|Home | About | Journals | Submit | Contact Us | Français|
The genotoxicity testing battery is highly sensitive for detection of chemical carcinogens. However, it features a low specificity and provides only limited mechanistic information required for risk assessment of positive findings. This is especially important in case of positive findings in the in vitro chromosome damage assays, because chromosome damage may be also induced secondarily to cell death. An increasing body of evidence indicates that toxicogenomic analysis of cellular stress responses provides an insight into mechanisms of action of genotoxicants. To evaluate the utility of such a toxicogenomic analysis we evaluated gene expression profiles of TK6 cells treated with four model genotoxic agents using a targeted high density real-time PCR approach in a multilaboratory project coordinated by the Health and Environmental Sciences Institute Committee on the Application of Genomics in Mechanism-based Risk Assessment. We show that this gene profiling technology produced reproducible data across laboratories allowing us to conclude that expression analysis of a relevant gene set is capable of distinguishing compounds that cause DNA adducts or double strand breaks from those that interfere with mitotic spindle function or that cause chromosome damage as a consequence of cytotoxicity. Furthermore, our data suggest that the gene expression profiles at early time points are most likely to provide information relevant to mechanisms of genotoxic damage and that larger gene expression arrays will likely provide richer information for differentiating molecular mechanisms of action of genotoxicants. Although more compounds need to be tested to identify a robust molecular signature, this study confirms the potential of toxicogenomic analysis for investigation of genotoxic mechanisms.
The standard genotoxicity testing battery consisting of a bacterial mutation assay and chromosome damage assays in vitro and in vivo provides a relatively simple and economical assessment of genetic hazard, namely the ability of chemicals to cause genetic damage manifested as point mutations and/or chromosome damage. Because of the mechanistic connection between genetic damage and cancer development and also the practical limitations of long-term in vivo carcinogenicity assays, the genotoxicity testing battery has been widely applied as a surrogate for the prediction of carcinogenicity. Currently, the genotoxicity testing battery is an integral part of the safety assessment paradigm for drugs and industrial or environmental chemicals.
The detailed analysis of genotoxicity testing results revealed a high sensitivity but low specificity for detection of carcinogens (Kirkland et al., 2005, 2006). In the case of pharmaceuticals 50% of noncarcinogens had some genotoxicity findings in at least one test particularly in the mammalian in vitro chromosome damage assays (Snyder and Green, 2001). Consequently, the relevance of positive genotoxicity findings to cancer risk, especially those occurring in the in vitro chromosome damage assays at concentrations above the therapeutic exposure and in presence of substantial cellular toxicity, has been questioned (Kirkland et al., 2006).
In theory, a single mutation can initiate cancer formation. Therefore, the risk assessment of positive genotoxicity findings is typically based on linear extrapolation methods. On the other hand, a substantial body of evidence for thresholded dose responses exists particularly in cases where the genotoxicity arises as a consequence of secondary effects, without any direct impact on DNA. Mechanisms producing such responses are considered as non-DNA reactive and include inhibition of enzymes involved in protein and DNA synthesis or DNA repair, inhibition of Na+/K+ transport, inhibition of topoisomerases, interference with processes leading to imbalance of DNA precursors, energy depletion, production of active oxygen species, lipid peroxidation, nuclease release from lysosomes, interaction with the mitotic apparatus and the in vitro experimental conditions of extreme changes in pH and osmolarity (Scott et al., 1991). Therefore, the assessment of direct genotoxicity is frequently complemented with a variety of other genotoxicity assays and endpoints in vitro and in vivo, for example, DNA adduct detection assays such as 32P labeling, Comet assay, and micronucleus tests, providing a “weight of evidence approach” (Anonymous, 2004; Dearfield and Moore, 2005; Thybaud et al., 2007a). Nevertheless, these assays provide only limited insights into the mechanisms underlying non-DNA reactive genotoxins. Investigating genotoxic mechanisms using available assays requires lengthy experimental follow-up strategies with uncertain outcome, which often lead to significant delays in the introduction of new medicines to patients. Therefore, the development of alternative mechanism-based testing strategies is critical.
