Chronic inflammation represents a major risk factor for tumor formation, but the underlying mechanisms have remained largely unknown. Epigenetic mechanisms can record the effects of environmental challenges on the genome level and could therefore play an important role in the pathogenesis of inflammation-associated tumors. Using single-base methylation maps and transcriptome analyses of a colitis-induced mouse colon cancer model, we identified a novel epigenetic program that is characterized by hypermethylation of DNA methylation valleys that are characterized by low CpG density and active chromatin marks. This program is conserved and functional in mouse intestinal adenomas and results in silencing of active intestinal genes that are involved in gastrointestinal homeostasis and injury response. Further analyses reveal that the program represents a prominent feature of human colorectal cancer and can be used to correctly classify colorectal cancer samples with high accuracy. Together, our results show that inflammatory signals establish a novel epigenetic program that silences a specific set of genes that contribute to inflammation-induced cellular transformation. Cancer Res; 75(10); 2120–30. ©2015 AACR.

Inflammation has been linked to the pathogenesis of tumors in a substantial fraction of human cancers (1, 2). Inflammatory processes play a causal role in cancer development through cellular pathways that involve genotoxicity, aberrant tissue repair, proliferative responses, invasion, and metastasis (2). The inflammatory microenvironment is an essential component of almost all tumors, including some for which a causal relationship with inflammation is not yet proven (3). One of the most widely studied and most prevalent inflammation-induced cancers is inflammatory bowel disease (IBD)-associated colorectal cancer. Patients suffering from long-standing IBD have a significantly higher risk of developing colorectal cancer (4).

Altered DNA methylation patterns have long been associated with tumor formation and represent one of the earliest and most consistent molecular markers of human cancers (5–7). Traditionally, most studies of DNA methylation have focused mainly on CpG islands, which are unmethylated and generally associated with promoters. It is well known that subsets of CpG islands can become aberrantly hypermethylated in cancer (6–10). More recently, genome-wide methylation pattern analyses have begun to systematically uncover key features of tumor methylomes (8, 9, 11). These include not only hypermethylated CpG islands, but also large (several 100 kb to several Mb) partially methylated domains (PMD) that are gene-poor and colocalize with lamina-associated domains (8, 9). Another novel feature of methylomes that shows altered DNA methylation patterns in tumors has been termed DNA methylation valleys (DMV). These regions extend over several kilobases of DNA, are strongly hypomethylated in most tissues, and are enriched for transcription factors and developmental genes (12). DMVs have been shown to become hypermethylated in colorectal cancer and may thus contribute to the aberrant epigenetic programming of tumor cells (12). Regions similar to DMVs, termed methylation canyons, were recently identified in hematopoietic stem cells and were shown to be distinct from CpG islands and CpG island–associated features (13).

While a few initial studies have suggested aberrant methylation patterns in mouse models of IBD (14, 15), both the effect of inflammation on DNA methylation and the functional outcome have not been investigated in detail yet. We have now used whole-genome bisulfite sequencing (16) to characterize the methylomes of the AOM/DSS mouse model at single-base resolution. In this model, mice are treated with dextran sodium sulfate (DSS) to induce colitis (17). When this treatment is preceded by injections of the weak carcinogen azoxymethane (AOM), the mice develop intestinal tumors (18). Our results identify hypermethylated DMVs as a prominent feature of the colitis methylome that is conserved in intestinal adenocarcinomas and in human colon cancer. Our findings provide strong support for the hypothesis that inflammatory signals induce a higher risk for cancer development by establishing a novel epigenetic program in enterocytes.

Preparation of samples

Male mice (C57BL/6) were obtained from Harlan Laboratory at 6 to 7 weeks of age. The mice were left for one week for adaptation and then treated to induce inflammation and/or cancer. In the adenoma analysis, both male and female mice at 4 to 6 months of age were used. For inflammation induction, mice were given 2% of 36 to 50 kDa DSS (MP Biomedicals) for one week in their drinking water, and then left for 2 weeks on regular drinking water. This cycle was repeated 3 times. To induce inflammation-associated cancer, the same protocol was used but the DSS cycles were preceded by two intraperitoneal doses of 12.5 mg/kg AOM (Sigma) given with one week difference (Supplementary Fig. S1). Mice were housed and cared for under specific pathogen-free conditions and all the animal procedures were approved by the Animal Care and Use Committee of the Hebrew University of Jerusalem.

For epithelial cell preparation from colon, we followed a previously published protocol (19). Briefly, the colon was flushed with ice-cold PBS and then cut open longitudinally. Distal and proximal colon regions were processed separately. Each part was minced into small pieces in cold PBS. Enterocytes were mechanically isolated by shaking in PBS with 30 mmol/L EDTA at 37°C. Crypts were recovered and then stored at −80°C for DNA and RNA extraction. Only samples from the distal colon were used in our study. For histologic analysis of colon and tumor samples, representative pieces of distal and proximal colon were used. Cancer cells were extracted from distal colon crypts of samples showing a high percentage of tumor tissue using histologic assessments, as described above. Preparation and maintenance of organoid cultures were performed as previously reported (19, 20). 4-methylpyrazol treatment was initiated directly after organoid derivation using 1 mmol/L final concentration. Organoids were monitored daily and size was measured using NIS software. Gene expression analysis was performed on organoids grown for 8 days in culture.

Analysis of DMVs

DMVs were identified as described previously (12) and further analyzed by visual inspection of methylation tracks in the UCSC Genome Browser. The refSeq gene annotation database was used for the identification of DMV-associated genes. A DMV that overlapped the gene body or promotor region of a gene was assigned to that gene. Analyses of the chromatin status of DMVs and DMV surrounding regions were performed using the cistrome online platform (21). Intestinal ChIP-seq data from the mouse ENCODE project were used in our analysis (GSE31039).

454 DNA bisulfite sequencing

Deep DNA bisulfite sequencing was performed as described previously (22). See Supplementary Materials and Methods for details.

Support vector machine–based classification of tumor and control samples

For the 30 genes that were found to be associated with hypermethylated DMVs and downregulated in the AOM/DSS model, we identified 28 human homologs. Of these, 18 were associated with a DMV or a DMV-like structure in a normal human colon methylome (23). We then used a support vector machine with a linear kernel (software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm; ref. 24) to classify the RNA-seq data of the The Cancer Genome Atlas (TCGA) colon cancer set, based on 10-fold cross-validation. To identify a minimum gene set for sample classification, the gene with the lowest F-score was removed, as long as the classification accuracy did not drop below 100%.

