Multiple Genes Exhibit Phenobarbital-Induced Constitutive Active/Androstane Receptor–Mediated DNA Methylation Changes during Liver Tumorigenesis and in Liver Tumors

Abstract

The constitutive active/androstane receptor (CAR) mediates responses to the nongenotoxic rodent liver tumor promoter phenobarbital (PB), including certain gene expression changes, hepatomegaly, and tumor formation. Aberrant DNA methylation represents epigenetic events that can play multiple roles in tumorigenesis. Previously, 146 unique PB-induced regions of altered DNA methylation (RAMs) were observed in liver tumor–susceptible CAR wild-type (WT) mice (in 23 weeks, precancerous tissue, and 32 weeks, tumor tissue), as compared to the resistant knockout (KO). We believe that at least some of these might be key for tumorigenesis. In the current study, cloning and annotation of a subset (82%) of the unique RAMs revealed 47 genes exhibiting altered methylation; 17 are already implicated in cancer or related processes and, thus, we have identified 30 “new” candidate genes that might be involved in carcinogenesis due to an epigenetic alteration. These may contribute to tumor development through their involvement in angiogenesis, apoptosis, epithelial-mesenchymal cell transition, growth/survival, and invasion/migration/metastasis. We have also, previously, discerned unique PB-elicited RAMs in liver tumor-prone B6C3F1 mice, as compared to the relatively resistant C57BL/6 strain, at 2 or 4 weeks, and identified 51 genes exhibiting altered methylation. Importantly, 11 of these genes were identified from identical, unique RAMs discerned in both the sensitive B6C3F1 and CAR WT mice, thus representing an initial, potential candidate “fingerprint” which might serve as a biomarker for PB-induced tumorigenesis. These two studies reveal “new” genes whose epigenetic statuses changed uniquely in liver tumor–susceptible mice (B6C3F1 and CAR WT), as compared to their resistant counterparts (C57BL/6 and CAR KO, respectively), within a continuum of PB-induced tumorigenesis.

The constitutive active/androstane nuclear receptor (CAR) is expressed primarily in the liver and mediates transcription of drug-metabolizing enzymes, for example, cytochrome P450 (CYP) 2B10 and CYP3A11, plus NADPH-cytochrome reductase (Ueda et al., 2002), the UDP-glucuronosyltransferase UGT1A1 (Sugatani et al., 2001), and glutathione S-transferases (Huang et al., 2003). Thus, CAR plays a key role in the metabolism and excretion of xenobiotics and endobiotics (e.g., bilirubin and bile acids), in addition to xenobiotic-induced changes in energy metabolism (reviewed in Konno et al., 2008).

Phenobarbital (PB), the prototypical nongenotoxic rodent liver tumor promoter, causes liver hyperplasia and hypertrophy, and induces xenobiotic-metabolizing enzymes (Whysner et al., 1996). A promoting dose of PB increases DNA synthesis and decreases apoptosis in murine hepatocytes (Kolaja et al., 1996b), and initially stimulates hepatocyte proliferation (Counts et al., 1996; Kolaja et al., 1996a). Ha-ras mutations are infrequent in PB-induced mouse liver tumors (Fox et al., 1990; Rumsby et al., 1991), however, hypomethylation of Ha-ras (Vorce and Goodman, 1991), and increased expression occurs (Counts et al., 1997). Additionally, connexin 32 (the major gap junction-forming protein in liver) is required for promotion by PB (Moennikes et al., 2000).

PB and PB-like compounds (e.g., 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene, TCPOBOP) activate CAR (reviewed in Swales and Negishi, 2004). CAR is required for PB-induced hepatomegaly and Cyp2b10 gene expression in mouse liver (Wei et al., 2000), and chronic CAR activation in response to PB or TCPOBOP results in hepatocarcinogenesis (Huang et al., 2005). Importantly, CAR is essential for tumor promotion by PB in diethylnitrosamine (DEN)–initiated C3H/He mice (Yamamoto et al., 2004). Microarray data analysis of PB-treated CAR wild-type (WT) and knockout (KO) mice indicated that of 138 genes (out of 8736 total genes/expressed sequence tags whose expression was altered, only approximately half of these changes were CAR dependent (Ueda et al., 2002). In response to PB treatment, CAR translocates from the cytoplasm to the nucleus, heterodimerizes with retinoid X receptor, and binds to and activates transcriptional elements (e.g., PB-responsive enhancer modules) to affect gene expression (Honkakoski et al., 1998; Kawamoto et al., 1999; Sueyoshi et al., 1999). Nuclear translocation of CAR can be blocked by an inhibitor of protein phosphatase 2A (PP2A) (Kawamoto et al., 1999), whereas a subunit of protein phosphatase 1, PPP1R16A, can inhibit protein phosphatase 1-beta (PP1β), resulting in CAR translocation (Sueyoshi et al., 2008). Additionally, CAR-mediated induction of the Cyp2b10 gene can be blocked by a Ca²⁺/calmodulin-dependent kinase inhibitor, without affecting nuclear accumulation (Yamamoto et al., 2003). These results suggest that both phosphorylation and dephosphorylation events contribute to CAR activation. 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.

DNA methylation is an epigenetic mechanism regulating transcription which, when altered, may lead to tumorigenesis. For instance, hypomethylation can activate oncogenes, whereas hypermethylation can silence tumor suppressors (Esteller, 2008; Goodman and Watson, 2002). Therefore, aberrant methylation, in addition to mutation, can play critical roles during all stages of tumor formation, for example, by facilitating the progressive clonal expansion of subpopulations which possess growth advantages over neighboring cells (Goodman and Watson, 2002). Although the detailed mechanisms of PB-induced altered DNA methylation remain to be elucidated, liver tumor-sensitive B6C3F1 mice, as compared with resistant C57BL/6, appear to be “defective” with regard to the ability to preserve normal methylation patterns (Watson and Goodman, 2002). Indeed, PB causes the formation of unique regions of altered DNA methylation (RAMs) in B6C3F1, compared with C57BL/6, mice at 2 and 4 weeks of treatment (Bachman et al., 2006b). Cloning and annotation of these unique RAMs revealed changes in methylation which occurred within genes that are known to play important roles in tumorigenesis (e.g., angiogenesis, invasion, metastasis, and epithelial-mesenchymal cell transition), plus genes which had not previously been linked to cancer, thus providing insight regarding specific genes which may play a role in PB-induced tumorigenesis due to altered DNA methylation (Phillips and Goodman, 2008).

In an analogous fashion to Bachman et al. (2006b), Phillips et al. (2007) compared DNA methylation patterns in liver tumor–susceptible C3H/He CAR WT mice and resistant CAR KO mice. Unique RAMs in the livers of CAR WT mice initiated with DEN and treated with PB for 23 (precancerous tissue) or 32 (tumor tissue) weeks, compared with CAR KO mice initiated with DEN and treated with PB for 23 weeks, were identified. Methylation changes also occurred in the CAR KO, PB-treated mice, suggesting, analogous to gene expression changes observed in Ueda et al. (2002), that DNA methylation changes are both CAR dependent and independent. We hypothesize that a subset of the unique RAMs in the precancerous and tumor tissue are important for facilitating tumorigenesis. In this study, unique RAMs detected in the precancerous and tumor tissue (Phillips et al., 2007) were cloned and annotated to discern the particular genes involved and how they might contribute to tumorigenesis, for example, common cellular targets of the genes of interest were identified in order to picture how they might interact to affect critical signaling pathways, leading to key alterations in phenotype. Although Phillips and Goodman (2008) focused upon genes involved in PB-induced tumor formation at very early treatment times (i.e., 2 and 4 weeks), the current study elucidated genes involved at later times, when foci (23 weeks, PB-treated CAR WT) and tumors (32 weeks, PB-treated CAR WT) are apparent. Taken together, these two studies (the aforementioned B6C3F1-C57BL/6 study, and the current CAR study) lead to the identification of genes whose methylation statuses changed uniquely in liver tumor–susceptible mice (B6C3F1 and CAR WT), as compared with their resistant counterparts (C57BL/6 and CAR KO, respectively), within a continuum of PB-induced tumorigenesis.

MATERIALS AND METHODS

Animals, Treatments, and Tissue Samples

The DNA employed for these studies was isolated from the same liver samples used by Phillips et al. (2007), and these samples were 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) 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 tissue), and CAR WT, 32-week PB (tumor tissue) (Yamamoto et al., 2004).

