PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of toxsciLink to Publisher's site
 
Toxicol Sci. 2009 August; 110(2): 319–333.
Published online 2009 May 29. doi:  10.1093/toxsci/kfp108
PMCID: PMC2708600

The Constitutive Active/Androstane Receptor Facilitates Unique Phenobarbital-Induced Expression Changes of Genes Involved in Key Pathways in Precancerous Liver and Liver Tumors

Abstract

Our overall goal is to elucidate progressive changes, in expression and methylation status, of genes which play key roles in phenobarbital (PB)–induced liver tumorigenesis, with an emphasis on their potential to affect signaling through critical pathways involved in the regulation of cell growth and differentiation. PB-elicited unique expression changes of genes, including some of those identified previously as exhibiting regions of altered DNA methylation, were discerned in precancerous liver tissue and/or individual liver tumors from susceptible constitutive active/androstane receptor (CAR) wild-type (WT) compared with resistant CAR knockout (KO) mice. Many of these function in crucial cancer-related processes, for example, angiogenesis, apoptosis, cell cycle, DNA methylation, Hedgehog signaling, invasion/metastasis, Notch signaling, and Wnt signaling. Furthermore, a subset of the uniquely altered genes contained CAR response elements (CAREs). This included Gadd45b, a coactivator of CAR and inhibitor of apoptosis, and two DNA methyltransferases (Dnmt1, Dnmt3a). The presence of CAREs in Dnmts suggests a potential direct link between PB and altered DNA methylation. The current data are juxtaposed with the effects of PB on DNA methylation and gene expression which occurred uniquely in liver tumor-prone B6C3F1 mice, as compared with the resistant C57BL/6, following 2 or 4 weeks of treatment. Collectively, these data reveal a comprehensive view of PB-elicited molecular alterations (i.e., changes in gene expression and DNA methylation) that can facilitate hepatocarcinogenesis. Notably, candidate genes for initial “fingerprints” of early and late stages of PB-induced tumorigenesis are proposed.

Keywords: CAR, DNA methylation, gene expression, mouse liver, tumors

Epigenetic alterations (e.g., DNA methylation), in addition to mutation, play crucial roles during the multistep/multistage process of tumorigenesis (Esteller, 2007; Loeb and Harris, 2008). Indeed, both hypermethylation and hypomethylation are critical mechanisms that facilitate tumor development (Counts and Goodman, 1995; Grønbaek et al., 2007). Furthermore, genes with diverse functions (e.g., cell cycle control, differentiation) participate in signaling that varies dramatically in cancer cells, as compared with normal cells (Weinstein, 2000). For example, physiological dependence upon (i.e., addiction to) increased levels and/or activity of specific genes can be essential for leading to and maintaining the transformed phenotype (Weinstein, 2002). Ultimately, the progressive accumulation of changes results in the acquisition of six major features of a tumor that control its malignant growth (Hanahan and Weinberg, 2000).

Our overall goal is to elucidate progressive changes, in expression and methylation status, of genes which play key roles in phenobarbital (PB)–induced liver tumorigenesis, with an emphasis on their potential to affect signaling through critical pathways involved in the regulation of cell growth and differentiation. Treatment with a 0.05% (wt/wt) promoting dose of PB, a nongenotoxic rodent hepatocarcinogen (Whysner et al., 1996), causes liver tumors in 100% of susceptible B6C3F1 and C3H/He mice after 12 months, but none in the resistant C57BL/6 mice after 18 months (Becker, 1982).

In order to shed light upon potential mechanisms involved in PB-induced tumorigenesis at very early times, we performed an extensive microarray analysis of RNA from tumor-prone B6C3F1 and relatively resistant C57BL/6 mice, treated for 2 and 4 weeks with PB, focusing upon genes whose expression was altered uniquely in the susceptible group (Phillips et al., 2009). A subset of the genes can affect angiogenesis, the epithelial-mesenchymal cell transition, and invasion/metastasis, plus the cell cycle and apoptosis, which is consistent with the observations that PB stimulates proliferation (Counts et al., 1996; Kolaja et al., 1996a), and suppresses apoptosis (Kolaja et al., 1996b). We hypothesize that at least some of these initial changes could impart selective clonal advantages to cells.

PB causes the formation of unique regions of altered DNA methylation (RAMs) in the same livers of these tumor-prone B6C3F1 mice versus the relatively resistant C57BL/6, at 2 and 4 weeks (Bachman et al., 2006). Annotation of these regions uncovered 51 genes, many of which are involved in cancer-related processes (Phillips and Goodman, 2008). Taken together, our expression and methylation analyses have provided a more comprehensive picture of unique PB-elicited molecular alterations in susceptible B6C3F1 mice which might facilitate hepatocarcinogenesis as early as 2 and 4 weeks.

The constitutive active/androstane receptor (CAR) mediates many of the effects of PB. However, microarray analysis of C3H/He CAR wild-type (WT) and knockout (KO) mice revealed that only approximately half of the PB-elicited alterations in gene expression observed were dependent upon CAR, indicating that PB also functions through CAR-independent mechanisms (Ueda et al., 2002). Importantly, CAR upregulates growth arrest and DNA-damage–inducible 45 beta (GADD45B) (Yamamoto and Negishi, 2008), a gene which can inhibit tumor necrosis factor-α/jun N-terminal kinase-induced apoptosis (Papa et al., 2004).

CAR is required for liver tumor formation elicited by PB (Huang et al., 2005; Yamamoto et al., 2004). C3H/He CAR WT mice, initiated with diethylnitrosamine (DEN), develop precancerous lesions after 23 weeks of PB promotion and tumors after 32 weeks, whereas CAR KO mice do not (Yamamoto et al., 2004). PB also induces unique RAMs in CAR WT mice (precancerous liver and/or individual liver tumors), as compared with the CAR KO (Phillips et al., 2007). Annotation of the regions that exhibited these unique methylation changes revealed 47 genes, many of which participate in crucial cancer-associated pathways (e.g., growth, angiogenesis, epithelial-mesenchymal cell transition) (Phillips and Goodman, 2009).

Here, we have expanded upon our assessment of unique methylation changes in CAR precancerous liver and/or individual liver tumors by evaluating the expression of genes identified from these RAMs, plus three DNA methyltransferases (Dnmts). Additionally, an extensive microarray analysis was performed using RNA isolated from the same tissues of the susceptible CAR WT and resistant CAR KO mice, in order to discern genes, as well as overarching cellular processes and signaling pathways, that were perturbed uniquely during later stages of PB-induced tumor development. Finally, a search for putative CAR response elements (CAREs), a subset of PB response elements, within genes whose methylation and/or expression patterns were altered uniquely in the CAR WT mice, including three Dnmt genes, was carried out.

The gene expression results presented here complement data obtained from the annotation of genes identified from unique RAMs in precancerous liver and/or liver tumors from PB-treated CAR WT mice (Phillips and Goodman, 2009), as well as the elucidation of genes whose expression and/or methylation statuses were altered uniquely in PB-treated susceptible B6C3F1 mice, at 2 and/or 4 weeks (Phillips and Goodman, 2008; Phillips et al., 2009). Combined, these data sets provide insight into progressive alterations of genes that are involved in signaling pathways and cellular processes that might function throughout a continuum of PB-induced liver tumorigenesis. Significantly, candidate genes for initial “fingerprints” of early, as well as late, stages of PB-induced tumor formation are proposed, based upon the expression and methylation changes of genes ascertained in these studies.

MATERIALS AND METHODS

Animals, Treatments, and Tissue Samples

The RNA employed for these studies was isolated from the same liver samples used by Phillips et al. (2007) for analysis of DNA methylation. These livers were originally provided by Yamamoto et al. (2004). CAR WT or CAR KO mice, on a C3H/He background (which is highly susceptible to liver tumorigenesis (Buchmann et al., 1991) were injected with a single intraperitoneal dose of DEN, 90 mg/kg, at 5 weeks of age and then administered drinking water (control) or 0.05% PB (wt/wt), a liver tumor-promoting dose, in drinking water starting at 7 weeks of age and continuing for 23 or 32 weeks, resulting in the following groups: CAR KO, 23-week control, CAR KO, 23-week PB, CAR WT, 23-week control, CAR WT, 23-week PB (precancerous liver tissue), and CAR WT, 32-week PB (individual liver tumors) (Yamamoto et al., 2004). Whole liver from four of the five (not including the CAR WT, 32-week PB) groups was utilized; the precancerous liver tissue (CAR WT, 23-week PB) contained no tumors, however, based upon histology of adjacent tissue, there are expected to be numerous microscopic foci of cellular alteration diffused throughout the tissue. Importantly, RNA was isolated from individual liver tumors that developed in the CAR WT, 32-week PB group.