Analysis of TCGA datasets

We obtained expression and methylation data for colon adenocarcinoma from the TCGA database (as of February/March 2014). Expression data comprised normalized gene expression values, from Illumina HighSeq2000 RNASeqV2 experiments, for 300 samples (259 tumor, 41 normal). Methylation analysis was performed on genome-mapped beta values (third data level) derived from the TCGA Infinium HumanMethylation450 dataset consisting of 329 samples (291 tumor, 38 normal). For each gene (NR5A2, ECH1, and FOXP2), three adjacent Infiniums probes were selected with regard to their positioning in DMVs. Box plots were computed using the statistical computing language R (version 3.0.2). The additional package rgl provided tools for visualizing the combined gene expression. Welch two-sample t tests were performed to assess statistical significance.

Data access

Sequencing data have been deposited in the GEO database under the accession number GSE57569.

Whole-genome DNA methylation analysis of a mouse model for inflammation-induced intestinal carcinogenesis

To gain insight into the potential contribution of DNA methylation to intestinal inflammation and tumorigenesis, we used the AOM/DSS mouse model of inflammation-induced colon cancer (Supplementary Fig. S1). In this model, IBD was induced by adding 2% DSS to the drinking water during weeks 2, 5, and 8. In parallel experiments, intestinal tumorigenesis was induced by additional injections of AOM. We determined base-resolution DNA methylomes of normal as well as inflamed (DSS-treated) and cancer (AOM/DSS-treated) intestinal epithelial cells. To obtain sufficient amounts of DNA for whole-genome bisulfite sequencing (WGBS), equimolar sample pools were prepared from intestinal epithelial cells or tumors from 3 mice for each group, resulting in 3 libraries (control, IBD, and tumor) for sequencing. Paired-end sequencing was performed on an Illumina HiSeq 2000 platform (see Supplementary Materials and Methods for details), with average genome coverage of 19.4× (control), 13.8× (IBD), and 11.4× (tumor). We also determined the bisulfite conversion rate by analyzing mitochondrial sequences that were copurified during the sample preparation and that are known to be unmethylated. These sequences showed a bisulfite conversion rate of >99.5% (see Supplementary Materials and Methods for details), suggesting highly effective bisulfite treatment.

Initial data analysis revealed many basal features of mammalian methylomes. This included the specificity for CpG dinucleotides and a characteristic bimodal enrichment for completely unmethylated and completely methylated CpG residues (data not shown). The average methylation ratios of the three datasets were very similar (Supplementary Fig. S2A), and a direct comparison of the methylation landscapes showed a high degree of similarity between the three datasets (Supplementary Fig. S2B). Partially methylated domains, which represent large regions spanning several 100 kb with reduced methylation levels, and which have been observed in a variety of human and mouse tumor methylomes (8, 9, 11, 25), were not observed in any of the datasets (Supplementary Fig. S2C).

Inflammation-induced hypermethylation of DMVs

Our results indicate that AOM/DSS treatments induce localized aberrations rather than the formation of large-scale DNA methylation changes. We therefore segmented our datasets according to gene substructure and observed a progressive (from control to IBD and then to tumor) and highly significant (P < 0.01) increase in the methylation levels of promoters, 5′-UTRs, and exons (Fig. 1A). This effect was not observed for 3′-UTRs and introns, which had highly similar methylation levels in all three datasets (Fig. 1A). Furthermore, a minor but significant (P < 0.01) decrease in methylation levels was observed for intergenic regions in the IBD and tumor datasets (Fig. 1A). These findings indicate that hypermethylation in AOM/DSS inflammation-induced cancer occurs at 5′ gene regions.

Figure 1.

Characterization of DMVs in the AOM/DSS model. A, average DNA methylation ratios of various intragenic subsegments are shown for the control (blue), IBD (green), and tumor sample (red), respectively. Visible methylation differences between individual samples were highly significant (P < 0.01), as determined by a paired t test. B, box plots showing the distribution of methylation ratios for CpG islands (left) and DMVs (right) in the control (blue), IBD (green), and tumor sample (red), respectively. Only DMVs with >20 CpGs were included in this analysis to ensure comparability and analytical robustness. Differences between samples were not significant (P > 0.05, two-tailed t test) for CpG islands, but highly significant for DMVs (as indicated). C, Venn diagram illustrating the numbers and conservation of DMVs between the samples. D, average sequence conservation level around DMVs. E, chromatin profile of DMVs. The highest enrichment is observed for H3K4me3 (green), followed by H3K27ac (purple), H3K4me1 (red), and H3K27me3 (black).

Figure 1.

Characterization of DMVs in the AOM/DSS model. A, average DNA methylation ratios of various intragenic subsegments are shown for the control (blue), IBD (green), and tumor sample (red), respectively. Visible methylation differences between individual samples were highly significant (P < 0.01), as determined by a paired t test. B, box plots showing the distribution of methylation ratios for CpG islands (left) and DMVs (right) in the control (blue), IBD (green), and tumor sample (red), respectively. Only DMVs with >20 CpGs were included in this analysis to ensure comparability and analytical robustness. Differences between samples were not significant (P > 0.05, two-tailed t test) for CpG islands, but highly significant for DMVs (as indicated). C, Venn diagram illustrating the numbers and conservation of DMVs between the samples. D, average sequence conservation level around DMVs. E, chromatin profile of DMVs. The highest enrichment is observed for H3K4me3 (green), followed by H3K27ac (purple), H3K4me1 (red), and H3K27me3 (black).

Close modal

The two epigenetic features that are commonly associated with 5′ regions of genes are CpG islands and DMVs. We therefore compared the methylation levels of all annotated CpG islands and DMVs in our datasets. This analysis failed to reveal any significant methylation changes in CpG islands, but interestingly showed a progressive and highly significant (P < 0.01, two-tailed t test) hypermethylation of DMVs in the IBD and tumor datasets (Fig. 1B). Of the 1753 DMVs that were identified in the normal intestinal methylome, 1,048 were shared between all three samples, while several hundred were unique for specific samples (Fig. 1C), suggesting that a subset of DMVs undergoes dynamic methylation changes in our model. In agreement with previous findings (12), the DMVs that we identified in the control intestinal methylome showed robust sequence conservation (Fig. 1D), and a moderate enrichment for the H3K4me1 and H3K27me3 histone modifications (Fig. 1E). However, a stronger and more defined enrichment was observed for H3K4me3 and H3K27ac (Fig. 1E), two modifications that are usually associated with active gene expression. This finding is in agreement with the notion that DMVs are often associated with active genes (12).

