Mouse methylation profiles for leukocyte cell types, and estimation of leukocyte fractions in inflamed gastrointestinal DNA samples.

Precise analysis of tissue DNA and RNA samples is often hampered by contaminating non-target cells whose amounts are highly variable. DNA methylation profiles are specific to cell types, and can be utilized for assessment of the fraction of such contaminating non-target cells. Here, we aimed 1) to identify methylation profiles specific to multiple types of mouse leukocytes, and 2) to estimate the fraction of leukocytes infiltrating inflamed tissues using DNA samples. First, genome-wide DNA methylation analysis was conducted for three myeloid-lineage cells and four lymphoid-lineage cells isolated by fluorescence-activated cell sorting after magnetic-activated cell sorting from leukocytes in the spleen. Clustering analysis using CpG sites within enhancers separated the three myeloid-lineage cells and four lymphoid-lineage cells while that using promoter CpG islands (TSS200CGIs) did not. Among the 266,108 CpG sites analyzed, one CpG site was specifically hypermethylated (ß value ≥ 0.7) in B cells, and four, seven, 183, and 34 CpG sites were specifically hypomethylated (ß value < 0.2) in CD4+ T cells, CD8+ T cells, B cells, and NK cells, respectively. Importantly, cell type-specific hypomethylated CpG sites were located at genes involved in cell type-specific biological functions. Then, marker CpG sites to estimate the leukocyte fraction in a tissue with leukocyte infiltration were selected, and an estimation algorithm was established. The fractions of infiltrating leukocytes were estimated to be 1.6-12.4% in the stomach (n = 10) with Helicobacter pylori-induced inflammation and 1.5-4.3% in the colon with dextran sulfate sodium-induced colitis (n = 4), and the fractions were highly correlated with those estimated histologically using Cd45-stained tissue sections [R = 0.811 (p = 0.004)]. These results showed that mouse methylation profiles at CpG sites within enhancers reflected leukocyte cell lineages, and the use of marker CpG sites successfully estimated the leukocyte fraction in inflamed gastric and colon tissues.

A quantification method of somatic mutations in normal tissues and their accumulation in pediatric patients with chemotherapy.

Somatic mutations are accumulated in normal human tissues with aging and exposure to carcinogens. If we can accurately count any passenger mutations in any single DNA molecule, since their quantity is much larger than driver mutations, we can sensitively detect mutation accumulation in polyclonal normal tissues. Duplex sequencing, which tags both DNA strands in one DNA molecule, enables accurate count of such mutations, but requires a very large number of sequencing reads for each single sample of human-genome size. Here, we reduced the genome size to 1/90 using the BamHI restriction enzyme and established a cost-effective pipeline. The enzymatically cleaved and optimal sequencing (EcoSeq) method was able to count somatic mutations in a single DNA molecule with a sensitivity of as low as 3 × 10-8 per base pair (bp), as assessed by measuring artificially prepared mutations. Taking advantages of EcoSeq, we analyzed normal peripheral blood cells of pediatric sarcoma patients who received chemotherapy (n = 10) and those who did not (n = 10). The former had a mutation frequency of 31.2 ± 13.4 × 10-8 per base pair while the latter had 9.0 ± 4.5 × 10-8 per base pair (P < 0.001). The increase in mutation frequency was confirmed by analysis of the same patients before and after chemotherapy, and increased mutation frequencies persisted 46 to 64 mo after chemotherapy, indicating that the mutation accumulation constitutes a risk of secondary leukemia. EcoSeq has the potential to reveal accumulation of somatic mutations and exposure to environmental factors in any DNA samples and will contribute to cancer risk estimation.

Combination of a synthetic retinoid and a DNA demethylating agent induced differentiation of neuroblastoma through retinoic acid signal reprogramming.

