Profiles of Metabolites and Gene Expression in Rats with Chemically Induced Hepatic Necrosis

Abstract

This study investigated whether integrated analysis of transcriptomics and metabolomics data increased the sensitivity of detection and provided new insight in the mechanisms of hepatotoxicity. Metabolite levels in plasma or urine were analyzed in relation to changes in hepatic gene expression in rats that received bromobenzene to induce acute hepatic centrilobular necrosis. Bromobenzene-induced lesions were only observed after treatment with the highest of 3 dose levels. Multivariate statistical analysis showed that metabolite profiles of blood plasma were largely different from controls when the rats were treated with bromobenzene, also at doses that did not elicit histopathological changes. Changes in levels of genes and metabolites were related to the degree of necrosis, providing putative novel markers of hepatotoxicity. Levels of endogenous metabolites like alanine, lactate, tyrosine and dimethylglycine differed in plasma from treated and control rats. The metabolite profiles of urine were found to be reflective of the exposure levels. This integrated analysis of hepatic transcriptomics and plasma metabolomics was able to more sensitively detect changes related to hepatotoxicity and discover novel markers. The relation between gene expression and metabolite levels was explored and additional insight in the role of various biological pathways in bromobenzene-induced hepatic necrosis was obtained, including the involvement of apoptosis and changes in glycolysis and amino acid metabolism.

The complete Table 2 is available as a supplemental file online at http://taylorandfrancis.metapress.com/openurlasp?genre=journal&issn=0192-6233. To access the file, click on the issue link for 33(4), then select this article. A download option appears at the bottom of this abstract. In order to access the full article online, you must either have an individual subscription or a member subscription accessed through www.toxpath.org.

Introduction

Previous toxicogenomic studies have shown that both large-scale measurement of gene expression (transcriptomics) and metabolite levels (metabolomics) complement the current methods to identify and discriminate different types of toxicity. To date, most toxicogenomic studies concentrated on hepatic toxicity. Transcriptomics using DNA-microarrays enabled the discrimination of responses in animals exposed to different classes of hepatotoxicants (Burczynski et al., 2000; Bulera et al., 2001; Waring et al., 2001; Hamadeh et al., 2002).

Metabolite profiling by NMR combined with pattern recognition techniques (metabolomics) have been used to classify urine samples of rats treated with either a liver or a kidney toxicant (Holmes et al., 1998). Others analyzed metabolites in liver, plasma, and urine of rats treated with the model hepatotoxicant alpha-naphthylisothiocynanate (ANIT) (Waters et al., 2002). Urine profiles were analyzed in time upon single dosage of ANIT, galactosamine, and butylated hydroxytoluene (Beckwith-Hall et al., 1998). Time-related differences in metabolite contents were related to the stage of the lesions, and specific changes in metabolite levels were identified for each compound.

In contrast to most routinely applied methods, the new technologies enabled investigation of molecular mechanisms that lead to toxicity. Gene expression changes influence biochemical reactions and metabolite levels are determined by those biochemical reactions. Therefore, a systems toxicology approach that combines transcriptomics, proteomics and/or metabolomics may be helpful to collect complementary information to analyze toxicity in a systematic manner.

Few experiments have integrated results from transcriptomics and metabolomics. Recently, Coen and colleagues (2004) reported transcriptomics and metabolomics analyses in mice treated with acetaminophen (paracetamol). This study demonstrated that separate analysis of gene expression and metabolite profiles provided complementary insights in APAP-induced hepatic effects. In earlier studies in our laboratory, we used transcriptomics and proteomics in parallel to analyze hepatotoxicity induced by bromobenzene (BB). Bromobenzene is a well-studied model toxicant that causes necrosis in the liver (centrilobular) and kidney. Hepatic bio-transformation and toxicity of BB in rats have been reported in detail (Thor et al., 1981; Monks et al., 1982; Casini et al., 1985; Lau and Monks, 1988; Miller et al., 1990). Because the liver is the target for toxicity induced by many compounds including bulk chemicals, drugs, and food ingredients, the characteristics of the response induced by BB could also be helpful in understanding hepatotoxicity induced by a variety of xenobiotics.

