Integrating transcriptomics and metabonomics to unravel modes-of-action of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in HepG2 cells

Presently, where the current testing strategy for identifying carcinogenic properties of novel chemicals has been designed to effectively capture genotoxic carcinogens, for the identification of important class of non-genotoxic carcinogens no suitable short-term test model, neither in vivo nor in vitro, is available yet. Till date, discovery of non-genotoxic carcinogenicity is completely depending on the two-year chronic rodent bioassays. These in vivo assays not only require a large number of animals, are costly, laborious, and time-consuming, but also cause many false-positive responses and their relevance towards human is questionable. Therefore, alternative assays are needed.

Transcriptomics has been well established in toxicogenomics research [1] and initial studies have indicated that gene expression profiling in both in vivo and in vitro systems is of value for predicting genotoxicity and carcinogenicity. As the liver is the major target organ for non-genotoxic carcinogens, short term in vivo studies have been targeted towards rat liver and these have generated promising gene expression profiles capable of reliably predicting non-genotoxic carcinogenesis [2–4]. Furthermore, promising results have been achieved in our previous studies discriminating genotoxic carcinogens from non-genotoxic carcinogens in the human liver cell line HepG2 [5, 6]. The genes identified in these short term in vivo and HepG2 studies are involved in processes related to cell cycle, cell signaling, (anti-)apoptosis, oxidative DNA/protein damage response, proliferation, cancer, transcription and protein synthesis. In addition, results of in vivo rodent studies, in which a single non-genotoxic carcinogen is examined, e.g. WY-14,643 [7] and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) [8–10], indicated changes in amino acid metabolism, lipid metabolism and oxidative stress mediated through ligand-activated receptors (e.g. aryl hydrocarbon receptor (AHR), peroxisome proliferator activated receptor (PPAR), pregnane × receptor (PXR; NR1I2), constitutive androstane receptor (CAR; NR1I3)).

Because reported transcriptomic changes induced by TCDD and other non-genotoxic carcinogens in short term in vivo studies, referred to the possibility that the endogenous metabolism is affected, we decided to particularly explore the added value of metabonomics analysis in identifying molecular responses in vitro caused by non-genotoxic carcinogens. Metabonomics is defined as the study of metabolic responses to drugs, environmental changes and diseases http://www.medterms.com/script/main/art.asp?articlekey=38634. It combines the application of analytical technologies (e.g. nuclear magnetic resonance spectroscopy (NMR), mass spectrometry (MS)) for metabolic profiling with multivariate statistical methods. In general, metabonomics is considered to be complementary to transcriptomics and proteomics [11]. However, this technique shows some distinct features in comparison to these other 'omics' technologies: a) It analyses the end point in a series of changes following exposure to a chemical compound; b) many of the metabolites have a known function; c) changes are detectable intracellularly as well as in extracellular fluids [12]. Various studies illustrate the applications of metabonomics in e.g. toxic risk assessment, biomarker discovery, and studies on toxic mechanisms [13, 14], but so far have been performed in vivo.

Such in vivo studies have also demonstrated that the integration of different 'omics' technologies offers a complementary insight into cellular responses to toxic processes [15–19]. For example, in the study of Coen et al. [15] the integration of transcriptomics and metabonomics provided new insight into the toxic consequences for the well-studied therapeutic agent acetaminophen in mice. In adition, Heijne et al. [17] showed that the cross-omics analysis provides new insight in the mechanisms of hepatotoxicity in rats exposed to bromobenzene, and also appears to be more sensitive than conventional techniques for indicating induced liver injury. We, therefore, hypothesize that combining of transcriptomics and metabonomics data improves the understanding of molecular responses to non-genotoxic carcinogens also in in vitro cellular models, thus potentially contributing to developing cellular assays which are capable of predicting non-genotoxic carcinogenesis in vivo[20]. To test this hypothesis, the human hepatocarcinoma cell line HepG2 was exposed to the well-known environmental pollutant TCDD.

