-exposed rats with DNA microarray technology and pattern recognition algo- the results obtained by scientists using other methodologies rithms to generate distinguishable gene expression profiles, over many years of study. This concordance is illustrated in a despite the existence of differences in response across similarly display for gene expression changes in the form of an "effector treated animals. Although we have observed similarities as pathway map" (EPM) for chemical action. The phenobarbital high as about 90% between animals exposed to the same gene expression profile data set was mapped into previously agents, this similarity showed chemical dependence, reaching defined cellular response pathways to demonstrate the infor- about 80% with some of the compounds. The best interanimal mational power of this type of data presentation (Fig. 7).
correlation was observed among Wyeth 14,643-treated ani- Organization of data in this integrated diagram facilitates clear mals, where we saw the greatest number of statistically signif- visualization of regulatory and molecular events occurring as a icant gene expression modulations, suggesting that compound response to chemical exposure and visualization of pathways potency may be positively related to decreased variation in that were affected by phenobarbital. At a glance, this diagram gene expression response.
illustrates how the gene expression profile data has corrobo-

GENE EXPRESSION ANALYSIS AND TOXIC MECHANISMS gene function. The majority of genes that were expressed in atime-independent fashion in response to phenobarbital or peroxi-some proliferator treatment corresponded to enzymes that func-tion in compound metabolism (Fig. 5A; p450 2B2, epoxide hy-drolase 1, GST, aflatoxin B1 aldehyde reductase) or cellbiochemical processes (Figs. 5A–5C; Acyl CoA oxidase, Carni-tine palmitoyltransferase 1, histidine decarboxylase, stearyl-CoAdesaturase). This makes sense when one considers that animalswere treated with the studied chemicals on a daily basis, furnish-ing recurring surges in blood levels of the compounds in theexposed animals, and thus affecting compound metabolism genesor downstream effects.
Metabolism-related genes were notably absent from the tran- siently altered transcripts, the majority of which representedsignaling related genes (Fig. 6A). This is consistent with the The Partek Pro 2000 software package was used for visual PCA transient nature of the response and these genes probably of the data for genes that were altered in a statistically significant manner with constituted an initial response in the liver upon exposure to the any of the treatments used. The first principal component for this data setaccounted for 37.4% of the variation present, the second component for an specific toxicants for the first time. Alterations in gene expres- additional 18.7%, and the third for 12.1%. Each colored point represents data sion that were present at 2 weeks of exposure but not at the from an individual animal treated with the respective agent for 24 h.
24-h time point (Fig. 6B) constituted delayed responses totoxicant exposure. These responses could be associated withadaptation events or with the relation to the histopathological rated past findings (Fig. 7, yellow targets) and may contribute observation of hypertrophy noted in all animals treated by new mechanistic insights (Fig. 7, blue targets).
chemicals for 2 weeks. These changes could also be due to Chronic cell proliferation is proposed as a major mechanism overt toxicity that may be manifested in gene expression re- for tumor promotion by phenobarbital (Barbason et al., 1983; sponses but not necessarily detected by histopathological ex- Feldman et al., 1981; Whysner et al., 1996). Histopathological amination until a much longer period of exposure.
analysis indicated liver enlargement at 2 weeks in phenobar-bital-treated animals. Several studies have demonstrated thatphenobarbital induced increases in DNA synthesis in rats andmice of various strains (Busser and Lutz, 1987). Our datacorroborates phenobarbital-induced increase in cell prolifera-tion. Transcript levels for DNA polymerase b, AIRC-SAICARsynthase, cyclin B1, and ARF-like factor (ARL5) were in-creased at the 24-h and 2-week time points, relative to controls,by phenobarbital exposure, indicating increased DNA synthe-sis and liver proliferation. This observation is further supportedby the modulation in cytoskeleton rearrangement-relatedgenes, such as rho-associated protein kinase, which is essentialfor the rearrangement of actin cytoskeleton (Ohashi et al.,2000; Watanabe et al., 1999), suggesting a role for thesealterations in the hypertrophy observed in the livers of theexposed animals. We also observed the upregulation of phos-phoprotein stathmin (p19) at the 2 time points in response tophenobarbital exposure. Phosphoprotein stathmin (p19), whichis abundant in neuroendocrine tumor cells, showed a 15-fold The ratios of exposed to control animals corresponding to tran- script levels of genes whose expression was significantly altered in any greater abundance in newborn than in adult brain and its levels toxicant-exposed animal at the 24-h time point, were compared via set pairwise increased after two-thirds partial rat hepatectomy (Koppel et correlation analysis using Spotfire software (Spotfire, Inc., Cambridge, MA).