The cellular response to stress sets off a complex process of molecular pathways involved in repair, survival, and/or cell death. The components of the response can be regulated at the transcriptional or translational levels, during posttranslational modifications, and/or by specific molecular interactions. Historically, the treatment of bacterial, yeast and mammalian cells using ionizing radiation and prototypical DNA-damaging agents led to identification of DNA-damage inducible genes. Those genes play an important role in cell cycle arrest and apoptosis both regulated via the p53 pathway, and activation of mitogen-activated protein kinase cascades, nuclear factor kappaB and the activator protein 1 transcription complex (Liu et al., 1996) (see reviews Amundson et al., 2003a,b; Snyder and Morgan, 2004). Furthermore, gene expression patterns consisting of stress response genes characterized in ex vivo experiments were also reported in humans after total body irradiation procedures (Amundson et al., 2004).
The availability of oligonucleotide microarrays enabling to evaluate cellular responses to damage at the level of the genome via monitoring the global changes of mRNA expression gave rise to the field of Toxicogenomics (TGx). The application of TGx approaches to genetic toxicology has been extensively reviewed in (Aubrecht and Caba, 2005; Thybaud et al., 2007b). The published results suggest that the toxicogenomic analysis of genotoxic stress responses could differentiate DNA reactive and non-DNA reactive mechanisms of genotoxicity at cytotoxic compound concentrations in two widely used cell lines for genotoxicity assessments: the p53 proficient human TK6 cells (Amundson et al., 2005; Islaih et al., 2004; Le Fevre et al., 2007) and the p53 deficient mouse L5178Y cells (Dickinson et al., 2004; Hu et al., 2004). The most comprehensive studies so far performed in these cell lines characterized stress-specific gene expression profiles for 13 DNA-damaging and cytotoxic (non-DNA damaging) agents in TK6 cells and in the p53 deficient isogenic cell line (Amundson et al., 2005) and 14 anticancer drugs representing DNA reactive and non-DNA reactive in vitro genotoxins in TK6 cells (Le Fevre et al., 2007).
Being involved from the early stages of TGx application to genotoxicity testing, the Health and Environmental Sciences Institute (HESI) Committee on the Application of Genomics in Mechanism-based Risk Assessment from the International Life Science Institute has focused one of its collaborative research programs on the evaluation of the utility of TGx analysis to improve risk assessment of genotoxicants (Aubrecht and Caba, 2005; Newton et al., 2004; Pennie et al., 2004), more specifically trying to advance the understanding of mechanisms of action/toxicity of DNA reactive versus non-DNA reactive genotoxins and to develop classification models to discriminate them between each other. Towards this goal, we have set out to evaluate the gene expression profile of TK6 cells treated with four model genotoxic agents, namely the DNA reactive cisplatin, the topoisomerase inhibitor etoposide, the mitotic spindle poison taxol and the cytotoxic clastogen NaCl, using a targeted TaqMan real-time PCR approach based on a set of 47 preselected genes, with the selection being made on the basis of data previously obtained by the group and reported in the literature. The transcriptomic analysis was performed immediately and 20 h after 4 h of treatment with each drug. Furthermore, we have evaluated the robustness of the gene expression data across nine independent laboratories.
The cell culture and treatment and microarray analyses were conducted by the following independent laboratories as part of the HESI Committee on Genomics: Bayer Healthcare AG, Hoffman-La Roche, Inc., Institute de Recherches Internationales SERVIER, Meiji Seika Kaisha, Ltd, Novartis Pharma AG, Pfizer, Inc., Sanofi aventis R&D, Schering-Plough Research Institute, Taiho Pharmaceutical Co., Ltd., and Mitsubishi Tanabe Pharma Corporation.
Cis-Platinum(II)-Diamine Dichloride (cisplatin), Paclitaxel (taxol), and etoposide (all from Sigma Aldrich, St Louis, MO) were dissolved in dimethylsulfoxide (DMSO) (Sigma Aldrich). Final concentration of DMSO in the cell culture medium did not exceed 1%. Sodium chloride (NaCl) (J.T. Baker, Phillipsburg, NJ) was dissolved in culture medium and was filter sterilized prior to use. Test solutions were prepared immediately prior to use.