TCDD is considered a non-genotoxic human carcinogen (IARC group 1 classification) that activates AHR and induces a broad spectrum of toxic responses including death, immunosuppression, carcinogenicity, and impaired reproduction and development [21, 22]. TCDD has a long half-life of 5-10 years in human beings as a result of its high lipophilicity and shows little or no metabolism [23]. The past decades, its effect on the endogenous metabolism in vivo as well as in vitro has been explored, showing changes in e.g. amino acid metabolism, lipid metabolism and polyamine synthesis [8–10, 24–29].

HepG2 was chosen as in vitro cell model since it has frequently been applied in toxicogenomics studies [5, 6, 30–32]. These cells possess a liver-like enzyme pattern [33], expressing many phase I, II and III drug metabolizing enzymes and are competent for regulating these enzymes by ligand-activated transcription factors [34–38]. Furthermore, its metabolizing capacities are evident [39], albeit that metabolic activation does not play a role for TCDD. Furthermore, we have previously observed changes in expression of genes related to the amino acid and lipid metabolism as well as other metabolic processes especially after 48 h of exposure to TCDD in HepG2 [40], which underlines the relevance of this hepatocellular model for the present study.

For integrating read-outs from different 'omics' platforms, i.e. transcriptomics, proteomics and metabonomics, in general, there is no lack of tools that are capable of visualizing data in a biological network context [43]. However, tools that directly integrate e.g. transcriptomics and metabonomics data in a pathway ranking analysis are available only to a limited extent. In this study, MetaCore was used for performing such an integrative analysis based on a combined list of transcriptomics and metabonomics data from TCDD-exposed HepG2 cells. We, thereby, hypothesize that combining of these 'omics' data will improve the understanding of molecular responses caused by TCDD in this in vitro heptocellular model. This cross-omics analysis demonstrated specific involvement of pathways related to receptor-mediated processes, cell signaling and endogenous metabolism.

Although different mechanisms may be involved in the carcinogenic action of non-genotoxic compounds, receptor-mediated processes are definitely eligible as possible mechanisms for non-genotoxic carcinogenicity of TCDD [44]. As an AHR-ligand, TCDD induces the transcription of CYP1A1, CYP1B1 and NQO1, which are all significantly up-regulated in this study, and this may lead to the generation of reactive oxygen species (ROS; includes superoxide anion and hydrogen peroxide) and thereby oxidative stress/DNA damage [45–47]. The latter was identified as a specific process for non-genotoxic carcinogenicity [47]. In the study by Knerr et al. [45] a slight increase of oxidative DNA damage was measured after incubation of HepG2 cells with TCDD. Furthermore, the increase of GSH and decrease of ATP as found in our study are indicative of oxidative stress. GSH participates in the regulated defense against oxidative stress [48, 49], which can be induced by TCDD in vivo[46, 50, 51] as well as in vitro[52, 53]. ATP was previously shown to be decreased whereas oxidative stress increased upon exposure to TCDD in mice [50, 51, 54].

Besides the AHR-mediated processes, several GPCR signaling pathway maps as well as other signal transduction pathway maps were significantly modulated in our study (Table 3). These maps include many of the 16 pathway maps uniquely identified by the integrated 'omics' analysis, e.g. "G-Proteins mediated regulation MARK-ERK signaling", "Beta-adrenergic receptors regulation of ERK", "G-Protein alpha-i signaling cascades", "Rap2A regulation pathway" and "PDGF signaling via MAPK cascades". Some of the underlying genes of these pathways belong to the RAS superfamily, a group of small GTPases of which their proteins are activated in a significant fraction of tumors [55, 56]. Therefore, these genes could be of importance in the mechanism for non-genotoxic carcinogenicity of TCDD. For example, muscle RAS oncogene homolog (MRAS), related RAS viral (r-ras) oncogene homolog (RRAS), v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) and v-ral simian leukemia viral oncogene homolog B (RALB) were found to be significantly up-regulated in our study. Members of the RAS superfamily are positively regulated by guanine nucleotide exchange factors (GEFs) [55, 56], e.g. son of sevenless (SOS) 1 and ral guanine nucleotide dissociation stimulator-like 1 (RGL1), which are both significantly up-regulated in the present study (Figure 1). GEF-activated RAS interact with downstream effectors, which regulate signaling networks that control gene expression and regulation of cell proliferation, growth arrest, senescence, differentiation, apoptosis and survival [55, 56]. As such, SOS1 is important in the cell growth-regulatory MAPK/ERK pathway through the activation of RAS family members [56]. In primary rat hepatocytes SOS has already been shown to be influenced by TCDD [57] while recently, Pierre et al. [58] demonstrated that SOS1 is not only controlled by GPCRs, but also by AHR in TCDD-exposed HepG2 (Figure 1). In particular, Pierre et al. [58] found an AHR-dependent increase of SOS1 at the mRNA and protein level in TCDD-exposed HepG2.