Protocol Employed for the Annotation of PB-Induced Unique Rams in Precancerous and Tumor Tissue

The RAMs cloned and annotated in the current study were previously detected via an approach involving methylation-sensitive restriction digestion, arbitrarily primed PCR (AP-PCR), and capillary electrophoresis (CE) (Supplemental Fig. S1, originally from Phillips et al., 2007), a technique described in detail by Bachman et al. (2006a). The following comparisons were previously made between experimental groups: (1) the CAR KO, 23-week PB data were compared with the CAR KO, 23-week control data, and (2) both the CAR WT, 23-week PB (precancerous tissue) and CAR WT, 32-week PB (tumor tissue) data were compared with the CAR WT, 23-week control data. For each specific PCR product size that was observed in control versus PB-treated an individual Student's t-test was performed to evaluate whether or not there was a statistical (p < 0.05) difference in the peak area. Additionally, new methylations (PCR products which were observed in the PB-treated and not in control), as well as 100% hypomethylations (PCR products which were observed in control but not seen at all in PB treated), were viewed as being “significant” (Bachman et al., 2006a).

For the methylation analysis performed by Phillips et al. (2007), whole liver from four of the five (not including the WT, 32-week PB) groups was utilized; the precancerous liver tissue (WT, 23-week PB) contained no tumors, however, based upon histology of adjacent tissue, there are expected to be very numerous microscopic foci of cellular alteration diffused throughout the tissue. Importantly, DNA was isolated from individual liver tumors that developed in the WT, 32-week PB group.

The 23-week PB-treated mice are 30 weeks of age (PB treatment started when the animals were 7 weeks old), and the 32-week mice are 39 weeks of age. Thus, the mice were past the juvenile development stage and not into old age, and at an age where a reasonable degree of stability of methylation might be anticipated over a 9-week period.

Cloning and sequencing of AP-PCR products.

AP-PCR products were first cloned using an in-gel approach to identify in what regions of the genome the PB-induced unique precancerous and tumor RAMs occurred. The AP-PCR products and a 100 base pair DNA ladder (Invitrogen, Carlsbad, CA) were electrophoresed through a 3% NuSieve GTG low melting temperature agarose gel (Lonza Biosciences, Basel, Switzerland). Portions of the gel that contained PCR products within 100 base pair size ranges were excised, melted and used for in-gel cloning reactions prepared with the pGEM-T Easy Vector Kit (Promega, Madison, WI). Clones that contained PCR product inserts were purified and sequenced at the Research Technology and Support Facility at Michigan State University. Sequencing reactions were prepared using either SP6 or T7 sequencing primers (as described in the pGEM-T Easy Vector Technical Manual; Promega, Madison, WI) and subsequently run on an ABI 3730×lGenetic Analyzer.

Comparison of the sizes of cloned and sequenced AP-PCR products to the sizes of unique RAMs.

After the sequences were obtained, the sizes of the cloned products were compared with the sizes of the unique RAMs in order to determine which cloned products represented unique precancerous and tumor RAMs. In many instances, it can be confidently stated that a particular cloned product represents a single RAM and as such, the methylation status of that RAM is unambiguous. However, the raw data analysis performed to establish if a RAM occurred in a treatment group as compared with its respective control group (Phillips et al., 2007) is based upon the understanding that the ABI 3130 Genetic Analyzer capillary electrophoresis instrument does not detect PCR product sizes with 100% accuracy. Therefore, in certain instances during the analysis of the raw data, PCR product sizes were combined. Six animals per experimental group were used and restriction digestions were performed in duplicate, followed by AP-PCR, for a total of 12 reactions. If multiple PCR product sizes within two base pairs of one another displayed product in less than half of the 12 AP-PCR reactions, these products were considered to be “identical” and were subsequently combined. This procedure has implications for analysis of the cloning data. For example, two RAMs occurred uniquely in the tumor tissue as a result of RsaI/HpaII digestion: a hypomethylation at 402 bp and a carry forward new methylation at 404 bp (Supplemental Fig. S1). A carry forward methylation change is a unique RAM that was observed in both precancerous and tumor tissue. A PCR product of 404 bp was cloned, and a BLAT search showed that the product spans an intronic region within the transmembrane protein 132d (Tmem132d) gene (Table 1). Due to our basic data analysis ± 2 bp assumption, the methylation status of Tmem132d is ambiguous; it could represent the hypomethylated RAM at 402 bp or the newly methylated carry forward RAM at 404 bp.

TABLE 1.

Genes and Genomic Regions Identified from Unique PB-induced RAMs in C3H/He CAR WT (Precancerous Liver and/or Liver Tumor), as compared with Resistant PB-Treated KO Mice, were Cloned and Subjected to BLAST-like Alignment Tool (BLAT) Searches

