Genome-wide promoter DNA Methylation Profiling of Hepatocellular Carcinomas Arising Either Spontaneously or due to Chronic Exposure to Ginkgo biloba Extract (GBE) in B6C3F1/N Mice

Abstract

Epigenetic modifications, such as DNA methylation, play an important role in carcinogenesis. In a recent NTP study, chronic exposure of B6C3F1/N mice to Ginkgo biloba extract (GBE) resulted in a high incidence of hepatocellular carcinomas (HCC). Genome-wide promoter methylation profiling on GBE-exposed HCC, spontaneous HCC, and age-matched vehicle control liver was performed to identify differentially methylated genes in GBE-exposed HCC and spontaneous HCC. DNA methylation alterations were correlated to the corresponding global gene expression changes. Compared to control liver, 1296 gene promoters (719 hypermethylated, 577 hypomethylated) in GBE-exposed HCC and 738 (427 hypermethylated, 311 hypomethylated) gene promoters in spontaneous HCC were significantly differentially methylated, suggesting an impact of methylation on GBE-exposed HCC. Differential methylation of promoter regions in relevant cancer genes (cMyc, Spry2, Dusp5) and their corresponding differential gene expression was validated by quantitative pyrosequencing and qRT-PCR, respectively. In conclusion, we have identified differentially methylated promoter regions of relevant cancer genes altered in GBE-exposed HCC compared to spontaneous HCC. Further study of unique sets of differentially methylated genes in chemically-exposed mouse HCC could potentially be used to differentiate treatment-related tumors from spontaneous-tumors in cancer bioassays and provide additional understanding of the underlying epigenetic mechanisms of chemical carcinogenesis.

INTRODUCTION

The liver is one of the most commonly affected target sites for spontaneous or chemically-exposed tumors in the B6C3F1/N mouse (Kim et al. 2005; Maronpot et al. 1987). Hepatocellular carcinoma (HCC) is a heterogeneous neoplasm that likely arises as a result of the accumulation of both genetic and epigenetic aberrations, involving dysregulation of numerous cancer related pathways (Niu et al. 2016; Rashid et al. 1999). Gene expression studies on mouse HCC have shown significant differences between spontaneous HCC (SPNT-HCC) in vehicle control mice and Ginkgo biloba leaf extract-exposed (GBE-HCC) B6C3F1/N mice (Hoenerhoff et al. 2013). Some of these changes in gene expression may be related to gene-specific differential promoter

DNA methylation.

Ginkgo biloba extract (GBE) has been used in traditional Chinese medicine and is one of the most commonly used botanical supplement in the United States (Chan et al. 2007). Now, it is being used for its purported effects in improving brain function, anticancer, and antioxidant effects (Chan et al. 2007; Moon et al. 2006). Purported anticancer effects of GBE have been attributed to its antioxidant, anticlastogenic and gene-regulatory actions (DeFeudis et al. 2003; Moon et al. 2006). GBE has been reported to possess hepatoprotective, cardioprotective, antiasthmatic, antidiabetic and antiangiogenic properties (DeFeudis et al. 2003; Naik and Panda 2007). However, the exact mechanism of GBE’s reported beneficial effects is still unclear. The National Toxicology Program (NTP) previously performed subchronic and chronic carcinogenicity bioassays on GBE to determine the cancer hazard associated with chronic exposure to a GBE botanical supplement (NTP 2013). A result of this study indicated that lifetime exposure to doses of 200, 600, 2000mg/kg of GBE resulted in a dose-related increase in the incidence of HCC in B6C3F1/N mice (NTP 2013).

In addition to genetic mutations, epigenetic alterations (e.g., DNA methylation, miRNAs, and histone tail modifications) also play an important role in cancer development and progression (Jones and Baylin 2007). Cancer gene activity can be significantly altered by various epigenetic modifications including promoter DNA methylation, histone acetylation, and differential expressions of miRNAs, among others (Ally et al. 2017; Calin and Croce 2006; Feinberg and Vogelstein 1983; Jones and Martienssen 2005; Zardo et al. 2002). Aberrant DNA methylation is a stable and early event in carcinogenesis, and specific DNA methylation signatures have great potential to become biomarkers of exposure for early detection and possibly for prediction of toxicity for unknown compounds (Ladd-Acosta 2015). DNA methylation is characterized by the methylation of cytosine in CpG dinucleotides within distinct GC-rich regions (“CpG islands”) in the mammalian genome (Doerfler 1983; Plass 2002). Multiple genome-wide DNA methylation studies in various cancers have identified genes with differentially methylated regions (DMRs) in malignant tissue compared with normal tissue (Costello et al. 2000). Understanding the effects of chemical treatment on DNA methylation is important in understanding the mechanisms of chemically-induced tumorigenesis.

In human HCCs, a number of mutations have been observed, most frequently in TP53, CTNNB1, APOB, ARID1A and other cancer genes. (Ally et al. 2017; Imbeaud et al. 2010). Additionally, several epigenetic alterations including differential DNA methylation of several cancer related gene promoters are known to play an important role in hepatocarcinogenesis. Some of the most commonly differentially methylated genes in human HCC include CDKN2A/INK4 (p16) (Hui et al. 1996), RASSF1A (Yeo et al. 2005), GSTP1 (Zhang et al. 2005), MGMT (Yang et al. 2003), cMYC (Nambu et al. 1987; Wang et al. 2013), TERT, APC, BMP4, HOXA9, FLT4 (Hernandez-Vargas et al. 2010), SPRY2 (Lo et al. 2006), and DUSP5 (Fu et al. 2006).

We evaluated the differential methylation changes in spontaneous HCC arising in vehicle control animals and HCC from animals with chronic GBE exposure, in B6C3F1/N mice using NimbleGen® mouse DNA promoter methylation arrays. In order to identify genes that are potentially epigenetically regulated, we have compared the differential promoter DNA methylation data with the corresponding transcriptomic alterations.

MATERIALS AND METHODS

Test Material and treatment

The National Toxicology Program (NTP) used the Gingko biloba extract (GBE) obtained from Shanghai Xing Ling Science and Technology Pharmaceutical Company Ltd. The composition of the the extract included 31.2% flavonol glycosides, 15.4% terpene lactones, and 10 ppm ginkgolic acids. The range of concentrations of the various constituents in Shanghai Xing Ling extract and EGB 761® were found to besimilar to other GBE products available in marketplace. Male and female B6C3F1/N mice were exposed to 0, 200, 600, and 2000 mg/kg GBE by corn-oil gavage, 5 days a week for 104 weeks (NTP 2013).