RNA Isolation and Preparation

In a dounce homogenizer, frozen liver samples were homogenized in 1 ml of TRIzol Reagent (Invitrogen, Carlsbad, CA) per 100 mg tissue. Total RNA from each group (N = 6) was isolated via the manufacturer's protocol and stored at −80°C until use. Upon removal from −80°C, the RNA samples were purified with the Qiagen RNeasy Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's protocol. Samples were eluted in diethyl pyrocarbonate (DEPC)–treated water, and RNA integrity was evaluated on the Agilent 2100 Bioanalyzer using 5 ng sample aliquots and the RNA 6000 Pico Lab-on-a-Chip (Agilent Technologies, Santa Clara, CA). The presence of two distinct peaks, representing 18S and 28S rRNA levels, were indicative of high quality samples. Data from the Agilent Bioanalyzer quality control measure revealed that three RNA samples (one CAR WT, 23-week control mouse, one CAR KO 23-week control mouse, and one CAR KO, 23-week PB-treated mouse) were degraded, and these three samples were excluded from further analysis. The purity (A260/A280 ratios) and concentrations of the RNA samples were determined via the NanoDrop 8000 spectrophotometer (Thermo Scientific, Wilmington, DE).

Microarray Analysis

All procedures were performed according to standard protocols found within the Affymetrix Genechip Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA).

RNA labeling and fragmentation.

The One-Cycle Target Labeling and Control Reagent kit (Affymetrix) was utilized for first- and second- strand cDNA synthesis plus double-stranded cDNA sample cleanup, and synthesis plus cleanup of biotin-labeled cRNA, of 27 samples (N = 6, for the WT, 23-week PB-treated (precancerous liver tissue), and CAR WT, 32-week PB-treated (individual liver tumors) groups, plus N = 5 for the WT, 23-week control, KO, 23-week control, and KO, 23-week PB-treated groups). To start, 1 μg of total RNA was used for the generation of double-stranded cDNA. The cDNA was then used as a template for the synthesis of biotinylated cRNA. The size distribution and yield of the labeled cRNA products were evaluated on the Agilent 2100 Bioanalyzer using 5 ng sample aliquots and the RNA 6000 Pico Lab-on-a-Chip (Agilent Technologies). Subsequently, 15 μg of labeled cRNA was fragmented to a range of 35–200 bp in a 40 μl of volume reaction (40mM Tris-acetate at pH 8.1, 100mM potassium acetate, and 30mM magnesium acetate) at 94°C for 35 min. The size distribution of the fragmented cRNA was assessed on the Agilent 2100 Bioanalyzer using 5 ng sample aliquots and the RNA 6000 Pico Lab-on-a-Chip (Agilent Technologies).

Hybridization, washing, staining, and scanning.

Fifteen micrograms of fragmented cRNA was hybridized to a GeneChip Mouse Genome 430 2.0 Array (Affymetrix, Santa Clara, CA), containing more than 45,000 probe sets representing over 34,000 genes. The instrumentation utilized for the washing and scanning of the chips is operated by the GeneChip Operating Software (GCOS) (Affymetrix), version 3.1. After hybridization cocktails were removed, arrays were washed and stained on an Affymetrix Fluidics 450 station, and subsequently scanned using the Affymetrix GeneChip Scanner 3000 7G, in order to detect hybridization signals. From the resulting image files (DAT file), GCOS computes cell intensity data (CEL file), which is further analyzed to determine differential gene expression patterns.

Data analysis.

Data from Affymetrix GeneChip CEL files were normalized using GC Robust Multi-array Average (Wu et al., 2004). Boxplots and principle component analysis (PCA) were employed to assess data quality. A CDF file established by Dai et al. (2005), containing more accurate gene/transcript definitions (as compared with those from Affymetrix) based on up-to-date Entrez Gene information, was utilized (Mouse430, version 9.0, from (http://brainarray.mbni.med.umich.edu/Brainarray/Database/CustomCDF/CDF_download_v9.asp). Based upon these new probe set definitions, data was generated for 16,475 genes (as compared with more than 45,000 features representing over 34,000 genes, as indicated by Affymetrix). All statistical analyses were performed in R (v2.6.1) using Bioconductor (2.1).

Identification of uniquely active genes in liver tumor-susceptible CAR WT mice, as compared with the resistant CAR KO.

A “consensus” list of CAR WT, 23-weeks control genes was created in order to focus upon those that demonstrated a high degree of internal consistency of expression within the group. For each gene, the 99% confidence interval (CI) of normalized intensity values of the five samples from the CAR WT, 23-week control group was calculated. If at least four out of the five samples in the control group exhibited expression levels that fell within the 99% CI, these genes were deemed to have the potential to be differentially expressed (i.e., active) in precancerous liver tissue and/or individual liver tumors (vs. CAR WT, 23-week control). In this fashion, the original 16,475 genes derived from the CDF file (for which expression data was initially generated, see above), was reduced to a pool of 13,733. For each gene (out of the 13,733), the average expression level in the CAR WT, 23-week control group was calculated using the normalized expression values from all five control samples, and this value was compared with its expression level in each individual precancerous liver tissue sample and individual liver tumor. A gene was considered to be active in the precancerous liver tissue or individual liver tumors (vs. CAR WT, 23-week control) based upon meeting two criteria: (1) at least three, of the six total, samples in each PB-treated group exhibited expression levels that fell outside (all above or all below) the 99% CI of the control group data, and (2) it exhibited a ≥ 2-fold change in expression (up- or downregulated) as compared with the average expression level of the control group. The average expression level in precancerous liver tissue or individual liver tumors was calculated for each active gene.

In an analogous fashion, a “consensus” list of CAR KO, 23-week control group genes was created. First, for each gene, the 99% confidence interval (CI) of normalized intensity values of the five samples from the CAR KO, 23-week control group was calculated. If at least four out of the five samples in the control group exhibited expression levels that fell within the 99% CI, these genes were deemed to have the potential to be active in liver tissue from the CAR KO, 23-week PB-treated mice (vs. CAR KO, 23-week control). In this fashion, the original 16,475 genes derived from the CDF file (for which expression data was initially generated, see above), was reduced to a pool of 12,264. For each gene (out of the 12,264), the average expression level in the CAR KO, 23-week control group was calculated using the normalized expression values from all five control samples, and this value was compared with its expression level in each liver sample from the CAR KO, 23-week PB-treated group. A gene was considered to be active in the CAR KO, 23-week PB-treated group based upon meeting two criteria: (1) at least three, of the five total, samples in the PB-treated group exhibited expression levels that fell outside (all above or all below) the 99% CI of the control group, and (2) it exhibited a ≥ 2-fold change in expression (up- or downregulated) as compared with the average expression level of the control group. The average expression level in the CAR KO, 23-week PB-treated group was calculated for each active gene.

Once the active genes in the precancerous liver tissue (vs. CAR WT, 23-week control), individual liver tumors (vs. CAR WT, 23-week control), and CAR KO, 23-week PB-treated (vs. CAR KO, 23-week control) groups were ascertained, uniquely active genes in precancerous and tumor tissue, as compared with liver tissue from the CAR KO, 23-week PB-treated group, were discerned. The active genes in the precancerous liver tissue and liver tissue from the CAR KO, 23-week PB-treated group were compared with one another, and similarly, active genes in the individual liver tumors and CAR KO, 23-week PB-treated group were compared with one another. In order to focus on major differences, common genes observed in both the PB-treated CAR WT (precancerous liver tissue or individual liver tumors) and CAR KO groups that exhibited expression changes in the same direction (i.e., induced or repressed in both) were not given further consideration. Thus, the genes considered to be uniquely active in precancerous liver tissue or individual liver tumors, as compared with liver tissue from CAR KO, 23-week PB-treated, included (1) common active genes whose expression changes were opposite (e.g., induction in the precancerous liver tissue or individual liver tumors, and repression in liver from the CAR KO, 23-week PB-treated group), and (2) genes which were active only in the precancerous liver tissue or individual liver tumors. In an analogous manner, uniquely active genes in liver tissue from the CAR KO, 23-week PB-treated group, as compared with the precancerous liver tissue and individual liver tumors, were elucidated.