The human lymphoblastoid cell line TK6 was obtained from ATCC (ATCC, Manassas, VA). Cells were routinely maintained as exponentially growing suspensions in complete RPMI 1640 medium containing 10% heat inactivated fetal bovine serum, 4mM L-glutamine, 1mM MEM sodium pyruvate, 100 units/ml penicillin and 100 μg/ml streptomycin (all from Gibco, Grand Island, NY). Cultures were incubated at 37°C in a humidified atmosphere of 5% CO2. TK6 cells have a doubling time of approximately 24 h under the conditions used in these studies. Exponentially growing cells were seeded at a concentration of 5 × 105 cells/ml. Cultures were exposed to medium containing dilutions of test compound or solvent for 4 h, after which medium was replaced with fresh, complete RPMI medium and incubated for an additional 20 h. A 2-h and 7-h post-treatment samples were also collected by two labs, and 5 labs, respectively but these time points were not considered for the cross-laboratory evaluation. At each time point (4 and 24 h after the beginning of exposure), an aliquot from each dose level and solvent control was taken; the sample was pelleted by centrifugation, washed with PBS, snap-frozen in liquid nitrogen and stored at −80°C until processing. Each experiment was run independently a minimum of three times.
Most of the nine participating laboratories selected taxol and cisplatin as test compounds; etoposide and NaCl were evaluated by fewer laboratories (Table 1). Each participating laboratory carried out dose-ranging studies to determine the concentration of each compound that would result in an approximate 50% reduction in cell viability, which was then used for the high dose in that laboratory. The growth inhibition determined by a Coulter counter served as a measure of cell viability. Relative viability calculations were made based on concurrent solvent cultures. Because the data were comparable in several laboratories, a fixed low (L, no cytotoxicity observed), medium (M) and high (H) concentration was suggested for use in the gene expression experiments in case of taxol and cisplatin, which were 1, 10, and 50 μg/ml, and 1, 10, and 30 μg/ml, respectively. The high doses were based on ca. 50% cytotoxicity at the 24-h time point.
RNA was extracted with the Qiagen RNeasy Mini Kit (Qiagen, Valencia, CA) using the RNeasy Mini Protocol for Isolation of Total RNA from Animal Cells and Qiashredders to homogenize samples. The purified RNA was used to generate cDNA by reverse transcription using the ABI High Capacity cDNA Archive Kit and protocol (Applied Biosystems, Foster City, CA). The expression of a panel of 47 genes (Table 2) was measured using either a custom-made Taqman card with these 47 genes plus 18S rRNA as control (three laboratories) or individual TaqMan Assays obtained from ABI (Applied Biosystems, Foster City, CA). In the latter case some laboratories also measured G3BP2 mRNA levels. Reverse transcription and TaqMan PCR reactions were performed according to the manufacturer's instructions (Applied Biosystems). Genes for which the expression level was below detection in both treated and control cells were excluded from further analysis. Gene deregulation ratios were calculated based on the ΔΔCt method:
Deregulation ratio = 2−ΔΔCt; > 1: Upregulation of a gene in the treated versus the control sample; < 1: downregulation of a gene in the treated versus the control sample.
The data have been deposited in the Chemical Effects in Biological Systems database at the National Institute of Environmental Health Sciences, Research Triangle Park, NC. Both the raw data and the final normalized dataset are available currently from https://dir-apps.niehs.nih.gov/arc/ and later from www.cebs.niehs.nih.gov. With the exception of genes with detection near the limits of sensitivity, the variability among technical replicates was near zero and among biological replicates within 1.4%. Because the different labs combined the data at different stages of the calculation of delta Ct, it was difficult to compute an overall statistic for the data. The mean value of all replicates was computed to give a single delta-delta Ct value for each gene-sample pair, where the samples were within the range of detection; for samples with undetectable levels of RNA in either control or treated, a value of 50 or 1/50 was arbitrarily assigned, respectively.
Although the gene expression response to treatment with DNA-damaging agents is quite complex (Amundson and Fornace, 2003; Fornace et al., 1999), including multiple pathways (Liu et al., 1996; Snyder and Morgan, 2004) it has been suggested that the differentiation of DNA reactive vs. non-DNA reactive genotoxic mechanisms might require a relatively small number of genes (Aubrecht et al., 1999; Dickinson et al., 2004; Hu et al., 2004; Le Fevre et al., 2007). Therefore, we decided to design a transcriptomic approach based on TaqMan quantitative real-time PCR technology using a gene set identified from published data (Table 2). The relative simplicity of the this real-time PCR technology allowed for evaluating time courses and dose responses of gene expression changes further to treatment to model genotoxicants. In addition, our unique experimental set-up as a consortium of participating laboratories enabled us to assess interlaboratory variability of the approach and consistency of compound evaluation with respect to differentiation of genotoxic mechanisms.