Sampling of hepatocellular tumors and DNA isolation

At necropsy, liver from vehicle control animals with no gross or microscopic lesions (CNTL), SPNT-HCC, and GBE-HCC were fixed in 10% neutral-buffered formalin, processed routinely, embedded in paraffin, and 5μm sections were taken and stained with hematoxylin and eosin (H&E). Portion of selected HCCs and normal liver from vehicle control animals were also collected and flash-frozen in liquid nitrogen at the time of necropsy for molecular analysis. Upon examination of the H & E slides made out of FFPE sections representing the frozen counterpart, five CNTL liver (3 males and 2 females), five SPNT-HCC (4 males and 1 female), and five GBE-HCC (4 males and 1 female) frozen tissue samples were selected for DNA isolation. DNA and RNA were isolated from these selected samples using the DNeasy kit® (Qiagen, Valencia, CA) and Invitrogen PureLink Mini Kit® (Life Technologies Corporation, Carlsbad, CA), respectively, and stored at −80C. DNA and RNA were quantified by using a NanoDrop® (Thermo Scientific, Wilmington, MA), and RNA quality was assessed using an Agilent Bioanalyzer®. (Agilent Technologies, Inc; Santa Clara, CA).

Methylation Array

Roche NimbleGen® mouse DNA methylation microarray ‘3×720K Promoter Plus CpG Plus RefSeq Promoter’ (Roche NimbleGen, Madison, Wisconsin), which includes 20,404 promoter regions, 22,881 transcripts, and 15,988 annotated CpG islands, was used to identify differentially methylated promoter CpG regions between groups. First, the high molecular weight genomic DNA was sonicated to generate 200–500bp sized DNA fragments using a Bioruptor sonicator® (Diagenode, Denville, NJ) using cycle conditions of 30 sec on/90 sec off for a total duration of 12.5 minutes. Methylated DNA molecules were enriched by binding to the MBD2B/MBD3L1 complex using the MethylCollector Ultra kit® (Active Motif, Carlsbad, CA) according to the manufacturer’s protocol. Input and enriched methylated-DNA fragments from the same sample were whole genome amplified (WGA) using a WGA2 kit® (Sigma-Aldrich, St. Louis, MO) according to the manufacturer’s protocol. The labeling of WGA IP and input DNA, microarray hybridization, and scanning were performed at the Florida State University genomics core laboratory following the described protocol (Roche NimbleGen (2010)). Data were extracted from scanned images using NimbleScan 2.4 extraction software (NimbleGen Systems, Inc., Madison, WI).

Methylation Array Normalization

The channel wise signal from pair file was log2 transformed and QC plots were generated to detect per sample systematic variations. Both, M (log2(cy5/cy3)) vs. A (log2(cy5*cy3)/2) and M vs. GC density (ratio of the G and C nucleotides count and probe sequence length) plot revealed presence of systematic non-linear bias. To correct for these biases simultaneously, we first binned all probes according to their GC density. The overall variability in GC density values was used to compute bin width according to zero-stage rule (Wand 1997). These bin widths are proven to be approximate L2 optimal; i.e., they minimize mean integrated square error. The bins with fewer probes were then merged so that each bin contains at least 500 probes. Within each bin, lowess regression (Cleveland and Devlin 1988; Hastie and Tibshirani 1990) was used to predict M values as a smooth function of A values. The difference between observed and predicted M values was used as within array normalized signal. Next, we employed quantile normalization to correct for between- sample systematic variation. The visual inspection of quality control plots (global PCA plot, cluster plot and replicate comparison scatter plots) identified three samples; #6 (SPNT-HCC, replicate 1), #12 (GBE-HCC, replicate #2) and #14 (GBE-HCC, Replicate #4) as potential outliers. These samples were excluded from all downstream analysis and remaining samples were re-normalized.

Detection of Differentially Methylated Regions (DMRs)

We developed a two-stage peak detection algorithm which simultaneously assesses within-sample local enrichment and between-treatment differential methylation. In the first stage, we computed a within-array enrichment statistic for each sample (defined as Mann-Whitney Z-score) by comparing signal values inside a genomic window of size (w=500 base pairs) centered at probe with signal values outside of window. In the second stage, we computed probe-specific significance p-values for each comparison of interest using a conventional two sample mean comparison test, using the enrichment metric derived in the first stage. This test utilizes the assumption that the enrichment statistics are independently distributed normal variates with mean 0 and variance 1. The probes with significant p-values less than the p-value threshold (p=0.05) were tagged as gained or lost based on the mean difference in enrichment statistics. Finally, the DMRs were identified as runs of consecutive gained (lost) probes that were separated by less than 1000 base pairs. Note that the values used for window size (w) and p-value (p) thresholds were identified using a grid search method. The p-value for DMRs was computed using a modification of Stouffer’s p-value combination method that incorporates the correlation between experiments. This method assumed that probit (inversion Guassian) transformed, probe-specific p-values within DMRs follow multivariate standard normal with autoregressive correlation matrix. Furthermore, we estimated autogressive correlation using quasi least squares (QLS) method (for QLS details see (Chaganty 1997).

Gene Expression Microarray Analysis

Affymetrix Mouse Genome 430 2.0 GeneChip arrays (Affymetrix, Santa Clara, CA) were used to assess differential gene expression profiles of vehicle CNTL liver, SPNT-HCC, and GBE-HCC. One-cycle cDNA synthesis protocol was used to amplify 1 µg of RNA as per the manufacturer’s instructions. For each array, 15 µg of amplified biotinylated cDNAs were fragmented and hybridized for 16 hours at 45°C as per Affymetrix protocol. Array slides were processed and obtained data for further analysis as previously described (Hoenerhoff et al. 2013). The raw PM intensity from Affymetrix .CEL files were normalized using the established Robust Multiarray Normalization (RMA) method (Bolstad et al. 2003). Next, we performed one-way analysis of variance (ANOVA) and post-hoc pairwise comparison t-tests to assess significance of differentially expressed genes (DEGs). The Benjamini-Hochberg multiple test correction method was used to control for false discovery rate.

DNA Methylation and Gene Expression Correlation Analysis

To perform a systematic methylation vs. gene expression comparison, it was necessary to construct standardized comparable gene level signal metrics. Towards this goal, we utilized the RefSeq genes annotation table for Mus musculus mm9 genome (downloaded from UCSC Genome Browser, http://hgdownload.cse.ucsc.edu/goldenPath/mm9/database/refFlat.txt.gz) (Casper et al. 2018). We first generated the gene level methylation signal by averaging previously normalized methylation signals from probes that overlap promoter regions, defined as +/− 2Kb of the transcription start site (TSS). Next, we computed the gene level expression signal by averaging the previously normalized probe set level signal of probe set annotated to the same gene. Finally, these gene level methylation (expression) signals were utilized to identify differentially methylated (expressed) genes using the conventional one-way ANOVA and subsequent post-hoc pairwise comparison.

Validation of Methylation by Quantitative Bisulfite Pyrosequencing

Based on the array data, the promoter regions of selected cancer-related genes were validated for differential promoter methylation between treatment groups using pyrosequencing methodology. Pyrosequencing was carried out with primers designed with the Pyromark Assay Design Software (Qiagen, Valenica, CA) version 2.0.2.15 (Supplemental Table S3). Bisulfite-modified DNA was then amplified using PCR in preparation for pyrosequencing, with either the forward or reverse biotinylated primers (Tost et al. 2006). The biotinylated PCR products were captured with streptavidin sepharose beads, denatured to single strand, and annealed to the sequencing primer for the pyrosequencing assay. Universal Methylated Mouse DNA Standard (Zymo Research, location) was used as positive control and water substituted for DNA was used as a negative control. Pyrosequencing was carried out using the PyroMark Q96 MD System (Qiagen, Valencia, CA) according to the manufacturer instructions. The percentage methylation was quantified using the Pyro Q-CpG Software (Qiagen, Valencia, CA).

