Gene Ontology Mapping as an Unbiased Method for Identifying Molecular Pathways and Processes Affected by Toxicant Exposure: Application to Acute Effects Caused by the Rodent Non-Genotoxic Carcinogen Diethylhexylphthalate

Richard A. Currie*,1,2, Vincent Bombail*,2, Jason D. Oliver*, David J. Moore*, Fei Ling Lim*, Victoria Gwilliam*, Ian Kimber*, Kevin Chipman{dagger}, Jonathan G. Moggs* and George Orphanides*

* Syngenta Central Toxicology Laboratory, Alderley Park, Cheshire, SK10 4TJ, United Kingdom; {dagger} School of Biosciences, University of Birmingham, Birmingham, B15 2TT, United Kingdom

1 To whom correspondence should be addressed at Syngenta Central Toxicology Laboratory, Alderley Park, Cheshire, SK10 4TJ, UK. Fax: +44 (0) 1625 585715. E-mail: richard.currie{at}syngenta.com.

Received April 4, 2005; accepted May 13, 2005


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
Toxicogenomics has the potential to reveal the molecular pathways and cellular processes that mediate the adverse responses to a toxicant. However, the initial output of a toxicogenomic experiment often consists of large lists of genes whose expression is altered after toxicant exposure. To interpret gene expression changes in the context of underlying biological pathways and processes, new bioinformatics methods must be developed. We have used global gene expression profiling combined with an evaluation of Gene Ontology (GO) and pathway mapping tools as unbiased methods for identifying the molecular pathways and processes affected upon toxicant exposure. We chose to use the acute effects caused by the non-genotoxic carcinogen and peroxisome proliferator (PP) diethylhexylphthalate (DEHP) in the mouse liver as a model system. Consistent with what is known about the mode of action of DEHP, our GO analysis of transcript profiling data revealed a striking overrepresentation of genes associated with the peroxisomal cellular component, together with genes involved in carboxylic acid and lipid metabolism. Furthermore we reveal gene expression changes associated with additional biological functions, including complement activation, hemostasis, the endoplasmic reticulum overload response, and circadian rhythm. Together, these data reveal potential new pathways of PP action and shed new light on the mechanisms by which non-genotoxic carcinogens control hepatocyte hypertrophy and proliferation. We demonstrate that GO mapping can identify, in an unbiased manner, both known and novel DEHP-induced molecular changes in the mouse liver and is therefore a powerful approach for elucidating modes of toxicity based on toxicogenomic data.

Key Words: toxicogenomics; gene ontology; pathway mapping; non-genotoxic carcinogenesis; diethylhexylphthalate.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
The recent sequencing of whole genomes has accelerated the development of tools, including microarray-based gene expression profiling, that make it possible to profile simultaneously the expression of all the genes expressed in an organism. The application of these techniques to toxicology, termed toxicogenomics, provides an unprecedented opportunity to identify holistically the biological pathways and processes affected by toxicant exposure (Fielden and Zacharewski, 2001Go; Irwin et al., 2004Go; Lovett, 2000Go; Nuwaysir et al., 1999Go; Orphanides, 2003Go). However, there is an urgent need to translate this information on individual genes into knowledge of the biological processes and pathways affected. The Gene Ontology (GO) Consortium (Ashburner et al., 2000Go) has developed a controlled vocabulary that describes the biological processes, molecular functions, and cellular components associated with a particular gene product, and so acts as a repository of the known functional biological information on each gene. Similarly, the GenMAPP organization is developing the Gene Map Annotator and Pathway Profiler (GenMAPP) to visualize metabolic and signaling pathways (Dahlquist et al., 2002Go). Several commercial organizations have also created useful tools that perform similar analyses (for example Applied Biosystems PANTHER https://panther.appliedbiosystems.com or Ingenuity Pathways Analysis http://www.ingenuity.com). The ability to determine which Gene Ontology terms or biological pathways are associated with differentially expressed genes from a microarray experiment, hereinafter referred to as GO or pathway mapping, would be an ideal way to gain an understanding of the molecular processes affected by the gene expression changes. Recently, many groups have developed methods and tools for pathway and GO mapping that reveal statistically significant annotations associated with microarray data (Beissbarth and Speed, 2004Go; Cheng et al., 2004Go; Doniger et al., 2003Go). In addition, GO or pathway mapping has recently been used to facilitate the interpretation of gene expression data derived from complex biological processes and systems, see Anderson et al. (2004a)Go for an example of GO-guided analysis applied to a peroxisome proliferator, and for a review see Li et al. (2004)Go. A major challenge in the interpretation of toxicogenomic data is the need to define how chemically induced changes in gene expression relate to conventional toxicological endpoints (Cunningham et al., 2003Go; Heinloth et al., 2004Go; Moggs et al., 2004Go; Paules, 2003Go). GO-based or pathway-based mapping of toxicant-induced gene expression changes to cellular pathways and processes represents a key step in this process.

We have examined the utility of GO and pathway mapping as an unbiased approaches for analyzing toxicogenomic data to identify holistically the biological processes and pathways affected by toxicant exposure. We chose to use the acute effects caused by the non-genotoxic carcinogen and peroxisome proliferator DEHP in the rodent liver, as these have been characterized at the molecular, cellular, and phenotypic levels, thus providing established biological endpoints from which to validate our GO-mapping approach. Previous studies have suggested that DEHP induces cancer through alterations in the control of cell growth, proliferation, and apoptosis, allowing the development and expansion of pre-neoplastic foci under the influence of paracrine signaling from non-parenchymal cells (Hasmall et al., 2000Go). The hepatocarcinogenic effects of DEHP in rodents are dependent on the presence of the peroxisome proliferator (PP)-activated receptor alpha PPAR{alpha} (Gonzalez, 2002Go), a member of the nuclear receptor family that functions as a ligand-regulated transcription factor, which in turn controls gene expression by binding to specific response elements (PPREs) within target gene promoters (Gonzalez, 2002Go). However, different responses of species to these non-genotoxic carcinogens call into question the relevance of such rodent studies for human risk assessment (Klaunig et al., 2003Go). Thus, the identification of the molecular pathways through which these compounds control cell proliferation should facilitate the assessment of the likely risk posed to humans by these compounds. Our GO mapping of DEHP-responsive genes not only reveals known and novel pathways of DEHP action in the rodent liver but also provides insights into the mechanisms by which non-genotoxic carcinogens control hepatocyte growth and proliferation.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
Animal experimentation.
B6C3F1 male mice (7–10 weeks old on arrival) were obtained from Harlan Olac Ltd. (Bicester, UK). Upon arrival at the laboratory, all animals were allowed at least 5 days to acclimatize. Animals were housed at up to 5/cage and received RM1 (Rat and Mouse No 1; Special Diet Services Ltd., Witham, UK) and water ad libitum. Animal experiments were conducted in accord with accepted standards (local and national regulations) of humane animal care. Mice (n = 6) were dosed by oral gavage (10 ml/kg body weight) with the model rodent non-genotoxic carcinogen diethylhexylphthalate (DEHP), every 24 h for 3 days (1150 mg/kg/day), or with an equivalent volume of vehicle control (corn oil). This dose of DEHP was found previously to cause a significant increase in liver weight after a 2-day exposure (James et al., 1998Go). The body weight of each mouse was recorded daily before dosing and prior to termination at 1, 2, 4, 8, 24, 48, and 72 h after first dose. Mice were killed by an overdose of anesthetic (Halothane Ph.Eur. vapor) followed by exsanguination via cardiac puncture. All animals were subject to a macroscopic examination post mortem, and liver tissue samples were fixed in formol saline and processed for histopathology. The livers were removed, blotted, and weighed. The comparison between the liver to body weight ratios for treated and control animals was made using a two-tailed t-test, assuming equal variance, in Microsoft Excel. The tissue samples for RNA analysis were collected in RNase-free tubes prior to being snap frozen in liquid nitrogen and stored at –70°C. The rate of hepatocyte DNA synthesis was measured in parallel groups of animals (n = 5) treated with corn oil or DEHP. 5-Bromo-2'-deoxyuridine (BrdU, 15 mg/ml in saline) was administered by Alzet osmotic minipumps (Cupertino, CA), and incorporation was assessed by immunohistochemistry after 48 h and 72 h of treatment. These animals received ketoprofen via injection, for pain relief.

Histology.
Formol–saline-fixed liver samples were embedded in paraffin, and 5-µm-thick sections were cut and stained with hematoxylin and eosin (H&E) before examination by light microscopy for phenotypic features including centrilobular hepatocyte eosinophilia, indicative of the development of smooth endoplasmic reticulum (peroxisomes), and hepatocyte basophilia, indicative of protein production in the endoplasmic reticulum.