Methylation statusa		Gene name	Gene symbol	Accession number	Genomic locationd	Normal function	Potential role in tumorigenesis
Precan.b	Tumorc
	Hypo (M)e	Cellular retinoic acid binding protein 1	Crabp1	NM_013496	1.A.ii	Retinoic acid (RA) metabolism (1)f	Methylation-associated silencing in esophageal carcinoma (2)
	Hypo (M)	NudC domain containing 3	Nudcd3	NM_173748	1.A.ii	Dynein stabilization and cell viability (3)	—
	Hypo (M)	Protein kinase C, epsilon	Prkce	NM_011104	1.A.iii	Inhibition of apoptosis (4) and cell proliferation (5)	Transformation (6) and cancer cell proliferation (7)
	Hypo (M)	p53 and DNA damage regulated 1	Pdrg1	NM_178939	1.C.ii	Facilitates ultraviolet radiation-induced cell death (8)	—
	Hypo (M)	Solute carrier family 11 (proton-coupled divalent metal ion transporters), member 2	Slc11a2	NM_008732	1.A.iii	Divalent cation transport (9)	Overexpressed during the progression of esophageal adenocarcinoma (10)
	Hypo (M)	Spermatogenesis associated 21	Spata21	NM_177867	1.B.i	Unknown	—
	Hypo (M)	Tubulin tyrosine ligase-like family, member 9 (381 bp)	Ttll9	NM_029064	1.A.ii	α-Tubulin-preferring glutamyl ligase (11)	—
	Hypo (B)	Solute carrier family 38, member 9	Slc38a9	NM_178746	1.A.iii	Unknown	—
Hypo (B)g	Hypo (B)	Annexin A4	Anxa4	NM_013471	1.A.i	Ca2+-regulated protein involved in ion conductance (12) and membrane permeability (13)	Promotes migration in a model system of renal carcinoma (14)
Hypo (B)	Hypo (B)	Coiled-coil domain containing 134	Ccdc134	NM_172428	1.A.i	Secretory protein that inhibits ERK and JNK activation (15)	Expressed in a variety of human tumor tissues (15)
Hypo (B)	Hypo (B)	Exosome component 2	Exosc2	NM_144886	1.C.i	Putative exosome component (16)	—
Hypo (M)	Hypo (M)	RIKEN cDNA 1700027D21 gene	1700027D21Rik	NM_029661	1.C.i	Unknown	—
Hypo (B)	Hypo (B)	c-abl oncogene 1, receptor tyrosine kinase	Abl1	NM_009594	1.B.ii	Cell cycle arrest (17), apoptosis (18), and DNA synthesis in response to growth factors (19)	Various oncogenes are derived from c-abl (20)
Hypo (H)	Hypo (H) or Hypo (H)h	DNA segment, Chr 13, Wayne State University 177, expressed	D13Wsu177e	NM_178605	1.B.i	Unknown	—
Hypo (H)	Hypo (H) or Hypo (H)h	HIG1 domain family, member 2A	Higd2a	NM_025933	1.A.i	Unknown	—
Hyper (M)	Hypo (B)	Dead box polypeptide 54	Ddx54	NM_028041	1.A.iii	Corepressor of nuclear receptors (21)	—
Hyper (H)	Hyper (M)	Yip1 interacting factor homolog A (S. cerevisiae)	Yif1a	NM_026553	1.A.ii	In yeast, functions in endoplasmic reticulum-to-Golgi transport (22)	—
New (M)		Branched chain aminotransferase 2, mitochondrial	Bcat2	NM_009737	1.A.ii	Metabolism of branched chain amino acid (23)	—
New (M)		Ubiquinol-cytochrome c reductase (6.4 kD) subunit	Uqcr	NM_025650	1.A.i	Unknown	—
New (M)		Zinc finger and BTB domain containing 8 opposite strand	Zbtb8os	NM_025970	1.A.iii	Unknown	—
New (B)	New (B)	Chimerin (chimaerin) 2	Chn2	NM_023543	1.A.iii	Regulation of smooth muscle cell proliferation/migration (24) and T-cell responses (25)	Inhibition of proliferation in breast cancer cells (26)
New (B)	New (B)	RIKEN cDNA 1520401A03 gene	1520401A03Rik	NM_177132	1.A.iii	Unknown	—
New (M)	New (M)	Sine oculis-related homeobox 3 homolog (Drosophila)	Six3	NM_011381	1.A.ii	Homeobox gene involved in brain (27) and ocular (28) development	—
	New (B)	A disintegrin-like and metallopeptidase (reprolysin type) with thrombospondin type 1 motif, 17	Adamts17	NM_001033877	1.A.iii	Unknown	—
	New (H)	Annexin A2	Anxa2	NM_007585	1.A.iii	Osteoblastic mineralization (29), signal transduction (30) and proliferation (31)	Expression in prostate cancer cells inhibits migration (32)
	New (H)	BTB (POZ) domain containing 11	Btbd11	NM_028709	1.A.iii	Potential role in nervous system development (33)	—
	New (H)	Chemokine (C-C- motif) receptor 4	Ccr4	NM_009916	1.A.iii	Normal immunity (34)	Tumor immunity (34)
	New (B,H)i	Calsyntenin 2	Clstn2	NM_022319	1.A.iii	Potential role in synaptic transmission (35)	—
	New (B,H)i	Folliculin	Flcn	NM_146018	1.A.iii	Potential role in energy and/or nutrient sensing (36)	Potential tumor suppressor (37)
	New (M)	Glutamate decarboxylase-like 1	Gadl1	XM_135211	1.A.iii	Unknown	—
	New (B)	Hect domain and RLD 3	Herc3	NM_028705	1.A.ii	Binds to and is regulated by ubiquitin (38)	—
	New (B)	Like-glycosyltransferase	Large	NM_010687	1.A.iii	Posttranslational protein modification (39), brain development (40)	Located within a critical region that is deleted in meningiomas (41)
	New (B)	Leucine rich repeat and Ig domain containing 2	Lingo2	NM_175516	1.A.iii	Unknown	—
	New (H,M)i	Src-related kinase lacking C- and N-terminal myristylation sites	Srms	NM_011481	1.A.i	Nonreceptor tyrosine kinase (42)	—
	New (B,H)i	Transcription factor 4	Tcf4	NM_013685	1.A.iii	Wnt signaling (43)	Expression increased in cancer (44,45)
	New (B)	Tubulin tyrosine ligase-like family, member 9 (208 bp)	Ttll9	NM_029064	1.A.ii	α-Tubulin–preferring glutamyl ligase (11)	—
	New (M)	WD repeat domain 17	Wdr17	NM_028220	1.A.iii	Potential role in eye development (46)	—
	New (H)	Zinc finger and SCAN domain containing 22	Zscan22	NM_001001447	1.A.i	Unknown	—
	New (H)	RIKEN cDNA A430078G23 gene	A430078G23Rik	NM_001033378	1.B.i	Unknown	—
	Hypo (B,H,M) and/or New (H)h,i	Protein tyrosine phosphatase, receptor type O	Ptpro	NM_011216	1.A.iii	Transmembrane protein tyrosine phosphatase (47)	Tumor suppressor (47)
New (M)	Hypo (M)	WSC domain containing 1 (407 bp)	Wscd1	NM_177618	1.A.iii	Unknown	—
Hypo (H) and/or Hyper (M)i	Hyper (M)	Dipeptidylpeptidase 10	Dpp10	NM_199021	1.A.iii	Regulation of K+-gated voltage channels (48)	—
Hypo (H) and/or Hyper (M)i	Hyper (M)	Tyrosine kinase nonreceptor 2	Tnk2	NM_016788	1.A.iii	Intracellular kinase that might be involved in growth/movement (49)	Invasion and metastasis (50)
Hypo (B) and/or Hyper (M)i	Hypo (B) and/or Hypo (M)i	Ephrin B2	Efnb2	NM_010111	1.A.iii	Vascular development, angiogenesis in adult mice (51)	Angiogenesis and invasion(51–53)
New (H)	New (H) or Hypo (H)h	Predicted gene, EG622408	EG622408	NM_001037914	1.C.i	Unknown	—
Hypo (H) or Hyper (H)h	Hypo (B,H) and/or New (M)i	Prickle-like 2 (Drosophila)	Prickle2	NM_001081146	1.B.ii	Tissue/planar polarity via the WNT/PCP signaling pathway (54)	Expression detected in gastric and uterine cancers (55)
Hypo (B) and/or Hyper (M)i	Hypo (B) and/or Hypo (M)i	Triple functional domain (PTPRF interacting)	Trio	XM_001474968	1.A.ii	Coordination of cytoskeleton remodeling (56)	Growth, invasion, tumorigenicity (57)
New (M) and/or New (H)i	Hypo (M,H) and/or New (H)h,i	Transmembrane protein 132d	Tmem132d	NM_172885	1.A.iii	Unknown	—
Hyper (H) and/or New (B)i	Hypo (B,H,M) and/or New (H)h,i	WSC domain containing 1 (359–361 bp)	Wscd1	NM_177618	1.A.iii	Unknown	—
Hypo in precan. or tumor (B/H/M)		Uncharacterizedj	5 RAMs	—	2	—	—
Hyper in precan. or tumor (B/H/M)		Uncharacterizedj	3 RAMs	—	2	—	—
New in precan. or tumor (B/H/M)		Uncharacterizedj	9 RAMs	—	2	—	—
Methylation statuses are opposite in precan versus tumor		Uncharacterizedj	1 RAM	—	2	—	—
Precan. status is ambiguoush,i		Uncharacterizedj	2 RAMs	—	2	—	—
	Tumor status is ambiguoush,i	Uncharacterizedj	3 RAMs	—	2	—	—
Hypo (B)	Hypo (M)	Multiple gene hitsk	442–445 bp	Various	3	Various	Various
New (M)		Multiple gene hitsk	462 bp	Various	3	Various	Various
	New (M)	Uncharacterizedj	491 bp	—	4	—	—

^aThe methylation status of a region of the genome as determined by Phillips et al. (2007), in the treatment group as compared with the control group: hypomethylations are decreases which were statistically significant (Student's t-test, p < 0.05) plus complete 100% decreases, hypermethylations are increases which were statistically significant (Student's t-test, p < 0.05), and new methylations were increases, in which specific PCR products formed in the treatment but not in the control group.

^bMethylation statuses of regions of the genome in precancerous tissue (23 weeks of PB treatment).

^cMethylation statuses of regions of the genome in tumor tissue (32 weeks of PB treatment).

^dGenomic location relative to the transcriptional start site.

^eThe letter in parentheses indicates which methylation-sensitive restriction enzyme (BssHII, HpaII, or MspI) led to the identification of the RAM.

^fThe numbers in parentheses correspond to references which are listed in Supplemental Table S1.

^gCarry forward RAM (listed in bold) was observed in both the precancerous and tumor tissue.

^hIndicates that the region of the genome might be represented by one of 2 RAMs, that formed via the same restriction digestion and the reason for this uncertainty is explained in the methods.

ⁱThe region of the genome might be represented by more than one RAM because different restriction digestions (e.g. HpaII, and/or MspI and/or BssHII) led to the identification of multiple RAMs represented by PCR products that aligned to the same region of the genome. With the exception of Efnb2 and Trio, “and/OR” indicates that at least 2 of these RAMs exhibited changes in the oppoite direction.

^jThe uncharacterized genomic regions that these RAMs represent are greater than 10 kb away from an annotated gene.

^kBLAT showed multiple top hits for the particular PCR product and the sequence associates with a specific repeat element.

Additionally, the use of three different methylation-sensitive enzyme pairs could also reveal multiple methylation statuses of a particular gene. In the case of dipeptidylpeptidase 10 (Dpp10), RsaI/HpaII digestion revealed a hypomethylated RAM, whereas RsaI/MspI digestion demonstrated a carry forward hypermethylated RAM in the precancerous tissue (Table 1). Thus, Dpp10 could represent one or both of these RAMs because different restriction digestions were utilized, each of which reveals information regarding the methylation patterns of different cytosines at and within the sites of primer annealing. The “methylation status” data column in Table 1 reflects these 2 situations (e.g., Hypo/New for a single gene, in a particular tissue).

Analysis of sequenced AP-PCR products.

The sequences were subjected to BLAT database searches of the mouse genome (UCSC Genome Browser, July 2007 mouse assembly (http://genome.ucsc.edu/cgi-bin/hgBlat?command=start&org=mouse) in order to ascertain in which regions of the genome the unique PB-induced precancerous and tumor RAMs occurred. The BLAT program aligns a nucleotide or amino acid sequence to an index of an entire animal genome. For DNA sequence queries, BLAT can detect sequence alignments of 95% or greater similarity of regions with lengths of 25 or more base pairs. Additional information about the genomic region/gene is listed, including, but not limited to: gene information, sequence conservation between species, GC percentage, and the location of single nucleotide polymorphisms and repeat elements. The unique RAMs were classified according to a scheme that indicates where, in relation to a gene (e.g., within an intron, within an exon, upstream of the transcriptional start site), they are located. RAMs were also categorized by chromosomal location and gene function. Gene Ontology information for Supplemental Figure S5 was obtained from http://www.geneontology.org/.