Validation of Gene Expression by Quantitative Real-time (qRT) PCR

RNA was extracted from each sample using the Invitrogen PureLink Mini Kit (Invitrogen cat# 12183–018A, Carlsbad, CA) according to the manufacture’s protocol. One microgram of total RNA was subjected to reverse-transcription PCR to generate complementary DNA (cDNA). Relative quantitative gene expression levels were detected using real-time PCR with the ABI PRISM 7900HT Sequence Detection System (Life Technologies, Grand Island, NY) using SYBR green detection methodology. Primers were designed using Primer3Plus software (Rozen and Skaletsky 2000) to span exon-exon junctions with an annealing temperature of 60°C and amplification size of less than 150 bp (Table S4). Briefly, 25 ng of cDNA were added to a 25 μl PCR reaction to get a final concentration of 1.00 ng/μl of cDNA. Forward and reverse primer final concentrations were 100nM in the SYBR green assay. The reactions were performed using the Power SYBR® Green PCR Master Mix (Life Technologies, Grand Island, NY). 18s RNA was chosen as the endogenous control gene in our qPCR experiments. Relative quantification of gene expression changes was recorded after normalizing for 18s RNA expression, computed by using the 2^−(ΔΔCt) method (user manual #2, ABI Prism 7700 SDS). To determine this normalized value, 2^−(ΔΔCt) values were compared between tumor and control samples, where the changes in crossing threshold (ΔCt) = Ct_{Target gene}- Ct_{18s RNA}, and ΔΔCt=ΔCt^control-ΔCt^target.

RESULTS

Genome-wide promoter DNA methylation profiles were obtained for five CNTL mouse liver, five SPNT-HCC, and five GBE-HCC from animals exposed to 2000 mg/kg GBE. Gene expression profiles were obtained from six CNTL mouse liver, six SPNT-HCC and six GBE-HCC. Once all array data was confirmed to be without confounding batch effects, the array data was subjected to further statistical analysis.

Methylation Profiles and Differentially Expressed Genes (DEGs) Differentiate Hepatocellular Carcinomas from Non-tumor Liver

All samples exhibited clear clustering within groups in the principal component analysis (PCA), indicating significant similarities in global gene expression profile and global promoter methylation within groups (Figure 1A and 1B). Further, there was significant discernible separation of experimental groups from one another, indicating significant differences in global gene expression (Figure 1A) and differential promoter methylation between groups (Figure 1B). Further, hierarchical clustering analysis idenfied significantly differentially methylated genes in GBE-HCC (n=1296) and SPNT-HCC (n=738) which correlated with global gene expression profiles (2037 DEGs in GBE-HCCs and 1632 DEGs in SPNT-HCC) and promoter methylation profiles between groups (Figure 2A–2D).

Figure 1.

A. Differential gene expression profiling of spontaneous and GBE-exposed HCC. Principal component analysis (PCA) demonstrated significant clustering of normal control liver (CNTL, red), spontaneous HCC (SPNT-HCC, blue), and GBE-exposed HCC (GBE-HCC, green), based on global gene expression. B. Profiling of differentially methylated regions (DMRs). Principal component analysis (PCA) of the global differential promoter methylation data demonstrated clustering of normal liver (red), spontaneous HCC (blue), and GBE-exposed HCC (green). Number of samples for gene expression study are 6 per each group and for methylation project, n=5 per each group.

Figure 2.

Differential methylation and differential gene expression profiling of spontaneous and GBE-exposed HCC. A. GBE-HCC vs. CNTL: RefSeq Genes with differential expression profile (N=2037) and their corresponding promoter methylation profile. Differential methylation criteria: Absolute foldchange >= 1.2 and p-value <= 0.05. Differential expression criteria Absolute foldchange >= 2.0 and p-value<=0.05. Genes with at least one promoter overlapping NimbleGen® methylation probe and at least one Affymetrix® probe set were considered. The p-value for comparison of two cohorts is derived by performing paired t-test of mean log2 expression values across all genes with gained (lost) methylation.

B. GBE-HCC vs. CNTL: Mean expression of hypo(hyper) methylated genes. Gained(lost)=1 if promoter methylation fold change >= 1.2 and p-value <= 0.05. Genes with at least one promoter overlapping NimbleGen® methylation probe and at least one Affymetrix® probe set were considered. The p-value for comparison of two cohorts is derived by performing paired t-test of mean log2 expression values across all genes with gained (lost) methylation.

C. SPNT-HCC vs. CNTL: RefSeq Genes with differential expression profile (N=1632) and their corresponding promoter methylation profile. Differential expression criteria: Absolute fold change >= 2.0 and p-value <=0.05. Differential methylation criteria: Absolute fold change >= 1.2 and p-value <= 0.05.

D. SPNT-HCC vs. CNTL: Mean expression of hypo(hyper) methylated refseq genes. Gained (Lost)=1 if promoter methylation fold change >= 1.2 and p-value <= 0.05. Genes with at least one promoter overlapping NimbleGen® methylation probe and at least one Affymetrix® probe set were considered. The p-value for comparison of two cohorts is derived by performing paired t-test of mean log2 expression values across all genes with gained (lost) methylation.

From the analysis of 20,404 promoters and 15,980 annotated CpG island regions on this array, a set of 1296 DMGs(hypomethylated or hypermethylated) in GBE-HCC (Figure 3A) and 738 DMGs in SPNT-HCC, were identified compared to CNTL liver (Figure 3B). Gene expression data from these tumors revealed a set of 2037 differentially expressed genes (DEGs, up- or down-regulated) in GBE-HCC and 1632 DEGs in SPNT-HCCs compared to control liver. A transcript was tagged as gene expression up (or down) if at least one microarray probe had an absolute fold change of ≥ +/−2.0 and a p-value of ≤ 0.05. A transcript was tagged as methylation gained (or lost) if the absolute fold change was ≥ +/−1.2 and the p-value was ≤ 0.05.

Figure 3.

Concordance of differentially methylated genes (DMGs) and differentially expressed genes (DEGs) in spontaneous and GBE-exposed HCC. For each refseq gene transcript that is present on Affymetrix® gene expression we first computed gene expression fold change and corresponding p-value. For each refseq gene, the methylation foldchange and corresponding p-values were computed signal by averaging previously normalized methylation signal from probes that overlap promoter region, defined as +/− 2Kb of TSS. The Venn-diagram represents co-occurrences of differential gene expression and differentially methylated genes within promoter regions of refseq genes by applying thresholds specified below. Transcript is tagged as methylation gained(lost)= 1 if promoter methylation foldchange >= 1.2 and p-value <=0.05. Transcript is tagged as gene expression up(down)=1 if transcript absolute expression fold change >=2.0 and p-value <=0.05. Genes with at least one promoter overlaying NimbleGen® methylation probe and at least on Affymetrix® probe set were considered.