BACKGROUND: The CpG island methylator phenotype of neuroblastoma (NBL) is strongly associated with poor prognosis and can be targeted by 5-aza-2'-deoxycytidine (5-aza-dC). Differentiation therapy is a standard maintenance therapy for high-risk NBLs. However, the in vivo effect of tamibarotene, a synthetic retinoic acid, and the efficacy of its combination with 5-aza-dC have not been studied. Here, we conducted a preclinical study to assess the in vivo tamibarotene effect and the combination. METHODS: Treatment effects were analysed by in vitro cell growth and differentiation state and by in vivo xenograft suppression. Demethylated genes were analysed by DNA methylation microarrays and geneset enrichment. RESULTS: Tamibarotene monotherapy induced neural extension and upregulation of differentiation markers of NBL cells in vitro, and tumour regression without severe side effects in vivo. 5-Aza-dC monotherapy suppressed tumour growth both in vitro and in vivo, and induced demethylation of genes related to nervous system development and function. Pre-treatment with 5-aza-dC in vitro enhanced upregulation of differentiation markers and genes involved in retinoic acid signaling. Pre-treatment with 5-aza-dC in vivo significantly suppressed tumour growth and reduced the variation in tumour sizes. CONCLUSIONS: Epigenetic drug-based differentiation therapy using 5-aza-dC and TBT is a promising strategy for refractory NBLs.

Influence of degree of DNA degradation in formalin-fixed and paraffin-embedded tissue samples on accuracy of genome-wide DNA methylation analysis.

Aim: Depending upon the degree of DNA degradation of formalin-fixed and paraffin-embedded tissue samples, accuracy of measurement by Infinium MethylationEPIC BeadChip assay (Illumina, CA, USA) was assessed. Materials & methods: DNA quality of six formalin-fixed and paraffin-embedded lung tissue samples with different formalin fixation periods was assessed by Illumina quality control, DNA copy number and DNA integrity number value. Infinium data from restored bisulfite treated DNA were compared with datum from a fresh-frozen sample. Results: The correlation coefficient decreased from 0.993 to 0.970 depending upon DNA degradation, even if the Illumina quality control was met. Exclusion of specific probes improved the correlation regardless of tissue. Conclusion: Poor DNA quality can be assessed as an amplifiable DNA copy number and DNA integrity number value. Probe filtering has the potential to improve assay accuracy.

Low-dose DNA demethylating therapy induces reprogramming of diverse cancer-related pathways at the single-cell level.

BACKGROUND: Epigenetic reprogramming using DNA demethylating drugs is a promising approach for cancer therapy, but its efficacy is highly dependent on the dosing regimen. Low-dose treatment for a prolonged period shows a remarkable therapeutic efficacy, despite its small demethylating effect. Here, we aimed to explore the mechanisms of how such low-dose treatment shows this remarkable efficacy by focusing on epigenetic reprograming at the single-cell level. METHODS: Expression profiles in HCT116 cells treated with decitabine (DAC) were analyzed by single-cell RNA-sequencing (scRNA-seq). Functional consequences and DNA demethylation at the single-cell level were analyzed using cloned HCT116 cells after DAC treatment. RESULTS: scRNA-seq revealed that DAC-treated cells had highly diverse expression profiles at the single-cell level, and tumor-suppressor genes, endogenous retroviruses, and interferon-stimulated genes were upregulated in random fractions of cells. DNA methylation analysis of cloned HCT116 cells revealed that, while only partial reduction of DNA methylation levels was observed in bulk cells, complete demethylation of specific cancer-related genes, such as cell cycle regulation, WNT pathway, p53 pathway, and TGF-ß pathway, was observed, depending upon clones. Functionally, a clone with complete demethylation of CDKN2A (p16) had a larger fraction of cells with tetraploid than parental cells, indicating induction of cellular senescence due to normalization of cell cycle regulation. CONCLUSIONS: Epigenetic reprogramming of specific cancer-related pathways at the single-cell level is likely to underlie the remarkable efficacy of low-dose DNA demethylating therapy.

TET repression and increased DNMT activity synergistically induce aberrant DNA methylation.

Chronic inflammation is deeply involved in various human disorders, such as cancer, neurodegenerative disorders, and metabolic disorders. Induction of epigenetic alterations, especially aberrant DNA methylation, is one of the major mechanisms, but how it is induced is still unclear. Here, we found that expression of TET genes, methylation erasers, was downregulated in inflamed mouse and human tissues, and that this was caused by upregulation of TET-targeting miRNAs such as MIR20A, MIR26B, and MIR29C, likely due to activation of NF-κB signaling downstream of IL-1ß and TNF-α. However, TET knockdown induced only mild aberrant methylation. Nitric oxide (NO), produced by NOS2, enhanced enzymatic activity of DNA methyltransferases (DNMTs), methylation writers, and NO exposure induced minimal aberrant methylation. In contrast, a combination of TET knockdown and NO exposure synergistically induced aberrant methylation, involving genomic regions not methylated by either alone. The results showed that a vicious combination of TET repression, due to NF-κB activation, and DNMT activation, due to NO production, is responsible for aberrant methylation induction in human tissues.

