DNA methyltransferase inhibition reduces inflammation-induced colon tumorigenesis

ABSTRACT

Chronic inflammation is strongly associated with an increased risk of developing colorectal cancer. DNA hypermethylation of CpG islands alters the expression of genes in cancer cells and plays an important role in carcinogenesis. Chronic inflammation is also associated with DNA methylation alterations and in a mouse model of inflammation-induced colon tumorigenesis, we previously demonstrated that inflammation-induced tumours have 203 unique regions with DNA hypermethylation compared to uninflamed epithelium. To determine if altering inflammation-induced DNA hypermethylation reduces tumorigenesis, we used the same mouse model and treated mice with the DNA methyltransferase (DNMT) inhibitor decitabine (DAC) throughout the tumorigenesis time frame. DAC treatment caused a significant reduction in colon tumorigenesis. The tumours that did form after DAC treatment had reduced inflammation-specific DNA hypermethylation and alteration of expression of associated candidate genes. When compared, inflammation-induced tumours from control (PBS-treated) mice were enriched for cell proliferation associated gene expression pathways whereas inflammation-induced tumours from DAC-treated mice were enriched for interferon gene signatures. To further understand the altered tumorigenesis, we derived tumoroids from the different tumour types. Interestingly, tumoroids derived from inflammation-induced tumours from control mice maintained many of the inflammation-induced DNA hypermethylation alterations and had higher levels of DNA hypermethylation at these regions than tumoroids from DAC-treated mice. Importantly, tumoroids derived from inflammation-induced tumours from the DAC-treated mice proliferated more slowly than those derived from the inflammation-induced tumours from control mice. These studies suggest that inhibition of inflammation-induced DNA hypermethylation may be an effective strategy to reduce inflammation-induced tumorigenesis.

Introduction

Chronic inflammation, including that induced by inflammatory bowel disease (IBD), is strongly associated with a greater risk of developing colorectal cancer (CRC), the third most common cancer in the United States in both men and women [1,2]. While overall cases of CRC are decreasing, cases in adults under the age of 50 increased by 22% from 2000 to 2013 likely due to inflammation-inducing lifestyle factors such as obesity, unhealthy diets, and lack of physical activity [3,4]. The increase in CRC incidence in younger adults suggests that while we understand that inflammation causes CRC, better interventions are still needed. Cancer cells have aberrant gains in DNA methylation in promoter CpG islands of genes, which results in their transcriptional repression. Some of these genes are tumour suppressor genes (TSGs), and their silencing plays a direct role in the tumorigenesis process [5]. At sites of chronic inflammation, epithelial cells undergo DNA methylation alterations that are similar to those found in cancer [6,7]. Reversing or preventing inflammation-induced DNA hypermethylation represents a potential strategy for reducing inflammation-induced tumorigenesis.

DNA methyltransferase inhibitors (DNMTis), including azacytidine (AZA) and decitabine (5-aza-2ʹ-deoxycytidine, DAC), are approved for treatment of myelodysplastic syndromes (MDS) and acute myeloid leukaemia (AML) and are in clinical trials for treatment of solid tumours [8]. DNMTis are cytidine analogues that incorporate into nucleic acids trapping DNMTs which results in their degradation. Loss of DNMT1 causes passive DNA demethylation during cell division. Low doses of DNMTis that avoid cytotoxicity are effective in reversing aberrant cancer-specific DNA hypermethylation leading to reactivation of silenced TSGs [9]. Recent studies have demonstrated that treatment with DNMTis also results in re-expression of endogenous retroviruses (ERVs) that are normally epigenetically silenced [10–13]. Expression of ERVs results in cytoplasmic double-stranded RNA (dsRNA), activation of an antiviral pathway, and induction of an interferon (INF) response. Treatment of cancer cells ex vivo with DNMTis activates this pathway and inhibits development of tumours following implantation into mice. Similarly, treatment of mice injected with tumour cells with DNMTis increases the immunogenicity of tumours and results in tumour reduction.

To study how inflammation causes DNA hypermethylation during tumorigenesis we use a mouse model of inflammation-driven colon tumorigenesis. In this model, multiple intestinal neoplasia (Min) mice, which are heterozygous for loss of the tumour suppressor gene Adenomatous polyposis coli (Apc^Δ716), are infected with the human commensal enterotoxigenic Bacteroides fragilis (ETBF) [14]. ETBF infection causes an innate inflammatory response in the distal colon, which peaks at two days post-infection, and then transitions to an IL-17 adaptive immune response [15]. Normally, Min mice predominantly develop tumours in their small intestine [16]. However, ETBF infection of Min mice specifically drives tumorigenesis in the distal colon with an average of 13 colon tumours per mouse. Mock-infected mice only occasionally develop colon tumours (Mock tumours – 0.3 tumours per mouse). Tumorigenesis after ETBF infection is driven by the IL-17 adaptive immune response and blocking this response significantly reduces tumorigenesis [15]. Previously, we have demonstrated that the acute epigenetic response to oxidative damage at the two-day timepoint initiates DNA hypermethylation of TSGs in inflammation-induced tumours [17]. Interestingly, DNA hypermethylation was specific for inflammation-driven tumours as mock colon tumours in uninfected Min mice had few regions with DNA hypermethylation relative to uninflamed colon epithelium.

DNMTis have been extensively evaluated for their potential to kill cancer cells in vitro and in vivo. They have also been used in transgenic mice to reduce background tumour formation. For example, treatment of Min mice with AZA was shown to reduce the number of small intestinal tumours per mouse [18,19]. However, limited studies have examined the efficacy of DNMTis as chemopreventive agents during inflammation-associated tumorigenesis. This setting is particularly relevant in regard to the ability of DNMTis to induce IFN and immune responses, which are also integral to the inflammatory response [20]. We report here that DAC treatment reduces inflammation-induced colon tumorigenesis and, in tumours that do form, reduces inflammation-induced DNA hypermethylation and increases IFN pathway gene expression. Tumoroids derived from tumours from DAC-treated mice maintain lower proliferation rates ex vivo than those derived from tumours from PBS-treated mice suggesting a prolonged response to DAC treatment that is independent of the immune response. This work provides experimental support for the development of chemoprevention strategies to reduce inflammation-induced tumorigenesis by decreasing inflammation-induced DNA hypermethylation.