Other prominent pathway maps significantly affected in HepG2 by TCDD exposure appeared involved in metabolic processes, mainly relating to amino acid and lipid metabolism but also vitamin and cofactor metabolism and metabolism of mediators (Table 3).

From the set of 20 essential amino acids at least 11 were significantly decreased after exposure to TCDD (Table 2). In addition, many genes involved in the metabolism of these amino acids were down-regulated. For example, glycine and creatine appeared decreased and GATM down-regulated which may affect the synthesis of creatine, an anti-oxidant against ROS, via two reactions [59]. In the first reaction, ornithine and guanidinoacetic acid are formed from glycine and arginine by glycine amidinotransferase (GATM). This reaction is the rate-limiting step in creatine biosynthesis and part of the polyamine metabolism (Figure 2). In the second reaction, catabolyzed by guanidinoacetate N-methyltransferase (GAMT), creatine is produced from guanidinoacetic acid and S-adenosyl-L-methionine.

Another deregulated route involves the contribution of glycine, aspartate and glutamine in the biosynthesis of nucleotides http://themedicalbiochemistrypage.org/nucleotide-metabolism.html. Levels of all three amino acids are decreased upon TCDD-exposure and the gene in the initial step of the purine metabolism, phosphoribosyl pyrophosphate amidotransferase (PPAT) is down-regulated.

Furthermore, aminoacyl-tRNA biosynthesis is down-regulated with 12 amino acids decreased and 6 genes down-regulated, thereby most likely inhibiting global protein synthesis. In previous rodent studies [8] it was suggested that genes involved in glutamate, cysteine and glycine metabolism, may be down-regulated to conserve the building blocks of GSH, whereas the genes involved in the synthesis of GSH are up-regulated. Results obtained in our HepG2 model, confirm this suggestion as both GSH and glutathione disulfide (GSSG) are increased upon TCDD-challenge, with a consistent increase in gene-expression of glutamate-cysteine ligase (GCL) which catalyses the production of γ-glutamyl-cysteine, a precursor of GSH (Figure 3).

A further process that is decreased in HepG2 after exposure to TCDD is the production of N-acetylaspartate (NAA) from aspartate and acetyl-CoA. This reaction is catalyzed by aspartate N-acetyltransferase, which recently has been identified as NAT8L [60]. This gene is down-regulated (log2 ratio of -0.83), but not significantly (p value = 0.08). NAA is a brain specific protein, but is also found in liver [60, 61]. It acts as an acetyl donor for lipid metabolism. For that, NAA is converted to aspartate and acetate. The acetate released in this process, may serve as a precursor in the synthesis of fatty acids by its conversion to acetyl-CoA. The decrease of NAA corresponds well with the decrease of most saturated and unsaturated fatty acids in HepG2 exposed to TCDD. In addition, the gene involved in the first reaction in the fatty acid synthesis from acetyl CoA, i.e. acetyl CoA carboxylase (ACACA), is significantly down-regulated.