al., 1993), suggesting a role in regeneration and growth of Comparison of rats treated with the sample compound showed the highest various tissues. These data support the observation that phe- correlation (Wyeth 14,643-treated animals). Comparison of Wyeth 14,643 with nobarbital produces liver enlargement due to proliferation of either clofibrate- or gemfibrozil-treated animals yielded appreciable correla-tion, which was in agreement with the fact that both of those compounds liver cells and offers new insights on its molecular basis.
belonged to the same peroxisome proliferator class of toxicants. The correla- Analysis of transient, delayed, and time-independent alterations tion dropped sharply when Wyeth 14,643- and phenobarbital-treated animals suggested a relationship between the pattern of expression and were compared, and indicated poor correlation.

HAMADEH ET AL.
Two-dimensional, hierarchical clustering analysis of genes that were altered in a statistically significant manner at 95% confidence level in at least 2 of the replicate hybridizations performed on each sample. Data from 9 hybridizations, representing 3 replicates of 3 independent biological samples derived from rats treatedwith the same compound at each time point, were averaged, and those values were used for clustering analysis. Clustering analysis was performed across genes but notanimals. Genes are represented on the vertical axis while animals are on the horizontal axis. (A) Highlighted nodes from a hierarchical clustering tree, showing a suiteof genes whose transcripts were increased in phenobarbital-exposed animals over controls in a time-independent fashion. (B) Genes modulated by gemfibrozil treatmentin a similar fashion at 24 h and 2 weeks of exposure. (C) Nodes showing different genes that were modulated by peroxisome proliferators collectively in atime-independent fashion. Red, gene induction; green, gene repression in the treated samples, relative to control.
We have successfully generated gene expression profiles cals from the same class of compounds give rise to discern- for 3 peroxisome proliferators and an enzyme inducer, and able gene expression profiles that bear more similarity to we were able to show, through the application of pattern each other than to patterns corresponding to exposure to recognition algorithms and computational analyses, that compounds from a different class. Systematic development these patterns were distinct. We demonstrated that chemi- of an expanded database for gene expression responses to

GENE EXPRESSION ANALYSIS AND TOXIC MECHANISMS Illustration of transient ver- sus delayed responses in gene expres-sion. Nodes from 4 hierarchical cluster-ing trees corresponding to the chemicalsused in this study were highlighted toshow examples of (A) transiently alteredtranscripts or (B) genes that required adelayed period of time for up- or down-regulation, implicating those genes aspossible biomarkers of toxicant-associ-ated delayed toxicity or adaptation to ex-posure. Red, gene induction; green, generepression in the treated samples, relativeto control.
drugs and environmental pollutants may yield compound- cals that we utilized. In addition, our data revealed gene specific signature patterns that would also provide insights expression that has not been previously associated with the into affected regulatory and proliferative and repair/adap- compounds we used and suggest that such results will tive pathways. We demonstrated the validity of our expres- provide valuable targets for further investigations of the sion data by corroborating published reports on the chemi- mechanism of action of chemical hazards.

HAMADEH ET AL.
Genes modulated with phe- nobarbital exposure, displayed as a mapof effector pathways of toxicant (MEPT),a schematic diagram of phenobarbital-modulated gene expression events. En-zymes enclosed in yellow boxes werealtered in this study and were previouslyreported in the literature to be modulatedby phenobarbital treatment, while thosein blue were not previously associatedwith phenobarbital exposure and arenovel observations. The colored circlesindicate the type of modulation. The cir-cle before the slash in each groupingdenotes the 24-h time point; the circleafter the slash, the 2-week time point.
Red, upregulation; green, downregula-tion; black, no change.
GENE EXPRESSION ANALYSIS AND TOXIC MECHANISMS tor, on the activities of rat liver glutathione-requiring enzymes and catalase
in comparison to the action of phenobarbital. Cancer Res. 45, 5011–5019.