Having established that DNA methylation is indeed dynamic at specific DMVs during inflammation, we performed a more detailed analysis of the methylome data. Interestingly, this survey revealed that DMVs found in normal intestinal methylome were globally hypermethylated in the IBD and tumor samples (Supplementary Fig. S3). We also identified a set of 99 DMVs that showed robust hypermethylation (methylation difference >0.1) in the IBD or the tumor sample (see Fig. 2A, for an example). Remarkably, the majority of these DMVs became hypermethylated in epithelial cells from IBD mice (Fig. 2B). A closer analysis of the hypermethylated DMVs revealed additional novel features that expand previous analyses (12). For example, average DMV methylation profiles showed that hypermethylation was confined to valley floors and did not spread beyond the borders of the DMVs (Fig. 2C). Furthermore and of note, hypermethylated DMVs were strongly depleted for CpG islands and also showed reduced CpG densities, when compared with the complete set of DMVs (Fig. 2D and E). Finally, DMVs that became hypermethylated were also enriched for the active chromatin marks H3K27ac and H3K4me3 in normal mouse intestine (Fig. 2F). Together, these data suggest that inflammation-related methylation changes are targeted to a specific subset of DMVs that are associated with active genes and thus define an epigenetic program that is clearly distinct from the previously described H3K27me3-associated de novo methylation of CpG island-associated cancer genes (26–28).

Figure 2.

Inflammation-induced hypermethylation of DMVs. A, Adh1 as a representative example for a DMV-associated gene. DNA methylation and histone modifications are shown as UCSC Genome Browser tracks. The DMV (shaded in orange) overlaps the promotor, TSS, and two exons. B, heatmap showing methylation ratios of DMVs that became hypermethylated (methylation difference >0.1) in the IBD or the tumor sample. C, average methylation profile of hypermethylated DMVs. Hypermethylated DMVs were normalized for their different lengths. D, hypermethylated DMVs are depleted for CpG islands (CGI). Bars show the fraction of DMVs that contain CpG islands. Results are shown for all DMVs (All) and for hypermethylated DMVs (Hyper). E, hypermethylated DMVs show reduced CpG density. Bars show the CpG density for all DMVs (All) and for hypermethylated DMVs (Hyper). Red line, the average CpG density of the mouse genome. F, chromatin profile of hypermethylated DMVs. Robust enrichment is observed for H3K4me3 (green) and H3K27ac (purple). No enrichment could be observed for H3K4me1 (red) and H3K27me3 (black). G, Ingenuity Pathway Analysis of the 102 genes that are associated with DMVs that were lost (due to hypermethylation) and became downregulated in the IBD or tumor samples. In addition to “Cancer,” the five most highly enriched functions are shown.

Figure 2.

Inflammation-induced hypermethylation of DMVs. A, Adh1 as a representative example for a DMV-associated gene. DNA methylation and histone modifications are shown as UCSC Genome Browser tracks. The DMV (shaded in orange) overlaps the promotor, TSS, and two exons. B, heatmap showing methylation ratios of DMVs that became hypermethylated (methylation difference >0.1) in the IBD or the tumor sample. C, average methylation profile of hypermethylated DMVs. Hypermethylated DMVs were normalized for their different lengths. D, hypermethylated DMVs are depleted for CpG islands (CGI). Bars show the fraction of DMVs that contain CpG islands. Results are shown for all DMVs (All) and for hypermethylated DMVs (Hyper). E, hypermethylated DMVs show reduced CpG density. Bars show the CpG density for all DMVs (All) and for hypermethylated DMVs (Hyper). Red line, the average CpG density of the mouse genome. F, chromatin profile of hypermethylated DMVs. Robust enrichment is observed for H3K4me3 (green) and H3K27ac (purple). No enrichment could be observed for H3K4me1 (red) and H3K27me3 (black). G, Ingenuity Pathway Analysis of the 102 genes that are associated with DMVs that were lost (due to hypermethylation) and became downregulated in the IBD or tumor samples. In addition to “Cancer,” the five most highly enriched functions are shown.

Close modal

Inflammation-induced gene expression changes

To better understand the role of this unique set of hypermethylated DMVs, we performed transcriptome sequencing of control, IBD, and tumor samples. mRNA was prepared from 3 to 4 individual mice, then barcoded and sequenced on an Illumina HiSeq 2000 platform (see Supplementary Experimental Procedures for details). Data analysis showed high expression levels of epithelial marker genes, such as Epcam and Cdh1 (Supplementary Table S1). Immune genes were only expressed at low levels, with no significant increase in the IBD or tumor samples (Supplementary Table S1), reflecting the high purity of our epithelial cell preparations. Furthermore, expression levels for genes encoding DNA methylation enzymes, such as Dnmt1-3 and Tet1-3, and their known cofactors Dnmt3l and Uhrf1 also appeared very similar in all samples (Supplementary Table S1), consistent with the absence of large-scale methylation changes in our mouse model.

To identify genes that were differentially expressed in the IBD and tumor samples, we used DESeq (29). This analysis revealed 791 and 1,597 differentially expressed (Q < 0.05) genes in the IBD and tumor samples, respectively (Supplementary Fig. S4A). Of these, a large fraction showed relatively minor quantitative changes (Supplementary Fig. S4A). We therefore applied an additional cutoff of ≥2-fold expression change, which revealed that the majority of genes that were deregulated in the IBD samples also showed concordant deregulation in the tumor samples (Supplementary Fig. S4B), thus illustrating the close relationship between intestinal inflammation and tumorigenesis. Pathway analysis of the 299 commonly upregulated genes revealed a highly significant enrichment of genes involved in immunologic, gastrointestinal, and inflammatory diseases categories (Supplementary Fig. S4C), consistent with the phenotypic changes observed in mice. Interestingly, a closer analysis of the downregulated genes revealed several genes that were associated with a differentially methylated DMV. To systematically investigate the relationship between differential expression and differential DNA methylation of nearby DMVs in the intestinal epithelium of the inflamed and tumorigenic tissue, we integrated our methylome and transcriptome datasets. This identified 102 genes that are associated with DMVs that disappeared (due to hypermethylation) and became significantly (Q < 0.05) downregulated in the IBD or tumor samples. Pathway analysis of these genes revealed a significant enrichment of genes involved in biologic processes directly relevant to tumor cell phenotypes (Fig. 2G). These results provide additional insight into the characteristic features of the inflammation-induced epigenetic program.