A review of the published data on cellular gene expression changes upon treatment with various genotoxicants revealed genes with known or suggested biological activities acting via both direct or indirect DNA damage (Table 2). Because p53 is a central transcriptional regulator activated upon DNA damage (e.g., Amundson et al., 2001), the gene set included the known p53 target genes cyclin-dependent kinase inhibitor 1A (CDKN1A), DNA damage inducible transcript 3 (DDIT3), transformed 3T3 cell double minute 2 protein (MDM2), pleckstrin homology-like domain A3 (PHDA3), protein phosphatase 1D magnesium-dependent protein (PPM1D, WIP1); tumor protein p53 inducible protein 3 (TP53I3), cyclin G1 (CCNG1), and Bcl2-associated protein X protein (BAX). Those were found deregulated in one or several of the studies reviewed (Akerman et al., 2004; Amundson et al., 2005; Aubrecht and Caba, 2005; Dickinson et al., 2004; Islaih et al., 2004). Further biological functions represented in this gene set were a general or oxidative stress response, DNA repair, apoptosis, cell cycle progression, and inflammation. The general and oxidative stress response category contained activating transcription factor 3 (ATF3), growth differentiation factor 15 (GDF15); transcription factors cFOS and cJUN, growth arrest and DNA damage inducible proteins alpha, beta, and gamma (GADD45 A, B, C), glutathione S-transferase A2 (GSTA2), endoplasmic reticulum stress-inducible gene (HERPUD1), in addition to eukaryotic translation initiation factor 1 (EIF1) and stress genes involved in protein folding encompassing DnaJ (Hsp40) homolog B1 (DNAJB1 or HSP40), heat shock 70-kDa protein 14 (HSPA14), and heat shock 70-kDa protein 5 (HSPA5). Among the stress-responsive genes GDF15 and GADD45A are known to be induced by p53, among other transcription factors (Riley et al., 2008). The DNA damage repair genes included APEX nuclease (APEX1) and xeroderma pigmentosum complementation group C gene (XPC). Genes involved primarily in cell cycle regulation comprised aurora kinase A (AURKA), inner centromere protein antigen (INCEP), proliferating cell nuclear antigen (PCNA), polo-like kinase 1 (PLK1), SAC3 domain containing 1 (SAC3D1,) cyclin-dependent kinase inhibitor 1C (CDKN1C/kip2), transducer of ERBB2 (TOB1), male cell-associated kinase (MAK) and the topoisomerases TOP1 and TOP2A. Genes with a role in chromatin organization such as matrin 3 (MATR3), H2A histone family member X (H2AFX), and histones 1H1c and H2ac (HIST1H1C and HIST1H2AC) may be connected to cell cycle progression or to changes in the DNA structure upon damage. Other genes reported to respond to DNA damage were involved in apoptosis such as B-cell CLL/lymphoma 2 (BCL2), lymphoid enhancer-binding factor 1 (LEF1), myeloid cell leukemia sequence 1 (MCL1), caspase 1 (CASP1), and granzymes A and B (GZMA and GZMB). The inflammatory response was represented by interleukin 1 beta (IL1B) and vesicle-associated membrane protein 3 (VAMP3) and mitochondrial function was represented by NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 1, 7.5 kDa (NDUFA1).