The functions of the genes identified using BLAT searches were investigated via Pathway Studio 5.0 (Ariadne Genomics, Rockville, MD). In this fashion, connections between each individual gene and other genes, cellular processes, or disease states were elucidated. For a subset of the genes, examples of these analyses are located in Supplemental Figures S6-S9. In addition, common targets and common regulators of genes identified from unique RAMs in both the precancerous and tumor tissue were discerned. Pathway Studio 5.0 was also utilized to uncover documented links between unique precancerous and tumor RAMs and cancer-related processes, including angiogenesis, apoptosis, epithelial-mesenchymal cell transition (EMT), migration/invasion/metastasis and growth and survival.

Comparison of Genes/Genomic Regions Identified from Unique PB-Induced RAMs that Formed in both CAR WT (Precancerous Liver and/or Liver Tumor) to Genes/Genomic Regions Identified from Unique PB-induced RAMs in Liver Tumor–Susceptible B6C3F1 Mice (2 and/or 4 Weeks)

We previously identified 170 total unique RAMs in livers of tumor-susceptible B6C3F1 mice treated with 0.05% (wt/wt) PB for 2 or 4 weeks, as compared with the resistant C57BL/6 stock (Bachman et al., 2006b), and PCR products representing 90 of these 170 (53%) RAMs were cloned and subjected to BLAT searches that resulted in 51 annotated genes (Phillips and Goodman, 2008). Unique B6C3F1 RAMs at 2 and 4 weeks, which corresponded to identical genes and uncharacterized regions (i.e., regions of DNA greater than 10 kb away from an annotated gene) observed in the current study, were compared with the unique CAR precancerous and tumor RAMs, in order to determine whether the methylation patterns of common genes/uncharacterized regions observed in both studies were altered similarly or differently by PB treatment. Because three different restriction digestions were utilized, multiple RAMs might represent the same gene/uncharacterized region.

The following criteria needed to be met in order for a gene/uncharacterized genomic region identified from a PB-induced RAM to be viewed as being in common between the studies: (1) the RAM must have been cloned in both the B6C3F1 (2 and/or 4 weeks PB) and CAR WT (precancerous and/or tumor) mice, and (2) the unique RAMs in the B6C3F1 and CAR WT mice must have aligned to the same region of the genome. Hypermethylated RAMs (significant increases, Student's t-test, p < 0.05) and newly methylated RAMs are considered to be increases in methylation, whereas hypomethylated RAMs (both 100% decreases, and those which are significant, Student's t-test, p < 0.05) are considered to be decreases. Genes and uncharacterized genomic regions, identified from identical, unique PB-induced RAMs that formed in both CAR WT (precancerous liver and/or liver tumor) and B6C3F1 (2 and/or 4 weeks treated), are listed in Table 3 and Supplemental Table S3, respectively.

TABLE 3.

Genes Identified from Identical, Unique PB-induced RAMs that Formed in both CAR WT (Precancerous Liver and/or Liver Tumor) and B6C3F1 (2- and/or 4-Week Treated) Mice.

Gene (genomic classification)		Methylation status
	Considered to be one RAM	B6C3F1a		CAR WTb
		2 weeks	4 weeks	Precancerous	Tumor
Bcat2: Branched chain aminotransferase 2, mitochondrial (1.A.ii: Exonic)	M464-468c		↑N (M464)d	↑N (M468)
Ddx54: Dead box polypeptide 54 (1.A.iii: Intronic)	M315-317g B312-315g		↓ (M317)e	↑ (M315)	↓ (B312)
Efnb2: Ephrin B2 (1.A.iii: Intronic)	M564-566	↓ (M566) or ↑N (M565)f	↓ (B315)	↑ (M564)	↓ (M564)
Prickle2: Prickle-like 2 (Drosophila) (1.B.ii: >2 and ≤10 kb upstream from TSS)	H310-312g	↓ (H312)	↓ (H310)	↓ (H310) or ↑ (H312)f	↓ (H312)
	B310-312g	↑N (B310)	↓ (B310)		↓ (B312)
Ptpro: Protein tyrosine phosphatase, receptor type O (1.A.iii: Intronic)	H340-343g	↓ (H341) or ↑N (H343)f			↓ (H342) or ↑N (H340)f
	B341-342g	↓ (B341)	↓ (B341)		↓ (B342)
Srms: src-related kinase lacking C-terminal regulatory tyrosine and N-terminal myristylation sites (1.A.i: Spans TSS and/or 5′ UTR)	M202-206	↑ (M202) or ↓ (M206)f			↑N (M205)
Tcf4: Transcription factor 4 (1.A.iii: Intronic)	H200g	↓ (H200)			↑N (H200)
	B200g	↑ (B200)	↓ (B200)		↑N (B200)
Tnk2: Tyrosine kinase nonreceptor 2 (1.A.iii: Intronic)	M275–276		↓ (M275)	↑ (M276)	↑ (M276)
Trio: Triple functional domain (PTPRF interacting) (1.A.ii: Exonic)	M564-566	↓ (M566) or ↑N (M565)f		↑ (M564)	↓ (M564)
Wscd1: WSC domain containing 1 (359–361 bp) (1.A.iii: Intronic)	M358–359g		↓ (M358)		↓ (M359)
	B357–358g		↑ (B357)	↑N (B357)	↓ (B358)
Zscan22: Zinc finger and SCAN domain containing 22 (1.A.i.: Spans TSS and/or 5′ UTR)	H238–239		↑N (H238 or H239)h		↑N (H238)

^aUnique RAMs that formed in livers of B6C3F1 mice upon 2 or 4 weeks of 0.05% (wt/wt) PB treatment.

^bUnique RAMs that formed in C3H/He CAR WT (precancerous or tumor) liver tissue.

^cA unique RAM (e.g., H200) is listed as a single letter, which indicates the methylation-sensitive restriction enzyme (BssHII, HpaII, or MspI) which led to the identification of the RAM, plus the size(s), in base pairs, of the unique RAMs detected. If more than 1 size is listed, (e.g., M564–566), these represent the span of the RAM sizes in the B6C3F1 and CAR WT groups, as noted in parentheses in the “methylation status” column. The rationale for considering multiple RAMs of similar sizes to be one RAM is detailed within the “Materials/Methods.”

^dThe presence of an arrow, plus the RAM size, indicates that the unique RAM was discerned in a specific group of mice. Increases in methylation (hypermethylated RAMs: those which are significant, Student's t-test, p < 0.05, and newly methylated RAMs, where PCR product formed in the treatment but not the control group) are collectively represented by an upwards arrow (↑); A^N depicts a new methylation, whereas hypermethylations remain as unlabeled upwards arrows.

^eDecreases in methylation (hypomethylated RAMs, both 100% decreases, and those which are significant, Student's t-test, p < 0.05) are collectively represented by a downwards arrow (↓).

^fThe methylation status is ambiguous (e.g., newly methylated M565 or hypomethylated M566 might represent Efnb2).

^gBecause three different restriction digestions were utilized, more than 1 RAM might represent a single gene.

^hThe gene might be represented by 1 of 2 RAMs which exhibited the same directiion of change (e.g., both are new methylations).

Cloning and sequencing revealed two observations regarding criterion number 2 (above). First, following digestion with the same methylation-sensitive restriction enzyme, PCR products occasionally formed which (1) represented distinct RAMs that differed by more than 2 bp in multiple groups (B6C3F1, 2 and/or 4 weeks, and CAR WT, precancerous liver and/or liver tumor) and (2) aligned to the same region of the genome. Second, following digestion with the same methylation-sensitive restriction enzyme, PCR products occasionally formed which (1) represented distinct RAMs that were within 2 bp of one another (e.g., unique RAMs at 315 and 317 bp) in multiple groups (B6C3F1, 2 and/or 4 weeks, and CAR WT, precancerous liver and/or liver tumor) and (2) aligned to the same region of the genome. Thus, for a particular restriction digestion, unique RAMs in the B6C3F1 and CAR WT mice that were within 2 bp of one another or more than 2 bp apart, which were represented by PCR products that aligned to the same region of the genome, were considered to be one RAM. This is evident in Table 3 and Supplemental Table S3. The following is an example of situation number 1, above. A PCR product, representing a 464-bp RAM that was identified after RsaI/MspI digestion and AP-PCR, was cloned in the B6C3F1 mice at 4 weeks, and aligned to Bcat2 (Table 3). Similarly, in the CAR precancerous tissue, a PCR product, representing a 468-bp RAM that was identified after RsaI/MspI digestion and AP-PCR, was cloned in the CAR precancerous tissue, and aligned to the same region of Bcat2 (Table 3). Although the two RAMs were not within 2 bp of one another, the sequencing results clearly demonstrated that the PCR products aligned to the same region of the genome. Therefore, these RAMs were considered to be one RAM. The following is an example of situation number 2, above. For Ddx54 (Table 3), two distinct RAMs (M315 in precancerous tissue, and M317 in the B6C3F1 mice at 4 weeks) that the cloned PCR products represent were within 2 bp of one another. Therefore, these RAMs were considered to be one RAM.