Methylation Profiles and Differentially Expressed Genes Differentiate Spontaneous HCCs from GBE-exposed HCCs

To analyze the co-occurrences of differential gene expression and differential methylation changes, an overlap summary was generated using the lists of DMGs and DEGs in respective groups, using fold change in gene expression of +/− 2.0 and p-value of ≤ 0.05. From the set of 1296 DEGs in GBE-HCC, 61 genes were correlated with differential promoter DNA methylation when stringent feature analysis was performed. A set of 32 out of 61 DMGs were hypermethylated, with correlating downregulation of gene expression, and 29 genes hypomethylated with correlating upregulation of gene expression (Figure 3A). Similarly, in SPNT-HCCs, 44 genes had co-occurrences of differential methylation and gene expression; 28 genes gained promoter methylation with downregulation and 16 genes lost promoter methylation and showed gene overexpression (Figure 3B). SPNT-HCC and GBE-HCC tumors are indistinguishable morphologically; however, their global differential promoter methylation profile and differential gene expression are very different from one another.

To understand the global overlap of DMRs and DEGs in different comparisons, all the probe sets in the microarray and methylation arrays were taken into consideration for calculating the overlap summary. The transcripts are tagged as methylation gained (or lost) =1 if promoter methylation fold change ≥1.2 and p-value ≤0.05 and transcripts are tagged as gene expression up (or down) =1 if transcript absolute expression fold change ≥2.0 and p-value ≤0.05. Genes with at least one promoter overlapping NimbleGen® methylation probe and at least one Affymetrix® probe set were considered. Only the set of probes with gained promoter methylation and corresponding repressed/downregulated gene expressions (DEGs) are considered statistically significant in both GBE-HCCs and SPNT-HCCs if the p-value is <0.0001 (Table 1). The comparison of gene promoter hypomethylation and correlated gene expression upregulation was was not statistically significant in both HCCs (p-value of 1.177 for GBE-HCC and 1.643 for SPNT-HCCs). Regardless of hypo- or hypermethylation, a higher number of genes were differentially methylated in GBE-HCC (1296) compared to SPNT-HCC (738) (Figure 3A–3B). The top 20 genes with concordance of hyper- or hypo-methylated promoter regions and down- or up-regulated gene expression profile respectively, ranked by methylation and gene expression fold change are listed in Table 2A–2B for GBE-HCCs and Table 2C–2D for SPNT-HCCs. The top 20 genes with hypo or hypermethylated promoter regions in each group are listed in supplementary Tables S1A−S1D, and the top 20 downregulated or upregulated genes are listed in supplementary Table S2A−S2D. Among the top statistically significant hypermethylated and downregulated genes in GBE-HCC were several members of the Major Urinary Protein gene family (MUP7, 8, 9, 10, 12, 13, 19), apoptosis genes (Perp), xenobiotic metabolism genes (Cyps, Asic5, Hsd17b13) and immune response related genes (C8a, C8b)(Table 2B). In SPNT-HCC, top statistically significant hypermethylated and downregulated genes included transcription factors (Id3, Lrtm1), and xenobiotic metabolism genes (Cyps) (Table 2D).

Table 1.

Differentially Methylated Regions (DMRs) and Differentially Expressed Genes (DEGs) Promoter Regions Overlap Summary. This displays odds ratio and corresponding p-value to assess co-occurrence of differential methylation and differential gene expression. The observed overlap between promoter regions (+/− 2kb around transcriptional start site, TSS) of differentially expressed refseq genes and differentially methylated sites was computed and compared with expected overlap using 10000 chromosome bound circular shift permutations.

Comparison				GBE-HCC vs. CNTL		SPNT-HCC vs. CNTL
Type				Repressed	Upregulated	Repressed	Upregulated
Comparison	Methylation	Number of Probes	Stat	6375	13692	9817	13782
GBE-HCC vs. CNTL	Lost	7264	Odds Ratio	0.248	1.177	0.530	0.951
			P-value	0.0105	0.4686	0.0679	0.8550
	Gained	10807	Odds Ratio	3.146	0.518	3.354	0.914
			P-value	<0.0001	0.0057	<0.0001	0.6466
SPNT-HCC vs. CNTL	Lost	4796	Odds Ratio	0.387	1.315	0.411	1.643
			P-value	0.1043	0.3206	.0635	0.0465
	Gained	6509	Odds Ratio	1.563	0.669	3.544	0.623
			P-value	0.1565	0.1513	<0.0001	0.1085

Number of features = 713332

Number of circular shift permutations used = 10000

DMRs were identified using W=500, P=0.05 and nProbes>=4 criteria and false discovery rate of 0.001

DEGs were identified using post-ANOVA adhoc tests using Foldchange>=2 and p-value<=0.005 criteria

Promoter region was defined as 2kb upstream and 2kb downstream of txStart

Table 2A.

GBE-HCC vs CNTL: Top 20 genes with concordance of hypomethylated promoter regions and upregulation of gene expression. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to gene expression fold change.

Gene Symbols	Gene Name	RefSeq Genes	Methylation Foldchange	Methylation P-value	Expression Foldchange	Expression P-value
Esm1	Endothelial Cell Specific Molecule 1	NM_023612	−1.3494	0.0000	16.1201	0.0000
Mmp12	Metallopeptidase 12	NM_008605	−1.2883	0.0091	7.5521	0.0031
Casp1	Caspase 1	NM_009807	−1.2671	0.0022	5.4685	0.0000
Casp12	Caspase 12	NM_009808	−1.2243	0.0054	5.1692	0.0000
Clec4a3	C-type lectin domain family 4, member a3	NM_001204241	−1.2085	0.0313	4.4933	0.0001
Cd48	CD48	NM_007649	−1.3816	0.0000	3.6570	0.0007
Lilr4b	Leukocyte immunoglobulin-like receptor, subfamily B, member 4B	NM_001291892 /// NM_001291893 /// NM_008147	−1.2219	0.0053	3.3254	0.0047
Cd34	CD34	NM_001111059 /// NM_133654	−1.2185	0.0011	3.3165	0.0231
Ctla2a	Cytotoxic T lymphocyte-associated protein 2α	NM_001145799	−1.2203	0.0006	3.3057	0.0087
Casp4	Caspase 4	NM_007609	−1.2271	0.0279	2.9870	0.0012
Ear12	Eosinophil-associated, ribonuclease A family, mem 12	NM_001012766	−1.3268	0.0050	2.9168	0.0003
Ear2	Eosinophil-associated, ribonuclease A family, member 2	NM_007895	−1.3419	0.0038	2.9168	0.0003
C1qc	Complement C1qC chain	NM_007574	−1.2026	0.0011	2.7334	0.0035
Sdcbp2	Syndecan binding protein 2	NM_145535	−1.2690	0.0028	2.6564	0.0001
Plek	Pleckstrin	NM_019549	−1.2298	0.0002	2.6347	0.0025
Gbp2b	Guanylate binding protein 2b	NM_010259	−1.2215	0.0026	2.6149	0.0041
Sema3b	Semaphorin 3B	NM_009153	−1.2184	0.0007	2.5364	0.0000
Sema3b	Semaphorin 3B	NR_111986	−1.2161	0.0010	2.5364	0.0000

Table 2B.