Results

DAC treatment reduces inflammation-driven tumorigenesis

We have previously demonstrated that 2 days post-infection there are higher levels of oxidative damage in the distal colon of ETBF relative to mock-infected mice [17]. Oxidative damage results in the recruitment of epigenetic proteins, including DNMT1, to chromatin in a mismatch repair protein (MSH2 and MSH6) dependent manner [21,22]. Loss of this recruitment in mice lacking MSH2 expression in intestinal epithelial cells results in a lack of DNA hypermethylation in tumours that form at sites of inflammation suggesting that the acute epigenetic response to oxidative damage drives DNA hypermethylation [17]. Here, to determine if treatment with DNMTis alters inflammation-induced tumorigenesis, mice were treated with low dose DAC (0.2 mg/kg) or PBS for 5 days on and 2 days off for 8 weeks beginning 2 days after ETBF inoculation (Figure 1(a)). Because DNMTs play a role in the immune response [23], we began DAC treatment after infection so as not to perturb the initial immune response to ETBF. DAC treatment significantly reduced ETBF-induced tumorigenesis with ETBF/PBS and ETBF/DAC treated mice having a mean of 14 and 4 colon tumours, respectively (Figure 1(b)). In our model, ETBF infection induces microadenoma formation as early as 5 days post-infection. DAC treatment of ETBF-infected mice did not alter microadenoma formation suggesting that while DAC treatment did reduce overall tumour numbers, DAC did not alter tumour initiation (Figure 1(c)).

Figure 1.

DAC treatment reduces inflammation-induced colon tumour formation.

(a) Schematic of treatment regimen. Mice were given antibiotic water 5 days prior to infection. Low dose DAC treatment (0.2 mg/kg) (blue bars) was begun 2 days after ETBF infection and was given 5 days on, 2 days off for the duration of the study. (b) Number of colon tumours per mouse 8 weeks post-ETBF infection. Data are from three independent cohorts of mice. Mean ± SEM. N = 16. *P < 0.05. (c) Number of microadenomas per H&E tissue section of rolled colons 14 days after ETBF infection. Mean ± SEM. N = 16. (d) Western blot of protein isolated from indicated tissue type 8 weeks post-infection. (e,f) Cytokine expression in distal colon sections from mice treated as indicated 7 or 14 days post-mock or ETBF inoculation. Mean ± SEM. N ≥ 6. *P < 0.05 compared to mock epithelium. (g) ERV expression in samples as in e. Mean ± SEM. N ≥ 6. *P < 0.05 compared to mock epithelium.

DAC treatment is thought to reduce DNA methylation by depletion of cellular levels of DNMT1 protein [24]. We did not observe reduction in DNMT1 protein levels in the colon epithelium of DAC-treated mice at early (7 or 14 days post-ETBF) or late time points (8 weeks post-ETBF) (Supplemental Figure 1A and data not shown). However, DNMT1 protein levels were reduced in tumours from ETBF/DAC treated mice relative to ETBF/PBS mice suggesting that DAC may specifically reduce DNMT1 protein levels in tumour cells or cells with higher than normal levels of DNMT1 protein (Figure 1(d)). ETBF/DAC tumours also had reduced protein levels of PCNA (proliferating cell nuclear antigen), a marker of proliferation, relative to ETBF/PBS tumours. Because PCNA levels often correlate with DNMT1 levels, lower levels of PCNA, likely associated with a reduction in proliferation, could also be an indirect cause for the reduced DNMT1 levels. LAMB (Lamin-B) serves as a loading control.

To determine if DAC alters the inflammatory response to ETBF infection we examined cytokine expression levels in the distal colons of treated mice. Previously, it has been demonstrated that ETBF infection induces an IL-17 adaptive immune response that peaks around 7 days post-infection [14]. As expected ETBF/PBS infection resulted in increased Tnf, Il1b, Il6, and Il17a expression relative to mock/PBS infection at 7 and 14 days post-infection (Figure 1(e)). DAC treatment alone did not alter expression of the cytokines Il1b, Il6, and Il17a relative to mock/PBS infection, but did increase expression of Tnf at the 14-day time point. The expression of Tnf, Il1b, Il6 or Il17a was not significantly different between ETBF/PBS and ETBF/DAC mice.

Treatment of cancer cells with DNMT is has been shown to induce type I IFN responses (IFN-α/β) [10,12]. Ifnb1 (type I) and Ifng (type II) expression was increased in ETBF/PBS and ETBF/DAC infected samples to similar amounts relative to mock/PBS (Figure 1(f)). DAC treatment alone did not alter Ifng levels. However, DAC treatment alone did significantly increase levels of Ifnb1 in mock relative to mock/PBS treated mice at 7 and 14 days post-infection. At both time points the expression of Ifnb1 in mock/DAC samples was similar to that of the ETBF-infected samples suggesting that DAC had no additional effect on Ifnb1 expression in ETBF-infected mice. Increased Ifnb1 expression in the mock/DAC samples suggests that DAC may be inducing the expression of ERVs at these timepoints [10,12]. At 7 and 14 days, post-infection distal colon tissue from ETBF/PBS and ETBF/DAC treated mice had similar levels of increased expression of the ERVs Xmv43 and Xmv45, which have been shown to be inducible by lipopolysaccharide (LPS) treatment [25] (Figure 1(g)). There was also a significant increase in expression of Xmv43 and Xmv45 in mock/DAC relative to mock/PBS 7-day samples. However, only Xmv45 was significantly increased in mock/DAC samples at the 14-day time point. At 7 days post-infection the ERV MMTV had increased expression in mock/DAC and ETBF/DAC samples relative to mock/PBS. IAP had significantly reduced expression in ETBF/DAC and ETBF/PBS samples relative to mock/PBS at 7 and 14 days post-infection, respectively. We also assayed several additional ERVs that have been demonstrated to be induced in cancer cells in vivo after Aza treatment [10,12] but detected no other significant increases (data not shown).