RESULTS

PCR products were cloned that represent 119 (82%) of the 146 total unique RAMs previously detected in the livers of male C3H/He CAR WT mice treated with a tumor-promoting dose of PB for 23 (precancerous tissue) or 32 (tumor tissue) weeks, as compared with resistant CAR KO mice, including 14 carry forward RAMs (i.e., those which were observed in both the precancerous and tumor tissue) (Supplemental Fig. S1) (Phillips et al., 2007). These RAMs were annotated via the BLAT sequence alignment tool, and 47 genes were discerned (Table 1). Two genes (Ttll9 and Wscd1) are listed twice because in each situation, two PCR products of different sizes aligned to two distinct, separate regions of the same gene. Literature references in Table 1 are listed in Supplemental Table S1. RAMs were also classified based on their location relative to an annotated gene (Table 1 and Supplemental Figs. S2 and S3) and chromosomal distribution (Supplemental Fig. S4). All of the annotated genes were subsequently investigated via an informatic approach in order to determine the function of the gene, in addition to any genes which are common targets and/or regulators of the unique RAMs of interest in the precancerous and tumor tissue. A functional summary of these RAMs (Supplemental Fig. S5), plus four representative pathways of individual unique precancerous and tumor RAMs, are presented as Supplemental Information (Figs. S6–S9).

Many (59 of the 75 total genes and genomic regions) annotated RAMs could unambiguously be assigned a specific methylation status (Table 1). Unambiguous carry forward hypomethylations were observed in both the precancerous and tumor tissue: annexin A4 (Anxa4), coiled-coil domain containing 134 (Ccdc134), exosome component 2 (Exosc2), RIKEN cDNA 1700027D21 gene (1700027D21Rik), and c-abl oncogene 1, receptor tyrosine kinase (Abl1). Hypomethylated RAMs that occurred only in the tumor tissue included cellular retinoic acid binding protein 1 (Crabp1), nudC domain containing 3 (Nudcd3), protein kinase C, epsilon (Prkce), p53 and DNA damage regulated 1 (Pdrg1), solute carrier family 11 (proton-coupled divalent metal ion transporters), member 2 (Slc11a2), spermatogenesis associated 21 (Spata21), tubulin tyrosine ligase-like family, member 9, 381 bp (Ttll9), solute carrier family 38, member 9 (Slc38a9), dead box polypeptide 54 (Ddx54), and WSC domain containing 1, 407 bp (Wscd1).

RAMs that were unambiguously assigned a hypermethylated status in the precancerous tissue included dead box polypeptide 54 (Ddx54) and Yip1 interacting factor homolog A (Saccharomyces cerevisiae) (Yif1a). In the tumor tissue, Yip1 interacting factor homolog A (S. cerevisiae) (Yif1a) was hypermethylated. Furthermore, unambiguous new methylations in the precancerous tissue were branched chain aminotransferase 2, mitochondrial (Bcat2), ubiquinol-cytochrome c reductase (6.4 kDa) subunit (Uqcr), zinc finger and BTB domain containing eight opposite strand (Zbtb8os), chimerin (chimaerin) 2 (Chn2), RIKEN cDNA 1520401A03 gene (1520401A03Rik), sine oculis-related homeobox 3 homolog (Drosophila) (Six3), and WSC domain containing 1, 407 bp (Wscd1). Newly methylated RAMs in the tumor tissue included chimerin (chimaerin) 2 (Chn2), RIKEN cDNA 1520401A03 gene (1520401A03Rik), sine oculis-related homeobox 3 homolog (Drosophila) (Six3), A disintegrin-like and metallopeptidase (reprolysin type) with thrombospondin type 1 motif, 17 (Adamts17), annexin A2 (Anxa2), BTB (POZ) domain containing 11 (Btbd11), chemokine (C-C- motif) receptor 4 (Ccr4), calsyntenin 2 (Clstn2), folliculin (Flcn), glutamate decarboxylase-like 1 (Gadl1), Hect domain and RLD 3 (Herc3), like-glycosyltransferase (Large), leucine rich repeat and Ig domain containing 2 (Lingo2), src-related kinase lacking C- and N-terminal myristylation sites (Srms), transcription factor 4 (Tcf4), tubulin tyrosine ligase-like family, member 9, 208 bp (Ttll9), WD repeat domain 17 (Wdr17), zinc finger and SCAN domain containing 22 (Zscan22) and RIKEN cDNA A430078G23 gene (A430078G23Rik). There were no unambiguous hypermethylated or newly methylated RAMs that carried forward from the precancerous to the tumor tissue.

As described in the methods, some genes and genomic regions, which are greater than 10 kb away from a known gene, are tentatively classified as having multiple methylation statuses in either the precancerous or tumor tissue (21%, 16 of the 75 total genes and genomic regions). Table 1 indicates the methylation statuses and particular treatment periods at which these genes displayed altered methylation. The genes were: DNA segment, Chr 13, Wayne State University 177, expressed (D13Wsu177e), HIG1 domain family, member 2A (Higd2a), protein tyrosine phosphatase, receptor type O (Ptpro), dipeptidylpeptidase 10 (Dpp10), tyrosine kinase nonreceptor 2 (Tnk2), ephrin B2 (Efnb2), predicted gene, EG622408 (EG622408), prickle-like 2 (Drosophila) (Prickle2), triple functional domain, PTPRF interacting (Trio), transmembrane protein 132d (Tmem132d), and WSC domain containing 1, 359–361 bp (Wscd1).

There were three PCR products which each associated with multiple “hits” that displayed the highest BLAT scores (i.e., numerous genomic regions showed the fewest mismatches as compared with the sequence of interest), indicating that any of one of the various regions could correspond to the PCR product sequence. Although all of the top hits for a particular PCR product represented regions on different chromosomes, the product aligned to the identical repetitive element. Therefore, these three RAMs (442–445, 462, and 491 bp) appear to be linked to a different, specific repetitive element, and we are unable to determine with certainty which genomic region the unique RAM represents. BLAT searches revealed that regions of the genome (including the aforementioned 491 bp region), represented by 24 RAMs, are uncharacterized (i.e., located more than 10 kb away from an annotated gene). Of these 24 RAMs, the methylation statuses of 18 were unambiguous and therefore could be classified as either hypo- (five RAMs), hyper- (three RAMs), or new- (ten RAMs) methylations at a particular time point. One of the 24 uncharacterized RAMs could be assigned a specific methylation in both the precancerous and tumor tissue; the changes were opposite in direction (i.e., the precancerous RAM was hypomethylated and the tumor RAM was newly methylated). Finally, for 5 out of the 24 uncharacterized RAMs, the methylation status in the precancerous tissue (two RAMs) or the tumor tissue (three RAMs) was ambiguous.

The Pathway Studio 5.0 informatic program was used to identify common targets and regulators of these genes in precancerous and tumor tissue. Figure 1 depicts the common targets of several of the genes of interest in the precancerous tissue. Genes identified from PB-induced RAMs discerned in precancerous liver (Table 1) which do not possess targets that are in common with any of the other precancerous genes are not represented in Figure 1. Common targets of the genes identified from PB-induced RAMs discerned in tumor tissue are depicted in Figure 2. For example, Cyclin D1 (Ccnd1) is a common target of Abl1, Crabp1, Prkce and Tcf4, and similarly, v-akt murine thymoma viral oncogene homolog 1 (Akt1) is a common target of Abl1, Efnb2, Prkce, and Tcf4. In this way, potential targets of multiple genes of interest were ascertained in order to determine whether common cellular processes might be affected which could more efficiently facilitate tumorigenesis. Additionally, common regulators of genes of interest in the precancerous (Supplemental Fig. S10) and tumor (Supplemental Fig. S11) tissue were identified. In order to elucidate how these genes might be contributing to tumorigenesis, interconnections between them and five key cancer-related processes (angiogenesis, apoptosis, epithelial-mesenchymal cell transition, migration/invasion/metastasis and growth/survival) were discerned (Figs. 3–7). The genes of interest that were identified in the precancerous and tumor tissue which are linked to a particular process are depicted separately within each figure. Supplemental Information (Figs. S12-S21) contains color versions of these figures (plus the corresponding legends); the colors of the RAMs represent their methylation statuses. Additionally, the relationships depicted are referenced; numbers are noted in the figures, and the citations corresponding to the numbers are listed in Supplemental Table S2. Table 2 summarizes the potential effects of these genes on tumorigenesis, based on their connections to the aforementioned processes and their methylation statuses, assuming that methylation changes affect gene expression (e.g., increases in methylation silence expression and decreases in methylation facilitate expression).

FIG. 1.