Hypermethylation-induced silencing of DMV-associated genes

Further data analysis identified 58 genes that are associated with strongly hypermethylated DMVs (methylation difference >0.1) in the IBD or tumor samples. Among these, 30 genes showed relatively high expression levels in the control sample and reduced expression levels in the IBD or tumor samples, indicating methylation-dependent silencing in the inflamed and cancer tissues (Supplementary Table S2). To validate and further analyze the methylation changes at DMVs chosen from the above described subgroup, we used targeted deep bisulfite sequencing. DNA was prepared from 3 individual mice per group and bisulfite converted. Eight regions with predicted differential methylation (as observed in our WGBS datasets) were then PCR amplified and sequenced by 454 technology, which routinely generates sequencing coverages exceeding 100×. The resulting methylation profiles provided strong confirmation for the WGBS-based results and also confirmed DMV hypermethylation in tumors from AOM mice (Supplementary Table S3). A prominent example is Nr5a2, which encodes a nuclear receptor that has been shown to have a protective function against IBD (30). Our 454 bisulfite sequencing data show that the Nr5a2 DMV was largely unmethylated in intestinal epithelial cells from 3 independent control mice (Fig. 3A). In contrast, epithelial cells from 3 independent DSS-treated mice and intestinal tumors from 3 independent AOM/DSS-treated mice showed robust de novo methylation in the Nr5a2 DMV (Fig. 3A). These findings are consistent with the notion that hypermethylation of specific DMVs is an early event in inflammation-associated intestinal tumorigenesis. Similar effects were observed for 5 additional loci from our validation set (Adh1, B3gnt7, Kcne3, Sorbs2, Vdr), with little variation between individual mice (Fig. 3B). Only 2 DMVs (Cdx1 and Prelp) showed a less pronounced hypermethylation and appeared to be more strongly affected by individual variation (Fig. 3B).

Figure 3.

DMV hypermethylation is associated with gene silencing. A, 454 bisulfite sequencing results for the Nr5a2 DMV in three different groups of mice (control, DSS, AOM/DSS). PCR amplification was performed on DNA samples from individual mice and sequencing results are shown as heatmaps. Each row represents one sequence read, blue boxes represent methylated CpG dinucleotides, and yellow boxes represent unmethylated CpG dinucleotides. Sequencing coverage ranged from 312 to 495 reads, as indicated. B, 454 bisulfite sequencing results for the complete validation set. The heatmap shows average methylation ratios of 8 DMV amplicons from 9 individual mice. C, relative expression levels of genes associated with the differentially methylated DMVs shown in B. Gene expression data was extracted from RNAseq datasets; gene expression levels are indicated relative to the maximum gene-specific expression level across all individual samples. Q values (as determined by DESeq) are indicated relative to controls.

Figure 3.

DMV hypermethylation is associated with gene silencing. A, 454 bisulfite sequencing results for the Nr5a2 DMV in three different groups of mice (control, DSS, AOM/DSS). PCR amplification was performed on DNA samples from individual mice and sequencing results are shown as heatmaps. Each row represents one sequence read, blue boxes represent methylated CpG dinucleotides, and yellow boxes represent unmethylated CpG dinucleotides. Sequencing coverage ranged from 312 to 495 reads, as indicated. B, 454 bisulfite sequencing results for the complete validation set. The heatmap shows average methylation ratios of 8 DMV amplicons from 9 individual mice. C, relative expression levels of genes associated with the differentially methylated DMVs shown in B. Gene expression data was extracted from RNAseq datasets; gene expression levels are indicated relative to the maximum gene-specific expression level across all individual samples. Q values (as determined by DESeq) are indicated relative to controls.

Close modal

To specifically investigate the effect of the experimentally confirmed DMV methylation changes on gene expression, we reanalyzed our RNA-seq datasets. Expression data for the set of 8 genes included in our 454 bisulfite sequencing analysis were available from 10 mice and showed a noticeably consistent reduction in gene expression levels for DSS-treated and AOM/DSS-treated mice, relative to controls (Fig. 3C). This effect was observed for all 8 DMV-associated genes that were included in our analysis and was highly significant (Q < 0.05) in the majority of cases (Fig. 3C). Together, these findings indicate coordinated hypermethylation-induced gene silencing of specific loci during the inflammatory process.

Functional significance of DMV-associated gene silencing in intestinal cancer

To test whether hypermethylation of our unique set of DMVs is a general process occurring in other intestinal mouse malignancies, we analyzed their DNA methylation pattern in adenoma samples derived from APC (Min) mice (31), using 454 technology. Interestingly, we found increased methylation levels in three tested DMVs relative to their wild-type (WT) counterparts (Fig. 4A and B and Supplementary Fig. S5). Consistent with this methylation increase, a significant downregulation was observed in their associated genes (Fig. 4C). This suggests that the inflammation-associated epigenetic changes are conserved in mouse premalignant adenomas.

Figure 4.

DNA methylation-induced silencing of inflammation-associated DMVs in adenomas from APC (Min) mice. A, representative 454 bisulfite sequencing results for the Vdr DMV in wild-type (4 mice) and adenoma-derived (4 mice) epithelium. PCR amplification was performed on DNA samples from individual mice and sequencing results are shown as heatmaps. Each row represents one sequence read, dark gray boxes represent methylated CpG dinucleotides, and light gray boxes represent unmethylated CpG dinucleotides. Sequencing coverage ranged from 646 to 896 reads, as indicated. B, quantification of 454 bisulfite sequencing results for three DMV amplicons from 9 individual mice. Normal intestinal epithelial cells were derived from wild-type mice (n = 4) and adenoma epithelial cells from Min mice (n = 5). Values are means and error bars indicate SEM. P values were calculated using a t test (*, P ≤ 0.05; **, P ≤ 0.001). C, normalized expression levels of genes associated with the differentially methylated DMVs in wild-type- (n = 6) and adenoma (n = 4) -derived epithelium. Values are means and error bars indicate SEM. P values were calculated using a t test (*, P ≤ 0.01; **, P ≤ 0.001).

Figure 4.

DNA methylation-induced silencing of inflammation-associated DMVs in adenomas from APC (Min) mice. A, representative 454 bisulfite sequencing results for the Vdr DMV in wild-type (4 mice) and adenoma-derived (4 mice) epithelium. PCR amplification was performed on DNA samples from individual mice and sequencing results are shown as heatmaps. Each row represents one sequence read, dark gray boxes represent methylated CpG dinucleotides, and light gray boxes represent unmethylated CpG dinucleotides. Sequencing coverage ranged from 646 to 896 reads, as indicated. B, quantification of 454 bisulfite sequencing results for three DMV amplicons from 9 individual mice. Normal intestinal epithelial cells were derived from wild-type mice (n = 4) and adenoma epithelial cells from Min mice (n = 5). Values are means and error bars indicate SEM. P values were calculated using a t test (*, P ≤ 0.05; **, P ≤ 0.001). C, normalized expression levels of genes associated with the differentially methylated DMVs in wild-type- (n = 6) and adenoma (n = 4) -derived epithelium. Values are means and error bars indicate SEM. P values were calculated using a t test (*, P ≤ 0.01; **, P ≤ 0.001).