Additionally, the sulphur containing amino acid taurine was found to be increased in TCDD-exposed HepG2. Taurine plays an important role in several physiological functions, such as central nervous system neuromodulation, cardiovascular effects, endocrine and metabolic effects and antioxidant/detoxifying activity. It is mainly synthesized in liver and brain, either from cysteine or from methionine, which is first converted to cysteine [62]. However, the genes involved in taurine production, were down-regulated or not changed at all in TCDD-exposed HepG2. Furthermore, taurine is involved in the conjugation of bile acids by bile acid CoA:amino acid N-acyltransferase (BAAT) [63]. The expression of BAAT in this study was down-regulated. These findings indicate that the increase of taurine in HepG2 exposed to TCDD is not the result from an increased taurine production, but rather from a decreased metabolism. The study of Yanagita et al. [64] showed that an increase of taurine in HepG2 results in a decrease of triglycerides and cholesteryl ester. In the same study it was suggested that acyl-CoA:cholesterol acyltransferases or sterol O-acyltransferases (SOAT1 and SOAT2), which catalyzes the production of choleteryl ester from cholesterol, is inhibited. The decrease of the triglycerides and cholesteryl ester was also observed in this study and SOAT2 was down-regulated. In another study of taurine-exposed HepG2 [65] microarray data analysis showed taurine-mediated gene expression changes in processes involved in e.g. amino acid metabolism, fatty acid and lipid metabolism, regulation of transcription, immune response, cell cycle, apoptosis and DNA repair. However, from the results of this study it is unclear in what way taurine plays a role in these processes.

From the maps identified by cross-omics analysis to be involved in metabolism of mediators, the polyamine metabolism map also appeared the number 1 ranked pathway map in the transcriptomics gene list and furthermore, is linked to the GABA metabolism [66] (Figure 2). Both polyamine and GABA metabolism have a function in cancer-related processes, whereby an increase of polyamines and/or GABA were observed in various cancer types [67–71]. Our findings suggest that TCDD-exposure in HepG2 decreases polyamine metabolism. In the polyamine metabolism putrescine is the first polyamine produced from which spermidine and subsequently spermine are produced [67, 68]. The production of putrescine from ornithine is the rate-limiting reaction. This reaction is catalyzed by ornithine decarboxylase (ODC; DCOR). ODC was significantly down-regulated in this study, as well as arginase 1 and 2 (ARG1, ARG2) or GATM that catalyze the preceding reaction in which ornithine is produced from arginine or from glycine and arginine, respectively. Spermidine and N1-acetylspermidine, an N-acetylated polyamine, were found to be significantly decreased. N1-acetylspermidine is an intermediate for the recycling of spermidine into putrescine [67, 68]. These in vitro results correspond well with the in vivo study of Thomas et al. [28], who showed a decrease in the concentration of polyamines in mouse liver after exposure to TCDD. In addition, depletion of polyamines in HepG2 is shown to prevent apoptosis [66]. In our study, exposure of the HepG2 cells to TCDD did not affect apoptosis levels, which were low (< 1%; data not shown).

With regard to the GABA metabolism our results indicated this metabolism to be enhanced. In the GABA metabolism, putrescine is used to form GABA through 4-aminobutanal [66]. The latter reaction is catalyzed by several aldehyde dehydrogenases of which ALDH1A1 (or AL1A1), ALDH1B1 (or AL1B1) and ALDH2 had a significantly up-regulated gene expression after TCDD exposure in HepG2. Furthermore, GABA can be formed from glutamate by glutamate decarboxylase 1 or 2 (GAD1, GAD2), of which GAD1 also has significantly up-regulated gene expression in this study.

The major result from this study is that TCDD-mediated 'omics' responses in HepG2 are comparable with data from in vivo studies on transcriptomics responses induced by non-genotoxic carcinogens in rodent liver. The integrated 'omics' analysis resulted in the identification of unique pathway maps involved in receptor-mediated mechanisms. Such mechanisms have been identified as possible mechanisms for non-genotoxic carcinogenicity in vivo[44]; in particular up-regulation of SOS1 seems to play an important role. This resemblance with the in vivo situation opens a venue for further developing HepG2 as an in vitro model for non-genotoxic carcinogenicity, by applying cross-omics analysis.

Furthermore, integration of transcriptomics and metabonomics data provided novel insights into response pathways to TCDD exposure in HepG2, in particular related to changes in processes involved in amino acid and lipid metabolism as well as glutathione metabolism From these metabolic processes the metabolites taurine, creatine and NAA as well as the genes involved in their metabolism seem to be important for TCDD-mediated carcinogenicity as the elevation of taurine and absence/decrease of creatine and NAA is also observed in human brain and liver tumors [72–76].

Where novel 'omics' technologies such as microRNA analysis and epigenomics, are becoming increasingly available, integrated data analysis seems to represent the way forward in our quest for better understanding human health risks in relation to exposure to toxic agents.