Common targets of genes identified from unique regions of altered DNA methylation (RAMs) in PB-induced precancerous liver tissue from C3H/He CAR WT mice. An informatic approach was utilized to discern targets of two or more genes identified from unique PB-induced RAMs in precancerous tissue (23 weeks). Red symbols are common targets of the unique RAMs. The arrows point away from the unique RAM and towards the common target; positive arrows () indicate that the RAM positively affects the target. Unique RAMs are hypomethylated (green) or hypermethylated (orange). A combination of colors (i.e., green and orange) depicts a RAM with an ambiguous methylation status in the precancerous tissue. A RAM with a pink center depicts a carry forward RAM from precancerous to tumor tissue. The shapes of the entities represent the specific class of molecules to which the RAM or common target belongs: extracellular proteins or nuclear receptors (), ligands (), kinases (), and transcription factors ().

FIG. 2.

Common targets of genes identified from unique regions of altered DNA methylation (RAMs) in PB-induced liver tumor tissue from C3H/He CAR WT mice. An informatic approach was utilized to discern targets of two or more genes identified from unique PB-induced RAMs in tumor tissue (32 weeks). Red symbols are common targets of the unique RAMs. The arrows point away from the unique RAM and towards the common target; positive arrows () indicate that the RAM positively affects the target, whereas negative arrows () denote a negative effect. Unique RAMs are hypomethylated (green), hypermethylated (orange), or newly methylated (blue). A RAM with a pink center depicts a carry forward RAM from precancerous to tumor tissue. The shapes of the entities represent the specific class of molecules to which the RAM or common target belongs: extracellular proteins or nuclear receptors (), ligands (), kinases (), and transcription factors ().

FIG. 3.

Genes which exhibited altered methylation uniquely in PB-induced precancerous and liver tumor tissue and which are potentially involved in the regulation of angiogenesis. An informatic approach was utilized to discern relationships between angiogenesis and unique RAMs in precancerous (A) and tumor (B) tissue. Unique RAMs are denoted by asterisks (*). Positive arrows () indicate that an entity positively affects another gene, and/or angiogenesis directly. The shapes of the entities represent the specific class of molecules to which the RAM or connecting entity belongs: extracellular proteins or nuclear receptors (), ligands (), and kinases (). Figures S12 and S13 contain color versions of (A) and (B), respectively; the colors of the RAMs represent their methylation statuses. Additionally, the relationships depicted are referenced; numbers are noted in the figures, and the citations corresponding to the numbers are listed in Supplemental Table S2.

FIG. 4.

Genes which exhibited altered methylation uniquely in PB-induced precancerous and liver tumor tissue and which are potentially involved in the regulation of apoptosis. An informatic approach was utilized to discern relationships between apoptosis and unique RAMs in precancerous (A) and tumor (B) tissue. Unique RAMs are denoted by asterisks (*). Positive arrows () indicate whether an entity positively affects another gene, and/or apoptosis directly, and the negative arrows () denote a negative effect. The shapes of the entities represent the specific class of molecules to which the RAM or connecting entity belongs: extracellular proteins or nuclear receptors (), ligands (), kinases (), and phosphatases (). Figures S14 and S15 contain color versions of (A) and (B), respectively; the colors of the RAMs represent their methylation statuses. Additionally, the relationships depicted are referenced; numbers are noted in the figures, and the citations corresponding to the numbers are listed in Supplemental Table S2.

FIG. 5.

Genes which exhibited altered methylation uniquely in PB-induced precancerous and liver tumor tissue and which are potentially involved in the regulation of EMT. An informatic approach was utilized to discern relationships between EMT and unique RAMs in precancerous (A) and tumor (B) tissue. Unique RAMs are denoted by asterisks (*). The positive arrows () indicate whether an entity positively affects another gene, and/or EMT directly; the negative arrows () denote a negative effect. The shapes of the entities from the modified section of the diagram represent the specific class of molecules to which the RAM or connecting entity belongs: extracellular proteins or nuclear receptors (), ligands (), and kinases (). Figures S16 and S17 contain color versions of (A) and (B), respectively; the colors of the RAMs represent their methylation statuses. Additionally, the relationships depicted are referenced; numbers are noted in the figures, and the citations corresponding to the numbers are listed in Supplemental Table S2. The annotated RAMs depicted have been superimposed on a modification of a schematic of the EMT program (Zvaifler, 2006).

FIG. 6.

Genes which exhibited altered methylation uniquely in PB-induced precancerous and liver tumor tissue and which are potentially involved in the regulation of migration, invasion and/or metastasis. An informatic approach was utilized to discern relationships between migration/invasion/metastasis and unique RAMs in precancerous (A) and tumor (B) tissue. Unique RAMs are denoted by asterisks (*). The positive arrows () indicate whether an entity positively affects another gene, and/or migration/invasion/metastasis directly, and the negative arrows () denote a negative effect. The shapes of the entities represent the specific class of molecules to which the RAM or connecting entity belongs: extracellular proteins or nuclear receptors (), ligands (), transcription factors (), and kinases (). Figures S18 and S19 contain color versions of (A) and (B), respectively; the colors of the RAMs represent their methylation statuses. Additionally, the relationships depicted are referenced; numbers are noted in the figures, and the citations corresponding to the numbers are listed in Supplemental Table S2.

FIG. 7.

Genes which exhibited altered methylation uniquely in PB-induced precancerous and liver tumor tissue and which are potentially involved in the regulation of growth/survival. An informatic approach was utilized to discern relationships between growth/survival and unique RAMs in precancerous (A) and tumor (B) tissue. Unique RAMs are denoted by asterisks (*). The positive arrows () indicate whether an entity positively affects another gene, and/or growth/survival directly, and the negative arrows () denote a negative effect. The shapes of the entities represent the specific class of molecules to which the RAM or connecting entity belongs: extracellular proteins or nuclear receptors (), ligands (), kinases (), phosphatases (), and transcription factors (). Figures S20 and S21 contain color versions of (A) and (B), respectively; the colors of the RAMs represent their methylation statuses. Additionally, the relationships depicted are referenced; numbers are noted in the figures, and the citations corresponding to the numbers are listed in Supplemental Table S2.

TABLE 2.

Genes Identified from Unique PB-induced RAMs in C3H/He CAR WT (Precancerous Liver and/or Liver Tumor), as compared with Resistant CAR KO Mice, and their Potential Functional Significance in Tumorigenesis

Cell process or genea	Methylation status in precancerousb	Methylation status in tumorc	Potential effect of the RAM on indicated processd
			Precancerous	Tumor
Angiogenesis
Abl1e	Carry forward HypoM	Carry forward HypoM	↑f	↑
Efnb2	Carry forward HypoM and/or HyperM	Carry forward HypoM and/or HypoM	↑ or ↓g	↑
Prkce		HypoM		↑
Apoptosis
Abl1	Carry forward HypoM	Carry Forward HypoM	↓	↓
Efnb2	Carry forward HypoM and/or HyperM	Carry forward HypoM and/or HypoM	↓ or ↑	↓
Tnk2	HypoM and/or carry forward HyperM	Carry forward HyperM	↓ or ↑	↑
Prkce		HypoM		↓
Ptpro		HypoM and/or NewM		↑ or ↓
Epithelial-mesenchymal transition
Abl1	Carry forward HypoM	Carry Forward HypoM	↑	↑
Ddx54	HyperM	HypoM	↓	↑
Efnb2	Carry forward HypoM and/or HyperM	Carry forward HypoM and/or HypoM	↑ or ↓	↑
Prkce		HypoM		↑
Slc11a2		HypoM		↑
Growth/survival
Abl1	Carry forward HypoM	Carry Forward HypoM	↑	↑
Chn2	NewM	NewM	↑	↑
Efnb2	Carry forward HypoM and/or HyperM	Carry forward HypoM and/or HypoM	↑ or ↓	↑
Tnk2	HypoM and/or carry forward HyperM	Carry forward HyperM	↑ or ↓	↓
Crabp1		HypoM		↑
Nudcd3		HypoM		↑
Prkce		HypoM		↑
Ptpro		HypoM and/or NewM		↓ or ↑
Slc11a2		HypoM		↑
Tcf4		NewM		↑
Migration/invasion/metastasis
Abl1	Carry forward HypoM	Carry forward HypoM	↑	↑
Anxa4	Carry forward HypoM	Carry forward HypoM	↑	↑
Chn2	NewM	NewM	↑	↑
Efnb2	Carry forward HypoM and/or HyperM	Carry forward HypoM and/or HypoM	↑ or ↓	↑
Tnk2	HypoM and/or carry forward HyperM	Carry forward HyperM	↑ or ↓	↓
Trio	HyperM and/or carry forward HypoM	HypoM and/or carry forward HypoM	↓ or ↑	↑
Anxa2		NewM		↑
Prkce		HypoM		↑
Slc11a2		HypoM		↑
Tcf4		NewM		↑

^aGenes which were identified from one or more unique RAMs discerned in precancerous and tumor tissue.

^bMethylation statuses of genes in precancerous tissue (23 weeks of PB treatment).

^cMethylation statuses of genes in tumor tissue (32 weeks of PB treatment).