Gene expression profiling.
Gene expression levels were measured 2, 8, 24, and 72 h after first exposure using Affymetrix (High Wycombe UK) Mouse Genome 430 2.0 GeneChip arrays (45,101 probe sets). Three animals used were from both treatment and control groups for each time point. Total liver RNA was isolated with RNeasy Maxi kits (Qiagen Ltd, Crawley, West Sussex, UK), and 10 µg was converted to cDNA and amplified with Affymetrix GeneChip 3' Amplification reagents (Invitrogen for Affymetrix). Biotin-labeled complementary RNAs were synthesized with the Enzo Bioarray HighYield RNA Transcript Labeling Kit, and 15 µg was hybridized to Affymetrix Mouse Genome 430 2.0 GeneChips as described in the Affymetrix GeneChip Expression Analysis Technical Manual (http://www.affymetrix.com/support/technical/manual/expression.manual.affx). Probe arrays were scanned, and the intensities were averaged with Microarray Suite 5.0 (Affymetrix). The mean of each array was globally normalized to 500. These data are submitted to Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/). Data were imported into GeneSpring v6.0 (SiliconGenetics), and the following normalizations were applied: data transformation setting measurements less than 0.01 to 0.01; per-gene normalization, where the signal strength of each gene is normalized to the median of its signal strength in all samples. The 45,101 probe sets on the array were then analyzed by means of a series of data filters and statistical tests. First, genes were selected for further analysis on the following basis: genes that had probe sets for which the expression value was greater than 60 during at least one time point and one treatment (which in our study constitutes the average background reading of all probe sets), and that had a present flag call in half the samples (3 of 6). To detect significant changes in the expression levels, a 2-way analysis of variance (ANOVA: parametric; assuming equal variances; Benjamini and Hochberg multiple testing correction at a False Discovery rate <0.05) on time and treatment was then applied to the resulting 21,317 genes, generating three lists of probe sets changing by "time" (1863 probe sets), "treatment" (899 probe sets), and "time–treatment interaction" (76 probe sets). Ratios of changes in gene expression were calculated by normalizing each DEHP-treated sample to the median value of the three corresponding time-matched vehicle controls. Because there was a possibility that the "time" gene list may include genes whose expression levels only changed with time (i.e., changed with time similarly in both treatment and control groups), to identify only those probe sets whose expression was altered by DEHP treatment during the time course, we performed a one-way ANOVA (parametric; assuming equal variances; Benjamini and Hochberg multiple testing correction at a False Discovery rate <0.05) on time on the 1863 "time" probe sets, using the aforementioned ratios of changes in gene expression. The resultant 1084 probe sets were merged with the "treatment" and "time–treatment interaction" lists identified in the two-way ANOVA to form a final list of 1786 probe sets whose expression was significantly altered by DEHP during the 72-h time course. Theoretically, the use of the Benjamini and Hochberg multiple testing correction at a False Discovery rate <0.05 limits the number likely to have been selected by chance to 5% of the 1786 probe sets.

Gene Ontology and GenMAPP analysis of DEHP-regulated genes.
To identify the biological processes and pathways altered by DEHP, we used complementary methods (see Fig. 3C): mapping the DEHP-responsive Affymetrix probe sets to Gene Ontology (Cheng et al., 2004Go) and secondly to GenMAPPs (Doniger et al., 2003Go). Statistically significantly (Benjamini and Hochberg False discovery rate <0.1) overrepresented Gene Ontology Biological Process and Cellular Component annotations from the 1786 differentially regulated probe sets were identified by comparison with the 21,317 probe sets identified as expressed in the liver in our experiment (hereafter called the liver transcriptome) using the Web-based tool GOStat (Beissbarth and Speed, 2004Go; http://gostat.wehi.edu.au/). GOStat counts the number of appearances of each GO term for the genes inside the DEHP-altered group and for the reference (liver transcriptome) genes. Fisher's exact test is performed to judge whether the observed difference is significant. To quantify the extent to which a Gene Ontology annotation was overrepresented, an alternative complementary GO-mapping tool was employed, in which the representation index (Ri) was calculated for each annotation as described by Karpinets et al. (2004)Go with one modification. In their original method they assumed that if a set of genes in a cluster was chosen at random from the chip, the percentage of genes with a given annotation would equal the percentage of genes on the array with that annotation. As not all of the probe sets on the GeneChip Mouse Genome 430 2.0 Array were found to be expressed in our study, this assumption does not hold. Therefore the "universe" of annotations for the comparison was reduced to the liver transcriptome rather than the entire group of probe sets on the chip. The Gene Ontology Biological Process and Biological Component annotations associated with the liver transcriptome were extracted via the NetAffx Gene Ontology Mining tool (Affymetrix,http://www.affymetrix.com/analysis/query/go_analysisaffix; Chen et al., 2004) and exported into Microsoft Excel, where Ri values were calculated (Karpinets et al., 2004Go) for each statistically significant GO term identified using GOStat. Of the 1786 DEHP-regulated probe sets, 890 had Process annotations associated with their target genes, and 960 had Component annotations. A Ri value of +1 would indicate that all of the probe sets associated with a particular annotation are included in the DEHP-regulated gene set, whereas a value of –1 would indicate that none are included. A value of 0 indicates that the proportion of probe sets with that annotation in the DEHP-regulated gene set is the same as the proportion in the liver transcriptome. The pathway visualization and analysis tool GenMAPP/MAPPFinder (Doniger et al., 2003Go) was used to search out and visualize statistically significant pathways that were altered by DEHP. For PANTHER analysis (https://panther.appliedbiosystems.com/), the Affymetrix probe set identifiers for both the "DEHP-altered" genes and the "liver transcriptome" were converted to the representative NCBI identifiers with GeneSpring and then submitted to the PANTHER "compare gene lists" tool.



View larger version (29K):
[in this window]
[in a new window]
 
FIG. 3. Gene Ontology Biological Process and Cellular Component annotations associated with DEHP-regulated probe sets. Pie charts showing GO Biological Process (A) and Cellular Component (B and inset) annotations associated with the DEHP-regulated probe sets. Sizes of slices are indicative of number of probe sets with that annotation, and the inset shows the "children" annotations of the GO term "intracellular." (C) Schematic of Gene Ontology analysis methods used to determine the statistically significantly overrepresented annotation terms.

 
Quantitative real-time reverse transcriptase–polymerase chain reaction (PCR).
To confirm and extend the alterations in gene expression identified using Affymetrix microarrays, two-step quantitative real-time RT-PCR was performed on RNA isolated, as described above, from the livers of animals dosed with DEHP for 1, 2, 4, 8, 24, and 48 h or their time-matched vehicle-treated controls. Contaminating genomic DNA was removed with DNA-free (Ambion), and 200 ng total RNA was reverse transcribed using random hexamers and the Superscript III first strand cDNA-synthesis kit (Invitrogen). Quantitative PCR (qPCR) was performed on an ABI Prism 7700 (Applied Biosystems) using the DyNAmo HS SYBR Green qPCR kit (Finnzymes) and the following primers: 900 nM G0s2 forward 5'- GCCCAGAGCTCAGATGGAAA, 900 nM G0s2 reverse 5'-TTCTGCGCCATCATCTCCTT; 50 nM Gadd45b forward 5'-TGGTCGCACGGGAAGGTTTT, 900 nM Gadd45b ß reverse 5'-GGTGAAGTGAATCTGCAGAGCGTATA ; 900 nM Nola3 forward 5'-CACCGCGTCCTGTCCTCTGA, and 900 nM Nola3 reverse 5'-TTGACGCGTTTGGTTGCTGATAATAG. All mouse gene symbols are in accordance with the conventions of the Mouse Genome Informatics (MGI; see http://www.informatics.jax.org/mgihome/nomen/gene.shtml).


    RESULTS AND DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
Peroxisome Proliferator Responses Induced by DEHP in the Mouse Liver
Our aim was to determine the extent to which GO mapping of global gene expression profiling data can be used in an unbiased and holistic way to identify the biological processes or pathways associated with responses to toxicants. The mouse liver growth response to the peroxisome proliferator DEHP was used as a model system. We first confirmed that our experimental system reproduced the acute biological effects of DEHP in the mouse liver described previously: Male mice (7–8 weeks old) were dosed by oral gavage with DEHP (1150 mg/kg/day) or vehicle (corn oil; 10 ml/kg/day), and liver weights and histological parameters were measured at seven different times after exposure (Fig. 1A). As expected, DEHP treatment induced a statistically significant (p < 0.005) increase in liver weight 48 h and 72 h after the first exposure (Fig. 1B), consistent with previous observations (James et al., 1998Go). Furthermore, as previously reported (Macdonald et al., 2001Go), histological analysis (data not shown) revealed an increase in eosinophilic staining of the smooth endoplasmic reticulum in centrilobular hepatocytes at both these times (indicative of peroxisome proliferation) and hypertrophy of these cells. We also observed an increased rate of hepatocyte DNA synthesis 48 h and 72 h after the first dosing in a satellite group of animals containing BrdU minipumps (Fig. 1C), consistent with an increase in S-phase progression in the periportal area of livers from mice exposed to DEHP. Members of the Cyp4a gene subfamily had previously been reported to be upregulated by DEHP in the mouse liver (Kroetz et al., 1998Go). We also observed a robust increase in Cyp4a10 expression from the earliest time point sampled (Fig. 1D). We conclude that the dose of DEHP used in our study induced conventional liver gene expression, cell growth, and proliferative responses.