Establishment of a high-throughput detection system for DNA demethylating agents.

Epigenetic alterations underlie various human disorders, including cancer, and this has resulted in the development of drugs targeting epigenetic alterations. Although DNA demethylating agents are one of the major epigenetic drugs, only two compounds-5-azacytidine (5-aza-CR, azacitidine) and 5-aza-2'-deoxycytidine (5-aza-dC, decitabine)-have obtained clinical approval. Here, we aimed to establish a detection system for DNA demethylating agents suitable for a high-throughput screening (HTS) in mammalian cells. We inserted luciferase and EGFP reporter genes under the UCHL1 promoter, which is methylation-silenced in human colon cancers and can be readily demethylated to drive strong expression. Methylated UCHL1 promoter was introduced into HCT116 colon cancer cells, and transfectants with methylated exogenous UCHL1 promoter were obtained. By screening subclones from each of the epigenetically heterogeneous transfectant clones, we finally obtained three optimal subclones that expressed luciferase and EGFP after 5-aza-dC treatment with high signal-to-noise ratios. Nucleosomes with H3K9me2 were present around the exogenous UCHL1 promoter in all three subclones. Using one of the subclones (HML58-3), HTS was conducted using 19,840 small molecules. Two hit compounds were obtained, and these turned out to be 5-aza-dC and 5-aza-CR. The assay system constructed here demonstrates a robust response to DNA demethylating agents, along with high specificity, and will be useful for screening and biological assays in epigenetics.

Degree of methylation burden is determined by the exposure period to carcinogenic factors.

Aberrant DNA methylation accumulated in normal tissues, namely methylation burden, is associated with risk of carcinogenesis. The levels of methylation burden are known to be influenced by multiple factors, such as genetic factors and strengths of carcinogenic factors. However, the impact of the degree of exposure to a carcinogenic factor is still unclear. Here, using a Mongolian gerbil model of Helicobacter pylori (H. pylori)-induced gastritis, we aimed to clarify the impact of the degree of exposure on methylation burden in normal gastric tissues. DNA methylation levels of four CpG islands, HE6, SA9, SB5, and SD2, increased by H. pylori infection, depending upon the infection period. After eradication of H. pylori, DNA methylation levels decreased, but tended to be higher in gastric mucosae with a longer infection period. DNA molecules with dense methylation, but not those with sparse methylation, increased depending upon the infection period. DNA methylation levels of one of the four CpG islands, SA9, tended to be higher in gastric mucosae of gerbils infected with H. pylori, even 50 weeks after eradication than in those of non-infected gerbils. These results showed for the first time that the levels of methylation burden in normal tissues are influenced by the degree of exposure to a carcinogenic factor.

Frequent involvement of chromatin remodeler alterations in gastric field cancerization.

A field for cancerization, or a field defect, is formed by the accumulation of genetic and epigenetic alterations in normal-appearing tissues, and is involved in various cancers, especially multiple cancers. Epigenetic alterations are frequently present in chronic inflammation-exposed tissues, but information on individual genes involved in the formation of a field defect is still fragmental. Here, using non-cancerous gastric tissues of cancer patients, we isolated 16 aberrantly methylated genes, and identified chromatin remodelers ACTL6B and SMARCA1 as novel genes frequently methylated in non-cancerous tissues. SMARCA1 was expressed at high levels in normal gastric tissues, but was frequently silenced by aberrant methylation in gastric cancer cells. Moreover, somatic mutations of additional chromatin remodelers, such as ARID1A, SMARCA2, and SMARCA4, were found in 30% of gastric cancers. Mutant allele frequency suggested that the majority of cancer cells harbored a mutation when present. Depletion of a chromatin remodeler, SMARCA1 or SMARCA2, in cancer cell lines promoted their growth. These results showed that epigenetic and genetic alterations of chromatin remodelers are induced at an early stage of carcinogenesis and are frequently involved in the formation of a field defect.