Close modal

We also sought to investigate the functional significance of genes that were regulated by our specific set of hypermethylated DMVs. To this end, we used organoid cultures, derived from adenomas developing in APC (Min) mice (adenoma organoids) or WT mice (WT organoids; ref. 20). These cultures provide a pure system to investigate whether the effect of these epigenetic lesions are maintained independently of stroma and infiltrating immune cells, as well as a mean to study the role of these lesions in cellular transformation. For the first time, we have adopted this system to decipher the role of DNA methylation in inflammation-induced cancer. Adenoma organoids showed an increased activity of Wnt signaling (Fig. 5A), a hallmark of intestinal adenomas, exemplifying their independence from external Wnt ligands, as previously reported (20). Interestingly, in these organoids 4 of 5 DMV-associated genes were significantly hypermethylated (Fig. 5B) and their expression level was strongly reduced (Fig. 5C). This clearly shows that the organoid cultures mimic the in vivo methylation and expression patterns of our DMV-associated genes, and thus are suitable for studying their role in inflammation-induced intestinal cancer.

Figure 5.

Functional silencing of DMV-associated genes in adenoma-derived organoids. A, normalized expression levels of Wnt target genes in wild-type (WT) and adenoma-derived intestinal organoids. B, 454 bisulfite sequencing results for four DMVs in freshly isolated (8 days in culture) WT- and adenoma-derived organoids. C, expression levels of DMV-associated genes in WT- and adenoma-derived intestinal organoids. Values are means and error bars indicate SD. The results are representative of two independent derivations of organoids in each group. D, representative images of WT intestinal organoids treated or not treated with 1 mmol/L 4-methylpyrazole (4-MP) for 4 days (see Materials and Methods). Scale bar, 50 μm. E, quantification of organoid size using area size in μm2. **, P ≤ 0.05, as calculated by a t test. Normalized expression levels of Wnt signaling target genes (F) and proliferation genes (G) in untreated (u) or 8 days 4–MP-treated WT intestinal organoids. Values are means and error bars indicate SD. The results are representative of two independent derivations of organoids in each group.

Figure 5.

Functional silencing of DMV-associated genes in adenoma-derived organoids. A, normalized expression levels of Wnt target genes in wild-type (WT) and adenoma-derived intestinal organoids. B, 454 bisulfite sequencing results for four DMVs in freshly isolated (8 days in culture) WT- and adenoma-derived organoids. C, expression levels of DMV-associated genes in WT- and adenoma-derived intestinal organoids. Values are means and error bars indicate SD. The results are representative of two independent derivations of organoids in each group. D, representative images of WT intestinal organoids treated or not treated with 1 mmol/L 4-methylpyrazole (4-MP) for 4 days (see Materials and Methods). Scale bar, 50 μm. E, quantification of organoid size using area size in μm2. **, P ≤ 0.05, as calculated by a t test. Normalized expression levels of Wnt signaling target genes (F) and proliferation genes (G) in untreated (u) or 8 days 4–MP-treated WT intestinal organoids. Values are means and error bars indicate SD. The results are representative of two independent derivations of organoids in each group.

Close modal

The availability of organoid cultures also provided an excellent opportunity to functionally characterize a novel candidate gene that is subjected to DMV-associated hypermethylation and gene silencing. Treatment of WT organoids with 4-methylpyrazole (4-MP), a competitive inhibitor of alcohol dehydrogenases, including Adh1, led to enhanced growth of freshly derived organoids (Fig. 5D and E). This loss-of-function phenotype was also observed in long-term cultured organoids (Supplementary Fig. S6). Remarkably, inhibition of Adh1 activity was associated with increased Wnt signaling in WT organoids (Fig. 5F), as well as with a moderately but significantly (P < 0.05) increased expression of the proliferation-associated genes Ki67 and Top2A (Fig. 5G). Altogether, these results further confirm the functional relevance of our novel set of DMV-associated genes, that are targeted for efficient hypermethylation and silencing, to facilitate inflammation-induced cellular transformation.

Hypermethylation-induced silencing of DMV-associated genes is conserved in human colon cancer samples

To further explore the conservation of our set of hypermethylated DMVs, we analyzed the expression of DMV-associated genes in human colorectal cancer datasets. Of the 30 genes that we found to be associated with hypermethylated DMVs and downregulated in the AOM/DSS model, we identified 28 human homologs. Of these, 18 were associated with a DMV or a DMV-like structure (Supplementary Fig. S7) in a human healthy colon methylome (23). We then analyzed the expression patterns of these 18 genes in a published RNA-seq dataset from TCGA (32) that was available for 259 colon cancer samples and 41 control mucosa samples. This revealed a distinct clustering of control samples that was defined by relatively high expression levels for the majority of DMV-associated genes (data not shown). Furthermore, we used a support vector machine approach to define a minimum subset of genes for sample classification. After starting with all 18 DMV-associated genes, we iteratively removed the gene with the lowest F-score as long as the classification accuracy did not drop below 100% (Supplementary Fig. S8). This identified an optimized set of 3 genes, NR5A2, FOXP2, and ECH1, with remarkable classification accuracy (10-fold cross-validation) of 100%. Interestingly, the expression levels of all three genes were significantly reduced in tumor samples when compared with control tissues (Fig. 6A). Moreover, low expression levels were coordinately observed for all three genes in the majority of tumor samples (Fig. 6B), which is in agreement with the notion that they are regulated by a common epigenetic silencing mechanism.

Figure 6.

DNA methylation-induced silencing of inflammation-associated DMVs in human colon cancer. A, box plots showing mRNA expression levels of the minimum gene set for tumor classification in normal mucosa (N, n = 41) and tumor (T, n = 259) samples from the TCGA colon cancer set. B, combined expression levels of NR5A2, ECH1, and FOXP2 in colon adenocarcinomas. Each dot represents a tumor sample from the TCGA colon cancer set. C, box plots showing methylation levels of three adjacent probes, each from the NR5A2, ECH1, and FOXP2 DMVs in control (n = 38) and tumor (n = 291) samples from the TCGA colon cancer set. cg numbers, probe numbers (see Materials and Methods for details).

Figure 6.

DNA methylation-induced silencing of inflammation-associated DMVs in human colon cancer. A, box plots showing mRNA expression levels of the minimum gene set for tumor classification in normal mucosa (N, n = 41) and tumor (T, n = 259) samples from the TCGA colon cancer set. B, combined expression levels of NR5A2, ECH1, and FOXP2 in colon adenocarcinomas. Each dot represents a tumor sample from the TCGA colon cancer set. C, box plots showing methylation levels of three adjacent probes, each from the NR5A2, ECH1, and FOXP2 DMVs in control (n = 38) and tumor (n = 291) samples from the TCGA colon cancer set. cg numbers, probe numbers (see Materials and Methods for details).