View larger version (34K):
[in this window]
[in a new window]
 
FIG. 1. Diethylhexylphthalate (DEHP) treatment induces a stereotypical PP-response. A. Schematic of the study design showing three oral doses of DEHP at 24-h intervals. Livers were harvested for histological, S phase, and induction of Cyp4a10 at the indicated times. B. increase in liver/body weight ratio, data expressed as mean ± standard deviation for n = 6 mice, **p < 0.01 as determined by the Student t-test. C. BrdU incorporation in control (CO) and DEHP-treated mouse livers 72 h after the initial dosing. The bar represents 200 µm. D. Increased expression of Cyp4a10, measured by qPCR, relative to a control gene (NOLA3, nucleolar protein family A, member 3 which the Affymetrix microarray data indicated did not change across the time course).

 
Transcript Profiling of Temporal Patterns of DEHP-Responsive Gene Expression During Liver Cell Growth and Proliferation
To identify the global gene expression changes associated with DEHP treatment, liver RNA from each of the four time points (2, 8, 24, and 72 h) was processed for microarray analysis using Affymetrix MG-430 2.0 GeneChips (see Fig. 2A for a schematic). Transcript profiling was performed separately on three individual animals within each treatment group (generating a total of 24 microarray data sets), allowing a rigorous statistical analysis of the gene expression data to identify probe sets that were significantly altered in expression in response to DEHP treatment (see Materials and Methods).



View larger version (25K):
[in this window]
[in a new window]
 
FIG. 2. Staged transcriptional response of the liver in response to DEHP. A. Schematic of the microarray experimental design and subsequent data analysis to illustrate the comparison of DEHP-treated animals with requisite time-matched control animals. B. A heat map illustrating the staged response of the 1786 DEHP-altered probe sets (rows) for each individual animal (column). The genes (rows) and treatment groups (columns) are clustered by Pearson correlation, but the gene tree has been omitted for clarity.

 
We used a multi-step method to analyze the microarray gene expression data. First, data were filtered and subjected to ANOVA to identify the 1786 genes (see Table 1 in the Supplementary Data online) with altered expression in DEHP-treated mice at at least one of the four time points (see Materials and Methods). The heat map in Figure 2B represents the DEHP-regulated probe sets, clustered by the Pearson correlation coefficient according to their similarity in expression pattern by gene and by condition (time after first dose). This analysis reveals a complex, multistage transcriptional response to DEHP in the liver, and it also illustrates the high reproducibility of response among the replicate animals (at each time point). Comparison of the DEHP-induced transcript profiles across the four different time points by unsupervised hierarchical clustering revealed two distinct phases of the DEHP-induced transcriptional cascade: early (2 h and 8 h) and late (24 h and 72 h). These transcriptional responses indicate that DEHP induces a complex program of gene regulatory events that precede the onset of liver growth and proliferation, and thus may provide novel insights into the molecular mechanisms associated with peroxisome proliferation.

To confirm that the microarrays were capable of detecting known DEHP-responsive genes, we examined the expression profiles of classical PP-responsive genes. As mentioned above, members of the Cyp4a gene subfamily are direct targets of PPAR{alpha} (Mandard et al., 2004Go) and have been reported to be upregulated by DEHP in mouse liver (Kroetz et al., 1998Go). Our data reveal that two members of the Cyp4a subfamily (Cyp4a10 and Cyp4a14) were consistently upregulated by DEHP from the 2 h time point onward, ranging from 4.6 ± 1.4-fold induction to 73.2 ± 32.9-fold induction during the 72 h time course (see Table 1 in the Supplementary Data online), changes similar in magnitude to the those observed in Cyp4a10 gene expression as measured by qPCR) (Fig. 1D). In addition we compared our data to published transcript profiling studies of a variety of peroxisome proliferators (Anderson et al. 2004bGo; Hamadeh et al., 2002Go; McMillian et al., 2004Go; Wong and Gill, 2002Go; Yadetie et al., 2003Go). Comparison of gene expression data with those obtained in other studies on peroxisome proliferators necessarily has to be made with caution because the experimental conditions used (species, chemical, dose, time points) may differ. Nevertheless, in the rat, gene expression changes associated with fatty acid metabolism, cell cycle, and acute phase proteins were altered at 24 h and 2 weeks after daily dosing with peroxisome proliferators (Hamadeh et al., 2002Go; McMillian et al., 2004Go). Furthermore, Yadetie et al. (2003)Go treated rats with ciprofibrate daily for 60 days and found changes in gene expression associated with lipid metabolism and inflammatory responses. Expression profiles were also altered in relation to cell cycle and stress responses, but these changes may reflect secondary changes associated with altered pathology at this late time point. It is important to note that in mouse hepatocytes, the vast majority of gene expression changes seen after 7 days of exposure to WY-14,643 were dependent on PPAR{alpha} (Anderson et al., 2004bGo). The marked effect on expression of proteasomal genes seen in that in vitro study were reflected in our shorter term in vivo study with DEHP. Thus, our gene expression profiling data confirm a stereotypical liver cell growth and proliferation response to DEHP at the transcriptional level, as well as at the organ and histological levels (Fig. 1B and 1C).

Gene Ontology and Pathway Mapping of DEHP-Induced Gene Expression Changes
Next, we employed a supervised Gene Ontology–driven clustering approach, using both Gene Ontology and pathway mapping tools, as an unbiased method for identifying the predominant biological processes and pathways represented among the 1786 DEHP-responsive genes. The major Gene Ontology Biological Process and Cellular Component terms represented in the DEHP-regulated probe sets and their relative abundance are illustrated in Figure 3A and B, respectively. DEHP regulates the expression of genes associated with a broad range of biological processes. The largest group of genes is associated with "metabolism," while "organismal/cellular physiological processes" and "cell communication" are the next most abundant groups of gene annotations. DEHP also alters the expression of genes in a variety of cellular component locations, with "intracellular" genes being the largest single group, and with significant numbers of probe sets being associated with "membranes" and the "extracellular space." The inset pie chart in Figure 3B further illustrates the subdivision of the "intracellular" GO component annotations, showing that the predominant annotations are "mitochondrion," "endoplasmic reticulum," and "peroxisome." This rather simplistic GO analysis merely quantifies the number of probe sets with a particular annotation, and thus it may simply reflect the distribution of annotations associated with genes expressed in the liver (i.e., it does not allow us to determine which GO annotation(s) are overrepresented in the DEHP-regulated set).

To overcome this limitation, we next used a series of more advanced Gene Ontology and pathway mapping tools (Fig. 3C) to identify the biological pathways and processes targeted by DEHP. Both DEHP-regulated and liver transcriptome probe sets were submitted to GOStat (Beissbarth and Speed, 2004Go), to determine the most statistically significantly (Benjamini and Hochberg False Discovery Rate p < 0.1) overrepresented Cellular Component annotations (Table 1) and Biological Process annotations (Table 2). To identify the extent to which a particular term was overrepresented, we calculated the representation index (Ri; Karpinets et al., 2004Go) for each annotation relative to the liver transcriptome (as described in Materials and Methods). These values are also shown in Tables 1 and 2. To identify known biological pathways that were altered by DEHP, we analyzed the DEHP-regulated and liver transcriptome gene lists, using MAPPFinder (Doniger et al., 2003Go) and PANTHER (https://panther.appliedbiosystem.com). Significantly overrepresented GenMAPPs ranked by Z-score (as defined in Doniger et al., 2003Go) are listed in Table 3. Significantly overrepresented PANTHER pathway and biological process groups are listed in Tables 2 and 3 in the Supplementary Data online. Importantly, the most significantly overrepresented Gene Ontology component annotation (Ri = 0.310) is "peroxisome." Given that DEHP is known to target this organelle at the molecular level, resulting in peroxisome proliferation, our data demonstrate that GO mapping of global gene expression profiling data can be used in an unbiased manner to identify the principal biological processes associated with cellular responses to a toxicant. Indeed, inspection of the gene lists associated with this term reveals genes associated with both peroxisomal organization and biogenesis, in addition to peroxisomal fatty acid ß-oxidation. This observation confirms the validity of our GO analysis method, as induction of peroxisomal genes is expected to feature prominently in the gene expression response to PPs (Mandard et al., 2004Go). The significant overrepresentation of genes annotated as microsomal (Ri = 0.147) probably reflects the regulation of xenobiotic metabolism enzymes, as well as the PPAR{alpha}-induced Cyp4a genes (Cyp4a10 and Cyp4a14), which are also expressed in microsomes. The observed overrepresentation of mitochondrial genes is likely to reflect the induction of fatty acid metabolism genes involved in the PP response.