Finally, the uniquely active genes in precancerous liver tissue were compared with those in individual liver tumors. Genes whose expression was altered uniquely in precancerous liver tissue, as compared with individual liver tumors, included (1) those whose expression changes were opposite (e.g., induction in the precancerous tissue, and repression in the tumors), and (2) those that were active only in the precancerous tissue. Similarly, genes whose expression was altered uniquely in individual liver tumors, as compared with precancerous liver tissue, included (1) those whose expression changes were opposite (e.g., induction in the tumors, and repression in the precancerous tissue), and (2) those that were active only in the individual liver tumors. Finally, uniquely active “carry forward” genes were those whose expression changed in the same direction in both precancerous liver tissue and individual liver tumors.

Additionally, in order to be progressively more stringent, core groups of uniquely active genes were identified in precancerous liver tissue and individual liver tumors based on at least 5/6 and 6/6 samples exhibiting expression levels outside of the 99% CI of the control group data, in addition to being ≥2-fold up- or downregulated.

qRT-PCR Analysis

Expression levels of three groups of genes were investigated via quantitative real-time PCR (qRT-PCR): (1) a subset of 22 genes (listed in Supplementary Table S5) which were uniquely active in precancerous liver tissue and/or individual liver tumors, including four uniquely active carry forward genes, two genes that were active in both precancerous liver tissue and individual liver tumors that exhibited opposite expression changes, and 16 genes that were active only in individual liver tumors; (2) a subset of 16 genes (listed in Supplementary Table S8) identified from unique RAMs in precancerous liver tissue and/or individual liver tumors, as compared with 23-week PB-treated CAR KO mice, as discerned by Phillips and Goodman (2009); and (3) three Dnmt genes (Dnmt1, Dnmt3a, Dnmt3b). The genes in Groups 1 and 3, above, were chosen for analysis based on their interesting, potentially cancer related, documented functions, and the possibility that they affect pertinent signaling pathways (e.g., the three genes from Group 3 are involved in establishing DNA methylation patterns). The genes in Group 2 can also influence key pathways, and many were selected because they exhibited identical PB-induced RAMs in liver tumor-susceptible CAR WT and B6C3F1 mice (Phillips and Goodman, 2009).

Reverse transcription of RNA.

RNA, from the same samples used for microarray analysis (with the exception of 1 tumor sample, for which the RNA was entirely depleted during microarray analysis, and could not be utilized for qRT-PCR, thus the N of the tumor samples decreased from 6 to 5), were treated with amplification grade DNAse I enzyme (Invitrogen, Carlsbad, CA) to eliminate contaminating genomic DNA. Each reaction, containing 1 μg of RNA, 10X DNase I Reaction Buffer, 2 U DNase I, and DEPC-treated water (Ambion, Inc., Austin, TX) up to 10 μl, was incubated for 15 min at room temperature. DNAse I was inactivated by adding 1 μl of 25mM ethylenediaminetetraacetic acid solution and subsequently heating the reactions for 10 min at 65°C. Reverse transcription reactions containing DNAse I–treated RNA were prepared with reagents from the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Each reaction contained 1 μg RNA, 10X reverse transcription buffer, 25× deoxyribonucleotide triphosphate mix (100mM), 10× random primers, 50 U MultiScribe reverse transcriptase, and RNAse-free water, up to total 20 μl. Reactions were incubated at 25°C for 10 min, 37°C for 120 min, and 85°C for 5 s. All samples were stored at 4°C until needed.

Primers for qRT-PCR.

Primers were designed using the web-based Primer3 program, v. 0.4.0 (http://frodo.wi.mit.edu/primer3/input.htm) and synthesized by the Macromolecular Structure Facility at Michigan State University. The sizes of the amplicons ranged from 100 to 140 bp and the primers were 20 mers; all other parameters (e.g., melting temperature) remained on the default settings. The majority of primers were designed such that the amplicon spanned an intron-exon junction near the 3′ end of the gene of interest. This was confirmed using GenBank sequence information, and also, in most instances, the UCSC In-Silico PCR web-based tool (July 2007 build, http://genome.ucsc.edu/cgi-bin/hgPcr?command = start). This was done so as to preclude the possibility that the expression data could be attributed to contamination with genomic DNA. Furthermore, we are attempting to not simply confirm the microarray data but are extending this to evaluate proper splicing, that is, this represents a stringent attempt to evaluate changes in expression of functional mRNAs. Names and symbols, accession numbers, forward and reverse primer sequences of genes chosen for qRT-PCR analysis, plus amplicon sizes, are listed in Supplementary Tables S1A (selected genes that were uniquely active in precancerous liver tissue and/or individual liver tumors, based on microarray analysis) and S1B (selected genes identified from unique RAMs, plus three Dnmt genes).

qRT-PCR assays.

Reactions were prepared with Power SYBR Green PCR Master Mix (Applied Biosystems), according to the manufacturer's protocol, with each reaction containing 1 μl of cDNA from the aforementioned reverse transcription reaction (with the exception of 18S reactions, which contained 1 μl of 1:100 diluted cDNA), 1X Power SYBR® Green PCR Master Mix, 0.3μM of both the forward and reverse primers (with the exception of Dnmt3b, 0.1μM), and DEPC-treated water (Ambion, Inc.) up to 50 μl. PCR amplification of duplicate reactions was conducted in MicroAmp 96-well Optical Plates (Applied Biosystems) using the Applied Biosystems 7500 Real-Time PCR System, with the following thermal cycling conditions: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 s plus 60°C for 1 min. A dissociation protocol was run for each primer pair to ensure that a single product formed, and agarose gel electrophoresis of amplified products was performed to verify amplicon size.

Before samples were analyzed, standard curves for each gene were generated from purified amplicons. Standard curve samples (spanning 102–108 copies) for a particular gene were included on the appropriate sample plate so that the absolute quantitation method, which compares the threshold cycle of an unknown sample against a standard curve with known copy numbers, could be used to determine mRNA expression levels. The copy number of the gene of interest for each sample was standardized to that of the 18S rRNA gene to control for differences in RNA quantity, quality, and reverse transcription efficiency. Finally, fold changes in the treatment groups (vs. controls), were calculated. The following comparisons were made: (1) precancerous liver tissue (N = 6) versus CAR WT, 23-week control group (N = 5), (2) individual liver tumors (N = 5) versus CAR WT, 23-week control group, and (3) CAR KO, 23-week PB-treated group (N = 5) versus CAR KO, 23-week control group (N = 5). Statistical outliers were excluded from the final fold change calculations following their identification by the Grubbs’ test (p < 0.05, http://www.graphpad.com/quickcalcs/Grubbs1.cfm). Significance was determined via a Student's t-test (p < 0.05). A change was considered to be unique in precancerous liver tissue and/or individual liver tumors (as compared with liver tissue from the CAR KO, 23-week PB-treated group) if (1) it was in an opposite direction, or (2) it changed to a greater extent (Student's t-test, p < 0.05). However, in those cases where there was not a statistically significant change in expression in the treated group, if the expression (either up or down) of the gene in at least three samples was outside of the 95% CI of the control group, it was considered to show an “indication of a change” in expression.

Functional Analysis of Uniquely Active Genes in Precancerous Liver Tissue and/or Individual Liver Tumors

The Database for Annotation, Visualization and Integrated Discovery (DAVID), version 6 (Dennis et al., 2003; http://david.abcc.ncifcrf.gov/), was utilized to obtain functional information about the uniquely active genes in precancerous liver tissue and individual liver tumors. Functional annotation clustering (focused on Gene Ontology terms), one of several DAVID tools available, grouped genes in the precancerous liver tissue and individual liver tumors, separately, into distinct categories which were based upon cellular processes (e.g., angiogenesis) or activity (e.g., oxidoreductase activity). Furthermore, DAVID was used to overlay these genes onto pathways from the Kyoto Encyclopedia of Genes and Genomes database, thus enhancing our understanding of these data. The Pathway Studio 6.0 informatics program (Ariadne Genomics, Rockville, MD) was used to confirm selected information obtained from DAVID (e.g., multiple relationships depicted in Figures 225 and Supplementary Figs. S4S14), and to discern additional connections between genes and pathways/processes of interest.