Since Ifnb1 activates the janus kinase/signal transducers and activators of transcription (JAK/STAT) pathway and STAT3 is activated by ETBF infection, we also examined phosphorylated STAT1 (pSTAT1) and pSTAT3 protein levels in the distal colon epithelium. We detected no phosphorylation of STAT1 or STAT3 in the DAC/PBS samples and only a modest, non-significant increase in both in the ETBF/DAC relative to the ETBF/PBS samples 7 days post-infection (Supplemental Figure 1A). All together these results suggest that DAC treatment does not significantly alter the inflammatory response to ETBF infection.

Inflammation-induced tumours from DAC-treated mice have reduced inflammation-associated DNA hypermethylation compared to tumours from PBS-treated mice

Methyl-binding domain sequencing (MBD-seq) allows for the examination of DNA methylation levels genome-wide in a manner that is specific for methylated CpGs. Previously we have demonstrated that inflammation-induced tumours have 203 unique DNA hypermethylated regions when compared to mock epithelium [17]. To determine DNA methylation alterations in the remaining tumours that are present after DAC treatment we performed MBD-seq on individual samples of mock epithelium, ETBF/PBS tumours, or ETBF/DAC tumours (N = 3, 8, and 8, respectively). Autosomal regions within CpG islands with altered DNA methylation in either tumour type compared to the mock epithelium were determined using a 500-bp window. ETBF/DAC tumours had fewer regions with DNA hypermethylation relative to mock epithelium than ETBF/PBS tumours (151 versus 208) (Figure 2(a)). Alternatively, ETBF/DAC tumours had more regions with DNA hypomethylation relative to mock epithelium than ETBF/PBS tumours (264 versus 168). Because of the known connection between inflammation and DNA hypermethylation, we further focused on the 279 regions with DNA hypermethylation in ETBF/PBS or ETBF/DAC tumours relative to mock epithelium (Supplemental Table S1). We also included MBD-seq data for three mock tumours from our previous study for comparison [17]. PCA analysis using the methylation data for these 279 regions resulted in distinct clustering of the different sample types (Figure 2(b)). Unsupervised clustering of these regions also resulted in distinct clustering of the sample types (columns) with mock tumours being more similar to mock epithelium than the ETBF/PBS or ETBF/DAC tumours (Figure 2(c)). One hundred and ninety-nine regions (rows) were DNA hypermethylated in ETBF/PBS tumours relative to mock epithelium (Figure 2(c) – above horizontal black line) and the majority of these regions had reduced DNA hypermethylation in ETBF/DAC tumours and no DNA hypermethylation in mock tumours. Interestingly, 80 regions were DNA hypermethylated in ETBF/DAC tumours relative to mock epithelium (Figure 2(c) – below horizontal black line) and appeared to have somewhat reduced methylation levels in ETBF/PBS tumours. The genes associated with both sets of regions were enriched for similar gene ontology (GO) biological processes, namely processes associated with positive regulation of transcription and development (Supplemental Table S2).

Figure 2.

Tumours from DAC-treated mice have reduced inflammation-induced methylation changes. (a) Venn diagrams of the number of regions with significantly increased (upper panel) or decreased (lower panel) MBD-seq z-scores in the indicated tumour type relative to mock epithelium. (b) PCA analysis of MBD-seq z-scores of the 279 regions with DNA hypermethylation in one of the tumour groups relative to mock epithelium. Unit variance scaling is applied to rows; SVD with imputation is used to calculate principal components. Prediction ellipses are such that with probability 0.95, a new observation from the same group will fall inside the ellipse. All tissue samples were collected 8 weeks after inoculation (N = 3 mock epithelia; N = 8 ETBF or ETBF/DAC tumours). (c) Unsupervised hierarchical clustering of regions (rows) and samples (columns) from b. Rows are centred; unit variance scaling is applied to rows. Both rows and columns are clustered using Euclidean distance and Ward linkage. The colour of each cell reflects the degree of DNA methylation. (d) qMSP of bisulfite treated DNA from indicated tissue in CpG islands of candidate genes. Mean ± SEM. N ≥ 6. *P < 0.05 compared to mock epithelium. ^#P < 0.05 compared to ETBF/PBS tumours. (e) Change in z-scores in ET tumours relative to mock epithelium of regions with increased methylation in ETBF relative to ETBF/DAC tumours. (f) Global DNA methylation relative to mock epithelium in the indicated sample types. Mean ± SEM. N ≥ 5. *P < 0.05 compared to mock epithelium.

To confirm the reduction in inflammation-induced DNA hypermethylation in ETBF/DAC tumours, we assayed DNA methylation in 14 genic CpG island regions using additional tumour samples and site-specific qMSP assays [17]. Regions that were DNA hypermethylated in ETBF/PBS tumours relative to mock epithelium tended to have an intermediate level of DNA methylation in ETBF/DAC tumours (e.g., CD37, Cldn4, Fam229a, Fut4, Hoxa5, Mex3a, Polg, Rasip, Runx1 – Figure 2(d)). Some regions were not DNA hypermethylated in ETBF/DAC tumours relative to mock epithelium but were DNA hypermethylated in ETBF/PBS tumours (e.g., Ano1, Lgr6, Wnt6). Only two of the regions assayed (Bcl6b, Gpc2) showed significant DNA hypermethylation in both tumour types relative to the mock epithelium but no difference in methylation between the two tumour types (Supplemental Figure 1B). To demonstrate that DAC treatment was not simply reducing DNA methylation at all regions in the genome we also assayed several control regions (Spsb4, Chst2, Kbtbd13) that have relatively high levels of DNA methylation in mock epithelium and based on the MBD-seq data had no significant changes in DNA methylation in tumours. These regions had no significant changes in DNA methylation in either tumour type by qMSP (Figure 2(d)).