Close modal

To investigate the methylation patterns of these DMVs in human samples, we analyzed array-based methylation datasets from TCGA (32) that were available for 291 colon cancer samples and 38 control mucosa samples. Data analysis showed that several methylation probes from the array were located in the DMVs associated with NR5A2, ECH1, and FOXP2. The NR5A2 DMV probes showed a robust and significant methylation increase in the tumor samples (Fig. 6C). For the ECH1 DMV, significant hypermethylation was also detected in tumors (Fig. 6C), but the effect size appeared smaller, which may indicate that hypermethylation is restricted to a subset of tumor cells. The FOXP2 probes again showed markedly increased methylation levels in tumors (Fig. 6C). Together, our data clearly demonstrate that hypermethylation and silencing of our novel inflammation-associated set of DMVs is a general process conserved in mouse and human colon cancers.

The incidence of IBD is increasing among adults and children in the developed and developing world. Current estimates attribute roughly 10% of the IBD heritability to identified genetic variants (33), while environmental factors interacting with genetic predisposition are discussed as major determinants for disease manifestation (34). Importantly, IBD significantly increases the risk of colorectal cancer (4). Because epigenetic patterns are highly tissue-specific and can be influenced by environmental factors, they might explain key features of IBD and IBD-associated cancers (35, 36). Indeed, two previous studies have identified candidate risk loci for IBD, using low-coverage DNA methylation arrays (37, 38). We have now used whole-genome bisulfite sequencing of an established mouse model for IBD and IBD-associated intestinal cancer to investigate disease-associated DNA methylation changes at single-base resolution. Our results identify the hypermethylation of DMVs as a dynamic and prominent feature, not only of the cancer methylome (12), but also of the inflamed intestinal methylome. Moreover, our results clearly show that DMV hypermethylation is an early event in inflammation-induced intestinal tumorigenesis. Over the past years, many studies have associated the mechanisms of epigenetic deregulation with cancer (39). Cancer-specific DNA methylation patterns are often explained through H3K27me3-associated de novo methylation of CpG islands in the vicinity of genes that may already be repressed in the corresponding normal cell types (26–28). This mechanism has also been implicated in the etiology of mouse intestinal adenomas (40). However, the molecular mechanism identified in our study appears to be different because the hypermethylated DMVs showed a robust enrichment for the activating chromatin marks H3K27ac and H3K4me3, were depleted of CpG islands, and were characterized by intermediate to low CpG densities (Fig. 7). This indicates that during the inflammatory process aberrant DMV methylation occurs in a programmed and coordinated manner, and, as a consequence, silences the expression of nearby genes, which are active in the normal cells. Furthermore, our data clearly demonstrate that inflammation-associated hypermethylation and silencing of a specific set of DMVs is a general process, conserved in other intestinal mouse and human colon malignancies.

Figure 7.

Model illustrating the differences between the epigenetic program induced by intestinal inflammation (right) and the known epigenetic program associated with gene silencing in cancer (left). While the known mechanism of epigenetic gene silencing in cancer is associated with poorly expressed genes that are characterized by high CpG density and Polycomb-dependent repressive chromatin modifications, inflammation-induced epigenetic gene silencing is targeted to highly expressed genes with low CpG density and active chromatin modifications.

Figure 7.

Model illustrating the differences between the epigenetic program induced by intestinal inflammation (right) and the known epigenetic program associated with gene silencing in cancer (left). While the known mechanism of epigenetic gene silencing in cancer is associated with poorly expressed genes that are characterized by high CpG density and Polycomb-dependent repressive chromatin modifications, inflammation-induced epigenetic gene silencing is targeted to highly expressed genes with low CpG density and active chromatin modifications.

Close modal

Recent epigenomic analyses have indicated the existence of two distinct gene silencing mechanisms during cellular differentiation (12, 41): while the majority of lineage specific genes of differentiating stem cells were found to be CpG-rich and silenced by PcG-mediated repression in nonexpressing lineages, gene silencing in later stages of development was mostly characterized by DNA methylation of CpG-poor genes. Our results are consistent with the latter mechanism and suggest that PcG-independent gene silencing is relevant for epigenetic deregulation in inflammation and creates a signature that is maintained through malignant tumor stages. Our data also raise the possibility of additional, inflammation-induced methylation changes, including the establishment of de novo DMVs in AOM/DSS-treated mice (Fig. 1B). These findings will have to be validated and further analyzed in future studies.

Remarkably, inflammation-induced DMV hypermethylation affected a number of genes that have previously been reported to be methylated or/and silenced in human cancers, such as B3GNT7, CDX1, NR5A2, and VDR (42–45). This confirms previous findings of DNA hypermethylation in DMVs that are associated with cancer genes (12), and further establishes that hypermethylated DMVs define a novel epigenetic program that links intestinal inflammation to colon cancer.

It should be noted that our novel set of inflammation-induced DMV-silenced genes is also hypermethylated and silenced in adenoma organoid cultures, which recapitulate the complete intestinal stem cell differentiation hierarchy. This in vitro system allows confirmation of the functional relevance of these genes. Indeed, we found that pharmacologic inhibition of Adh1, an enzyme that is encoded by a DMV-associated gene, significantly enhanced intestinal organoid proliferation, possibly via the activation of Wnt-dependent signaling. Moreover, the functional relevance of this program is clearly illustrated by the 3 DMV-associated genes that correctly classified human colon cancer: NR5A2 encodes a nuclear receptor with protective functions against inflammatory bowel disease in mice (30). ECH1 encodes enoyl coenzyme A hydratase, an essential enzyme for fatty acid metabolism. Fatty acid metabolism has been shown to play an important role in IBD and in colon cancer (46). FOXP2 encodes a transcription factor that is commonly associated with brain function but that has also been shown to be involved in gut development (47). Thus, using an integrated approach, we were able to provide evidence that inflammation-associated epigenetic lesions could have causal effects in cancer development.

The diagnostic potential of aberrant DNA methylation patterns in colorectal cancer is well established (48, 49), and our data suggest that DMV hypermethylation might be a promising biomarker for early cancer detection. Comprehensive methylation profiling of IBD samples and polyps will be required to systematically identify inflammation-related epigenetic biomarkers for the risk stratification of patients. Moreover, human diseases known to be associated with increased cancer incidence, including obesity, are known to involve a systemic low-grade inflammation (2). However, biomarkers that identify patients at risk for cancer development are lacking. We suggest that the methylation status of a minimal gene subset can be used in large retrospective studies to evaluate their potential as risk stratification biomarkers.

No potential conflicts of interest were disclosed.