Identification of coexistence of DNA methylation and H3K27me3 specifically in cancer cells as a promising target for epigenetic therapy.

Alterations of epigenetic modifications are promising targets for cancer therapy, and several epigenetic drugs are now being clinically utilized. At the same time, individual epigenetic modifications have physiological functions in normal cells, and cancer cell specificity is considered difficult to achieve using a drug against a single epigenetic modification. To overcome this limitation, a combination of epigenetic modifications specifically or preferentially present in cancer cells is a candidate target. In this study, we aimed to demonstrate (i) the presence of a cancer cell-specific combination of epigenetic modifications by focusing on DNA methylation and trimethylation of histone H3 lysine 27 (H3K27me3) and (ii) the therapeutic efficacy of a combination of DNA demethylation and EZH2 inhibition. Analyses of DNA methylation and H3K27me3 in human colon, breast and prostate cancer cell lines revealed that 24.7±4.1% of DNA methylated genes had both DNA methylation and H3K27me3 (dual modification) in cancer cells, while it was 11.8±7.1% in normal cells. Combined treatment with a DNA demethylating agent, 5-aza-2'-deoxycytidine (5-aza-dC) and an EZH2 inhibitor, GSK126, induced marked re-expression of genes with the dual modification, including known tumor-suppressor genes such as IGFBP7 and SFRP1, and showed an additive inhibitory effect on growth of cancer cells in vitro. Finally, an in vivo combined treatment with 5-aza-dC and GSK126 inhibited growth of xenograft tumors more efficiently than a single treatment with 5-aza-dC. These results showed that the dual modification exists specifically in cancer cells and is a promising target for cancer cell-specific epigenetic therapy.

Interleukin-1ß induced by Helicobacter pylori infection enhances mouse gastric carcinogenesis.

Interleukin-1ß (Il1b) is considered to be involved in Helicobacter pylori (HP)-induced human gastric carcinogenesis, while the role of its polymorphisms in gastric cancer susceptibility remains controversial. Here, we aimed to clarify the role of HP infection-induced IL1B in gastric inflammation and carcinogenesis using Il1b(-/-) (Il1b-null) mice. In gastric mucosa of the Il1b(+/+) (WT) mice, HP infection induced Il1b expression and severe inflammation. In contrast, in Il1b-null mice, recruitment of neutrophils and macrophages by HP infection was markedly suppressed. In a carcinogenicity test, the multiplicity of gastric tumors was significantly suppressed in theIl1b-null mice (58% of WT; P<0.005). Mechanistically, HP infection induced NF-κB activation both in the inflammatory and epithelial cells in gastric mucosae, and the activation was attenuated in the Il1b-null mice. Accordingly, increased proliferation and decreased apoptosis of gastric epithelial cells induced by HP infection in the WT mice were attenuated in the Il1b-null mice. These results demonstrated that the IL1B physiologically induced by HP infection enhanced gastric carcinogenesis by affecting both inflammatory and epithelial cells.

Induction of aberrant trimethylation of histone H3 lysine 27 by inflammation in mouse colonic epithelial cells.

A field for cancerization (field defect), where genetic and epigenetic alterations are accumulated in normal-appearing tissues, is involved in human carcinogenesis, especially cancers associated with chronic inflammation. Although aberrant DNA methylation is involved in the field defect and induced by chronic inflammation, it is still unclear for trimethylation of histone H3 lysine 27 (H3K27me3), which is involved in gene repression independent of DNA methylation and functions as a pre-mark for aberrant DNA methylation. In this study, using a mouse colitis model induced by dextran sulfate sodium (DSS), we aimed to clarify whether aberrant H3K27me3 is induced by inflammation and involved in a field defect. ChIP-on-chip analysis of colonic epithelial cells revealed that H3K27me3 levels were increased or decreased for 266 genomic regions by aging, and more extensively (23 increased and 3574 decreased regions) by colitis. Such increase or decrease of H3K27me3 was induced as early as 2 weeks after the initiation of DSS treatment, and persisted at least for 16 weeks even after the inflammation disappeared. Some of the aberrant H3K27me3 in colonic epithelial cells was carried over into colon tumors. Furthermore, H3K27me3 acquired at Dapk1 by colitis was followed by increased DNA methylation, supporting its function as a pre-mark for aberrant DNA methylation. These results demonstrated that aberrant H3K27me3 can be induced by exposure to a specific environment, such as colitis, and suggested that aberrant histone modification, in addition to aberrant DNA methylation, is involved in the formation of a field defect.