GBE-HCC vs CNTL: Top 20 genes with concordance of hypermethylated promoter regions and downregulation of gene expression. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to gene expression fold change.

Gene Symbols	Gene Name	RefSeq Gene	Methylation Foldchange	Methylation P-value	Expression Foldchange	Expression P-value
Mup9	Major urinary protein 9	NM_001281979	1.2411	0.0308	−20.2562	0.0008
Mup10	Major urinary protein 10	NM_001122647	1.3041	0.0168	−15.7589	0.0021
Mup12	Major urinary protein 12	NM_001199995	1.5084	0.0012	−15.5871	0.0049
Cyp2f2	Cytochrome P450, family 2, subfamily f, polypeptide 2	NM_007817	1.2696	0.0025	−12.7422	0.0138
C8a	Complement C8 alpha chain	NM_001290645 /// NM_001316667 /// NM_146148	1.2965	0.0004	−7.8226	0.0001
Mup13	Major urinary protein 13	NM_001134674	1.2892	0.0259	−7.4205	0.0376
Mup19	Major urinary protein 19	NM_001135127	1.2331	0.0037	−7.4205	0.0376
Mup7	Major urinary protein 7	NM_001134675	1.3431	0.0019	−7.4205	0.0376
Mup8	Major urinary protein 8	NM_001134676	1.3356	0.0159	−7.4205	0.0376
Perp	P53 Apoptosis Effector Related to PMP22	NM_022032	1.2024	0.0034	−5.7755	0.0003
Cyp8b1	Cytochrome P450, family 8, subfamily b, polypeptide 1	NM_010012	1.2519	0.0002	−4.8848	0.0009
Asic5	Acid-sensing ion channel family member 5	NM_021370	1.2243	0.0007	−4.1873	0.0005
Slc3a1	Solute carrier family 3 member 1	NM_009205	1.2355	0.0011	−3.6061	0.0089
S1pr5	Sphingosine-1-phosphate receptor 5	NM_053190	1.2277	0.0049	−3.5562	0.0000
Pigr	Polymeric immunoglobulin receptor	NM_011082	1.5604	0.0001	−3.5354	0.0317
C8b	Complement C8 beta chain	NM_001316671 /// NM_133882	1.2578	0.0006	−3.5304	0.0005
Cyp2u1	Cytochrome P450, family 2, subfamily u, polypeptide 1	NM_027816	1.2508	0.0047	−3.3892	0.0001
Hsd17b13	Hydroxysteroid 17-beta dehydrogenase 13	NM_001163486 /// NM_198030	1.2361	0.0012	−3.2950	0.0105
Lrtm1	Leucine-rich repeats and transmembrane domains 1	NM_176920	1.2940	0.0010	−3.2036	0.0000
Cyp2c23	Cytochrome P450, family 2, subfamily c, polypeptide 23	NM_001001446 /// NM_001167905	1.2291	0.0007	−2.8640	0.0008

Table 2C.

SPNT-HCC vs CNTL: Top 20 genes with concordance of hypomethylated promoter regions and upregulation of gene expression. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to gene expression fold change.

Gene Symbols	Gene Name	RefSeq Gene	Methylation Foldchange	Methylation P-value	Expression Foldchange	Expression P-value
D17H6S56E-5	DNA segment, Chr 17, human D6S56E 5	NM_033075	−1.2201	0.0034	22.2847	0.0000
Tspan8	Tetraspanin 8	NM_001168679 /// NM_001168680 /// NM_146010	−1.4400	0.0008	14.6112	0.0000
Tnfsf13	TNF superfamily member 13	NM_001159505 /// NM_023517	−1.3919	0.0000	13.0511	0.0000
Tlr1	Toll like receptor 1	NM_030682	−1.3132	0.0000	13.0444	0.0000
Tmem71	Transmembrane protein 71	NM_172514	−1.2584	0.0051	12.1900	0.0000
Btg2	BTG anti-proliferation factor 2	NM_007570	−1.2715	0.0072	7.1245	0.0000
Casp12	Caspase 12	NM_009808	−1.2002	0.0060	4.8237	0.0001
B4galt6	Beta-1,4-galactosyltransferase 6	NM_019737	−1.2585	0.0000	4.5903	0.0000
Mmp12	Matrix metalloproteinase 12	NM_008605	−1.3047	0.0043	4.2464	0.0235
Casp4	Caspase 4	NM_007609	−1.2710	0.0087	4.0187	0.0001
Pold4	DNA polymerase delta 1, catalytic subunit	NM_027196	−1.3624	0.0018	3.9744	0.0000
Wfdc15b	WAP four-disulfide core domain 15B	NM_001045554 /// NM_138685	−1.2233	0.0067	3.5051	0.0393
Casp1	Caspase 1	NM_009807	−1.2136	0.0045	3.1855	0.0001
Cd48	CD48	NM_007649	−1.2149	0.0002	2.3810	0.0122

Table 2D.

SPNT-HCC vs CNTL: Top 20 genes with concordance of hypermethylated promoter regions and downregulation of gene expression. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to gene expression fold change.

Gene Symbols	Gene Name	RefSeq Gene	Methylation Foldchange	Methylation P-value	Expression Foldchange	Expression P-value
Gna14	G protein subunit alpha 14	NM_008137	1.2931	0.0000	−4.4985	0.0000
Mgll	Monoglyceride lipase	NM_001166249 /// NM_011844	1.2345	0.0000	−2.5745	0.0001
Aldh5a1	Aldehyde dehydrogenase 5 family member A1	NM_172532	1.2043	0.0000	−2.5135	0.0000
Id3	Inhibitor of DNA binding 3	NM_008321	1.7932	0.0000	−3.4771	0.0060
Csrp3	Cysteine and glycine-rich protein 3	NM_013808	1.3016	0.0000	−4.9180	0.0000
Cyp8b1	Cytochrome P450, family 8, subfamily b, polypeptide 1	NM_010012	1.2787	0.0001	−39.3978	0.0000
Kcnk5	Potassium channel, subfamily K, member 5	NM_021542	1.2521	0.0001	−2.0600	0.0000
Them7	Thioesterase superfamily member 7	NM_001159638 /// NM_028747	1.3703	0.0001	−2.3785	0.0004
Lppos	LIM domain containing preferred translocation partner in lipoma, opposite strand	NR_037954	1.2513	0.0002	−2.2850	0.0000
Slc46a3	Solute carrier family 46 member 3	NM_027872	1.2321	0.0004	−2.6685	0.0000
Abcc3	ATP binding cassette subfamily C member 3	NM_029600	1.2294	0.0005	−2.1311	0.0126
Tat	Tyrosine aminotransferase	NM_146214	1.4520	0.0005	−2.1282	0.0033
Asl	Argininosuccinate lyase	NM_133768	1.2134	0.0005	−2.5087	0.0000
Cyp2c23	Cytochrome P450, family 2, subfamily c, polypeptide 23	NM_001001446 /// NM_001167905	1.2108	0.0007	−3.9800	0.0001
Fam214a	Family with sequence similarity 214, member A	NM_153584	1.3726	0.0007	−3.2562	0.0000
Prlr	Prolactin receptor	NM_011169	1.2647	0.0008	−2.2455	0.0042
Pigr	Polymeric immunoglobulin receptor	NM_011082	1.3323	0.0011	−6.3871	0.0034
Hpgd	Hydroxyprostaglandin dehydrogenase 15	NM_008278	1.2429	0.0015	−2.0069	0.0018
Dpys	Dihydropyrimidinase	NM_001164466 /// NM_022722	1.2045	0.0016	−2.3561	0.0005
Lrtm1	Leucine-rich repeats and transmembrane domains 1	NM_176920	1.2303	0.0023	−4.8227	0.0000