Transcriptomics and proteomics analyses of hepatotoxicity were evaluated 24 hours after a single ip dose of BB (Heijne et al., 2003). A new study was designed to determine the acute hepatotoxic effects at the gene expression level in time, after oral dosage of several concentrations of BB. Hepatic necrosis was observed only at the high-dose level after 24 hours, though gene expression changes characteristic for BB exposure were observed at 2.5 times lower-dose level. The expression of several genes was detected to changed at 10 times lower-dose levels. Expression of several genes was found to change 6 hours after dosage. Genes that had statistically significant differential expression after BB were involved in processes such as drug metabolism, oxidative stress, GSH synthesis, and the acute phase response (Heijne et al., 2004).

The aim of the research presented here was to investigate whether an integrated “systems toxicology”—analysis with transcriptomics and metabolomics could increase the sensitivity of detection of hepatotoxicity and provide new insights in mechanisms of chemically induced hepatotoxicity. This approach could identify relationships between gene expression changes and altered metabolite levels after exposure to a toxicant. For this, NMR-based metabolite profiles of plasma and urine samples, collected from the study described before (Heijne et al., 2004) were combined with the transcriptomics data of that same study. With metabolomics, concentrations of both endogenous as well as exogenous metabolites were measured. The endogenous metabolites related to the hepatotoxic effects could potentially serve as biomarkers of effect. In addition, the exogenous bromobenzene-derived metabolites in urine and plasma could be useful as markers of exposure. Effects on gene expression and metabolite levels were combined with preexisting biochemical knowledge to assess the relevance of the changes in the context of biological pathways involved in bromobenzene-induced hepatic necrosis.

Materials and Methods

Urine and plasma samples were collected from a study of which the transcriptomics and toxicity data were reported earlier (Heijne et al., 2004). Briefly, 3 doses of bromobenzene (0.5, 2.0, and 5.0 mmol/kg body weight, dissolved in corn oil, 40% v/v) were administered to male Wistar rats by oral gavage. Animals were kept under controlled conditions, and the welfare of the animals was maintained in accordance with the general principles governing the use of animals in toxicity experiments of the European Communities (Directive 86/609/EEC) and Dutch legislation (The Experiments on Animals Act, 1997). Nine rats per dose group were treated with BB or corn oil, while an additional group was not treated. Three rats per group were sacrificed after 6, 24, and 48 hours and blood and livers were collected. Urine was collected for metabolomics between dosing and sacrifice for the 6-hour group, and during the last 16 hours before sacrifice for the 24- and 48-hour groups. During the time urine was collected, rats received water ad libitum, but no food. Histopathological abnormalities in the liver were determined and given a score within the range of 0–10, with 0 in the abscence of necrotic cells, below 3 when only mononuclear aggregates and/or necrotic hepatocytes were found, and a score of 3–10 for centrilobular necrosis, depending on the severity (see Table 1).

Transcriptomics

cDNA microarray preparation and hybridization was described previously (Heijne et al., 2003, 2004). A reference RNA was used, and hybridizations were replicated with swapped fluorophore incorporation (Cy3 and Cy5) in the sample and reference RNA. After quality filtering, LOWESS normalization and log(base 2) transformation, a set of about 2700 cDNAs was obtained. In the present study, we required a correlation higher than 0.6 between the duplicate sets of dye-swap measurements, resulting in 400 genes in the data set.

NMR analysis

NMR spectra of urine of individual animals were recorded in triplicate, according to (Lamers et al., 2003). Plasma samples were deproteinized by filtration. Filters with a cutoff of 10 kDa (Microcon YM-10, Millipore) were spin-rinsed with 0.5 ml of 0.05 M NaOH followed by 2 × 0.5 ml de-ionized water to avoid contamination of the ultrafiltrate with glycerin. Centrifugation (1 hour at 10,000 rpm) of 0.5 ml plasma over a filter was followed by the centrifugation (1 hour at 10,000 rpm) of 0.5 ml deionized water. Filtrates were freeze-dried and reconstituted in 750 μl sodium phosphate buffer (pH 6.0, made up with D₂O) containing 1 mM sodium trimethylsilyl-[2,2,3,3,-2H4]-1-propionate (TMSP) as an internal standard. NMR spectra were recorded in a fully automated manner on a Varian UNITY 400 MHz spectrometer (Palo Alto, CA, USA) according to (Lamers et al., 2003).