View this table:
[in this window]
[in a new window]
 
TABLE 1 Statistically Significantly Altered Gene Ontology Cellular Component Terms in the Diethylhexylphthalate (DEHP)-Induced Gene List When Compared to the Liver Transcriptome

 

View this table:
[in this window]
[in a new window]
 
TABLE 2 Statistically Significantly Altered Gene Ontology Biological Process Terms in the DEHP-Induced Gene List When Compared to the Liver Transcriptome

 

View this table:
[in this window]
[in a new window]
 
TABLE 3 Statistically Significantly GenMAPPs Associated with the DEHP-Induced Gene List When Compared to the Liver Transcriptome

 
Another well-characterized effect of peroxisome proliferators is the coordinate induction of genes involved in fatty acid metabolism via the activation of PPAR{alpha} (reviewed in Mandard et al., 2004Go). This effect is confirmed by the overrepresentation of Gene Ontology terms and PANTHER and GenMAPP pathways involved in metabolism of lipids (e.g., acyl-coA metabolism, fatty acid ß-oxidation, pantothenate and coA biosynthesis, or coenzyme metabolism; Tables 2 and 3). Other gene expression changes, such as those related to amino acid metabolism (Kersten et al., 2001Go), have also been reported (reviewed in Mandard et al., 2004Go). These Biological Processes are component parts of the "metabolism" annotation in Figure 3A (left) and were therefore not illustrated individually.

A comparison of the statistically significantly overrepresented GO terms in Table 1 with the Cellular Components identified, as shown in Figure 3B, implies that the majority of GO annotations identified as associated with the DEHP-regulated probe sets are not overrepresented when compared to the liver transcriptome. Although some genes that are not overrepresented may be biologically important (e.g., see DEHP-responsive genes related to epigenetic status discussed below), our analysis illustrates the utility of overrepresentation analysis, when compared to a mere listing of GO terms, for the identification of the collective effects of DEHP treatment on biological processes.

There are many tools in the public domain that perform the overrepresentation analysis of Gene Ontology described above, we chose to use GOStat because of its Web-based interface and ease of use. The analyses performed by PANTHER use equivalent statistical tools to compare two gene lists; however, Applied Biosystems (ABI) has created its own ontology as the basis for the comparative analysis. A comparison of Tables 2 and 3 with Tables 2 and 3 of the Supplementary Data online indicates the high degree of concordance among the three methods; especially with regard the identification of the overrepresentation of genes in lipid (fatty acid) metabolism, steroid metabolism, blood clotting, complement activation, and the circadian clock. PANTHER also reproduces identification of the "Defense response" genes by GOStat, under the term "Immunity and Defense." Therefore the broad conclusions drawn from these different tools are the same, and either of them could be used for an initial analysis. However, there are some differences between the outputs of these tools.

PANTHER analysis also identified as overrepresented a number of genes involved in transporting the components of those metabolic pathways identified as overrepresented by all three methods—e.g., "lipid and fatty acid transport" and "small molecule transport." Also, genes involved in proteolysis and monosaccharide metabolism were identified as overrepresented by PANTHER but not by GOStat or MAPPFinder. The reason for this probably lies in the structure of the ontology used as the basis for the comparisons, because these four PANTHER biological process terms have many (GO terms), whereas only one (PANTHER term) maps in the PANTHER ontology, which may indicate that there is an advantage to using PANTHER for identifying increases in a number of related biological processes. Indeed, searching for overrepresented "molecular function" Gene Ontology terms using GOStat does indicate an overrepresentation of a number of diverse transport genes (data not shown). Clearly, an analysis involving multiple tools gives useful complementary information.

This mapping of GO processes to a single term may also be a disadvantage if the aggregated processes are, in fact, always biologically distinct. For example, the ER-overload GO term is mapped, along with a number of other specific stress responses, to a much more general "Stress Response" biological process. Given that stress-responses are discrete biological processes, the PANTHER ontology may be less useful for the identification of cellular stress through overrepresentation analysis. This problem may be inherent in the design of flat ontologies, and it demonstrates that great care should be taken in the design of ontologies for the purpose of microarray data analysis to avoid losing important biological information. It also points to the importance of using complementary data analysis tools for a thorough evaluation of genomic data.

We conclude that GO mapping of transcript profiling data can correctly identify known molecular pathways and cellular processes targeted by a PP. These include acyl-CoA and fatty acid metabolism pathways and the peroxisomal cellular component, findings consistent with previous studies of DEHP-induced liver cell growth and proliferation. In addition, our GO mapping analyses significantly extend previous observations of DEHP-dependent gene expression changes associated with a wide range of biological responses, including amino acid metabolism, hemostasis, complement activation, and steroid metabolism (Tables 2 and 3; Fig. 4), and they reveal novel insights into the mechanisms of DEHP action, including roles for the endoplasmic reticulum overload response and circadian rhythms (described in more detail below). Therefore, GO and pathway mapping are powerful approaches to generating an unbiased view of the biological processes and cellular components that are regulated by DEHP at the transcriptional level.




View larger version (51K):
[in this window]
[in a new window]
 
FIG. 4. GenMAPPs illustrating some classes of DEHP-responsive genes. (A) GenMAPP pathways of the blood clotting cascade and complement activation, with the Affymetrix GeneChip data for DEHP-responsive probe sets overlaid. Gray boxes indicate genes whose probe sets were classed as expressed; blue boxes indicate those genes whose probe sets expression was significantly altered upon DEHP treatment. (B) GenMAPP pathway of steroid biosynthesis colored as above. The graph illustrates the downregulation of 3-ß-hydroxy-{Delta}steroid dehydrogenase isoforms 2, 3, 5, and 6 at early time points, plus alterations in other genes in this pathway.

 
Hemostasis and Complement Genes
Surprisingly, our GO and GenMAPP analyses revealed extensive DEHP-dependent transcriptional responses associated with complement activation and the blood clotting cascade (i.e., hemostasis) pathways (Tables 2 and 3, see also Fig. 4A for a GenMAPP and Table 4 for quantitative gene expression data). Although the functional consequences of DEHP-induced alterations in blood clotting and complement activation pathways are unclear (see below), our gene expression data are consistent with previous experiments in which peroxisome proliferators were shown to reduce the expression levels of many complement component genes, including complement c9, complement c4b-binding protein, and serine protease inhibitor 2 (Anderson et al., 2004bGo; Corton et al., 2004Go; McMillian et al., 2004Go; Wong and Gill, 2002Go; Yamazaki et al., 2002Go). Some of these gene expression changes were shown to be dependent on PPAR{alpha} (Anderson et al., 2004bGo). A comparison of the components altered in these published studies consistently shows alterations in complement c9 expression, with most studies also showing alterations in C4bp. Although the overlap between the present study and that of Anderson et al. (2004b)Go is not extensive (complement component 9), in aggregate the gene expression changes observed in our study (Fig. 4A) with DEHP and that of Anderson et al. with WY-14643 show that peroxisome proliferators can alter the expression of almost all components of the complement cascade.


View this table:
[in this window]
[in a new window]
 
TABLE 4 Quantitative Data for Distinct Functional Classes of DEHP-Responsive Genes in the Mouse at Time Points 2, 8, 24, and 72

 
Other PPAR{alpha} ligands, such as fibrates, have also been reported to downregulate the expression of genes coding for components ({alpha}, ß, and {gamma}) of fibrinogen in both rodents (Corton et al., 1998Go; confirmed by Kockx et al., 1999Go and McMillian et al., 2004Go) and humans (Gervois et al., 2001Go). Our data show that the genes encoding fibrinogen ß and {gamma} are also downregulated by DEHP, along with a number of additional genes in these pathways (Fig. 4A; see also Table 1 of the Supplementary Data online), thus extending the mechanistic link between DEHP-induced liver responses and alterations in the expression levels of hemostasis genes. In Crl:CD BR (CD) rats treated with the peroxisome proliferator WY-14643, Hurtt et al. (1997)Go observed increased pro-thrombin times in diluted (but not undiluted) serum from some rats exhibiting clinical signs of anemia, suggesting a subclinical abnormality of coagulation. In their study this condition was resolved by dietary supplementation with vitamin K, demonstrating that the abnormality involved some vitamin K–dependent coagulation factor deficiency in the extrinsic pathway. Given that Factors VII, IX, and X are vitamin K–dependent, our observations of decreased expression for each of these factors may explain, in part, the data of Hurtt et al. (1997)Go in the CD rat. It should be noted that this effect has only been reported in the CD rat strain and that this is probably a strain-specific difference. It is also noteworthy that fibrate drugs have been reported to show beneficial effects with respect to cardiovascular disease (reviewed by Elisaf, 2002Go), and thus alterations in complement and hemostasis pathways might be part of this mechanism.