Validation of Promoter DNA Methylation and Gene Expression of Relevant Cancer-Related Genes

Differentially methylated and differentially expressed genes in GBE-HCCs and SPNT-HCCs include genes related to xenobiotic signaling, metabolism, immune signaling, and cancer related genes. A set of cancer related genes including cMyc, and tumor suppressor genes (Spry2, Dusp5) were selected based on their differential promoter methylation in human hepatocellular carcinomas (Lee et al. 2010; Nambu et al. 1987; Wang et al. 2013). Differential methylation and gene expression profiles from the methylation array and gene expression microarray are depicted in Table 3. The fold change and p-values for these genes (cMyc; −1.04, p-value 0.3841: Spry2; 1.21, p-value <0.0001) are outside the cut off value when stringent feature analysis was applied for methylation data. Technical validation of the methylation array platform used in this study was performed via bisulfite conversion and pyrosequencing of 5’ gene promoter regions containing differentially methylated probes. Percentage methylation of the targeted region for bisulfite conversion and pyrosequencing was analyzed, cMyc promoter region was significantly hypomethylated in GBE-HCC compared to either vehicle control or SPNT-HCC (Figure 4A). In order to correlate the coordinated DNA methylation-based deregulation of cMyc, gene expression of cMyc was measured by real-time quantitative PCR, which revealed a significant increase in cMyc gene expression in GBE-HCC compared to SPNT-HCC (Figure 4B). Similarly, % methylation and gene expression levels by qRT-PCR of two candidate tumor suppressor genes, Spry2 and Dusp5, had statistically significant inverse correlation between the percentage methylation and level of gene expression (Figure 5A–5D).

Table 3.

Differential methylation and differential gene expression profiles of c-Myc, Spry2 and Dusp5 from array data.

	GBE-HCC vs. CNTL				SPNT-HCC vs. CNTL
Genes	Promoter Methylation		Gene Expression		Promoter Methylation		Gene Expression
	Fold change	p-value	Fold change	p-value	Fold change	p-value	Fold change	p-value
c-Myc	−1.04	0.3841	2.22	<0.0001	1.01	0.7517	1.48	0.0051
Myct1	−1.32	<0.0001	−1.05	0.7180	−1.23	0.0001	1.01	0.9458
Spry2	1.21	<0.0001	1.11	0.5966	−1.02	0.4496	1.70	0.0120
Dusp5	1.11	0.0161	1.14	0.0865	−1.07	0.0630	3.48	<0.0001

Figure 4.

Validation of promoter DNA methylation and gene expression of an oncogene. Methylation analysis of oncogene, cMyc by quantitative bisulphite pyrosequencing (A), gene expression analysis by RQ-PCR (B), correlation of cMyc percentile methylation and its gene expression. The promoter region of cMyc is hypomethylated and the gene expression of the oncogene mRNA is overexpressed in GBE-HCC.

Figure 5.

Validation of promoter methylation and gene expression of tumor suppressor genes. Methylation analysis of a tumor suppressor gene Spry2 (A) and Dusp5 (C) by quantitative bisulphite pyrosequencing, gene expression analysis of Spry2 (B) and Dusp5 (D) by RQ-PCR. Sprouty 2 (Spry2) and Dusp5 gene promoters are significantly hypermethylated and the corresponding mRNA expressions are downregulated in GBE-HCC, whereas in SPNT-HCC, there is no significant change in the % methylation of Spry2.

Identification of Pathways That Are Altered by DNA Methylation Changes in Mouse HCC

Since there were significant differences in DMGsand DEGs, and their overall overlap analysis between SPNT-HCC and GBE-HCC, we performed a core and comparison analysis with GBE-HCC or SPNT-HCC in relation to CNTL samples using Ingenuity® Pathway Analysis (IPA) software. The results of comparison analysis of DEGs indicated overrepresentation of genes associated with pathways involved in cell cycle control and chromosome replication, AHR, ATM, DNA damage, and NRF2-mediated oxidative stress response signaling pathways in GBE-HCCs. In SPNT-HCC, EIF2, PPARα/RXRα, and mTOR pathways were overrepresented (Figure 6A). Comparison analysis based on unique DMGs indicated enrichment of genes associated with LPS/IL-1 mediated inhibition of RXR function, ERK5 signaling, and LXR/RXR activation in GBE-HCC, whereas PTEN signaling, non-small cell lung cancer signaling genes were enriched in SPNT-HCCs (Figure 6B). When DEGs and DMGs were superimposed by overlay function in IPA, there was an inverse relationship between expression and promoter methylation of cMyc (Figure 6C, inset). The rat sarcoma virus (Ras)/Raf/MEK/ERK signaling pathway contributes a core effect in regulating cell proliferation, differentiation, survival and apoptosis. In addition to the Ras/Raf/MEK/ERK signaling pathway, the phosphoinositide 3-kinase (PI3K)/Akt signaling pathway has been previously implicated in the pathogenesis of HCC (Zhou et al. 2011). In GBE-HCC, there are enrichment of genes in PTEN signaling pathway (Figure 6B).

Figure 6.

Ingenuity pathway analysis (IPA) comparison analysis of DEGs between SPNT-HCCs and GBE-HCCs (A). Enriched genes in pathways associated with DMGs in GBE-HCCs and SPNT-HCCs (B). The overlay of DEGs and DMGs in comparison analysis of GBE-HCCs and SPNT-HCCs (C) with an inset table indicate inverse correlation of cMyc gene promoter methylation (Line 1) and cMyc gene expression (Line 2). cMyc gene promoter is hypomethylated indicated by green bar and there is corresponding overexpression of cMyc mRNA.

Several cancer related genes in GBE-HCC had differential promoter methylation, including Arhgap15, Smc4, Perp, Egfr, Tbr1, Rcbtb2, Dirc2, Lifr, Sox2, Eif4e, Pdcd4, Socs5, Runx1, cMyc, Rasgrf1, Cdh8, Wt1, Mdc1, Nras, Pigr, Spry2, Hnf1a, Socs2, Dusp6, and Pten. In SPNT-HCCs, DMGs included Bmpr1a, Cdkn1b, Pik3R1, Id3, Casp12, Casp4, Cd48, Mmp12, Aldh5a1, Pigr, and Lrtm1.