To evaluate the potential of our gene set to differentiate genotoxic mechanisms and to assess interlaboratory variability of the toxicogenomic approach, we designed a relatively simple study measuring gene expression changes using quantitative real-time PCR after treatment of TK6 cells with four model agents including cisplatin as a DNA cross-linking agent, etoposide as a topoisomerase II inhibitor, taxol as an aneugen, or sodium chloride as a cytotoxic clastogen (Fig. 1). Our testing protocol consisted of treating exponentially growing TK6 cells with the test agent for 4 h followed by medium changes and subsequent incubation in medium without test agent for a total of 24 h. The cytotoxicity was assessed as a relative decrease of cell count at 4-, 7-, and 24-h post-treatment. The genotoxic stress associated alteration of mRNA levels was detected using RT-PCR (TaqMan) at 4- and 24-h time points. Dose responses were evaluated via exposing the cells to a low (L), medium (M), and high (H) concentration of the test agent. The doses used in the gene expression studies were determined in a prior dose-range finding study in each laboratory using the cell count as a cytotoxicity indicator. For cisplatin and taxol, which were used as test agents by most laboratories, specific doses were suggested to be used in the gene expression studies, which were followed by five and three laboratories, respectively. The other laboratories (three for cisplatin and five for taxol) used doses which produced the agreed-upon cytotoxicity levels in their laboratories. The high dose was selected on the basis of ca. 50% cytotoxicity at the 24-h time point. Although we did not measure chromosome damage, the selected concentrations in our toxicogenomic studies for cisplatin and etoposide were reported to induce three- to fourfold induction of micronuclei frequency in TK-6 cells and 3- to 14-fold induction was reported for taxol (Le Fevre et al., 2007). Furthermore, sodium chloride exposures used in our studies produced reportedly two- to threefold induction of micronuclei frequency in L5178Y cells (Dickinson et al., 2004). Overall the common study design included three biological replicates for the gene expression studies, and at least the 4- and 24-h time point. Table 1 lists laboratories together with the agents tested, and the doses and time points employed.
Although the data from participating laboratories follow a similar trend there was a significant variability with respect to individual data points and to the extent of cytotoxicity reached at the 24-h time point (Fig. 2). Cytotoxicity, which was defined as a decrease in cell number of greater than 50–60%, was seen at the 24-h time points in most laboratories that tested cisplatin, etoposide and taxol, although laboratory 1 did not achieve this level of cytotoxicity for cisplatin and taxol, and laboratory 9 did not achieve it for etoposide. Three laboratories studied sodium chloride, and in this case the high dose produced a decrease in cell number also at early time point as well as at 24 h. Not all laboratories consistently achieved a cytotoxic dose for sodium chloride. Thus with the DNA damage inducing compounds (cisplatin, etoposide) and the aneugen (taxol), clear cytotoxicity is observed only at 24 h, whereas for the cytotoxic clastogen sodium chloride, dose-dependent cytotoxicity is already seen 4 h after treatment start, yet with higher variability between laboratories.
The gene expression profile analysis was performed using the TaqMan technology, with three laboratories running TaqMan Low Density Arrays (TQcards), and six laboratories employing TaqMan Assays-on-Demand. It should be noted that, although a general common protocol was agreed upon, the specific protocol details were set according to individual lab protocols, thus representing real life experiments. As outlined above, the gene set included genes with biological relevance to the genotoxic stress response (Table 2) such as p53 target genes and genes involved in DNA repair, chromatin formation, pro- and antiapoptosis, cell cycle progression, general or oxidative stress responses and inflammatory responses.
Among the 47 genes selected for gene expression measurements, 7 were below the detection limit in both vehicle control and treated samples in all laboratories (Table 2, those not checked in column 6). These were excluded from further comparisons. To evaluate the similarities among individual gene expression profiles across laboratories, the gene expression ratios at 4 and 24 h were used for principal components analysis, employing only the 40 genes expressed across all studies out of the list of 47 genes (Table 2 and Fig. 3). In this analysis, plotted data points represent expression profiles of individual samples. Increased similarity among gene expression profiles is represented as closer proximity. The PCA is composed of 134 expression profiles. Considering the facts that each expression profile consists of 40 genes, and that the ratios shown are the means of three biological replicates, this corresponds to 402 single replicate ratio calculations and > 800 (treated and untreated, and 18S rRNA control) measurements, not considering the technical replicates. Therefore, this PCA approach is helpful to identify potential outliers in our relatively large interlaboratory evaluation study. The PCA analysis showed that the majority of the data clustered in a tight cluster suggesting robustness of the real-time PCR measurements. Nevertheless, data from three participating laboratories (#5, #6, and #7) were scattered outside this cluster. Thus these gene expression profiles were considered as outliers, and removed from further analysis.