^dBased on methylation status(es) of the genes (assuming that increases in methylation silence expression and decreases in methylation facilitate expression), and the relationships between the genes and a cellular process (Figs. 3–7, and Supplemental Figs. S12-S21).

^eSpecific relationships between a gene and cellular process are depicted in Figures 3–7 and Supplemental Figures S12-S21, and the corresponding references are listed in Supplemental Table S2.

^fA ↑ symbol indicates a potential positive effect (including positive regulation, and inhibition of negative regulation) of the RAM on a cellular process.

^gA ↓ symbol indicates a potential negative effect (including negative regulation, and inhibition of positive regulation) of the RAM on a cellular process.

As described in the Methods, unique B6C3F1 RAMs at 2 and 4 weeks (identified in Bachman et al., 2006b; annotated in Phillips and Goodman, 2008), which corresponded to identical genes and uncharacterized genomic regions (i.e., regions of DNA greater than 10 kb away from an annotated gene observed in the current study, were compared with the unique CAR precancerous and tumor RAMs, in order to determine whether the methylation patterns of common genes/genomic regions seen in both studies were altered similarly or differently by PB treatment. Eleven genes (Bcat2, Ddx54, Efnb2, Prickle2, Ptpro, Srms, Tcf4, Tnk2, Trio, Wscd1, and Zscan22), identified from unique RAMs, were observed in both groups of mice (Table 3). Five genes were identified from RAMs whose methylation statuses either clearly increase (Bcat2: M464-468 and Zscan22: H238–239) or clearly decrease (Ddx54:B312-315, Ptpro: B341–342 and Wscd1: M358–359) in both the B6C3F1 and CAR WT mice. Three genes were identified from RAMs whose methylation statuses are clearly opposite (e.g., increases in one group and decreases in another) in at least 1 B6C3F1 group and one CAR WT group: Ddx54 (M315-317), Tcf4 (H200) and Tnk2 (M275–276). Prickle2 (B310-312), and Tcf4 (B200) were identified from RAMs in the B6C3F1 whose methylation statuses increase at 2 weeks, decrease at 4 weeks, and either increase (Tcf4) or decrease (Prickle2) in the CAR tumor tissue. Wscd1 (B357–358) was identified from RAMs in the CAR precancerous and tumor tissue whose methylation statuses increase and decrease, respectively, and also increase in the B6C3F1 at 4 weeks. Finally, Efnb2 (M564–566), Prickle2 (H310-312), Ptpro (H340–343), Srms (M202–206), and Trio (M564–566) were identified from a RAM in at least one group whose methylation status is unclear after “combining” multiple RAMs (i.e., two RAMs formed in the same group and the methylation changes are opposite in direction), and thus, it is uncertain whether the altered methylation patterns in these five genes, identified from five RAMs, are similar across groups of B6C3F1 and CAR WT mice.

Six uncharacterized genomic regions on 6 different chromosomes (271–272, 288, 324, 370, 378–379, and 575–576 bp), identified from unique RAMs, were observed in both groups of mice (Supplemental Table S3). Two regions were identified from RAMs whose methylation statuses clearly decrease (271–272 bp: M270–272 and 378–379 bp: M378-380) in both the B6C3F1 and CAR WT mice. Two regions were identified from RAMs whose methylation statuses are clearly opposite (i.e., increases in one group and decreases in another) in at least one B6C3F1 group and one CAR WT group: 288 bp (M286) and 324 bp (M323–326). Regions 271–272 bp (B269-272) and 575–576 bp (M577–579) were identified from RAMs in the B6C3F1 whose methylation statuses are opposite at 2 and 4 weeks, and either increase (271–272 bp) or decrease (575–576 bp) in the CAR tumor tissue. Finally, regions 272–271 bp (H269–272) and 370 bp (B368–370) were identified from a RAM in one group whose methylation status is unclear after “combining” multiple RAMs (e.g., two RAMs formed in the same group and the methylation changes are opposite in direction), and thus, it is uncertain whether the altered methylation in these two cases is similar across groups of B6C3F1 and CAR WT mice.

DISCUSSION

The methodology employed in this study to evaluate PB-induced altered methylation represents an unbiased approach in the sense that the genes/genomic regions evaluated are not predetermined, that is, methylation changes are detected first, followed by the identification of the genes/genomic regions exhibiting the changes. RAMs were detected using DNA isolated from whole liver tissue for four of the experimental groups (CAR WT 23-week control, CAR KO 23-week control, CAR WT 23-week PB (precancerous) and CAR KO 23-week PB). DNA from individual liver tumors was analyzed for the fifth group, CAR WT 32-week PB. It is vital to focus upon whole animal studies, as compared with in vitro models only. Firstly, interactions between various cell types are likely to be essential for tumorigenesis, and whole animals also possess an intact immune system, for example, secretion of IL-6 by macrophages facilitates DEN-induced hepatocarcinogenesis (Naugler et al., 2007) and stromal fibroblasts can promote tumor growth (Orimo et al., 2005). Secondly, the origin of liver tumor cells is unclear. Cancer cell precursors are likely hepatocytes, which can proliferate in response to treatment with mitogens such as PB (Counts et al., 1996; Kolaja et al., 1996a), or partial hepatectomy (Steer, 1995), and adult human (Takahashi et al., 2007) and murine embryonic (Wernig et al., 2007) fibroblasts can be reprogrammed, producing induced pluripotent stem (iPS) cells. Importantly, iPS cells were generated from adult murine hepatocytes following transduction with only four genes, Oct 3/4, Sox2, Klf4, and c-Myc (Aoi et al., 2008). The reprogramming of adult pancreatic exocrine cells into induced β-cells via expression of three transcription factors: Ngn3, Pdx1, and Mafa (Zhou et al., 2008) provides a basis for speculating that liver cancer progenitor cells could arise directly from normal liver cells without reverting to a stem cell state. Additionally, PB treatment can inhibit hepatocyte proliferation, providing an opportunity for those cells which “lose” responsiveness to this inhibition to, possibly, form the initial clones of precancerous cells (reviewed in Counts et al., 1996). The suggestion that hepatocytes are liver tumor cell precursors is not incompatible with the notion that tumor cells might originate from stem cell transformation (Reya et al., 2001).

A limited number of genes have been associated with mouse liver tumorigenesis, that is, CAR (Huang et al., 2005; Yamamoto et al., 2004), β-catenin (Aydinlik et al., 2001; Strathmann et al., 2006), c-Myc (Vorce and Goodman, 1991), Ha-ras (Counts et al., 1997; Vorce and Goodman, 1991; Wiseman et al., 1986), Ki-ras (Vorce and Goodman, 1991), and Raf (Ray et al., 1994). Previously, 146 total unique RAMs were observed in liver tumor–susceptible PB-treated CAR WT mice (precancerous and tumor tissue), as compared with their resistant KO counterpart and we suggested that, at least some of, these changes might be playing critical roles in the carcinogenesis process (Phillips et al., 2007). In this study, cloning and annotation of a subset (82%) of these unique RAMs revealed 47 genes which might contribute to tumorigenesis due to their altered DNA methylation statuses; 17 have previously been implicated in cancer or related processes (Table 1). Thus, we have identified 30 “new” candidate genes which might be involved in PB-induced carcinogenesis.

We recently identified 170 unique RAMs in liver tumor-prone B6C3F1, as compared with relatively resistant C57BL/6, mice upon treatment with 0.05% PB (wt/wt) for 2 or 4 weeks (Bachman et al., 2006b); subsequent cloning and annotation of these RAMs uncovered 51 genes that exhibited altered methylation (Phillips and Goodman, 2008). The combined data from these two model systems have revealed the formation of unique RAMs in liver tumor-sensitive mice at early (i.e., 2 and 4 weeks) as well as later (i.e., 23 weeks, precancerous and 32 weeks, tumor tissue) treatment time points, and thus, we have (1) evaluated methylation changes which occur within a continuum of PB-induced liver tumorigenesis, and (2) identified the genes/genomic regions that harbor these modifications.

Strikingly, there were multiple RAMs cloned in both the CAR WT and B6C3F1 mice which aligned to the exact same genomic regions and exhibited unique alterations in methylation in livers of both groups of tumor-susceptible mice, that is, B6C3F1 and CAR WT. Specifically, 11 genes were identified from identical, unique PB-induced RAMs that formed in both C3H/He CAR WT (precancerous liver and/or liver tumor) and B6C3F1 (2- and/or 4-week treated) mice (Table 3). Importantly, these similarities were observed regardless of the stock/strain of mice. Although six of these genes (Efnb2, Prickle2, Ptpro, Tcf4, Tnk2, and Trio) have previously been shown to be involved in cancer or cancer-related processes (Table 1; Supplemental Figs. S22-S27), the remaining 5 (Bcat2, Ddx54, Srms, Wscd1, and Zscan22) are of particular interest because they represent “new” candidate genes potentially involved in both relatively early and late stages of tumorigenesis. The androgen receptor (AR) and Akt1 are common targets (Supplemental Fig. S28), whereas Egf is a common regulator (Supplemental Fig. S29), of a subset of the 6 cancer-related genes. For some of these common genes (Ddx54: M315-317, Tcf4: H200, and Tnk2: M275–76), the change in methylation status at the 2/4-week time point(s) does not occur in the same direction (e.g., hypomethylation vs. hypermethylation) as that seen in the precancerous/tumor tissue, indicating the possibility for them to have a different role during early, as compared with late, stages of tumorigenesis.