Conception and design: Y. Bergman, F. Lyko

Development of methodology: Y. Bergman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Abu-Remaileh, I. Ansari, T. Musch, H. Linhart, A. Breiling, Y. Bergman, F. Lyko

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Abu-Remaileh, S. Bender, G. Raddatz, I. Ansari, D. Cohen, J. Gutekunst, A. Breiling, E. Pikarsky, Y. Bergman

Writing, review, and/or revision of the manuscript: I. Ansari, Y. Bergman, F. Lyko

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Ansari, D. Cohen, A. Breiling, Y. Bergman

Study supervision: Y. Bergman, F. Lyko

The authors thank Haya Hamza, Yinon Ben-Neriah, and Guy Ludwig, The Hebrew University Medical School, for providing organoids and adenoma samples, respectively. The authors also thank the DKFZ Genomics and Proteomics Core Facility for sequencing services.

This work was supported in part by the Cooperation Program in Cancer Research of the Deutsches Krebsforschungszentrum (DKFZ) and Israel's Ministry of Science, Technology and Space (MOST; F. Lyko and Y. Bergman). This work was further supported by a research grant from the Israel Cancer Research Foundation (Y. Bergman). M. Abu-Remaileh was supported by the Adams Fellowship Program of the Israel Academy of Sciences and Humanities.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Kuper
H
,
Adami
HO
,
Trichopoulos
D
. 
Infections as a major preventable cause of human cancer
.
J Intern Med
2000
;
248
:
171
83
.
2.
Elinav
E
,
Nowarski
R
,
Thaiss
CA
,
Hu
B
,
Jin
C
,
Flavell
RA
. 
Inflammation-induced cancer: crosstalk between tumours, immune cells and microorganisms
.
Nat Rev Cancer
2013
;
13
:
759
71
.
3.
Hanahan
D
,
Weinberg
RA
. 
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
4.
Soderlund
S
,
Brandt
L
,
Lapidus
A
,
Karlen
P
,
Brostrom
O
,
Lofberg
R
, et al
Decreasing time-trends of colorectal cancer in a large cohort of patients with inflammatory bowel disease
.
Gastroenterology
2009
;
136
:
1561
7
.
5.
Baylin
SB
,
Jones
PA
. 
A decade of exploring the cancer epigenome - biological and translational implications
.
Nat Rev Cancer
2011
;
11
:
726
34
.
6.
Heyn
H
,
Esteller
M
. 
DNA methylation profiling in the clinic: applications and challenges
.
Nat Rev Genet
2012
;
13
:
679
92
.
7.
Timp
W
,
Feinberg
AP
. 
Cancer as a dysregulated epigenome allowing cellular growth advantage at the expense of the host
.
Nat Rev Cancer
2013
;
13
:
497
510
.
8.
Hansen
KD
,
Timp
W
,
Bravo
HC
,
Sabunciyan
S
,
Langmead
B
,
McDonald
OG
, et al
Increased methylation variation in epigenetic domains across cancer types
.
Nat Genet
2011
;
43
:
768
75
.
9.
Berman
BP
,
Weisenberger
DJ
,
Aman
JF
,
Hinoue
T
,
Ramjan
Z
,
Liu
Y
, et al
Regions of focal DNA hypermethylation and long-range hypomethylation in colorectal cancer coincide with nuclear lamina-associated domains
.
Nat Genet
2012
;
44
:
40
6
.
10.
Nejman
D
,
Straussman
R
,
Steinfeld
I
,
Ruvolo
M
,
Roberts
D
,
Yakhini
Z
, et al
Molecular rules governing de novo methylation in cancer
.
Cancer Res
2014
;
74
:
1475
83
.
11.
Hon
GC
,
Hawkins
RD
,
Caballero
OL
,
Lo
C
,
Lister
R
,
Pelizzola
M
, et al
Global DNA hypomethylation coupled to repressive chromatin domain formation and gene silencing in breast cancer
.
Genome Res
2012
;
22
:
246
58
.
12.
Xie
W
,
Schultz
MD
,
Lister
R
,
Hou
Z
,
Rajagopal
N
,
Ray
P
, et al
Epigenomic analysis of multilineage differentiation of human embryonic stem cells
.
Cell
2013
;
153
:
1134
48
.
13.
Jeong
M
,
Sun
D
,
Luo
M
,
Huang
Y
,
Challen
GA
,
Rodriguez
B
, et al
Large conserved domains of low DNA methylation maintained by Dnmt3a
.
Nat Genet
2014
;
46
:
17
23
.
14.
Hahn
MA
,
Hahn
T
,
Lee
DH
,
Esworthy
RS
,
Kim
BW
,
Riggs
AD
, et al
Methylation of polycomb target genes in intestinal cancer is mediated by inflammation
.
Cancer Res
2008
;
68
:
10280
9
.
15.
Katsurano
M
,
Niwa
T
,
Yasui
Y
,
Shigematsu
Y
,
Yamashita
S
,
Takeshima
H
, et al
Early-stage formation of an epigenetic field defect in a mouse colitis model, and non-essential roles of T- and B-cells in DNA methylation induction
.
Oncogene
2012
;
31
:
342
51
.
16.
Lister
R
,
Ecker
JR
. 
Finding the fifth base: genome-wide sequencing of cytosine methylation
.
Genome Res
2009
;
19
:
959
66
.
17.
Rosenberg
DW
,
Giardina
C
,
Tanaka
T
. 
Mouse models for the study of colon carcinogenesis
.
Carcinogenesis
2009
;
30
:
183
96
.
18.
Wirtz
S
,
Neufert
C
,
Weigmann
B
,
Neurath
MF
. 
Chemically induced mouse models of intestinal inflammation
.
Nat Protoc
2007
;
2
:
541
6
.
19.
Sato
T
,
Clevers
H
. 
Primary mouse small intestinal epithelial cell cultures
.
Methods Mol Biol
2013
;
945
:
319
28
.
20.
Sato
T
,
Stange
DE
,
Ferrante
M
,
Vries
RG
,
Van Es
JH
,
Van den Brink
S
, et al
Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett's epithelium
.
Gastroenterology
2011
;
141
:
1762
72
.
21.
Liu
T
,
Ortiz
JA
,
Taing
L
,
Meyer
CA
,
Lee
B
,
Zhang
Y
, et al
Cistrome: an integrative platform for transcriptional regulation studies
.
Genome Biol
2011
;
12
:
R83
.
22.
Gronniger
E
,
Weber
B
,
Heil
O
,