FIG. 2.
The expression and/or methylation statuses of genes involved in the cell cycle, apoptosis, and MAPK signaling were altered in precancerous and tumor tissue. Microarray analysis was employed to identify uniquely active genes in precancerous liver tissue ...
FIG. 3.
The expression and/or methylation statuses of genes involved in angiogenesis, cytoskeleton organization/dynamics, and adhesion were altered in precancerous and tumor tissue. Microarray analysis was employed to identify uniquely active genes in precancerous ...
FIG. 4.
The expression and/or methylation statuses of genes involved in pathways associated with development and differentiation were altered in precancerous and tumor tissue. Microarray analysis was employed to identify uniquely active genes in precancerous ...
FIG. 5.
Summary of signaling pathways and processes involving genes that exhibited unique expression and/or methylation changes in precancerous and tumor tissue. PB induced unique regions of altered DNA methylation (RAMs), including hypomethylations (HypoM), ...

Identification of CAREs

CAREs are a subset of PB-responsive elements because (1) CAR-independent effects of PB have been observed (Ueda et al., 2002), and thus it is presumed that other classes of cis elements modulate those effects, and (2) there might exist elements which are necessary, but not sufficient, for transcriptional regulation by CAR. Seven bona fide CAREs were identified from the literature (Chen et al., 2003; Gerbal-Chaloin et al., 2002; Goodwin et al., 2002; Honkakoski et al., 1998; Sueyoshi et al., 1999; Sugatani et al., 2001) and used to construct a position weight matrix. The criteria for bona fide CAREs are that the elements are functional in a reporter gene assay and have been shown to bind to CAR by either gel shift or chromatin immunoprecipitation assays. Additionally, there had to be a match between the sequences of the elements reported in the literature and the most recent build, that is, compilation of genome sequence and annotation data at a particular point in time, of the human, mouse, or rat genome. The matrix similarity score threshold used was 0.80, that is, the identified CAREs were at least 80% similar to the consensus cis element.

The search for CAREs was performed within genes from the three groups described above (a subset of uniquely active genes in precancerous liver tissue and/or individual liver tumors that were selected for confirmation by qRT-PCR, a subset of genes identified from unique RAMs in precancerous liver tissue and/or individual liver tumors whose expression levels were evaluated by qRT-PCR, and three Dnmt genes). Potential/possible regulatory regions (−10,000 bp relative to the transcription start site through the 5′-untranslated region) for the genes selected for analysis were obtained from the UCSC genome database (http://genome.ucsc.edu). These regulatory regions were analyzed using previously described methods (Sun et al., 2004).

RESULTS

The complete microarray data set is presented as Supplementary Table S2A. For the CAR WT, 23-week control and CAR KO, 23-week control mice, “consensus” gene lists were made (Supplementary Tables S2B and S2C, respectively) and active genes in the PB-treated mice were identified from them. PCA revealed distinct differences in gene expression patterns, based upon treatment and strain (Supplementary Fig. S1). Generally, mice within a particular experimental group (e.g., precancerous liver tissue) exhibited similar expression patterns (Supplementary Fig. S1). The microarrays show consistent distributions, based on box plots, which is an indication that the experiments are reproducible, whereas some outlying data points for each individual mouse were observed (Supplementary Fig. S2).

Identification of Uniquely Active Genes

Active genes in the three PB-treated groups, as compared with their respective controls (precancerous liver tissue vs. WT, 23-week control; individual liver tumors vs. WT, 23-week control; and KO, 23-week PB-treated vs. KO, 23-week control), were discerned (Fig. 1). Lists of the active genes represented in the Venn diagrams (Fig. 1) can be found in Supplementary Tables S3AI.

FIG. 1.
Identification of uniquely active genes in precancerous liver tissue and/or individual liver tumors, as compared with liver tissue from tumor-resistant CAR KO mice. Microarray analysis was utilized to discern genes which were differentially expressed, ...

PB treatment altered the expression of 116 and 19 genes in precancerous liver tissue and liver tissue from CAR KO 23-week PB-treated mice, respectively (Fig. 1). Comparison of these active genes revealed 3 that exhibited the same direction of fold change (two induced and one repressed) in both groups, and therefore, these were not considered further (Supplementary Table S3A). Thus, 113 genes (88 induced and 25 repressed) were uniquely active in precancerous liver tissue (Supplementary Table S3B). Conversely, 16 genes (seven induced and nine repressed) were uniquely active in liver tissue from CAR KO, 23-week PB-treated mice, as compared with precancerous liver tissue (Supplementary Table S3C).

PB altered the expression of 891 genes in individual liver tumors (Fig. 1). When these were compared with the 19 active genes in CAR KO, 23-week PB-treated mice, three genes were in common (Fig. 1). Two exhibited fold changes that occurred in the same direction (one induced and one repressed) in both groups, and therefore, were not considered further (Supplementary Table S3D). The third gene exhibited opposite changes, and hence, was considered to be uniquely active in each group. Furthermore, the expression of 888 and 16 genes was altered in individual liver tumors, and liver tissue from CAR KO, 23-week PB-treated mice, respectively. Thus, 889 uniquely active genes (589 induced and 300 repressed) were discerned in individual liver tumors, whereas 17 (eight induced and nine repressed) were identified in liver tissue from CAR KO, 23-week PB-treated mice (Fig. 1; Supplementary Tables S3E and S3F, respectively).

Comparison of the 113 (precancerous liver tissue) and 889 (individual liver tumors) uniquely active genes revealed 77 that overlapped (Fig. 1). Sixty genes were upregulated in both groups and nine were downregulated in both groups (Supplementary Table S3G), whereas eight exhibited opposite expression changes: four were induced in precancerous liver tissue and repressed in individual liver tumors, and four were repressed in precancerous liver tissue and induced in individual liver tumors. These eight genes were considered to be uniquely active in each group (Supplementary Tables S3H and S3I). In summary, 69 uniquely active genes carried forward (i.e., were induced or repressed in both groups), whereas 44 genes were altered uniquely in precancerous liver tissue and 820 were altered uniquely in individual liver tumors (Fig. 1; Supplementary Tables S3GI). In total, PB elicited unique expression changes of 933 genes in precancerous liver tissue and/or individual liver tumors (Fig. 1).

The validity of these microarray data is substantiated by the observation that the expression of PB-induced genes, encoding drug-metabolizing enzymes and drug transporters, whose transcription is mediated by CAR, for example, Cyp2b10, Por, Abcc4 (Assem et al., 2004; Ueda et al., 2002), was upregulated in treated CAR WT, but not KO mice. Additional PB-inducible drug-metabolizing enzymes, for example, Cyp2b9 and Cyp2b13 (Nemoto and Sakurai, 1995; Stupans et al., 1984) were also upregulated only in CAR WT mice. Importantly, as would be expected, all of these genes comprised a subset of the 69 uniquely active carry forward genes (Supplementary Table S3G).

Venn diagrams, showing the numbers of genes that were uniquely active in precancerous liver tissue and/or individual liver tumors, based on at least 5/6 and 6/6 samples exhibiting expression levels outside of the 99% CI of the normalized intensity values of the control group, in addition to being ≥2-fold up- or downregulated, are located in Supplementary Figure S3, and genes are listed in Supplementary Tables S4AJ. Genes that were uniquely active in at least 5/6 precancerous liver tissue samples and 5/6 individual liver tumors, 51 and 385, respectively, were identified (Supplementary Fig. S3) and are listed in Supplementary Tables S4A and S4B. Comparison of these genes revealed 402 total active genes: 34 carried forward, whereas 17 were altered uniquely in precancerous liver tissue and 351 were altered uniquely in individual liver tumors (Supplementary Tables S4CE).