It has been demonstrated in cell culture that DNMTi treatment of cancer cells has some specificity for reducing DNA methylation of regions with cancer-specific DNA hypermethylation [26]. Of the regions with decreased methylation in ETBF/DAC tumours relative to ETBF/PBS tumours the majority had higher levels of DNA methylation in ETBF/PBS tumours than mock epithelium (171 vs 67) suggesting that the majority of decreased methylation associated with DAC treatment is occurring in regions with tumour-specific gains (Figure 2(e)). It has been widely reported that there is a global loss of DNA methylation in cancer [5]. Therefore, we examined the effect of DAC on global CpG DNA methylation. Consistent with what is observed in humans, the ETBF/PBS tumours exhibited a global loss of DNA CpG methylation. Interestingly, mock tumours and tumours from ETBF/DAC treated mice had total CpG methylation levels similar to mock epithelium (Figure 2(f)). Altogether the DNA methylation data suggests that in tumours that do form in DAC-treated mice, CpG islands with inflammation-associated DNA hypermethylation have reduced levels of DNA methylation compared to tumours from mice infected with ETBF but not treated with DAC.

Tumours from ETBF/DAC treated mice have increased interferon transcription programs

To further elucidate the effect of DAC treatment on tumorigenesis we performed RNA-seq on 5 single tumours each from mock, ETBF/PBS or ETBF/DAC treated mice and 5 mock epithelium samples. All tumour types had around 10,000 genes with significantly altered expression relative to mock epithelium (Figure 3(a) and Supplemental Table S3). There were more genes with significant expression changes between ETBF/PBS and ETBF/DAC tumours (927 genes – FDR ≤ 0.05) than between ETBF/PBS and mock tumours (453 genes). When comparing the genes with alterations between each tumour type and the mock epithelium there were 7,286 genes that were significantly altered in all tumours types (Figure 3(b)). ETBF/PBS tumours had the most uniquely altered genes relative to mock epithelium (1,603 ETBF/PBS, 783 ETBF/DAC, 675 mock – Figure 3(b)).

Figure 3.

ETBF/PBS and ETBF/DAC tumours have proliferative and inflammatory gene expression profiles, respectively. (a) Number of genes significantly differentially expressed at 5% FDR in the indicated comparison. (b) Venn diagram of the number genes with significantly different expression in the indicated tumour type relative to mock epithelium. (c-f) Normalized enrichment score in the indicated comparison from GSEA using all genes from RNA-seq data. Pathways listed have an FDR ≤ 0.001.

To determine pathways with altered gene expression in the different tumour types we utilized unbiased gene set enrichment analysis (GSEA) using all expressed genes and Hallmark gene sets [27]. ETBF/DAC relative to ETBF/PBS tumours were enriched for gene sets associated with inflammation, including the IFNG and IFNA response and complement (Figure 3(c) and Supplemental Figure 2A). ETBF/PBS relative to ETBF/DAC tumours were enriched for pathways associated with proliferation, including E2F targets, MYC targets, G2M checkpoint, and mitotic spindle (Figure 3(d) and Supplemental Figure 2A). When comparing ETBF/DAC tumours to mock tumours, ETBF/DAC tumours were again enriched for IFNG and IFNA responses (Figure 3(e)). Relative to mock tumours ETBF/PBS tumours were also enriched for proliferation-associated pathways, including MYC and E2F targets and G2M checkpoint (Figure 3(f)). At the false discovery rate (FDR) used as a cutoff (FDR < 0.001), mock tumours were not enriched in any pathways relative to either other tumour type. The GSEA analysis supports previous findings that DNMTi treatment results in upregulation of IFN-stimulated genes [20,28]. It also suggests that the increase in tumorigenesis driven by ETBF is through driving cell proliferation pathways, including those regulated by MYC and that DAC treatment reduces the expression of these pathways as has been shown in non-small cell lung cancer (NSCLC) [13].

As mentioned previously, treatment of cancer cells in vitro with DNMTis shows specificity for altering the expression of genes which have altered expression in the cancer cells relative to normal cells [26]. To determine if the same holds true for expression in our in vivo model, we analyzed if genes that have increased expression in ETBF/DAC tumours relative to ETBF/PBS tumours are expressed in mock epithelium and repressed in ETBF/PBS tumours as these would represent genes with specific changes. Nonspecific effects would be any other gene with increased expression in ETBF/DAC tumours. Using a cutoff of a log2 fold change of 1, there were 600 genes with significantly altered expression between ETBF/DAC and ETBF/PBS tumours. The majority of these (569 genes, 95%) had increased expression in ETBF/DAC relative to ETBF/PBS tumours as predicted (Supplemental Figure 2B). Three hundred and sixty-three of these genes (64%) also had significantly altered expression in ETBF/PBS tumours relative to mock epithelium supporting the idea that DAC is mostly targeting cancer-specific changes.

We further examined expression of genes known to be upregulated in the anti-viral response triggered by DNMTi-induced expression of ERVs [12]. The majority of the genes examined, including Ifitm1, Ifitm3, Ifit1, Ifi44, Irf9, Isg15, Mx2, and Oas2 had significantly higher expression in all three tumour types relative to mock epithelium, and their expression was not significantly different between the tumour types (Supplemental Figure 2C). Ifit2 and Stat1 had significantly higher expression in all tumour types relative to mock epithelium and higher expression in ETBF/DAC tumours than ETBF/PBS tumours. Irf7 had significantly higher expression only in ETBF/DAC tumours relative to mock epithelium. Mx1 and Oasl had increased expression in ETBF/DAC and mock tumours relative to mock epithelium but not in ETBF/PBS tumours. This data suggests that the anti-viral response pathway is upregulated in all tumours in this model but expression of some components is further increased in tumours from mice treated with DAC.