DISCUSSION

Epigenetic alterations including promoter DNA methylation are implicated in the development of human HCCs. Epigenetic changes contribute to hepatocellular tumorigenesis resulting from various etiologies (Wahid et al. 2017). Given that HCCs are driven by various known environmental factors, understanding genetic and epigenetic alterations would help better understand the underlying molecular mechanism of carcinogenesis of these tumors. The application of integrated epigenomic and transcriptomic profiling of environmental chemical exposure in animal models has enabled enhanced mechanistic interpretation and discovery of biomarkers of exposure (Terranova et al. 2017).

Hepatocarcinogenesis is a multistep process involving the deregulation of several signaling pathways (Aravalli et al. 2008). Broader understanding of the pathogenesis of human HCC has been difficult to elucidate due to the multiplicity of underlying liver disease etiologies (Woo et al. 2017). Epigenetic changes impact genetic networks, and a network-based integrative analysis of epigenetic alterations and resultant transcriptomic changes is ideally suited for exploring mechanisms of carcinogenesis. Using Ingenuity pathway analysis (IPA), we correlated data from methylation arrays and cDNA microarrays, and performed an integrative analysis in order to identify the most common aberrantly methylated genes and associated molecular pathways. Signaling pathways that are altered due to differential promoter DNA methylation in GBE-HCC include LPS/IL-1 mediated inhibition of RXR function, mouse embryonic stem cell pluripotency, ERK5 signaling, and LXR/RXR activation pathways, whereas in SPNT-HCC, the altered pathways include PTEN signaling, colorectal cancer metastasis signaling and non-small cell lung cancer signaling. cAMP-mediated signaling is the common pathway affected in both SPNT-HCCs and GBE-HCCs. Aberrant alterations in some of these pathways are reported in human HCCs (Calvisi et al. 2008; Hu et al. 2003; Hwang et al. 2004; Lo et al. 2006; Zhou et al. 2011). The mitogen-activated extracellular signal-regulated kinase 5 (ERK5) is involved in regulation of growth and development of human hepatocellular carcinoma (Rovida et al. 2015).

Activation of the RAS/MAPK pathway is commonly observed in both human and rodent hepatocellular carcinomas (Calvisi et al. 2006; Coleman 2003; Kim et al. 2005). Various cellular responses including cell proliferation, survival and differentiation are regulated by RAS activation through RAF-MEK-MAPK cascade. Point mutations in either RAS or its downstream effector B-RAF are the most common targets for constitutive gain-of-function mutations in human and rodent cancers (Hoenerhoff et al. 2013; Kim et al. 2005; Schubbert et al. 2007). Hras mutations are commonly observed in SPNT-HCC, however, in GBE-HCCs, there are decreased incidences in Hras mutations with a reciprocal increase in the incidences of Ctnnb1 mutations (Hoenerhoff et al. 2013). Crosstalk between Wnt/β-catenin and Ras/MAP kinase pathways has been reported in many different tumors (Zeller et al. 2013). The contributions of epigenetic alterations including promoter DNA methylation of genes in Wnt/β-catenin and RAS/MAP kinase pathways in mouse hepatocellular carcinoma is not completely understood; however, in human HCC, overexpression of the cMET protooncogene, and the loss of the MAPK inhibitor SPRY2 has been implicated as alternative mechanisms leading to constitutive induction of the RAS/MAPK pathway (Aravalli et al. 2008; Calvisi et al. 2006). Previous studies in human HCCs also showed that RAS pathway can be upregulated by multiple factors, including downregulation of RAS inhibitors RASSF1A (Calvisi et al. 2006) or loss of ERK inhibitor DUSP1 (Calvisi et al. 2008; Fu et al. 2006). One of the most effective mode of inactivation of tumor suppressor genes in cancer is by promoter hypermethylation. Based on the DMR analysis, hypermethylation and corresponding downregulation of gene expression of MAPK inhibitor Spry2 and ERK inhibitor Dusp5 was observed in GBE-HCC, but not in SPNT-HCCs. Thus, the present study provides evidence that the coordinated deregulation of Dusp5 and Spry2 may represent an important epigenetic mechanism for the uncontrolled activation of the RAS pathway in the absence of activating B-Raf and Ras mutations in mouse GBE-HCCs. However, further investigations at the phosphorylation sites of downstream effectors would be required to to further define this potential mechanism of chemically-induced hepatocarcinogenesis in the B6C3F1/N mouse.

cMyc is one of the most commonly overexpressed oncogenes in human cancers; its expression in mammalian cells is tightly regulated and closely linked to cell growth, apoptosis and differentiation (Dang et al. 2006). Overexpression of cMyc was reported in chronic liver disease and HCC in humans (Chan et al. 2004). Similarly, cMyc expression is increased in chemical-exposed liver cancer models as well as in hepatic injury models (Fang et al. 2004). Promoter hypomethylation of the cMYC oncogene has been observed in human HCCs (Nambu et al., 1987; Wang et al. 2013). In this study, based on methylation array data and subsequent validation by quantitative bisulfite conversion and pyrosequencing, there was significant hypomethylation of the cMyc promoter site, with concordant up-regulation of cMyc gene expression, in GBE-HCC compared to SPNT-HCC and normal age-matched CTRL liver.

In this study, using genome-wide interrogation of promoter DNA methylation, about 32 and 28 genes were coordinately hypermethylated and down-regulated in GBE-HCCs and in SPNT-HCCs, respectively. Among these genes, Saa4, Pdcd4, Acadsb, C8a, C8b, Asic5, Dmgdh, Cbs, Slc3a1, Hsd17b13, Ido2, Mup7, Lipc, Perp, Kyat1, Mup19, Cyp2u1, Dbt, S1pr5, Galt, Mup8, Sdr42e1, Mup13, and Mup9 were uniquely altered only in GBE-HCCs, and Gna14, MglI, Aldh5a1, Id3, Csrp3, Kcnk5, Them7, Lppos, Slc46a3, Abcc3, Asl, Fam214a, Prlr, Hpgd, Dpys, Corin, Nnmt, Csrp3, and Tcf21 were uniquely altered only in SPNT-HCCs. Within the statistically significant top hypermethylated and downregulated concordant genes, there are several members of Major urinary proteins (Mup7, 8, 9, 10, 12, 13, 19), apoptosis (Perp), xenobiotic metabolism (Cyps, Asic5, Hsd17b13) and immune response related genes (C8a, C8b) in GBE-HCCs group, and transcription factors (Id3, Lrtm1), and xenobiotic metabolism genes (Cyps), in SPNT-HCC groups.