Data preprocessing and multivariate data analysis

The NMR data file was imported into Winlin (V1.12, TNO, Zeist, The Netherlands). Minor variations from comparable signals in different NMR spectra were adjusted and aligned without loss of resolution. The intensities of signals present in each NMR spectrum were normalized, so that the sum of all intensities was equal to 1. This data set was imported into Matlab (Version 6.5, The MathWorks Inc., Natick, MA, USA) together with the transcriptomics data in order to obtain an overall data set. This set was used for preprocessing and multivariate data analysis. The data matrix was centered across time and dose. The sum of squares per variable over time and dose was scaled to 1, and PCA was performed (PLS tool-box Version 3.0, Eigenvector Research Inc., Manson, WA, USA). PCA is a multivariate statistical analysis that reduces the many dimensions of a dataset to few dimensions that describe the majority of the variance. PCA was performed on the overall data set to visualize differences in rats on the basis of profiles of gene expression and metabolite levels and assess relationships between gene expression changes and metabolite levels. The contribution of each variable to the trend observed in the score plot was calculated.

PCA was also performed on plasma and urine NMR data separately. When score plots revealed differences between groups, the contributions of the original NMR signals to these difference between treated and control were displayed in a factor spectrum. Metabolites were identified using an in-house reference database, and chemical shifts of characteristic metabolites in the NMR spectra are listed in Table 3.

Results

Rats exposed to bromobenzene developed hepatic necrosis 24 hours after dosing with the high concentration. In parallel, hepatic gene transcription and profiles of plasma and urine metabolites were analyzed.

Toxicological examinations

No macroscopic abnormalities of the liver or other organs were observed in any of the rats sacrificed 6 hours after dosage. Histopathology of liver tissue showed no abnormalities in the controls and low-dose rats. Only in rats that received the high concentration of BB did livers have a patchy appearance and gross lesions after 24 hours with focal discoloration after 48 hours. Centrilobular necrosis was present in livers of all high dose rats, with interindividual variation in the degree of response (Table 1). Plasma levels of ASAT, ALAT, and bilirubin were markedly elevated, also with interindividual variation. To correlate the conventional markers of hepatotoxicity with the degree of necrosis in the individual rats, a semiquantitative score was developed for the hepatocellular necrosis ranging from 0 (no effects) to 10 (very severe centrilobular necrosis) (Table 1). This score was also used to correlate gene expression levels to necrosis. Figure 1 depicts the correlation of ASAT, ALAT, bilirubin, and the relative liver weight with the observed degree of hepatocellular damage.

BB also significantly decreased plasma levels of glucose at the mid and high dose, after 24 and 48 hours. Cholesterol (n.s.) and phospholipid levels increased by high BB treatment at all time points. Hepatic GSH levels, which play a pivotal role in the hepatotoxicity induced by BB, were slightly decreased 6 hours after administration of BB. The mid and high dose depleted GSH levels to ~25% of control levels. After 24 hours, GSH levels were nearly restored.

Transcriptomics analysis and parallel metabolite profiling

BB elicited specific changes in expression of many rat liver genes, as reported before (Heijne et al., 2004). In the present study, the transcriptomics measurements were combined with the data obtained by NMR, describing the metabolite contents of plasma. Consensus PCA (Smilde et al., 2003) was performed using both types of data in 1 integrated analysis, and results are shown in Figure 2. This plot indicates that the samples from the high and mid dose groups, collected after 24 and 48 hours were distant from the others, having lower PC1 scores. Most distinct from all the other samples were the samples from rats #84, #82, and #90, that received a high dose of BB. Microscopic examination revealed (very) severe hepatic centrilobular necrosis in those rats. Profiles of rats #80, #86, and #88 were less distinct from the controls. Correspondingly, moderate centrilobular necrosis was observed in rat #88, and (very) slight necrosis in rats #86 and #80. The profiles of the rats treated with a mid dose of BB were distinct from the controls after 24 hours. Routine markers were not able to indicate hepatotoxicity in those rats. After 48 hours, rats treated with the mid dose were not distinct from controls. Samples from rats treated with the low dose of BB were not distinguishable from the controls, after 24 or 48 hours. All samples collected after 6 hours were distinct from the other time points in the lower right corner of the plot.

Genes and metabolites were sorted according to their contribution to the observed trend, reflecting the degree of hepatic necrosis (Full Table 2 available as supplementary material online—see end of Abstract for URL). Tables 2A and 2B list the genes and metabolites with the highest and lowest scores, that therefore putatively correlate with the degree of hepatotoxicity. The levels of gene expression and metabolites listed in Table 2A are positively, and in Table 2B negatively correlated to the degree of necrosis. Many genes with a significant contribution to pattern differences in the PCA were identified to be in- or decreased by BB with high significance in univariate statistical tests, and the toxicological relevance of these changes was discussed before (Heijne et al., 2004).