As it has been demonstrated that DNMTis alter the immune cells in the tumour microenvironment we used IHC to examine immune cells in tissue sections from the treated mice [12,13]. There were no differences in CD4+ helper T-cells, CD8+ cytotoxic T-cells, or F4/80+ macrophages between ETBF/PBS or ETBF/DAC tumours (data not shown). There was however a significant decrease in regulatory T-cells marked by FoxP3 in ETBF/DAC relative to ETBF/PBS tumours (Supplemental Figure 2D). Altogether these findings suggest that there is a modest effect of DAC treatment on the expression of genes in the anti-viral response and the presence of tumour-associated lymphocytes.

Genes with altered DNA methylation have altered expression

It has been well established that increases in promoter CpG island DNA hypermethylation are associated with transcriptional repression. It is also becoming more understood that DNA hypermethylation of gene body and 3ʹ exon CpG islands can be associated with increased expression of tumour-promoting genes [29]. To understand how DNA methylation alterations effect gene expression in our model we examined the expression of genes with significantly altered methylation between ETBF/PBS and ETBF/DAC tumours. Of the 332 regions with altered methylation between the tumour types, 281 regions are associated with genes (Supplemental Table S4). Only 13 of these genes had significantly different expression between ETBF/PBS and ETBF/DAC tumours. However, 193 of the 281 regions were in 144 unique genes that had significantly altered expression in ETBF/PBS tumours relative to mock epithelium. A heatmap of the mean DNA methylation data for the 193 regions demonstrates that a large portion of the genes have increased levels of DNA methylation in ETBF/PBS tumours relative to mock epithelium, which is reduced in the ETBF/DAC tumour samples (Figure 4(a)). Importantly, for the 140 genes with significantly altered expression between ETBF/PBS tumours and mock epithelium, DAC tumours tend to have a mean expression level between the other two sample types (Figure 4(b)).

Figure 4.

Expression changes in genes with significant DNA methylation changes in tumours relative to mock epithelium. Heatmaps of the mean (a) z-ratios (Methylation, 193 regions, rows) and (a) RPKM values (Expression, 144 genes, rows) for the genes with significant changes in DNA methylation between ETBF and ETBF/DAC tumours and significant expression changes between ETBF/PBS tumours and mock epithelium. Each column corresponds to the indicated epithelium or tumour sample. The colour of each cell reflects the degree of DNA methylation or expression. (c and d) Gene expression by qPCR of candidate genes with DNA hypermethylation in ETBF tumours relative to mock epithelium. Mean ± SEM. N ≥ 6. *P < 0.05 compared to mock epithelium. ^#P < 0.05 compared to ETBF tumours.

The expression levels of a set of candidate TSGs (Hoxa5, Ano1, Fut4, Polg) that have increased promoter CpG island DNA methylation and decreased expression in ETBF/PBS tumours relative to mock epithelium were verified to be partially restored in ETBF/DAC tumours although the changes in Ano1 did not reach significance (Figure 4(c)). Additionally, genes that have increased DNA methylation in 3ʹ exon or gene body CpG islands and increased expression in ETBF/PBS tumours relative to mock epithelium have significantly reduced expression in ETBF/DAC tumours relative to ETBF/PBS tumours (Gbc2, Lgr6, Mex3a, Wnt6 – Figure 4(d)). All of these genes encode proteins that are involved in Wnt pathway activation and/or stemness and are associated with carcinogenesis suggesting that their hypermethylation and increased expression are tumour promoting in our model [30–33].

Tumoroids derived from inflammation-induced tumours from DAC-treated mice have reduced proliferation

Colon organoids can be derived from crypts isolated from mouse or human colons or from tumour cells and are grown in vitro in 3D culture [34,35]. They lack other cell types and therefore provide an effective approach to study effects of genetic and epigenetic changes in epithelial and tumours cells. We have previously used organoids derived from ETBF and mock tumours (tumoroids) to demonstrate that DNA methylation and silencing of DNA polymerase gamma (Polg) the only mitochondrial DNA polymerase results in altered metabolism in ETBF tumoroids [36]. Here, we derived organoids from mock epithelium (colonoids) or ETBF/PBS or ETBF/DAC tumours. Tumours from Min mice have lost the second copy of Apc allowing tumoroids derived from them to grow in media lacking Wnt growth factors [34]. We confirmed that while mock colonoids had one copy each of WT and mutant Apc, ETBF/PBS and ETBF/DAC tumoroids no longer have the WT copy (Figure 5(d)). Organoids derived from human CRC have been shown to retain the genetic characteristics of the original cancer from which they were derived, however less is known about retention of epigenetic status [37]. DNA hypermethylation of most of the tumorigenesis-related candidate genes assayed was maintained in the ETBF/PBS tumoroids and reduced in the ETBF/DAC tumoroids (Ano1, Fam229a, Fut4, Gata2, Hoxa5, Lgr6, Polg, Runx3, Wnt6 – Figure 5(b)). Additionally, five candidate genes that had differential methylation in the tumours (Figure 2(d)) when assayed in the organoids displayed variable DNA methylation and/or did not have higher levels of DNA methylation in ETBF/PBS versus ETBF/DAC tumoroids (Supplemental Figure 3A). Furthermore, candidate genes with DNA hypermethylation and decreased expression in ETBF/PBS tumours had persistent-reduced expression in the ETBF/PBS tumoroids, which was partially restored in the ETBF/DAC tumoroids (Ano1, Hoxa5, Polg – Figure 5(c)). Genes with increased methylation and increased expression in ETBF/PBS tumours also maintained the relative expression differences in the tumoroids (Bcl6b, Lgr6, Mex3a, Wnt6 – Figure 5(d)). This data suggests that at least some of the tumour-specific DNA hypermethylation and associated transcriptional changes were maintained in the ETBF/PBS tumoroids. Because the ETBF/DAC tumours had IFN gene signatures, we examined expression of genes in the IFN and anti-viral pathways but observed no significant differential expression of Ifnb1, Irf7, Oasl1 or Mx1 between the ETBF/DAC or ETBF/PBS tumoroids or the mock colonoids (data not shown).