Steroid Metabolism
The liver is a major center for steroid hormone metabolism and is vital for the correct functioning of a number of hormonal systems. Interestingly, our GO and GenMAPP analyses (Tables 2 and 3) identified genes associated with steroid hormone metabolism as being overrepresented in the DEHP-altered gene list. Consistent with our observations, Wong and Gill (2002)Go demonstrated altered expression of some genes involved in steroid hormone metabolism in the livers of mice dosed for 13 weeks. Additionally Fan et al. (1998)Go showed that peroxisome proliferators can increase the expression of estrogen-metabolizing enzymes. Our data extend these observations, by showing that four isoforms of 3-ß-hydroxysteroid dehydrogenase (2, 3, 5, and 6) and 3-isoforms of 17-ß-hydroxysteroid dehydrogenase are coordinately regulated by DEHP. The Hsd3b-isoforms are all downregulated from 2 h after DEHP treatment (Fig. 4B), with peak decreases seen at 8 h (see Fig 4B inset graph). Interestingly, Wong and Gill (2002)Go reported large sustained alterations in Hsd3b gene expression after 13 weeks of dosing, whereas we observed only transient changes in Hsd3b isoforms. Therefore, it seems likely that any dramatic alteration in their expression occurs only with longer term dosing with DEHP.

Early Signaling Responses to DEHP
Several GO terms associated with stress, defense, and regulatory responses were identified as overrepresented in response to DEHP: for example, "endoplasmic reticulum (ER)-overload response," "response to stress," and "negative regulation of protein kinase activity" (Table 2).

ER Overload Response and Apoptosis
The ER-overload response is defined by AmiGO (www.godatabase.org), as "the series of molecular signals generated by the accumulation of normal or misfolded proteins in the endoplasmic reticulum and leading to activation of transcription by NF-kappaB." The GO cellular component analysis (Table 1) indicates alterations in the expression of many genes whose products are resident in the ER. Interestingly a specialized ER compartment has been proposed to be important for peroxisome biosynthesis (Geuze et al., 2003Go). Our histological studies showed increased eosin staining after 48 h (data not shown), which may be representative of increased smooth ER or peroxisome proliferation. Also, the ER-overload response has been reported to be activated by overexpression of cytochrome P450s (Szczesna-Skorupa et al., 2004Go). Interestingly both Macdonald et al. (2000)Go and Anderson et al. (2004a)Go have shown that peroxisome proliferators induce proteins involved in proteome maintenance in a PPAR{alpha}-dependent manner and so may provide evidence for ER or other stress responses. Irrespective of the mechanism involved, activation of the ER-overload response would be expected to lead to a block in apoptosis via activation of NF-{kappa}B (nuclear factor kappaB) (Szczesna-Skorupa et al., 2004Go). It is therefore possible that the DEHP-induced ER-overload response drives an anti-apoptotic mechanism that may play a role in DEHP-induced carcinogenesis by preventing the normal apoptotic clearance of tumorigenic cells (Oliver and Roberts, 2002Go).

Signaling Pathways
Inspection of the genes associated with the overrepresented GO cluster "response to stress" reveals the presence of a number of known components of the TNF/IL-1 (tumor necrosis factor/interleukin-1) signaling pathways, including Irak2, Myd88, mitogen-activated protein kinase kinase kinase 7 interacting protein 2 (Map3k7ip2, TAB2), and Ikbkg (Fig. 5A). These genes have a similar time course of expression (see Table 4), with increases early (2 h) after DEHP treatment that decline at later times (24 h). Interestingly, it has been shown that, in the presence of PPs, non-parenchymal cells (probably Kupffer cells) secrete factors (e.g., TNF{alpha}, IL-1) that stimulate hepatocytes to enter S phase (Hasmall et al., 2001Go). For PP-induced S-phase entry, PPAR{alpha} is required in the hepatocytes but not in the non-parenchymal cells (Hasmall et al., 2001Go), indicating that a PPAR{alpha}-dependent gene expression change in the hepatocytes is required to respond to the levels of cytokines secreted by the non-parenchymal cells. Additionally, there is evidence that the p38 and PPAR{alpha} signaling pathways are both involved in the proliferative response of rodent hepatocytes upon PP exposure (reviewed by Roberts et al., 2002Go)



View larger version (22K):
[in this window]
[in a new window]
 
FIG. 5. Tumor necrosis factor/interleukin-1 (TNF/IL1) pathway components and their alterations by DEHP. (A) Genes associated with the "response to stress" cluster from Table 2 and additional literature-derived genes (Janssens and Beyaert, 2003Go) (De Smaele et al., 2001Go; Johnson, 2002Go) in the TLR-signaling pathways are illustrated. Genes colored in red are upregulated at the 2 h or 8 h time points, and genes in gray are unaltered components of the pathway. (B) Graph shows the extended time course of Gadd45b expression upon DEHP or CO treatment. Data points are expressed relative to a control gene Nola3, as in Figure 1D.

 
The activation of both the NF-{kappa}B (which could be activated by either or both of the TNF/IL-1 pathways and the ER-overload response) and p38 signaling pathways by DEHP is supported by our observations of a similar temporal pattern of DEHP-induced gene expression changes for both NF-{kappa}B and p38 signaling pathway target genes (Fig. 5A). Gadd45ß is a known NF-{kappa}B target gene (De Smaele et al., 2001Go), and thioredoxin-interacting protein (Txnip) has been described as a downstream target of p38 activity (Schulze et al., 2004Go). Both genes are induced at early time points (Table 4 and see also Fig. 5B), a finding consistent with an increase in signaling through NF-kB and p38. The upregulation of Gadd45b may result in a downstream suppressive effect on apoptosis via the Jun N-terminal kinase (JNK) pathway (Fig. 5A; De Smaele et al., 2001Go).

These alterations in the expression levels of key signaling components may therefore create a DEHP-induced sensitization of hepatocytes to the levels of cytokines prevailing in the hepatic milieu. Overall, the rapid (within 2 h) DEHP-induced alterations in the expression of various stress-response genes are consistent with the activation of NF-{kappa}B and p38 signaling, leading to a pro-proliferative and anti-apoptotic phenotype in hepatocytes (Roberts et al., 2002Go) that drives subsequent liver weight increases and proliferation (Fig. 1B).

Circadian Genes
A novel finding of both our Gene Ontology and GenMAPP analyses of DEHP-responsive genes was the overrepresentation of genes with a "circadian" GO annotation (Fig. 6A and Tables 2 and 3). This GO term is defined as "the specific actions or reactions of an organism that recur with a regularity of approximately 24 hours" (AmiGO, www.godatabase.org). DEHP-responsive genes with this annotation include genes that encode circadian rhythm transcriptional regulators (e.g., Per2, Cry1, Clock, and Dbp; Schibler and Sassone-Corsi, 2002Go; Oishi et al., 2003Go) and genes that exhibit a circadian rhythm in their gene expression pattern (e.g., G0s2, Map3k7ip2; Oishi et al., 2003Go; Zambon et al., 2003Go). Our GO mapping data therefore suggest that DEHP regulates the circadian timing of cellular responses in the liver at the transcriptional level. Importantly, recent observations indicate a direct link between components of the circadian rhythm regulatory machinery and the timing of cell division in the liver (Matsuo et al., 2003Go). Thus, the DEHP-induced hepatocyte proliferation and liver enlargement observed in the present study (Fig. 1) may be caused in part by alterations in the circadian rhythm gene pathways. Whether circadian rhythm genes are under direct control of PPAR{alpha} remains to be determined, but it is noteworthy that Ppara expression levels have been reported to be regulated with a circadian rhythm (Lemberger et al., 1996Go), which in turn may regulate the circadian expression patterns of certain PPAR{alpha}-target genes (Patel et al., 2001Go).



View larger version (30K):
[in this window]
[in a new window]
 
FIG. 6. GenMAPP of circadian genes illustrating DEHP-altered probe sets. (A) GenMAPP pathway of genes with a circadian expression pattern with the expressed and DEHP-responsive genes highlighted as described in Figure 4. (B) Graph shows the extended time course of the G0/S-phase switch protein G0s2 expression upon DEHP or CO treatment. Data points are expressed relative to the control gene Nola3, as in figure 1D.

 
There is evidence that a number of environmental changes, such as exposure to glucocorticoids and dexamethasone (Balsalobre et al., 2000Go), as well as food restriction (Hara et al., 2001Go), may alter the entrainment of the clock in peripheral tissues such as the liver. The significant overrepresentation of genes identified in "glucocorticoid mineralcorticoid metabolism" (Table 3) suggests that an alteration in genes with a circadian rhythm may be mediated via alteration in glucocorticoid metabolism. Alternatively, given (1) that there are overlapping PPAR{alpha}-dependent gene expression changes regulated by food restriction and peroxisome proliferator treatment, and (2) that there are close links between metabolism, food intake, and the entrainment of circadian rhythms (Corton et al., 2004Go; Rutter et al., 2002Go; Wong and Gill, 2002Go), it is possible that the complex interrelationships between these metabolic and transcriptional consequences of PP treatment alter the expression of circadian genes.