Genes with high scores in the integrated analysis include structure and cytoskeleton-related genes (beta actin, weakly similar to pervin, tubulin), many ribosomal subunits, and other factors involved in protein synthesis (e.g., nucleophosmin). Also oxidative stress induced genes (HO-1, TIMP1, peroxiredoxin1, ferritins), hepatic acute phase response genes (orosomucoid 1, fibrinogen gamma) and enzymes involved in glucose metabolism (GAPDH, phosphoglycerate mutase 1, aldolase A) have high rankings. Drug metabolizing enzymes such as Ephx1, AFAR, GSTA, and aldo-keto reductases, likely involved in the hepatic biotransformation of bromobenzene, appeared in the upper part of the ranking. Several cell cycle and apoptosis related genes (BCL2-related protein A1, PCNA, p53, p21 (WAF), EST, highly similar to p53-regulated PA29-T2, cyclin G1) had coordinately increased expression. High ranked genes with others functions include casein kinase II, VL30 element, and RAN. Plasma metabolites with a high score in the analysis include acetate, choline, phenylalanine, and several metabolites that could not be characterized using the reference database.

Genes with low scores include hepatic acute phase response genes such as alpha-1-inhibitor, serine protease inhibitor, fibrinogen beta, complement components, drug metabolizing enzymes such as CYPs, aldehyde dehydrogenases, FMO3, enzymes involved in fatty acid and cholesterol metabolism (HMG-CoA synthase, LCAT, STAR, fatty acid CoA ligase, acyl CoA dehydrogenases), and glucose metabolism (G6pt1, alanineglyoxylate aminotransferase) Many genes with other functions, such as asialoglycoprotein receptor 2, Cathepsin S, and dimethylglycine dehydrogenase had a low score, indicating that they were decreased—compared to the controls. Plasma metabolites with a low score in the analysis include dimethylglycine, tyrosine, and glucose.

Gene expression markers

The correlation between the level of gene expression and the degree of necrosis in the individual rats was calculated (Table 2). Figure 3, upper panels illustrate expression of ESTs highly similar to actin and pervin, and orosomucoid 1 in relation to hepatic necrosis. Expression levels of asialoglycoprotein receptor 2 and lecithin-cholesterol acyltransferase (LCAT) decreased in concordance with the degree of hepatic damage, shown in Figure 3, lower panels. In total, 14 genes were found with a positive correlation between 0.80 and 0.89, which was the highest coefficient. The correlation of the average expression level of these 14 genes with necrosis was 0.969. In parallel, 20 negatively correlated genes were found, with an individual correlation to necrosis varying from −0.80 to −0.90. The correlation of the average gene expression of these 20 genes with necrosis was −0.959. These observations suggest that valuable markers of hepatocellular necrosis may be formed by a combination of genes’ expressions.

Time- and dose-dependent changes in plasma metabolites

The plasma NMR data were also analyzed separately to visualize time and dose specific changes in metabolite contents (Figure 4). The factor spectra show that lipid levels increased after treatment with a high or a mid dose of BB. Increased plasma phospholipid levels were also found with clinical chemistry after treatment with the high dose, but not after treatment with the mid or low dose. Levels of glucose were greater 6 hours after a high dose of BB, but lower after 24 and 48 hours, as seen both with clinical chemistry measurements and metabolomics analysis. NMR of plasma showed higher levels of creatine and/or creatinine in BB-treated rats, though clinical chemistry could not reveal significant changes in creatinine. The levels of tyrosine were lower 6 and 24 hours after BB, while higher after 48 hours. Methionine, alanine and lactate levels in plasma of BB-treated rats were lower 6 hours after dosage but higher 24 and 48 hours after dosage. Dimethylglycine and taurin levels were increased compared to controls 6 hours after the BB treatment, and decreased after 24 hours. Choline levels were decreased after treatment to mid or high dose of BB.

Profiles of urine metabolites

The urine metabolite contents were analyzed by NMR profiling and PCA. Analysis per time point demonstrated the effects of the BB treatment on the composition of the urine. All rat urines collected during the first 6 hours could be distinguished from each other, dependent on the exposure dose. By 48 hours after dosage, urine profiles of rats treated with the high concentration of BB could still be distinguished from controls. Factor spectra were constructed to determine which NMR signals most significantly differed between the high-dose and control group. Figure 4C shows the factor spectrum for rat urine collected during the 6 hours after dosage. The identity of several peaks was established using a reference database. The factor spectra revealed the marked presence of BB-derived metabolites such as bromphenols, bromcatechols, and quinones in urine. It was not possible to discriminate and identify these metabolites. Markedly elevated levels of mercapturic acids, derived from GSH-conjugates, were observed after treatment. Methionine levels in urine were higher in the treated rats compared to controls. Formate levels increased after 24 hours in the treated rats, and elevated levels were observed of urocanate and (methyl)histidine, as well as decreased levels of nicotinate, hippurate, phenylalanine/tyrosine, and glucose/fructose.