Peters
N
,
Stab
F
,
Wenck
H
, et al
Aging and chronic sun exposure cause distinct epigenetic changes in human skin
.
PLoS Genet
2010
;
6
:
e1000971
.
23.
Ziller
MJ
,
Gu
H
,
Muller
F
,
Donaghey
J
,
Tsai
LT
,
Kohlbacher
O
, et al
Charting a dynamic DNA methylation landscape of the human genome
.
Nature
2013
;
500
:
477
81
.
24.
Chang
CC
,
Lin
CJ
. 
LIBSVM: A library for support vector machines
.
ACM T Intel Syst Tech
2011
;
2
:
27
.
25.
Raddatz
G
,
Gao
Q
,
Bender
S
,
Jaenisch
R
,
Lyko
F
. 
Dnmt3a protects active chromosome domains against cancer-associated hypomethylation
.
PLoS Genet
2012
;
8
:
e1003146
.
26.
Schlesinger
Y
,
Straussman
R
,
Keshet
I
,
Farkash
S
,
Hecht
M
,
Zimmerman
J
, et al
Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer
.
Nat Genet
2007
;
39
:
232
6
.
27.
Ohm
JE
,
McGarvey
KM
,
Yu
X
,
Cheng
L
,
Schuebel
KE
,
Cope
L
, et al
A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing
.
Nat Genet
2007
;
39
:
237
42
.
28.
Widschwendter
M
,
Fiegl
H
,
Egle
D
,
Mueller-Holzner
E
,
Spizzo
G
,
Marth
C
, et al
Epigenetic stem cell signature in cancer
.
Nat Genet
2007
;
39
:
157
8
.
29.
Anders
S
,
Huber
W
. 
Differential expression analysis for sequence count data
.
Genome Biol
2012
;
11
:
R106
.
30.
Coste
A
,
Dubuquoy
L
,
Barnouin
R
,
Annicotte
JS
,
Magnier
B
,
Notti
M
, et al
LRH-1-mediated glucocorticoid synthesis in enterocytes protects against inflammatory bowel disease
.
Proc Natl Acad Sci U S A
2007
;
104
:
13098
103
.
31.
Moser
AR
,
Pitot
HC
,
Dove
WF
. 
A dominant mutation that predisposes to multiple intestinal neoplasia in the mouse
.
Science
1990
;
247
:
322
4
.
32.
Cancer Genome Atlas Network
. 
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
33.
Jostins
L
,
Ripke
S
,
Weersma
RK
,
Duerr
RH
,
McGovern
DP
,
Hui
KY
, et al
Host-microbe interactions have shaped the genetic architecture of inflammatory bowel disease
.
Nature
2012
;
491
:
119
24
.
34.
Rosenstiel
P
,
Sina
C
,
Franke
A
,
Schreiber
S
. 
Towards a molecular risk map–recent advances on the etiology of inflammatory bowel disease
.
Semin Immunol
2009
;
21
:
334
45
.
35.
Chiba
T
,
Marusawa
H
,
Ushijima
T
. 
Inflammation-associated cancer development in digestive organs: mechanisms and roles for genetic and epigenetic modulation
.
Gastroenterology
2012
;
143
:
550
63
.
36.
Ventham
NT
,
Kennedy
NA
,
Nimmo
ER
,
Satsangi
J
. 
Beyond gene discovery in inflammatory bowel disease: the emerging role of epigenetics
.
Gastroenterology
2013
;
145
:
293
308
.
37.
Lin
Z
,
Hegarty
JP
,
Cappel
JA
,
Yu
W
,
Chen
X
,
Faber
P
, et al
Identification of disease-associated DNA methylation in intestinal tissues from patients with inflammatory bowel disease
.
Clin Genet
2011
;
80
:
59
67
.
38.
Hasler
R
,
Feng
Z
,
Backdahl
L
,
Spehlmann
ME
,
Franke
A
,
Teschendorff
A
, et al
A functional methylome map of ulcerative colitis
.
Genome Res
2012
;
22
:
2130
7
.
39.
Bergman
Y
,
Cedar
H
. 
DNA methylation dynamics in health and disease
.
Nat Struct Mol Biol
2013
;
20
:
274
81
.
40.
Grimm
C
,
Chavez
L
,
Vilardell
M
,
Farrall
AL
,
Tierling
S
,
Bohm
JW
, et al
DNA-methylome analysis of mouse intestinal adenoma identifies a tumour-specific signature that is partly conserved in human colon cancer
.
PLoS Genet
2013
;
9
:
e1003250
.
41.
Gifford
CA
,
Ziller
MJ
,
Gu
H
,
Trapnell
C
,
Donaghey
J
,
Tsankov
A
, et al
Transcriptional and epigenetic dynamics during specification of human embryonic stem cells
.
Cell
2013
;
153
:
1149
63
.
42.
Lu
CH
,
Wu
WY
,
Lai
YJ
,
Yang
CM
,
Yu
LC
. 
Suppression of B3GNT7 gene expression in colon adenocarcinoma and its potential effect in the metastasis of colon cancer cells
.
Glycobiology
2014
;
24
:
359
67
.
43.
Pilozzi
E
,
Onelli
MR
,
Ziparo
V
,
Mercantini
P
,
Ruco
L
. 
CDX1 expression is reduced in colorectal carcinoma and is associated with promoter hypermethylation
.
J Pathol
2004
;
204
:
289
95
.
44.
Naumov
VA
,
Generozov
EV
,
Zaharjevskaya
NB
,
Matushkina
DS
,
Larin
AK
,
Chernyshov
SV
, et al
Genome-scale analysis of DNA methylation in colorectal cancer using infinium humanmethylation450 beadchips
.
Epigenetics
2013
;
8
:
921
34
.
45.
Shabahang
M
,
Buras
RR
,
Davoodi
F
,
Schumaker
LM
,
Nauta
RJ
,
Evans
SR
. 
1,25-Dihydroxyvitamin D3 receptor as a marker of human colon carcinoma cell line differentiation and growth inhibition
.
Cancer Res
1993
;
53
:
3712
8
.
46.
Heimerl
S
,
Moehle
C
,
Zahn
A
,
Boettcher
A
,
Stremmel
W
,
Langmann
T
, et al
Alterations in intestinal fatty acid metabolism in inflammatory bowel disease
.
Biochim Biophys Acta
2006
;
1762
:
341
50
.
47.
Shu
W
,
Lu
MM
,
Zhang
Y
,
Tucker
PW
,
Zhou
D
,
Morrisey
EE
. 
Foxp2 and Foxp1 cooperatively regulate lung and esophagus development
.
Development
2007
;
134
:
1991
2000
.
48.
Draht
MX
,
Riedl
RR
,
Niessen
H
,
Carvalho
B
,
Meijer
GA
,
Herman
JG
, et al
Promoter CpG island methylation markers in colorectal cancer: the road ahead
.
Epigenomics
2012
;
4
:
179
94
.
49.
Lange
CP
,
Laird
PW
. 
Clinical applications of DNA methylation biomarkers in colorectal cancer
.
Epigenomics
2013
;
5
:
105
8
.