One of the earliest DEHP-induced alterations in circadian gene expression was observed for the G0/G1 switch gene 2 (G0s2). The mRNA levels of G0s2 have been reported to follow a circadian pattern of expression in the mouse liver and skeletal muscle (Zambon et al., 2003Go). Consistent with circadian regulation, we found that G0s2 gene expression levels varied (up to fivefold) with time in vehicle-treated control animals (Fig 6B). Strikingly, DEHP treatment induced a much larger upregulation in G0s2 expression that was maximal (29-fold relative to the control gene) at 8 h (Fig. 6B).

G0s2 was initially identified as an upregulated transcript in human T lymphocytes upon lectin-induced proliferation (Siderovski et al., 1990Go), and its expression is repressed by treatment with the immunosuppressant cyclosporin A (Cristillo et al., 1997Go). Although these data point toward a role of G0s2 in the mechanism of G0 to G1 transition, the gene product has also been reported to be upregulated upon retinoic acid-mediated cell growth arrest in the human acute promyelocytic leukemia cell line, NB4 (Tamayo et al., 1999Go). These apparent contradictions of function could be explained by our observations that GOs2 has a similar DEHP-induced expression pattern to that of the Cdkn1a cell cycle regulator (see Table 1 in the Supplementary Data online), and thus may have a similar function (i.e., to inhibit cell cycle progression; Ekholm and Reed, 2000Go). However, we have recently shown that a transient increase in Cdkn1a precedes nuclear hormone receptor–driven proliferation in vivo (Moggs et al., 2004Go) and therefore increases in the expression of negative cell-cycle regulators may stall the cell cycle until the preparations for progression into S phase have been completed.

Genes associated with epigenetic status.
Our initial unbiased Gene Ontology mapping identified genes associated with the term "regulation of gene expression, epigenetic" (Fig. 3A) as being DEHP regulated, although our subsequent overrepresentation analysis (Table 2) did not highlight this biological process. Given that a growing body of evidence suggests that alterations in epigenetic status, particularly DNA methylation patterns, accompany, and may even promote, carcinogenesis induced by non-genotoxic chemicals (Bombail et al., 2004Go; Watson and Goodman, 2002Go), we explored whether DEHP might also be associated with early alterations in the expression of genes associated with epigenetic status. Therefore, we also interrogated the 1786 DEHP-responsive genes, using both hand-edited gene lists and GO terms associated with known mechanisms of epigenetic regulation of gene expression, including "DNA methylation," "epigenetic," "chromatin," "one-carbon compound metabolism."

Using this approach, we identified methyl CpG-binding domain protein (Mbd1) as an early DEHP-responsive gene that was initially upregulated and was then repressed at later times (see Table 1 of the Supplementary Data online). Mbd1 binds to methylated DNA sequences and is involved in transcriptional repression and the epigenetic regulation of gene expression (Fujita et al., 2003Go). The DEHP-dependent repression of Mbd1 may thus function to activate the transcription of downstream target genes at later times. Furthermore, we also identified DEHP-induced changes in the expression of a number of genes involved in one-carbon compound metabolism, including enzymes involved in cellular methylation reactions (e.g., methionine adenosyltransferase 2A (Mat2a), serine hydroxymethyltransferase 1 and 2 (Shmt1 and Shmt2), formyltetrahydrofolate dehydrogenase (Fthfd), and S-adenosylhomocysteine hydrolyase (Ahcy). Epigenetic regulation of gene expression is intimately associated with both DNA and histone methylation, and these processes require methyl donors in the form of S-adenosyl methionine (SAM). Thus, the observed DEHP-dependent alterations in cellular methylation reactions linked to SAM synthesis may modulate the activities of DNA and histone methyltransferases, and they might lead to alterations in the epigenetic status of the genome. The functional consequences of these changes may promote carcinogenesis, as it has been demonstrated that dietary manipulation of compounds involved in one-carbon compound metabolism (and SAM production) are associated with DNA hypomethylation and can lead to cancer in rodent liver (Wilson et al., 1984Go).

Our studies demonstrate that, although unsupervised GO mapping and subsequent overrepresentation analysis constitute a powerful unbiased approach for holistically defining the biological pathways and processes associated with DEHP-induced proliferation, complementary supervised GO mapping approaches, perhaps guided by the initial unsupervised analysis and our a priori knowledge of those biological pathways known to be important, will provide additional complementary insights.


    CONCLUSIONS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
This study demonstrates that Gene Ontology and pathway analysis of toxicogenomic data constitute a powerful approach to determining the mechanism of action of a toxicant. Using genome-wide transcript profiling of early gene expression changes associated with DEHP treatment, a model rodent non-genotoxic carcinogen and peroxisome proliferator, we have evaluated several GO and pathway-mapping methods and shown that these complementary analysis techniques can identify, in an unbiased manner, both known and novel DEHP-induced molecular changes in the mouse liver. These include previously described changes in fatty acid metabolism and peroxisome formation–associated genes, as well as more extensive knowledge on gene expression changes in hemostasis, complement activation, and steroid hormone biosynthesis. In addition, we have identified several novel biological pathways that are perturbed by DEHP, including the ER-overload response and circadian rhythm regulators. These data provide novel insights into the mechanisms of action of DEHP and may help to elucidate the molecular basis for PP-induced hepatocarcinogenesis in the rodent liver.

In this study we have focused on early gene expression changes that may play important roles in the development of altered pathology. The persistence of critical changes that determine alteration of apoptosis and cell proliferation needs further assessment. Furthermore, the pathways identified in this study might be examined in cultured hepatocytes, including those of humans, to identify potential species differences. The application of GO-mapping approaches similar to those described here are also likely to become increasingly important for the interpretation of toxicogenomic data generated during the development, risk assessment, and regulation of novel pharmaceuticals and chemicals (Cunningham et al., 2003Go; Freeman, 2004Go; Frueh et al., 2004Go; Orphanides, 2003Go; Pettit, 2004Go).


    SUPPLEMENTARY MATERIAL
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
Supplementary Data are available online at www.toxsci.oupjournals.org.


    NOTES
 
2 These authors made an equal contribution to this work. Back


    ACKNOWLEDGMENTS
 
This work was partially supported by a CEFIC LRI grant. We thank A. Soames for the histopathology data, N. Macdonald for reprographic assistance, and R. Roberts for help and advice in the design of this study. Conflict of interest: none declared.


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 SUPPLEMENTARY MATERIAL
 REFERENCES
 
Anderson, S. P., Howroyd, P., Liu, J., Qian, X., Bahnemann, R., Swanson, C., Kwak, M.-K., Kensler, T. W., and Corton, J. C. (2004a). The transcriptional response to a peroxisome proliferator-activated receptor a agonist includes increase expression of proteome maintenance genes. J. Biol. Chem. 279, 52390–52398.[Abstract/Free Full Text]

Anderson, S. P., Dunn, C., Laughter, A., Yoon, L., Swanson, C., Stulnig, T. M., Steffensen, K. R., Chandraratna, R. A. S., Gustaffsson, J.-A., and Corton, J. C. (2004b). Overlapping transcriptional programs regulated by the nuclear receptors peroxisome proliferator-activated receptor a, retinoid X receptor, and liver X receptor in mouse liver. Mol. Pharmacol. 66, 1440–1452.[Abstract/Free Full Text]

Ashburner, M., Ball, C. A., Blake, J. A., Botstein, D., Butler, H., Cherry, J. M., Davis, A. P., Dolinski, K., Dwight, S. S., Eppig, J. T. et al. (2000). Gene ontology: Tool for the unification of biology. The Gene Ontology Consortium. Nat. Genet. 25, 25–29.[CrossRef][ISI][Medline]

Balsalobre, A., Brown, S. A., Marcacci, L., Tronche, F., Kellendonk, C., Reichardt, H. M., Schutz, G., and Schibler, U. (2000). Resetting of circadian time in peripheral tissues by glucocorticoid signaling. Science 289, 2344–2347.[Abstract/Free Full Text]

Beissbarth, T., and Speed, T. P. (2004). GOstat: Find statistically overrepresented gene ontologies within a group of genes. Bioinformatics 20, 1464–1465.[Abstract]

Bombail, V., Moggs, J. G., and Orphanides, G. (2004). Perturbation of epigenetic status by toxicants. Toxicol. Lett. 149, 51–58.[CrossRef][ISI][Medline]

Cheng, J., Sun, S., Tracy, A., Hubbell, E., Morris, J., Valmeekam, V., Kimbrough, A., Cline, M. S., Liu, G., Shigeta, R. et al. (2004). NetAffx Gene Ontology Mining Tool: A visual approach for microarray data analysis. Bioinformatics 20, 1462–1463.[Abstract]