Development of a novel approach, the epigenome-based outlier approach, to identify tumor-suppressor genes silenced by aberrant DNA methylation.

Identification of tumor-suppressor genes (TSGs) silenced by aberrant methylation of promoter CpG islands (CGIs) is important, but hampered by a large number of genes methylated as passengers of carcinogenesis. To overcome this issue, we here took advantage of the fact that the vast majority of genes methylated in cancers lack, in normal cells, RNA polymerase II (Pol II) and have trimethylation of histone H3 lysine 27 (H3K27me3) in their promoter CGIs. First, we demonstrated that three of six known TSGs in breast cancer and two of three in colon cancer had Pol II and lacked H3K27me3 in normal cells, being outliers to the general rule. BRCA1, HOXA5, MLH1, and RASSF1A had high Pol II, but were expressed only at low levels in normal cells, and were unlikely to be identified as outliers by their expression statuses in normal cells. Then, using epigenome statuses (Pol II binding and H3K27me3) in normal cells, we made a genome-wide search for outliers in breast cancers, and identified 14 outlier promoter CGIs. Among these, DZIP1, FBN2, HOXA5, and HOXC9 were confirmed to be methylated in primary breast cancer samples. Knockdown of DZIP1 in breast cancer cell lines led to increases of their growth, suggesting it to be a novel TSG. The outliers based on their epigenome statuses contained unique TSGs, including DZIP1, compared with those identified by the expression microarray data. These results showed that the epigenome-based outlier approach is capable of identifying a different set of TSGs, compared to the expression-based outlier approach.

Methylation silencing of angiopoietin-like 4 in rat and human mammary carcinomas.

Aberrant DNA methylation is deeply involved in the development and progression of human breast cancers, but its inducers and molecular mechanisms are still unclear. To reveal such inducers and clarify the molecular mechanisms, animal models are indispensable. Here, to identify genes silenced by promoter DNA methylation in rat mammary carcinomas, we took a combined approach of methylated DNA immunoprecipitation (MeDIP)-CpG island (CGI) microarray analysis and expression microarray analysis after treatment with epigenetic drugs. MeDIP-CGI microarray revealed that among 5031 genes with promoter CGI, 465 were methylated in a carcinoma cell line induced by 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), but not in normal mammary epithelial cells. By treatment of the cell line with 5-aza-2'-deoxycytidine and trichostatin A, 29 of the 465 genes were shown to be re-expressed. In primary mammary carcinomas, five (Angptl4, Coro1a, RGD1304982, Tmem37 and Ndn) of the 29 genes were methylated in one or more of 25 samples. Quantitative expression analysis revealed that Angptl4 had high expression in normal mammary glands, but low expression in primary carcinomas. Also in humans, ANGPTL4 was unmethylated and expressed in normal mammary epithelial cells, but was methylated in 11 of 91 (12%) primary breast cancers. This is the first study to identify genes aberrantly methylated in rat mammary carcinomas, and Angptl4 is a novel methylation-silenced gene both in rat and human mammary carcinomas. The combination of the MeDIP-CGI microarray analysis and expression microarray analysis after treatment with epigenetic drugs was effective in reducing the number of methylated genes that are not methylation silenced.

Lysosomal turnover of GABARAP-phospholipid conjugate is activated during differentiation of C2C12 cells to myotubes without inactivation of the mTor kinase-signaling pathway.