A detailed analysis of the gene expression profiles on an individual gene level revealed several interesting patterns (Fig. 4). The cells treated with the cross-linking agent cisplatin and the topoisomerase inhibitor etoposide showed a strong relationship between induction of cytotoxicity at 24 h and dose-dependent gene expression changes detected at both the 4- and 24-h time points. In the studies where relevant cytotoxicity exposures were achieved, most of the known p53 target genes in our panel showed increased expression at both early and late time points. The p53 target gene CCNG1 was not found induced in our experiments in the TK6 cells. Additionally, there is a decrease in expression of genes involved in cell cycle progression at both or at least the later time point (AURKA, PLK1, SHD1 = SAC3D1, TOP1, TOP2A, INCENP, HIST1H1C, HIST1H2AC) suggesting a cell growth arrest. Treatment with the aneugen taxol resulted in a homogeneous cross-laboratory increase in p53 target genes only 24-h postinitiation of treatment. The cytotoxic clastogen sodium chloride resulted in highly variable changes in gene expression with no clear pattern with respect to dose and time.
Figure 5 highlights the expression changes of genes in three categories, p53 target genes, cell cycle progression/antiapoptosis, and stress response genes, in a time-compound combination-dependent manner. At both 4- and 24-h time points (Figs. 5A and 5B, respectively), the expression of p53 target genes increased after treatment with cisplatin and etoposide. Treatment with taxol, a mitotic spindle poison and thus an aneugen, caused increased expression of this class of genes at 24 h, but not at 4 h. Treatment with the cytotoxic clastogen sodium chloride resulted in inconsistent expression changes. Growth differentiation factor 15 (GDF15) shows a profile very similar to known p53 target genes. The expression of genes with a known or suggested function in cell cycle progression, decreased at both 4 and 24 h following treatment with cisplatin and etoposide. There was little effect on this class of genes by taxol, except that expression of a specific histone gene, HIST1H2AC, was increased at the 24-h time point. The responses of these cell cycle genes were inconsistent after exposure to sodium chloride. Expression of stress response genes increased following all the treatment classes tested.
Genetic toxicology testing provides a relatively simple and accurate assessment of genotoxic hazard namely the chemical's potential to cause DNA damage in experimental test systems. Because of the mechanistic connection between DNA damage and cancer development and practical considerations such as lack of short term assays for the detection of carcinogenicity, genotoxicity testing is used as a surrogate endpoint for predicting of oncogenic potential in early stages of drug development and for the evaluation of industrial and environmental chemicals (Jacobson-Kram and Jacobs, 2005). Although the genotoxicity testing battery has a high 93% sensitivity for detection of carcinogens, the testing paradigm features a low specificity (Kirkland et al., 2005). For instance, 50% of noncarcinogens among pharmaceuticals had some positive findings in at least one genotoxicity assay. The in vitro chromosome damage assays are particularly sensitive but also of low specificity (Snyder and Green, 2001). Therefore, the development of new mechanism-based approaches to assess the relevance of positive findings in the in vitro chromosome damage assays is very important in light of human cancer risk. Recently, emerging TGx approaches consisting of monitoring gene expression changes on a transcriptome level have enabled investigating molecular mechanisms of toxicity (Hamadeh et al., 2002) including genotoxicity (Aubrecht and Caba, 2005). In this study, we set out to assess the potential of a real-time PCR-based gene expression profile analysis for differentiating genotoxic mechanisms and evaluate the robustness of the approach in a cross-laboratory evaluation study.
To evaluate the potential of gene expression profile analysis to differentiate genotoxic mechanisms we have selected four model agents: Cisplatin as a DNA reactive genotoxicant which induces direct DNA damage as a consequence of cross-linking (Zwelling and Kohn, 1979), etoposide which causes DNA double strand breaks via inhibition of topoisomerases II (Ross et al., 1984) taxol as an aneugen which produces chromosome damage via interaction with spindle microtubules during cell division (Schiff et al., 1979), and NaCl as an example of an agent producing structural chromosome damage in vitro as a consequence of cytotoxicity (Ashby and Ishidate, 1986). The selection of these model compounds provided an opportunity to assess the gene expression signature across a broad range of genotoxic mechanisms. Because we wanted to evaluate the responses in a relevant in vitro model we have selected the TK-6 lymphoblastoid cell line that carries a functional p53 gene and has been widely used in published TGx studies investigating diverse genotoxins (Amundson et al., 2005; Hu et al., 2004; Le Fevre et al., 2007).