Genes that were active in all six of the precancerous liver tissue samples and all six of individual tumors, 35 and 94, respectively, were identified (Supplementary Fig. S3) and are listed in Supplementary Tables S4F and S4G. Comparison of these genes revealed 106 total active genes: 23 carried forward, whereas 12 were altered uniquely in precancerous liver tissue, and 71 were altered in individual liver tumors (Supplementary Tables S4HJ). Functional annotation and pathway analysis was performed on all 933 active genes in precancerous liver tissue and individual liver tumors, based on at least 3/6 total samples from PB-treated mice exhibiting a change. Additionally, genes that were active in 5/6 total samples are noted in the pathways (Figs. 255 and Supplementary Figs. S4S14).

Twenty-two of the 933 uniquely active genes identified by microarray analysis in the precancerous liver tissue and/or individual liver tumors were selected for confirmation of their expression statuses by qRT-PCR (Supplementary Table S5). The expression statuses of four carry forward genes were confirmed. Two genes that exhibited changes in opposite directions in the precancerous liver tissue and individual liver tumors were evaluated. The expression of Cdh1 was confirmed only in precancerous liver tissue, and that of Fgf21 was confirmed only in individual liver tumors. However, Fgf21 exhibited a statistically significant decrease in precancerous liver tissue (vs. its control). Of 16 genes that were uniquely active only in individual liver tumors, all of their expression statuses were confirmed. Although none of these was identified as being uniquely active in precancerous liver tissue based on microarray analysis, unique expression changes were detected by qRT-PCR for 4 of them (Bmp8b, Gadd45b, Mdm2, Tgfbr2), and two genes (Mapk3 and Pdgfa) exhibited indications of unique expression changes (Supplementary Table S5).

Functional Analysis of Uniquely Active Genes

Functional annotation clustering by DAVID segregated the 933 total uniquely active genes in precancerous liver tissue and/or individual liver tumors into groups based on common biological processes and molecular functions (selected examples in Supplementary Tables S6AS). Numerous uniquely active genes participate in key cellular processes and signaling pathways that can influence tumor development. The genes are depicted in pathways/processes in Figures 225 and in Supplementary Information (Figs. S4S14).

PB altered the expression of genes that can positively or negatively regulate cell cycle progression (e.g., Cdc2a, Ccnd1, Gadd45a, Gadd45b, Gmnn, Mdm2, Wee1) (Fig. 2). Additional genes demonstrate connections to the cell cycle and growth (Supplementary Fig. S4). The induction and repression of genes involved in apoptosis (e.g., Casp1, Gadd45a, Gadd45b, Gas1, Myc, and p21) was observed (Fig. 2; Supplementary Fig. S5)

Furthermore, genes linked to MAPK signaling were altered (Fig. 2; Supplementary Fig. S6). Notably, Mapk3 (also known as Erk1) was induced. Growth factors which are known to stimulate MAPK cascades (e.g., Fgf21 and Pdgfa), as well as downstream targets of MAPK activation (e.g., Ddit3, Jun, Mknk2, Myc), were also affected.

Figure 3 depicts uniquely active genes that are involved in multiple cancer-related processes: angiogenesis (e.g., Angptl6, Cxcr7, Cyr61, Esam1, Pdgfa), cytoskeleton organization/dynamics (e.g., Diap2, F2r, Gna12, Rras, Rock2), and cell-to-cell or cell-to-matrix adhesion (Cd36, Cd44, Cdh1, Icam1, Lmo7). Additional genes that contribute to these three processes are illustrated in Supplementary Figures S7S9.

Significantly, genes involved in five major developmental pathways (epithelial-mesenchymal cell transition, EMT; Hedgehog; Notch; transforming growth factor-beta, TGF-β, and Wnt) were uniquely active in precancerous liver tissue and/or individual liver tumors (Fig. 4; Supplementary Figs. S10S14). Additionally, uniquely active kinase, phosphatase and phosphatase regulatory (inhibitor) subunit genes are presented in Supplementary Table S7.

Expression of Dnmt Genes

Expression data of Dnmt1, Dnmt3a, and Dnmt3b are presented in Table 1. There was a twofold increase in the expression of Dnmt1 in individual liver tumors. Though not statistically significant, there was an indication of upregulation of Dnmt1 expression in precancerous liver tissue and in liver tissue from CAR KO, 23-week PB-treated mice, as evidenced by 3/6 total samples in the precancerous group, and 3/5 total samples in the CAR KO, 23-week PB-treated group exhibiting expression levels that fell above the 95% CI of their respective control group data. The expression of Dnmt3a was increased in both precancerous liver tissue (1.27-fold) and individual liver tumors (2.29-fold), whereas there was no expression change in the CAR KO, 23-weeks PB-treated group. The expression of Dnmt3b was not altered.

TABLE 1
qRT-PCR Analysis of the Expression of Three Dnmt Genes

Expression of Genes Identified from Unique PB-Induced RAMs

In addition to evaluating the effects of PB on global gene expression, we were also interested in assessing the expression of a subset of genes previously shown to exhibit unique PB-induced RAMs in precancerous liver tissue and/or individual liver tumors (Phillips and Goodman, 2009). qRT-PCR was employed to evaluate expression of 16 of these genes, using RNA isolated from the same tissues in which the unique RAMs were ascertained (Supplementary Table S8). In this manner, complementary data from our gene expression and DNA methylation studies were depicted (Figs. 225). Genes that exhibited only a PB-induced change in methylation,are shown in Supplementary Figures S4S5, S7, and S9.

Presence of CAREs

Thirty genes, whose expression and/or methylation statuses were altered uniquely (with the exception of Dnmt3b) in precancerous liver tissue and/or individual liver tumors, contained at least one putative CARE within 10-kb upstream of their transcriptional start sites (Table 2; base sequences of the putative CAREs are listed in Supplementary Table S9). Of these, 26 (including Gadd45a, Gadd45b, Tgfbr2, Dnmt1, Dnmt3a, and Dnmt3b) were subjected to qRT-PCR analysis (Table 1; Supplementary Tables S5 and S8).

TABLE 2
Expression of Genes that Contain Putative CAREsa,b

Comparison of Current Data with Analogous Data Obtained from B6C3F1 Mice

In a previous series of studies, we evaluated unique gene expression changes and the presence of CAREs, and performed functional annotation of genes of interest in liver tumor-prone B6C3F1 mice, as compared with the relatively resistant C57BL/6, following treatment with a tumor-promoting dose of PB for 2 and 4 weeks (Phillips et al., 2009). It is instructive to now compare these data with the analogous data reported in the current manuscript: the expression of three Dnmt genes (Table 3), genes identified from identical RAMs (Supplementary Table S10), uniquely active genes, based on microarray analysis (Supplementary Tables S11AC), the presence of CAREs (Table 2), and pathways and processes (Table 4 and Supplementary Table S12).

TABLE 3
qRT-PCR Analysis of Expression of Three Dnmt Genes in Two Types of Liver Tumor-Susceptible Mice
TABLE 4
Common Signaling Pathways and Cellular Processes Involving Genes that Exhibited Methylation and/or Expression Changes in Two Types of Liver-Tumor Susceptible Micea

DISCUSSION

Our overall goal is to elucidate progressive changes, in expression and methylation status, of genes which play key roles during PB-induced liver tumorigenesis, with an emphasis on their potential to affect signaling through critical pathways involved in the regulation of cell growth and differentiation. In a series of key studies (Phillips and Goodman, 2008, 2009; Phillips et al., 2009) plus the current one, we have discerned unique PB-elicited alterations, in DNA methylation and gene expression, that occurred in two types of liver tumor-prone mice. This was accomplished by “subtracting” out those which were also observed in their tumor-resistant counterparts. Indeed, PB induced RAMs in tissue from kidney (a nontarget organ) as well as liver tissue from relatively tumor-resistant C57BL/6 mice (Bachman et al., 2006) and tumor-resistant CAR KO mice (Phillips et al., 2007). Therefore, the use of this “subtraction approach” was necessary in order to provide novel insight regarding key molecular events that transpired during tumorigenesis across a broad time-line of PB treatment (i.e., 2 and 4 weeks, to precancerous liver tissue and, finally, individual liver tumors). This permitted us to focus on critical genes and how they might interact to contribute to tumorigenesis.