Although conjugation of overexpressed GABARP to phospholipid has been reported during starvation-induced autophagy, it is unclear whether endogenous GABARAP-phospholipid conjugation is also activated under starvation conditions. We observed little accumulation of GABARAP-phospholipid conjugate (GABARAP-PL) in mouse liver and kidney under starvation conditions, whereas endogenous LC3-phospholipid conjugate (LC3-II) accumulated. A small amount of endogenous GABARAP-PL was observed in the heart, independent of starvation. In rapamycin-treated HEK293 cells, there was little accumulation of endogenous GABARAP-PL, even in the presence of lysosomal protease-inhibitors, whereas there was significant accumulation of endogenous LC3-II, together with inactivation of the mTor kinase-signaling pathway. In HeLa and C2C12 cells, GABARAP-PL accumulation in the presence of lysosomal protease inhibitors was independent of starvation-induced autophagy, whereas LC3-II accumulation was significant during starvation-induced autophagy. Interestingly, we observed activation of lysosomal turnover of GABARAP-PL during the differentiation of C2C12 cells to myotubes, along with increased lysosomal turnover of LC3-II. Under these conditions, S6 ribosomal protein was still phosphorylated, suggesting that the mTor kinase-signaling pathway is active during the differentiation of C2C12 cells to myotubes, in contrast to starvation-induced autophagy. These results indicated that lysosomal turnover of GABARAP-PL was activated during the differentiation of C2C12 cells to myotubes without inactivation of the mTor kinase-signaling pathway, whereas little lysosomal turnover of GABARAP-PL was activated during starvation-induced autophagy.

Silencing of the UCHL1 gene in human colorectal and ovarian cancers.

Aberrant DNA methylation is associated with many types of human cancers. To identify genes silenced in human colorectal cancers, we performed a microarray analysis for genes whose expression was induced by treatment of HCT116 human colon cancer cells with a demethylating agent, 5-aza-2'-deoxycitidine (5-aza-dC). Seven known genes were identified as being upregulated (> or =8-fold) and expressed at more than twice as high as the average level. Among these was the UCHL1 gene (also known as PGP9.5), which is involved in regulation of cellular ubiquitin levels. A dense CpG island in its promoter region was completely methylated in HCT116 cells, and no mRNA was detected. 5-Aza-dC treatment of HCT116 cells induced dose-dependent demethylation of the CpG island, and restored UCHL1 mRNA and protein expression. UCHL1 silencing was observed in 11 of 12 human colorectal cancer cell lines, and its methylation was detected in 8 of 17 primary colorectal cancers. Further, UCHL1 silencing was observed in 6 of 13 ovarian cancer cell lines, and its methylation was detected in 1 of 17 primary ovarian cancers. These results showed that UCHL1 is inactivated in human colorectal and ovarian cancers by its promoter methylation, and suggest that disturbance of cellular ubiquitin levels is present.

Methylation and expression analysis of 15 genes and three normally-methylated genes in 13 Ovarian cancer cell lines.

Aberrant methylation of CpG islands (CGIs) in promoter regions of tumor-suppressor genes causes their silencing, and aberrant demethylation of normally methylated CGIs in promoter regions causes aberrant expression of cancer-testis antigens. Here, we comprehensively analyzed aberrant methylation of 15 genes and demethylation of three normally methylated genes in 13 ovarian cancer cell lines. RASSF1A was most frequently methylated (complete methylation in 7 and partial methylation in 4 cell lines), followed by ESR1 (5 and 2, respectively), FLNC (4 and 4), HAND1 (4 and 2), LOX (3 and 2), HRASLS (3 and 2), MGMT (3 and 0), CDKN2A (3 and 0), THBD (2 and 1), hMLH1 (2 and 0), CDH1 (1 and 1) and GSTP1 (1 and 0). hTERC and TIMP3 were only partially methylated in 7 and 2 cell lines, respectively. BRCA1 was not methylated at all. Aberrant demethylation of MAGE-A3, -B2 and -A1 was detected in 8, 4 and 3 cell lines, respectively. Gene expression was consistently absent in cell lines without unmethylated DNA molecules. Aberrant methylation was frequently observed in MCAS, RMUG-L (mucinous cell carcinomas), RTSG (poorly-differentiated carcinoma) and TYK-nu (undifferentiated carcinoma) while infrequent in HTOA, JHOS-2, and OV-90 (serous cell carcinomas). Aberrant demethylation was frequently observed in OV-90, OVK-18, and ES-2 cell lines. It was shown that aberrant methylation and demethylation were frequently observed in ovarian cancer cell lines, and these data will provide a basis for further epigenetic analysis in ovarian cancers.