The treatment of TK6 cells with model agents resulted in a measurable dose-dependent suppression of cellular growth for cisplatin, etoposide and taxol 24-h post-treatment. Although there was a significant variability among the participating laboratories the cytotoxicity profile followed a similar trend across all sites, and the discrepancies might be attributed to general biological variability and minor differences in cell culture techniques. The fact that we did not detect any measurable decrease of cell count at the earlier time point 4 h after initiation of treatment with cisplatin, etoposide and taxol could be explained by a 24-h long cell cycle time of TK6 cells, and the absence of actual cell lysis. On the other hand, the treatment of TK6 cells with sodium chloride produced a decrease of cell numbers also at the early time point, however, it was highly variable. This suggests that the toxicity of sodium chloride leads to cell lysis or other cellular responses due to osmotic stress in an inconsistent manner highly dependent on the exact cell culture conditions. This was also observed in previous study with sodium chloride (Dickinson et al., 2004) and could be attributed to a narrow concentration window between impairment of cellular function and cell lysis caused by osmotic stress. Because the gene expression provides a snapshot of a molecular interaction at the time the mRNA was isolated, the anchoring of gene expression data using a cytotoxicity parameter such as relative cell growth is challenging. This also points to a need to develop a better strategy for anchoring the gene expression data across compounds sets.
The mRNA expression changes of selected genes were measured using real-time PCR. Three laboratories used TaqMan card technology that allow for multiplexing assays and six laboratories used “Assay on demand” single gene PCR assays. The advantage of the TaqMan real-time PCR technology is that it utilizes quality controlled, validated PCR assays available world-wide. Although the cellular stress response to DNA damage caused by a prototypical agent such as gamma radiation displayed considerable complexity (Fornace et al., 1999), relatively small sets of genes were able to differentiate genotoxic mechanisms in other studies (for instance, Dickinson et al., 2004; Le Fevre et al., 2007). Therefore we selected a limited set of genes that were reported to alter their mRNA level in response to variety of stresses including DNA damage responses which should allow to at least partially distinguishing different genotoxic mechanisms. The resulting gene set thus consisted of genes involved in a variety cellular stress pathways such as DNA damage, oxidative stress response, apoptosis, cell cycle progression and inflammation (Table 1).
The analysis of the gene expression data revealed that 40 genes out of 47 selected resulted in measurable mRNA levels. Considering that the gene set was selected using literature data reported for a variety of cell lines and conditions, our experimental conditions showed a reasonable agreement with published data. The combined data set consisted of 134 individual gene expression profiles across the treatment conditions. We used a PCA to assess the reproducibility of data across experiments and participating laboratories (Figs. 2 and and3).3). The majority of the gene expression profiles clustered in a tight cluster. However, the data from three laboratories were outside this cluster indicating differences in the gene expression profiles in comparison to the majority of the data. These differences were likely caused by treating cells with different doses of model agents and possibly by minor differences in the experimental protocols. These outlier experiments were excluded from further analysis.