The present report expands upon our previous work involving CAR WT and KO mice, in which genes that exhibited unique PB-induced RAMs in precancerous liver tissue and/or individual liver tumors were identified (Phillips and Goodman, 2009). An extensive analysis of expression culminated in the elucidation of genes (including several identified from unique RAMs) whose transcription was altered uniquely in precancerous liver tissue and/or individual liver tumors versus liver tissue from resistant CAR KO, 23-week PB-treated mice. Remarkably, these data sets are complementary, as genes whose expression and/or methylation statuses changed uniquely in response to PB are involved in many of the same cancer-associated signaling pathways and cellular processes (Fig. 5; Supplementary Figs. S4S14).

In an analogous fashion, we have demonstrated that unique methylation (Phillips and Goodman, 2008) and gene expression (Phillips et al., 2009) changes occurred in tumor-susceptible B6C3F1 mice versus their relatively resistant C57BL/6 counterpart, at 2 and/or 4 weeks of PB treatment. Importantly, genes exhibiting unique alterations in both types of susceptible mice are involved in many of the same fundamental signaling pathways and cellular processes at both early (i.e., 2 and 4 weeks) and later (23 weeks, precancerous liver tissue and 32 weeks, individual liver tumors) times (Table 4). Early events, including DNA methylation changes, might cause a cell to become addicted to (i.e., dependent on) oncogenic signaling pathways that are necessary for transformation (Baylin and Ohm, 2006), and this strengthens support for our assertion that changes observed at 2 and 4 weeks could facilitate selective clonal expansion, which is characteristic of tumor formation (Feinberg et al., 2006). We have provided a view of the progressive nature of the multistage model of carcinogenesis, as genes whose expression (e.g., Gadd45b, Cxcr7, Kcnk1) or methylation statuses (e.g., Prickle2, Ptpro, Zscan22) changed at 2 and/or 4 weeks remained deregulated in precancerous liver tissue and/or individual liver tumors (Supplementary Table S14). Those genes whose expression changed only in individual liver tumors are listed in Supplementary Table S3I.

Indeed, genes that were altered at both the early and late time points are particularly compelling candidates for direct involvement in tumorigenesis. Furthermore, in a hypothesis driven fashion, these genes can serve as the substrate for developing a “fingerprint” for PB-like, nongenotoxic rodent hepatocarcinogens. Efforts have been made to identify gene expression signatures of nongenotoxic rodent hepatocarcinogenesis in mice (Iida et al., 2005), and rats (Fielden et al., 2007), which are not as responsive to PB-induced tumorigenesis as the susceptible mouse strains. Our approach for elucidating gene expression changes that might be useful biomarker candidates is more appropriate than what is in the literature currently because we “subtracted out” alterations that occurred in the PB-treated resistant CAR KO mice, and thus, are focusing on changes in precancerous liver tissue and/or individual liver tumors that are likely to be related directly to tumorigenesis.

The phosphorylation and de-phosphorylation of proteins is intimately involved in numerous signaling pathways. Thus, it is interesting to note that genes which encode kinases, phosphatases, protein phosphatase regulatory (inhibitor) subunits, and/or protein phosphatase activators (regulatory subunit) were uniquely active in precancerous liver tissue and/or individual liver tumors (Supplementary Table S7) and in B6C3F1 mice following 2 or 4 weeks of treatment with PB (Supplementary Table S8 in Phillips et al., 2009). In this context, it is important to reflect on how the activation of CAR is dependent on kinases and phosphatases. CAR is sequestered in the cytoplasm by HSP90 in mouse hepatocytes (Yoshinari et al., 2003) and HSP90 plus the cochaperone cytoplasmic CAR retention protein in human hepatocytes (Kobayashi et al., 2003). In response to PB treatment, CAR translocates from the cytoplasm to the nucleus, heterodimerizes with retinoid X receptor, and binds to transcriptional elements (e.g., PB-responsive enhancer modules) to affect gene expression (Honkakoski et al., 1998; Kawamoto et al., 1999; Sueyoshi et al., 1999). Specifically, PB causes recruitment of protein phosphatase 2A (PP2A) to the CAR:HSP90 cytoplasmic complex (Yoshinari et al., 2003). Okadaic acid, a phosphatase inhibitor which can inhibit PP2A, blocks the PB-induced nuclear translocation of CAR (Kawamoto et al., 1999). CAR-mediated induction of the Cyp2b10 gene can be blocked by KN-62 (a Ca2+/calmodulin-dependent kinase inhibitor) without affecting nuclear accumulation (Yamamoto et al., 2003), which suggests that CAR is activated in the nucleus by a phosphorylation event. The mechanism underlying PB-induced translocation of CAR from the cytoplasm to the nucleus appears to entail interaction of CAR with a homodimer of the membrane-associated subunit of protein phosphatase 1, PPP1R16A (R16A). The R16A residues dimerize at a region containing protein kinase A phosphorylation sites. Additionally, R16A can inhibit protein phosphatase 1-beta (PP1β), which can block CAR translocation (Sueyoshi et al., 2008). We speculate that enhancement of PP2A and/or inhibition of PP1β plays a role in the mechanism by which PB stimulates nuclear translocation of CAR.

Rodents are more prone to tumor development than humans (Rangarajan and Weinberg, 2003). However, in our opinion, the basic genes involved in tumorigenesis are likely to be largely the same in both species. Therefore, it is reasonable to hypothesize that differences in regulation of gene activity can contribute to this species disparity in susceptibility. A fundamental difference between murine and human CAR is their responsiveness to a range of agonists and inverse agonists, which might be attributable to structural variations (Stanley et al., 2006). Thus, the relative sensitivity of mice to carcinogenesis might, in part, be due to distinct regulatory mechanisms and/or structural features of mCAR. Moreover, it has been demonstrated that methylation patterns in rodent cells are less stable than those in human cells (reviewed in Goodman and Watson, 2002), so differences in epigenetic control (e.g., DNA methylation) between the species could, in part, underlie the enhanced propensity of rodents, as compared with humans, to develop cancer.

A key question is, what is responsible for the difference in susceptibility between the tumor-prone and tumor-resistant mice? Proliferation alone is not the answer, as PB treatment for 1 and 2 weeks increases levels in both the susceptible B6C3F1 and relatively resistant C57BL/6 mice (Counts et al., 1996). CAR is clearly involved, because PB-treated CAR KO mice do not develop liver tumors (Yamamoto et al., 2004). However, there appears to be other downstream events that are requisite for transformation because the induction of genes encoding drug-metabolizing enzymes, indicative of a functional CAR protein, occurred in both the B6C3F1 and C57BL/6 mice at 2 and 4 weeks (Phillips et al., 2009). Gadd45b is a likely candidate gene, as its overexpression can enhance CAR-mediated transcription (Yamamoto and Negishi, 2008). At 2 weeks of PB treatment, Gadd45b appears to be upregulated to a greater extent in B6C3F1 mice versus the C57BL/6 (Phillips et al., 2009), and it is also uniquely upregulated in precancerous liver tissue and individual liver tumors (Supplementary Table S5). Therefore, increased Gadd45b expression might be a crucial factor involved in the ability of PB to promote tumorigenesis, especially because it can function as an inhibitor of apoptosis (Papa et al., 2004) which is one of the hallmark effects of PB (Kolaja et al., 1996b; Whysner et al., 1996). Furthermore, expression of Gadd45b has been reported to lead to demethylation of DNA (Ma et al., 2009). Additional genes of interest in this regard are those that harbor CAREs and exhibit PB-elicited altered methylation and/or expression statuses (Table 2; Table 2 in Phillips et al., 2009).

Differential expression of Dnmts (Table 3: decreases in Dnmt1, 3a, and 3b at 2 and 4 weeks; increase in Dnmt3a in precancerous liver tissue; and increases in Dnmt1 and 3a in individual liver tumors) is suggestive of distinct roles that seem to be dependent on the stage of tumorigenesis. Furthermore, all 3 Dnmts contain at least 1 CARE (Table 2), which implies that CAR potentially regulates their transcription, thus suggesting a direct link between PB and altered DNA methylation. However, Dnmts in all mice are expected to contain CAREs. Therefore, other factors (e.g., single nucleotide polymorphisms in the CAREs, proteins that target CAR to DNA, coactivators and corepressors of CAR) might contribute to the differences in expression. Nevertheless, the unique PB-elicited expression changes of Dnmts in susceptible mice might influence the formation of RAMs, a subset of which could foster tumor development.