Figure 5.

Tumoroids derived from ETBF/DAC treated mice have reduced proliferation. (a) Apc wildtype versus Min allele ratio determined by qPCR using genomic DNA from indicated organoid types. Organoids are derived from the indicated tissue type 8 weeks post-infection. Individual values are plotted and the bars indicate mean ± SEM. N = 6. *P < 0.05. (b) qMSP of bisulfite treated DNA from indicated organoids in CpG islands of candidate genes. Mean ± SEM. N ≥ 6. *P < 0.05 compared to mock colonoids. ^#P < 0.05 compared to ETBF/PBS tumoroids. (c and d) Gene expression by qPCR of candidate genes with DNA hypermethylation in ETBF/PBS and ETBF/DAC tumoroids relative to mock colonoids. Mean ± SEM. N ≥ 6. *, # as in b. (e) White light images of tumoroids derived from the indicated tumour type. (f) MTT assay of tumoroids. Data presented as absorbance relative to ETBF/PBS tumoroids. Mean ± SEM. N = 12. *P < 0.05. (g) Gene expression by qPCR of Wnt/β-catenin target genes as in c. (h) MTT assay of tumoroids treated with 50 μM CHIR99021 or DMSO. Data presented as absorbance relative to DMSO treated sample for the respective tumoroid type. Mean ± SEM. N = 8. *P < 0.05 compared to respective DMSO treated. (i) MTT assay of tumoroids treated daily for 3 days with 100 nM DAC or PBS. Data presented as absorbance relative to PBS-treated sample for the respective tumoroid type. Mean ± SEM. N = 8. *P < 0.05 compared to respective PBS treated.

Our GSEA results demonstrated that ETBF/DAC tumours have reduced expression of genes in cell proliferation pathways relative to ETBF/PBS tumours (Figure 3(d)). When passaging the tumoroids, we noted that the ETBF/DAC tumoroids grew slower than ETBF/PBS tumoroids (Figure 5(e)). To quantify this result, we performed MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assays on the mock, ETBF/PBS, and ETBF/DAC tumoroids and demonstrated that mock and ETBF/DAC tumoroids grew significantly slower than ETBF/PBS tumoroids (Figure 5(f)). Similar to the tumours, ETBF/DAC tumoroids had reduced DNMT1 and PCNA protein levels relative to ETBF/PBS tumoroids (Supplemental Figure 3B). DNMT1 levels in mock tumoroids were mixed, but PCNA levels were similar to ETBF/PBS tumoroids.

Because we observed persistent DNA hypermethylation in ETBF/PBS tumoroids, we considered that epigenetic changes could contribute to the increased proliferation. Several genes with inflammation-induced DNA hypermethylation in ETBF/PBS relative to ETBF/DAC tumours are involved in the Wnt/β-catenin pathway, including Hoxa5, Lgr6, and Wnt6 (Figure 2(d)) [33,38,39]. Wnt signalling has important roles in proliferation, cell differentiation and apoptosis. Furthermore, it has been demonstrated that mutant Apc alone is insufficient to sustain Wnt pathway activity, and epigenetic inactivation of Wnt inhibitors in Apc mutant CRC cell lines further activates Wnt/β-catenin signalling [40,41]. We determined that ETBF/DAC relative to ETBF/PBS tumoroids had significantly reduced expression of several target genes of Wnt/β-catenin signalling, including Lef1, Axin2, Myc, and Lgr5 (Figure 5(g)) [42]. Additional Wnt target genes, Ccnd1 and Cd44, were significantly elevated in both ETBF/PBS and ETBF/DAC tumoroids relative to mock colonoids but were not differently expressed between the two tumoroid types (Figure 5(g)). We hypothesized that the decrease of inflammation-induced DNA hypermethylation in ETBF/DAC tumoroids may result in decreased Wnt/β-catenin signalling. CHIR99021 is a potent inhibitor of glycogen synthase kinase 3 (GSK3) and functions as a Wnt activator even in Apc mutant organoids [43]. Treatment of ETBF/DAC but not ETBF/PBS tumoroids with CHIR99021 led to a significant increase in growth suggesting that differences in baseline Wnt pathway activity may be contributing to the growth differences between the types of tumoroids (Figure 5(h)). Additionally, in vitro, low dose DAC treatment (100 nM) of the tumoroids for three consecutive days resulted in significantly less growth of ETBF/PBS but not ETBF/DAC tumoroids (Figure 5(i)). However, since DAC’s effect on DNA methylation is dependent on cell division it is possible that the difference in proliferation was the main driver of this effect.

Discussion

Chronic inflammation can drive DNA hypermethylation and carcinogenesis [6]. DNMTis have been used in vitro to prevent growth of cancer cells and in vivo to reduce tumour burden [8]. However, limited preclinical studies have used DNMTis as chemopreventive agents. Herein we demonstrated that in vivo DAC treatment abrogates inflammation-induced DNA hypermethylation, induces an IFN response, and significantly reduces inflammation-driven colon tumorigenesis.