We thank Julie Foley, Jeffrey Tucker, Kelli Reynolds, Jeffrey Reece, Jen- Griffin, M. J., Kirsten, E., Carubelli, R., Palakodety, R. B., McLick, J., and nifer Collins, and Dr. Johnathan Ladapo for assistance in analyzing and Kun, E. (1984). The in vivo effect of benzamide and phenobarbital on liver compiling the data from this study. In addition, we acknowledge and thank enzymes: Poly(ADP-ribose) polymerase, cytochrome P-450, styrene oxide Drs. Jeff Trent, Mike Bittner, and Pat Hurban, and Donald Cox, for advice in hydrolase, cholesterol oxide hydrolase, glutathione S-transferase and UDP- development of our microarray center. We also thank Nigel Walker and Chris glucuronyl transferase. Biochem. Biophys. Res. Commun. 122, 770 –775.
Miller for their help and guidance in the RT-PCR analysis. Finally, we thank Hamadeh, H. K., Bushel, P. B., Paules, R., and Afshari, C. A. (2001).
Drs. Thomas Kunkel, Robert Maronpot, Samuel Wilson, Paul Nettesheim, and Discovery in toxicology: Mediation by gene expression array technology.
Masahiko Negishi for helpful comments on the preparation of this manuscript.
J. Biochem. Mol. Toxicol. 15, 231–242.
Hug, G., McGraw, C. A., Bates, S. R., and Landrigan, E. A. (1991). Reduction of serum carnitine concentrations during anticonvulsant therapy with phe-nobarbital, valproic acid, phenytoin, and carbamazepine in children. J. Pe- Amacher, D. E., Beck, R., Schomaker, S. J., and Kenny, C. V. (1997). Hepatic diatr. 119, 799 – 802.
microsomal enzyme induction, ␤-oxidation, and cell proliferation following IARC (1977). IARC monographs on the evaluation of the carcinogenic risk of administration of clofibrate, gemfibrozil, or bezafibrate in the CD rat. Toxi- chemicals to man: Some miscellaneous pharmaceutical substances. IARC col. Appl. Pharmacol. 142, 143–150.
Monogr. Eval. Carcinog. Risk Chem. Man. 13, 1–255.
Argaud, D., Halimi, S., Catelloni, F., and Leverve, X. M. (1991). Inhibition of IARC (1987). Overall evaluations of carcinogenicity: An updating of IARC gluconeogenesis in isolated rat hepatocytes after chronic treatment with Monographs volumes 1 to 42. IARC Monogr. Eval. Carcinog. Risks Hum. 7
phenobarbital. Biochem. J. 280, 663– 669.
(Suppl.), 1– 440.
Barbason, H., Rassenfosse, C., and Betz, E. H. (1983). Promotion mechanism Koppel, J., Loyer, P., Maucuer, A., Rehak, P., Manceau, V., Guguen-Guil- of phenobarbital and partial hepatectomy in DENA hepatocarcinogenesis louzo, C., and Sobel, A. (1993). Induction of stathmin expression during cell-kinetics effect. Br. J. Cancer 47, 517–525.
liver regeneration. FEBS Lett. 331, 65–70.
Biegel, L. B., Hurtt, M. E., Frame, S. R., Applegate, M., O⬘Connor, J. C., and Melnick, R. L., Morrissey, R. E., and Tomaszewski, K. E. (1987). Studies by Cook, J. C. (1992). Comparison of the effects of Wyeth 14,643 in Crl:CD the National Toxicology Program on di(2-ethylhexyl)phthalate. Toxicol. BR and Fisher-344 rats. Fundam. Appl. Toxicol. 19, 590 –597.
Ind. Health 3, 99 –118.
Burchiel, S. K., Knall, C. M. Davis, J. W., III, Paules, R. S. Boggs, S. E.
Nuwaysir, E. F., Bittner, M., Trent, J., Barrett, J. C., and Afshari, C. A. (1999).
Afshari, C. A. (2001). Analysis of genetic and epigenetic mechanisms of Microarrays and toxicology: The advent of toxicogenomics. Mol. Carcinog.
toxicity: Potential roles of toxicogenomics and proteomics in toxicology.
Toxicol. Sci. 59, 193–195.
Ohashi, K., Nagata, K., Maekawa, M., Ishizaki, T., Narumiya, S., and Mizuno, Bushel, P. R., Hamadeh, H., Bennett, L., Sieber, S., Martin, K., Nuwaysir, K. (2000). Rho-associated kinase ROCK activates LIM-kinase 1 by phos- E. F., Johnson, K., Reynolds, K., Paules, R. S., and Afshari, C. A. (2001).
phorylation at threonine 508 within the activation loop. J. Biol. Chem. 275,
MAPS: A microarray project system for gene expression experiment infor- mation and data validation. Bioinformatics 17, 564 –565.