Several members of major urinary proteins (MUPs) were coordinately hypermethylated and downregulated in GBE-HCC compared to SPNT-HCC. The MUPs are extracellular proteins in the lipocalin family, which are mainly synthesized in the liver. MUP expression levels were significantly lower in SPNT–HCC, and has previously been shown to be reduced in dimethylaminoazobenzene and urethane-induced mouse HCC (Dragani et al. 1989). Based on decreased RNA and protein levels of MUP in diethyl nitrosamine-induced microscopic liver nodules and HCCs, it was suggested that the decreased MUP gene expression was an early event in mouse hepatocarcinogenesis in the male mouse (Dragani et al. 1989). However, the biological significance of epigenetic silencing of this set of genes in mouse HCC is largely unknown, and further characterization of this potential tumor suppressor is needed to understand its role in mouse hepatocarcinogenesis.

In addition to epigenetic silencing of apoptotic genes such as Perp, and xenobiotic genes including several Cyps (Cyp8b1, Cyp2c23, Cyp2f2, Cyp2u1), a novel tumor suppressor, Programmed cell death 4 (Pdcd4) was one of the top 20 hypermethylated and down-regulated genes in GBE-HCC. Pdcd4 expression is frequently downregulated in several types of cancers including human HCC and colon cancer (Ding et al. 2016; Wang et al. 2008). Downregulation of the Pdcd4 tumor suppressor gene is associated with activation of β-catenin/Tcf and AP-1-dependent transcription via cMyc and MAP4K1 (Wang et al. 2008; Wang et al. 2012).

Similarly, 29 genes in GBE-HCC and 16 genes in SPNT-HCCs had hypomethylation of promoter DNA and concordant up-regulation of gene expression. Amongst these genes, a unique set of genes in GBE-HCC included Esm1, Gpnmb, Ctla2b, Spic, Plek, Nckap1l, c1qb, Ctla2a, Sema3b, Cd34, C1qc, Treml4, Nedd4, Gbp2b, Sdcbp2, Ear2, Ear12, Ear3, Lilr4b, Clec4a3 and Ms4a7. Uniquely hypomethylated and upregulated genes in SPNT-HCC included Tnfsf13, B4galt6, Tlr1, Tspan8, Tmem71, Wrdc15b and Btg2. There is evidence that promoter hypomethylation of some genes may be associated with the development of cancers by regulating the activity of other genes (Pulukuri et al. 2007). Some of these hypomethylated genes, including Gpnmb (Tian et al. 2013), Esm1 (Kang et al. 2011), and Clec4a3 (Hou et al. 2018), have been shown to be upregulated in HCC and other liver diseases including NAFLD (Katayama et al. 2015) in humans, and promoter hypomethylation of immune-related genes is thought to possibly promote carcinogenesis (Son et al. 2010). Several immune-related genes were hypomethylated in GBE-HCC, including Ctla2b, Ctla2a, C1qb, Cd34, C1qc, Nedd4, Tnfsf13, Tlr1 and Gpb2b.

In addition to the concordant genes in GBE-HCC, there were several cancer-related genes that were differentially methylated without concordance with associated gene expression levels. These genes included Arhgap15, Smc4, Perp, Egfr, Tbr1, Rcbtb2, Dirc2, Lifr, Sox2, Eif4e, Pdcd4, Socs5, Runx1, cMyc, Rasgrf1, Cdh8, Wt1, Mdc1, Nras, Pigr, Spry2, Hnf1a, Socs2, Dusp6, Esm1, Gpnmb, Ctla2b and Pten. Similar genes showed lack of concordance between promoter methylation and gene expressoin in SPNT-HCC, including Bmpr1a, Cdkn1b, Pik3R1, Id3, Casp12, Casp4, Cd48, Mmp12, Aldh5a1, Tnfsf13, B4galt6, Tlr1, Btg2, Pigr and Lrtm1. Expression of several of these genes, including Socs2 and Socs5 (Cui et al. 2016), Pdcd4 (Wang et al. 2012), Spry 2, Dusp6, and Pten, is significantly decreased in human HCC. Recently, a unique set of six HCC-specific CpGs mapped to four genes (NEBL, FAM55c, GALNT3, DSE) were identified as biomarkers for diagnosing human HCCs (Cheng et al. 2018).

In summary, our data demonstrates that GBE-HCC harbors distinct promoter DNA methylation profiles compared to SPNT-HCC. For example Saa4, Pdcd4, Acadsb, C8a, C8b, Asic5, Dmgdh, Cbs, Slc3a1, Hsd17b13, Ido2, Mup7, Lipc, Perp, Kyat1, Mup19, Cyp2u1, Dbt, S1pr5, Galt, Mup8, Sdr42e1, Mup13 and Mup9 had unique methylation alterations only in GBE-HCC, whereas methylation alterations of Gna14, MglI, Aldh5a1, Id3, Csrp3, Kcnk5, Them7, Lppos, Slc46a3, Abcc3, Asl, Fam214a, Prlr, Hpgd, Dpys, Corin, Nnmt, Csrp3 and Tcf21 were unique to SPNT-HCC. Several of these uniquely altered methylated genes may serve as biomarkers for exposure or disease, and could potentially help in differentiating chemically-induced tumors from those arising spontaneously. In order to validate these findings, we are further characterizing promoter methylation profiles of mouse HCCs arising either spontaneously or due to chronic exposure to several genotoxic and non-genotoxic carcinogens. These additional studies will provide a more detailed understanding of unique methylation changes in chemically induced HCC compared to spontaneously arising HCC in B6C3F1/N mice as models of human disease.

Supplementary Material

1

Table S1A. GBE-HCC vs CNTL: Top 20 genes with Hypomethylated promoter regions. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to fold change.

Table S1B. GBE-HCC vs CNTL: Top 20 genes with hypermethylated promoter regions. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to fold change.

Table S1C. SPNT-HCC vs CNTL: Top 20 genes with hypomethylated promoter regions. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to fold change.

Table S1D. SPNT-HCC vs CNTL: Top 20 genes with hypermethylated promoter regions. Genes with at least 4 probes overlapping +/− 2Kb of TSS and P-value <=0.005 ordered according to fold change.

Table S2A: GBE-HCC vs CNTL: Top 20 downregulated genes. Genes with P-value <=0.005 ordered according to fold change.

Table S2B. GBE-HCC vs CNTL: Top 20 upregulated genes. Genes with P-value <=0.005 ordered according to fold change.

Table S2C. SPNT-HCC vs CNTL: Top 20 downregulated genes. Genes with P-value <=0.005 ordered according to fold change.

Table S2D. SPNT-HCC vs CNTL: Top 20 upregulated genes. Genes with p-value <=0.005 ordered according to fold change.

Table S3. List of primers for pyrosequencing

Table S4. List of qRT-PCR Primers

Abbreviations

GBE

Ginkgo biloba extract

HCC

Hepatocellular carcinoma

SPNT

Spontaneous

CNTL

Control

Myc

Myelocytomatosis

Spry2

Sprouty RTK signaling antagonist 2

Dusp5

Dual specificity phosphatase 5

DEGs

Differentially expressed genes

DMR

Differentially methylated regions

DMGs

Differentially methylated genes

H & E

Hematoxylin and eosin

FFPE

Formalin-fixed paraffin embedded

RMA

Robust Multiarray Normalization

ANOVA

Analysis of variance