Discussion

This study presents one of the first integrated toxicogenomics studies to analyze acute hepatotoxicity at the transcriptome and metabolite level in a time- and dose-dependent manner. The transcriptomics and metabolomics analyses were integrated to increase the sensitivity of detection of hepatotoxicity and to assess relationships between gene expression and metabolite level changes.

When rats were treated with BB, hepatic centrilobular necrosis was observed after 24 hours at the high, but not at lower doses. The interindividual response varied from very slight to very severe hepatic centrilobular necrosis, which was supported by clinical chemistry parameters. In addition to these routine toxicological observations, the hepatic transcriptomics and plasma metabolomics showed that differences between molecular profiles were dependent on the dose and time after dosage. Molecular profiles from the 6 hr time point were distinguishable from other time points. BB treatment at the high dose resulted in highly distinct profiles, while the mid dose altered the profiles up to 24 hours after dosage. At this dose level, conventional signs of hepatotoxicity were not observed. Thus, the toxicogenomics approach was more sensitive than conventional techniques. However, it was not possible to distinguish samples treated with the low dose from the controls, even after combining transcriptomics and metabolite profiling.

Gene expression and metabolite changes were identified in correlation with the degree of hepatic necrosis. These genes and metabolites may be used to objectively quantify the degree of necrosis. If these markers prove to be predictive at earlier time points or lower dose levels, they may improve the current detection of hepatotoxicity. The observed changes could be explained from a mechanistic point of view. The induction of cytoskeleton constituents (actin and pervin amongst others) with the degree of necrosis indicates remodeling of the cytoskeleton. Presumably, necrosis and repair occur simultaneously in different liver cells, but the experiments using whole liver do not allow to localize the events. The negative correlation of genes such as alpha-1-inhibitor and serine protease inhibitor is likely related to the acute phase response, involving a shift in protein production in the liver.

By taking the average expression level of a set of positively or negatively correlated genes, a higher coefficient of correlation with the degree of necrosis was obtained. This also suggests that a model consisting of a combination of positively and negatively correlated genes and metabolites could be to used to more accurately assess degree of hepatocellular necrosis.

Xenobiotic compounds such as BB are degraded into many metabolites, and ultimately excreted in urine. The xenobiotic-derived metabolites could be suitable to monitor exposure and to elucidate routes of biotransformation. Levels of endogenous metabolites that changed after treatment may be putative biomarkers of toxicity, and could also help to identify the mechanism of hepatotoxicity.

In agreement with the levels of exposure to BB, the metabolite contents of urines of different rats were found to differ. Especially shortly after dosage, many water-soluble BB-metabolites were found, including bromphenols, -catechols, and -quinones, and mercapturic acids. The identification of all BB-derived metabolites (e.g., signals that were detected around 6 ppm in the spectra) would require further analysis with techniques like liquid chromatography and mass spectrometry (LC-MS).

Few endogenous metabolites that could be putative markers of hepatotoxicity were discovered in urine. Levels of methionine were higher 24 hours after BB dosage. Urocanate, related to histidine metabolism, and histidine itself displayed elevated levels. Notably, elevated urocanate levels were also found with galactosamine-induced hepatotoxicity (Beckwith-Hall et al., 1998).

In contrast to the urine analysis, distinct signals of BB-derived metabolites were not found in plasma. However, various endogenous metabolites were detected in plasma that could potentially be used as biomarkers of toxicity. Treatment-induced decreases in the levels of glucose and increases in lipid levels as measured by NMR were corroborated by clinical chemistry. The levels of formate in plasma and urine were increased after 24 and 48 hours. Formate may be produced from dimethylglycine via sarcosine and formaldehyde. Formate is also a product of oxalate in the glyoxylate catabolism, and possibly related to folate synthesis in the 1-carbon metabolism.