Schoonjans, K., Staels, B., and Auwerx, J. (1996). Role of the peroxisome Busser, M. T., and Lutz, W. K. (1987). Stimulation of DNA synthesis in rat proliferator-activated receptor (PPAR) in mediating the effects of fibrates and mouse liver by various tumor promoters. Carcinogenesis 8, 1433–1437.
and fatty acids on gene expression. J. Lipid Res. 37, 907–925.
Tavoloni, N., Jones, M. J., and Berk, P. D. (1983). Dose-related effects of Butterworth, B. E., Conolly, R. B., and Morgan, K. T. (1995). A strategy for phenobarbital on hepatic microsomal enzymes. Proc. Soc. Exp. Biol. Med.
establishing mode of action of chemical carcinogens as a guide for ap- 174, 20 –27.
proaches to risk assessments. Cancer Lett. 93, 129 –146.
Thomas, R. S., Rank, D. R., Penn, S. G., Zastrow, G. M., Hayes, K. R., Pande, Chen, Y., Dougherty, E, R., and Bittner, M. L. (1997). Ratio-based decisions K., Glover, E., Silander, T., Craven, M. W., Reddy, J. K., Jovanovich, S. B., and the quantitative analysis of cDNA microarray images. J. Biomed. Optics and Bradfield, C. A. (2001). Identification of toxicologically predictive gene 2, 364 –374.
sets using cDNA microarrays. Mol. Pharmacol. 60, 1189 –1194.
DeRisi, J., Penland, L., Brown, P. O., Bittner, M. L., Meltzer, P. S., Ray, M., Thurman, R. G., and Marazzo, D. P. (1975). Mixed-function oxidation and Chen, Y., Su, Y. A., and Trent, J. M. (1996). Use of a cDNA microarray to intermediary metabolism: Metabolic interdependencies in the liver. Adv. analyse gene expression patterns in human cancer. Nat. Genet. 14, 457– 460.
Exp. Med. Biol. 58, 355–367.
Duggan, D. J., Bittner, M., Chen, Y., Meltzer, P., and Trent, J. M. (1999).
Waring, J. F., Ciurlionis, R., Jolly, R.A., Heindel, M., and Ulrich, R. G.
Expression profiling using cDNA microarrays. Nat. Genet. 21(Suppl. 1),
(2001a). Microarray analysis of hepatotoxins in vitro reveals a correlation between gene expression profiles and mechanisms of toxicity. Toxicol. Lett.
Eisen, M. B., Spellman, P. T., Brown, P. O., and Botstein, D. (1998). Cluster 120, 359 –368.
analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Waring, J. F., Jolly, R. A., Ciurlionis, R., Lum, P. Y., Praestgaard, J. T., Sci. U.S.A 95, 14863–14868.
Morfitt, D. C., Buratto, B., Roberts, C., Schadt, E., and Ulrich, R. G.
Feldman, D., Swarm, R. L., and Becker, J. (1981). Ultrastructural study of rat (2001b). Clustering of hepatotoxins based on mechanism of toxicity using liver and liver neoplasms after long-term treatment with phenobarbital.
gene expression profiles. Toxicol. Appl. Pharmacol. 175, 28 – 42.
Cancer Res. 41, 2151–2162.
Watanabe, N., Kato, T., Fujita, A., Ishizaki, T., and Narumiya, S. (1999).
Fielden, M. R., and Zacharewski, T. R. (2001). Challenges and limitations of Cooperation between mDia1 and ROCK in Rho-induced actin reorganiza- gene expression profiling in mechanistic and predictive toxicology. Toxicol. tion. Nat. Cell Biol. 1, 136 –143.
Sci. 60, 6 –10.
Whysner, J., Ross, P. M., and Williams, G. M. (1996). Phenobarbital mecha- Furukawa, K., Numoto, S., Furuya, K., Furukawa, N. T., and Williams, G. M.
nistic data and risk assessment: Enzyme induction, enhanced cell prolifer- (1985). Effects of the hepatocarcinogen nafenopin, a peroxisome prolifera- ation, and tumor promotion. Pharmacol. Ther. 71, 153–191.

Source: http://blumberg-serv.bio.uci.edu/past%20teaching/bio145b-sp2004/required%20reading/hamadeh-2002.pdf