Based on these data, it is conceivable that an “altered methylation fingerprint,” consisting of a subset of the genes which exhibited PB-induced altered methylation uniquely in the liver tumor-sensitive mice, could be developed and utilized as a biomarker to identify PB-like nongenotoxic liver tumor promoters at relatively early times following treatment. Preliminary candidates for the “altered methylation fingerprint” are those genes, presented in Table 3, which revealed identical methylation changes in both the B6C3F1 and CAR WT mice: Bcat2: M464-468, Ddx54: B312-315, Ptpro: B341–342, Wscd1: M358–359, and Zscan22: 238–239. Pathway diagrams of Bcat2 and Ptpro are located in Supplemental Information (Figs. S30 and S24, respectively). Due to the dearth of literature information on Wscd1 and Zscan22, pathways could not be generated for them. A second tier of potential genes that might contribute to an “altered methylation fingerprint” includes three genes (identified from four RAMs) which occurred in both the B6C3F1 and CAR WT mice; RAMs from at least one time point in each group exhibited similar methylation changes, however, an additional group showed an opposite or ambiguous change (e.g., from Table 3: Prickle2: H310-312, excluding the ambiguous precancerous RAM, Prickle2: B310-312, excluding the B6C3F1, 2-week RAM, Tcf4: B200, excluding the B6C3F1, 4-week RAM, and Wscd1: B357–358, excluding the tumor RAM). Pathway diagrams of Prickle2 and Tcf4 are located in Supplemental Information (Figs. S23 and S25, respectively).

Although the genes exhibiting unique PB-induced altered methylation patterns in the precancerous and tumor tissue may individually be playing crucial roles during tumorigenesis (e.g., hypermethylation of a growth suppressor or hypomethylation of an oncogene), analysis of total hepatic DNA, or DNA isolated from individual tumors, does not allow us to conclude whether the changes occurred in one, or multiple, cell populations. Because it is possible that these modifications arose in different populations, it is constructive to consider the overarching pathways (e.g., angiogenesis) in which these genes participate (Table 2; Figs. 3–7; Supplemental Figs. S12–S21). Similar observations were made by Jones et al. (2008) and Parsons et al. (2008), who showed that core signaling pathways were genetically altered in pancreatic cancer and glioblastoma, respectively. Importantly, although the particular mutations varied among tumor samples of a specific cancer type (i.e., different genes were affected, depending on the sample), the altered genes participate in the same pathways. Likewise, epigenetic deregulation of certain pathways by numerous genes identified from unique RAMs might more efficiently facilitate tumor formation. Importantly, several unique precancerous RAMs are involved in five distinct processes, however, more unique RAMs in the tumor tissue potentially affect these same pathways (Figs. 3–7), suggesting that altered methylation plays progressively critical roles during tumor development. Additionally, common targets of multiple unique RAMs in the precancerous (Fig. 1) and tumor (Fig. 2) tissue suggest that common processes might be influenced, for example, AKT1 is a common target of 4 genes identified from unique RAMs in the tumor tissue (Fig. 2: Abl1, Efnb2, Prkce, and Tcf4).

Of the five the processes in which multiple genes identified from unique precancerous and tumor RAMs are involved (Figs. 3–7), four of them (angiogenesis, EMT, growth/survival, and invasion/metastasis) are also converged upon by unique B6C3F1 RAMs at 2 or 4 weeks of PB treatment (Fig. 4 in Phillips and Goodman, 2008). Furthermore, common targets (AKT1, CCND1, FOS, JUN, PTK2, and VEGF) of unique precancerous and/or tumor RAMs (Figs. 1 and 2) are also targets of unique B6C3F1, 4-week RAMs (Fig. 3 in Phillips and Goodman, 2008). These data, combined with the finding that several common genes in the B6C3F1 (2 and 4 weeks) and CAR WT (23 and 32 weeks) mice exhibited PB-induced altered methylation (Table 3), indicate the importance of early modifications which might ultimately influence expression and drive tumorigenesis.

Many genes exhibit point mutations, deletions or amplifications in various cancers (Jones et al., 2008; Parsons et al., 2008). However, the integration of genetic and epigenetic data are crucial to understanding the molecular basis of the disease (McLendon et al., 2008), and our data highlight the significance of DNA methylation as an epigenetic mechanism underlying tumorigenesis. It is known that C3H/He and B6C3F1 mice are more susceptible to liver tumors as compared with the C57BL/6 strain (Becker, 1982). In response to PB treatment, methylation patterns are altered to a greater extent in the sensitive groups, including higher levels of hypermethylation (C3H/He >> B6C3F1 > C57BL/6), and thus, it is hypothesized that the inability of these animals to maintain normal methylation patterns is responsible, at least in part, for their differential liver tumor susceptibilities (Watson and Goodman, 2002). Although the fundamental genes underlying liver cancer in mice and humans are likely the same, rodents exhibit an increased susceptibility to tumorigenesis (Rangarajan and Weinberg, 2003). Also, it has been observed that methylation patterns in rodent cells are less stable than in human cells (reviewed in Goodman and Watson, 2002). Therefore, a difference between humans and rodents might exist with regard to the regulation of epigenetic control, including DNA methylation, resulting in enhanced sensitivity to tumor formation in the latter.

PB causes hepatic hypertrophy and hyperplasia (Whysner et al., 1996), and PB-treated CAR WT, but not CAR KO, mice exhibit an increase in liver mass due to mitogenesis and cellular hypertrophy (Wei et al., 2000). Thus, hypomethylation could result from replication-dependent, passive demethylation (Howlett and Reik, 1991). Hypomethylated RAMs might occur via physical blocking of cis-elements and/or trans-acting factors (e.g., DNA methyltransferases, DNMTs, or methyl-binding proteins) by CAR itself, for example, the presence of CAR on DNA could sterically hinder the binding of DNMT1. Active demethylation (reviewed in Ooi and Bestor, 2008) might also decrease 5-methylcytosine levels. Increases in methylation (i.e., hyper- and new methylations) could be mediated by the recruitment of de novo DNMTs (i.e., 3A and 3B) and/or methyl-binding proteins to DNA by CAR, similar to the orphan nuclear receptor GCNF recruitment of de novo methyltransferases, resulting in Oct-3/4 silencing (Sato et al., 2006). Moreover, CAR might mediate PB-induced expression changes in proteins that could alter methylation status: DNMTs (Goll and Bestor, 2005), proteins involved in demethylation (Ooi and Bestor, 2008), methyl-binding proteins (Lopez-Serra and Esteller, 2008) and/or enzymes involved in regulating levels of S-adenosylmethionine (SAM), the primary methyl-donor for DNMT reactions (Ulrey et al., 2005).

The current study focused upon PB-induced RAMs that formed uniquely in liver tumor-sensitive CAR WT (precancerous liver and/or liver tumor tissue), as compared with resistant CAR KO, mice, and the genes identified from these RAMs can function in processes that are important for tumor development, including angiogenesis, apoptosis, EMT, growth/survival, and migration/invasion/metastasis. These data, plus the results from an analogous study which discerned PB-elicited unique RAMs at 2 and 4 weeks in susceptible B6C3F1, as compared with relatively resistant C57BL/6, mice (Bachman et al., 2006b; Phillips and Goodman, 2008), allowed us to identify genes whose methylation statuses were altered uniquely in sensitive animals within a continuum of PB-induced liver tumorigenesis. A subset of these, as discussed above, represent our initial, preliminary candidate “altered methylation fingerprint” which might be employed as a biomarker to identify PB-like nongenotoxic liver tumor promoters at relatively early times following treatment. Common genes observed at both early (B6C3F1) and later (CAR WT) time points of treatment (Table 3), which exhibited unique PB-induced alterations in methylation, participate in the same key processes, indicating that methylation changes at early times are critical for tumor formation. Thus, the current study represents an important step forward towards our goal of enhancing our understanding of genes and pathways involved in PB-induced tumorigenesis. Expression analysis (both quantitative real-time PCR and microarrays) is currently being performed on RNA isolated from the same liver tissue used in the current study and our study involving a comparison of the effects of PB (following 2 or 4 weeks of treatment) on liver tumor–susceptible B6C3F1 mice as compared with the relatively resistant C57BL/6 (Phillips and Goodman, 2008).

SUPPLEMENTARY DATA

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

FUNDING

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