In vivo studies are essential for advancing our understanding of the complex carcinogenesis process. Numerous stromal components, such as matrix metalloproteinases, can influence malignant behavior (Marx, 2008). Moreover, inflammation-related signaling can contribute to tumorigenesis, for example, IL-6 production by macrophages facilitates DEN-initiated liver tumor formation (Naugler et al., 2007). Significantly, our use of whole liver tissue (including precancerous liver) and individual liver tumors derived from PB-treated mice enabled the detection of changes that occurred within the context of the microenvironment, and potentially, within multiple cell types.

In summary, we have performed extensive investigations (spanning six studies: Bachman et al., 2006; Phillips and Goodman, 2008, 2009; Phillips et al., 2007, 2009; and the current report) of PB-elicited molecular alterations, that is, DNA methylation and gene expression changes, that occurred uniquely in liver tumor-susceptible mice, as compared with their tumor-resistant counterparts. This approach, involving two distinct model systems comprised of PB-treated mice whose exposure periods spanned a continuum of PB-induced hepatocarcinogenesis, has proven to be effective for discerning genes that might be mechanistically critical for tumor development. As a direct result of these analyses, numerous genes are presented that could contribute to tumorigenesis by virtue of their deregulation in response to PB. Many of these genes participate in processes (Table 4, Supplementary Table S12, and listed in parentheses below) that are capable of contributing to the acquisition of the six basic features of a tumor (Hanahan and Weinberg, 2000): (1) self-sufficiency in growth signals (cell cycle, growth, survival, and MAPK signaling), (2) insensitivity to anti-growth signals (cell cycle, growth, survival, TGF-β signaling, and differentiation-related pathways, i.e., Hedgehog, Notch, Wnt), (3) evasion of apoptosis (apoptosis and survival), (4) sustained angiogenesis (angiogenesis), (5) tissue invasion and metastasis (invasion, metastasis, EMT, adhesion and cytoskeleton organization/dynamics), and (6) limitless replicative potential (cell cycle, growth, and survival), plus an inflammatory microenvironment (acute inflammatory response, in Supplementary Table S6O), as emphasized by Mantovani (2009). Significantly, DNA methylation and gene expression changes that could be utilized in the development of a “fingerprint” of PB exposure are presented (Supplementary Table S14). Collectively, these data provide novel insight regarding the detailed picture of gene expression and DNA methylation changes that underlie PB-induced liver tumorigenesis.

FUNDING

National Institutes of Health/National Institute of Environmental Health Sciences training grant (no. T32-ES-7255) predoctoral fellowship support for J.M.P.

SUPPLEMENTARY DATA

Supplementary data are available online at http://toxsci.oxfordjournals.org/.

[Supplementary Data]

Acknowledgments

Research support, in the form of an unrestricted gift, from the R.J. Reynolds Tobacco Company is acknowledged gratefully. We thank the Michigan State University Research Technology and Support Facility for technical assistance with the microarray experiments.