The inflammation-driven tumorigenesis in our model involves the immune system causing tumour-promoting inflammation in response to a bacterial infection and inflammation driving DNA hypermethylation changes in epithelial cells. DNMTis are thought to decrease tumorigenesis by promoting tumour immunogenicity and by reducing aberrant DNA hypermethylation of TSGs [28,44–46]. Therefore, DAC treatment could alter tumorigenesis in our model by altering the immune response and/or by reducing/preventing inflammation-induced DNA hypermethylation. In cancer cells, DNMTis induce the expression of ERVs leading to dsRNA that activates an anti-viral response and IFN signalling. Less is known about the effect of DNMTis on ERVs and IFN signalling in normal cells and normal cells undergoing inflammation. In our model, we detected increased expression of ERVs in the distal colon after ETBF infection and/or DAC treatment. Uninfected mice treated with DAC had increased Ifnb1 expression levels relative to mock/PBS treated mice, but these expression levels were similar to those in colon tissue from ETBF-infected mice with and without DAC treatment. IFNB activates JAK/STAT1 signalling. However, we did not detect phosphorylation of STAT1 or STAT3 in colon tissue from uninfected DAC-treated mice. DAC treatment also did not alter the Il6 or Il17a induction caused by ETBF infection. These findings importantly suggest that DAC treatment did not significantly alter the normal inflammatory response to infection. However, when examining tumours from DAC-treated mice there was an induction of IFNG and IFNA gene expression pathways suggesting that the effect of DAC on the IFN response is somewhat selective for tumour versus normal cells.

While tumorigenesis was significantly reduced in ETBF/DAC treated mice, the tumours that did form had altered DNA methylation compared to those from ETBF/PBS mice. This finding suggests that DAC treatment reduced or prevented aberrant inflammation-induced DNA hypermethylation. These tumours also had some regions with increased DNA methylation relative to mock epithelium and mock and ETBF/PBS tumours. The hypermethylation in tumours from DAC-treated mice is likely not direct and could be caused by effects of DAC on the tumour microenvironment, including surrounding fibroblasts and immune cells. The cause and implications of these DNA methylation changes need further study. Since we could only assay the transformed cells that became tumours in DAC-treated mice, it is tempting to speculate that in transformed cells that did not become tumours, DAC treatment prevented inflammation-induced DNA hypermethylation at either more regions or to a greater extent at the identified regions. We are continuing to investigate the mechanism and timing between the initial chromatin-mediated transcriptional repression that occurs two days post-infection when oxidative damage levels are high and the more stable silencing by DNA hypermethylation that occurs later [17].

Interestingly, the DAC associated reduction in inflammation-induced DNA hypermethylation persisted when the different tumour types were grown in 3D culture in vitro as tumoroids. There was no observable persistent activation of the IFN response in the ETBF/DAC tumoroids. Tumoroids only consist of epithelial cells so once they were derived there was no additional influence on them from the immune system. Because the DNA hypermethylation and associated gene expression changes persisted in the tumoroids it suggests that the reduced proliferation in the ETBF/DAC tumoroids is at least in part due to the effects of DAC on inflammation-induced DNA hypermethylation. We demonstrated that increased activation of the Wnt pathway in ETBF/DAC tumoroids by treatment with CHIR99021 increased their growth but did not alter the growth of ETBF/PBS tumoroids. Epigenetic inactivation of Wnt inhibitors in Apc mutant CRC cell lines has been shown to further activate Wnt/β-catenin signalling [40,41]. Therefore, we hypothesize that inflammation-induced DNA hypermethylation is driving full activation of the Wnt/β-catenin pathway in ETBF/PBS tumoroids, which prevents Wnt activity from being further increased by CHIR99021. Elevated Wnt signalling can drive the cell cycle and E2F and MYC related transcriptional programs that are elevated in ETBF/PBS relative to ETBF/DAC tumours based on our RNA-seq data. The decreased MYC pathway gene signature in ETBF/DAC tumours is consistent with previous work that demonstrated that in vitro treatment of NSCLC cells with AZA resulted in an MYC depletion signature that sensitized cells to histone deacetylase inhibitor-induced reduction in proliferation [13]. Additional work needs to be performed to determine which genes or sets of genes with inflammation-induced DNA hypermethylation and expression changes drive the growth and signalling pathway differences between ETBF/PBS and ETBF/DAC tumoroids.

In vivo treatment with DAC also results in the DAC exposure of immune cells, which are similarly responsive to DNMTis. De novo methylation has been shown to promote CD8+ T-cell exhaustion, and demethylation is important for their reactivation [23]. Furthermore, DNMTis have been shown to alter the immune cell population in the tumour microenvironment [13]. In support of this previous work, we observe a decrease in immune suppressive FoxP3+ cells in ETBF/DAC tumours compared to ETBF/PBS tumours suggesting that DAC treatment allows immune cells to better target the tumours.

Inflammation is induced by many lifestyle factors and plays a key role in the development of CRC. Preventing or reversing inflammation-induced DNA hypermethylation may be a way to reduce inflammation-associated CRC. Here we use the DNMTi DAC to provide proof-of-concept for repurposing available epigenetic inhibitors and demonstrate that DAC treatment significantly reduces inflammation-induced colon tumorigenesis in vivo. However, DNMTis have side effects that limit their use for chemoprevention [8]. Additionally, reactivation of ERVs may increase genomic instability and therefore be detrimental. Further studies are needed to find safe strategies to modulate inflammation-induced DNA hypermethylation to reduce CRC occurrence.