The most significant effects identified with the integrated analysis of transcriptomics and plasma metabolomics were categorized according to biochemical pathways. Changes in gene expression related to several pathways were described previously (Heijne et al., 2004). Other changes, e.g., in apoptosis and cell cycle were not recognized before. Pathways like glycolysis, GSH and amino acid metabolism were disturbed both at the gene expression and metabolite level, and are described next. Figure 5 presents a schematic overview of changes in GSH and amino acid metabolism that were found in association with bromobenzene-induced hepatic necrosis.

Glycolysis

Glucose levels in plasma transiently decreased after BB treatment. This could be ascribed to enhanced glycolysis, in order to increase the energy levels needed to restore homeostasis after the toxic insult. The decreasing glucose levels were concurrent with increasing plasma levels of alanine and lactate, products that may be formed by breakdown of glucose when the oxidation of pyruvate is incomplete. The expression of many enzymes involved in glycolysis, gluconeogenesis, and glucose transport were altered. The expression of a glucose transport protein was decreased by BB. In summary, the results suggested that glycolysis was enhanced (GAPDH, aldolase A, pyruvate kinase, G6PD and PGAM), and gluconeogenesis was reduced through reduction of expression of of G-6-phosphatase, transport protein 1 (G6pt1), alanine-glyoxylate aminotransferase and pyruvate carboxylase. In agreement with the present findings for bromobenzene, APAP was found to decrease glucose levels and was suggested to induce glycolysis based on gene expression and metabolite profile changes, suggestively as a reaction to decreased ATP availability from beta oxidation of fatty acids (Coen et al., 2004). This might corroborate the connection between these effects and toxicity, as high doses of acetaminophen are known to induce hepatic centrilobular necrosis in the same manner as bromobenzene.

GSH and amino acid metabolism

A pivotal process in the chemically-induced hepatic necrosis is the depletion of GSH levels. GSH normally protects cells by scavenging hazardous, reactive molecules. GSH levels decreased to around 25% of controls, 6 hours after oral BB dosage, (Heijne et al., 2004), while total depletion of hepatic GSH was observed 24 hours after ip administration of BB (Heijne et al., 2003). GSH is conjugated to BB-derived metabolites in a reaction catalyzed by GSTs. The NMR analysis showed that the reduction of GSH levels was accompanied by a decrease of plasma methionine. The depleted GSH levels were probably restored through a mechanism of induction of GSH synthase protein (Heijne et al., 2003) and GCLC gene expression (Heijne et al., 2004).

Along with the changes in the levels of GSH and methionine levels, other enzymes and metabolites related to GSH and methionine were found to change. GSH and methionine levels are interconnected via cysteine and homocysteine levels, involving enzyme activity of BHMT. Gene expression of BHMT initially increased and later decreased upon BB treatment. The expression of S-adenosyl homocysteine hydrolase was reduced by treatment. Plasma levels of dimethylglycine, produced in the reaction catalyzed by BHMT were found to correlate with the BHMT mRNA levels in time, and also the hepatic dimethylglycine dehydrogenase gene expression levels followed this pattern. Dimethylglycine can be catalyzed in a multistep reaction to formate. The levels of formate, in turn, were increased both in plasma and urine after treatment. Induced levels of cysteine in plasma were observed after BB treatment, along with increased gene expression of cysteine dioxygenase, while increased levels of cysteine sulfinic acid decarboxylase were observed before (Heijne et al., 2003). Plasma tyrosine levels show a characteristic pattern, decreasing drastically 24 hours after high BB, while 48 hours after high BB, levels were highly increased compared to controls. Protein levels of HPD, an enzyme involved in tyrosine metabolism, were found to decrease 24 hours after BB (Heijne et al., 2003). The level of phenylalanine is also related to tyrosine levels, and the data suggested a decrease in plasma phenylalanine levels due to the treatment.

Conclusions

The present study presents one of the first integrated analyses of transcriptomics and metabolite profiling. This approach more sensitively detected effects related to chemically induced hepatic necrosis. Through integration of the datasets, changes were observed before histopathology or clinical chemistry showed indications of necrosis. Moreover, this analysis provided new insights in cellular processes related to this type of toxicity. This study also highlights that the full integration of the methods awaits technical optimization, especially in the process of identification of metabolites. Nevertheless, corroborating findings from liver transcriptomics and plasma metabolite profiling enabled the generation of new hypotheses concerning cellular mechanisms putatively related to necrosis, such as changes related to apoptosis, glycolysis and amino acid metabolism. Both liver gene and plasma metabolite markers were discovered to correlate with the degree of hepatocellular necrosis in individual animals.