References

  • Assem M, Schuetz EG, Leggas M, Sun D, Yasuda K, Reid G, Zelcer N, Adachi M, Strom S, Evans RM, et al. Interactions between hepatic Mrp4 and Sult2a as revealed by the constitutive androstane receptor and Mrp4 knockout mice. J. Biol. Chem. 2004;279:22250–22257. [PubMed]
  • Bachman AN, Phillips JM, Goodman JI. Phenobarbital induces progressive patterns of GC-rich and gene-specific altered DNA methylation in the liver of tumor-prone B6C3F1 mice. Toxicol. Sci. 2006;91:393–405. [PubMed]
  • Baylin SB, Ohm JE. Epigenetic gene silencing in cancer—A mechanism for early oncogenic pathway addiction? Nat. Rev. Cancer. 2006;6:107–116. [PubMed]
  • Becker FF. Morphological classification of mouse liver tumors based on biological characteristics. Cancer Res. 1982;42:3918–3923. [PubMed]
  • Buchmann A, Bauer-Hofmann R, Mahr J, Drinkwater NR, Luz A, Schwarz M. Mutational activation of the c-Ha-ras gene in liver tumors of different rodent strains: Correlation with susceptibility to hepatocarcinogenesis. Proc. Natl. Acad. Sci. U. S. A. 1991;88:911–915. [PubMed]
  • Chen Y, Ferguson SS, Negishi M, Goldstein JA. Identification of constitutive androstane receptor and glucocorticoid receptor binding sites in the CYP2C19 promoter. Mol. Pharmacol. 2003;64:316–324. [PubMed]
  • Counts JL, Goodman JI. Alterations in DNA methylation may play a variety of roles in carcinogenesis. Cell. 1995;83:13–15. [PubMed]
  • Counts JL, Sarmiento JI, Harbison ML, Downing JC, McClain RM, Goodman JI. Cell proliferation and global methylation status changes in mouse liver after phenobarbital and/or choline-devoid, methionine-deficient diet administration. Carcinogenesis. 1996;17:1251–1257. [PubMed]
  • Dai M, Wang P, Boyd AD, Kostov G, Athey B, Jones EG, Bunney WE, Myers RM, Speed TP, Akil H, et al. Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data. Nucleic Acids Res. 2005;33:e175. [PMC free article] [PubMed]
  • Dennis G, Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA. DAVID: Database for Annotation, Visualization. Integrated Discovery. Genome Biol. 2003;4:P3. [PubMed]
  • Esteller M. Epigenetic gene silencing in cancer: the DNA hypermethylome. Hum. Mol. Genet. 2007;16(Spec No 1):R50–R59. [PubMed]
  • Feinberg AP, Ohlsson R, Henikoff S. The epigenetic progenitor origin of human cancer. Nat. Rev. Genet. 2006;7:21–33. [PubMed]
  • Fielden MR, Brennan R, Gollub J. A gene expression biomarker provides early prediction and mechanistic assessment of hepatic tumor induction by non-genotoxic chemicals. Toxicol. Sci. 2007;99:90–100. [PubMed]
  • Gerbal-Chaloin S, Daujat M, Pascussi JM, Pichard-Garcia L, Vilarem MJ, Maurel P. Transcriptional regulation of CYP2C9 gene. Role of glucocorticoid receptor and constitutive androstane receptor. J. Biol. Chem. 2002;277:209–217. [PubMed]
  • Goodman JI, Watson RE. Altered DNA methylation: A secondary mechanism involved in carcinogenesis. Annu. Rev. Pharmacol. Toxicol. 2002;42:501–525. [PubMed]
  • Goodwin B, Hodgson E, D'Costa DJ, Robertson GR, Liddle C. Transcriptional regulation of the human CYP3A4 gene by the constitutive androstane receptor. Mol. Pharmacol. 2002;62:359–365. [PubMed]
  • Grønbaek K, Hother C, Jones PA. Epigenetic changes in cancer. APMIS. 2007;115:1039–1059. [PubMed]
  • Hanahan D, Weinberg RA. The hallmarks of cancer. Cell. 2000;100:57–70. [PubMed]
  • Honkakoski P, Zelko I, Sueyoshi T, Negishi M. The nuclear orphan receptor CAR-retinoid X receptor heterodimer activates the phenobarbital-responsive enhancer module of the CYP2B gene. Mol. Cell. Biol. 1998;18:5652–5658. [PMC free article] [PubMed]
  • Huang W, Zhang J, Washington M, Liu J, Parant JM, Lozano G, Moore DD. Xenobiotic stress induces hepatomegaly and liver tumors via the nuclear receptor constitutive androstane receptor. Mol. Endocrinol. 2005;19:1646–1653. [PubMed]
  • Iida M, Anna CH, Holliday WM, Collins JB, Cunningham ML, Sills RC, Devereux TR. Unique patterns of gene expression changes in liver after treatment of mice for 2 weeks with different known carcinogens and non-carcinogens. Carcinogenesis. 2005;26:689–699. [PubMed]
  • Kawamoto T, Sueyoshi T, Zelko I, Moore R, Washburn K, Negishi M. Phenobarbital-responsive nuclear translocation of the receptor CAR in induction of the CYP2B gene. Mol. Cell. Biol. 1999;19:6318–6322. [PMC free article] [PubMed]
  • Kobayashi K, Sueyoshi T, Inoue K, Moore R, Negishi M. Cytoplasmic accumulation of the nuclear receptor CAR by a tetratricopeptide repeat protein in HepG2 cells. Mol. Pharmacol. 2003;64:1069–1075. [PubMed]
  • Kolaja KL, Stevenson DE, Johnson JT, Walborg EF, Jr, Klaunig JE. Subchronic effects of dieldrin and phenobarbital on hepatic DNA synthesis in mice and rats. Fundam. Appl. Toxicol. 1996a;29:219–228. [PubMed]
  • Kolaja KL, Stevenson DE, Walborg EF, Jr, Klaunig JE. Dose dependence of phenobarbital promotion of preneoplastic hepatic lesions in F344 rats and B6C3F1 mice: Effects on DNA synthesis and apoptosis. Carcinogenesis. 1996b;17:947–954. [PubMed]
  • Loeb LA, Harris CC. Advances in chemical carcinogenesis: A historical review and prospective. Cancer Res. 2008;68:6863–6872. [PMC free article] [PubMed]
  • Ma DK, Jang MH, Guo JU, Kitabatake Y, Chang ML, Pow-Anpongkul N, Flavell RA, Lu B, Ming GL, Song H. Neuronal activity-induced Gadd45b promotes epigenetic DNA demethylation and adult neurogenesis. Science. 2009;323:1074–1077. [PMC free article] [PubMed]
  • Mantovani A. Cancer: Inflaming metastasis. Nature. 2009;457:36–37. [PubMed]
  • Marx J. Cancer biology. All in the stroma: Cancer's Cosa Nostra. Science. 2008;320:38–41. [PubMed]
  • Naugler WE, Sakurai T, Kim S, Maeda S, Kim K, Elsharkawy AM, Karin M. Gender disparity in liver cancer due to sex differences in MyD88-dependent IL-6 production. Science. 2007;317:121–124. [PubMed]
  • Nemoto N, Sakurai J. Glucocorticoid and sex hormones as activating or modulating factors for expression of Cyp2b-9 and Cyp2b-10 in the mouse liver and hepatocytes. Arch. Biochem. Biophys. 1995;319:286–292. [PubMed]
  • Papa S, Zazzeroni F, Bubici C, Jayawardena S, Alvarez K, Matsuda S, Nguyen DU, Pham CG, Nelsbach AH, Melis T, et al. Gadd45β mediates the NF-κB suppression of JNK signalling by targeting MKK7/JNKK2. Nat. Cell. Biol. 2004;6:146–153. [PubMed]
  • Phillips JM, Goodman JI. Identification of genes that may play critical roles in phenobarbital (PB)-induced liver tumorigenesis due to altered DNA methylation. Toxicol. Sci. 2008;104:86–99. [PubMed]
  • Phillips JM, Goodman JI. Multiple genes exhibit phenobarbital (PB)-induced constitutive active/androstane receptor (CAR)-mediated DNA methylation changes during liver tumorigenesis and in liver tumors. Toxicol. Sci. 2009;108:273–289. [PMC free article] [PubMed]
  • Phillips JM, Burgoon LD, Goodman JI. Phenobarbital (PB) elicits unique, early changes in the expression of hepatic genes that affect critical pathways in tumor-prone B6C3F1 mice. Toxicol. Sci. 2009;109:193–205. [PMC free article] [PubMed]
  • Phillips JM, Yamamoto Y, Negishi M, Maronpot RR, Goodman JI. Orphan nuclear receptor constitutive active/androstane receptor-mediated alterations in DNA methylation during phenobarbital promotion of liver tumorigenesis. Toxicol. Sci. 2007;96:72–82. [PubMed]
  • Rangarajan A, Weinberg RA. Opinion: Comparative biology of mouse versus human cells: Modelling human cancer in mice. Nat. Rev. Cancer. 2003;3:952–959. [PubMed]
  • Stanley LA, Horsburgh BC, Ross J, Scheer N, Wolf CR. PXR and CAR: Nuclear receptors which play a pivotal role in drug disposition and chemical toxicity. Drug Metab. Rev. 2006;38:515–597. [PubMed]
  • Stupans I, Ikeda T, Kessler DJ, Nebert DW. Characterization of a cDNA clone for mouse phenobarbital-inducible cytochrome P-450b. DNA. 1984;3:129–137. [PubMed]
  • Sueyoshi T, Kawamoto T, Zelko I, Honkakoski P, Negishi M. The repressed nuclear receptor CAR responds to phenobarbital in activating the human CYP2B6 gene. J. Biol. Chem. 1999;274:6043–6046. [PubMed]
  • Sueyoshi T, Moore R, Sugatani J, Matsumura Y, Negishi M. PPP1R16A, the membrane subunit of protein phosphatase 1β, signals nuclear translocation of the nuclear receptor constitutive active/androstane receptor. Mol. Pharmacol. 2008;73:1113–1121. [PubMed]
  • Sugatani J, Kojima H, Ueda A, Kakizaki S, Yoshinari K, Gong QH, Owens IS, Negishi M, Sueyoshi T. The phenobarbital response enhancer module in the human bilirubin UDP-glucuronosyltransferase UGT1A1 gene and regulation by the nuclear receptor CAR. Hepatology. 2001;33:1232–1238. [PubMed]
  • Sun YV, Boverhof DR, Burgoon LD, Fielden MR, Zacharewski TR. Comparative analysis of dioxin response elements in human, mouse and rat genomic sequences. Nucleic Acids Res. 2004;32:4512–4523. [PMC free article] [PubMed]
  • Tomida S, Yanagisawa K, Koshikawa K, Yatabe Y, Mitsudomi T, Osada H, Takahashi T. Identification of a metastasis signature and the DLX4 homeobox protein as a regulator of metastasis by combined transcriptome approach. Oncogene. 2007;26:4600–4608. [PubMed]
  • Ueda A, Hamadeh HK, Webb HK, Yamamoto Y, Sueyoshi T, Afshari CA, Lehmann JM, Negishi M. Diverse roles of the nuclear orphan receptor CAR in regulating hepatic genes in response to phenobarbital. Mol. Pharmacol. 2002;61:1–6. [PubMed]
  • Weinstein IB. Disorders in cell circuitry during multistage carcinogenesis: The role of homeostasis. Carcinogenesis. 2000;4:857–864. [PubMed]
  • Weinstein IB. Addiction to oncogenes—The Achilles heal of cancer. Science. 2002;297:63–64. [PubMed]
  • Whysner J, Ross PM, Williams GM. Phenobarbital mechanistic data and risk assessment: Enzyme induction, enhanced cell proliferation, and tumor promotion. Pharmacol. Ther. 1996;71:53–91. [PubMed]
  • Wu Z, Irizarry RA, Gentleman R, Martinez-Murillo F, Spencer F. A model based background adjustment for oligonucleotide expression arrays. J. Am. Stat. Assoc. 2004;99:909–917.
  • Yamamoto Y, Kawamoto T, Negishi M. The role of the nuclear receptor CAR as a coordinate regulator of hepatic gene expression in defense against chemical toxicity. Arch. Biochem. Biophys. 2003;409:207–211. [PubMed]
  • Yamamoto Y, Moore R, Goldsworthy TL, Negishi M, Maronpot RR. The orphan nuclear receptor constitutive active/androstane receptor is essential for liver tumor promotion by phenobarbital in mice. Cancer Res. 2004;64:7197–7200. [PubMed]
  • Yamamoto Y, Negishi M. The antiapoptotic factor growth arrest and DNA-damage-inducible 45 β regulates the nuclear receptor constitutive active/androstane receptor-mediated transcription. Drug Metab. Dispos. 2008;36:1189–1193. [PMC free article] [PubMed]
  • Yoshinari K, Kobayashi K, Moore R, Kawamoto T, Negishi M. Identification of the nuclear receptor CAR:HSP90 complex in mouse liver and recruitment of protein phosphatase 2A in response to phenobarbital. FEBS Lett. 2003;548:17–20. [PubMed]

Articles from Toxicological Sciences are provided here courtesy of Oxford University Press