Materials and methods

Animal model

C57BL/6J (Jackson labs) and Min^ApcΔ716± mice were handled and inoculated with ETBF as in Wu et al. (2009) [14]. Low dose decitabine (DAC; 0.2 mg/kg; Sigma; A3656) in PBS or PBS alone was given by intraperitoneal injection, started 2 days post-infection and performed for cycles of 5 days consecutive days on followed by 2 days off for the duration of the study. For all experiments, both males and females were used and mice were randomized between treatment groups. Individual tumours were removed from dissected colons with the aid of a dissecting microscope and stored at −80°C until further analysis. Distal (0–2 cm measured from the rectum) epithelium was collected by scraping the mucosal surface of the dissected colon (after removal of any tumours), washed three times in PBS, and then subjected to the indicated protocol. Such scraping has been shown by others to be an effective method to obtain samples of intestinal epithelial cells [47]. Whole distal pieces of colon (0–1 cm measured from the rectum) were used for cytokine expression analysis. All mouse experiments were covered under protocol number 16–027, which was approved by the Indiana University Bloomington Animal Care and Use Committee in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care International.

Gene expression

RNA was prepared from colon, epithelium, tumours or organoids using Trizol followed by cleanup with an RNeasy kit (Qiagen). cDNA were prepared and qPCR was done using TaqMan assays (see Supplemental Table S5 for assays used). Expression of candidate genes was normalized to expression of a housekeeping gene (PPIA or 18S).

Protein extraction and western blotting

Protein was extracted from tissue using 4% SDS followed by lysis in a qiashredder (Qiagen, 79654). Equal amounts of protein were loaded on 4–12% Bis-Tris protein gels (Invitrogen), transferred to PVDF (Millipore), and blotted with the indicated antibodies (anti-DNMT1 – Sigma, D4692; anti-PCNA – CST, 2586; anti-pSTAT3 – CST, 9145; anti-pSTAT1- CST, 9167; LAMB – SCBT, sc-6216).

Methyl CpG binding domain (MBD)-seq

DNA was isolated from tumours or epithelium using the QIAamp DNA mini kit (Qiagen, 51306) following the manufacturer’s protocol. To identify differentially methylated regions (DMRs), MBD enrichment was performed from DNA from epithelium collected from individual tumours (n = 3 ETBF/PBS, n = 8 ETBF/DAC) using Diagenode’s MethylCap kit. Libraries were prepared following Bioo Scientific’s Methyl Sequencing kit. Single-end 75 bp sequencing was performed using an Illumina Nextseq. Our previous MBD-seq data were included in the analysis (n = 3 mock epithelia, n = 3 mock tumours, n = 5 ETBF tumours) [17]. Z-scores were calculated using a 500 bp fixed sized bin spanning CpG islands based on the distribution of coverage from uniquely mapped reads. Z-ratios were derived from the comparison of z-scores for the different sample types for the 500 bp regions. See Maiuri et al. [17] for more detail.

Quantitative methylation-specific PCR (qMSP)

DNA was bisulfite treated (EZ DNA methylation-Gold kit, Zymo Research, D5005) and used for qMSP. qMSP assays were first tested using a standard curve of bisulphite-treated mixtures of unmethylated and methylated DNA (data not shown). Only unmethylated and methylated DNA assays with little to no amplification of methylated or unmethylated samples, respectively, close to 100% efficiency and R² close to 1 were used further. To calculate relative DNA methylation, qMSP assays were performed for each gene using primer sets specific for unmethylated (U) and methylated (M) alleles. The delta Cq method was used to calculate M/U. The mean M/U for the control samples (mock epithelium or mock colonoids) was then calculated. All data points were then divided by that mean to calculate M/U relative to the mock sample. Data are presented as the mean of the relative M/U + SEM for all the samples in that group. See Supplemental Table S5 for primers used.

Total CpG methylation

Total CpG methylation in DNA from epithelium and tumour samples was determined using an ELISA-based assay (MethylFlash Global DNA Methylation (5-mC) – EpiGentek, P-1030).

Organoids

Organoids were derived from mock distal colon epithelium (colonoids) or tumours from mock, ETBF/PBS or ETBF/DAC treated Min mice as demonstrated previously and grown in basal growth media containing EGF (epidermal growth factor), Noggin, R-Spondin1, and Wnt3a (colonoids) or EGF and Noggin (tumoroids) [43,48]. 2.5 μM CHIR99021 (Sigma Aldrich) or DMSO was added to growth media when tumoroids were split and added again 2 days post-plating when the media was changed. For low-dose DAC treatment, 100 nM DAC (Sigma Aldrich) or PBS was added to tumoroids in fresh growth media for three consecutive days beginning the day after plating.

Tumoroid viability

Media was removed from organoid cultures and 250 μl of KRPH (Krebs Ringer Phosphate) buffer was added plus MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide) to a final concentration of 500 μg/ml and incubated at 37°C for 1 h. Matrigel and MTT were solubilized as previously described [49].

Apc allele PCR

Apc allele-specific PCR was performed by using Min genotyping primers in qPCR assays (Apc-B and Apc-D for Apc^Δ716; Apc-C and Apc-D for WT Apc) and then calculating the delta Cq of the WT to the Min allele.

Heatmaps, PCA, venn diagrams and GSEA

Heatmaps and Principal Component Analysis (PCA) blots were generated using ClustVis [50]. Venn diagrams were generated using the Venn diagram generator (http://barc.wi.mit.edu/tools/venn/). GSEA was performed using gene set enrichment analysis, GSEA software, and Molecular Signature Database (MSigDB) ([27], http://www.broad.mit.edu/gsea/).

Statistical analysis

Tumour counts, expression data, qMSP, and MTT absorbance are presented as the mean ± standard error (SEM). These data are evaluated by one-tail t-test or Mann-Whitney U test when variance is different between groups and considered statistically significant with a p-value <0.05. Sample sizes are indicated in associated figure legends. For all assays performed with tissue or organoids, experiments were performed 3 times with 2–3 biological replicates per experiment. MBD-seq statistical analysis is detailed in Maiuri et al. [17].

Funding Statement

This work was supported by the National Institute of Environmental Health Sciences under Grant [R01ES023183]to H.M. O’Hagan.

Acknowledgments

We thank the Indiana University Center for Genomics and Bioinformatics for their assistance. We also thank Sue Childress for her assistance with tissue processing.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.