Corton, J. C., Fan, L.-Q., Brown, S., Anderson, S. P., Bocos, C., Cattley, R. C., Mode, A., and Gustafsson J.-A. (1998). Down-regulation of cytochrome P450 2C family members and positive acute-phase response gene expression by peroxisome proliferator chemicals. Mol. Pharmacol. 54, 463–473.[Abstract/Free Full Text]

Corton, J. C., Apte, U., Anderson, S. P., Limaye, P., Yoon, L., Latendresse, J., Dunn, C., Everitt, J. I., Voss, K. A., Swanson, C. et al. (2004). Mimetics of caloric restriction include agonists of lipid-activated nuclear receptors. J. Biol. Chem. 279, 46204–46212.[Abstract/Free Full Text]

Cristillo, A. D., Heximer, S. P., Russell, L., and Forsdyke, D. R. (1997). Cyclosporin A inhibits early mRNA expression of G0/G1 switch gene 2 (G0S2) in cultured human blood mononuclear cells. DNA Cell Biol 16, 1449–1458.[ISI][Medline]

Cunningham, M. L., Bogdanffy, M. S., Zacharewski, T. R., and Hines, R. N. (2003). Workshop overview: Use of genomic data in risk assessment. Toxicol. Sci. 73, 209–215.[Abstract/Free Full Text]

Dahlquist, K. D., Salomonis, N., Vranizan, K., Lawlor S. C., and Conklin, B. R. (2002). GenMAPP, a new tool for viewing and analyzing microarray data on biological pathways. Nat. Genet. 31, 19–20.[CrossRef][ISI][Medline]

De Smaele, E., Zazzeroni, F., Papa, S., Nguyen, D. U., Jin, R., Jones, J., Cong, R., and Franzoso, G. (2001). Induction of gadd45beta by NF-kappaB downregulates pro-apoptotic JNK signalling. Nature 414, 308–313.[CrossRef][ISI][Medline]

Doniger, S. W., Salomonis, N., Dahlquist, K. D., Vranizan, K., Lawlor, S. C., and Conklin, B. R. (2003). MAPPFinder: Using Gene Ontology and GenMAPP to create a global gene-expression profile from microarray data. Genome Biol. 4, R7.[CrossRef][Medline]

Ekholm, S. V., and Reed, S. I. (2000). Regulation of G1 cyclin-dependent kinases in the mammalian cell cycle. Curr. Opin. Cell Biol.12, 676–684.[CrossRef][ISI][Medline]

Elisaf, M. (2002). Effects of fibrates on serum metabolic parameters. Curr. Med. Res. Opin. 18, 269–276.[CrossRef][ISI][Medline]

Fan, L. Q., Cattley, R. C., and Corton, J. C. (1998). Tissue-specific induction of 17 beta-hydroxysteroid dehydrogenase type IV by peroxisome proliferator chemicals is dependent on the peroxisome proliferator-activated receptor alpha. J. Endocrinol. 158, 237–246.[Abstract/Free Full Text]

Fielden, M. R., and Zacharewski, T. R. (2001). Challenges and limitations of gene expression profiling in mechanistic and predictive toxicology. Toxicol. Sci. 60, 6–10.[Abstract/Free Full Text]

Freeman, K. (2004). Toxicogenomics data: The road to acceptance. Environ. Health Perspect. 112, A678–A685.[ISI][Medline]

Frueh, F. W., Huang, S. M., and Lesko, L. J. (2004). Regulatory acceptance of toxicogenomics data. Environ. Health Perspect. 112, A663–A664.[ISI][Medline]

Fujita, N., Watanabe, S., Ichimura, T., Tsuruzoe, S., Shinkai, Y., Tachibana, M., Chiba, T., and Nakao, M. (2003). Methyl-CpG binding domain 1 (MBD1) interacts with the Suv39h1-HP1 heterochromatic complex for DNA methylation-based transcriptional repression. J. Biol. Chem. 278, 24132–24138.[Abstract/Free Full Text]

Gervois, P., Vu-Dac, N., Kleemann, R., Kockx, M., Dubois, G., Laine, B., Kosykh, V., Fruchart, J. C., Kooistra, T., and Staels, B. (2001). Negative regulation of human fibrinogen gene expression by peroxisome proliferator-activated receptor alpha agonists via inhibition of CCAAT box/enhancer-binding protein beta. J. Biol. Chem. 276, 33471–33477.[Abstract/Free Full Text]

Geuze, H. J., Murk, J. L., Stroobants, A. K., Griffith, J. M., Kleijmeer, M. J., Koster, A. J., Verkleij, A. J., Distel, B., and Tabak, H. F. (2003). Involvement of the endoplasmic reticulum in peroxisome formation. Mol. Biol. Cell 14, 2900–2907.[Abstract/Free Full Text]

Gonzalez, F. J. (2002). The peroxisome proliferator-activated receptor alpha (PPARalpha): Role in hepatocarcinogenesis. Mol. Cell Endocrinol. 193, 71–79.[CrossRef][ISI][Medline]

Hamadeh, H. K., Bushel, P. R., Jayadev, S., Martin, K., DiSorbo, O., Sieber, S., Bennett, L., Tennant, R., Stoll, R., Barrett, J. C. et al. (2002). Gene expression analysis reveals chemical-specific profiles. Toxicol. Sci. 67, 219–231.[Abstract/Free Full Text]

Hara, R., Wan, K., Wakamatsu, H., Aida, R., Moriya, T., Akiyama, M., and Shibata, S. (2001). Restricted feeding entrains liver clock without participation of the suprachiasmatic nucleus. Genes Cells 6, 269–78.[Abstract/Free Full Text]

Hasmall, S., James, N., Hedley, K., Olsen, K., and Roberts, R. (2001). Mouse hepatocyte response to peroxisome proliferators: Dependency on hepatic nonparenchymal cells and peroxisome proliferator activated receptor alpha (PPARalpha). Arch. Toxicol. 75, 357–361.[CrossRef][ISI][Medline]

Hasmall, S. C., West, D. A., Olsen, K., and Roberts, R. A. (2000). Role of hepatic non-parenchymal cells in the response of rat hepatocytes to the peroxisome proliferator nafenopin in vitro. Carcinogenesis 21, 2159–2165.[Abstract/Free Full Text]

Heinloth, A. N., Irwin, R. D., Boorman, G. A., Nettesheim, P., Fannin, R. D., Sieber, S. O., Snell, M. L., Tucker, C. J., Li, L., Travlos, G. S. et al. (2004). Gene expression profiling of rat livers reveals indicators of potential adverse effects. Toxicol. Sci. 80, 193–202.[Abstract/Free Full Text]

Hurtt, M. E., Elloit, G. S., Cook, J. C., Obourn, J. D., Frame, S. R., and Biegel, L. B. (1997). Induction of coagulation effects by Wyeth-14643 in CRL:CD® BR rats. Drug Chem. Toxicol. 20, 1–10.[ISI][Medline]

Irwin, R. D., Boorman, G. A., Cunningham, M. L., Heinloth, A. N., Malarkey, D. E., and Paules, R. S. (2004). Application of toxicogenomics to toxicology: Basic concepts in the analysis of microarray data. Toxicol. Pathol. 32 (Suppl 1), 72–83.[CrossRef]

James, N. H., Soames, A. R., and Roberts, R. A. (1998). Suppression of hepatocyte apoptosis and induction of DNA synthesis by the rat and mouse hepatocarcinogen diethylhexylphlathate (DEHP) and the mouse hepatocarcinogen 1,4-dichlorobenzene (DCB). Arch. Toxicol. 72, 784–90.[CrossRef][ISI][Medline]

Janssens, S., and Beyaert, R. (2003). Functional diversity and regulation of different interleukin-1 receptor-associated kinase (IRAK) family members. Mol. Cell 11, 293–302.[CrossRef][ISI][Medline]

Johnson, G. (2002). Scaffolding proteins—More than meets the eye. Science 295, 1249–1250.[Abstract/Free Full Text]

Karpinets, T. V., Foy, B. D., and Frazier, J. M. (2004). Tailored gene array databases: Applications in mechanistic toxicology. Bioinformatics 20, 507–517.[Abstract/Free Full Text]

Kersten, S., Mandard, S., Escher, P., Gonzalez, F. J., Tafuri, S., Desvergne, B., and Wahli, W. (2001). The peroxisome proliferator-activated receptor alpha regulates amino acid metabolism. FASEB J. 15, 1971–1978.[Abstract/Free Full Text]

Klaunig, J. E., Babich, M. A., Baetcke, K. P., Cook, J. C., Corton, J. C., David, R. M., DeLuca, J. G., Lai, D. Y., McKee, R. H., Peters, J. M. et al. (2003). PPARalpha agonist-induced rodent tumors: Modes of action and human relevance. Crit. Rev. Toxicol. 33, 655–780.[ISI][Medline]

Kockx, M., Gervois, P. P., Poulain, P., Derudas, B., Peters, J. M., Gonzalez, F. J., Princen, H. M., Kooistra, T., and Staels, B. (1999). Fibrates suppress fibrinogen gene expression in rodents via activation of the peroxisome proliferator-activated receptor-alpha. Blood 93, 2991–2998.[Abstract/Free Full Text]

Kroetz, D. L., Yook, P., Costet, P., Bianchi, P., and Pineau, T. (1998). Peroxisome proliferator-activated receptor alpha controls the hepatic CYP4A induction adaptive response to starvation and diabetes. J. Biol. Chem. 273, 31581–31589.[Abstract/Free Full Text]

Lemberger, T., Saladin, R., Vazquez, M., Assimacopoulos, F., Staels, B., Desvergne, B., Wahli, W., and Auwerx, J. (1996). Expression of the peroxisome proliferator-activated receptor alpha gene is stimulated by stress and follows a diurnal rhythm. J. Biol. Chem. 271, 1764–1769.[Abstract/Free Full Text]

Li, S., Becich, M. J., and Gilbertson, J. (2004). Microarray data mining using gene ontology. Medinfo 2004, 778–782.