The detailed analysis of gene expression profiles revealed that the treatment with agents causing DNA damage such as the cross-linking agent cisplatin and the topoisomerase II poison etoposide, increased mRNA levels of the p53 target genes CDKN1A, MDM2, PHDA3, TP53I3, PPM1D, DDIT3, with the latter two being more clearly induced at the later time point. The very similar expression profile of GDF15 compared with p53 target genes can be reconciled with the fact that it may be induced by p53 via a p53-response element in response to DNA damage (Osada et al., 2007). Cisplatin and etoposide also decreased the expression of genes involved in cell cycle progression and survival including AURKA, PLK1, TOP2A, HIST1H2AC, INCENP, MCL1, and BCL2 at both early (4 h) and late (24 h) time points (Fig. 5). Furthermore, the increase in expression of these genes correlated with the cytotoxicity level at 24-h post-treatment. The increase of mRNA levels of p53 target genes is indicative of activation of the p53 pathway that together with growth arrest has been well characterized as a prototypical response to DNA damage (Zhan et al., 1993). This is also in accordance with previously published TGx studies investigating gene expression responses with prototypic DNA-damaging agents (Akerman et al., 2004; Dickinson et al., 2004; Islaih et al., 2004; Le Fevre et al., 2007). Although cisplatin directly interacts with DNA forming crosslinks whereas etoposide produces double strand DNA breaks via stabilization of DNA-topoisomerase cleavable complexes, both treatments provided a sufficient signal for activation of the p53 pathway. Interestingly, the gene expression profile of taxol-treated cells displayed induction of p53 target genes only 24-h post-treatment. The primary genotoxic effect of taxol is aneugenicity manifesting as chromosome loss or gain (Digue et al., 1999). The cytotoxic activity of taxol is at least partially due to its ability to induce apoptosis via activation of cell cycle dependent kinases leading to activation of caspases (Moos and Fitzpatrick, 1998) and then further to DNA fragmentation. In addition a role of the p53 pathway in taxol induced apoptosis has been reported (Bachman et al., 1998; Barboule et al., 1997). The activation of p53 pathway observed only at the later time point upon taxol treatment suggests that it is a secondary process associated with induction of apoptosis. In our study, the treatment of TK-6 cells with sodium chloride produced highly variable responses in both cytotoxicity and gene expression changes. This is in line with previously published data (Dickinson et al., 2004) and likely caused by the nature of osmotic stress that leads to cell lysis in a relative narrow concentration range. The fact that the treatment with agents producing DNA damage induced the prototypical p53-mediated DNA damage response early after treatment suggests that gene expression changes at early time points provide information related to the mechanism of damage. This is supported by the fact that many of these responses are of immediate early nature and robust gene expression changes can be detected within several hours and prior to the onset of appreciable toxicity (Amundson et al., 2005). In contrast, the gene expression profiles at later time points (24 h) likely mirror subsequent processes such as apoptosis inducing DNA strand breaks among other things and therefore the interpretation of such data in light of the mechanisms of the initial damage might be difficult.
In general, our study provides evidence for good reproducibility of detection of mRNA levels in cells in vitro using quantitative real-time PCR. The study design included nine participating laboratories across three continents. Although the individual experiments were done according to a common protocol, each laboratory used its own standard operating procedures for cell culture and mRNA level analysis. This experimental set-up provides a real life example of applying real-time PCR for gene expression analysis to test the response to a certain compound family in an in vitro model. The principal component analysis led to exclusion of outlier data sets on the basis of dissimilar gene expression profiles compared with the data that clustered more tightly together. This highlights a need to develop approaches for more meaningful phenotypic anchoring of the data than possible by simple cytotoxicity measurement.
In summary, we have shown that a set of relevant genes is capable of differentiating compounds that cause DNA adducts (cis platin) or DNA double strand breaks (etoposide) from a mitotic spindle poison (taxol) and a compound causing chromosome damage as a consequence of cytotoxicity (sodium chloride). We have shown that gene expression profiles at early time points are most likely to provide information relevant to mechanisms of genotoxic damage. We have also shown a need for developing better methods for phenotypic anchoring of gene expression profiles. Our study provides evidence that the gene expression profile technology produces reproducible data across laboratories. Although more compounds need to be tested our study further confirms the potential value of TGx analysis for investigation of genotoxic mechanisms.
Contributions by J. Fostel were supported by the Division of Intramural research of the National Institute of Environmental Health Science, under contract (HHSN273200700046U).
We would like to thank Nicole Hellwig and Kerstin Albrecht for technical assistance and Grazyna Wasinska-Kempka for supporting the cell culture experiments conducted at Bayer Healthcare AG.
We would like to thank Monika Haiker for technical assistance at Hoffman-La Roche, Inc.
We would like to express our special gratitude to Josiane Bringel for her invaluable technical assistance throughout the whole study and Marc DeCristofaro for crucial contribution to selection of the gene list and setting up the TaqMan Cards design and Deborah Garcia for continuing cell culture support at Novartis Pharma AG. Furthermore we would like to deeply thank Manoli Flor and Jean-Pierre Marchandeau for performing the real-time PCR experiments at Sanofi Aventis.