Liu, G., Loraine, A. E., Shigeta, R., Cline, M., Cheng, J., Valmeekam, V., Sun, S., Kulp, D., and Siani-Rose, M. A. (2003). NetAffx: Affymetrix probe sets and annotations. Nucleic Acids Res. 31, 82–86.[Abstract/Free Full Text]

Lovett, R. A. (2000). Toxicogenomics. Toxicologists brace for genomics revolution. Science 289, 536–537.[Free Full Text]

Macdonald, N., Barrow, K., Tonge, R., Davidson, M., Roberts, R. A., and Chevalier, S. (2000). PPARa-dependent alterations of GRP95 expression in mouse hepatocytes. Biochem. Biophys. Res. Commun. 277, 699–704.[CrossRef][ISI][Medline]

Macdonald, N., Chevalier, S., Tonge, R., Davison, M., Rowlinson, R., Young, J., Rayner, S., and Roberts, R. (2001). Quantitative proteomic analysis of mouse liver response to the peroxisome proliferator diethylhexylphthalate (DEHP). Arch. Toxicol. 75, 415–424.[CrossRef][ISI][Medline]

Mandard, S., Muller, M., and Kersten, S. (2004). Peroxisome proliferator-activated receptor alpha target genes. Cell Mol. Life Sci. 61, 393–416.[CrossRef][ISI][Medline]

Matsuo, T., Yamaguchi, S., Mitsui, S., Emi, A., Shimoda, F., and Okamura, H. (2003). Control mechanism of the circadian clock for timing of cell division in vivo. Science 302, 255–259.[Abstract/Free Full Text]

McMillian, M., Nie, A. Y., Parker, J. B., Leone, A., Kemmerer, M., Bryant, S., Herlich, J., Yieh, L., Bittner, A. et al. (2004). Inverse gene expression patterns for macrophage activating hepatotoxicants and peroxisome proliferators in rat liver. Biochem. Pharmacol. 67, 2141–2167.[CrossRef][ISI][Medline]

Moggs, J. G., Tinwell, H., Spurway, T., Chang, H. S., Pate, I., Lim, F. L., Moore, D. J., Soames, A., Stuckey, R. Currie, R. et al. (2004). Phenotypic anchoring of gene expression changes during estrogen-induced uterine growth. Environ. Health Perspect. 112, 1589–1606.[ISI][Medline]

Nuwaysir, E. F., Bittner, M., Trent, J., Barrett J. C., and Afshari, C. A. (1999). Microarrays and toxicology: The advent of toxicogenomics. Mol. Carcinog. 24, 153–159.[CrossRef][ISI][Medline]

Oishi, K., Miyazaki, K., Kadota, K., Kikuno, R., Nagase, T., Atsumi, G., Ohkura, N., Azama, T., Mesaki, M., Yukimasa, S. et al. (2003). Genome-wide expression analysis of mouse liver reveals CLOCK-regulated circadian output genes. J. Biol. Chem. 278, 41519–41527.[Abstract/Free Full Text]

Oliver, J. D., and Roberts, R. A. (2002). Receptor-mediated hepatocarcinogenesis: role of hepatocyte proliferation and apoptosis. Pharmacol. Toxicol. 91, 1–7.[CrossRef][ISI][Medline]

Orphanides, G. (2003). Toxicogenomics: Challenges and opportunities. Toxicol. Lett. 140–141, 145–148.[ISI]

Patel, D. D., Knight, B. L., Wiggins, D., Humphreys, S. M., and Gibbons, G. F. (2001). Disturbances in the normal regulation of SREBP-sensitive genes in PPAR alpha-deficient mice. J. Lipid Res. 42, 328–337.[Abstract/Free Full Text]

Paules, R. (2003). Phenotypic anchoring: linking cause and effect. Environ. Health Perspect. 111, A338–A339.[ISI][Medline]

Pettit, S. D. (2004). Toxicogenomics in risk assessment: communicating the challenges. Environ. Health Perspect. 112, A662.[ISI][Medline]

Roberts, R. A., Chevalier, S., Hasmall, S. C., James, N. H., Cosulich, S. C., and Macdonald, N. (2002). PPAR alpha and the regulation of cell division and apoptosis. Toxicology 181–182, 167–170.[ISI]

Rutter, J., Reick, M., and McKnight, S. L. (2002). Metabolism and the control of circadian rhythms. Annu. Rev. Biochem. 71, 307–331.[CrossRef][ISI][Medline]

Schibler, U., and Sassone-Corsi, P. (2002). A web of circadian pacemakers. Cell 111, 919–922.[CrossRef][ISI][Medline]

Schulze, P. C., Yoshioka, J., Takahashi, T., He, Z., King, G. L., and Lee, R. T. (2004). Hyperglycemia promotes oxidative stress through inhibition of thioredoxin function by thioredoxin-interacting protein. J. Biol. Chem. 279, 30369–30374.[Abstract/Free Full Text]

Siderovski, D. P., Blum, S., Forsdyke, R. E., and Forsdyke, D. R. (1990). A set of human putative lymphocyte G0/G1 switch genes includes genes homologous to rodent cytokine and zinc finger protein-encoding genes. DNA Cell Biol. 9, 579–587.[ISI][Medline]

Szczesna-Skorupa, E., Chen, C. D., Liu, H., and Kemper, B. (2004). Gene expression changes associated with the endoplasmic reticulum stress response induced by microsomal cytochrome p450 overproduction. J. Biol. Chem. 279, 13953–13961.[Abstract/Free Full Text]

Tamayo, P., Slonim, D., Mesirov, J., Zhu, Q., Kitareewan, S., Dmitrovsky, E., Lander, E. S., and Golub, T. R. (1999). Interpreting patterns of gene expression with self-organizing maps: Methods and application to hematopoietic differentiation. Proc. Natl. Acad. Sci. U. S. A. 96, 2907–2912.[Abstract/Free Full Text]

Watson, R. E., and Goodman, J. I. (2002). Epigenetics and DNA methylation come of age in toxicology. Toxicol. Sci. 67, 11–16.[Abstract/Free Full Text]

Wilson, M. J., Shivapurkar, N., and Poirier, L. A. (1984). Hypomethylation of hepatic nuclear DNA in rats fed with a carcinogenic methyl-deficient diet. Biochem. J. 218, 987–990.[ISI][Medline]

Wong, J. S., and Gill, S. S. (2002). Gene expression changes induced in mouse liver by di(2-ethylhexyl) phthalate. Toxicol. Appl. Pharmacol. 185, 180–196.[CrossRef][ISI][Medline]

Yadetie, F., Laegreid, A., Bakke, I., Kusnierczyk, W., Komorowski, J., Waldum, H. L., and Sandvik, A. K. (2003). Liver gene expression in rats in response to the peroxisome proliferator-activated receptor-{alpha} agonist ciprofibrate Physiol. Genomics 15, 9–19.[Abstract/Free Full Text]

Yamazaki, K., Kuromitsu, J., and Tanaka, I. (2002). Microarray analysis of gene expression changes in mouse liver induced by peroxisome proliferator- activated receptor alpha agonists. Biochem. Biophys. Res. Commun. 290, 1114–1122.[CrossRef][ISI][Medline]

Zambon, A. C., McDearmon, E. L., Salomonis, N., Vranizan, K. M., Johansen, K. L., Adey, D., Takahashi, J. S., Schambelan, M., and Conklin, B. R. (2003). Time- and exercise-dependent gene regulation in human skeletal muscle. Genome Biol. 4, R61.[CrossRef][Medline]





This Article
Abstract
Full Text (PDF)
Supplementary Data
All Versions of this Article:
86/2/453    most recent
kfi207v1
Alert me when this article is cited
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Disclaimer
Request Permissions
Google Scholar
Articles by Currie, R. A.
Articles by Orphanides, G.
PubMed
PubMed Citation
Articles by Currie, R. A.
Articles by Orphanides, G.